[llvm] [X86] isTargetShuffleEquivalent - only use "known zero" element if its known never poison/undef as well (PR #217030)
Simon Pilgrim via llvm-commits
llvm-commits at lists.llvm.org
Wed Aug 19 01:58:48 PDT 2026
https://github.com/RKSimon updated https://github.com/llvm/llvm-project/pull/217030
>From 49da57ad47c89ce08c07a066e67dfebd08adc61d Mon Sep 17 00:00:00 2001
From: Simon Pilgrim <llvm-dev at redking.me.uk>
Date: Tue, 18 Aug 2026 09:41:18 +0100
Subject: [PATCH 1/2] [X86] Add test coverage for #214388
---
llvm/test/CodeGen/X86/pr214388.ll | 118 ++++++++++++++++++++++++++++++
1 file changed, 118 insertions(+)
create mode 100644 llvm/test/CodeGen/X86/pr214388.ll
diff --git a/llvm/test/CodeGen/X86/pr214388.ll b/llvm/test/CodeGen/X86/pr214388.ll
new file mode 100644
index 0000000000000..77917e41380e7
--- /dev/null
+++ b/llvm/test/CodeGen/X86/pr214388.ll
@@ -0,0 +1,118 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6
+; RUN: llc < %s -mtriple=x86_64-- -mcpu=x86-64 | FileCheck %s --check-prefixes=SSE2
+; RUN: llc < %s -mtriple=x86_64-- -mcpu=x86-64-v2 | FileCheck %s --check-prefixes=SSE42
+; RUN: llc < %s -mtriple=x86_64-- -mcpu=x86-64-v3 | FileCheck %s --check-prefixes=AVX
+; RUN: llc < %s -mtriple=x86_64-- -mcpu=x86-64-v4 | FileCheck %s --check-prefixes=AVX
+
+define i64 @PR214388(i8 %arg) {
+; SSE2-LABEL: PR214388:
+; SSE2: # %bb.0: # %entry
+; SSE2-NEXT: movzbl %dil, %eax
+; SSE2-NEXT: movd %eax, %xmm0
+; SSE2-NEXT: por {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; SSE2-NEXT: movl $20, %eax
+; SSE2-NEXT: movd %eax, %xmm1
+; SSE2-NEXT: psubb %xmm0, %xmm1
+; SSE2-NEXT: movd %xmm1, %eax
+; SSE2-NEXT: movw %ax, -{{[0-9]+}}(%rsp)
+; SSE2-NEXT: movdqa %xmm1, -{{[0-9]+}}(%rsp)
+; SSE2-NEXT: cmpb $0, -{{[0-9]+}}(%rsp)
+; SSE2-NEXT: je .LBB0_1
+; SSE2-NEXT: # %bb.2: # %for.body8.preheader
+; SSE2-NEXT: leaq -{{[0-9]+}}(%rsp), %rcx
+; SSE2-NEXT: xorl %esi, %esi
+; SSE2-NEXT: leaq -{{[0-9]+}}(%rsp), %rdx
+; SSE2-NEXT: .p2align 4
+; SSE2-NEXT: .LBB0_3: # %for.body8
+; SSE2-NEXT: # =>This Inner Loop Header: Depth=1
+; SSE2-NEXT: movq %rsi, %rax
+; SSE2-NEXT: movl $1, %esi
+; SSE2-NEXT: cmpb $0, 1(%rcx)
+; SSE2-NEXT: movq %rdx, %rcx
+; SSE2-NEXT: jne .LBB0_3
+; SSE2-NEXT: # %bb.4: # %for.end11
+; SSE2-NEXT: retq
+; SSE2-NEXT: .LBB0_1:
+; SSE2-NEXT: xorl %eax, %eax
+; SSE2-NEXT: retq
+;
+; SSE42-LABEL: PR214388:
+; SSE42: # %bb.0: # %entry
+; SSE42-NEXT: movdqa {{.*#+}} xmm0 = [u,1,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
+; SSE42-NEXT: pinsrb $0, %edi, %xmm0
+; SSE42-NEXT: movl $20, %eax
+; SSE42-NEXT: movd %eax, %xmm1
+; SSE42-NEXT: psubb %xmm0, %xmm1
+; SSE42-NEXT: pextrw $0, %xmm1, -{{[0-9]+}}(%rsp)
+; SSE42-NEXT: movd %xmm1, %eax
+; SSE42-NEXT: testl %eax, %eax
+; SSE42-NEXT: je .LBB0_1
+; SSE42-NEXT: # %bb.2: # %for.body8.preheader
+; SSE42-NEXT: leaq -{{[0-9]+}}(%rsp), %rcx
+; SSE42-NEXT: xorl %esi, %esi
+; SSE42-NEXT: leaq -{{[0-9]+}}(%rsp), %rdx
+; SSE42-NEXT: .p2align 4
+; SSE42-NEXT: .LBB0_3: # %for.body8
+; SSE42-NEXT: # =>This Inner Loop Header: Depth=1
+; SSE42-NEXT: movq %rsi, %rax
+; SSE42-NEXT: movl $1, %esi
+; SSE42-NEXT: cmpb $0, 1(%rcx)
+; SSE42-NEXT: movq %rdx, %rcx
+; SSE42-NEXT: jne .LBB0_3
+; SSE42-NEXT: # %bb.4: # %for.end11
+; SSE42-NEXT: retq
+; SSE42-NEXT: .LBB0_1:
+; SSE42-NEXT: xorl %eax, %eax
+; SSE42-NEXT: retq
+;
+; AVX-LABEL: PR214388:
+; AVX: # %bb.0: # %entry
+; AVX-NEXT: vpbroadcastd {{.*#+}} xmm0 = [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1]
+; AVX-NEXT: vpinsrb $0, %edi, %xmm0, %xmm0
+; AVX-NEXT: movl $20, %eax
+; AVX-NEXT: vmovd %eax, %xmm1
+; AVX-NEXT: vpsubb %xmm0, %xmm1, %xmm0
+; AVX-NEXT: vpextrw $0, %xmm0, -{{[0-9]+}}(%rsp)
+; AVX-NEXT: vmovd %xmm0, %eax
+; AVX-NEXT: testl %eax, %eax
+; AVX-NEXT: je .LBB0_1
+; AVX-NEXT: # %bb.2: # %for.body8.preheader
+; AVX-NEXT: leaq -{{[0-9]+}}(%rsp), %rcx
+; AVX-NEXT: xorl %esi, %esi
+; AVX-NEXT: leaq -{{[0-9]+}}(%rsp), %rdx
+; AVX-NEXT: .p2align 4
+; AVX-NEXT: .LBB0_3: # %for.body8
+; AVX-NEXT: # =>This Inner Loop Header: Depth=1
+; AVX-NEXT: movq %rsi, %rax
+; AVX-NEXT: movl $1, %esi
+; AVX-NEXT: cmpb $0, 1(%rcx)
+; AVX-NEXT: movq %rdx, %rcx
+; AVX-NEXT: jne .LBB0_3
+; AVX-NEXT: # %bb.4: # %for.end11
+; AVX-NEXT: retq
+; AVX-NEXT: .LBB0_1:
+; AVX-NEXT: xorl %eax, %eax
+; AVX-NEXT: retq
+entry:
+ %i0 = alloca [4 x i8], align 1
+ %i1 = insertelement <2 x i8> splat (i8 1), i8 %arg, i64 0
+ %i2 = sub nuw <2 x i8> <i8 20, i8 poison>, %i1
+ %i3 = insertelement <2 x i8> %i2, i8 0, i64 1
+ %i4 = getelementptr i8, ptr %i0, i64 2
+ store <2 x i8> %i3, ptr %i4, align 1
+ %i5 = extractelement <2 x i8> %i2, i64 0
+ %i6 = icmp eq i8 %i5, 0
+ br i1 %i6, label %for.end11, label %for.body8
+
+for.body8:
+ %i7 = phi ptr [ %i0, %for.body8 ], [ %i4, %entry ]
+ %i8 = phi i64 [ 1, %for.body8 ], [ 0, %entry ]
+ %i9 = getelementptr i8, ptr %i7, i64 1
+ %i10 = load i8, ptr %i9, align 1
+ %i11 = icmp eq i8 %i10, 0
+ br i1 %i11, label %for.end11, label %for.body8
+
+for.end11:
+ %i12 = phi i64 [ 0, %entry ], [ %i8, %for.body8 ]
+ ret i64 %i12
+}
>From b018828a52606c057d7ab96f2541dd13b09ee300 Mon Sep 17 00:00:00 2001
From: Simon Pilgrim <llvm-dev at redking.me.uk>
Date: Tue, 18 Aug 2026 14:34:43 +0100
Subject: [PATCH 2/2] [X86] isTargetShuffleEquivalent - only use "known zero"
element if its known never poison/undef as well
We're starting to hit cases where we've used AND/ANDNP masking to zero out vector elements, but SimplifyDemandedElts assumes the element is unused, folds to poison and then later folds AND(poison,0) -> poison.
This needs a more thorough cleanup of a number of x86 folds that do this (similar to #215538), but that would result in a great deal of churn, and it looks like a #214388 fix requires backporting to 23.x.
So initially I'm taking the approach that @xyyy1420 found so we can get this backported and I can then address the issue more thoroughly in trunk. I'm not convinced that this will address all poison cases, but I'm reluctant to attempt a larger patch (happy to put it up for comparison).
Fixes #214388
---
llvm/lib/Target/X86/X86ISelLowering.cpp | 8 +-
llvm/test/CodeGen/X86/pr214388.ll | 10 +-
llvm/test/CodeGen/X86/psadbw.ll | 16 +-
llvm/test/CodeGen/X86/shrink_vmul.ll | 16 +-
.../vector-interleaved-load-i8-stride-6.ll | 80 +-
.../vector-interleaved-store-i16-stride-7.ll | 1496 +++++++++--------
.../vector-shuffle-combining-avx512bwvl.ll | 4 +
.../CodeGen/X86/vector-shuffle-combining.ll | 1 +
8 files changed, 825 insertions(+), 806 deletions(-)
diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index e89afce282602..6f78baf7e86da 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -10422,8 +10422,12 @@ static bool isTargetShuffleEquivalent(MVT VT, ArrayRef<int> Mask,
}
return false;
}
- return (ZeroV1.isZero() || DAG.MaskedVectorIsZero(V1, ZeroV1)) &&
- (ZeroV2.isZero() || DAG.MaskedVectorIsZero(V2, ZeroV2));
+ return (ZeroV1.isZero() ||
+ (DAG.MaskedVectorIsZero(V1, ZeroV1) &&
+ DAG.isGuaranteedNotToBeUndefOrPoison(V1, ZeroV1))) &&
+ (ZeroV2.isZero() ||
+ (DAG.MaskedVectorIsZero(V2, ZeroV2) &&
+ DAG.isGuaranteedNotToBeUndefOrPoison(V2, ZeroV2)));
}
// Check if the shuffle mask is suitable for the AVX vpunpcklwd or vpunpckhwd
diff --git a/llvm/test/CodeGen/X86/pr214388.ll b/llvm/test/CodeGen/X86/pr214388.ll
index 77917e41380e7..b53ba07406fb1 100644
--- a/llvm/test/CodeGen/X86/pr214388.ll
+++ b/llvm/test/CodeGen/X86/pr214388.ll
@@ -13,7 +13,9 @@ define i64 @PR214388(i8 %arg) {
; SSE2-NEXT: movl $20, %eax
; SSE2-NEXT: movd %eax, %xmm1
; SSE2-NEXT: psubb %xmm0, %xmm1
-; SSE2-NEXT: movd %xmm1, %eax
+; SSE2-NEXT: movdqa {{.*#+}} xmm0 = [255,0,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; SSE2-NEXT: pand %xmm1, %xmm0
+; SSE2-NEXT: movd %xmm0, %eax
; SSE2-NEXT: movw %ax, -{{[0-9]+}}(%rsp)
; SSE2-NEXT: movdqa %xmm1, -{{[0-9]+}}(%rsp)
; SSE2-NEXT: cmpb $0, -{{[0-9]+}}(%rsp)
@@ -43,7 +45,8 @@ define i64 @PR214388(i8 %arg) {
; SSE42-NEXT: movl $20, %eax
; SSE42-NEXT: movd %eax, %xmm1
; SSE42-NEXT: psubb %xmm0, %xmm1
-; SSE42-NEXT: pextrw $0, %xmm1, -{{[0-9]+}}(%rsp)
+; SSE42-NEXT: pextrb $0, %xmm1, %eax
+; SSE42-NEXT: movw %ax, -{{[0-9]+}}(%rsp)
; SSE42-NEXT: movd %xmm1, %eax
; SSE42-NEXT: testl %eax, %eax
; SSE42-NEXT: je .LBB0_1
@@ -72,7 +75,8 @@ define i64 @PR214388(i8 %arg) {
; AVX-NEXT: movl $20, %eax
; AVX-NEXT: vmovd %eax, %xmm1
; AVX-NEXT: vpsubb %xmm0, %xmm1, %xmm0
-; AVX-NEXT: vpextrw $0, %xmm0, -{{[0-9]+}}(%rsp)
+; AVX-NEXT: vpextrb $0, %xmm0, %eax
+; AVX-NEXT: movw %ax, -{{[0-9]+}}(%rsp)
; AVX-NEXT: vmovd %xmm0, %eax
; AVX-NEXT: testl %eax, %eax
; AVX-NEXT: je .LBB0_1
diff --git a/llvm/test/CodeGen/X86/psadbw.ll b/llvm/test/CodeGen/X86/psadbw.ll
index ccfb46a204800..a759acd080933 100644
--- a/llvm/test/CodeGen/X86/psadbw.ll
+++ b/llvm/test/CodeGen/X86/psadbw.ll
@@ -76,18 +76,10 @@ define void @combine_psadbw_demandedelt_store(<16 x i8> %0, <16 x i8> %1, ptr %p
; TODO: Each PSADBW source element has a maximum value of 3 - so max sum-of-diffs for each <8 x i8> should be 24.
define <2 x i64> @combine_psadbw_cmp_knownbits(<16 x i8> %a0) nounwind {
-; X86-SSE-LABEL: combine_psadbw_cmp_knownbits:
-; X86-SSE: # %bb.0:
-; X86-SSE-NEXT: xorps %xmm0, %xmm0
-; X86-SSE-NEXT: retl
-;
-; X64-SSE-LABEL: combine_psadbw_cmp_knownbits:
-; X64-SSE: # %bb.0:
-; X64-SSE-NEXT: pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; X64-SSE-NEXT: pxor %xmm1, %xmm1
-; X64-SSE-NEXT: psadbw %xmm1, %xmm0
-; X64-SSE-NEXT: pcmpgtd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; X64-SSE-NEXT: retq
+; SSE-LABEL: combine_psadbw_cmp_knownbits:
+; SSE: # %bb.0:
+; SSE-NEXT: xorps %xmm0, %xmm0
+; SSE-NEXT: ret{{[l|q]}}
;
; AVX2-LABEL: combine_psadbw_cmp_knownbits:
; AVX2: # %bb.0:
diff --git a/llvm/test/CodeGen/X86/shrink_vmul.ll b/llvm/test/CodeGen/X86/shrink_vmul.ll
index 766d72fec0d8c..9c6ecf400ead8 100644
--- a/llvm/test/CodeGen/X86/shrink_vmul.ll
+++ b/llvm/test/CodeGen/X86/shrink_vmul.ll
@@ -1742,15 +1742,20 @@ define void @mul_2xi16_varconst1(ptr nocapture readonly %a, i64 %index) {
; X86-SSE-LABEL: mul_2xi16_varconst1:
; X86-SSE: # %bb.0: # %entry
; X86-SSE-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-SSE-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-SSE-NEXT: movl c, %edx
+; X86-SSE-NEXT: movl {{[0-9]+}}(%esp), %edx
+; X86-SSE-NEXT: movl c, %ecx
; X86-SSE-NEXT: movd {{.*#+}} xmm0 = mem[0],zero,zero,zero
; X86-SSE-NEXT: movd {{.*#+}} xmm1 = [0,65535,0,0,0,0,0,0]
; X86-SSE-NEXT: movdqa %xmm0, %xmm2
; X86-SSE-NEXT: pmulhuw %xmm1, %xmm2
; X86-SSE-NEXT: pmullw %xmm1, %xmm0
+; X86-SSE-NEXT: movdqa %xmm0, %xmm1
+; X86-SSE-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
+; X86-SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,1,1,1]
+; X86-SSE-NEXT: pxor %xmm2, %xmm2
; X86-SSE-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
-; X86-SSE-NEXT: movq %xmm0, (%edx,%eax,4)
+; X86-SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
+; X86-SSE-NEXT: movq %xmm0, (%ecx,%eax,4)
; X86-SSE-NEXT: retl
;
; X86-AVX-LABEL: mul_2xi16_varconst1:
@@ -1772,7 +1777,12 @@ define void @mul_2xi16_varconst1(ptr nocapture readonly %a, i64 %index) {
; X64-SSE-NEXT: movdqa %xmm0, %xmm2
; X64-SSE-NEXT: pmulhuw %xmm1, %xmm2
; X64-SSE-NEXT: pmullw %xmm1, %xmm0
+; X64-SSE-NEXT: movdqa %xmm0, %xmm1
+; X64-SSE-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
+; X64-SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,1,1,1]
+; X64-SSE-NEXT: pxor %xmm2, %xmm2
; X64-SSE-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
+; X64-SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; X64-SSE-NEXT: movq %xmm0, (%rax,%rsi,4)
; X64-SSE-NEXT: retq
;
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-6.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-6.ll
index 054df4c00e0a3..023f2241fd31f 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-6.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-6.ll
@@ -295,71 +295,73 @@ define void @load_i8_stride6_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; SSE-LABEL: load_i8_stride6_vf4:
; SSE: # %bb.0:
; SSE-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; SSE-NEXT: movdqa (%rdi), %xmm5
+; SSE-NEXT: movdqa (%rdi), %xmm2
; SSE-NEXT: movdqa 16(%rdi), %xmm1
; SSE-NEXT: movdqa {{.*#+}} xmm0 = [65535,0,65535,65535,65535,65535,65535,65535]
-; SSE-NEXT: movdqa %xmm5, %xmm2
-; SSE-NEXT: pand %xmm0, %xmm2
+; SSE-NEXT: movdqa %xmm2, %xmm3
+; SSE-NEXT: pand %xmm0, %xmm3
; SSE-NEXT: pandn %xmm1, %xmm0
-; SSE-NEXT: por %xmm2, %xmm0
+; SSE-NEXT: por %xmm3, %xmm0
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,1,3]
-; SSE-NEXT: movdqa {{.*#+}} xmm2 = [16711935,16711935,16711935,16711935]
-; SSE-NEXT: pand %xmm2, %xmm0
+; SSE-NEXT: movdqa {{.*#+}} xmm4 = [16711935,16711935,16711935,16711935]
+; SSE-NEXT: pand %xmm4, %xmm0
; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,3,2,1,4,5,6,7]
; SSE-NEXT: packuswb %xmm0, %xmm0
-; SSE-NEXT: pxor %xmm3, %xmm3
-; SSE-NEXT: movdqa %xmm5, %xmm7
-; SSE-NEXT: movdqa {{.*#+}} xmm4 = [255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255]
-; SSE-NEXT: pandn %xmm1, %xmm4
-; SSE-NEXT: movdqa %xmm1, %xmm6
-; SSE-NEXT: shufps {{.*#+}} xmm6 = xmm6[1,0],xmm5[0,0]
-; SSE-NEXT: shufps {{.*#+}} xmm6 = xmm6[2,0],xmm5[2,3]
-; SSE-NEXT: movdqa {{.*#+}} xmm8 = [65535,65535,0,65535,65535,0,65535,65535]
-; SSE-NEXT: pand %xmm8, %xmm1
-; SSE-NEXT: pandn %xmm5, %xmm8
-; SSE-NEXT: punpckhbw {{.*#+}} xmm5 = xmm5[8],xmm3[8],xmm5[9],xmm3[9],xmm5[10],xmm3[10],xmm5[11],xmm3[11],xmm5[12],xmm3[12],xmm5[13],xmm3[13],xmm5[14],xmm3[14],xmm5[15],xmm3[15]
-; SSE-NEXT: movdqa %xmm5, %xmm9
-; SSE-NEXT: psrld $16, %xmm9
-; SSE-NEXT: punpcklbw {{.*#+}} xmm7 = xmm7[0],xmm3[0],xmm7[1],xmm3[1],xmm7[2],xmm3[2],xmm7[3],xmm3[3],xmm7[4],xmm3[4],xmm7[5],xmm3[5],xmm7[6],xmm3[6],xmm7[7],xmm3[7]
+; SSE-NEXT: pxor %xmm5, %xmm5
+; SSE-NEXT: movdqa %xmm2, %xmm3
+; SSE-NEXT: punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm5[8],xmm3[9],xmm5[9],xmm3[10],xmm5[10],xmm3[11],xmm5[11],xmm3[12],xmm5[12],xmm3[13],xmm5[13],xmm3[14],xmm5[14],xmm3[15],xmm5[15]
+; SSE-NEXT: movdqa %xmm3, %xmm6
+; SSE-NEXT: psrld $16, %xmm6
+; SSE-NEXT: movdqa %xmm2, %xmm7
+; SSE-NEXT: punpcklbw {{.*#+}} xmm7 = xmm7[0],xmm5[0],xmm7[1],xmm5[1],xmm7[2],xmm5[2],xmm7[3],xmm5[3],xmm7[4],xmm5[4],xmm7[5],xmm5[5],xmm7[6],xmm5[6],xmm7[7],xmm5[7]
; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm7[0,1,0,3]
; SSE-NEXT: pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,5,7,6,7]
-; SSE-NEXT: punpckhdq {{.*#+}} xmm7 = xmm7[2],xmm9[2],xmm7[3],xmm9[3]
+; SSE-NEXT: punpckhdq {{.*#+}} xmm7 = xmm7[2],xmm6[2],xmm7[3],xmm6[3]
; SSE-NEXT: packuswb %xmm7, %xmm7
-; SSE-NEXT: por %xmm7, %xmm4
-; SSE-NEXT: pshuflw {{.*#+}} xmm7 = xmm6[2,1,2,3,4,5,6,7]
+; SSE-NEXT: movdqa {{.*#+}} xmm6 = [255,255,255,0,255,255,255,255,255,255,255,255,255,255,255,255]
+; SSE-NEXT: pand %xmm6, %xmm7
+; SSE-NEXT: pandn %xmm1, %xmm6
+; SSE-NEXT: por %xmm7, %xmm6
+; SSE-NEXT: movdqa %xmm1, %xmm8
+; SSE-NEXT: shufps {{.*#+}} xmm8 = xmm8[1,0],xmm2[0,0]
+; SSE-NEXT: shufps {{.*#+}} xmm8 = xmm8[2,0],xmm2[2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm7 = xmm8[2,1,2,3,4,5,6,7]
; SSE-NEXT: pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,7,6,7]
; SSE-NEXT: pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm7
; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm7[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm7 = xmm7[1,2,3,0,4,5,6,7]
; SSE-NEXT: packuswb %xmm7, %xmm7
-; SSE-NEXT: punpcklbw {{.*#+}} xmm6 = xmm6[0],xmm3[0],xmm6[1],xmm3[1],xmm6[2],xmm3[2],xmm6[3],xmm3[3],xmm6[4],xmm3[4],xmm6[5],xmm3[5],xmm6[6],xmm3[6],xmm6[7],xmm3[7]
-; SSE-NEXT: shufps {{.*#+}} xmm6 = xmm6[1,2],xmm5[0,3]
-; SSE-NEXT: shufps {{.*#+}} xmm6 = xmm6[2,0,1,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm6 = xmm6[3,1,2,3,4,5,6,7]
-; SSE-NEXT: pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,7,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm6[0,2,2,3]
-; SSE-NEXT: packuswb %xmm6, %xmm6
-; SSE-NEXT: por %xmm1, %xmm8
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm8[3,1,2,0]
-; SSE-NEXT: pand %xmm2, %xmm1
+; SSE-NEXT: punpcklbw {{.*#+}} xmm8 = xmm8[0],xmm5[0],xmm8[1],xmm5[1],xmm8[2],xmm5[2],xmm8[3],xmm5[3],xmm8[4],xmm5[4],xmm8[5],xmm5[5],xmm8[6],xmm5[6],xmm8[7],xmm5[7]
+; SSE-NEXT: shufps {{.*#+}} xmm8 = xmm8[1,2],xmm3[0,3]
+; SSE-NEXT: shufps {{.*#+}} xmm8 = xmm8[2,0,1,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm8 = xmm8[3,1,2,3,4,5,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm8 = xmm8[0,1,2,3,7,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm8[0,2,2,3]
+; SSE-NEXT: packuswb %xmm8, %xmm8
+; SSE-NEXT: movdqa {{.*#+}} xmm9 = [65535,65535,0,65535,65535,0,65535,65535]
+; SSE-NEXT: pand %xmm9, %xmm1
+; SSE-NEXT: pandn %xmm2, %xmm9
+; SSE-NEXT: por %xmm1, %xmm9
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm9[3,1,2,0]
+; SSE-NEXT: pand %xmm4, %xmm1
; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[2,1,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[2,1,0,3,4,5,6,7]
; SSE-NEXT: packuswb %xmm1, %xmm1
-; SSE-NEXT: punpcklbw {{.*#+}} xmm8 = xmm8[0],xmm3[0],xmm8[1],xmm3[1],xmm8[2],xmm3[2],xmm8[3],xmm3[3],xmm8[4],xmm3[4],xmm8[5],xmm3[5],xmm8[6],xmm3[6],xmm8[7],xmm3[7]
-; SSE-NEXT: shufps {{.*#+}} xmm5 = xmm5[1,0],xmm8[0,0]
-; SSE-NEXT: shufps {{.*#+}} xmm5 = xmm5[2,0],xmm8[2,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm5[3,1,2,3,4,5,6,7]
+; SSE-NEXT: punpcklbw {{.*#+}} xmm9 = xmm9[0],xmm5[0],xmm9[1],xmm5[1],xmm9[2],xmm5[2],xmm9[3],xmm5[3],xmm9[4],xmm5[4],xmm9[5],xmm5[5],xmm9[6],xmm5[6],xmm9[7],xmm5[7]
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[1,0],xmm9[0,0]
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[2,0],xmm9[2,3]
+; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm3[3,1,2,3,4,5,6,7]
; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,2,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[3,0,1,2,4,5,6,7]
; SSE-NEXT: packuswb %xmm2, %xmm2
; SSE-NEXT: movd %xmm0, (%rsi)
-; SSE-NEXT: movd %xmm4, (%rdx)
+; SSE-NEXT: movd %xmm6, (%rdx)
; SSE-NEXT: movd %xmm7, (%rcx)
-; SSE-NEXT: movd %xmm6, (%r8)
+; SSE-NEXT: movd %xmm8, (%r8)
; SSE-NEXT: movd %xmm1, (%r9)
; SSE-NEXT: movd %xmm2, (%rax)
; SSE-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 6e0d9efbb083c..c36e9a36bfe59 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
@@ -981,10 +981,11 @@ define void @store_i16_stride7_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; SSE-NEXT: movdqa %xmm6, %xmm13
; SSE-NEXT: psrldq {{.*#+}} xmm13 = xmm13[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
; SSE-NEXT: por %xmm11, %xmm13
+; SSE-NEXT: movdqa {{.*#+}} xmm0 = [65535,65535,65535,65535,65535,0,0,65535]
+; SSE-NEXT: pand %xmm0, %xmm13
; SSE-NEXT: psrld $16, %xmm10
; SSE-NEXT: movdqa %xmm4, %xmm11
; SSE-NEXT: punpckhdq {{.*#+}} xmm11 = xmm11[2],xmm10[2],xmm11[3],xmm10[3]
-; SSE-NEXT: movdqa {{.*#+}} xmm0 = [65535,65535,65535,65535,65535,0,0,65535]
; SSE-NEXT: movdqa %xmm0, %xmm12
; SSE-NEXT: pandn %xmm11, %xmm12
; SSE-NEXT: por %xmm13, %xmm12
@@ -1701,13 +1702,13 @@ define void @store_i16_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; SSE: # %bb.0:
; SSE-NEXT: subq $216, %rsp
; SSE-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; SSE-NEXT: movdqa 16(%rdi), %xmm6
+; SSE-NEXT: movdqa 16(%rdi), %xmm4
; SSE-NEXT: movdqa 16(%rsi), %xmm9
; SSE-NEXT: movdqa 16(%rdx), %xmm15
; SSE-NEXT: movdqa 16(%rcx), %xmm1
; SSE-NEXT: movdqa 16(%r8), %xmm8
; SSE-NEXT: movdqa 16(%r9), %xmm7
-; SSE-NEXT: movdqa 16(%rax), %xmm3
+; SSE-NEXT: movdqa 16(%rax), %xmm6
; SSE-NEXT: movdqa %xmm1, %xmm0
; SSE-NEXT: movdqa %xmm1, %xmm5
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -1716,37 +1717,39 @@ define void @store_i16_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; SSE-NEXT: punpckhqdq {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE-NEXT: movdqa {{.*#+}} xmm2 = [65535,65535,65535,0,0,65535,65535,65535]
; SSE-NEXT: pandn %xmm1, %xmm2
-; SSE-NEXT: movdqa %xmm6, %xmm0
-; SSE-NEXT: movdqa %xmm6, %xmm10
-; SSE-NEXT: movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm4, %xmm0
+; SSE-NEXT: movdqa %xmm4, %xmm10
+; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm9[4],xmm0[5],xmm9[5],xmm0[6],xmm9[6],xmm0[7],xmm9[7]
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: psrldq {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
; SSE-NEXT: por %xmm2, %xmm0
-; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm8[2,2,3,3]
; SSE-NEXT: movdqa {{.*#+}} xmm4 = [65535,65535,65535,65535,65535,0,65535,65535]
-; SSE-NEXT: movdqa %xmm4, %xmm2
-; SSE-NEXT: pandn %xmm1, %xmm2
-; SSE-NEXT: por %xmm0, %xmm2
+; SSE-NEXT: movdqa %xmm0, %xmm1
+; SSE-NEXT: pand %xmm4, %xmm1
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm8[2,2,3,3]
+; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm4, %xmm3
+; SSE-NEXT: pandn %xmm2, %xmm3
+; SSE-NEXT: por %xmm1, %xmm3
; SSE-NEXT: movdqa %xmm8, %xmm1
-; SSE-NEXT: movdqa %xmm7, %xmm6
+; SSE-NEXT: movdqa %xmm7, %xmm2
; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm7[0],xmm1[1],xmm7[1],xmm1[2],xmm7[2],xmm1[3],xmm7[3]
; SSE-NEXT: movdqa %xmm1, %xmm7
-; SSE-NEXT: movdqa %xmm6, %xmm1
+; SSE-NEXT: movdqa %xmm2, %xmm1
; SSE-NEXT: psrld $16, %xmm1
-; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[3,0],xmm2[2,0]
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[3,0],xmm3[2,0]
; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,0]
-; SSE-NEXT: movaps {{.*#+}} xmm6 = [0,65535,65535,65535,65535,65535,65535,0]
-; SSE-NEXT: andps %xmm6, %xmm0
-; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm3[3,3,3,3]
-; SSE-NEXT: andnps %xmm1, %xmm6
-; SSE-NEXT: orps %xmm0, %xmm6
-; SSE-NEXT: movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm3[0,1,0,1]
+; SSE-NEXT: movaps {{.*#+}} xmm3 = [0,65535,65535,65535,65535,65535,65535,0]
+; SSE-NEXT: andps %xmm3, %xmm0
+; SSE-NEXT: movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm6[3,3,3,3]
+; SSE-NEXT: andnps %xmm1, %xmm3
+; SSE-NEXT: orps %xmm0, %xmm3
+; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm6[0,1,0,1]
; SSE-NEXT: movdqa %xmm4, %xmm1
; SSE-NEXT: pandn %xmm0, %xmm1
; SSE-NEXT: movdqa %xmm7, %xmm0
@@ -3593,12 +3596,14 @@ define void @store_i16_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; SSE-NEXT: movdqa %xmm5, %xmm2
; SSE-NEXT: psrldq {{.*#+}} xmm2 = xmm2[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
; SSE-NEXT: por %xmm1, %xmm2
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm9[2,2,3,3]
; SSE-NEXT: movdqa {{.*#+}} xmm1 = [65535,65535,65535,65535,65535,0,65535,65535]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm9[2,2,3,3]
; SSE-NEXT: movdqa %xmm1, %xmm4
-; SSE-NEXT: movdqa %xmm1, %xmm8
; SSE-NEXT: pandn %xmm3, %xmm4
-; SSE-NEXT: por %xmm2, %xmm4
+; SSE-NEXT: movdqa %xmm2, %xmm3
+; SSE-NEXT: pand %xmm1, %xmm3
+; SSE-NEXT: movdqa %xmm1, %xmm5
+; SSE-NEXT: por %xmm3, %xmm4
; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[3,0],xmm4[2,0]
; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,1],xmm0[2,0]
; SSE-NEXT: movdqa {{.*#+}} xmm1 = [0,65535,65535,65535,65535,65535,65535,0]
@@ -3610,41 +3615,40 @@ define void @store_i16_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa (%rax), %xmm7
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm7[0,1,0,1]
-; SSE-NEXT: movdqa %xmm8, %xmm2
+; SSE-NEXT: movdqa %xmm5, %xmm4
+; SSE-NEXT: movdqa %xmm5, %xmm2
; SSE-NEXT: pandn %xmm0, %xmm2
; SSE-NEXT: movdqa (%r8), %xmm0
-; SSE-NEXT: movdqa (%r9), %xmm13
+; SSE-NEXT: movdqa (%r9), %xmm14
; SSE-NEXT: movdqa %xmm0, %xmm3
-; SSE-NEXT: movdqa %xmm0, %xmm12
+; SSE-NEXT: movdqa %xmm0, %xmm13
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm13[0],xmm3[1],xmm13[1],xmm3[2],xmm13[2],xmm3[3],xmm13[3]
-; SSE-NEXT: movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm14[0],xmm3[1],xmm14[1],xmm3[2],xmm14[2],xmm3[3],xmm14[3]
+; SSE-NEXT: movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa %xmm3, %xmm0
; SSE-NEXT: movdqa %xmm3, %xmm5
; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pslldq {{.*#+}} xmm0 = zero,zero,xmm0[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
-; SSE-NEXT: pand %xmm8, %xmm0
+; SSE-NEXT: pand %xmm4, %xmm0
; SSE-NEXT: por %xmm2, %xmm0
-; SSE-NEXT: movdqa {{.*#+}} xmm1 = [65535,65535,65535,0,0,0,65535,65535]
-; SSE-NEXT: movdqa %xmm1, %xmm2
-; SSE-NEXT: movdqa %xmm1, %xmm10
+; SSE-NEXT: movdqa {{.*#+}} xmm10 = [65535,65535,65535,0,0,0,65535,65535]
+; SSE-NEXT: movdqa %xmm10, %xmm2
; SSE-NEXT: pandn %xmm0, %xmm2
-; SSE-NEXT: movdqa (%rcx), %xmm1
-; SSE-NEXT: movdqa %xmm1, %xmm0
+; SSE-NEXT: movdqa (%rcx), %xmm8
+; SSE-NEXT: movdqa %xmm8, %xmm0
; SSE-NEXT: psrld $16, %xmm0
-; SSE-NEXT: movdqa (%rdx), %xmm4
-; SSE-NEXT: movdqa %xmm4, %xmm3
-; SSE-NEXT: movdqa %xmm4, %xmm14
+; SSE-NEXT: movdqa (%rdx), %xmm1
+; SSE-NEXT: movdqa %xmm1, %xmm3
; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
; SSE-NEXT: movdqa {{.*#+}} xmm15 = [65535,0,0,65535,65535,65535,65535,65535]
; SSE-NEXT: movdqa %xmm15, %xmm4
; SSE-NEXT: pandn %xmm3, %xmm4
-; SSE-NEXT: movdqa (%rdi), %xmm11
+; SSE-NEXT: movdqa (%rdi), %xmm12
; SSE-NEXT: movdqa (%rsi), %xmm9
; SSE-NEXT: movdqa %xmm9, %xmm3
; SSE-NEXT: movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm11[0],xmm3[1],xmm11[1],xmm3[2],xmm11[2],xmm3[3],xmm11[3]
-; SSE-NEXT: movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm12[0],xmm3[1],xmm12[1],xmm3[2],xmm12[2],xmm3[3],xmm12[3]
+; SSE-NEXT: movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[2,2,2,2,4,5,6,7]
; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,5,4]
; SSE-NEXT: pand %xmm15, %xmm3
@@ -3661,61 +3665,62 @@ define void @store_i16_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; SSE-NEXT: movdqa {{.*#+}} xmm0 = [65535,0,0,0,65535,65535,65535,65535]
; SSE-NEXT: movdqa %xmm0, %xmm2
; SSE-NEXT: pandn %xmm3, %xmm2
-; SSE-NEXT: movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm14[2,2,2,2]
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm1[2,2,2,2]
; SSE-NEXT: movdqa %xmm6, %xmm4
; SSE-NEXT: pandn %xmm3, %xmm4
-; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm1[0,1,2,3,4,4,4,4]
+; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm8[0,1,2,3,4,4,4,4]
; SSE-NEXT: pand %xmm6, %xmm3
; SSE-NEXT: por %xmm4, %xmm3
-; SSE-NEXT: punpckhwd {{.*#+}} xmm11 = xmm11[4],xmm9[4],xmm11[5],xmm9[5],xmm11[6],xmm9[6],xmm11[7],xmm9[7]
-; SSE-NEXT: movdqa %xmm11, %xmm4
+; SSE-NEXT: punpckhwd {{.*#+}} xmm12 = xmm12[4],xmm9[4],xmm12[5],xmm9[5],xmm12[6],xmm9[6],xmm12[7],xmm9[7]
+; SSE-NEXT: movdqa %xmm12, %xmm4
; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[0,1],xmm3[3,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm1[3,3,3,3,4,5,6,7]
-; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm8[3,3,3,3,4,5,6,7]
+; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[0,1],xmm4[0,2]
; SSE-NEXT: andps %xmm0, %xmm3
; SSE-NEXT: orps %xmm2, %xmm3
; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm7[2,3,2,3]
-; SSE-NEXT: movdqa {{.*#+}} xmm0 = [65535,0,65535,65535,65535,65535,65535,65535]
-; SSE-NEXT: movdqa %xmm0, %xmm3
+; SSE-NEXT: movdqa {{.*#+}} xmm4 = [65535,0,65535,65535,65535,65535,65535,65535]
+; SSE-NEXT: movdqa %xmm4, %xmm3
; SSE-NEXT: pandn %xmm2, %xmm3
-; SSE-NEXT: movdqa %xmm13, %xmm2
-; SSE-NEXT: punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm12[4],xmm2[5],xmm12[5],xmm2[6],xmm12[6],xmm2[7],xmm12[7]
+; SSE-NEXT: movdqa %xmm14, %xmm2
+; SSE-NEXT: punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm13[4],xmm2[5],xmm13[5],xmm2[6],xmm13[6],xmm2[7],xmm13[7]
; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[2,2,2,2,4,5,6,7]
; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,5,4]
-; SSE-NEXT: pand %xmm0, %xmm2
+; SSE-NEXT: pand %xmm4, %xmm2
; SSE-NEXT: por %xmm3, %xmm2
-; SSE-NEXT: movdqa %xmm14, %xmm4
-; SSE-NEXT: punpckhwd {{.*#+}} xmm4 = xmm4[4],xmm1[4],xmm4[5],xmm1[5],xmm4[6],xmm1[6],xmm4[7],xmm1[7]
+; SSE-NEXT: movdqa %xmm1, %xmm4
+; SSE-NEXT: punpckhwd {{.*#+}} xmm4 = xmm4[4],xmm8[4],xmm4[5],xmm8[5],xmm4[6],xmm8[6],xmm4[7],xmm8[7]
; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm11, %xmm3
+; SSE-NEXT: movdqa %xmm12, %xmm3
; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[2,2],xmm4[2,3]
; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[1,2],xmm2[0,3]
; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa 16(%rax), %xmm7
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm7[0,1,0,1]
-; SSE-NEXT: movdqa %xmm8, %xmm5
+; SSE-NEXT: movdqa {{.*#+}} xmm8 = [65535,65535,65535,65535,65535,0,65535,65535]
; SSE-NEXT: movdqa %xmm8, %xmm3
; SSE-NEXT: pandn %xmm2, %xmm3
; SSE-NEXT: movdqa 16(%r8), %xmm10
-; SSE-NEXT: movdqa 16(%r9), %xmm8
+; SSE-NEXT: movdqa 16(%r9), %xmm5
; SSE-NEXT: movdqa %xmm10, %xmm4
; SSE-NEXT: movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm8[0],xmm4[1],xmm8[1],xmm4[2],xmm8[2],xmm4[3],xmm8[3]
-; SSE-NEXT: movdqa %xmm8, (%rsp) # 16-byte Spill
+; SSE-NEXT: punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm5[0],xmm4[1],xmm5[1],xmm4[2],xmm5[2],xmm4[3],xmm5[3]
+; SSE-NEXT: movdqa %xmm5, (%rsp) # 16-byte Spill
; SSE-NEXT: movdqa %xmm4, %xmm2
-; SSE-NEXT: movdqa %xmm4, %xmm12
+; SSE-NEXT: movdqa %xmm4, %xmm11
; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pslldq {{.*#+}} xmm2 = zero,zero,xmm2[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
-; SSE-NEXT: pand %xmm5, %xmm2
+; SSE-NEXT: pand %xmm8, %xmm2
; SSE-NEXT: por %xmm3, %xmm2
; SSE-NEXT: movdqa {{.*#+}} xmm1 = [65535,65535,65535,0,0,0,65535,65535]
; SSE-NEXT: movdqa %xmm1, %xmm3
; SSE-NEXT: pandn %xmm2, %xmm3
-; SSE-NEXT: movdqa 16(%rcx), %xmm5
-; SSE-NEXT: movdqa %xmm5, %xmm2
+; SSE-NEXT: movdqa 16(%rcx), %xmm0
+; SSE-NEXT: movdqa %xmm0, %xmm2
+; SSE-NEXT: movdqa %xmm0, %xmm8
; SSE-NEXT: psrld $16, %xmm2
; SSE-NEXT: movdqa 16(%rdx), %xmm0
; SSE-NEXT: movdqa %xmm0, %xmm4
@@ -3741,7 +3746,7 @@ define void @store_i16_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; SSE-NEXT: movdqa {{.*#+}} xmm2 = [65535,65535,65535,0,65535,65535,65535,65535]
; SSE-NEXT: pandn %xmm7, %xmm2
; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm12, %xmm3
+; SSE-NEXT: movdqa %xmm11, %xmm3
; SSE-NEXT: psrldq {{.*#+}} xmm3 = xmm3[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
; SSE-NEXT: por %xmm2, %xmm3
; SSE-NEXT: movdqa {{.*#+}} xmm0 = [65535,0,0,0,65535,65535,65535,65535]
@@ -3751,14 +3756,14 @@ define void @store_i16_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm9[2,2,2,2]
; SSE-NEXT: movdqa %xmm6, %xmm4
; SSE-NEXT: pandn %xmm3, %xmm4
-; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm5[0,1,2,3,4,4,4,4]
+; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm8[0,1,2,3,4,4,4,4]
; SSE-NEXT: pand %xmm6, %xmm3
; SSE-NEXT: por %xmm4, %xmm3
; SSE-NEXT: punpckhwd {{.*#+}} xmm13 = xmm13[4],xmm14[4],xmm13[5],xmm14[5],xmm13[6],xmm14[6],xmm13[7],xmm14[7]
; SSE-NEXT: movdqa %xmm13, %xmm4
; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[0,1],xmm3[3,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm5[3,3,3,3,4,5,6,7]
-; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm8[3,3,3,3,4,5,6,7]
+; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[0,1],xmm4[0,2]
; SSE-NEXT: andps %xmm0, %xmm3
; SSE-NEXT: orps %xmm2, %xmm3
@@ -3767,13 +3772,13 @@ define void @store_i16_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; SSE-NEXT: movdqa {{.*#+}} xmm0 = [65535,0,65535,65535,65535,65535,65535,65535]
; SSE-NEXT: movdqa %xmm0, %xmm3
; SSE-NEXT: pandn %xmm2, %xmm3
-; SSE-NEXT: punpckhwd {{.*#+}} xmm8 = xmm8[4],xmm10[4],xmm8[5],xmm10[5],xmm8[6],xmm10[6],xmm8[7],xmm10[7]
-; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm8[2,2,2,2,4,5,6,7]
+; SSE-NEXT: punpckhwd {{.*#+}} xmm5 = xmm5[4],xmm10[4],xmm5[5],xmm10[5],xmm5[6],xmm10[6],xmm5[7],xmm10[7]
+; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm5[2,2,2,2,4,5,6,7]
; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,5,4]
; SSE-NEXT: pand %xmm0, %xmm2
; SSE-NEXT: por %xmm3, %xmm2
; SSE-NEXT: movdqa %xmm9, %xmm4
-; SSE-NEXT: punpckhwd {{.*#+}} xmm4 = xmm4[4],xmm5[4],xmm4[5],xmm5[5],xmm4[6],xmm5[6],xmm4[7],xmm5[7]
+; SSE-NEXT: punpckhwd {{.*#+}} xmm4 = xmm4[4],xmm8[4],xmm4[5],xmm8[5],xmm4[6],xmm8[6],xmm4[7],xmm8[7]
; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa %xmm13, %xmm3
; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[2,2],xmm4[2,3]
@@ -3799,8 +3804,8 @@ define void @store_i16_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; SSE-NEXT: movdqa {{.*#+}} xmm5 = [65535,65535,65535,0,0,0,65535,65535]
; SSE-NEXT: movdqa %xmm5, %xmm3
; SSE-NEXT: pandn %xmm2, %xmm3
-; SSE-NEXT: movdqa 32(%rcx), %xmm12
-; SSE-NEXT: movdqa %xmm12, %xmm2
+; SSE-NEXT: movdqa 32(%rcx), %xmm11
+; SSE-NEXT: movdqa %xmm11, %xmm2
; SSE-NEXT: psrld $16, %xmm2
; SSE-NEXT: movdqa 32(%rdx), %xmm0
; SSE-NEXT: movdqa %xmm0, %xmm4
@@ -3827,7 +3832,7 @@ define void @store_i16_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; SSE-NEXT: movdqa %xmm14, %xmm2
; SSE-NEXT: psrldq {{.*#+}} xmm2 = xmm2[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
; SSE-NEXT: por %xmm3, %xmm2
-; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm12[0,1,2,3,4,4,4,4]
+; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm11[0,1,2,3,4,4,4,4]
; SSE-NEXT: pand %xmm6, %xmm3
; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm8[2,2,2,2]
@@ -3837,7 +3842,7 @@ define void @store_i16_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; SSE-NEXT: movdqa %xmm1, %xmm3
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[0,1],xmm6[3,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm4 = xmm12[3,3,3,3,4,5,6,7]
+; SSE-NEXT: pshuflw {{.*#+}} xmm4 = xmm11[3,3,3,3,4,5,6,7]
; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[0,1],xmm3[0,2]
; SSE-NEXT: movaps {{.*#+}} xmm5 = [65535,0,0,0,65535,65535,65535,65535]
; SSE-NEXT: andps %xmm5, %xmm4
@@ -3855,7 +3860,7 @@ define void @store_i16_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; SSE-NEXT: movdqa %xmm0, %xmm9
; SSE-NEXT: por %xmm3, %xmm2
; SSE-NEXT: movdqa %xmm8, %xmm0
-; SSE-NEXT: punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm12[4],xmm0[5],xmm12[5],xmm0[6],xmm12[6],xmm0[7],xmm12[7]
+; SSE-NEXT: punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm11[4],xmm0[5],xmm11[5],xmm0[6],xmm11[6],xmm0[7],xmm11[7]
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[2,2],xmm0[2,3]
; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[1,2],xmm2[0,3]
@@ -3930,7 +3935,7 @@ define void @store_i16_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; SSE-NEXT: movdqa {{.*#+}} xmm6 = [65535,65535,65535,0,0,65535,65535,65535]
; SSE-NEXT: movdqa %xmm6, %xmm2
; SSE-NEXT: pandn %xmm0, %xmm2
-; SSE-NEXT: movdqa %xmm11, %xmm0
+; SSE-NEXT: movdqa %xmm12, %xmm0
; SSE-NEXT: psrldq {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
; SSE-NEXT: por %xmm2, %xmm0
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
@@ -3986,11 +3991,11 @@ define void @store_i16_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; SSE-NEXT: movdqa %xmm1, (%rsp) # 16-byte Spill
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: movdqa %xmm1, %xmm0
-; SSE-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm12[0],xmm1[1],xmm12[1],xmm1[2],xmm12[2],xmm1[3],xmm12[3]
+; SSE-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm11[0],xmm1[1],xmm11[1],xmm1[2],xmm11[2],xmm1[3],xmm11[3]
; SSE-NEXT: movdqa %xmm1, %xmm14
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: psrlq $48, %xmm12
-; SSE-NEXT: punpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm12[1]
+; SSE-NEXT: psrlq $48, %xmm11
+; SSE-NEXT: punpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm11[1]
; SSE-NEXT: movdqa %xmm6, %xmm1
; SSE-NEXT: pandn %xmm0, %xmm1
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
@@ -4031,10 +4036,10 @@ define void @store_i16_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[2,2],xmm5[1,1]
; SSE-NEXT: movaps %xmm5, %xmm3
-; SSE-NEXT: movaps {{.*#+}} xmm12 = [65535,65535,0,0,0,65535,65535,65535]
-; SSE-NEXT: movaps %xmm12, %xmm1
+; SSE-NEXT: movaps {{.*#+}} xmm11 = [65535,65535,0,0,0,65535,65535,65535]
+; SSE-NEXT: movaps %xmm11, %xmm1
; SSE-NEXT: andnps %xmm2, %xmm1
-; SSE-NEXT: pand %xmm12, %xmm0
+; SSE-NEXT: pand %xmm11, %xmm0
; SSE-NEXT: orps %xmm0, %xmm1
; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
@@ -4054,9 +4059,9 @@ define void @store_i16_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; SSE-NEXT: por %xmm2, %xmm0
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT: shufps {{.*#+}} xmm2 = xmm2[2,2],xmm10[1,1]
-; SSE-NEXT: movaps %xmm12, %xmm1
+; SSE-NEXT: movaps %xmm11, %xmm1
; SSE-NEXT: andnps %xmm2, %xmm1
-; SSE-NEXT: pand %xmm12, %xmm0
+; SSE-NEXT: pand %xmm11, %xmm0
; SSE-NEXT: orps %xmm0, %xmm1
; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
@@ -4078,9 +4083,9 @@ define void @store_i16_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; SSE-NEXT: movaps %xmm9, %xmm1
; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[2,2],xmm7[1,1]
; SSE-NEXT: movaps %xmm7, %xmm5
-; SSE-NEXT: movaps %xmm12, %xmm10
+; SSE-NEXT: movaps %xmm11, %xmm10
; SSE-NEXT: andnps %xmm1, %xmm10
-; SSE-NEXT: pand %xmm12, %xmm0
+; SSE-NEXT: pand %xmm11, %xmm0
; SSE-NEXT: orps %xmm0, %xmm10
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: movdqa %xmm1, %xmm0
@@ -4099,12 +4104,12 @@ define void @store_i16_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,1,2,3]
; SSE-NEXT: pand %xmm4, %xmm0
; SSE-NEXT: por %xmm1, %xmm0
-; SSE-NEXT: pand %xmm12, %xmm0
+; SSE-NEXT: pand %xmm11, %xmm0
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[2,2],xmm14[1,1]
-; SSE-NEXT: andnps %xmm1, %xmm12
-; SSE-NEXT: orps %xmm0, %xmm12
+; SSE-NEXT: andnps %xmm1, %xmm11
+; SSE-NEXT: orps %xmm0, %xmm11
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: shufps $42, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = xmm1[2,2],mem[2,0]
@@ -4125,20 +4130,20 @@ define void @store_i16_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; SSE-NEXT: pslldq {{.*#+}} xmm1 = zero,zero,zero,zero,zero,zero,xmm1[0,1,2,3,4,5,6,7,8,9]
; SSE-NEXT: movdqa %xmm4, %xmm0
; SSE-NEXT: pandn %xmm1, %xmm0
-; SSE-NEXT: pslldq {{.*#+}} xmm11 = zero,zero,xmm11[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
-; SSE-NEXT: pand %xmm4, %xmm11
-; SSE-NEXT: por %xmm0, %xmm11
+; SSE-NEXT: pslldq {{.*#+}} xmm12 = zero,zero,xmm12[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
+; SSE-NEXT: pand %xmm4, %xmm12
+; SSE-NEXT: por %xmm0, %xmm12
; SSE-NEXT: movdqa {{.*#+}} xmm2 = [65535,65535,65535,0,0,0,0,65535]
; SSE-NEXT: movdqa %xmm2, %xmm1
-; SSE-NEXT: pandn %xmm11, %xmm1
-; SSE-NEXT: movaps %xmm3, %xmm11
-; SSE-NEXT: shufpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
-; SSE-NEXT: # xmm11 = xmm11[1],mem[0]
+; SSE-NEXT: pandn %xmm12, %xmm1
+; SSE-NEXT: movaps %xmm3, %xmm12
+; SSE-NEXT: shufpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
+; SSE-NEXT: # xmm12 = xmm12[1],mem[0]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: punpckhwd {{.*#+}} xmm0 = xmm0[4,4,5,5,6,6,7,7]
-; SSE-NEXT: shufps {{.*#+}} xmm11 = xmm11[2,0],xmm0[2,1]
-; SSE-NEXT: andps %xmm2, %xmm11
-; SSE-NEXT: orps %xmm1, %xmm11
+; SSE-NEXT: shufps {{.*#+}} xmm12 = xmm12[2,0],xmm0[2,1]
+; SSE-NEXT: andps %xmm2, %xmm12
+; SSE-NEXT: orps %xmm1, %xmm12
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: pslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3,4,5,6,7,8,9]
; SSE-NEXT: movdqa %xmm4, %xmm1
@@ -4229,10 +4234,10 @@ define void @store_i16_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; SSE-NEXT: movdqa %xmm1, (%rax)
; SSE-NEXT: movdqa %xmm2, 288(%rax)
; SSE-NEXT: movaps %xmm13, 176(%rax)
-; SSE-NEXT: movaps %xmm11, 64(%rax)
+; SSE-NEXT: movaps %xmm12, 64(%rax)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 416(%rax)
-; SSE-NEXT: movaps %xmm12, 368(%rax)
+; SSE-NEXT: movaps %xmm11, 368(%rax)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 352(%rax)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
@@ -7573,751 +7578,749 @@ define void @store_i16_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; SSE: # %bb.0:
; SSE-NEXT: subq $1640, %rsp # imm = 0x668
; SSE-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; SSE-NEXT: movdqa 112(%rdi), %xmm15
-; SSE-NEXT: movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 112(%rsi), %xmm2
+; SSE-NEXT: movdqa 96(%rdi), %xmm13
+; SSE-NEXT: movdqa %xmm13, (%rsp) # 16-byte Spill
+; SSE-NEXT: movdqa 96(%rsi), %xmm2
; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 96(%rdx), %xmm5
+; SSE-NEXT: movdqa 112(%rdx), %xmm11
+; SSE-NEXT: movdqa 96(%rdx), %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 96(%rcx), %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 96(%r8), %xmm5
; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 112(%rdx), %xmm10
-; SSE-NEXT: movdqa 96(%rcx), %xmm12
-; SSE-NEXT: movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 112(%rcx), %xmm11
-; SSE-NEXT: movdqa 112(%r8), %xmm9
-; SSE-NEXT: movdqa 112(%r9), %xmm8
-; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movaps 112(%rax), %xmm7
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm10[2,2,2,2]
-; SSE-NEXT: movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{.*#+}} xmm13 = [65535,65535,65535,65535,65535,65535,0,65535]
+; SSE-NEXT: movdqa 96(%r9), %xmm7
+; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 96(%rax), %xmm4
+; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: psrlq $48, %xmm0
+; SSE-NEXT: punpckhqdq {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT: movdqa {{.*#+}} xmm3 = [65535,65535,65535,0,0,65535,65535,65535]
+; SSE-NEXT: movdqa %xmm3, %xmm0
+; SSE-NEXT: movdqa %xmm3, %xmm6
+; SSE-NEXT: pandn %xmm1, %xmm0
; SSE-NEXT: movdqa %xmm13, %xmm1
-; SSE-NEXT: pandn %xmm0, %xmm1
-; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm11[3,3,3,3,4,5,6,7]
+; SSE-NEXT: punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: psrldq {{.*#+}} xmm1 = xmm1[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; SSE-NEXT: por %xmm0, %xmm1
+; SSE-NEXT: movdqa {{.*#+}} xmm8 = [65535,65535,65535,65535,65535,0,65535,65535]
+; SSE-NEXT: movdqa %xmm1, %xmm0
+; SSE-NEXT: pand %xmm8, %xmm0
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm5[2,2,3,3]
+; SSE-NEXT: movdqa %xmm8, %xmm3
+; SSE-NEXT: movdqa %xmm8, %xmm5
+; SSE-NEXT: pandn %xmm2, %xmm3
+; SSE-NEXT: por %xmm0, %xmm3
+; SSE-NEXT: movdqa %xmm7, %xmm0
+; SSE-NEXT: psrld $16, %xmm0
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[3,0],xmm3[2,0]
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,0]
+; SSE-NEXT: movaps {{.*#+}} xmm10 = [0,65535,65535,65535,65535,65535,65535,0]
+; SSE-NEXT: andps %xmm10, %xmm1
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm4[3,3,3,3]
+; SSE-NEXT: movaps %xmm10, %xmm2
+; SSE-NEXT: andnps %xmm0, %xmm2
+; SSE-NEXT: orps %xmm1, %xmm2
+; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm11[2,2,2,2]
+; SSE-NEXT: movdqa {{.*#+}} xmm14 = [65535,65535,65535,65535,65535,65535,0,65535]
+; SSE-NEXT: movdqa %xmm14, %xmm1
+; SSE-NEXT: pandn %xmm0, %xmm1
+; SSE-NEXT: movdqa 112(%rcx), %xmm9
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm9[3,3,3,3,4,5,6,7]
+; SSE-NEXT: movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,4,4,4]
-; SSE-NEXT: pand %xmm13, %xmm0
+; SSE-NEXT: pand %xmm14, %xmm0
; SSE-NEXT: por %xmm1, %xmm0
-; SSE-NEXT: punpckhwd {{.*#+}} xmm15 = xmm15[4],xmm2[4],xmm15[5],xmm2[5],xmm15[6],xmm2[6],xmm15[7],xmm2[7]
-; SSE-NEXT: movdqa %xmm15, %xmm1
-; SSE-NEXT: movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 112(%r8), %xmm8
+; SSE-NEXT: movdqa {{.*#+}} xmm15 = [65535,0,65535,65535,65535,65535,65535,65535]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm8[1,1,1,1]
+; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm15, %xmm2
+; SSE-NEXT: pandn %xmm1, %xmm2
+; SSE-NEXT: movdqa 112(%rdi), %xmm3
+; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 112(%rsi), %xmm1
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: punpckhwd {{.*#+}} xmm3 = xmm3[4],xmm1[4],xmm3[5],xmm1[5],xmm3[6],xmm1[6],xmm3[7],xmm1[7]
+; SSE-NEXT: movdqa %xmm3, %xmm1
+; SSE-NEXT: movdqa %xmm3, %xmm4
+; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,0],xmm0[3,0]
-; SSE-NEXT: movdqa {{.*#+}} xmm3 = [65535,0,65535,65535,65535,65535,65535,65535]
-; SSE-NEXT: pand %xmm3, %xmm0
-; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm9[1,1,1,1]
-; SSE-NEXT: movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pandn %xmm2, %xmm3
-; SSE-NEXT: por %xmm0, %xmm3
-; SSE-NEXT: movdqa %xmm8, %xmm0
+; SSE-NEXT: pand %xmm15, %xmm0
+; SSE-NEXT: por %xmm0, %xmm2
+; SSE-NEXT: movdqa 112(%r9), %xmm0
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: psrld $16, %xmm0
-; SSE-NEXT: movdqa %xmm0, %xmm4
-; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[1,0],xmm3[0,0]
-; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[2,0],xmm1[0,2]
+; SSE-NEXT: movdqa %xmm0, %xmm3
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[1,0],xmm2[0,0]
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[2,0],xmm1[0,2]
; SSE-NEXT: movaps {{.*#+}} xmm1 = [65535,65535,65535,0,65535,65535,65535,65535]
-; SSE-NEXT: andps %xmm1, %xmm4
+; SSE-NEXT: andps %xmm1, %xmm3
+; SSE-NEXT: movaps 112(%rax), %xmm7
; SSE-NEXT: andnps %xmm7, %xmm1
; SSE-NEXT: movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: orps %xmm4, %xmm1
+; SSE-NEXT: orps %xmm3, %xmm1
; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm12, %xmm1
+; SSE-NEXT: movdqa %xmm9, %xmm1
; SSE-NEXT: psrlq $48, %xmm1
-; SSE-NEXT: movdqa %xmm5, %xmm3
+; SSE-NEXT: movdqa %xmm11, %xmm3
; SSE-NEXT: punpckhqdq {{.*#+}} xmm3 = xmm3[1],xmm1[1]
-; SSE-NEXT: movdqa {{.*#+}} xmm2 = [65535,65535,65535,0,0,65535,65535,65535]
-; SSE-NEXT: movdqa %xmm2, %xmm1
+; SSE-NEXT: movdqa %xmm6, %xmm1
; SSE-NEXT: pandn %xmm3, %xmm1
-; SSE-NEXT: movdqa 96(%rdi), %xmm4
-; SSE-NEXT: movdqa %xmm4, (%rsp) # 16-byte Spill
-; SSE-NEXT: movdqa 96(%rsi), %xmm3
-; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpckhwd {{.*#+}} xmm4 = xmm4[4],xmm3[4],xmm4[5],xmm3[5],xmm4[6],xmm3[6],xmm4[7],xmm3[7]
-; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa %xmm4, %xmm3
; SSE-NEXT: psrldq {{.*#+}} xmm3 = xmm3[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
; SSE-NEXT: por %xmm1, %xmm3
-; SSE-NEXT: movdqa 96(%r8), %xmm1
-; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[2,2,3,3]
-; SSE-NEXT: movdqa {{.*#+}} xmm5 = [65535,65535,65535,65535,65535,0,65535,65535]
-; SSE-NEXT: movdqa %xmm5, %xmm4
-; SSE-NEXT: movdqa %xmm5, %xmm14
-; SSE-NEXT: pandn %xmm1, %xmm4
-; SSE-NEXT: por %xmm3, %xmm4
-; SSE-NEXT: movdqa 96(%r9), %xmm1
-; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: psrld $16, %xmm1
-; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[3,0],xmm4[2,0]
-; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[0,1],xmm1[2,0]
-; SSE-NEXT: movdqa 96(%rax), %xmm1
-; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{.*#+}} xmm5 = [0,65535,65535,65535,65535,65535,65535,0]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[3,3,3,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm8[2,2,3,3]
; SSE-NEXT: movdqa %xmm5, %xmm4
; SSE-NEXT: pandn %xmm1, %xmm4
-; SSE-NEXT: andps %xmm5, %xmm3
-; SSE-NEXT: por %xmm3, %xmm4
-; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm11, %xmm1
-; SSE-NEXT: psrlq $48, %xmm1
-; SSE-NEXT: movdqa %xmm10, %xmm3
-; SSE-NEXT: punpckhqdq {{.*#+}} xmm3 = xmm3[1],xmm1[1]
-; SSE-NEXT: movdqa %xmm2, %xmm1
-; SSE-NEXT: pandn %xmm3, %xmm1
-; SSE-NEXT: movdqa %xmm15, %xmm3
-; SSE-NEXT: psrldq {{.*#+}} xmm3 = xmm3[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; SSE-NEXT: por %xmm1, %xmm3
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm9[2,2,3,3]
-; SSE-NEXT: movdqa %xmm14, %xmm4
-; SSE-NEXT: pandn %xmm1, %xmm4
-; SSE-NEXT: por %xmm3, %xmm4
+; SSE-NEXT: movdqa %xmm3, %xmm1
+; SSE-NEXT: pand %xmm5, %xmm1
+; SSE-NEXT: por %xmm1, %xmm4
; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[3,0],xmm4[2,0]
; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[0,1],xmm0[2,0]
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm7[3,3,3,3]
+; SSE-NEXT: movaps %xmm10, %xmm1
+; SSE-NEXT: andnps %xmm0, %xmm1
+; SSE-NEXT: andps %xmm10, %xmm3
+; SSE-NEXT: orps %xmm3, %xmm1
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa (%rax), %xmm12
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm12[0,1,0,1]
; SSE-NEXT: movdqa %xmm5, %xmm1
; SSE-NEXT: pandn %xmm0, %xmm1
-; SSE-NEXT: andps %xmm5, %xmm3
-; SSE-NEXT: por %xmm3, %xmm1
-; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa (%rax), %xmm11
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm11[0,1,0,1]
-; SSE-NEXT: movdqa %xmm14, %xmm1
-; SSE-NEXT: pandn %xmm0, %xmm1
-; SSE-NEXT: movdqa (%r8), %xmm8
-; SSE-NEXT: movdqa (%r9), %xmm7
-; SSE-NEXT: movdqa %xmm8, %xmm3
-; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm7[0],xmm3[1],xmm7[1],xmm3[2],xmm7[2],xmm3[3],xmm7[3]
-; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa (%r8), %xmm11
+; SSE-NEXT: movdqa (%r9), %xmm9
+; SSE-NEXT: movdqa %xmm11, %xmm3
+; SSE-NEXT: movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm9[0],xmm3[1],xmm9[1],xmm3[2],xmm9[2],xmm3[3],xmm9[3]
+; SSE-NEXT: movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa %xmm3, %xmm0
-; SSE-NEXT: movdqa %xmm3, %xmm6
+; SSE-NEXT: movdqa %xmm3, %xmm7
; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pslldq {{.*#+}} xmm0 = zero,zero,xmm0[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
-; SSE-NEXT: pand %xmm14, %xmm0
+; SSE-NEXT: pand %xmm5, %xmm0
; SSE-NEXT: por %xmm1, %xmm0
-; SSE-NEXT: movdqa {{.*#+}} xmm2 = [65535,65535,65535,0,0,0,65535,65535]
-; SSE-NEXT: movdqa %xmm2, %xmm1
-; SSE-NEXT: movdqa %xmm2, %xmm10
+; SSE-NEXT: movdqa {{.*#+}} xmm10 = [65535,65535,65535,0,0,0,65535,65535]
+; SSE-NEXT: movdqa %xmm10, %xmm1
; SSE-NEXT: pandn %xmm0, %xmm1
-; SSE-NEXT: movdqa (%rcx), %xmm5
-; SSE-NEXT: movdqa %xmm5, %xmm0
+; SSE-NEXT: movdqa (%rcx), %xmm8
+; SSE-NEXT: movdqa %xmm8, %xmm0
; SSE-NEXT: psrld $16, %xmm0
-; SSE-NEXT: movdqa (%rdx), %xmm2
-; SSE-NEXT: movdqa %xmm2, %xmm3
+; SSE-NEXT: movdqa (%rdx), %xmm5
+; SSE-NEXT: movdqa %xmm5, %xmm3
; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
-; SSE-NEXT: movdqa {{.*#+}} xmm0 = [65535,0,0,65535,65535,65535,65535,65535]
-; SSE-NEXT: movdqa %xmm0, %xmm4
-; SSE-NEXT: movdqa %xmm0, %xmm12
+; SSE-NEXT: movdqa {{.*#+}} xmm2 = [65535,0,0,65535,65535,65535,65535,65535]
+; SSE-NEXT: movdqa %xmm2, %xmm4
; SSE-NEXT: pandn %xmm3, %xmm4
-; SSE-NEXT: movdqa (%rdi), %xmm9
+; SSE-NEXT: movdqa (%rdi), %xmm13
; SSE-NEXT: movdqa (%rsi), %xmm0
; SSE-NEXT: movdqa %xmm0, %xmm3
-; SSE-NEXT: movdqa %xmm0, %xmm15
+; SSE-NEXT: movdqa %xmm0, %xmm6
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm9[0],xmm3[1],xmm9[1],xmm3[2],xmm9[2],xmm3[3],xmm9[3]
-; SSE-NEXT: movdqa %xmm9, %xmm0
-; SSE-NEXT: movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm13[0],xmm3[1],xmm13[1],xmm3[2],xmm13[2],xmm3[3],xmm13[3]
+; SSE-NEXT: movdqa %xmm13, %xmm0
+; SSE-NEXT: movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[2,2,2,2,4,5,6,7]
; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,5,4]
-; SSE-NEXT: pand %xmm12, %xmm3
-; SSE-NEXT: movdqa %xmm12, %xmm9
+; SSE-NEXT: pand %xmm2, %xmm3
; SSE-NEXT: por %xmm4, %xmm3
; SSE-NEXT: pand %xmm10, %xmm3
; SSE-NEXT: por %xmm1, %xmm3
; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{.*#+}} xmm1 = [65535,65535,65535,0,65535,65535,65535,65535]
-; SSE-NEXT: movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pandn %xmm11, %xmm1
-; SSE-NEXT: movdqa %xmm6, %xmm3
+; SSE-NEXT: movdqa {{.*#+}} xmm13 = [65535,65535,65535,0,65535,65535,65535,65535]
+; SSE-NEXT: movdqa %xmm13, %xmm1
+; SSE-NEXT: movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pandn %xmm12, %xmm1
+; SSE-NEXT: movdqa %xmm7, %xmm3
; SSE-NEXT: psrldq {{.*#+}} xmm3 = xmm3[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
; SSE-NEXT: por %xmm1, %xmm3
-; SSE-NEXT: movdqa {{.*#+}} xmm12 = [65535,0,0,0,65535,65535,65535,65535]
-; SSE-NEXT: movdqa %xmm12, %xmm1
+; SSE-NEXT: movdqa {{.*#+}} xmm4 = [65535,0,0,0,65535,65535,65535,65535]
+; SSE-NEXT: movdqa %xmm4, %xmm1
+; SSE-NEXT: movdqa %xmm4, %xmm10
; SSE-NEXT: pandn %xmm3, %xmm1
-; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm2[2,2,2,2]
-; SSE-NEXT: movdqa %xmm13, %xmm4
+; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm5[2,2,2,2]
+; SSE-NEXT: movdqa %xmm14, %xmm4
; SSE-NEXT: pandn %xmm3, %xmm4
-; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm5[0,1,2,3,4,4,4,4]
-; SSE-NEXT: pand %xmm13, %xmm3
+; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm8[0,1,2,3,4,4,4,4]
+; SSE-NEXT: pand %xmm14, %xmm3
; SSE-NEXT: por %xmm4, %xmm3
-; SSE-NEXT: punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm15[4],xmm0[5],xmm15[5],xmm0[6],xmm15[6],xmm0[7],xmm15[7]
+; SSE-NEXT: punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm6[4],xmm0[5],xmm6[5],xmm0[6],xmm6[6],xmm0[7],xmm6[7]
; SSE-NEXT: movdqa %xmm0, %xmm4
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[0,1],xmm3[3,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm5[3,3,3,3,4,5,6,7]
-; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm8[3,3,3,3,4,5,6,7]
+; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[0,1],xmm4[0,2]
-; SSE-NEXT: andps %xmm12, %xmm3
+; SSE-NEXT: andps %xmm10, %xmm3
; SSE-NEXT: orps %xmm1, %xmm3
; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm11[2,3,2,3]
-; SSE-NEXT: movdqa {{.*#+}} xmm13 = [65535,0,65535,65535,65535,65535,65535,65535]
-; SSE-NEXT: movdqa %xmm13, %xmm3
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm12[2,3,2,3]
+; SSE-NEXT: movdqa %xmm15, %xmm3
; SSE-NEXT: pandn %xmm1, %xmm3
-; SSE-NEXT: punpckhwd {{.*#+}} xmm7 = xmm7[4],xmm8[4],xmm7[5],xmm8[5],xmm7[6],xmm8[6],xmm7[7],xmm8[7]
-; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm7[2,2,2,2,4,5,6,7]
+; SSE-NEXT: punpckhwd {{.*#+}} xmm9 = xmm9[4],xmm11[4],xmm9[5],xmm11[5],xmm9[6],xmm11[6],xmm9[7],xmm11[7]
+; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm9[2,2,2,2,4,5,6,7]
; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,5,4]
-; SSE-NEXT: pand %xmm13, %xmm1
-; SSE-NEXT: movdqa %xmm13, %xmm15
+; SSE-NEXT: pand %xmm15, %xmm1
; SSE-NEXT: por %xmm3, %xmm1
-; SSE-NEXT: punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm5[4],xmm2[5],xmm5[5],xmm2[6],xmm5[6],xmm2[7],xmm5[7]
-; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,2],xmm2[2,3]
-; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[1,2],xmm1[0,3]
-; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 16(%rax), %xmm11
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm11[0,1,0,1]
-; SSE-NEXT: movdqa %xmm14, %xmm2
-; SSE-NEXT: movdqa %xmm14, %xmm3
-; SSE-NEXT: pandn %xmm1, %xmm3
-; SSE-NEXT: movdqa 16(%r8), %xmm14
-; SSE-NEXT: movdqa 16(%r9), %xmm13
-; SSE-NEXT: movdqa %xmm14, %xmm5
-; SSE-NEXT: movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm13[0],xmm5[1],xmm13[1],xmm5[2],xmm13[2],xmm5[3],xmm13[3]
-; SSE-NEXT: movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm5, %xmm1
+; SSE-NEXT: punpckhwd {{.*#+}} xmm5 = xmm5[4],xmm8[4],xmm5[5],xmm8[5],xmm5[6],xmm8[6],xmm5[7],xmm8[7]
; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm0, %xmm3
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[2,2],xmm5[2,3]
+; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[1,2],xmm1[0,3]
+; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 16(%rax), %xmm12
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm12[0,1,0,1]
+; SSE-NEXT: movdqa {{.*#+}} xmm2 = [65535,65535,65535,65535,65535,0,65535,65535]
+; SSE-NEXT: movdqa %xmm2, %xmm3
+; SSE-NEXT: pandn %xmm1, %xmm3
+; SSE-NEXT: movdqa 16(%r8), %xmm9
+; SSE-NEXT: movdqa 16(%r9), %xmm8
+; SSE-NEXT: movdqa %xmm9, %xmm4
+; SSE-NEXT: movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm8[0],xmm4[1],xmm8[1],xmm4[2],xmm8[2],xmm4[3],xmm8[3]
+; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm4, %xmm1
+; SSE-NEXT: movdqa %xmm4, %xmm0
+; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pslldq {{.*#+}} xmm1 = zero,zero,xmm1[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
; SSE-NEXT: pand %xmm2, %xmm1
; SSE-NEXT: por %xmm3, %xmm1
-; SSE-NEXT: movdqa %xmm10, %xmm3
+; SSE-NEXT: movdqa {{.*#+}} xmm11 = [65535,65535,65535,0,0,0,65535,65535]
+; SSE-NEXT: movdqa %xmm11, %xmm3
; SSE-NEXT: pandn %xmm1, %xmm3
-; SSE-NEXT: movdqa 16(%rcx), %xmm8
-; SSE-NEXT: movdqa %xmm8, %xmm1
+; SSE-NEXT: movdqa 16(%rcx), %xmm4
+; SSE-NEXT: movdqa %xmm4, %xmm1
+; SSE-NEXT: movdqa %xmm4, %xmm10
; SSE-NEXT: psrld $16, %xmm1
-; SSE-NEXT: movdqa 16(%rdx), %xmm7
-; SSE-NEXT: movdqa %xmm7, %xmm4
+; SSE-NEXT: movdqa 16(%rdx), %xmm5
+; SSE-NEXT: movdqa %xmm5, %xmm4
; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
-; SSE-NEXT: movdqa %xmm9, %xmm2
-; SSE-NEXT: movdqa %xmm9, %xmm1
+; SSE-NEXT: movdqa {{.*#+}} xmm2 = [65535,0,0,65535,65535,65535,65535,65535]
+; SSE-NEXT: movdqa %xmm2, %xmm1
; SSE-NEXT: pandn %xmm4, %xmm1
-; SSE-NEXT: movdqa 16(%rdi), %xmm0
+; SSE-NEXT: movdqa 16(%rdi), %xmm7
; SSE-NEXT: movdqa 16(%rsi), %xmm6
; SSE-NEXT: movdqa %xmm6, %xmm4
-; SSE-NEXT: movdqa %xmm6, %xmm9
; SSE-NEXT: movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1],xmm4[2],xmm0[2],xmm4[3],xmm0[3]
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm7[0],xmm4[1],xmm7[1],xmm4[2],xmm7[2],xmm4[3],xmm7[3]
+; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm4 = xmm4[2,2,2,2,4,5,6,7]
; SSE-NEXT: pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5,5,4]
; SSE-NEXT: pand %xmm2, %xmm4
; SSE-NEXT: por %xmm1, %xmm4
-; SSE-NEXT: pand %xmm10, %xmm4
-; SSE-NEXT: movdqa %xmm10, %xmm6
+; SSE-NEXT: pand %xmm11, %xmm4
; SSE-NEXT: por %xmm3, %xmm4
; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{.*#+}} xmm10 = [65535,65535,65535,0,65535,65535,65535,65535]
-; SSE-NEXT: movdqa %xmm10, %xmm1
-; SSE-NEXT: pandn %xmm11, %xmm1
-; SSE-NEXT: movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: psrldq {{.*#+}} xmm5 = xmm5[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; SSE-NEXT: por %xmm1, %xmm5
-; SSE-NEXT: movdqa %xmm12, %xmm1
-; SSE-NEXT: pandn %xmm5, %xmm1
-; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm7[2,2,2,2]
-; SSE-NEXT: movdqa {{.*#+}} xmm5 = [65535,65535,65535,65535,65535,65535,0,65535]
-; SSE-NEXT: movdqa %xmm5, %xmm4
+; SSE-NEXT: movdqa %xmm13, %xmm1
+; SSE-NEXT: pandn %xmm12, %xmm1
+; SSE-NEXT: movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: psrldq {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; SSE-NEXT: por %xmm1, %xmm0
+; SSE-NEXT: movdqa {{.*#+}} xmm13 = [65535,0,0,0,65535,65535,65535,65535]
+; SSE-NEXT: movdqa %xmm13, %xmm1
+; SSE-NEXT: pandn %xmm0, %xmm1
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm5[2,2,2,2]
+; SSE-NEXT: movdqa %xmm5, %xmm2
+; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa {{.*#+}} xmm0 = [65535,65535,65535,65535,65535,65535,0,65535]
+; SSE-NEXT: movdqa %xmm0, %xmm4
; SSE-NEXT: pandn %xmm3, %xmm4
-; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm8[0,1,2,3,4,4,4,4]
-; SSE-NEXT: pand %xmm5, %xmm3
+; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm10[0,1,2,3,4,4,4,4]
+; SSE-NEXT: pand %xmm0, %xmm3
; SSE-NEXT: por %xmm4, %xmm3
-; SSE-NEXT: punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm9[4],xmm0[5],xmm9[5],xmm0[6],xmm9[6],xmm0[7],xmm9[7]
-; SSE-NEXT: movdqa %xmm0, %xmm4
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: punpckhwd {{.*#+}} xmm7 = xmm7[4],xmm6[4],xmm7[5],xmm6[5],xmm7[6],xmm6[6],xmm7[7],xmm6[7]
+; SSE-NEXT: movdqa %xmm7, %xmm4
+; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[0,1],xmm3[3,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm8[3,3,3,3,4,5,6,7]
-; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm10[3,3,3,3,4,5,6,7]
+; SSE-NEXT: movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[0,1],xmm4[0,2]
-; SSE-NEXT: andps %xmm12, %xmm3
+; SSE-NEXT: andps %xmm13, %xmm3
; SSE-NEXT: orps %xmm1, %xmm3
; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm11[2,3,2,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm12[2,3,2,3]
; SSE-NEXT: movdqa %xmm15, %xmm3
; SSE-NEXT: pandn %xmm1, %xmm3
-; SSE-NEXT: movdqa %xmm13, %xmm1
-; SSE-NEXT: punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm14[4],xmm1[5],xmm14[5],xmm1[6],xmm14[6],xmm1[7],xmm14[7]
+; SSE-NEXT: movdqa %xmm8, %xmm1
+; SSE-NEXT: punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm9[4],xmm1[5],xmm9[5],xmm1[6],xmm9[6],xmm1[7],xmm9[7]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[2,2,2,2,4,5,6,7]
; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,5,4]
; SSE-NEXT: pand %xmm15, %xmm1
; SSE-NEXT: por %xmm3, %xmm1
-; SSE-NEXT: punpckhwd {{.*#+}} xmm7 = xmm7[4],xmm8[4],xmm7[5],xmm8[5],xmm7[6],xmm8[6],xmm7[7],xmm8[7]
-; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,2],xmm7[2,3]
-; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[1,2],xmm1[0,3]
-; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 32(%rax), %xmm7
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm7[0,1,0,1]
-; SSE-NEXT: movdqa {{.*#+}} xmm2 = [65535,65535,65535,65535,65535,0,65535,65535]
-; SSE-NEXT: movdqa %xmm2, %xmm3
+; SSE-NEXT: punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm10[4],xmm2[5],xmm10[5],xmm2[6],xmm10[6],xmm2[7],xmm10[7]
+; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: shufps {{.*#+}} xmm7 = xmm7[2,2],xmm2[2,3]
+; SSE-NEXT: shufps {{.*#+}} xmm7 = xmm7[1,2],xmm1[0,3]
+; SSE-NEXT: movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 32(%rax), %xmm5
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm5[0,1,0,1]
+; SSE-NEXT: movdqa {{.*#+}} xmm13 = [65535,65535,65535,65535,65535,0,65535,65535]
+; SSE-NEXT: movdqa %xmm13, %xmm3
; SSE-NEXT: pandn %xmm1, %xmm3
-; SSE-NEXT: movdqa 32(%r8), %xmm14
-; SSE-NEXT: movdqa 32(%r9), %xmm13
-; SSE-NEXT: movdqa %xmm14, %xmm0
-; SSE-NEXT: movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm13[0],xmm0[1],xmm13[1],xmm0[2],xmm13[2],xmm0[3],xmm13[3]
-; SSE-NEXT: movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 32(%r8), %xmm1
+; SSE-NEXT: movdqa 32(%r9), %xmm8
+; SSE-NEXT: movdqa %xmm1, %xmm0
+; SSE-NEXT: movdqa %xmm1, %xmm6
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1],xmm0[2],xmm8[2],xmm0[3],xmm8[3]
+; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa %xmm0, %xmm1
-; SSE-NEXT: movdqa %xmm0, %xmm11
+; SSE-NEXT: movdqa %xmm0, %xmm12
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pslldq {{.*#+}} xmm1 = zero,zero,xmm1[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
-; SSE-NEXT: pand %xmm2, %xmm1
+; SSE-NEXT: pand %xmm13, %xmm1
; SSE-NEXT: por %xmm3, %xmm1
-; SSE-NEXT: movdqa %xmm6, %xmm8
-; SSE-NEXT: movdqa %xmm6, %xmm3
+; SSE-NEXT: movdqa %xmm11, %xmm3
; SSE-NEXT: pandn %xmm1, %xmm3
-; SSE-NEXT: movdqa 32(%rcx), %xmm2
-; SSE-NEXT: movdqa %xmm2, %xmm1
+; SSE-NEXT: movdqa 32(%rcx), %xmm0
+; SSE-NEXT: movdqa %xmm0, %xmm1
+; SSE-NEXT: movdqa %xmm0, %xmm9
; SSE-NEXT: psrld $16, %xmm1
-; SSE-NEXT: movdqa 32(%rdx), %xmm5
-; SSE-NEXT: movdqa %xmm5, %xmm4
+; SSE-NEXT: movdqa 32(%rdx), %xmm0
+; SSE-NEXT: movdqa %xmm0, %xmm4
+; SSE-NEXT: movdqa %xmm0, %xmm14
; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
-; SSE-NEXT: movdqa {{.*#+}} xmm12 = [65535,0,0,65535,65535,65535,65535,65535]
-; SSE-NEXT: movdqa %xmm12, %xmm1
+; SSE-NEXT: movdqa {{.*#+}} xmm10 = [65535,0,0,65535,65535,65535,65535,65535]
+; SSE-NEXT: movdqa %xmm10, %xmm1
; SSE-NEXT: pandn %xmm4, %xmm1
-; SSE-NEXT: movdqa 32(%rdi), %xmm9
+; SSE-NEXT: movdqa 32(%rdi), %xmm7
; SSE-NEXT: movdqa 32(%rsi), %xmm0
; SSE-NEXT: movdqa %xmm0, %xmm4
-; SSE-NEXT: movdqa %xmm0, %xmm6
+; SSE-NEXT: movdqa %xmm0, %xmm2
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm9[0],xmm4[1],xmm9[1],xmm4[2],xmm9[2],xmm4[3],xmm9[3]
-; SSE-NEXT: movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm7[0],xmm4[1],xmm7[1],xmm4[2],xmm7[2],xmm4[3],xmm7[3]
+; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm4 = xmm4[2,2,2,2,4,5,6,7]
; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm4[0,1,2,3,4,5,5,4]
-; SSE-NEXT: pand %xmm12, %xmm0
+; SSE-NEXT: pand %xmm10, %xmm0
; SSE-NEXT: por %xmm1, %xmm0
-; SSE-NEXT: pand %xmm8, %xmm0
+; SSE-NEXT: pand %xmm11, %xmm0
; SSE-NEXT: por %xmm3, %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm10, %xmm1
-; SSE-NEXT: movdqa %xmm10, %xmm15
-; SSE-NEXT: pandn %xmm7, %xmm1
-; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm11, %xmm3
+; SSE-NEXT: movdqa {{.*#+}} xmm15 = [65535,65535,65535,0,65535,65535,65535,65535]
+; SSE-NEXT: movdqa %xmm15, %xmm1
+; SSE-NEXT: pandn %xmm5, %xmm1
+; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm12, %xmm3
; SSE-NEXT: psrldq {{.*#+}} xmm3 = xmm3[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
; SSE-NEXT: por %xmm1, %xmm3
; SSE-NEXT: movdqa {{.*#+}} xmm10 = [65535,0,0,0,65535,65535,65535,65535]
; SSE-NEXT: movdqa %xmm10, %xmm1
; SSE-NEXT: pandn %xmm3, %xmm1
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm5[2,2,2,2]
-; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm14[2,2,2,2]
+; SSE-NEXT: movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa {{.*#+}} xmm0 = [65535,65535,65535,65535,65535,65535,0,65535]
; SSE-NEXT: movdqa %xmm0, %xmm4
; SSE-NEXT: pandn %xmm3, %xmm4
-; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm2[0,1,2,3,4,4,4,4]
+; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm9[0,1,2,3,4,4,4,4]
; SSE-NEXT: pand %xmm0, %xmm3
; SSE-NEXT: por %xmm4, %xmm3
-; SSE-NEXT: movdqa %xmm9, %xmm0
-; SSE-NEXT: punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm6[4],xmm0[5],xmm6[5],xmm0[6],xmm6[6],xmm0[7],xmm6[7]
+; SSE-NEXT: movdqa %xmm7, %xmm0
+; SSE-NEXT: punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
; SSE-NEXT: movdqa %xmm0, %xmm4
-; SSE-NEXT: movdqa %xmm0, %xmm11
+; SSE-NEXT: movdqa %xmm0, %xmm12
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[0,1],xmm3[3,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm2[3,3,3,3,4,5,6,7]
-; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm9[3,3,3,3,4,5,6,7]
+; SSE-NEXT: movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,1],xmm4[0,2]
; SSE-NEXT: andps %xmm10, %xmm0
; SSE-NEXT: orps %xmm1, %xmm0
; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm7[2,3,2,3]
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm5[2,3,2,3]
; SSE-NEXT: movdqa {{.*#+}} xmm0 = [65535,0,65535,65535,65535,65535,65535,65535]
; SSE-NEXT: movdqa %xmm0, %xmm3
; SSE-NEXT: pandn %xmm1, %xmm3
-; SSE-NEXT: movdqa %xmm13, %xmm1
-; SSE-NEXT: punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm14[4],xmm1[5],xmm14[5],xmm1[6],xmm14[6],xmm1[7],xmm14[7]
+; SSE-NEXT: movdqa %xmm8, %xmm1
+; SSE-NEXT: punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm6[4],xmm1[5],xmm6[5],xmm1[6],xmm6[6],xmm1[7],xmm6[7]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[2,2,2,2,4,5,6,7]
; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,5,4]
; SSE-NEXT: pand %xmm0, %xmm1
; SSE-NEXT: por %xmm3, %xmm1
-; SSE-NEXT: punpckhwd {{.*#+}} xmm5 = xmm5[4],xmm2[4],xmm5[5],xmm2[5],xmm5[6],xmm2[6],xmm5[7],xmm2[7]
-; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm11, %xmm0
-; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,2],xmm5[2,3]
+; SSE-NEXT: movdqa %xmm14, %xmm3
+; SSE-NEXT: punpckhwd {{.*#+}} xmm3 = xmm3[4],xmm9[4],xmm3[5],xmm9[5],xmm3[6],xmm9[6],xmm3[7],xmm9[7]
+; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm12, %xmm0
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,2],xmm3[2,3]
; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[1,2],xmm1[0,3]
; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa 48(%rax), %xmm5
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm5[0,1,0,1]
-; SSE-NEXT: movdqa {{.*#+}} xmm14 = [65535,65535,65535,65535,65535,0,65535,65535]
-; SSE-NEXT: movdqa %xmm14, %xmm3
+; SSE-NEXT: movdqa %xmm13, %xmm4
+; SSE-NEXT: movdqa %xmm13, %xmm3
; SSE-NEXT: pandn %xmm1, %xmm3
-; SSE-NEXT: movdqa 48(%r8), %xmm0
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 48(%r9), %xmm12
-; SSE-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm12[0],xmm0[1],xmm12[1],xmm0[2],xmm12[2],xmm0[3],xmm12[3]
-; SSE-NEXT: movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa 48(%r8), %xmm13
+; SSE-NEXT: movdqa 48(%r9), %xmm14
+; SSE-NEXT: movdqa %xmm13, %xmm0
+; SSE-NEXT: movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm14[0],xmm0[1],xmm14[1],xmm0[2],xmm14[2],xmm0[3],xmm14[3]
+; SSE-NEXT: movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa %xmm0, %xmm1
-; SSE-NEXT: movdqa %xmm0, %xmm11
+; SSE-NEXT: movdqa %xmm0, %xmm12
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pslldq {{.*#+}} xmm1 = zero,zero,xmm1[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
-; SSE-NEXT: pand %xmm14, %xmm1
+; SSE-NEXT: pand %xmm4, %xmm1
; SSE-NEXT: por %xmm3, %xmm1
-; SSE-NEXT: movdqa %xmm8, %xmm7
-; SSE-NEXT: movdqa %xmm8, %xmm3
+; SSE-NEXT: movdqa %xmm11, %xmm7
+; SSE-NEXT: movdqa %xmm11, %xmm3
; SSE-NEXT: pandn %xmm1, %xmm3
-; SSE-NEXT: movdqa 48(%rcx), %xmm0
-; SSE-NEXT: movdqa %xmm0, %xmm1
-; SSE-NEXT: movdqa %xmm0, %xmm9
+; SSE-NEXT: movdqa 48(%rcx), %xmm10
+; SSE-NEXT: movdqa %xmm10, %xmm1
; SSE-NEXT: psrld $16, %xmm1
-; SSE-NEXT: movdqa 48(%rdx), %xmm0
-; SSE-NEXT: movdqa %xmm0, %xmm4
-; SSE-NEXT: movdqa %xmm0, %xmm10
+; SSE-NEXT: movdqa 48(%rdx), %xmm9
+; SSE-NEXT: movdqa %xmm9, %xmm4
; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
-; SSE-NEXT: movdqa {{.*#+}} xmm2 = [65535,0,0,65535,65535,65535,65535,65535]
-; SSE-NEXT: movdqa %xmm2, %xmm1
+; SSE-NEXT: movdqa {{.*#+}} xmm11 = [65535,0,0,65535,65535,65535,65535,65535]
+; SSE-NEXT: movdqa %xmm11, %xmm1
; SSE-NEXT: pandn %xmm4, %xmm1
-; SSE-NEXT: movdqa 48(%rdi), %xmm13
+; SSE-NEXT: movdqa 48(%rdi), %xmm8
; SSE-NEXT: movdqa 48(%rsi), %xmm0
; SSE-NEXT: movdqa %xmm0, %xmm4
-; SSE-NEXT: movdqa %xmm0, %xmm8
+; SSE-NEXT: movdqa %xmm0, %xmm2
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm13[0],xmm4[1],xmm13[1],xmm4[2],xmm13[2],xmm4[3],xmm13[3]
-; SSE-NEXT: movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm8[0],xmm4[1],xmm8[1],xmm4[2],xmm8[2],xmm4[3],xmm8[3]
+; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm4 = xmm4[2,2,2,2,4,5,6,7]
; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm4[0,1,2,3,4,5,5,4]
-; SSE-NEXT: pand %xmm2, %xmm0
+; SSE-NEXT: pand %xmm11, %xmm0
; SSE-NEXT: por %xmm1, %xmm0
; SSE-NEXT: pand %xmm7, %xmm0
+; SSE-NEXT: movdqa %xmm7, %xmm11
; SSE-NEXT: por %xmm3, %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pandn %xmm5, %xmm15
; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm11, %xmm3
+; SSE-NEXT: movdqa %xmm12, %xmm3
; SSE-NEXT: psrldq {{.*#+}} xmm3 = xmm3[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
; SSE-NEXT: por %xmm15, %xmm3
-; SSE-NEXT: movdqa {{.*#+}} xmm2 = [65535,0,0,0,65535,65535,65535,65535]
-; SSE-NEXT: movdqa %xmm2, %xmm1
+; SSE-NEXT: movdqa {{.*#+}} xmm7 = [65535,0,0,0,65535,65535,65535,65535]
+; SSE-NEXT: movdqa %xmm7, %xmm1
; SSE-NEXT: pandn %xmm3, %xmm1
-; SSE-NEXT: movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm10[2,2,2,2]
+; SSE-NEXT: movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm9[2,2,2,2]
; SSE-NEXT: movdqa {{.*#+}} xmm0 = [65535,65535,65535,65535,65535,65535,0,65535]
; SSE-NEXT: movdqa %xmm0, %xmm4
; SSE-NEXT: pandn %xmm3, %xmm4
-; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm9[0,1,2,3,4,4,4,4]
+; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm10[0,1,2,3,4,4,4,4]
; SSE-NEXT: pand %xmm0, %xmm3
; SSE-NEXT: por %xmm4, %xmm3
-; SSE-NEXT: movdqa %xmm13, %xmm0
-; SSE-NEXT: punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm8[4],xmm0[5],xmm8[5],xmm0[6],xmm8[6],xmm0[7],xmm8[7]
+; SSE-NEXT: movdqa %xmm8, %xmm0
+; SSE-NEXT: punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
; SSE-NEXT: movdqa %xmm0, %xmm4
; SSE-NEXT: movdqa %xmm0, %xmm6
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[0,1],xmm3[3,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm9[3,3,3,3,4,5,6,7]
-; SSE-NEXT: movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm10[3,3,3,3,4,5,6,7]
+; SSE-NEXT: movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,1],xmm4[0,2]
-; SSE-NEXT: andps %xmm2, %xmm0
+; SSE-NEXT: andps %xmm7, %xmm0
; SSE-NEXT: orps %xmm1, %xmm0
; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm5[2,3,2,3]
; SSE-NEXT: movdqa {{.*#+}} xmm0 = [65535,0,65535,65535,65535,65535,65535,65535]
; SSE-NEXT: movdqa %xmm0, %xmm3
; SSE-NEXT: pandn %xmm1, %xmm3
-; SSE-NEXT: punpckhwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
-; SSE-NEXT: # xmm12 = xmm12[4],mem[4],xmm12[5],mem[5],xmm12[6],mem[6],xmm12[7],mem[7]
-; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm12[2,2,2,2,4,5,6,7]
+; SSE-NEXT: movdqa %xmm14, %xmm1
+; SSE-NEXT: punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm13[4],xmm1[5],xmm13[5],xmm1[6],xmm13[6],xmm1[7],xmm13[7]
+; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[2,2,2,2,4,5,6,7]
; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,5,4]
; SSE-NEXT: pand %xmm0, %xmm1
; SSE-NEXT: por %xmm3, %xmm1
-; SSE-NEXT: movdqa %xmm10, %xmm3
-; SSE-NEXT: punpckhwd {{.*#+}} xmm3 = xmm3[4],xmm9[4],xmm3[5],xmm9[5],xmm3[6],xmm9[6],xmm3[7],xmm9[7]
-; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: punpckhwd {{.*#+}} xmm9 = xmm9[4],xmm10[4],xmm9[5],xmm10[5],xmm9[6],xmm10[6],xmm9[7],xmm10[7]
+; SSE-NEXT: movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa %xmm6, %xmm0
-; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,2],xmm3[2,3]
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,2],xmm9[2,3]
; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[1,2],xmm1[0,3]
; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa 64(%rax), %xmm5
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm5[0,1,0,1]
-; SSE-NEXT: movdqa %xmm14, %xmm3
+; SSE-NEXT: movdqa {{.*#+}} xmm4 = [65535,65535,65535,65535,65535,0,65535,65535]
+; SSE-NEXT: movdqa %xmm4, %xmm3
; SSE-NEXT: pandn %xmm1, %xmm3
; SSE-NEXT: movdqa 64(%r8), %xmm1
-; SSE-NEXT: movdqa 64(%r9), %xmm10
+; SSE-NEXT: movdqa 64(%r9), %xmm12
; SSE-NEXT: movdqa %xmm1, %xmm0
-; SSE-NEXT: movdqa %xmm1, %xmm7
+; SSE-NEXT: movdqa %xmm1, %xmm10
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm10[0],xmm0[1],xmm10[1],xmm0[2],xmm10[2],xmm0[3],xmm10[3]
-; SSE-NEXT: movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm12[0],xmm0[1],xmm12[1],xmm0[2],xmm12[2],xmm0[3],xmm12[3]
+; SSE-NEXT: movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa %xmm0, %xmm1
-; SSE-NEXT: movdqa %xmm0, %xmm2
+; SSE-NEXT: movdqa %xmm0, %xmm6
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pslldq {{.*#+}} xmm1 = zero,zero,xmm1[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
-; SSE-NEXT: pand %xmm14, %xmm1
+; SSE-NEXT: pand %xmm4, %xmm1
; SSE-NEXT: por %xmm3, %xmm1
-; SSE-NEXT: movdqa {{.*#+}} xmm14 = [65535,65535,65535,0,0,0,65535,65535]
-; SSE-NEXT: movdqa %xmm14, %xmm3
+; SSE-NEXT: movdqa %xmm11, %xmm3
; SSE-NEXT: pandn %xmm1, %xmm3
; SSE-NEXT: movdqa 64(%rcx), %xmm0
; SSE-NEXT: movdqa %xmm0, %xmm1
-; SSE-NEXT: movdqa %xmm0, %xmm11
+; SSE-NEXT: movdqa %xmm0, %xmm13
; SSE-NEXT: psrld $16, %xmm1
; SSE-NEXT: movdqa 64(%rdx), %xmm0
; SSE-NEXT: movdqa %xmm0, %xmm4
-; SSE-NEXT: movdqa %xmm0, %xmm13
+; SSE-NEXT: movdqa %xmm0, %xmm15
; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
-; SSE-NEXT: movdqa {{.*#+}} xmm12 = [65535,0,0,65535,65535,65535,65535,65535]
-; SSE-NEXT: movdqa %xmm12, %xmm1
+; SSE-NEXT: movdqa {{.*#+}} xmm7 = [65535,0,0,65535,65535,65535,65535,65535]
+; SSE-NEXT: movdqa %xmm7, %xmm1
; SSE-NEXT: pandn %xmm4, %xmm1
-; SSE-NEXT: movdqa 64(%rdi), %xmm8
+; SSE-NEXT: movdqa 64(%rdi), %xmm9
; SSE-NEXT: movdqa 64(%rsi), %xmm0
; SSE-NEXT: movdqa %xmm0, %xmm4
-; SSE-NEXT: movdqa %xmm0, %xmm6
+; SSE-NEXT: movdqa %xmm0, %xmm2
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm8[0],xmm4[1],xmm8[1],xmm4[2],xmm8[2],xmm4[3],xmm8[3]
-; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm9[0],xmm4[1],xmm9[1],xmm4[2],xmm9[2],xmm4[3],xmm9[3]
+; SSE-NEXT: movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm4 = xmm4[2,2,2,2,4,5,6,7]
; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm4[0,1,2,3,4,5,5,4]
-; SSE-NEXT: pand %xmm12, %xmm0
+; SSE-NEXT: pand %xmm7, %xmm0
; SSE-NEXT: por %xmm1, %xmm0
-; SSE-NEXT: pand %xmm14, %xmm0
+; SSE-NEXT: pand %xmm11, %xmm0
; SSE-NEXT: por %xmm3, %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{.*#+}} xmm15 = [65535,65535,65535,0,65535,65535,65535,65535]
-; SSE-NEXT: movdqa %xmm15, %xmm1
+; SSE-NEXT: movdqa {{.*#+}} xmm14 = [65535,65535,65535,0,65535,65535,65535,65535]
+; SSE-NEXT: movdqa %xmm14, %xmm1
; SSE-NEXT: pandn %xmm5, %xmm1
; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm2, %xmm3
+; SSE-NEXT: movdqa %xmm6, %xmm3
; SSE-NEXT: psrldq {{.*#+}} xmm3 = xmm3[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
; SSE-NEXT: por %xmm1, %xmm3
-; SSE-NEXT: movdqa {{.*#+}} xmm9 = [65535,0,0,0,65535,65535,65535,65535]
-; SSE-NEXT: movdqa %xmm9, %xmm1
+; SSE-NEXT: movdqa {{.*#+}} xmm7 = [65535,0,0,0,65535,65535,65535,65535]
+; SSE-NEXT: movdqa %xmm7, %xmm1
; SSE-NEXT: pandn %xmm3, %xmm1
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm13[2,2,2,2]
-; SSE-NEXT: movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm15[2,2,2,2]
+; SSE-NEXT: movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa {{.*#+}} xmm0 = [65535,65535,65535,65535,65535,65535,0,65535]
; SSE-NEXT: movdqa %xmm0, %xmm4
; SSE-NEXT: pandn %xmm3, %xmm4
-; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm11[0,1,2,3,4,4,4,4]
+; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm13[0,1,2,3,4,4,4,4]
; SSE-NEXT: pand %xmm0, %xmm3
; SSE-NEXT: por %xmm4, %xmm3
-; SSE-NEXT: movdqa %xmm8, %xmm0
-; SSE-NEXT: punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm6[4],xmm0[5],xmm6[5],xmm0[6],xmm6[6],xmm0[7],xmm6[7]
+; SSE-NEXT: movdqa %xmm9, %xmm0
+; SSE-NEXT: punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
; SSE-NEXT: movdqa %xmm0, %xmm4
-; SSE-NEXT: movdqa %xmm0, %xmm2
+; SSE-NEXT: movdqa %xmm0, %xmm6
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[0,1],xmm3[3,3]
-; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm11[3,3,3,3,4,5,6,7]
-; SSE-NEXT: movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm13[3,3,3,3,4,5,6,7]
+; SSE-NEXT: movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,1],xmm4[0,2]
-; SSE-NEXT: andps %xmm9, %xmm0
+; SSE-NEXT: andps %xmm7, %xmm0
; SSE-NEXT: orps %xmm1, %xmm0
; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm5[2,3,2,3]
; SSE-NEXT: movdqa {{.*#+}} xmm0 = [65535,0,65535,65535,65535,65535,65535,65535]
; SSE-NEXT: movdqa %xmm0, %xmm3
; SSE-NEXT: pandn %xmm1, %xmm3
-; SSE-NEXT: movdqa %xmm10, %xmm1
-; SSE-NEXT: punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm7[4],xmm1[5],xmm7[5],xmm1[6],xmm7[6],xmm1[7],xmm7[7]
+; SSE-NEXT: movdqa %xmm12, %xmm1
+; SSE-NEXT: punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm10[4],xmm1[5],xmm10[5],xmm1[6],xmm10[6],xmm1[7],xmm10[7]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[2,2,2,2,4,5,6,7]
; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,5,4]
; SSE-NEXT: pand %xmm0, %xmm1
; SSE-NEXT: por %xmm3, %xmm1
-; SSE-NEXT: movdqa %xmm13, %xmm3
-; SSE-NEXT: punpckhwd {{.*#+}} xmm3 = xmm3[4],xmm11[4],xmm3[5],xmm11[5],xmm3[6],xmm11[6],xmm3[7],xmm11[7]
+; SSE-NEXT: movdqa %xmm15, %xmm3
+; SSE-NEXT: punpckhwd {{.*#+}} xmm3 = xmm3[4],xmm13[4],xmm3[5],xmm13[5],xmm3[6],xmm13[6],xmm3[7],xmm13[7]
; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm2, %xmm0
+; SSE-NEXT: movdqa %xmm6, %xmm0
; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,2],xmm3[2,3]
; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[1,2],xmm1[0,3]
; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa 80(%rax), %xmm9
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm9[0,1,0,1]
+; SSE-NEXT: movdqa 80(%rax), %xmm5
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm5[0,1,0,1]
; SSE-NEXT: movdqa {{.*#+}} xmm12 = [65535,65535,65535,65535,65535,0,65535,65535]
; SSE-NEXT: movdqa %xmm12, %xmm3
; SSE-NEXT: pandn %xmm1, %xmm3
; SSE-NEXT: movdqa 80(%r8), %xmm1
-; SSE-NEXT: movdqa 80(%r9), %xmm11
-; SSE-NEXT: movdqa %xmm1, %xmm8
+; SSE-NEXT: movdqa 80(%r9), %xmm8
+; SSE-NEXT: movdqa %xmm1, %xmm0
; SSE-NEXT: movdqa %xmm1, %xmm10
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpcklwd {{.*#+}} xmm8 = xmm8[0],xmm11[0],xmm8[1],xmm11[1],xmm8[2],xmm11[2],xmm8[3],xmm11[3]
-; SSE-NEXT: movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm8, %xmm1
+; SSE-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1],xmm0[2],xmm8[2],xmm0[3],xmm8[3]
; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm0, %xmm1
+; SSE-NEXT: movdqa %xmm0, %xmm6
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pslldq {{.*#+}} xmm1 = zero,zero,xmm1[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
; SSE-NEXT: pand %xmm12, %xmm1
; SSE-NEXT: por %xmm3, %xmm1
-; SSE-NEXT: movdqa %xmm14, %xmm3
+; SSE-NEXT: movdqa %xmm11, %xmm3
; SSE-NEXT: pandn %xmm1, %xmm3
; SSE-NEXT: movdqa 80(%rcx), %xmm0
; SSE-NEXT: movdqa %xmm0, %xmm1
; SSE-NEXT: movdqa %xmm0, %xmm13
; SSE-NEXT: psrld $16, %xmm1
-; SSE-NEXT: movdqa 80(%rdx), %xmm2
-; SSE-NEXT: movdqa %xmm2, %xmm4
+; SSE-NEXT: movdqa 80(%rdx), %xmm0
+; SSE-NEXT: movdqa %xmm0, %xmm4
+; SSE-NEXT: movdqa %xmm0, %xmm15
; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
-; SSE-NEXT: movdqa {{.*#+}} xmm5 = [65535,0,0,65535,65535,65535,65535,65535]
-; SSE-NEXT: movdqa %xmm5, %xmm1
+; SSE-NEXT: movdqa {{.*#+}} xmm2 = [65535,0,0,65535,65535,65535,65535,65535]
+; SSE-NEXT: movdqa %xmm2, %xmm1
; SSE-NEXT: pandn %xmm4, %xmm1
-; SSE-NEXT: movdqa 80(%rdi), %xmm7
-; SSE-NEXT: movdqa 80(%rsi), %xmm0
-; SSE-NEXT: movdqa %xmm0, %xmm4
-; SSE-NEXT: movdqa %xmm0, %xmm6
-; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm7[0],xmm4[1],xmm7[1],xmm4[2],xmm7[2],xmm4[3],xmm7[3]
+; SSE-NEXT: movdqa 80(%rdi), %xmm9
+; SSE-NEXT: movdqa 80(%rsi), %xmm7
+; SSE-NEXT: movdqa %xmm7, %xmm4
; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm9[0],xmm4[1],xmm9[1],xmm4[2],xmm9[2],xmm4[3],xmm9[3]
+; SSE-NEXT: movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshuflw {{.*#+}} xmm4 = xmm4[2,2,2,2,4,5,6,7]
; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm4[0,1,2,3,4,5,5,4]
-; SSE-NEXT: pand %xmm5, %xmm0
+; SSE-NEXT: pand %xmm2, %xmm0
; SSE-NEXT: por %xmm1, %xmm0
-; SSE-NEXT: pand %xmm14, %xmm0
+; SSE-NEXT: pand %xmm11, %xmm0
; SSE-NEXT: por %xmm3, %xmm0
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm15, %xmm1
-; SSE-NEXT: movdqa %xmm9, %xmm5
-; SSE-NEXT: movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pandn %xmm9, %xmm1
-; SSE-NEXT: psrldq {{.*#+}} xmm8 = xmm8[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; SSE-NEXT: por %xmm1, %xmm8
-; SSE-NEXT: movdqa {{.*#+}} xmm9 = [65535,0,0,0,65535,65535,65535,65535]
-; SSE-NEXT: movdqa %xmm9, %xmm1
-; SSE-NEXT: pandn %xmm8, %xmm1
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm2[2,2,2,2]
-; SSE-NEXT: movdqa %xmm2, %xmm8
-; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{.*#+}} xmm0 = [65535,65535,65535,65535,65535,65535,0,65535]
-; SSE-NEXT: movdqa %xmm0, %xmm4
+; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pandn %xmm5, %xmm14
+; SSE-NEXT: movdqa %xmm6, %xmm3
+; SSE-NEXT: psrldq {{.*#+}} xmm3 = xmm3[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; SSE-NEXT: por %xmm14, %xmm3
+; SSE-NEXT: movdqa {{.*#+}} xmm14 = [65535,0,0,0,65535,65535,65535,65535]
+; SSE-NEXT: movdqa %xmm14, %xmm1
+; SSE-NEXT: pandn %xmm3, %xmm1
+; SSE-NEXT: movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm15[2,2,2,2]
+; SSE-NEXT: movdqa {{.*#+}} xmm6 = [65535,65535,65535,65535,65535,65535,0,65535]
+; SSE-NEXT: movdqa %xmm6, %xmm4
; SSE-NEXT: pandn %xmm3, %xmm4
; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm13[0,1,2,3,4,4,4,4]
-; SSE-NEXT: pand %xmm0, %xmm3
+; SSE-NEXT: pand %xmm6, %xmm3
; SSE-NEXT: por %xmm4, %xmm3
-; SSE-NEXT: movdqa %xmm7, %xmm0
-; SSE-NEXT: punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm6[4],xmm0[5],xmm6[5],xmm0[6],xmm6[6],xmm0[7],xmm6[7]
+; SSE-NEXT: movdqa %xmm9, %xmm0
+; SSE-NEXT: punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm7[4],xmm0[5],xmm7[5],xmm0[6],xmm7[6],xmm0[7],xmm7[7]
; SSE-NEXT: movdqa %xmm0, %xmm4
; SSE-NEXT: movdqa %xmm0, %xmm2
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[0,1],xmm3[3,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm13[3,3,3,3,4,5,6,7]
-; SSE-NEXT: movdqa %xmm13, %xmm6
; SSE-NEXT: movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,1],xmm4[0,2]
-; SSE-NEXT: andps %xmm9, %xmm0
-; SSE-NEXT: movaps %xmm9, %xmm13
+; SSE-NEXT: andps %xmm14, %xmm0
; SSE-NEXT: orps %xmm1, %xmm0
; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm5[2,3,2,3]
-; SSE-NEXT: movdqa {{.*#+}} xmm0 = [65535,0,65535,65535,65535,65535,65535,65535]
-; SSE-NEXT: movdqa %xmm0, %xmm3
+; SSE-NEXT: movdqa {{.*#+}} xmm14 = [65535,0,65535,65535,65535,65535,65535,65535]
+; SSE-NEXT: movdqa %xmm14, %xmm3
; SSE-NEXT: pandn %xmm1, %xmm3
-; SSE-NEXT: movdqa %xmm11, %xmm1
-; SSE-NEXT: punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm10[4],xmm1[5],xmm10[5],xmm1[6],xmm10[6],xmm1[7],xmm10[7]
-; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[2,2,2,2,4,5,6,7]
+; SSE-NEXT: punpckhwd {{.*#+}} xmm8 = xmm8[4],xmm10[4],xmm8[5],xmm10[5],xmm8[6],xmm10[6],xmm8[7],xmm10[7]
+; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm8[2,2,2,2,4,5,6,7]
; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5,5,4]
-; SSE-NEXT: pand %xmm0, %xmm1
-; SSE-NEXT: movdqa %xmm0, %xmm11
+; SSE-NEXT: pand %xmm14, %xmm1
; SSE-NEXT: por %xmm3, %xmm1
-; SSE-NEXT: punpckhwd {{.*#+}} xmm8 = xmm8[4],xmm6[4],xmm8[5],xmm6[5],xmm8[6],xmm6[6],xmm8[7],xmm6[7]
-; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm15, %xmm3
+; SSE-NEXT: punpckhwd {{.*#+}} xmm3 = xmm3[4],xmm13[4],xmm3[5],xmm13[5],xmm3[6],xmm13[6],xmm3[7],xmm13[7]
+; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa %xmm2, %xmm0
-; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,2],xmm8[2,3]
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,2],xmm3[2,3]
; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[1,2],xmm1[0,3]
; SSE-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm5[0,1,0,1]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm15[0,1,0,1]
; SSE-NEXT: movdqa %xmm12, %xmm3
; SSE-NEXT: pandn %xmm1, %xmm3
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm8, %xmm0
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm6[0],xmm0[1],xmm6[1],xmm0[2],xmm6[2],xmm0[3],xmm6[3]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm9, %xmm0
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
; SSE-NEXT: movdqa %xmm0, %xmm1
; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pslldq {{.*#+}} xmm1 = zero,zero,xmm1[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
; SSE-NEXT: pand %xmm12, %xmm1
; SSE-NEXT: por %xmm3, %xmm1
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm2, %xmm3
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm12, %xmm3
; SSE-NEXT: psrld $16, %xmm3
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm7, %xmm4
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm10, %xmm4
; SSE-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
-; SSE-NEXT: movdqa {{.*#+}} xmm14 = [65535,0,0,65535,65535,65535,65535,65535]
-; SSE-NEXT: movdqa %xmm14, %xmm3
+; SSE-NEXT: movdqa {{.*#+}} xmm13 = [65535,0,0,65535,65535,65535,65535,65535]
+; SSE-NEXT: movdqa %xmm13, %xmm3
; SSE-NEXT: pandn %xmm4, %xmm3
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; SSE-NEXT: punpcklwd (%rsp), %xmm4 # 16-byte Folded Reload
; SSE-NEXT: # xmm4 = xmm4[0],mem[0],xmm4[1],mem[1],xmm4[2],mem[2],xmm4[3],mem[3]
; SSE-NEXT: pshuflw {{.*#+}} xmm4 = xmm4[2,2,2,2,4,5,6,7]
; SSE-NEXT: pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5,5,4]
-; SSE-NEXT: pand %xmm14, %xmm4
+; SSE-NEXT: pand %xmm13, %xmm4
; SSE-NEXT: por %xmm3, %xmm4
-; SSE-NEXT: movdqa {{.*#+}} xmm10 = [65535,65535,65535,0,0,0,65535,65535]
-; SSE-NEXT: pand %xmm10, %xmm4
-; SSE-NEXT: pandn %xmm1, %xmm10
-; SSE-NEXT: por %xmm4, %xmm10
-; SSE-NEXT: movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm15, %xmm3
-; SSE-NEXT: movdqa %xmm15, %xmm12
-; SSE-NEXT: pandn %xmm5, %xmm3
-; SSE-NEXT: movdqa %xmm5, %xmm10
-; SSE-NEXT: movdqa %xmm0, %xmm1
-; SSE-NEXT: psrldq {{.*#+}} xmm1 = xmm1[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; SSE-NEXT: por %xmm3, %xmm1
-; SSE-NEXT: movdqa %xmm7, %xmm14
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm7[2,2,2,2]
-; SSE-NEXT: movdqa {{.*#+}} xmm9 = [65535,65535,65535,65535,65535,65535,0,65535]
-; SSE-NEXT: movdqa %xmm9, %xmm4
-; SSE-NEXT: pandn %xmm3, %xmm4
-; SSE-NEXT: pshufhw {{.*#+}} xmm5 = xmm2[0,1,2,3,4,4,4,4]
-; SSE-NEXT: pand %xmm9, %xmm5
+; SSE-NEXT: movdqa {{.*#+}} xmm11 = [65535,65535,65535,0,0,0,65535,65535]
+; SSE-NEXT: pand %xmm11, %xmm4
+; SSE-NEXT: pandn %xmm1, %xmm11
+; SSE-NEXT: por %xmm4, %xmm11
+; SSE-NEXT: movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa {{.*#+}} xmm1 = [65535,65535,65535,0,65535,65535,65535,65535]
+; SSE-NEXT: pandn %xmm15, %xmm1
+; SSE-NEXT: movdqa %xmm0, %xmm3
+; SSE-NEXT: psrldq {{.*#+}} xmm3 = xmm3[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; SSE-NEXT: por %xmm1, %xmm3
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm10[2,2,2,2]
+; SSE-NEXT: movdqa %xmm6, %xmm4
+; SSE-NEXT: pandn %xmm1, %xmm4
+; SSE-NEXT: pshufhw {{.*#+}} xmm5 = xmm12[0,1,2,3,4,4,4,4]
+; SSE-NEXT: pand %xmm6, %xmm5
; SSE-NEXT: por %xmm4, %xmm5
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, %xmm3
-; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[0,1],xmm5[3,3]
-; SSE-NEXT: movdqa %xmm7, %xmm4
-; SSE-NEXT: punpckhwd {{.*#+}} xmm4 = xmm4[4],xmm2[4],xmm4[5],xmm2[5],xmm4[6],xmm2[6],xmm4[7],xmm2[7]
+; SSE-NEXT: movaps %xmm0, %xmm1
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,1],xmm5[3,3]
+; SSE-NEXT: movdqa %xmm10, %xmm4
+; SSE-NEXT: punpckhwd {{.*#+}} xmm4 = xmm4[4],xmm12[4],xmm4[5],xmm12[5],xmm4[6],xmm12[6],xmm4[7],xmm12[7]
; SSE-NEXT: movdqa %xmm4, %xmm5
; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpcklwd {{.*#+}} xmm14 = xmm14[0],xmm2[0],xmm14[1],xmm2[1],xmm14[2],xmm2[2],xmm14[3],xmm2[3]
-; SSE-NEXT: movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshuflw {{.*#+}} xmm4 = xmm2[3,3,3,3,4,5,6,7]
-; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[0,1],xmm3[0,2]
-; SSE-NEXT: movaps %xmm13, %xmm3
-; SSE-NEXT: andps %xmm13, %xmm4
-; SSE-NEXT: andnps %xmm1, %xmm3
-; SSE-NEXT: orps %xmm4, %xmm3
-; SSE-NEXT: movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm10[2,3,2,3]
-; SSE-NEXT: movdqa %xmm11, %xmm1
+; SSE-NEXT: punpcklwd {{.*#+}} xmm10 = xmm10[0],xmm12[0],xmm10[1],xmm12[1],xmm10[2],xmm12[2],xmm10[3],xmm12[3]
+; SSE-NEXT: movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshuflw {{.*#+}} xmm4 = xmm12[3,3,3,3,4,5,6,7]
+; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[0,1],xmm1[0,2]
+; SSE-NEXT: movaps {{.*#+}} xmm1 = [65535,0,0,0,65535,65535,65535,65535]
+; SSE-NEXT: andps %xmm1, %xmm4
+; SSE-NEXT: andnps %xmm3, %xmm1
+; SSE-NEXT: orps %xmm4, %xmm1
+; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm15[2,3,2,3]
+; SSE-NEXT: movdqa %xmm14, %xmm1
; SSE-NEXT: pandn %xmm3, %xmm1
-; SSE-NEXT: movdqa %xmm8, %xmm2
-; SSE-NEXT: movdqa %xmm6, %xmm3
-; SSE-NEXT: punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm6[4],xmm2[5],xmm6[5],xmm2[6],xmm6[6],xmm2[7],xmm6[7]
-; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpckhwd {{.*#+}} xmm3 = xmm3[4],xmm8[4],xmm3[5],xmm8[5],xmm3[6],xmm8[6],xmm3[7],xmm8[7]
+; SSE-NEXT: movdqa %xmm9, %xmm4
+; SSE-NEXT: movdqa %xmm2, %xmm3
+; SSE-NEXT: punpckhwd {{.*#+}} xmm4 = xmm4[4],xmm2[4],xmm4[5],xmm2[5],xmm4[6],xmm2[6],xmm4[7],xmm2[7]
+; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: punpckhwd {{.*#+}} xmm3 = xmm3[4],xmm9[4],xmm3[5],xmm9[5],xmm3[6],xmm9[6],xmm3[7],xmm9[7]
; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[2,2,2,2,4,5,6,7]
; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,5,4]
-; SSE-NEXT: pand %xmm11, %xmm3
-; SSE-NEXT: movdqa %xmm11, %xmm15
+; SSE-NEXT: pand %xmm14, %xmm3
; SSE-NEXT: por %xmm1, %xmm3
; SSE-NEXT: movaps %xmm0, %xmm1
; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[2,2],xmm5[2,3]
; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[1,2],xmm3[0,3]
; SSE-NEXT: movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm9, %xmm1
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm15, %xmm1
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; SSE-NEXT: punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm4[4],xmm1[5],xmm4[5],xmm1[6],xmm4[6],xmm1[7],xmm4[7]
-; SSE-NEXT: movdqa %xmm9, %xmm3
-; SSE-NEXT: punpcklwd {{.*#+}} xmm9 = xmm9[0],xmm4[0],xmm9[1],xmm4[1],xmm9[2],xmm4[2],xmm9[3],xmm4[3]
+; SSE-NEXT: movdqa %xmm15, %xmm3
+; SSE-NEXT: punpcklwd {{.*#+}} xmm15 = xmm15[0],xmm4[0],xmm15[1],xmm4[1],xmm15[2],xmm4[2],xmm15[3],xmm4[3]
; SSE-NEXT: psrld $16, %xmm4
; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm13, %xmm4
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm6, %xmm4
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
; SSE-NEXT: punpcklwd {{.*#+}} xmm4 = xmm4[0],xmm10[0],xmm4[1],xmm10[1],xmm4[2],xmm10[2],xmm4[3],xmm10[3]
; SSE-NEXT: pshuflw {{.*#+}} xmm4 = xmm4[2,2,2,2,4,5,6,7]
; SSE-NEXT: pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5,5,4]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm6[0,0,0,0]
-; SSE-NEXT: pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,5,5,5,5]
-; SSE-NEXT: shufps {{.*#+}} xmm5 = xmm5[2,1],xmm4[3,3]
-; SSE-NEXT: movdqa {{.*#+}} xmm0 = [65535,0,0,65535,65535,65535,65535,65535]
-; SSE-NEXT: pand %xmm0, %xmm4
-; SSE-NEXT: pandn %xmm3, %xmm0
-; SSE-NEXT: por %xmm4, %xmm0
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm11[0,0,1,1]
+; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm11[0,0,0,0]
+; SSE-NEXT: pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,5,5,5,5]
+; SSE-NEXT: shufps {{.*#+}} xmm5 = xmm5[2,1],xmm4[3,3]
+; SSE-NEXT: pand %xmm13, %xmm4
+; SSE-NEXT: pandn %xmm3, %xmm13
+; SSE-NEXT: por %xmm4, %xmm13
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm7[0,0,1,1]
+; SSE-NEXT: movdqa {{.*#+}} xmm12 = [65535,65535,65535,0,65535,65535,65535,65535]
; SSE-NEXT: movdqa %xmm12, %xmm4
; SSE-NEXT: pandn %xmm3, %xmm4
-; SSE-NEXT: pand %xmm12, %xmm0
-; SSE-NEXT: por %xmm0, %xmm4
+; SSE-NEXT: pand %xmm12, %xmm13
+; SSE-NEXT: por %xmm13, %xmm4
; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[0,1],xmm5[0,2]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm2[0,1,0,1]
@@ -8327,10 +8330,10 @@ define void @store_i16_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; SSE-NEXT: andps %xmm5, %xmm4
; SSE-NEXT: por %xmm4, %xmm3
; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm11, %xmm4
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm11[2,2,2,2]
-; SSE-NEXT: movdqa %xmm11, %xmm0
-; SSE-NEXT: movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm7, %xmm4
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm7[2,2,2,2]
+; SSE-NEXT: movdqa %xmm7, %xmm0
+; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: shufps {{.*#+}} xmm4 = xmm4[3,0],xmm1[2,0]
; SSE-NEXT: movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: pslldq {{.*#+}} xmm1 = zero,zero,zero,zero,zero,zero,xmm1[0,1,2,3,4,5,6,7,8,9]
@@ -8346,58 +8349,56 @@ define void @store_i16_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; SSE-NEXT: pandn %xmm3, %xmm5
; SSE-NEXT: pand %xmm7, %xmm1
; SSE-NEXT: por %xmm1, %xmm5
-; SSE-NEXT: movdqa %xmm6, %xmm1
-; SSE-NEXT: movdqa %xmm6, %xmm7
+; SSE-NEXT: movdqa %xmm11, %xmm1
; SSE-NEXT: psrldq {{.*#+}} xmm1 = xmm1[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
-; SSE-NEXT: movdqa %xmm15, %xmm3
+; SSE-NEXT: movdqa %xmm14, %xmm3
; SSE-NEXT: pandn %xmm1, %xmm3
-; SSE-NEXT: pand %xmm15, %xmm5
+; SSE-NEXT: pand %xmm14, %xmm5
; SSE-NEXT: por %xmm5, %xmm3
-; SSE-NEXT: movdqa {{.*#+}} xmm5 = [65535,65535,0,65535,65535,65535,65535,65535]
-; SSE-NEXT: pand %xmm5, %xmm3
-; SSE-NEXT: movdqa %xmm2, %xmm6
+; SSE-NEXT: movdqa {{.*#+}} xmm8 = [65535,65535,0,65535,65535,65535,65535,65535]
+; SSE-NEXT: pand %xmm8, %xmm3
+; SSE-NEXT: movdqa %xmm2, %xmm5
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm2[2,2,3,3]
-; SSE-NEXT: pandn %xmm1, %xmm5
-; SSE-NEXT: por %xmm3, %xmm5
-; SSE-NEXT: movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: pandn %xmm1, %xmm8
+; SSE-NEXT: por %xmm3, %xmm8
+; SSE-NEXT: movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa %xmm10, %xmm3
; SSE-NEXT: movdqa %xmm10, %xmm1
; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm10[1,1,1,1,4,5,6,7]
; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm13[0],xmm3[1],xmm13[1],xmm3[2],xmm13[2],xmm3[3],xmm13[3]
+; SSE-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm6[0],xmm3[1],xmm6[1],xmm3[2],xmm6[2],xmm3[3],xmm6[3]
; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm13, %xmm3
+; SSE-NEXT: movdqa %xmm6, %xmm3
; SSE-NEXT: psrld $16, %xmm3
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
; SSE-NEXT: movdqa %xmm4, %xmm3
; SSE-NEXT: pandn %xmm1, %xmm3
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm9[0,1,2,3,4,5,6,6]
+; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm15[0,1,2,3,4,5,6,6]
; SSE-NEXT: pand %xmm4, %xmm1
; SSE-NEXT: por %xmm3, %xmm1
-; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[1,1],xmm9[2,3]
+; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[1,1],xmm15[2,3]
; SSE-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[2,3]
+; SSE-NEXT: movdqa %xmm12, %xmm2
; SSE-NEXT: andps %xmm12, %xmm0
-; SSE-NEXT: movdqa %xmm7, %xmm1
+; SSE-NEXT: movdqa %xmm11, %xmm1
; SSE-NEXT: pslld $16, %xmm1
-; SSE-NEXT: pandn %xmm1, %xmm12
-; SSE-NEXT: por %xmm0, %xmm12
+; SSE-NEXT: pandn %xmm1, %xmm2
+; SSE-NEXT: por %xmm0, %xmm2
; SSE-NEXT: movdqa {{.*#+}} xmm1 = [65535,65535,65535,65535,0,65535,65535,65535]
-; SSE-NEXT: pand %xmm1, %xmm12
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm6[0,1,1,3]
+; SSE-NEXT: pand %xmm1, %xmm2
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm5[0,1,1,3]
; SSE-NEXT: pandn %xmm0, %xmm1
-; SSE-NEXT: por %xmm12, %xmm1
+; SSE-NEXT: por %xmm2, %xmm1
; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm9, %xmm0
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm1, %xmm0
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
-; SSE-NEXT: movdqa %xmm1, %xmm13
-; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa %xmm2, %xmm1
+; SSE-NEXT: punpcklwd {{.*#+}} xmm9 = xmm9[0],xmm1[0],xmm9[1],xmm1[1],xmm9[2],xmm1[2],xmm9[3],xmm1[3]
+; SSE-NEXT: movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: psrlq $48, %xmm1
; SSE-NEXT: punpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm1[1]
-; SSE-NEXT: movdqa {{.*#+}} xmm8 = [65535,65535,65535,0,0,65535,65535,65535]
-; SSE-NEXT: movdqa %xmm8, %xmm1
+; SSE-NEXT: movdqa {{.*#+}} xmm7 = [65535,65535,65535,0,0,65535,65535,65535]
+; SSE-NEXT: movdqa %xmm7, %xmm1
; SSE-NEXT: pandn %xmm0, %xmm1
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: psrldq {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
@@ -8410,8 +8411,8 @@ define void @store_i16_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; SSE-NEXT: psrld $16, %xmm1
; SSE-NEXT: movdqa %xmm3, %xmm2
; SSE-NEXT: punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm1[2],xmm2[3],xmm1[3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm11, %xmm1
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm6, %xmm1
; SSE-NEXT: punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,4,7]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[3,1,2,1]
@@ -8422,15 +8423,15 @@ define void @store_i16_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; SSE-NEXT: pand %xmm5, %xmm0
; SSE-NEXT: por %xmm0, %xmm2
; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm6, %xmm0
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm10, %xmm0
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: punpcklwd {{.*#+}} xmm6 = xmm6[0],xmm1[0],xmm6[1],xmm1[1],xmm6[2],xmm1[2],xmm6[3],xmm1[3]
-; SSE-NEXT: movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: punpcklwd {{.*#+}} xmm10 = xmm10[0],xmm1[0],xmm10[1],xmm1[1],xmm10[2],xmm1[2],xmm10[3],xmm1[3]
+; SSE-NEXT: movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: psrlq $48, %xmm1
; SSE-NEXT: punpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm1[1]
-; SSE-NEXT: movdqa %xmm8, %xmm15
-; SSE-NEXT: movdqa %xmm8, %xmm1
+; SSE-NEXT: movdqa %xmm7, %xmm14
+; SSE-NEXT: movdqa %xmm7, %xmm1
; SSE-NEXT: pandn %xmm0, %xmm1
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: psrldq {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
@@ -8443,8 +8444,8 @@ define void @store_i16_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; SSE-NEXT: psrld $16, %xmm1
; SSE-NEXT: movdqa %xmm3, %xmm2
; SSE-NEXT: punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm1[2],xmm2[3],xmm1[3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm8, %xmm1
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm11, %xmm1
; SSE-NEXT: punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,4,7]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[3,1,2,1]
@@ -8461,7 +8462,7 @@ define void @store_i16_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: psrlq $48, %xmm1
; SSE-NEXT: punpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm1[1]
-; SSE-NEXT: movdqa %xmm15, %xmm1
+; SSE-NEXT: movdqa %xmm14, %xmm1
; SSE-NEXT: pandn %xmm0, %xmm1
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: psrldq {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
@@ -8474,8 +8475,8 @@ define void @store_i16_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; SSE-NEXT: psrld $16, %xmm1
; SSE-NEXT: movdqa %xmm3, %xmm2
; SSE-NEXT: punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm1[2],xmm2[3],xmm1[3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm14, %xmm1
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm13, %xmm1
; SSE-NEXT: punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,4,7]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[3,1,2,1]
@@ -8493,7 +8494,7 @@ define void @store_i16_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; SSE-NEXT: movdqa %xmm2, %xmm1
; SSE-NEXT: psrlq $48, %xmm1
; SSE-NEXT: punpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm1[1]
-; SSE-NEXT: movdqa %xmm15, %xmm1
+; SSE-NEXT: movdqa %xmm14, %xmm1
; SSE-NEXT: pandn %xmm0, %xmm1
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: psrldq {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
@@ -8506,8 +8507,8 @@ define void @store_i16_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; SSE-NEXT: psrld $16, %xmm1
; SSE-NEXT: movdqa %xmm3, %xmm2
; SSE-NEXT: punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm1[2],xmm2[3],xmm1[3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm3, %xmm1
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm8, %xmm1
; SSE-NEXT: punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,4,7]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[3,1,2,1]
@@ -8525,21 +8526,21 @@ define void @store_i16_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; SSE-NEXT: movdqa %xmm2, %xmm1
; SSE-NEXT: psrlq $48, %xmm1
; SSE-NEXT: punpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm1[1]
-; SSE-NEXT: movdqa %xmm15, %xmm1
+; SSE-NEXT: movdqa %xmm14, %xmm1
; SSE-NEXT: pandn %xmm0, %xmm1
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: psrldq {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
; SSE-NEXT: por %xmm1, %xmm0
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm10, %xmm2
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm3, %xmm2
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7]
; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: psrld $16, %xmm1
-; SSE-NEXT: movdqa %xmm10, %xmm2
+; SSE-NEXT: movdqa %xmm3, %xmm2
; SSE-NEXT: punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm1[2],xmm2[3],xmm1[3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm10, %xmm1
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm12, %xmm1
; SSE-NEXT: punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,4,7]
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[3,1,2,1]
@@ -8557,21 +8558,21 @@ define void @store_i16_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; SSE-NEXT: movdqa %xmm2, %xmm1
; SSE-NEXT: psrlq $48, %xmm1
; SSE-NEXT: punpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm1[1]
-; SSE-NEXT: movdqa %xmm15, %xmm1
+; SSE-NEXT: movdqa %xmm14, %xmm1
; SSE-NEXT: pandn %xmm0, %xmm1
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: psrldq {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
; SSE-NEXT: por %xmm1, %xmm0
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm12, %xmm2
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm3, %xmm2
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7]
; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: psrld $16, %xmm1
-; SSE-NEXT: movdqa %xmm12, %xmm2
+; SSE-NEXT: movdqa %xmm3, %xmm2
; SSE-NEXT: punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm1[2],xmm2[3],xmm1[3]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm15, %xmm1
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT: movdqa %xmm14, %xmm1
; SSE-NEXT: punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
; SSE-NEXT: pand %xmm5, %xmm0
; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,4,7]
@@ -8592,16 +8593,17 @@ define void @store_i16_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT: movdqa %xmm4, %xmm1
; SSE-NEXT: pandn %xmm0, %xmm1
-; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm13[0,1,2,3,4,5,6,6]
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm9[0,1,2,3,4,5,6,6]
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,1,2,3]
; SSE-NEXT: pand %xmm4, %xmm0
; SSE-NEXT: por %xmm1, %xmm0
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[2,2],xmm11[1,1]
-; SSE-NEXT: movaps {{.*#+}} xmm5 = [65535,65535,0,0,0,65535,65535,65535]
-; SSE-NEXT: movaps %xmm5, %xmm2
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[2,2],xmm6[1,1]
+; SSE-NEXT: movaps %xmm6, %xmm5
+; SSE-NEXT: movaps {{.*#+}} xmm3 = [65535,65535,0,0,0,65535,65535,65535]
+; SSE-NEXT: movaps %xmm3, %xmm2
; SSE-NEXT: andnps %xmm1, %xmm2
-; SSE-NEXT: pand %xmm5, %xmm0
+; SSE-NEXT: pand %xmm3, %xmm0
; SSE-NEXT: orps %xmm0, %xmm2
; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
@@ -8616,16 +8618,15 @@ define void @store_i16_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; SSE-NEXT: movdqa %xmm4, %xmm1
; SSE-NEXT: pandn %xmm0, %xmm1
-; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm6[0,1,2,3,4,5,6,6]
+; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm10[0,1,2,3,4,5,6,6]
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,1,2,3]
; SSE-NEXT: pand %xmm4, %xmm0
; SSE-NEXT: por %xmm1, %xmm0
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[2,2],xmm8[1,1]
-; SSE-NEXT: movaps %xmm8, %xmm12
-; SSE-NEXT: movaps %xmm5, %xmm2
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[2,2],xmm11[1,1]
+; SSE-NEXT: movaps %xmm3, %xmm2
; SSE-NEXT: andnps %xmm1, %xmm2
-; SSE-NEXT: pand %xmm5, %xmm0
+; SSE-NEXT: pand %xmm3, %xmm0
; SSE-NEXT: orps %xmm0, %xmm2
; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
@@ -8645,11 +8646,10 @@ define void @store_i16_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; SSE-NEXT: pand %xmm4, %xmm0
; SSE-NEXT: por %xmm1, %xmm0
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[2,2],xmm14[1,1]
-; SSE-NEXT: movaps %xmm14, %xmm11
-; SSE-NEXT: movaps %xmm5, %xmm2
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[2,2],xmm13[1,1]
+; SSE-NEXT: movaps %xmm3, %xmm2
; SSE-NEXT: andnps %xmm1, %xmm2
-; SSE-NEXT: pand %xmm5, %xmm0
+; SSE-NEXT: pand %xmm3, %xmm0
; SSE-NEXT: orps %xmm0, %xmm2
; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
@@ -8670,11 +8670,11 @@ define void @store_i16_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; SSE-NEXT: pand %xmm4, %xmm0
; SSE-NEXT: por %xmm1, %xmm0
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[2,2],xmm3[1,1]
-; SSE-NEXT: movaps %xmm3, %xmm14
-; SSE-NEXT: movaps %xmm5, %xmm2
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[2,2],xmm8[1,1]
+; SSE-NEXT: movaps %xmm8, %xmm7
+; SSE-NEXT: movaps %xmm3, %xmm2
; SSE-NEXT: andnps %xmm1, %xmm2
-; SSE-NEXT: pand %xmm5, %xmm0
+; SSE-NEXT: pand %xmm3, %xmm0
; SSE-NEXT: orps %xmm0, %xmm2
; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
@@ -8695,11 +8695,11 @@ define void @store_i16_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; SSE-NEXT: pand %xmm4, %xmm0
; SSE-NEXT: por %xmm1, %xmm0
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: movdqa %xmm10, %xmm13
-; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[2,2],xmm10[1,1]
-; SSE-NEXT: movaps %xmm5, %xmm2
+; SSE-NEXT: movdqa %xmm12, %xmm6
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[2,2],xmm12[1,1]
+; SSE-NEXT: movaps %xmm3, %xmm2
; SSE-NEXT: andnps %xmm1, %xmm2
-; SSE-NEXT: pand %xmm5, %xmm0
+; SSE-NEXT: pand %xmm3, %xmm0
; SSE-NEXT: orps %xmm0, %xmm2
; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
@@ -8720,10 +8720,10 @@ define void @store_i16_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; SSE-NEXT: pand %xmm4, %xmm0
; SSE-NEXT: por %xmm1, %xmm0
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[2,2],xmm15[1,1]
-; SSE-NEXT: movaps %xmm5, %xmm2
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[2,2],xmm14[1,1]
+; SSE-NEXT: movaps %xmm3, %xmm2
; SSE-NEXT: andnps %xmm1, %xmm2
-; SSE-NEXT: pand %xmm5, %xmm0
+; SSE-NEXT: pand %xmm3, %xmm0
; SSE-NEXT: orps %xmm0, %xmm2
; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa (%rsp), %xmm1 # 16-byte Reload
@@ -8743,48 +8743,48 @@ define void @store_i16_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,1,2,3]
; SSE-NEXT: pand %xmm4, %xmm0
; SSE-NEXT: por %xmm1, %xmm0
-; SSE-NEXT: pand %xmm5, %xmm0
+; SSE-NEXT: movaps %xmm3, %xmm2
+; SSE-NEXT: pand %xmm3, %xmm0
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[2,2],xmm3[1,1]
-; SSE-NEXT: andnps %xmm1, %xmm5
-; SSE-NEXT: orps %xmm0, %xmm5
-; SSE-NEXT: movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
-; SSE-NEXT: # xmm9 = xmm9[0],mem[0]
+; SSE-NEXT: andnps %xmm1, %xmm2
+; SSE-NEXT: orps %xmm0, %xmm2
+; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT: # xmm15 = xmm15[0],mem[0]
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: shufps $212, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = xmm0[0,1],mem[1,3]
-; SSE-NEXT: shufps {{.*#+}} xmm9 = xmm9[2,0],xmm0[0,2]
+; SSE-NEXT: shufps {{.*#+}} xmm15 = xmm15[2,0],xmm0[0,2]
; SSE-NEXT: movaps {{.*#+}} xmm2 = [65535,65535,65535,65535,65535,0,65535,65535]
-; SSE-NEXT: andps %xmm2, %xmm9
+; SSE-NEXT: andps %xmm2, %xmm15
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm1[0,1,2,3,5,6,6,7]
; SSE-NEXT: pslldq {{.*#+}} xmm1 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm1[0,1,2,3,4,5]
; SSE-NEXT: andnps %xmm1, %xmm2
-; SSE-NEXT: orps %xmm9, %xmm2
-; SSE-NEXT: movaps {{.*#+}} xmm5 = [65535,65535,65535,65535,65535,65535,0,65535]
-; SSE-NEXT: andps %xmm5, %xmm2
-; SSE-NEXT: movaps %xmm2, %xmm7
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm2[0,0,0,0]
-; SSE-NEXT: andnps %xmm1, %xmm5
-; SSE-NEXT: orps %xmm7, %xmm5
-; SSE-NEXT: movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT: shufps $42, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
-; SSE-NEXT: # xmm5 = xmm5[2,2],mem[2,0]
+; SSE-NEXT: orps %xmm15, %xmm2
+; SSE-NEXT: movaps {{.*#+}} xmm8 = [65535,65535,65535,65535,65535,65535,0,65535]
+; SSE-NEXT: andps %xmm8, %xmm2
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm10[0,0,0,0]
+; SSE-NEXT: andnps %xmm1, %xmm8
+; SSE-NEXT: orps %xmm2, %xmm8
+; SSE-NEXT: movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT: shufps $42, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT: # xmm2 = xmm2[2,2],mem[2,0]
; SSE-NEXT: movaps {{.*#+}} xmm1 = [0,65535,65535,65535,65535,65535,65535,0]
-; SSE-NEXT: andps %xmm1, %xmm5
+; SSE-NEXT: andps %xmm1, %xmm2
; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
; SSE-NEXT: andnps %xmm0, %xmm1
-; SSE-NEXT: orps %xmm5, %xmm1
-; SSE-NEXT: movaps {{.*#+}} xmm5 = [65535,0,65535,65535,65535,65535,65535,65535]
-; SSE-NEXT: andps %xmm5, %xmm1
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm2[2,3,2,3]
-; SSE-NEXT: andnps %xmm0, %xmm5
-; SSE-NEXT: orps %xmm1, %xmm5
-; SSE-NEXT: movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: orps %xmm2, %xmm1
+; SSE-NEXT: movaps {{.*#+}} xmm2 = [65535,0,65535,65535,65535,65535,65535,65535]
+; SSE-NEXT: andps %xmm2, %xmm1
+; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm10[2,3,2,3]
+; SSE-NEXT: andnps %xmm0, %xmm2
+; SSE-NEXT: orps %xmm1, %xmm2
+; SSE-NEXT: movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; SSE-NEXT: pslldq {{.*#+}} xmm1 = zero,zero,zero,zero,zero,zero,xmm1[0,1,2,3,4,5,6,7,8,9]
; SSE-NEXT: movdqa %xmm4, %xmm0
@@ -8797,136 +8797,138 @@ define void @store_i16_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; SSE-NEXT: movdqa {{.*#+}} xmm2 = [65535,65535,65535,0,0,0,0,65535]
; SSE-NEXT: movdqa %xmm2, %xmm1
; SSE-NEXT: pandn %xmm0, %xmm1
-; SSE-NEXT: movapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT: movapd %xmm10, %xmm5
-; SSE-NEXT: shufpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
-; SSE-NEXT: # xmm5 = xmm5[1],mem[0]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: punpckhwd {{.*#+}} xmm0 = xmm0[4,4,5,5,6,6,7,7]
-; SSE-NEXT: shufps {{.*#+}} xmm5 = xmm5[2,0],xmm0[2,1]
-; SSE-NEXT: andps %xmm2, %xmm5
-; SSE-NEXT: orps %xmm1, %xmm5
-; SSE-NEXT: movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: pslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3,4,5,6,7,8,9]
+; SSE-NEXT: movaps %xmm5, %xmm0
+; SSE-NEXT: movaps %xmm5, %xmm8
+; SSE-NEXT: shufpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; SSE-NEXT: # xmm8 = xmm8[1],mem[0]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: punpckhwd {{.*#+}} xmm5 = xmm5[4,4,5,5,6,6,7,7]
+; SSE-NEXT: shufps {{.*#+}} xmm8 = xmm8[2,0],xmm5[2,1]
+; SSE-NEXT: andps %xmm2, %xmm8
+; SSE-NEXT: orps %xmm1, %xmm8
+; SSE-NEXT: movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: pslldq {{.*#+}} xmm5 = zero,zero,zero,zero,zero,zero,xmm5[0,1,2,3,4,5,6,7,8,9]
; SSE-NEXT: movdqa %xmm4, %xmm1
-; SSE-NEXT: pandn %xmm0, %xmm1
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: pslldq {{.*#+}} xmm0 = zero,zero,xmm0[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
-; SSE-NEXT: pand %xmm4, %xmm0
-; SSE-NEXT: por %xmm1, %xmm0
+; SSE-NEXT: pandn %xmm5, %xmm1
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: pslldq {{.*#+}} xmm5 = zero,zero,xmm5[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
+; SSE-NEXT: pand %xmm4, %xmm5
+; SSE-NEXT: por %xmm1, %xmm5
; SSE-NEXT: movdqa %xmm2, %xmm1
-; SSE-NEXT: pandn %xmm0, %xmm1
+; SSE-NEXT: pandn %xmm5, %xmm1
+; SSE-NEXT: movdqa %xmm11, %xmm8
+; SSE-NEXT: movaps %xmm11, %xmm12
; SSE-NEXT: shufpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
; SSE-NEXT: # xmm12 = xmm12[1],mem[0]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: punpckhwd {{.*#+}} xmm0 = xmm0[4,4,5,5,6,6,7,7]
-; SSE-NEXT: shufps {{.*#+}} xmm12 = xmm12[2,0],xmm0[2,1]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: punpckhwd {{.*#+}} xmm5 = xmm5[4,4,5,5,6,6,7,7]
+; SSE-NEXT: shufps {{.*#+}} xmm12 = xmm12[2,0],xmm5[2,1]
; SSE-NEXT: andps %xmm2, %xmm12
; SSE-NEXT: orps %xmm1, %xmm12
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: pslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3,4,5,6,7,8,9]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: pslldq {{.*#+}} xmm5 = zero,zero,zero,zero,zero,zero,xmm5[0,1,2,3,4,5,6,7,8,9]
; SSE-NEXT: movdqa %xmm4, %xmm1
-; SSE-NEXT: pandn %xmm0, %xmm1
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: pslldq {{.*#+}} xmm0 = zero,zero,xmm0[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
-; SSE-NEXT: pand %xmm4, %xmm0
-; SSE-NEXT: por %xmm1, %xmm0
+; SSE-NEXT: pandn %xmm5, %xmm1
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: pslldq {{.*#+}} xmm5 = zero,zero,xmm5[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
+; SSE-NEXT: pand %xmm4, %xmm5
+; SSE-NEXT: por %xmm1, %xmm5
; SSE-NEXT: movdqa %xmm2, %xmm1
-; SSE-NEXT: pandn %xmm0, %xmm1
-; SSE-NEXT: movaps %xmm11, %xmm6
+; SSE-NEXT: pandn %xmm5, %xmm1
+; SSE-NEXT: movaps %xmm13, %xmm11
; SSE-NEXT: shufpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Folded Reload
; SSE-NEXT: # xmm11 = xmm11[1],mem[0]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: punpckhwd {{.*#+}} xmm0 = xmm0[4,4,5,5,6,6,7,7]
-; SSE-NEXT: shufps {{.*#+}} xmm11 = xmm11[2,0],xmm0[2,1]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: punpckhwd {{.*#+}} xmm5 = xmm5[4,4,5,5,6,6,7,7]
+; SSE-NEXT: shufps {{.*#+}} xmm11 = xmm11[2,0],xmm5[2,1]
; SSE-NEXT: andps %xmm2, %xmm11
; SSE-NEXT: orps %xmm1, %xmm11
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: pslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3,4,5,6,7,8,9]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: pslldq {{.*#+}} xmm5 = zero,zero,zero,zero,zero,zero,xmm5[0,1,2,3,4,5,6,7,8,9]
; SSE-NEXT: movdqa %xmm4, %xmm1
-; SSE-NEXT: pandn %xmm0, %xmm1
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: pslldq {{.*#+}} xmm0 = zero,zero,xmm0[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
-; SSE-NEXT: pand %xmm4, %xmm0
-; SSE-NEXT: por %xmm1, %xmm0
+; SSE-NEXT: pandn %xmm5, %xmm1
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: pslldq {{.*#+}} xmm5 = zero,zero,xmm5[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
+; SSE-NEXT: pand %xmm4, %xmm5
+; SSE-NEXT: por %xmm1, %xmm5
; SSE-NEXT: movdqa %xmm2, %xmm1
-; SSE-NEXT: pandn %xmm0, %xmm1
-; SSE-NEXT: movaps %xmm14, %xmm15
-; SSE-NEXT: movaps %xmm14, %xmm9
+; SSE-NEXT: pandn %xmm5, %xmm1
+; SSE-NEXT: movaps %xmm7, %xmm15
+; SSE-NEXT: movaps %xmm7, %xmm9
; SSE-NEXT: shufpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
; SSE-NEXT: # xmm9 = xmm9[1],mem[0]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: punpckhwd {{.*#+}} xmm0 = xmm0[4,4,5,5,6,6,7,7]
-; SSE-NEXT: shufps {{.*#+}} xmm9 = xmm9[2,0],xmm0[2,1]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: punpckhwd {{.*#+}} xmm5 = xmm5[4,4,5,5,6,6,7,7]
+; SSE-NEXT: shufps {{.*#+}} xmm9 = xmm9[2,0],xmm5[2,1]
; SSE-NEXT: andps %xmm2, %xmm9
; SSE-NEXT: orps %xmm1, %xmm9
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: pslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3,4,5,6,7,8,9]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: pslldq {{.*#+}} xmm5 = zero,zero,zero,zero,zero,zero,xmm5[0,1,2,3,4,5,6,7,8,9]
; SSE-NEXT: movdqa %xmm4, %xmm1
-; SSE-NEXT: pandn %xmm0, %xmm1
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: pslldq {{.*#+}} xmm0 = zero,zero,xmm0[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
-; SSE-NEXT: pand %xmm4, %xmm0
-; SSE-NEXT: por %xmm1, %xmm0
+; SSE-NEXT: pandn %xmm5, %xmm1
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: pslldq {{.*#+}} xmm5 = zero,zero,xmm5[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
+; SSE-NEXT: pand %xmm4, %xmm5
+; SSE-NEXT: por %xmm1, %xmm5
; SSE-NEXT: movdqa %xmm2, %xmm1
-; SSE-NEXT: pandn %xmm0, %xmm1
-; SSE-NEXT: movdqa %xmm13, %xmm7
+; SSE-NEXT: pandn %xmm5, %xmm1
+; SSE-NEXT: movdqa %xmm6, %xmm7
; SSE-NEXT: shufpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
; SSE-NEXT: # xmm7 = xmm7[1],mem[0]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: punpckhwd {{.*#+}} xmm0 = xmm0[4,4,5,5,6,6,7,7]
-; SSE-NEXT: shufps {{.*#+}} xmm7 = xmm7[2,0],xmm0[2,1]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: punpckhwd {{.*#+}} xmm5 = xmm5[4,4,5,5,6,6,7,7]
+; SSE-NEXT: shufps {{.*#+}} xmm7 = xmm7[2,0],xmm5[2,1]
; SSE-NEXT: andps %xmm2, %xmm7
; SSE-NEXT: orps %xmm1, %xmm7
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: pslldq {{.*#+}} xmm0 = zero,zero,zero,zero,zero,zero,xmm0[0,1,2,3,4,5,6,7,8,9]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: pslldq {{.*#+}} xmm5 = zero,zero,zero,zero,zero,zero,xmm5[0,1,2,3,4,5,6,7,8,9]
; SSE-NEXT: movdqa %xmm4, %xmm1
-; SSE-NEXT: pandn %xmm0, %xmm1
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: pslldq {{.*#+}} xmm0 = zero,zero,xmm0[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
-; SSE-NEXT: pand %xmm4, %xmm0
-; SSE-NEXT: por %xmm1, %xmm0
+; SSE-NEXT: pandn %xmm5, %xmm1
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT: pslldq {{.*#+}} xmm5 = zero,zero,xmm5[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
+; SSE-NEXT: pand %xmm4, %xmm5
+; SSE-NEXT: por %xmm1, %xmm5
; SSE-NEXT: movdqa %xmm2, %xmm1
-; SSE-NEXT: pandn %xmm0, %xmm1
-; SSE-NEXT: movapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT: movapd %xmm14, %xmm5
+; SSE-NEXT: pandn %xmm5, %xmm1
+; SSE-NEXT: movaps %xmm14, %xmm5
; SSE-NEXT: shufpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
; SSE-NEXT: # xmm5 = xmm5[1],mem[0]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: punpckhwd {{.*#+}} xmm0 = xmm0[4,4,5,5,6,6,7,7]
-; SSE-NEXT: shufps {{.*#+}} xmm5 = xmm5[2,0],xmm0[2,1]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT: punpckhwd {{.*#+}} xmm6 = xmm6[4,4,5,5,6,6,7,7]
+; SSE-NEXT: shufps {{.*#+}} xmm5 = xmm5[2,0],xmm6[2,1]
; SSE-NEXT: andps %xmm2, %xmm5
; SSE-NEXT: orps %xmm1, %xmm5
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT: pslldq {{.*#+}} xmm6 = zero,zero,zero,zero,zero,zero,xmm6[0,1,2,3,4,5,6,7,8,9]
; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT: pslldq {{.*#+}} xmm1 = zero,zero,zero,zero,zero,zero,xmm1[0,1,2,3,4,5,6,7,8,9]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: pslldq {{.*#+}} xmm0 = zero,zero,xmm0[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
-; SSE-NEXT: pand %xmm4, %xmm0
-; SSE-NEXT: pandn %xmm1, %xmm4
-; SSE-NEXT: por %xmm0, %xmm4
-; SSE-NEXT: shufpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT: # xmm3 = xmm3[1],mem[0]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: punpckhwd {{.*#+}} xmm0 = xmm0[4,4,5,5,6,6,7,7]
-; SSE-NEXT: shufps {{.*#+}} xmm3 = xmm3[2,0],xmm0[2,1]
-; SSE-NEXT: andps %xmm2, %xmm3
+; SSE-NEXT: pslldq {{.*#+}} xmm1 = zero,zero,xmm1[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
+; SSE-NEXT: pand %xmm4, %xmm1
+; SSE-NEXT: pandn %xmm6, %xmm4
+; SSE-NEXT: por %xmm1, %xmm4
+; SSE-NEXT: movaps %xmm3, %xmm1
+; SSE-NEXT: shufpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: # xmm1 = xmm1[1],mem[0]
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT: punpckhwd {{.*#+}} xmm3 = xmm3[4,4,5,5,6,6,7,7]
+; SSE-NEXT: shufps {{.*#+}} xmm1 = xmm1[2,0],xmm3[2,1]
+; SSE-NEXT: andps %xmm2, %xmm1
; SSE-NEXT: pandn %xmm4, %xmm2
-; SSE-NEXT: por %xmm3, %xmm2
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = xmm0[0],mem[0]
-; SSE-NEXT: shufps $98, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT: # xmm0 = xmm0[2,0],mem[2,1]
+; SSE-NEXT: por %xmm1, %xmm2
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; SSE-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT: # xmm6 = xmm6[0],mem[0]
+; SSE-NEXT: shufps $98, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
+; SSE-NEXT: # xmm6 = xmm6[2,0],mem[2,1]
; SSE-NEXT: pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[0,0,1,1]
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm10[0,0,1,1]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm0[0,0,1,1]
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
; SSE-NEXT: movdqa {{.*#+}} xmm4 = [65535,65535,65535,65535,0,0,0,65535]
; SSE-NEXT: movdqa %xmm4, %xmm10
; SSE-NEXT: pandn %xmm1, %xmm10
-; SSE-NEXT: andps %xmm4, %xmm0
-; SSE-NEXT: por %xmm0, %xmm10
+; SSE-NEXT: andps %xmm4, %xmm6
+; SSE-NEXT: por %xmm6, %xmm10
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: unpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; SSE-NEXT: # xmm0 = xmm0[0],mem[0]
@@ -8947,7 +8949,7 @@ define void @store_i16_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; SSE-NEXT: # xmm0 = xmm0[2,0],mem[2,1]
; SSE-NEXT: pshufd $80, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; SSE-NEXT: # xmm1 = mem[0,0,1,1]
-; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm6[0,0,1,1]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm13[0,0,1,1]
; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
; SSE-NEXT: movdqa %xmm4, %xmm6
; SSE-NEXT: pandn %xmm1, %xmm6
@@ -9102,17 +9104,17 @@ define void @store_i16_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 96(%rax)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 880(%rax)
-; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 816(%rax)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT: movaps %xmm0, 768(%rax)
+; SSE-NEXT: movaps %xmm0, 880(%rax)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 848(%rax)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 832(%rax)
; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; SSE-NEXT: movaps %xmm0, 800(%rax)
+; SSE-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT: movaps %xmm0, 768(%rax)
; SSE-NEXT: addq $1640, %rsp # imm = 0x668
; SSE-NEXT: retq
;
diff --git a/llvm/test/CodeGen/X86/vector-shuffle-combining-avx512bwvl.ll b/llvm/test/CodeGen/X86/vector-shuffle-combining-avx512bwvl.ll
index 1f7de6115fe51..e3e993fec41c4 100644
--- a/llvm/test/CodeGen/X86/vector-shuffle-combining-avx512bwvl.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-combining-avx512bwvl.ll
@@ -206,6 +206,8 @@ define void @PR46178(ptr %0) {
; X86-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0
; X86-NEXT: vpsllw $8, %ymm0, %ymm0
; X86-NEXT: vpsraw $8, %ymm0, %ymm0
+; X86-NEXT: vpxor %xmm1, %xmm1, %xmm1
+; X86-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
; X86-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,2,1,1]
; X86-NEXT: vmovdqu %ymm0, (%eax)
; X86-NEXT: vzeroupper
@@ -220,6 +222,8 @@ define void @PR46178(ptr %0) {
; X64-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0
; X64-NEXT: vpsllw $8, %ymm0, %ymm0
; X64-NEXT: vpsraw $8, %ymm0, %ymm0
+; X64-NEXT: vpxor %xmm1, %xmm1, %xmm1
+; X64-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
; X64-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,2,1,1]
; X64-NEXT: vmovdqu %ymm0, (%rdi)
; X64-NEXT: vzeroupper
diff --git a/llvm/test/CodeGen/X86/vector-shuffle-combining.ll b/llvm/test/CodeGen/X86/vector-shuffle-combining.ll
index a24b5508c1ac7..f31c1d5667f2f 100644
--- a/llvm/test/CodeGen/X86/vector-shuffle-combining.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-combining.ll
@@ -2034,6 +2034,7 @@ define <16 x i8> @combine_and_or_shuffle(<16 x i8> %x, <16 x i8> %y) {
; SSE2-NEXT: packuswb %xmm0, %xmm0
; SSE2-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3],xmm0[4],xmm3[4],xmm0[5],xmm3[5],xmm0[6],xmm3[6],xmm0[7],xmm3[7]
; SSE2-NEXT: por %xmm2, %xmm0
+; SSE2-NEXT: pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; SSE2-NEXT: retq
;
; SSSE3-LABEL: combine_and_or_shuffle:
More information about the llvm-commits
mailing list