[llvm] ef38880 - [X86 isel] Remove lane requirement from lowerShuffleAsUNPCKAndPermute
Han Zhu via llvm-commits
llvm-commits at lists.llvm.org
Wed Apr 12 17:11:27 PDT 2023
Author: Han Zhu
Date: 2023-04-12T17:11:18-07:00
New Revision: ef38880ce03bc1f1fb3606c5a629151f3d0e975e
URL: https://github.com/llvm/llvm-project/commit/ef38880ce03bc1f1fb3606c5a629151f3d0e975e
DIFF: https://github.com/llvm/llvm-project/commit/ef38880ce03bc1f1fb3606c5a629151f3d0e975e.diff
LOG: [X86 isel] Remove lane requirement from lowerShuffleAsUNPCKAndPermute
`lowerShuffleAsUNPCKAndPermute` requires the shuffle mask element to be
in the same lane in both the input and output vectors. This prevents it from
matching certain patterns for example in [GHI
61964](https://github.com/llvm/llvm-project/issues/61964). Removing the lane
requirement fixes the issue.
The change I'm targeting is in the test llvm/test/CodeGen/X86/pr61964.ll. The
codegen has improved notably with this patch. Otherwise, looks like some
broadcast instructions are replaced with unpck and perm. To check if there's
any other performance change, I ran llvm-test-suite benchmarks from the
SingleSource, MultiSource, and MicroBenchmarks directories:
```
Tests: 2665
Short Running: 2009 (filtered out)
Same hash: 140 (filtered out)
In Blacklist: 513 (filtered out)
Remaining: 3
Metric: exec_time
Program exec_time
lhs rhs diff
test-suite :: MultiSource/Benchmarks/ASCI_Purple/SMG2000/smg2000.test 1.64 1.64 0.1%
test-suite :: SingleSource/Benchmarks/Adobe-C++/loop_unroll.test 1.06 1.06 0.0%
test-suite :: MultiSource/Applications/JM/lencod/lencod.test 5.25 5.25 0.0%
Geomean difference nan nan 0.0%
exec_time
l/r lhs rhs diff
count 3.000000 3.000000 3.000000
mean 2.648300 2.649100 0.000462
std 2.269035 2.268849 0.000415
min 1.055500 1.055900 0.000095
25% 1.349300 1.350250 0.000237
50% 1.643100 1.644600 0.000379
75% 3.444700 3.445700 0.000646
max 5.246300 5.246800 0.000913
```
The patch only hits three cases and the result is neutral. (The 513 blacklisted
benchmarks are the ones under MicroBenchmarks, which `--filter-hash` does
not work and I manually verified their code did not change).
Differential Revision: https://reviews.llvm.org/D147668
Added:
Modified:
llvm/lib/Target/X86/X86ISelLowering.cpp
llvm/test/CodeGen/X86/oddshuffles.ll
llvm/test/CodeGen/X86/pr61964.ll
llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-8.ll
llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-5.ll
llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-8.ll
llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-5.ll
llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-7.ll
llvm/test/CodeGen/X86/vector-shuffle-256-v16.ll
llvm/test/CodeGen/X86/vector-shuffle-256-v8.ll
Removed:
################################################################################
diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index a621c83075bb4..15496d8a2fe37 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -13288,47 +13288,52 @@ static SDValue lowerShuffleAsUNPCKAndPermute(const SDLoc &DL, MVT VT,
SDValue Ops[2] = {DAG.getUNDEF(VT), DAG.getUNDEF(VT)};
// Determine UNPCKL/UNPCKH type and operand order.
- for (int Lane = 0; Lane != NumElts; Lane += NumLaneElts) {
- for (int Elt = 0; Elt != NumLaneElts; ++Elt) {
- int M = Mask[Lane + Elt];
- if (M < 0)
- continue;
+ for (int Elt = 0; Elt != NumElts; ++Elt) {
+ int M = Mask[Elt];
+ if (M < 0)
+ continue;
- SDValue &Op = Ops[Elt & 1];
- if (M < NumElts && (Op.isUndef() || Op == V1))
- Op = V1;
- else if (NumElts <= M && (Op.isUndef() || Op == V2))
- Op = V2;
- else
- return SDValue();
+ // Normalize the mask value depending on whether it's V1 or V2.
+ int NormM = M;
+ SDValue &Op = Ops[Elt & 1];
+ if (M < NumElts && (Op.isUndef() || Op == V1))
+ Op = V1;
+ else if (NumElts <= M && (Op.isUndef() || Op == V2)) {
+ Op = V2;
+ NormM -= NumElts;
+ } else
+ return SDValue();
+ bool MatchLoAnyLane = false, MatchHiAnyLane = false;
+ for (int Lane = 0; Lane != NumElts; Lane += NumLaneElts) {
int Lo = Lane, Mid = Lane + NumHalfLaneElts, Hi = Lane + NumLaneElts;
- MatchLo &= isUndefOrInRange(M, Lo, Mid) ||
- isUndefOrInRange(M, NumElts + Lo, NumElts + Mid);
- MatchHi &= isUndefOrInRange(M, Mid, Hi) ||
- isUndefOrInRange(M, NumElts + Mid, NumElts + Hi);
- if (!MatchLo && !MatchHi)
- return SDValue();
+ MatchLoAnyLane |= isUndefOrInRange(NormM, Lo, Mid);
+ MatchHiAnyLane |= isUndefOrInRange(NormM, Mid, Hi);
+ if (MatchLoAnyLane || MatchHiAnyLane) {
+ assert((MatchLoAnyLane ^ MatchHiAnyLane) &&
+ "Failed to match UNPCKLO/UNPCKHI");
+ break;
+ }
}
+ MatchLo &= MatchLoAnyLane;
+ MatchHi &= MatchHiAnyLane;
+ if (!MatchLo && !MatchHi)
+ return SDValue();
}
assert((MatchLo ^ MatchHi) && "Failed to match UNPCKLO/UNPCKHI");
- // Now check that each pair of elts come from the same unpack pair
- // and set the permute mask based on each pair.
- // TODO - Investigate cases where we permute individual elements.
+ // Element indices have changed after unpacking. Calculate permute mask
+ // so that they will be put back to the position as dictated by the
+ // original shuffle mask indices.
SmallVector<int, 32> PermuteMask(NumElts, -1);
- for (int Lane = 0; Lane != NumElts; Lane += NumLaneElts) {
- for (int Elt = 0; Elt != NumLaneElts; Elt += 2) {
- int M0 = Mask[Lane + Elt + 0];
- int M1 = Mask[Lane + Elt + 1];
- if (0 <= M0 && 0 <= M1 &&
- (M0 % NumHalfLaneElts) != (M1 % NumHalfLaneElts))
- return SDValue();
- if (0 <= M0)
- PermuteMask[Lane + Elt + 0] = Lane + (2 * (M0 % NumHalfLaneElts));
- if (0 <= M1)
- PermuteMask[Lane + Elt + 1] = Lane + (2 * (M1 % NumHalfLaneElts)) + 1;
- }
+ for (int Elt = 0; Elt != NumElts; ++Elt) {
+ int M = Mask[Elt];
+ if (NumElts <= M)
+ PermuteMask[Elt] = NumLaneElts * ((M - NumElts) / NumLaneElts) +
+ (2 * (M % NumHalfLaneElts)) + 1;
+ else if (0 <= M)
+ PermuteMask[Elt] =
+ NumLaneElts * (M / NumLaneElts) + (2 * (M % NumHalfLaneElts));
}
unsigned UnpckOp = MatchLo ? X86ISD::UNPCKL : X86ISD::UNPCKH;
@@ -13548,6 +13553,24 @@ static bool isNoopOrBroadcastShuffleMask(ArrayRef<int> Mask) {
return isNoopShuffleMask(Mask) || isBroadcastShuffleMask(Mask);
}
+/// Check if the Mask consists of the same element repeated multiple times.
+static bool isSingleElementRepeatedMask(ArrayRef<int> Mask) {
+ size_t NumUndefs = 0;
+ std::optional<int> UniqueElt;
+ for (int Elt : Mask) {
+ if (Elt == SM_SentinelUndef) {
+ NumUndefs++;
+ continue;
+ }
+ if (UniqueElt.has_value() && UniqueElt.value() != Elt)
+ return false;
+ UniqueElt = Elt;
+ }
+ // Make sure the element is repeated enough times by checking the number of
+ // undefs is small.
+ return NumUndefs <= Mask.size() / 2 && UniqueElt.has_value();
+}
+
/// Generic routine to decompose a shuffle and blend into independent
/// blends and permutes.
///
@@ -13623,9 +13646,17 @@ static SDValue lowerShuffleAsDecomposedShuffleMerge(
if (SDValue BlendPerm = lowerShuffleAsBlendAndPermute(DL, VT, V1, V2, Mask,
DAG, true))
return BlendPerm;
- if (SDValue UnpackPerm = lowerShuffleAsUNPCKAndPermute(DL, VT, V1, V2, Mask,
- DAG))
- return UnpackPerm;
+ // If either input vector provides only a single element which is repeated
+ // multiple times, unpacking from both input vectors would generate worse
+ // code. e.g. for
+ // t5: v16i8 = vector_shuffle<16,0,16,1,16,2,16,3,16,4,16,5,16,6,16,7> t2, t4
+ // it is better to process t4 first to create a vector of t4[0], then unpack
+ // that vector with t2.
+ if (!isSingleElementRepeatedMask(V1Mask) &&
+ !isSingleElementRepeatedMask(V2Mask))
+ if (SDValue UnpackPerm =
+ lowerShuffleAsUNPCKAndPermute(DL, VT, V1, V2, Mask, DAG))
+ return UnpackPerm;
if (SDValue RotatePerm = lowerShuffleAsByteRotateAndPermute(
DL, VT, V1, V2, Mask, Subtarget, DAG))
return RotatePerm;
diff --git a/llvm/test/CodeGen/X86/oddshuffles.ll b/llvm/test/CodeGen/X86/oddshuffles.ll
index 1bf1298fc6bad..b3b94e83f5c21 100644
--- a/llvm/test/CodeGen/X86/oddshuffles.ll
+++ b/llvm/test/CodeGen/X86/oddshuffles.ll
@@ -2437,35 +2437,35 @@ define void @D107009(ptr %input, ptr %output) {
;
; AVX2-LABEL: D107009:
; AVX2: # %bb.0:
-; AVX2-NEXT: vmovdqu 64(%rdi), %ymm0
-; AVX2-NEXT: vmovdqu 128(%rdi), %ymm1
-; AVX2-NEXT: vmovdqu 192(%rdi), %ymm2
+; AVX2-NEXT: vmovdqu (%rdi), %ymm0
+; AVX2-NEXT: vmovdqu 64(%rdi), %ymm1
+; AVX2-NEXT: vmovdqu 128(%rdi), %ymm2
+; AVX2-NEXT: vmovdqu 192(%rdi), %ymm3
+; AVX2-NEXT: vpunpckldq {{.*#+}} ymm3 = ymm3[0],mem[0],ymm3[1],mem[1],ymm3[4],mem[4],ymm3[5],mem[5]
; AVX2-NEXT: vpunpckldq {{.*#+}} ymm2 = ymm2[0],mem[0],ymm2[1],mem[1],ymm2[4],mem[4],ymm2[5],mem[5]
-; AVX2-NEXT: vpunpckldq {{.*#+}} ymm1 = ymm1[0],mem[0],ymm1[1],mem[1],ymm1[4],mem[4],ymm1[5],mem[5]
-; AVX2-NEXT: vpunpcklqdq {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
-; AVX2-NEXT: vpbroadcastd 48(%rdi), %xmm2
-; AVX2-NEXT: vpbroadcastd 16(%rdi), %ymm3
-; AVX2-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
+; AVX2-NEXT: vpunpcklqdq {{.*#+}} ymm2 = ymm2[0],ymm3[0],ymm2[2],ymm3[2]
; AVX2-NEXT: vpunpckldq {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[4],mem[4],ymm0[5],mem[5]
-; AVX2-NEXT: vpermq {{.*#+}} ymm0 = ymm0[2,2,2,2]
-; AVX2-NEXT: vpblendd {{.*#+}} xmm0 = xmm2[0,1],xmm0[2,3]
-; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm0
+; AVX2-NEXT: vpunpckldq {{.*#+}} ymm1 = ymm1[0],mem[0],ymm1[1],mem[1],ymm1[4],mem[4],ymm1[5],mem[5]
+; AVX2-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2]
+; AVX2-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
+; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
; AVX2-NEXT: vpsrld $16, %ymm0, %ymm0
-; AVX2-NEXT: vpmovzxdq {{.*#+}} ymm1 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero
-; AVX2-NEXT: vpshufd {{.*#+}} xmm2 = xmm0[3,3,3,3]
-; AVX2-NEXT: vpshufd {{.*#+}} xmm3 = xmm0[1,1,1,1]
-; AVX2-NEXT: vpbroadcastd %xmm0, %ymm4
-; AVX2-NEXT: vpshufd {{.*#+}} ymm5 = ymm0[3,3,3,3,7,7,7,7]
-; AVX2-NEXT: vpshufd {{.*#+}} ymm6 = ymm0[2,3,2,3,6,7,6,7]
-; AVX2-NEXT: vpshufd {{.*#+}} ymm7 = ymm0[1,1,1,1,5,5,5,5]
+; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
+; AVX2-NEXT: vpbroadcastd %xmm0, %ymm2
+; AVX2-NEXT: vpshufd {{.*#+}} xmm3 = xmm0[3,3,3,3]
+; AVX2-NEXT: vpmovzxdq {{.*#+}} ymm4 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero
+; AVX2-NEXT: vpshufd {{.*#+}} ymm5 = ymm0[1,1,1,1,5,5,5,5]
+; AVX2-NEXT: vpshufd {{.*#+}} ymm6 = ymm0[3,3,3,3,7,7,7,7]
+; AVX2-NEXT: vpshufd {{.*#+}} ymm7 = ymm0[2,3,2,3,6,7,6,7]
; AVX2-NEXT: vmovdqu %ymm0, 128(%rsi)
-; AVX2-NEXT: vmovdqu %ymm7, 160(%rsi)
-; AVX2-NEXT: vmovdqu %ymm6, 192(%rsi)
-; AVX2-NEXT: vmovdqu %ymm5, 224(%rsi)
-; AVX2-NEXT: vmovdqu %ymm4, (%rsi)
-; AVX2-NEXT: vmovdqa %xmm3, 48(%rsi)
-; AVX2-NEXT: vmovdqa %xmm2, 112(%rsi)
-; AVX2-NEXT: vmovdqu %ymm1, 64(%rsi)
+; AVX2-NEXT: vmovdqu %ymm7, 192(%rsi)
+; AVX2-NEXT: vmovdqu %ymm6, 224(%rsi)
+; AVX2-NEXT: vmovdqu %ymm5, 160(%rsi)
+; AVX2-NEXT: vmovdqu %ymm4, 64(%rsi)
+; AVX2-NEXT: vmovdqa %xmm3, 112(%rsi)
+; AVX2-NEXT: vmovdqu %ymm2, (%rsi)
+; AVX2-NEXT: vmovdqa %xmm1, 48(%rsi)
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
diff --git a/llvm/test/CodeGen/X86/pr61964.ll b/llvm/test/CodeGen/X86/pr61964.ll
index 6cc9d6d1f4ee1..24f150727a184 100644
--- a/llvm/test/CodeGen/X86/pr61964.ll
+++ b/llvm/test/CodeGen/X86/pr61964.ll
@@ -29,17 +29,11 @@ define { <8 x i32>, <8 x i32> } @splitTransposeDecode_8_avx2(<16 x i16> %a, <16
;
; AVX2-LABEL: splitTransposeDecode_8_avx2:
; AVX2: # %bb.0:
-; AVX2-NEXT: vpermq {{.*#+}} ymm2 = ymm1[0,2,0,2]
-; AVX2-NEXT: vmovdqa {{.*#+}} ymm3 = <0,1,8,9,2,3,10,11,u,u,u,u,u,u,u,u,4,5,12,13,6,7,14,15,u,u,u,u,u,u,u,u>
-; AVX2-NEXT: vpshufb %ymm3, %ymm2, %ymm2
-; AVX2-NEXT: vpermq {{.*#+}} ymm4 = ymm0[0,2,0,2]
-; AVX2-NEXT: vpshufb %ymm3, %ymm4, %ymm4
-; AVX2-NEXT: vpunpcklwd {{.*#+}} ymm2 = ymm4[0],ymm2[0],ymm4[1],ymm2[1],ymm4[2],ymm2[2],ymm4[3],ymm2[3],ymm4[8],ymm2[8],ymm4[9],ymm2[9],ymm4[10],ymm2[10],ymm4[11],ymm2[11]
-; AVX2-NEXT: vpermq {{.*#+}} ymm1 = ymm1[1,3,1,3]
-; AVX2-NEXT: vpshufb %ymm3, %ymm1, %ymm1
-; AVX2-NEXT: vpermq {{.*#+}} ymm0 = ymm0[1,3,1,3]
-; AVX2-NEXT: vpshufb %ymm3, %ymm0, %ymm0
-; AVX2-NEXT: vpunpcklwd {{.*#+}} ymm1 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11]
+; AVX2-NEXT: vpunpcklwd {{.*#+}} ymm2 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11]
+; AVX2-NEXT: vmovdqa {{.*#+}} ymm3 = [0,4,1,5,2,6,3,7]
+; AVX2-NEXT: vpermd %ymm2, %ymm3, %ymm2
+; AVX2-NEXT: vpunpckhwd {{.*#+}} ymm0 = ymm0[4],ymm1[4],ymm0[5],ymm1[5],ymm0[6],ymm1[6],ymm0[7],ymm1[7],ymm0[12],ymm1[12],ymm0[13],ymm1[13],ymm0[14],ymm1[14],ymm0[15],ymm1[15]
+; AVX2-NEXT: vpermd %ymm0, %ymm3, %ymm1
; AVX2-NEXT: vmovdqa %ymm2, %ymm0
; AVX2-NEXT: retq
;
@@ -72,17 +66,11 @@ define { <8 x i32>, <8 x i32> } @splitTransposeDecode_8_avx2(<16 x i16> %a, <16
;
; XOPAVX2-LABEL: splitTransposeDecode_8_avx2:
; XOPAVX2: # %bb.0:
-; XOPAVX2-NEXT: vpermq {{.*#+}} ymm2 = ymm1[0,2,0,2]
-; XOPAVX2-NEXT: vmovdqa {{.*#+}} ymm3 = <0,1,8,9,2,3,10,11,u,u,u,u,u,u,u,u,4,5,12,13,6,7,14,15,u,u,u,u,u,u,u,u>
-; XOPAVX2-NEXT: vpshufb %ymm3, %ymm2, %ymm2
-; XOPAVX2-NEXT: vpermq {{.*#+}} ymm4 = ymm0[0,2,0,2]
-; XOPAVX2-NEXT: vpshufb %ymm3, %ymm4, %ymm4
-; XOPAVX2-NEXT: vpunpcklwd {{.*#+}} ymm2 = ymm4[0],ymm2[0],ymm4[1],ymm2[1],ymm4[2],ymm2[2],ymm4[3],ymm2[3],ymm4[8],ymm2[8],ymm4[9],ymm2[9],ymm4[10],ymm2[10],ymm4[11],ymm2[11]
-; XOPAVX2-NEXT: vpermq {{.*#+}} ymm1 = ymm1[1,3,1,3]
-; XOPAVX2-NEXT: vpshufb %ymm3, %ymm1, %ymm1
-; XOPAVX2-NEXT: vpermq {{.*#+}} ymm0 = ymm0[1,3,1,3]
-; XOPAVX2-NEXT: vpshufb %ymm3, %ymm0, %ymm0
-; XOPAVX2-NEXT: vpunpcklwd {{.*#+}} ymm1 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11]
+; XOPAVX2-NEXT: vpunpcklwd {{.*#+}} ymm2 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11]
+; XOPAVX2-NEXT: vmovdqa {{.*#+}} ymm3 = [0,4,1,5,2,6,3,7]
+; XOPAVX2-NEXT: vpermd %ymm2, %ymm3, %ymm2
+; XOPAVX2-NEXT: vpunpckhwd {{.*#+}} ymm0 = ymm0[4],ymm1[4],ymm0[5],ymm1[5],ymm0[6],ymm1[6],ymm0[7],ymm1[7],ymm0[12],ymm1[12],ymm0[13],ymm1[13],ymm0[14],ymm1[14],ymm0[15],ymm1[15]
+; XOPAVX2-NEXT: vpermd %ymm0, %ymm3, %ymm1
; XOPAVX2-NEXT: vmovdqa %ymm2, %ymm0
; XOPAVX2-NEXT: retq
%shuffle.i = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 8, i32 24, i32 9, i32 25, i32 10, i32 26, i32 11, i32 27>
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-8.ll b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-8.ll
index 39238d7438b2a..5f6de8f4b8bf4 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-8.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-load-i32-stride-8.ll
@@ -92,9 +92,8 @@ define void @load_i32_stride8_vf2(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX2-ONLY-NEXT: vpshufd {{.*#+}} xmm5 = xmm2[1,1,1,1]
; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm5 = xmm5[0],xmm3[1],xmm5[2,3]
; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} xmm2 = xmm2[2],xmm3[2],xmm2[3],xmm3[3]
-; AVX2-ONLY-NEXT: vbroadcastss 48(%rdi), %xmm3
-; AVX2-ONLY-NEXT: vbroadcastss 16(%rdi), %ymm6
-; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm3 = xmm6[0],xmm3[0],xmm6[1],xmm3[1]
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm3 = ymm1[0],ymm0[0],ymm1[1],ymm0[1],ymm1[4],ymm0[4],ymm1[5],ymm0[5]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm3, %xmm3
; AVX2-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm1[1,1,1,1,5,5,5,5]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0],ymm0[1],ymm6[2,3,4],ymm0[5],ymm6[6,7]
; AVX2-ONLY-NEXT: vextractf128 $1, %ymm6, %xmm6
@@ -379,59 +378,56 @@ define void @load_i32_stride8_vf4(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
; AVX2-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX2-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %r10
; AVX2-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %r11
-; AVX2-ONLY-NEXT: vmovdqa 32(%rdi), %ymm0
-; AVX2-ONLY-NEXT: vmovdqa (%rdi), %ymm2
-; AVX2-ONLY-NEXT: vmovdqa 96(%rdi), %ymm1
-; AVX2-ONLY-NEXT: vmovdqa 64(%rdi), %ymm3
-; AVX2-ONLY-NEXT: vmovaps 96(%rdi), %xmm6
-; AVX2-ONLY-NEXT: vbroadcastss %xmm6, %xmm4
-; AVX2-ONLY-NEXT: vmovaps (%rdi), %xmm7
-; AVX2-ONLY-NEXT: vmovaps 32(%rdi), %xmm8
-; AVX2-ONLY-NEXT: vmovaps 64(%rdi), %xmm9
-; AVX2-ONLY-NEXT: vbroadcastss %xmm9, %xmm5
-; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm4 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
-; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm5 = xmm7[0],xmm8[0],xmm7[1],xmm8[1]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm5[0,1],xmm4[2,3]
-; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm5 = xmm9[0],xmm6[0],xmm9[1],xmm6[1]
-; AVX2-ONLY-NEXT: vshufps {{.*#+}} xmm10 = xmm7[1,1,1,1]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm10 = xmm10[0],xmm8[1],xmm10[2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm10[0,1],xmm5[2,3]
-; AVX2-ONLY-NEXT: vshufps {{.*#+}} xmm10 = xmm6[2,2,2,2]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm10 = xmm9[0,1,2],xmm10[3]
-; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm7 = xmm7[2],xmm8[2],xmm7[3],xmm8[3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm8 = xmm7[0,1],xmm10[2,3]
-; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm6 = xmm9[2],xmm6[2],xmm9[3],xmm6[3]
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm6 = xmm7[1],xmm6[1]
-; AVX2-ONLY-NEXT: vpbroadcastd 48(%rdi), %xmm7
-; AVX2-ONLY-NEXT: vpbroadcastd 16(%rdi), %ymm9
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} xmm7 = xmm9[0],xmm7[0],xmm9[1],xmm7[1]
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm9 = ymm3[0],ymm1[0],ymm3[1],ymm1[1],ymm3[4],ymm1[4],ymm3[5],ymm1[5]
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm9 = ymm9[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm7 = xmm7[0,1],xmm9[2,3]
-; AVX2-ONLY-NEXT: vpmovzxdq {{.*#+}} xmm9 = mem[0],zero,mem[1],zero
-; AVX2-ONLY-NEXT: vpbroadcastd 116(%rdi), %ymm10
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm9 = xmm9[0,1,2],xmm10[3]
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm10 = ymm2[1,1,1,1,5,5,5,5]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm10 = ymm10[0],ymm0[1],ymm10[2,3,4],ymm0[5],ymm10[6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm10, %xmm10
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm9 = xmm10[0,1],xmm9[2,3]
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm0 = ymm2[2],ymm0[2],ymm2[3],ymm0[3],ymm2[6],ymm0[6],ymm2[7],ymm0[7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm0, %xmm2
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm1 = ymm3[2],ymm1[2],ymm3[3],ymm1[3],ymm3[6],ymm1[6],ymm3[7],ymm1[7]
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm3 = ymm1[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm2 = xmm2[0,1],xmm3[2,3]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm1, %xmm1
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[2,3,2,3,6,7,6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm0, %xmm0
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
-; AVX2-ONLY-NEXT: vmovaps %xmm4, (%rsi)
-; AVX2-ONLY-NEXT: vmovaps %xmm5, (%rdx)
-; AVX2-ONLY-NEXT: vmovaps %xmm8, (%rcx)
-; AVX2-ONLY-NEXT: vmovaps %xmm6, (%r8)
-; AVX2-ONLY-NEXT: vmovdqa %xmm7, (%r9)
-; AVX2-ONLY-NEXT: vmovdqa %xmm9, (%r11)
-; AVX2-ONLY-NEXT: vmovdqa %xmm2, (%r10)
-; AVX2-ONLY-NEXT: vmovdqa %xmm0, (%rax)
+; AVX2-ONLY-NEXT: vmovaps 96(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovaps 64(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovaps 32(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vmovaps (%rdi), %ymm3
+; AVX2-ONLY-NEXT: vmovaps 96(%rdi), %xmm4
+; AVX2-ONLY-NEXT: vbroadcastss %xmm4, %xmm5
+; AVX2-ONLY-NEXT: vmovaps (%rdi), %xmm6
+; AVX2-ONLY-NEXT: vmovaps 32(%rdi), %xmm7
+; AVX2-ONLY-NEXT: vmovaps 64(%rdi), %xmm8
+; AVX2-ONLY-NEXT: vbroadcastss %xmm8, %xmm9
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm5 = xmm9[0],xmm5[0],xmm9[1],xmm5[1]
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm9 = xmm6[0],xmm7[0],xmm6[1],xmm7[1]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm9[0,1],xmm5[2,3]
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm9 = xmm8[0],xmm4[0],xmm8[1],xmm4[1]
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} xmm10 = xmm6[1,1,1,1]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm10 = xmm10[0],xmm7[1],xmm10[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm9 = xmm10[0,1],xmm9[2,3]
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} xmm10 = xmm4[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm10 = xmm8[0,1,2],xmm10[3]
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm6 = xmm6[2],xmm7[2],xmm6[3],xmm7[3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm7 = xmm6[0,1],xmm10[2,3]
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm4 = xmm8[2],xmm4[2],xmm8[3],xmm4[3]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm4 = xmm6[1],xmm4[1]
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm6 = ymm3[0],ymm2[0],ymm3[1],ymm2[1],ymm3[4],ymm2[4],ymm3[5],ymm2[5]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm6, %xmm6
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm8 = ymm1[0],ymm0[0],ymm1[1],ymm0[1],ymm1[4],ymm0[4],ymm1[5],ymm0[5]
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm10 = ymm8[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0,1],xmm10[2,3]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm8, %xmm8
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm3[1,1,1,1,5,5,5,5]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0],ymm2[1],ymm10[2,3,4],ymm2[5],ymm10[6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm10, %xmm10
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm8 = xmm10[0,1],xmm8[2,3]
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm2 = ymm3[2],ymm2[2],ymm3[3],ymm2[3],ymm3[6],ymm2[6],ymm3[7],ymm2[7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm3
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm0 = ymm1[2],ymm0[2],ymm1[3],ymm0[3],ymm1[6],ymm0[6],ymm1[7],ymm0[7]
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm1 = ymm0[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm3[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm0, %xmm0
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm2[2,3,2,3,6,7,6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm2
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm2[0,1],xmm0[2,3]
+; AVX2-ONLY-NEXT: vmovaps %xmm5, (%rsi)
+; AVX2-ONLY-NEXT: vmovaps %xmm9, (%rdx)
+; AVX2-ONLY-NEXT: vmovaps %xmm7, (%rcx)
+; AVX2-ONLY-NEXT: vmovaps %xmm4, (%r8)
+; AVX2-ONLY-NEXT: vmovaps %xmm6, (%r9)
+; AVX2-ONLY-NEXT: vmovaps %xmm8, (%r11)
+; AVX2-ONLY-NEXT: vmovaps %xmm1, (%r10)
+; AVX2-ONLY-NEXT: vmovaps %xmm0, (%rax)
; AVX2-ONLY-NEXT: vzeroupper
; AVX2-ONLY-NEXT: retq
;
@@ -701,117 +697,114 @@ define void @load_i32_stride8_vf8(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, ptr
;
; AVX2-ONLY-LABEL: load_i32_stride8_vf8:
; AVX2-ONLY: # %bb.0:
-; AVX2-ONLY-NEXT: vmovaps 32(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovaps 96(%rdi), %ymm0
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa (%rdi), %ymm2
-; AVX2-ONLY-NEXT: vmovdqa 96(%rdi), %ymm1
-; AVX2-ONLY-NEXT: vmovdqa 64(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vmovaps 64(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovaps 32(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vmovaps (%rdi), %ymm3
; AVX2-ONLY-NEXT: vmovaps 160(%rdi), %xmm7
-; AVX2-ONLY-NEXT: vmovaps 128(%rdi), %xmm9
-; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm4 = xmm9[0],xmm7[0],xmm9[1],xmm7[1]
+; AVX2-ONLY-NEXT: vmovaps 128(%rdi), %xmm11
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm4 = xmm11[0],xmm7[0],xmm11[1],xmm7[1]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm0, %ymm4
; AVX2-ONLY-NEXT: vmovaps 224(%rdi), %xmm8
; AVX2-ONLY-NEXT: vbroadcastss %xmm8, %xmm5
-; AVX2-ONLY-NEXT: vmovaps 192(%rdi), %xmm10
-; AVX2-ONLY-NEXT: vbroadcastss %xmm10, %xmm6
+; AVX2-ONLY-NEXT: vmovaps 192(%rdi), %xmm9
+; AVX2-ONLY-NEXT: vbroadcastss %xmm9, %xmm6
; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm5 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm5
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm5[6,7]
-; AVX2-ONLY-NEXT: vmovaps 96(%rdi), %xmm11
-; AVX2-ONLY-NEXT: vbroadcastss %xmm11, %xmm5
-; AVX2-ONLY-NEXT: vmovaps (%rdi), %xmm12
-; AVX2-ONLY-NEXT: vmovaps 32(%rdi), %xmm13
-; AVX2-ONLY-NEXT: vmovaps 64(%rdi), %xmm14
-; AVX2-ONLY-NEXT: vbroadcastss %xmm14, %xmm6
+; AVX2-ONLY-NEXT: vmovaps 96(%rdi), %xmm10
+; AVX2-ONLY-NEXT: vbroadcastss %xmm10, %xmm5
+; AVX2-ONLY-NEXT: vmovaps (%rdi), %xmm13
+; AVX2-ONLY-NEXT: vmovaps 32(%rdi), %xmm14
+; AVX2-ONLY-NEXT: vmovaps 64(%rdi), %xmm12
+; AVX2-ONLY-NEXT: vbroadcastss %xmm12, %xmm6
; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm5 = xmm6[0],xmm5[0],xmm6[1],xmm5[1]
-; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm6 = xmm12[0],xmm13[0],xmm12[1],xmm13[1]
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm6 = xmm13[0],xmm14[0],xmm13[1],xmm14[1]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm6[0,1],xmm5[2,3]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3],ymm4[4,5,6,7]
-; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm5 = xmm14[0],xmm11[0],xmm14[1],xmm11[1]
-; AVX2-ONLY-NEXT: vshufps {{.*#+}} xmm6 = xmm12[1,1,1,1]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0],xmm13[1],xmm6[2,3]
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm5 = xmm12[0],xmm10[0],xmm12[1],xmm10[1]
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} xmm6 = xmm13[1,1,1,1]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0],xmm14[1],xmm6[2,3]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm6[0,1],xmm5[2,3]
-; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm6 = xmm10[0],xmm8[0],xmm10[1],xmm8[1]
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm6 = xmm9[0],xmm8[0],xmm9[1],xmm8[1]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6
-; AVX2-ONLY-NEXT: vshufps {{.*#+}} xmm15 = xmm9[1,1,1,1]
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} xmm15 = xmm11[1,1,1,1]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm15 = xmm15[0],xmm7[1],xmm15[2,3]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm15, %ymm0, %ymm15
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm15 = ymm15[0,1,2,3,4,5],ymm6[6,7]
-; AVX2-ONLY-NEXT: vmovdqa 224(%rdi), %ymm6
+; AVX2-ONLY-NEXT: vmovaps 224(%rdi), %ymm6
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm15[4,5,6,7]
-; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm15 = xmm9[2],xmm7[2],xmm9[3],xmm7[3]
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm15 = xmm11[2],xmm7[2],xmm11[3],xmm7[3]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm15, %ymm0, %ymm7
-; AVX2-ONLY-NEXT: vshufps {{.*#+}} xmm9 = xmm8[2,2,2,2]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm9 = xmm10[0,1,2],xmm9[3]
-; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm9, %ymm0, %ymm9
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm9[6,7]
-; AVX2-ONLY-NEXT: vmovdqa 192(%rdi), %ymm9
-; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm0 = xmm12[2],xmm13[2],xmm12[3],xmm13[3]
-; AVX2-ONLY-NEXT: vshufps {{.*#+}} xmm12 = xmm11[2,2,2,2]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm12 = xmm14[0,1,2],xmm12[3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm12 = xmm0[0,1],xmm12[2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm12[0,1,2,3],ymm7[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqa 160(%rdi), %ymm12
-; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm8 = xmm10[2],xmm8[2],xmm10[3],xmm8[3]
-; AVX2-ONLY-NEXT: vmovdqa 128(%rdi), %ymm13
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} xmm11 = xmm8[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm11 = xmm9[0,1,2],xmm11[3]
+; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm11, %ymm0, %ymm11
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm11[6,7]
+; AVX2-ONLY-NEXT: vmovaps 192(%rdi), %ymm11
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm13 = xmm13[2],xmm14[2],xmm13[3],xmm14[3]
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} xmm14 = xmm10[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm14 = xmm12[0,1,2],xmm14[3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm14 = xmm13[0,1],xmm14[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm14[0,1,2,3],ymm7[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovaps 160(%rdi), %ymm14
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm8 = xmm9[2],xmm8[2],xmm9[3],xmm8[3]
+; AVX2-ONLY-NEXT: vmovaps 128(%rdi), %ymm0
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm8, %ymm0, %ymm8
-; AVX2-ONLY-NEXT: vshufps {{.*#+}} xmm10 = xmm15[2,3,2,3]
-; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm10, %ymm0, %ymm10
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm10[0,1,2,3,4,5],ymm8[6,7]
-; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm10 = xmm14[2],xmm11[2],xmm14[3],xmm11[3]
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm10[1]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm0[0,1,2,3],ymm8[4,5,6,7]
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm0 = ymm13[0],ymm12[0],ymm13[1],ymm12[1],ymm13[4],ymm12[4],ymm13[5],ymm12[5]
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm11 = ymm9[0],ymm6[0],ymm9[1],ymm6[1],ymm9[4],ymm6[4],ymm9[5],ymm6[5]
-; AVX2-ONLY-NEXT: vpunpcklqdq {{.*#+}} ymm0 = ymm0[0],ymm11[0],ymm0[2],ymm11[2]
-; AVX2-ONLY-NEXT: vpbroadcastd 48(%rdi), %xmm10
-; AVX2-ONLY-NEXT: vpbroadcastd 16(%rdi), %ymm14
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} xmm10 = xmm14[0],xmm10[0],xmm14[1],xmm10[1]
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm14 = ymm3[0],ymm1[0],ymm3[1],ymm1[1],ymm3[4],ymm1[4],ymm3[5],ymm1[5]
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm14 = ymm14[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm10 = xmm10[0,1],xmm14[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm10 = ymm10[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT: vpbroadcastd 148(%rdi), %ymm0
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm12[5],ymm0[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm11[6,7]
-; AVX2-ONLY-NEXT: vpmovzxdq {{.*#+}} xmm11 = mem[0],zero,mem[1],zero
-; AVX2-ONLY-NEXT: vpbroadcastd 116(%rdi), %ymm14
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm11 = xmm11[0,1,2],xmm14[3]
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm14 = ymm2[1,1,1,1,5,5,5,5]
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm14 = ymm14[0],ymm15[1],ymm14[2,3,4],ymm15[5],ymm14[6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm14, %xmm14
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm11 = xmm14[0,1],xmm11[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm11[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT: vpbroadcastd 248(%rdi), %ymm11
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm9 = ymm9[0,1,2,3,4,5,6],ymm11[7]
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm11 = ymm13[2],ymm12[2],ymm13[3],ymm12[3],ymm13[6],ymm12[6],ymm13[7],ymm12[7]
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm2 = ymm2[2],ymm15[2],ymm2[3],ymm15[3],ymm2[6],ymm15[6],ymm2[7],ymm15[7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm2, %xmm12
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm1 = ymm3[2],ymm1[2],ymm3[3],ymm1[3],ymm3[6],ymm1[6],ymm3[7],ymm1[7]
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm3 = ymm1[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm3 = xmm12[0,1],xmm3[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm9 = ymm11[0,1,2,3,4,5],ymm9[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm9[4,5,6,7]
-; AVX2-ONLY-NEXT: vpbroadcastd 220(%rdi), %ymm9
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm6 = ymm9[2],ymm6[2],ymm9[3],ymm6[3],ymm9[6],ymm6[6],ymm9[7],ymm6[7]
-; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm6 = ymm11[1],ymm6[1],ymm11[3],ymm6[3]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm1, %xmm1
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[2,3,2,3,6,7,6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm2, %xmm2
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm6[4,5,6,7]
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} xmm9 = xmm15[2,3,2,3]
+; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm9, %ymm0, %ymm9
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm9[0,1,2,3,4,5],ymm8[6,7]
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm9 = xmm12[2],xmm10[2],xmm12[3],xmm10[3]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm9 = xmm13[1],xmm9[1]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm9[0,1,2,3],ymm8[4,5,6,7]
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm9 = ymm0[0],ymm14[0],ymm0[1],ymm14[1],ymm0[4],ymm14[4],ymm0[5],ymm14[5]
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm10 = ymm3[0],ymm2[0],ymm3[1],ymm2[1],ymm3[4],ymm2[4],ymm3[5],ymm2[5]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm10, %xmm10
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm12 = ymm1[0],ymm15[0],ymm1[1],ymm15[1],ymm1[4],ymm15[4],ymm1[5],ymm15[5]
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm13 = ymm12[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm10 = xmm10[0,1],xmm13[2,3]
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm13 = ymm11[0],ymm6[0],ymm11[1],ymm6[1],ymm11[4],ymm6[4],ymm11[5],ymm6[5]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm9 = ymm9[0],ymm13[0],ymm9[2],ymm13[2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm10[0,1,2,3],ymm9[4,5,6,7]
+; AVX2-ONLY-NEXT: vbroadcastss 148(%rdi), %ymm10
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3,4],ymm14[5],ymm10[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3,4,5],ymm13[6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm12, %xmm12
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} ymm13 = ymm3[1,1,1,1,5,5,5,5]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0],ymm2[1],ymm13[2,3,4],ymm2[5],ymm13[6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm13, %xmm13
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm12 = xmm13[0,1],xmm12[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm12[0,1,2,3],ymm10[4,5,6,7]
+; AVX2-ONLY-NEXT: vbroadcastss 248(%rdi), %ymm12
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3,4,5,6],ymm12[7]
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm0 = ymm0[2],ymm14[2],ymm0[3],ymm14[3],ymm0[6],ymm14[6],ymm0[7],ymm14[7]
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm2 = ymm3[2],ymm2[2],ymm3[3],ymm2[3],ymm3[6],ymm2[6],ymm3[7],ymm2[7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm3
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm1 = ymm1[2],ymm15[2],ymm1[3],ymm15[3],ymm1[6],ymm15[6],ymm1[7],ymm15[7]
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm12 = ymm1[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm12[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm0[0,1,2,3,4,5],ymm11[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm11[4,5,6,7]
+; AVX2-ONLY-NEXT: vbroadcastss 220(%rdi), %ymm11
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm6 = ymm11[2],ymm6[2],ymm11[3],ymm6[3],ymm11[6],ymm6[6],ymm11[7],ymm6[7]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm6[1],ymm0[3],ymm6[3]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm2[2,3,2,3,6,7,6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm2
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovaps %ymm4, (%rsi)
; AVX2-ONLY-NEXT: vmovaps %ymm5, (%rdx)
; AVX2-ONLY-NEXT: vmovaps %ymm7, (%rcx)
; AVX2-ONLY-NEXT: vmovaps %ymm8, (%r8)
-; AVX2-ONLY-NEXT: vmovdqa %ymm10, (%r9)
+; AVX2-ONLY-NEXT: vmovaps %ymm9, (%r9)
; AVX2-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX2-ONLY-NEXT: vmovdqa %ymm0, (%rax)
+; AVX2-ONLY-NEXT: vmovaps %ymm10, (%rax)
; AVX2-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX2-ONLY-NEXT: vmovdqa %ymm3, (%rax)
+; AVX2-ONLY-NEXT: vmovaps %ymm3, (%rax)
; AVX2-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX2-ONLY-NEXT: vmovdqa %ymm1, (%rax)
+; AVX2-ONLY-NEXT: vmovaps %ymm0, (%rax)
; AVX2-ONLY-NEXT: vzeroupper
; AVX2-ONLY-NEXT: retq
;
@@ -1403,7 +1396,7 @@ define void @load_i32_stride8_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
;
; AVX2-ONLY-LABEL: load_i32_stride8_vf16:
; AVX2-ONLY: # %bb.0:
-; AVX2-ONLY-NEXT: subq $328, %rsp # imm = 0x148
+; AVX2-ONLY-NEXT: subq $456, %rsp # imm = 0x1C8
; AVX2-ONLY-NEXT: vmovaps 288(%rdi), %xmm0
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vmovaps 256(%rdi), %xmm2
@@ -1419,8 +1412,8 @@ define void @load_i32_stride8_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
; AVX2-ONLY-NEXT: vmovaps 416(%rdi), %xmm1
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 384(%rdi), %xmm13
-; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm1 = xmm13[0],xmm1[0],xmm13[1],xmm1[1]
+; AVX2-ONLY-NEXT: vmovaps 384(%rdi), %xmm14
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm1 = xmm14[0],xmm1[0],xmm14[1],xmm1[1]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX2-ONLY-NEXT: vmovaps 480(%rdi), %xmm3
; AVX2-ONLY-NEXT: vbroadcastss %xmm3, %xmm2
@@ -1433,247 +1426,249 @@ define void @load_i32_stride8_vf16(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps 160(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vmovaps 128(%rdi), %xmm1
-; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm2 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; AVX2-ONLY-NEXT: vmovaps %xmm1, %xmm4
+; AVX2-ONLY-NEXT: vmovaps 160(%rdi), %xmm1
+; AVX2-ONLY-NEXT: vmovaps 128(%rdi), %xmm2
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm0 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; AVX2-ONLY-NEXT: vmovaps %xmm2, %xmm4
+; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps %xmm1, %xmm5
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vmovaps %xmm0, %xmm5
-; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX2-ONLY-NEXT: vmovaps 224(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vbroadcastss %xmm0, %xmm3
-; AVX2-ONLY-NEXT: vmovaps %xmm0, %xmm10
-; AVX2-ONLY-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
+; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-ONLY-NEXT: vmovaps 224(%rdi), %xmm2
+; AVX2-ONLY-NEXT: vbroadcastss %xmm2, %xmm1
+; AVX2-ONLY-NEXT: vmovaps %xmm2, %xmm10
+; AVX2-ONLY-NEXT: vmovaps %xmm2, (%rsp) # 16-byte Spill
; AVX2-ONLY-NEXT: vmovaps 192(%rdi), %xmm11
; AVX2-ONLY-NEXT: vbroadcastss %xmm11, %xmm12
-; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm3 = xmm12[0],xmm3[0],xmm12[1],xmm3[1]
-; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm1 = xmm12[0],xmm1[0],xmm12[1],xmm1[1]
+; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm0[0,1,2,3,4,5],ymm1[6,7]
; AVX2-ONLY-NEXT: vmovaps 96(%rdi), %xmm12
-; AVX2-ONLY-NEXT: vbroadcastss %xmm12, %xmm2
-; AVX2-ONLY-NEXT: vmovaps 64(%rdi), %xmm14
-; AVX2-ONLY-NEXT: vbroadcastss %xmm14, %xmm3
-; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm0 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
+; AVX2-ONLY-NEXT: vbroadcastss %xmm12, %xmm1
+; AVX2-ONLY-NEXT: vmovaps 64(%rdi), %xmm13
+; AVX2-ONLY-NEXT: vbroadcastss %xmm13, %xmm2
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm0 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX2-ONLY-NEXT: vmovaps (%rdi), %xmm8
; AVX2-ONLY-NEXT: vmovaps 32(%rdi), %xmm7
; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm6 = xmm8[0],xmm7[0],xmm8[1],xmm7[1]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm6[0,1],xmm0[2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm8[1,1,1,1]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0],xmm7[1],xmm0[2,3]
-; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm1 = xmm14[0],xmm12[0],xmm14[1],xmm12[1]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
-; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm1 = xmm11[0],xmm10[0],xmm11[1],xmm10[1]
-; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm3 = xmm13[0],xmm12[0],xmm13[1],xmm12[1]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm3[2,3]
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm3 = xmm11[0],xmm10[0],xmm11[1],xmm10[1]
+; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
; AVX2-ONLY-NEXT: vshufps {{.*#+}} xmm6 = xmm4[1,1,1,1]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0],xmm5[1],xmm6[2,3]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm6[0,1,2,3,4,5],ymm1[6,7]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm6[0,1,2,3,4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm15[1,1,1,1]
; AVX2-ONLY-NEXT: vmovaps %xmm15, %xmm10
; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0],xmm5[1],xmm0[2,3]
; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm1 = xmm3[0],xmm15[0],xmm3[1],xmm15[1]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
-; AVX2-ONLY-NEXT: vmovaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm9[0],xmm2[1],xmm9[1]
-; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX2-ONLY-NEXT: vshufps {{.*#+}} xmm6 = xmm13[1,1,1,1]
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm3 = xmm2[0],xmm15[0],xmm2[1],xmm15[1]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm3[2,3]
+; AVX2-ONLY-NEXT: vmovaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm3 = xmm1[0],xmm9[0],xmm1[1],xmm9[1]
+; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} xmm6 = xmm14[1,1,1,1]
; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0],xmm4[1],xmm6[2,3]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm6[0,1,2,3,4,5],ymm1[6,7]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm6[0,1,2,3,4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm0 = xmm10[2],xmm5[2],xmm10[3],xmm5[3]
-; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm1 = xmm13[2],xmm4[2],xmm13[3],xmm4[3]
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm3 = xmm14[2],xmm4[2],xmm14[3],xmm4[3]
; AVX2-ONLY-NEXT: vshufps {{.*#+}} xmm6 = xmm9[2,2,2,2]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm2[0,1,2],xmm6[3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm1[0,1,2],xmm6[3]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6
-; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm13
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm13[0,1,2,3,4,5],ymm6[6,7]
-; AVX2-ONLY-NEXT: vshufps {{.*#+}} xmm13 = xmm15[2,2,2,2]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm13 = xmm3[0,1,2],xmm13[3]
-; AVX2-ONLY-NEXT: vmovaps %xmm3, %xmm9
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm13 = xmm0[0,1],xmm13[2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm13[0,1,2,3],ymm6[4,5,6,7]
+; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm14
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm14[0,1,2,3,4,5],ymm6[6,7]
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} xmm14 = xmm15[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm14 = xmm2[0,1,2],xmm14[3]
+; AVX2-ONLY-NEXT: vmovaps %xmm2, %xmm9
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm14 = xmm0[0,1],xmm14[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm14[0,1,2,3],ymm6[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm4 # 16-byte Folded Reload
-; AVX2-ONLY-NEXT: # xmm4 = xmm2[2],mem[2],xmm2[3],mem[3]
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm4 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT: # xmm4 = xmm1[2],mem[2],xmm1[3],mem[3]
; AVX2-ONLY-NEXT: vmovaps (%rsp), %xmm10 # 16-byte Reload
; AVX2-ONLY-NEXT: vshufps {{.*#+}} xmm5 = xmm10[2,2,2,2]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm11[0,1,2],xmm5[3]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm5
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm0, %ymm6
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4,5],ymm5[6,7]
-; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm2 = xmm8[2],xmm7[2],xmm8[3],xmm7[3]
-; AVX2-ONLY-NEXT: vshufps {{.*#+}} xmm3 = xmm12[2,2,2,2]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm14[0,1,2],xmm3[3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm2[0,1],xmm3[2,3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm5[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm3 = xmm9[2],xmm15[2],xmm9[3],xmm15[3]
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm3[1]
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
-; AVX2-ONLY-NEXT: # xmm3 = xmm3[2],mem[2],xmm3[3],mem[3]
-; AVX2-ONLY-NEXT: vmovdqa 96(%rdi), %ymm5
-; AVX2-ONLY-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm1 = xmm8[2],xmm7[2],xmm8[3],xmm7[3]
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} xmm2 = xmm12[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm13[0,1,2],xmm2[3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm1[0,1],xmm2[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm5[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm2 = xmm9[2],xmm15[2],xmm9[3],xmm15[3]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm2[1]
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT: # xmm2 = xmm2[2],mem[2],xmm2[3],mem[3]
+; AVX2-ONLY-NEXT: vmovaps 96(%rdi), %ymm5
+; AVX2-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} xmm3 = xmm3[2,3,2,3]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm3
-; AVX2-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[2,3,2,3]
-; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm3[6,7]
-; AVX2-ONLY-NEXT: vmovdqa 64(%rdi), %ymm6
-; AVX2-ONLY-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vmovaps 64(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 224(%rdi), %ymm3
-; AVX2-ONLY-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 32(%rdi), %ymm6
+; AVX2-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm0 = xmm11[2],xmm10[2],xmm11[3],xmm10[3]
-; AVX2-ONLY-NEXT: vmovdqa 192(%rdi), %ymm13
+; AVX2-ONLY-NEXT: vmovaps (%rdi), %ymm11
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm4[2,3,2,3]
-; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT: vmovdqa 160(%rdi), %ymm7
-; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm1 = xmm14[2],xmm12[2],xmm14[3],xmm12[3]
-; AVX2-ONLY-NEXT: vmovdqa 128(%rdi), %ymm12
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm2[1],xmm1[1]
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} xmm2 = xmm4[2,3,2,3]
+; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vmovaps 224(%rdi), %ymm4
+; AVX2-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm2 = xmm13[2],xmm12[2],xmm13[3],xmm12[3]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 48(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vpbroadcastd 16(%rdi), %ymm1
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm1 = ymm6[0],ymm5[0],ymm6[1],ymm5[1],ymm6[4],ymm5[4],ymm6[5],ymm5[5]
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm1 = ymm12[0],ymm7[0],ymm12[1],ymm7[1],ymm12[4],ymm7[4],ymm12[5],ymm7[5]
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm8 = ymm13[0],ymm3[0],ymm13[1],ymm3[1],ymm13[4],ymm3[4],ymm13[5],ymm3[5]
-; AVX2-ONLY-NEXT: vpunpcklqdq {{.*#+}} ymm1 = ymm1[0],ymm8[0],ymm1[2],ymm8[2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 304(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vpbroadcastd 272(%rdi), %ymm1
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; AVX2-ONLY-NEXT: vmovdqa 352(%rdi), %ymm9
-; AVX2-ONLY-NEXT: vmovdqa 320(%rdi), %ymm6
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm1 = ymm6[0],ymm9[0],ymm6[1],ymm9[1],ymm6[4],ymm9[4],ymm6[5],ymm9[5]
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm10 = xmm0[0,1],xmm1[2,3]
-; AVX2-ONLY-NEXT: vmovdqa 480(%rdi), %ymm15
-; AVX2-ONLY-NEXT: vmovdqa 448(%rdi), %ymm5
-; AVX2-ONLY-NEXT: vmovdqa 416(%rdi), %ymm2
-; AVX2-ONLY-NEXT: vmovdqa 384(%rdi), %ymm4
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm11 = ymm4[0],ymm2[0],ymm4[1],ymm2[1],ymm4[4],ymm2[4],ymm4[5],ymm2[5]
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm0 = ymm5[0],ymm15[0],ymm5[1],ymm15[1],ymm5[4],ymm15[4],ymm5[5],ymm15[5]
-; AVX2-ONLY-NEXT: vpunpcklqdq {{.*#+}} ymm11 = ymm11[0],ymm0[0],ymm11[2],ymm0[2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm10[0,1,2,3],ymm11[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm1, (%rsp) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 148(%rdi), %ymm11
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm11 = ymm11[0,1,2,3,4],ymm7[5],ymm11[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm8 = ymm11[0,1,2,3,4,5],ymm8[6,7]
-; AVX2-ONLY-NEXT: vpmovzxdq {{.*#+}} xmm11 = mem[0],zero,mem[1],zero
-; AVX2-ONLY-NEXT: vpbroadcastd 116(%rdi), %ymm14
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm11 = xmm11[0,1,2],xmm14[3]
-; AVX2-ONLY-NEXT: vmovdqa 32(%rdi), %ymm14
-; AVX2-ONLY-NEXT: vmovdqa (%rdi), %ymm1
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm10 = ymm1[1,1,1,1,5,5,5,5]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm10 = ymm10[0],ymm14[1],ymm10[2,3,4],ymm14[5],ymm10[6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm10, %xmm10
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm10 = xmm10[0,1],xmm11[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm10[0,1,2,3],ymm8[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 404(%rdi), %ymm8
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm8 = ymm8[0,1,2,3,4],ymm2[5],ymm8[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm8[0,1,2,3,4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT: vpmovzxdq {{.*#+}} xmm8 = mem[0],zero,mem[1],zero
-; AVX2-ONLY-NEXT: vpbroadcastd 372(%rdi), %ymm10
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm8 = xmm8[0,1,2],xmm10[3]
-; AVX2-ONLY-NEXT: vmovdqa 288(%rdi), %ymm10
-; AVX2-ONLY-NEXT: vmovdqa 256(%rdi), %ymm0
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm11 = ymm0[1,1,1,1,5,5,5,5]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm11 = ymm11[0],ymm10[1],ymm11[2,3,4],ymm10[5],ymm11[6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm11, %xmm11
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm8 = xmm11[0,1],xmm8[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm8 = ymm8[0,1,2,3],ymm3[4,5,6,7]
-; AVX2-ONLY-NEXT: vpbroadcastd 248(%rdi), %ymm3
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm13[0,1,2,3,4,5,6],ymm3[7]
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm7 = ymm12[2],ymm7[2],ymm12[3],ymm7[3],ymm12[6],ymm7[6],ymm12[7],ymm7[7]
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm11 = ymm1[2],ymm14[2],ymm1[3],ymm14[3],ymm1[6],ymm14[6],ymm1[7],ymm14[7]
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm12 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm12 = ymm1[2],mem[2],ymm1[3],mem[3],ymm1[6],mem[6],ymm1[7],mem[7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm11, %xmm1
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm13 = ymm12[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm13[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm7[0,1,2,3,4,5],ymm3[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
-; AVX2-ONLY-NEXT: vpbroadcastd 504(%rdi), %ymm3
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm5[0,1,2,3,4,5,6],ymm3[7]
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm2 = ymm4[2],ymm2[2],ymm4[3],ymm2[3],ymm4[6],ymm2[6],ymm4[7],ymm2[7]
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm0 = ymm0[2],ymm10[2],ymm0[3],ymm10[3],ymm0[6],ymm10[6],ymm0[7],ymm10[7]
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm4 = ymm6[2],ymm9[2],ymm6[3],ymm9[3],ymm6[6],ymm9[6],ymm6[7],ymm9[7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm0, %xmm5
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm6 = ymm4[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm5 = xmm5[0,1],xmm6[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm2[0,1,2,3,4,5],ymm3[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm5[0,1,2,3],ymm3[4,5,6,7]
-; AVX2-ONLY-NEXT: vpbroadcastd 220(%rdi), %ymm5
-; AVX2-ONLY-NEXT: vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm5 = ymm5[2],mem[2],ymm5[3],mem[3],ymm5[6],mem[6],ymm5[7],mem[7]
-; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm5 = ymm7[1],ymm5[1],ymm7[3],ymm5[3]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm12, %xmm6
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm7 = ymm11[2,3,2,3,6,7,6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm7, %xmm7
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm6 = xmm7[0,1],xmm6[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm5 = ymm6[0,1,2,3],ymm5[4,5,6,7]
-; AVX2-ONLY-NEXT: vpbroadcastd 476(%rdi), %ymm6
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm6 = ymm6[2],ymm15[2],ymm6[3],ymm15[3],ymm6[6],ymm15[6],ymm6[7],ymm15[7]
-; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm2 = ymm2[1],ymm6[1],ymm2[3],ymm6[3]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm4, %xmm4
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[2,3,2,3,6,7,6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm0, %xmm0
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm4[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm2, 32(%rsi)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm2, (%rsi)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm2, 32(%rdx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm2, (%rdx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm2, 32(%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm2, (%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm2, 32(%r8)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm2, (%r8)
-; AVX2-ONLY-NEXT: vmovups (%rsp), %ymm2 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm2, 32(%r9)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm2, (%r9)
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm0 = ymm11[0],ymm6[0],ymm11[1],ymm6[1],ymm11[4],ymm6[4],ymm11[5],ymm6[5]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm0, %xmm0
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm5 = ymm3[0],ymm5[0],ymm3[1],ymm5[1],ymm3[4],ymm5[4],ymm3[5],ymm5[5]
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm1 = ymm5[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT: vmovaps 192(%rdi), %ymm12
+; AVX2-ONLY-NEXT: vmovaps 160(%rdi), %ymm6
+; AVX2-ONLY-NEXT: vmovaps 128(%rdi), %ymm15
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm2 = ymm15[0],ymm6[0],ymm15[1],ymm6[1],ymm15[4],ymm6[4],ymm15[5],ymm6[5]
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm3 = ymm12[0],ymm4[0],ymm12[1],ymm4[1],ymm12[4],ymm4[4],ymm12[5],ymm4[5]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm2 = ymm2[0],ymm3[0],ymm2[2],ymm3[2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 288(%rdi), %ymm10
+; AVX2-ONLY-NEXT: vmovaps 256(%rdi), %ymm9
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm2 = ymm9[0],ymm10[0],ymm9[1],ymm10[1],ymm9[4],ymm10[4],ymm9[5],ymm10[5]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm2
+; AVX2-ONLY-NEXT: vmovaps 352(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 320(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm13 = ymm1[0],ymm0[0],ymm1[1],ymm0[1],ymm1[4],ymm0[4],ymm1[5],ymm0[5]
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm8 = ymm13[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0,1],xmm8[2,3]
+; AVX2-ONLY-NEXT: vmovaps 480(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 448(%rdi), %ymm8
+; AVX2-ONLY-NEXT: vmovaps 416(%rdi), %ymm7
+; AVX2-ONLY-NEXT: vmovaps 384(%rdi), %ymm14
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm1 = ymm14[0],ymm7[0],ymm14[1],ymm7[1],ymm14[4],ymm7[4],ymm14[5],ymm7[5]
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm4 = ymm8[0],ymm0[0],ymm8[1],ymm0[1],ymm8[4],ymm0[4],ymm8[5],ymm0[5]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm4[0],ymm1[2],ymm4[2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm1, (%rsp) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 148(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm6[5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm5, %xmm2
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm11[1,1,1,1,5,5,5,5]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0],ymm5[1],ymm3[2,3,4],ymm5[5],ymm3[6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm3, %xmm3
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 404(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm7[5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm4[6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm13, %xmm1
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm9[1,1,1,1,5,5,5,5]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm10[1],ymm2[2,3,4],ymm10[5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm2
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vbroadcastss 248(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm12[0,1,2,3,4,5,6],ymm0[7]
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm2 = ymm15[2],ymm6[2],ymm15[3],ymm6[3],ymm15[6],ymm6[6],ymm15[7],ymm6[7]
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm3 = ymm11[2],ymm5[2],ymm11[3],ymm5[3],ymm11[6],ymm5[6],ymm11[7],ymm5[7]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm4 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm4 = ymm1[2],mem[2],ymm1[3],mem[3],ymm1[6],mem[6],ymm1[7],mem[7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm3, %xmm1
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm6 = ymm4[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm6[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vbroadcastss 504(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3,4,5,6],ymm0[7]
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm5 = ymm14[2],ymm7[2],ymm14[3],ymm7[3],ymm14[6],ymm7[6],ymm14[7],ymm7[7]
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm6 = ymm9[2],ymm10[2],ymm9[3],ymm10[3],ymm9[6],ymm10[6],ymm9[7],ymm10[7]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm7 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm7 = ymm7[2],mem[2],ymm7[3],mem[3],ymm7[6],mem[6],ymm7[7],mem[7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm6, %xmm8
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm9 = ymm7[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm8 = xmm8[0,1],xmm9[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vbroadcastss 220(%rdi), %ymm8
+; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm8 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm8 = ymm8[2],mem[2],ymm8[3],mem[3],ymm8[6],mem[6],ymm8[7],mem[7]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm2 = ymm2[1],ymm8[1],ymm2[3],ymm8[3]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm4, %xmm4
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm3[2,3,2,3,6,7,6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm3, %xmm3
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm4[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT: vbroadcastss 476(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm3 = ymm3[2],mem[2],ymm3[3],mem[3],ymm3[6],mem[6],ymm3[7],mem[7]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm3 = ymm5[1],ymm3[1],ymm5[3],ymm3[3]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm7, %xmm4
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm6[2,3,2,3,6,7,6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm5, %xmm5
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm5[0,1],xmm4[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm4, 32(%rsi)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm4, (%rsi)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm4, 32(%rdx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm4, (%rdx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm4, 32(%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm4, (%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm4, 32(%r8)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm4, (%r8)
+; AVX2-ONLY-NEXT: vmovups (%rsp), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm4, 32(%r9)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm4, (%r9)
; AVX2-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX2-ONLY-NEXT: vmovdqa %ymm8, 32(%rax)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm2, (%rax)
+; AVX2-ONLY-NEXT: vmovaps %ymm13, 32(%rax)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm4, (%rax)
; AVX2-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX2-ONLY-NEXT: vmovdqa %ymm3, 32(%rax)
-; AVX2-ONLY-NEXT: vmovdqa %ymm1, (%rax)
+; AVX2-ONLY-NEXT: vmovaps %ymm0, 32(%rax)
+; AVX2-ONLY-NEXT: vmovaps %ymm1, (%rax)
; AVX2-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX2-ONLY-NEXT: vmovdqa %ymm0, 32(%rax)
-; AVX2-ONLY-NEXT: vmovdqa %ymm5, (%rax)
-; AVX2-ONLY-NEXT: addq $328, %rsp # imm = 0x148
+; AVX2-ONLY-NEXT: vmovaps %ymm3, 32(%rax)
+; AVX2-ONLY-NEXT: vmovaps %ymm2, (%rax)
+; AVX2-ONLY-NEXT: addq $456, %rsp # imm = 0x1C8
; AVX2-ONLY-NEXT: vzeroupper
; AVX2-ONLY-NEXT: retq
;
@@ -3101,7 +3096,7 @@ define void @load_i32_stride8_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
;
; AVX2-ONLY-LABEL: load_i32_stride8_vf32:
; AVX2-ONLY: # %bb.0:
-; AVX2-ONLY-NEXT: subq $1256, %rsp # imm = 0x4E8
+; AVX2-ONLY-NEXT: subq $1576, %rsp # imm = 0x628
; AVX2-ONLY-NEXT: vmovaps 288(%rdi), %xmm0
; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vmovaps 256(%rdi), %xmm2
@@ -3119,7 +3114,7 @@ define void @load_i32_stride8_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
; AVX2-ONLY-NEXT: vmovaps 416(%rdi), %xmm2
-; AVX2-ONLY-NEXT: vmovaps %xmm2, (%rsp) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vmovaps 384(%rdi), %xmm1
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
@@ -3200,7 +3195,7 @@ define void @load_i32_stride8_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vbroadcastss %xmm0, %xmm0
; AVX2-ONLY-NEXT: vmovaps 192(%rdi), %xmm8
; AVX2-ONLY-NEXT: vbroadcastss %xmm8, %xmm1
-; AVX2-ONLY-NEXT: vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps %xmm8, (%rsp) # 16-byte Spill
; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX2-ONLY-NEXT: vmovaps 160(%rdi), %xmm9
; AVX2-ONLY-NEXT: vmovaps 128(%rdi), %xmm7
@@ -3236,7 +3231,7 @@ define void @load_i32_stride8_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX2-ONLY-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; AVX2-ONLY-NEXT: # xmm2 = mem[1,1,1,1]
-; AVX2-ONLY-NEXT: vmovaps (%rsp), %xmm11 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0],xmm11[1],xmm2[2,3]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
@@ -3319,7 +3314,7 @@ define void @load_i32_stride8_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11, %xmm0 # 16-byte Folded Reload
; AVX2-ONLY-NEXT: # xmm0 = xmm11[2],mem[2],xmm11[3],mem[3]
-; AVX2-ONLY-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vshufps {{.*#+}} xmm14 = xmm12[2,2,2,2]
; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm14 = xmm12[0,1,2],xmm14[3]
@@ -3350,7 +3345,7 @@ define void @load_i32_stride8_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: # xmm9 = xmm4[2],mem[2],xmm4[3],mem[3]
; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
; AVX2-ONLY-NEXT: vshufps {{.*#+}} xmm6 = xmm14[2,2,2,2]
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps (%rsp), %xmm5 # 16-byte Reload
; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm5[0,1,2],xmm6[3]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm9, %ymm0, %ymm8
@@ -3384,7 +3379,7 @@ define void @load_i32_stride8_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12, %xmm1 # 16-byte Folded Reload
; AVX2-ONLY-NEXT: # xmm1 = xmm12[2],mem[2],xmm12[3],mem[3]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX2-ONLY-NEXT: vpermilps $238, (%rsp), %xmm2 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
; AVX2-ONLY-NEXT: # xmm2 = mem[2,3,2,3]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
@@ -3411,304 +3406,324 @@ define void @load_i32_stride8_vf32(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm7[1],xmm1[1]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 304(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vpbroadcastd 272(%rdi), %ymm1
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; AVX2-ONLY-NEXT: vmovdqa 352(%rdi), %ymm2
-; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 320(%rdi), %ymm1
-; AVX2-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[1],ymm2[1],ymm1[4],ymm2[4],ymm1[5],ymm2[5]
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
-; AVX2-ONLY-NEXT: vmovdqa 480(%rdi), %ymm2
-; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 448(%rdi), %ymm3
-; AVX2-ONLY-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 416(%rdi), %ymm9
-; AVX2-ONLY-NEXT: vmovdqa 384(%rdi), %ymm1
-; AVX2-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm9[0],ymm1[1],ymm9[1],ymm1[4],ymm9[4],ymm1[5],ymm9[5]
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm3 = ymm3[0],ymm2[0],ymm3[1],ymm2[1],ymm3[4],ymm2[4],ymm3[5],ymm2[5]
-; AVX2-ONLY-NEXT: vpunpcklqdq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[2],ymm3[2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 816(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vpbroadcastd 784(%rdi), %ymm1
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; AVX2-ONLY-NEXT: vmovdqa 864(%rdi), %ymm2
-; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 832(%rdi), %ymm1
-; AVX2-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[1],ymm2[1],ymm1[4],ymm2[4],ymm1[5],ymm2[5]
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
-; AVX2-ONLY-NEXT: vmovdqa 992(%rdi), %ymm1
-; AVX2-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 960(%rdi), %ymm4
-; AVX2-ONLY-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 928(%rdi), %ymm10
-; AVX2-ONLY-NEXT: vmovdqa 896(%rdi), %ymm2
-; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm2 = ymm2[0],ymm10[0],ymm2[1],ymm10[1],ymm2[4],ymm10[4],ymm2[5],ymm10[5]
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm1 = ymm4[0],ymm1[0],ymm4[1],ymm1[1],ymm4[4],ymm1[4],ymm4[5],ymm1[5]
-; AVX2-ONLY-NEXT: vpunpcklqdq {{.*#+}} ymm2 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 48(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vpbroadcastd 16(%rdi), %ymm2
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
-; AVX2-ONLY-NEXT: vmovdqa 96(%rdi), %ymm2
-; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 64(%rdi), %ymm4
-; AVX2-ONLY-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm2 = ymm4[0],ymm2[0],ymm4[1],ymm2[1],ymm4[4],ymm2[4],ymm4[5],ymm2[5]
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm2 = ymm2[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm2[2,3]
-; AVX2-ONLY-NEXT: vmovdqa 224(%rdi), %ymm2
-; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 192(%rdi), %ymm5
-; AVX2-ONLY-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 160(%rdi), %ymm12
-; AVX2-ONLY-NEXT: vmovdqa 128(%rdi), %ymm4
-; AVX2-ONLY-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm4 = ymm4[0],ymm12[0],ymm4[1],ymm12[1],ymm4[4],ymm12[4],ymm4[5],ymm12[5]
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm2 = ymm5[0],ymm2[0],ymm5[1],ymm2[1],ymm5[4],ymm2[4],ymm5[5],ymm2[5]
-; AVX2-ONLY-NEXT: vpunpcklqdq {{.*#+}} ymm4 = ymm4[0],ymm2[0],ymm4[2],ymm2[2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm4[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 560(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vpbroadcastd 528(%rdi), %ymm4
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
-; AVX2-ONLY-NEXT: vmovdqa 608(%rdi), %ymm4
-; AVX2-ONLY-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 576(%rdi), %ymm5
-; AVX2-ONLY-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm4 = ymm5[0],ymm4[0],ymm5[1],ymm4[1],ymm5[4],ymm4[4],ymm5[5],ymm4[5]
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm4 = ymm4[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm4 = xmm0[0,1],xmm4[2,3]
-; AVX2-ONLY-NEXT: vmovdqa 736(%rdi), %ymm0
-; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 704(%rdi), %ymm15
-; AVX2-ONLY-NEXT: vmovdqa 672(%rdi), %ymm8
-; AVX2-ONLY-NEXT: vmovdqa 640(%rdi), %ymm5
-; AVX2-ONLY-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm5 = ymm5[0],ymm8[0],ymm5[1],ymm8[1],ymm5[4],ymm8[4],ymm5[5],ymm8[5]
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm0 = ymm15[0],ymm0[0],ymm15[1],ymm0[1],ymm15[4],ymm0[4],ymm15[5],ymm0[5]
-; AVX2-ONLY-NEXT: vpunpcklqdq {{.*#+}} ymm5 = ymm5[0],ymm0[0],ymm5[2],ymm0[2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 404(%rdi), %ymm4
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4],ymm9[5],ymm4[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3,4,5],ymm3[6,7]
-; AVX2-ONLY-NEXT: vpmovzxdq {{.*#+}} xmm4 = mem[0],zero,mem[1],zero
-; AVX2-ONLY-NEXT: vpbroadcastd 372(%rdi), %ymm5
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm4 = xmm4[0,1,2],xmm5[3]
-; AVX2-ONLY-NEXT: vmovdqa 288(%rdi), %ymm7
-; AVX2-ONLY-NEXT: vmovdqa 256(%rdi), %ymm6
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm5 = ymm6[1,1,1,1,5,5,5,5]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm5 = ymm5[0],ymm7[1],ymm5[2,3,4],ymm7[5],ymm5[6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm5, %xmm5
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm4 = xmm5[0,1],xmm4[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 916(%rdi), %ymm3
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm10[5],ymm3[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm1[6,7]
-; AVX2-ONLY-NEXT: vpmovzxdq {{.*#+}} xmm3 = mem[0],zero,mem[1],zero
-; AVX2-ONLY-NEXT: vpbroadcastd 884(%rdi), %ymm4
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm3 = xmm3[0,1,2],xmm4[3]
-; AVX2-ONLY-NEXT: vmovdqa 800(%rdi), %ymm5
-; AVX2-ONLY-NEXT: vmovdqa 768(%rdi), %ymm4
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm11 = ymm4[1,1,1,1,5,5,5,5]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm11 = ymm11[0],ymm5[1],ymm11[2,3,4],ymm5[5],ymm11[6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm11, %xmm11
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm3 = xmm11[0,1],xmm3[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm1, (%rsp) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 148(%rdi), %ymm1
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm12[5],ymm1[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
-; AVX2-ONLY-NEXT: vpmovzxdq {{.*#+}} xmm2 = mem[0],zero,mem[1],zero
-; AVX2-ONLY-NEXT: vpbroadcastd 116(%rdi), %ymm3
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm11 = xmm2[0,1,2],xmm3[3]
-; AVX2-ONLY-NEXT: vmovdqa 32(%rdi), %ymm3
-; AVX2-ONLY-NEXT: vmovdqa (%rdi), %ymm2
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm13 = ymm2[1,1,1,1,5,5,5,5]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm13[0],ymm3[1],ymm13[2,3,4],ymm3[5],ymm13[6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm13, %xmm13
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm11 = xmm13[0,1],xmm11[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm11[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 660(%rdi), %ymm1
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm8[5],ymm1[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm11 = ymm1[0,1,2,3,4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT: vpmovzxdq {{.*#+}} xmm0 = mem[0],zero,mem[1],zero
-; AVX2-ONLY-NEXT: vpbroadcastd 628(%rdi), %ymm1
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm13 = xmm0[0,1,2],xmm1[3]
-; AVX2-ONLY-NEXT: vmovdqa 544(%rdi), %ymm1
-; AVX2-ONLY-NEXT: vmovdqa 512(%rdi), %ymm0
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm14 = ymm0[1,1,1,1,5,5,5,5]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm14 = ymm14[0],ymm1[1],ymm14[2,3,4],ymm1[5],ymm14[6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm14, %xmm14
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm13 = xmm14[0,1],xmm13[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm11 = ymm13[0,1,2,3],ymm11[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 504(%rdi), %ymm11
-; AVX2-ONLY-NEXT: vpblendd $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm14 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm14 = mem[0,1,2,3,4,5,6],ymm11[7]
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm11 = ymm11[2],ymm9[2],ymm11[3],ymm9[3],ymm11[6],ymm9[6],ymm11[7],ymm9[7]
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm6 = ymm6[2],ymm7[2],ymm6[3],ymm7[3],ymm6[6],ymm7[6],ymm6[7],ymm7[7]
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm13 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm13 = ymm7[2],mem[2],ymm7[3],mem[3],ymm7[6],mem[6],ymm7[7],mem[7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm6, %xmm7
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm9 = ymm13[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm7 = xmm7[0,1],xmm9[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm9 = ymm11[0,1,2,3,4,5],ymm14[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm9[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 1016(%rdi), %ymm7
-; AVX2-ONLY-NEXT: vpblendd $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm14 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm14 = mem[0,1,2,3,4,5,6],ymm7[7]
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm7 = ymm7[2],ymm10[2],ymm7[3],ymm10[3],ymm7[6],ymm10[6],ymm7[7],ymm10[7]
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm9 = ymm4[2],ymm5[2],ymm4[3],ymm5[3],ymm4[6],ymm5[6],ymm4[7],ymm5[7]
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm10 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm10 = ymm4[2],mem[2],ymm4[3],mem[3],ymm4[6],mem[6],ymm4[7],mem[7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm9, %xmm4
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm5 = ymm10[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm4 = xmm4[0,1],xmm5[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm5 = ymm7[0,1,2,3,4,5],ymm14[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 248(%rdi), %ymm4
-; AVX2-ONLY-NEXT: vpblendd $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm4 = mem[0,1,2,3,4,5,6],ymm4[7]
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm5 = ymm5[2],ymm12[2],ymm5[3],ymm12[3],ymm5[6],ymm12[6],ymm5[7],ymm12[7]
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm12 = ymm2[2],ymm3[2],ymm2[3],ymm3[3],ymm2[6],ymm3[6],ymm2[7],ymm3[7]
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm14 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm14 = ymm2[2],mem[2],ymm2[3],mem[3],ymm2[6],mem[6],ymm2[7],mem[7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm12, %xmm2
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm3 = ymm14[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm2 = xmm2[0,1],xmm3[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm5[0,1,2,3,4,5],ymm4[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm2[0,1,2,3],ymm3[4,5,6,7]
-; AVX2-ONLY-NEXT: vpbroadcastd 760(%rdi), %ymm2
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm15[0,1,2,3,4,5,6],ymm2[7]
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm8 = ymm4[2],ymm8[2],ymm4[3],ymm8[3],ymm4[6],ymm8[6],ymm4[7],ymm8[7]
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm1 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm15 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm15 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm1, %xmm0
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm4 = ymm15[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm4[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm8[0,1,2,3,4,5],ymm2[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm0[0,1,2,3],ymm2[4,5,6,7]
-; AVX2-ONLY-NEXT: vpbroadcastd 220(%rdi), %ymm0
-; AVX2-ONLY-NEXT: vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7]
-; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm0 = ymm5[1],ymm0[1],ymm5[3],ymm0[3]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm14, %xmm4
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm5 = ymm12[2,3,2,3,6,7,6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm5, %xmm5
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm4 = xmm5[0,1],xmm4[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT: vpbroadcastd 476(%rdi), %ymm0
-; AVX2-ONLY-NEXT: vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vmovaps 288(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovaps 256(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-ONLY-NEXT: vmovaps %ymm1, %ymm9
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm0, %xmm0
+; AVX2-ONLY-NEXT: vmovaps 352(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovups %ymm1, (%rsp) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 320(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm8 = ymm2[0],ymm1[0],ymm2[1],ymm1[1],ymm2[4],ymm1[4],ymm2[5],ymm1[5]
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm1 = ymm8[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT: vmovaps 480(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 448(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 416(%rdi), %ymm4
+; AVX2-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 384(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm1 = ymm1[0],ymm4[0],ymm1[1],ymm4[1],ymm1[4],ymm4[4],ymm1[5],ymm4[5]
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm5 = ymm3[0],ymm2[0],ymm3[1],ymm2[1],ymm3[4],ymm2[4],ymm3[5],ymm2[5]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm5[0],ymm1[2],ymm5[2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 800(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 768(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[1],ymm0[1],ymm1[4],ymm0[4],ymm1[5],ymm0[5]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm0, %xmm0
+; AVX2-ONLY-NEXT: vmovaps 864(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 832(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm14 = ymm2[0],ymm1[0],ymm2[1],ymm1[1],ymm2[4],ymm1[4],ymm2[5],ymm1[5]
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm1 = ymm14[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT: vmovaps 992(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 960(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 928(%rdi), %ymm4
+; AVX2-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 896(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm3 = ymm3[0],ymm4[0],ymm3[1],ymm4[1],ymm3[4],ymm4[4],ymm3[5],ymm4[5]
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm4 = ymm2[0],ymm1[0],ymm2[1],ymm1[1],ymm2[4],ymm1[4],ymm2[5],ymm1[5]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm3 = ymm3[0],ymm4[0],ymm3[2],ymm4[2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 32(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps (%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[1],ymm0[1],ymm1[4],ymm0[4],ymm1[5],ymm0[5]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm0, %xmm0
+; AVX2-ONLY-NEXT: vmovaps 96(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 64(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[1],ymm1[1],ymm2[4],ymm1[4],ymm2[5],ymm1[5]
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm3 = ymm1[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm0[0,1],xmm3[2,3]
+; AVX2-ONLY-NEXT: vmovaps 224(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 192(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 160(%rdi), %ymm6
+; AVX2-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 128(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm10 = ymm2[0],ymm6[0],ymm2[1],ymm6[1],ymm2[4],ymm6[4],ymm2[5],ymm6[5]
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm2 = ymm1[0],ymm0[0],ymm1[1],ymm0[1],ymm1[4],ymm0[4],ymm1[5],ymm0[5]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm10 = ymm10[0],ymm2[0],ymm10[2],ymm2[2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm10[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 544(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 512(%rdi), %ymm13
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm3 = ymm13[0],ymm0[0],ymm13[1],ymm0[1],ymm13[4],ymm0[4],ymm13[5],ymm0[5]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm3, %xmm10
+; AVX2-ONLY-NEXT: vmovaps 608(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 576(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm15 = ymm1[0],ymm0[0],ymm1[1],ymm0[1],ymm1[4],ymm0[4],ymm1[5],ymm0[5]
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm11 = ymm15[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm11 = xmm10[0,1],xmm11[2,3]
+; AVX2-ONLY-NEXT: vmovaps 736(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 704(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 672(%rdi), %ymm7
+; AVX2-ONLY-NEXT: vmovaps 640(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm1 = ymm1[0],ymm7[0],ymm1[1],ymm7[1],ymm1[4],ymm7[4],ymm1[5],ymm7[5]
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm12 = ymm3[0],ymm0[0],ymm3[1],ymm0[1],ymm3[4],ymm0[4],ymm3[5],ymm0[5]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm12[0],ymm1[2],ymm12[2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm11[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 404(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm10[5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm5[6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm8, %xmm5
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm11[1,1,1,1,5,5,5,5]
+; AVX2-ONLY-NEXT: vmovaps %ymm9, %ymm3
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0],ymm9[1],ymm8[2,3,4],ymm9[5],ymm8[6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm8, %xmm8
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm8[0,1],xmm5[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 916(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm6[5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm4[6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm14, %xmm4
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm9[1,1,1,1,5,5,5,5]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0],ymm14[1],ymm5[2,3,4],ymm14[5],ymm5[6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm5, %xmm5
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm5[0,1],xmm4[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 148(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm8[5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm2
+; AVX2-ONLY-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm4 = mem[1,1,1,1,5,5,5,5]
+; AVX2-ONLY-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm4 = ymm4[0],mem[1],ymm4[2,3,4],mem[5],ymm4[6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm4, %xmm4
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm4[0,1],xmm2[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 660(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm7[5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm12[6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm15, %xmm1
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm13[1,1,1,1,5,5,5,5]
+; AVX2-ONLY-NEXT: vmovaps %ymm13, %ymm15
+; AVX2-ONLY-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm2 = ymm2[0],mem[1],ymm2[2,3,4],mem[5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm2
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 504(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vblendps $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm1 = mem[0,1,2,3,4,5,6],ymm0[7]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm4 = ymm0[2],ymm10[2],ymm0[3],ymm10[3],ymm0[6],ymm10[6],ymm0[7],ymm10[7]
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm11 = ymm11[2],ymm3[2],ymm11[3],ymm3[3],ymm11[6],ymm3[6],ymm11[7],ymm3[7]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhps (%rsp), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: # ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7]
-; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm0 = ymm11[1],ymm0[1],ymm11[3],ymm0[3]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm13, %xmm5
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm6 = ymm6[2,3,2,3,6,7,6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm6, %xmm6
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm5 = xmm6[0,1],xmm5[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT: vpbroadcastd 732(%rdi), %ymm5
-; AVX2-ONLY-NEXT: vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm11, %xmm2
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm3 = ymm0[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0,1],xmm3[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm4[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 1016(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vblendps $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm2 = mem[0,1,2,3,4,5,6],ymm1[7]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm1 = ymm1[2],ymm6[2],ymm1[3],ymm6[3],ymm1[6],ymm6[6],ymm1[7],ymm6[7]
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm5 = ymm9[2],ymm14[2],ymm9[3],ymm14[3],ymm9[6],ymm14[6],ymm9[7],ymm14[7]
+; AVX2-ONLY-NEXT: vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm6 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm6 = ymm3[2],mem[2],ymm3[3],mem[3],ymm3[6],mem[6],ymm3[7],mem[7]
+; AVX2-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm5, %xmm3
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm5 = ymm6[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm5[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm1[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm2, (%rsp) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 248(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vblendps $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm2 = mem[0,1,2,3,4,5,6],ymm2[7]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm3 = ymm3[2],ymm8[2],ymm3[3],ymm8[3],ymm3[6],ymm8[6],ymm3[7],ymm8[7]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm9 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm9 = ymm5[2],mem[2],ymm5[3],mem[3],ymm5[6],mem[6],ymm5[7],mem[7]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm12 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm12 = ymm5[2],mem[2],ymm5[3],mem[3],ymm5[6],mem[6],ymm5[7],mem[7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm9, %xmm6
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm13 = ymm12[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0,1],xmm13[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT: vbroadcastss 760(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vblendps $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm2 = mem[0,1,2,3,4,5,6],ymm2[7]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm10 = ymm5[2],ymm7[2],ymm5[3],ymm7[3],ymm5[6],ymm7[6],ymm5[7],ymm7[7]
+; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm13 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm13 = ymm15[2],mem[2],ymm15[3],mem[3],ymm15[6],mem[6],ymm15[7],mem[7]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm14 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm14 = ymm5[2],mem[2],ymm5[3],mem[3],ymm5[6],mem[6],ymm5[7],mem[7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm13, %xmm15
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm5 = ymm14[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm15[0,1],xmm5[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm10[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm5[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT: vbroadcastss 220(%rdi), %ymm5
+; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: # ymm5 = ymm5[2],mem[2],ymm5[3],mem[3],ymm5[6],mem[6],ymm5[7],mem[7]
-; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm5 = ymm8[1],ymm5[1],ymm8[3],ymm5[3]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm15, %xmm6
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[2,3,2,3,6,7,6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm1, %xmm1
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm6[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm5[4,5,6,7]
-; AVX2-ONLY-NEXT: vpbroadcastd 988(%rdi), %ymm5
-; AVX2-ONLY-NEXT: vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm3 = ymm3[1],ymm5[1],ymm3[3],ymm5[3]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm12, %xmm5
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} ymm9 = ymm9[2,3,2,3,6,7,6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm9, %xmm9
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm9[0,1],xmm5[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm5[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-ONLY-NEXT: vbroadcastss 476(%rdi), %ymm5
+; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: # ymm5 = ymm5[2],mem[2],ymm5[3],mem[3],ymm5[6],mem[6],ymm5[7],mem[7]
-; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm5 = ymm7[1],ymm5[1],ymm7[3],ymm5[3]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm10, %xmm6
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm7 = ymm9[2,3,2,3,6,7,6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm7, %xmm7
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm6 = xmm7[0,1],xmm6[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm5 = ymm6[0,1,2,3],ymm5[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm6, 64(%rsi)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm6, (%rsi)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm6, 96(%rsi)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm6, 32(%rsi)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm6, 64(%rdx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm6, (%rdx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm6, 96(%rdx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm6, 32(%rdx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm6, 64(%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm6, (%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm6, 96(%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm6, 32(%rcx)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm6, 64(%r8)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm6, (%r8)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm6, 96(%r8)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm6, 32(%r8)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm6, 64(%r9)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm6, (%r9)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm6, 96(%r9)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm6, 32(%r9)
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm4 = ymm4[1],ymm5[1],ymm4[3],ymm5[3]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm0, %xmm0
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} ymm5 = ymm11[2,3,2,3,6,7,6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm5, %xmm5
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm5[0,1],xmm0[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm4[4,5,6,7]
+; AVX2-ONLY-NEXT: vbroadcastss 732(%rdi), %ymm4
+; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm4 = ymm4[2],mem[2],ymm4[3],mem[3],ymm4[6],mem[6],ymm4[7],mem[7]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm4 = ymm10[1],ymm4[1],ymm10[3],ymm4[3]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm14, %xmm5
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} ymm9 = ymm13[2,3,2,3,6,7,6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm9, %xmm9
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm9[0,1],xmm5[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3],ymm4[4,5,6,7]
+; AVX2-ONLY-NEXT: vbroadcastss 988(%rdi), %ymm5
+; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm5 = ymm5[2],mem[2],ymm5[3],mem[3],ymm5[6],mem[6],ymm5[7],mem[7]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm5[1],ymm1[3],ymm5[3]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm5, %xmm5
+; AVX2-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm7 = mem[2,3,2,3,6,7,6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm7, %xmm7
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm7[0,1],xmm5[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm5[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 64(%rsi)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, (%rsi)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 96(%rsi)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 32(%rsi)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 64(%rdx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, (%rdx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 96(%rdx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 32(%rdx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 64(%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, (%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 96(%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 32(%rcx)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 64(%r8)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, (%r8)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 96(%r8)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 32(%r8)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 64(%r9)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, (%r9)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 96(%r9)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 32(%r9)
; AVX2-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm6, 64(%rax)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm6, (%rax)
-; AVX2-ONLY-NEXT: vmovups (%rsp), %ymm6 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm6, 96(%rax)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX2-ONLY-NEXT: vmovaps %ymm6, 32(%rax)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 64(%rax)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, (%rax)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 96(%rax)
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm5, 32(%rax)
; AVX2-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX2-ONLY-NEXT: vmovdqa %ymm2, 64(%rax)
-; AVX2-ONLY-NEXT: vmovdqa %ymm3, (%rax)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovaps %ymm2, 64(%rax)
+; AVX2-ONLY-NEXT: vmovaps %ymm6, (%rax)
+; AVX2-ONLY-NEXT: vmovups (%rsp), %ymm2 # 32-byte Reload
; AVX2-ONLY-NEXT: vmovaps %ymm2, 96(%rax)
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
; AVX2-ONLY-NEXT: vmovaps %ymm2, 32(%rax)
; AVX2-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX2-ONLY-NEXT: vmovdqa %ymm5, 96(%rax)
-; AVX2-ONLY-NEXT: vmovdqa %ymm1, 64(%rax)
-; AVX2-ONLY-NEXT: vmovdqa %ymm0, 32(%rax)
-; AVX2-ONLY-NEXT: vmovdqa %ymm4, (%rax)
-; AVX2-ONLY-NEXT: addq $1256, %rsp # imm = 0x4E8
+; AVX2-ONLY-NEXT: vmovaps %ymm1, 96(%rax)
+; AVX2-ONLY-NEXT: vmovaps %ymm4, 64(%rax)
+; AVX2-ONLY-NEXT: vmovaps %ymm0, 32(%rax)
+; AVX2-ONLY-NEXT: vmovaps %ymm3, (%rax)
+; AVX2-ONLY-NEXT: addq $1576, %rsp # imm = 0x628
; AVX2-ONLY-NEXT: vzeroupper
; AVX2-ONLY-NEXT: retq
;
@@ -6779,9 +6794,9 @@ define void @load_i32_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
;
; AVX2-ONLY-LABEL: load_i32_stride8_vf64:
; AVX2-ONLY: # %bb.0:
-; AVX2-ONLY-NEXT: subq $3272, %rsp # imm = 0xCC8
+; AVX2-ONLY-NEXT: subq $3464, %rsp # imm = 0xD88
; AVX2-ONLY-NEXT: vmovaps 288(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vmovaps 256(%rdi), %xmm12
; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm0 = xmm12[0],xmm0[0],xmm12[1],xmm0[1]
; AVX2-ONLY-NEXT: vmovaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
@@ -6827,7 +6842,7 @@ define void @load_i32_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vbroadcastss %xmm1, %xmm1
; AVX2-ONLY-NEXT: vmovaps 960(%rdi), %xmm2
-; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps %xmm2, (%rsp) # 16-byte Spill
; AVX2-ONLY-NEXT: vbroadcastss %xmm2, %xmm2
; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; AVX2-ONLY-NEXT: vmovaps 928(%rdi), %xmm2
@@ -7021,7 +7036,7 @@ define void @load_i32_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm3[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm12[1,1,1,1]
-; AVX2-ONLY-NEXT: vblendps $2, (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT: vblendps $2, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX2-ONLY-NEXT: # xmm0 = xmm0[0],mem[1],xmm0[2,3]
; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; AVX2-ONLY-NEXT: vunpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
@@ -7047,7 +7062,7 @@ define void @load_i32_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: # xmm1 = xmm11[0],mem[0],xmm11[1],mem[1]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps (%rsp), %xmm1 # 16-byte Reload
; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm1 = xmm1[0],xmm11[0],xmm1[1],xmm11[1]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX2-ONLY-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
@@ -7169,13 +7184,13 @@ define void @load_i32_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX2-ONLY-NEXT: vunpckhps (%rsp), %xmm0, %xmm3 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm3 # 16-byte Folded Reload
; AVX2-ONLY-NEXT: # xmm3 = xmm0[2],mem[2],xmm0[3],mem[3]
; AVX2-ONLY-NEXT: vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm2 # 16-byte Folded Reload
; AVX2-ONLY-NEXT: # xmm2 = xmm0[2],mem[2],xmm0[3],mem[3]
-; AVX2-ONLY-NEXT: vmovaps %xmm2, (%rsp) # 16-byte Spill
+; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm12[2,2,2,2]
; AVX2-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX2-ONLY-NEXT: # xmm0 = mem[0,1,2],xmm0[3]
@@ -7198,7 +7213,7 @@ define void @load_i32_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: # xmm2 = xmm0[2],mem[2],xmm0[3],mem[3]
; AVX2-ONLY-NEXT: vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm11[2,2,2,2]
-; AVX2-ONLY-NEXT: vblendps $7, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT: vblendps $7, (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
; AVX2-ONLY-NEXT: # xmm0 = mem[0,1,2],xmm0[3]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm1
@@ -7340,7 +7355,7 @@ define void @load_i32_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm14 # 16-byte Folded Reload
; AVX2-ONLY-NEXT: # xmm14 = xmm5[2],mem[2],xmm5[3],mem[3]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm14, %ymm0, %ymm14
-; AVX2-ONLY-NEXT: vpermilps $238, (%rsp), %xmm12 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
; AVX2-ONLY-NEXT: # xmm12 = mem[2,3,2,3]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm12, %ymm0, %ymm12
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm14[6,7]
@@ -7350,7 +7365,7 @@ define void @load_i32_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm12 = xmm5[2],xmm11[2],xmm5[3],xmm11[3]
; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm11 = xmm5[1],xmm12[1]
-; AVX2-ONLY-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX2-ONLY-NEXT: vmovaps (%rsp), %xmm5 # 16-byte Reload
; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm12 # 16-byte Folded Reload
; AVX2-ONLY-NEXT: # xmm12 = xmm5[2],mem[2],xmm5[3],mem[3]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm12, %ymm0, %ymm12
@@ -7443,462 +7458,479 @@ define void @load_i32_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm1 = xmm15[1],xmm1[1]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 48(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vpbroadcastd 16(%rdi), %ymm1
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; AVX2-ONLY-NEXT: vmovdqa 96(%rdi), %ymm1
-; AVX2-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 64(%rdi), %ymm2
-; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[1],ymm1[1],ymm2[4],ymm1[4],ymm2[5],ymm1[5]
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
-; AVX2-ONLY-NEXT: vmovdqa 224(%rdi), %ymm2
-; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 192(%rdi), %ymm3
-; AVX2-ONLY-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 160(%rdi), %ymm11
-; AVX2-ONLY-NEXT: vmovdqa 128(%rdi), %ymm1
-; AVX2-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm11[0],ymm1[1],ymm11[1],ymm1[4],ymm11[4],ymm1[5],ymm11[5]
-; AVX2-ONLY-NEXT: vmovdqa %ymm11, %ymm10
-; AVX2-ONLY-NEXT: vmovdqu %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm7 = ymm3[0],ymm2[0],ymm3[1],ymm2[1],ymm3[4],ymm2[4],ymm3[5],ymm2[5]
-; AVX2-ONLY-NEXT: vpunpcklqdq {{.*#+}} ymm1 = ymm1[0],ymm7[0],ymm1[2],ymm7[2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 304(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vpbroadcastd 272(%rdi), %ymm1
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; AVX2-ONLY-NEXT: vmovdqa 352(%rdi), %ymm1
-; AVX2-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 320(%rdi), %ymm2
-; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[1],ymm1[1],ymm2[4],ymm1[4],ymm2[5],ymm1[5]
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm1 = ymm1[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
-; AVX2-ONLY-NEXT: vmovdqa 480(%rdi), %ymm1
-; AVX2-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 448(%rdi), %ymm3
-; AVX2-ONLY-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 416(%rdi), %ymm12
-; AVX2-ONLY-NEXT: vmovdqa 384(%rdi), %ymm2
-; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm2 = ymm2[0],ymm12[0],ymm2[1],ymm12[1],ymm2[4],ymm12[4],ymm2[5],ymm12[5]
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm1 = ymm3[0],ymm1[0],ymm3[1],ymm1[1],ymm3[4],ymm1[4],ymm3[5],ymm1[5]
-; AVX2-ONLY-NEXT: vpunpcklqdq {{.*#+}} ymm2 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 560(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vpbroadcastd 528(%rdi), %ymm2
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
-; AVX2-ONLY-NEXT: vmovdqa 608(%rdi), %ymm3
-; AVX2-ONLY-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 576(%rdi), %ymm2
-; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm2 = ymm2[0],ymm3[0],ymm2[1],ymm3[1],ymm2[4],ymm3[4],ymm2[5],ymm3[5]
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm2 = ymm2[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm2[2,3]
-; AVX2-ONLY-NEXT: vmovdqa 736(%rdi), %ymm2
-; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 704(%rdi), %ymm4
-; AVX2-ONLY-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 672(%rdi), %ymm14
-; AVX2-ONLY-NEXT: vmovdqa 640(%rdi), %ymm3
-; AVX2-ONLY-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm3 = ymm3[0],ymm14[0],ymm3[1],ymm14[1],ymm3[4],ymm14[4],ymm3[5],ymm14[5]
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm2 = ymm4[0],ymm2[0],ymm4[1],ymm2[1],ymm4[4],ymm2[4],ymm4[5],ymm2[5]
-; AVX2-ONLY-NEXT: vpunpcklqdq {{.*#+}} ymm3 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 816(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vpbroadcastd 784(%rdi), %ymm3
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
-; AVX2-ONLY-NEXT: vmovdqa 864(%rdi), %ymm4
-; AVX2-ONLY-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 832(%rdi), %ymm3
-; AVX2-ONLY-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm3 = ymm3[0],ymm4[0],ymm3[1],ymm4[1],ymm3[4],ymm4[4],ymm3[5],ymm4[5]
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm3 = ymm3[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm3[2,3]
-; AVX2-ONLY-NEXT: vmovdqa 992(%rdi), %ymm3
-; AVX2-ONLY-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 960(%rdi), %ymm5
-; AVX2-ONLY-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 928(%rdi), %ymm6
-; AVX2-ONLY-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 896(%rdi), %ymm4
-; AVX2-ONLY-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm4 = ymm4[0],ymm6[0],ymm4[1],ymm6[1],ymm4[4],ymm6[4],ymm4[5],ymm6[5]
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm3 = ymm5[0],ymm3[0],ymm5[1],ymm3[1],ymm5[4],ymm3[4],ymm5[5],ymm3[5]
-; AVX2-ONLY-NEXT: vpunpcklqdq {{.*#+}} ymm4 = ymm4[0],ymm3[0],ymm4[2],ymm3[2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm4[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 1072(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vpbroadcastd 1040(%rdi), %ymm4
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
-; AVX2-ONLY-NEXT: vmovdqa 1120(%rdi), %ymm5
-; AVX2-ONLY-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 1088(%rdi), %ymm4
-; AVX2-ONLY-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm4 = ymm4[0],ymm5[0],ymm4[1],ymm5[1],ymm4[4],ymm5[4],ymm4[5],ymm5[5]
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm4 = ymm4[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm4[2,3]
-; AVX2-ONLY-NEXT: vmovdqa 1248(%rdi), %ymm4
-; AVX2-ONLY-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 1216(%rdi), %ymm6
-; AVX2-ONLY-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 1184(%rdi), %ymm8
-; AVX2-ONLY-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 1152(%rdi), %ymm5
-; AVX2-ONLY-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm5 = ymm5[0],ymm8[0],ymm5[1],ymm8[1],ymm5[4],ymm8[4],ymm5[5],ymm8[5]
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm4 = ymm6[0],ymm4[0],ymm6[1],ymm4[1],ymm6[4],ymm4[4],ymm6[5],ymm4[5]
-; AVX2-ONLY-NEXT: vpunpcklqdq {{.*#+}} ymm5 = ymm5[0],ymm4[0],ymm5[2],ymm4[2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm5[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 1328(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vpbroadcastd 1296(%rdi), %ymm5
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm5[0],xmm0[0],xmm5[1],xmm0[1]
-; AVX2-ONLY-NEXT: vmovdqa 1376(%rdi), %ymm5
-; AVX2-ONLY-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 1344(%rdi), %ymm6
-; AVX2-ONLY-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm5 = ymm6[0],ymm5[0],ymm6[1],ymm5[1],ymm6[4],ymm5[4],ymm6[5],ymm5[5]
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm5 = ymm5[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm5[2,3]
-; AVX2-ONLY-NEXT: vmovdqa 1504(%rdi), %ymm5
-; AVX2-ONLY-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 1472(%rdi), %ymm8
-; AVX2-ONLY-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 1440(%rdi), %ymm9
-; AVX2-ONLY-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 1408(%rdi), %ymm6
-; AVX2-ONLY-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm6 = ymm6[0],ymm9[0],ymm6[1],ymm9[1],ymm6[4],ymm9[4],ymm6[5],ymm9[5]
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm5 = ymm8[0],ymm5[0],ymm8[1],ymm5[1],ymm8[4],ymm5[4],ymm8[5],ymm5[5]
-; AVX2-ONLY-NEXT: vpunpcklqdq {{.*#+}} ymm6 = ymm6[0],ymm5[0],ymm6[2],ymm5[2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm6[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 1584(%rdi), %xmm0
-; AVX2-ONLY-NEXT: vpbroadcastd 1552(%rdi), %ymm6
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm6[0],xmm0[0],xmm6[1],xmm0[1]
-; AVX2-ONLY-NEXT: vmovdqa 1632(%rdi), %ymm6
-; AVX2-ONLY-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 1600(%rdi), %ymm8
-; AVX2-ONLY-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm6 = ymm8[0],ymm6[0],ymm8[1],ymm6[1],ymm8[4],ymm6[4],ymm8[5],ymm6[5]
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm6 = ymm6[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm6 = xmm0[0,1],xmm6[2,3]
-; AVX2-ONLY-NEXT: vmovdqa 1760(%rdi), %ymm0
-; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 1728(%rdi), %ymm9
-; AVX2-ONLY-NEXT: vmovdqu %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 1696(%rdi), %ymm11
-; AVX2-ONLY-NEXT: vmovdqu %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 1664(%rdi), %ymm8
-; AVX2-ONLY-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm8 = ymm8[0],ymm11[0],ymm8[1],ymm11[1],ymm8[4],ymm11[4],ymm8[5],ymm11[5]
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm13 = ymm9[0],ymm0[0],ymm9[1],ymm0[1],ymm9[4],ymm0[4],ymm9[5],ymm0[5]
-; AVX2-ONLY-NEXT: vpunpcklqdq {{.*#+}} ymm8 = ymm8[0],ymm13[0],ymm8[2],ymm13[2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm8[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 1840(%rdi), %xmm6
-; AVX2-ONLY-NEXT: vpbroadcastd 1808(%rdi), %ymm8
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} xmm6 = xmm8[0],xmm6[0],xmm8[1],xmm6[1]
-; AVX2-ONLY-NEXT: vmovdqa 1888(%rdi), %ymm8
-; AVX2-ONLY-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 1856(%rdi), %ymm9
-; AVX2-ONLY-NEXT: vmovdqu %ymm9, (%rsp) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm8 = ymm9[0],ymm8[0],ymm9[1],ymm8[1],ymm9[4],ymm8[4],ymm9[5],ymm8[5]
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm8 = ymm8[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm8 = xmm6[0,1],xmm8[2,3]
-; AVX2-ONLY-NEXT: vmovdqa 2016(%rdi), %ymm6
-; AVX2-ONLY-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 1984(%rdi), %ymm11
-; AVX2-ONLY-NEXT: vmovdqu %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 1952(%rdi), %ymm15
-; AVX2-ONLY-NEXT: vmovdqa 1920(%rdi), %ymm0
-; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm9 = ymm0[0],ymm15[0],ymm0[1],ymm15[1],ymm0[4],ymm15[4],ymm0[5],ymm15[5]
-; AVX2-ONLY-NEXT: vmovdqu %ymm15, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm6 = ymm11[0],ymm6[0],ymm11[1],ymm6[1],ymm11[4],ymm6[4],ymm11[5],ymm6[5]
-; AVX2-ONLY-NEXT: vpunpcklqdq {{.*#+}} ymm9 = ymm9[0],ymm6[0],ymm9[2],ymm6[2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm8 = ymm8[0,1,2,3],ymm9[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 148(%rdi), %ymm8
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm8 = ymm8[0,1,2,3,4],ymm10[5],ymm8[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm7[6,7]
-; AVX2-ONLY-NEXT: vpmovzxdq {{.*#+}} xmm7 = mem[0],zero,mem[1],zero
-; AVX2-ONLY-NEXT: vpbroadcastd 116(%rdi), %ymm9
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm9 = xmm7[0,1,2],xmm9[3]
-; AVX2-ONLY-NEXT: vmovdqa 32(%rdi), %ymm11
-; AVX2-ONLY-NEXT: vmovdqa (%rdi), %ymm0
-; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm10 = ymm0[1,1,1,1,5,5,5,5]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm10 = ymm10[0],ymm11[1],ymm10[2,3,4],ymm11[5],ymm10[6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm10, %xmm10
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm9 = xmm10[0,1],xmm9[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm8 = ymm9[0,1,2,3],ymm8[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 404(%rdi), %ymm8
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm8 = ymm8[0,1,2,3,4],ymm12[5],ymm8[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm8[0,1,2,3,4,5],ymm1[6,7]
-; AVX2-ONLY-NEXT: vpmovzxdq {{.*#+}} xmm8 = mem[0],zero,mem[1],zero
-; AVX2-ONLY-NEXT: vpbroadcastd 372(%rdi), %ymm9
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm8 = xmm8[0,1,2],xmm9[3]
-; AVX2-ONLY-NEXT: vmovdqa 288(%rdi), %ymm10
-; AVX2-ONLY-NEXT: vmovdqa 256(%rdi), %ymm0
-; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm9 = ymm0[1,1,1,1,5,5,5,5]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm9 = ymm9[0],ymm10[1],ymm9[2,3,4],ymm10[5],ymm9[6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm9, %xmm9
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm8 = xmm9[0,1],xmm8[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm8[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 660(%rdi), %ymm1
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm14[5],ymm1[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
-; AVX2-ONLY-NEXT: vpmovzxdq {{.*#+}} xmm2 = mem[0],zero,mem[1],zero
-; AVX2-ONLY-NEXT: vpbroadcastd 628(%rdi), %ymm8
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm2 = xmm2[0,1,2],xmm8[3]
-; AVX2-ONLY-NEXT: vmovdqa 544(%rdi), %ymm0
-; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 512(%rdi), %ymm7
-; AVX2-ONLY-NEXT: vmovdqu %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm8 = ymm7[1,1,1,1,5,5,5,5]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm8 = ymm8[0],ymm0[1],ymm8[2,3,4],ymm0[5],ymm8[6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm8, %xmm8
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm2 = xmm8[0,1],xmm2[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 916(%rdi), %ymm1
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm8[5],ymm1[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm3[6,7]
-; AVX2-ONLY-NEXT: vpmovzxdq {{.*#+}} xmm2 = mem[0],zero,mem[1],zero
-; AVX2-ONLY-NEXT: vpbroadcastd 884(%rdi), %ymm3
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm2 = xmm2[0,1,2],xmm3[3]
-; AVX2-ONLY-NEXT: vmovdqa 800(%rdi), %ymm0
-; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 768(%rdi), %ymm3
-; AVX2-ONLY-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm3 = ymm3[1,1,1,1,5,5,5,5]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0],ymm0[1],ymm3[2,3,4],ymm0[5],ymm3[6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm3, %xmm3
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 1172(%rdi), %ymm1
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm9[5],ymm1[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm4[6,7]
-; AVX2-ONLY-NEXT: vpmovzxdq {{.*#+}} xmm2 = mem[0],zero,mem[1],zero
-; AVX2-ONLY-NEXT: vpbroadcastd 1140(%rdi), %ymm3
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm2 = xmm2[0,1,2],xmm3[3]
-; AVX2-ONLY-NEXT: vmovdqa 1056(%rdi), %ymm0
-; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 1024(%rdi), %ymm3
-; AVX2-ONLY-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm3 = ymm3[1,1,1,1,5,5,5,5]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0],ymm0[1],ymm3[2,3,4],ymm0[5],ymm3[6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm3, %xmm3
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 1428(%rdi), %ymm1
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm7[5],ymm1[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm5[6,7]
-; AVX2-ONLY-NEXT: vpmovzxdq {{.*#+}} xmm2 = mem[0],zero,mem[1],zero
-; AVX2-ONLY-NEXT: vpbroadcastd 1396(%rdi), %ymm3
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm2 = xmm2[0,1,2],xmm3[3]
-; AVX2-ONLY-NEXT: vmovdqa 1312(%rdi), %ymm5
-; AVX2-ONLY-NEXT: vmovdqa 1280(%rdi), %ymm0
-; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm3 = ymm0[1,1,1,1,5,5,5,5]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0],ymm5[1],ymm3[2,3,4],ymm5[5],ymm3[6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm3, %xmm3
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 1684(%rdi), %ymm1
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm4[5],ymm1[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm1[0,1,2,3,4,5],ymm13[6,7]
-; AVX2-ONLY-NEXT: vpmovzxdq {{.*#+}} xmm1 = mem[0],zero,mem[1],zero
-; AVX2-ONLY-NEXT: vpbroadcastd 1652(%rdi), %ymm2
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0,1,2],xmm2[3]
-; AVX2-ONLY-NEXT: vmovdqa 1568(%rdi), %ymm13
-; AVX2-ONLY-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa 1536(%rdi), %ymm0
-; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm2 = ymm0[1,1,1,1,5,5,5,5]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0],ymm13[1],ymm2[2,3,4],ymm13[5],ymm2[6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm2, %xmm2
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm3[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 1940(%rdi), %ymm0
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm15[5],ymm0[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm6[6,7]
-; AVX2-ONLY-NEXT: vpmovzxdq {{.*#+}} xmm1 = mem[0],zero,mem[1],zero
-; AVX2-ONLY-NEXT: vpbroadcastd 1908(%rdi), %ymm2
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0,1,2],xmm2[3]
-; AVX2-ONLY-NEXT: vmovdqa 1824(%rdi), %ymm15
-; AVX2-ONLY-NEXT: vmovdqa 1792(%rdi), %ymm2
-; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[1,1,1,1,5,5,5,5]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0],ymm15[1],ymm2[2,3,4],ymm15[5],ymm2[6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm2, %xmm2
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 248(%rdi), %ymm0
-; AVX2-ONLY-NEXT: vpblendd $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vmovaps 32(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovaps (%rdi), %ymm2
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm0 = ymm2[0],ymm1[0],ymm2[1],ymm1[1],ymm2[4],ymm1[4],ymm2[5],ymm1[5]
+; AVX2-ONLY-NEXT: vmovaps %ymm2, %ymm13
+; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps %ymm1, %ymm12
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm0, %xmm0
+; AVX2-ONLY-NEXT: vmovaps 96(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 64(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[1],ymm1[1],ymm2[4],ymm1[4],ymm2[5],ymm1[5]
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT: vmovaps 224(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 192(%rdi), %ymm4
+; AVX2-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 160(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vmovaps 128(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX2-ONLY-NEXT: vmovaps %ymm3, %ymm10
+; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm15 = ymm4[0],ymm2[0],ymm4[1],ymm2[1],ymm4[4],ymm2[4],ymm4[5],ymm2[5]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm15[0],ymm1[2],ymm15[2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 288(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 256(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[1],ymm0[1],ymm1[4],ymm0[4],ymm1[5],ymm0[5]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm0, %xmm0
+; AVX2-ONLY-NEXT: vmovaps 352(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 320(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[1],ymm1[1],ymm2[4],ymm1[4],ymm2[5],ymm1[5]
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT: vmovaps 480(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 448(%rdi), %ymm4
+; AVX2-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 416(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 384(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm8 = ymm4[0],ymm2[0],ymm4[1],ymm2[1],ymm4[4],ymm2[4],ymm4[5],ymm2[5]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm8[0],ymm1[2],ymm8[2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 544(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 512(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm0, %xmm0
+; AVX2-ONLY-NEXT: vmovaps 608(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 576(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[1],ymm1[1],ymm2[4],ymm1[4],ymm2[5],ymm1[5]
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT: vmovaps 736(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 704(%rdi), %ymm4
+; AVX2-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 672(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 640(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm7 = ymm4[0],ymm2[0],ymm4[1],ymm2[1],ymm4[4],ymm2[4],ymm4[5],ymm2[5]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm7[0],ymm1[2],ymm7[2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 800(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 768(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm0, %xmm0
+; AVX2-ONLY-NEXT: vmovaps 864(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 832(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[1],ymm1[1],ymm2[4],ymm1[4],ymm2[5],ymm1[5]
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT: vmovaps 992(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 960(%rdi), %ymm4
+; AVX2-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 928(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 896(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm5 = ymm4[0],ymm2[0],ymm4[1],ymm2[1],ymm4[4],ymm2[4],ymm4[5],ymm2[5]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm5[0],ymm1[2],ymm5[2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 1056(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 1024(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm0, %xmm0
+; AVX2-ONLY-NEXT: vmovaps 1120(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 1088(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[1],ymm1[1],ymm2[4],ymm1[4],ymm2[5],ymm1[5]
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT: vmovaps 1248(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 1216(%rdi), %ymm4
+; AVX2-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 1184(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 1152(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm4 = ymm4[0],ymm2[0],ymm4[1],ymm2[1],ymm4[4],ymm2[4],ymm4[5],ymm2[5]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm4[0],ymm1[2],ymm4[2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 1312(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 1280(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm0, %xmm0
+; AVX2-ONLY-NEXT: vmovaps 1376(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 1344(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[1],ymm1[1],ymm2[4],ymm1[4],ymm2[5],ymm1[5]
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT: vmovaps 1504(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 1472(%rdi), %ymm6
+; AVX2-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 1440(%rdi), %ymm3
+; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 1408(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm3 = ymm6[0],ymm2[0],ymm6[1],ymm2[1],ymm6[4],ymm2[4],ymm6[5],ymm2[5]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[2],ymm3[2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 1568(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 1536(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm0, %xmm0
+; AVX2-ONLY-NEXT: vmovaps 1632(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 1600(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[1],ymm1[1],ymm2[4],ymm1[4],ymm2[5],ymm1[5]
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT: vmovaps 1760(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 1728(%rdi), %ymm11
+; AVX2-ONLY-NEXT: vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 1696(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 1664(%rdi), %ymm6
+; AVX2-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm6 = ymm6[0],ymm2[0],ymm6[1],ymm2[1],ymm6[4],ymm2[4],ymm6[5],ymm2[5]
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm2 = ymm11[0],ymm1[0],ymm11[1],ymm1[1],ymm11[4],ymm1[4],ymm11[5],ymm1[5]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm6 = ymm6[0],ymm2[0],ymm6[2],ymm2[2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm6[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 1824(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 1792(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm0, %xmm0
+; AVX2-ONLY-NEXT: vmovaps 1888(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 1856(%rdi), %ymm6
+; AVX2-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm1 = ymm6[0],ymm1[0],ymm6[1],ymm1[1],ymm6[4],ymm1[4],ymm6[5],ymm1[5]
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm14 = ymm1[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm14 = xmm0[0,1],xmm14[2,3]
+; AVX2-ONLY-NEXT: vmovaps 2016(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 1984(%rdi), %ymm11
+; AVX2-ONLY-NEXT: vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 1952(%rdi), %ymm9
+; AVX2-ONLY-NEXT: vmovaps 1920(%rdi), %ymm1
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm1 = ymm1[0],ymm9[0],ymm1[1],ymm9[1],ymm1[4],ymm9[4],ymm1[5],ymm9[5]
+; AVX2-ONLY-NEXT: vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm0 = ymm11[0],ymm0[0],ymm11[1],ymm0[1],ymm11[4],ymm0[4],ymm11[5],ymm0[5]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm14[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 148(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm10[5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm14
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} ymm15 = ymm13[1,1,1,1,5,5,5,5]
+; AVX2-ONLY-NEXT: vmovaps %ymm12, %ymm6
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm15 = ymm15[0],ymm12[1],ymm15[2,3,4],ymm12[5],ymm15[6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm15, %xmm15
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm14 = xmm15[0,1],xmm14[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm14[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 404(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm11 # 32-byte Reload
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm11[5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm8[6,7]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm8
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm10[1,1,1,1,5,5,5,5]
+; AVX2-ONLY-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm14 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm14 = ymm14[0],mem[1],ymm14[2,3,4],mem[5],ymm14[6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm14, %xmm14
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm8 = xmm14[0,1],xmm8[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 660(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm12[5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm7[6,7]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm7
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm13[1,1,1,1,5,5,5,5]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0],ymm14[1],ymm8[2,3,4],ymm14[5],ymm8[6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm8, %xmm8
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm7 = xmm8[0,1],xmm7[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm7[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 916(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm8[5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm5[6,7]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm5
+; AVX2-ONLY-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm7 = mem[1,1,1,1,5,5,5,5]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0],ymm15[1],ymm7[2,3,4],ymm15[5],ymm7[6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm7, %xmm7
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm7[0,1],xmm5[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 1172(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm7[5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm4[6,7]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm4
+; AVX2-ONLY-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm5 = mem[1,1,1,1,5,5,5,5]
+; AVX2-ONLY-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm5 = ymm5[0],mem[1],ymm5[2,3,4],mem[5],ymm5[6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm5, %xmm5
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm5[0,1],xmm4[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 1428(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm5[5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm3
+; AVX2-ONLY-NEXT: vpermilps $85, (%rsp), %ymm4 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm4 = mem[1,1,1,1,5,5,5,5]
+; AVX2-ONLY-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm4 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm4 = ymm4[0],mem[1],ymm4[2,3,4],mem[5],ymm4[6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm4, %xmm4
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm4[0,1],xmm3[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 1684(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm4[5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm2
+; AVX2-ONLY-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm3 = mem[1,1,1,1,5,5,5,5]
+; AVX2-ONLY-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm3 = ymm3[0],mem[1],ymm3[2,3,4],mem[5],ymm3[6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm3, %xmm3
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm3[0,1],xmm2[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 1940(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm9[5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps $192, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1
+; AVX2-ONLY-NEXT: vpermilps $85, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm2 = mem[1,1,1,1,5,5,5,5]
+; AVX2-ONLY-NEXT: vblendps $34, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm2 = ymm2[0],mem[1],ymm2[2,3,4],mem[5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm2
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 248(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vblendps $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: # ymm0 = mem[0,1,2,3,4,5,6],ymm0[7]
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm6 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm6 = ymm1[2],mem[2],ymm1[3],mem[3],ymm1[6],mem[6],ymm1[7],mem[7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm2 = ymm1[2],ymm11[2],ymm1[3],ymm11[3],ymm1[6],ymm11[6],ymm1[7],ymm11[7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm3 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm3 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: # ymm3 = ymm1[2],mem[2],ymm1[3],mem[3],ymm1[6],mem[6],ymm1[7],mem[7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm2, %xmm1
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm2 = ymm3[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm6[0,1,2,3,4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 504(%rdi), %ymm0
-; AVX2-ONLY-NEXT: vpblendd $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm2 = ymm1[2],ymm6[2],ymm1[3],ymm6[3],ymm1[6],ymm6[6],ymm1[7],ymm6[7]
+; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm6 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm6 = ymm1[2],mem[2],ymm1[3],mem[3],ymm1[6],mem[6],ymm1[7],mem[7]
+; AVX2-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm1
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm2 = ymm6[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 504(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vblendps $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: # ymm0 = mem[0,1,2,3,4,5,6],ymm0[7]
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm3 = ymm1[2],ymm12[2],ymm1[3],ymm12[3],ymm1[6],ymm12[6],ymm1[7],ymm12[7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm2 = ymm1[2],ymm10[2],ymm1[3],ymm10[3],ymm1[6],ymm10[6],ymm1[7],ymm10[7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm6 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm3 = ymm1[2],ymm11[2],ymm1[3],ymm11[3],ymm1[6],ymm11[6],ymm1[7],ymm11[7]
+; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm2 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm2 = ymm10[2],mem[2],ymm10[3],mem[3],ymm10[6],mem[6],ymm10[7],mem[7]
+; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm6 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: # ymm6 = ymm1[2],mem[2],ymm1[3],mem[3],ymm1[6],mem[6],ymm1[7],mem[7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm2, %xmm1
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm2 = ymm6[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 760(%rdi), %ymm0
-; AVX2-ONLY-NEXT: vpblendd $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm1
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm2 = ymm6[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 760(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vblendps $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: # ymm0 = mem[0,1,2,3,4,5,6],ymm0[7]
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm3 = ymm1[2],ymm14[2],ymm1[3],ymm14[3],ymm1[6],ymm14[6],ymm1[7],ymm14[7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm2 = ymm1[2],mem[2],ymm1[3],mem[3],ymm1[6],mem[6],ymm1[7],mem[7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm6 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm3 = ymm1[2],ymm12[2],ymm1[3],ymm12[3],ymm1[6],ymm12[6],ymm1[7],ymm12[7]
+; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm2 = ymm13[2],ymm14[2],ymm13[3],ymm14[3],ymm13[6],ymm14[6],ymm13[7],ymm14[7]
+; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm6 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: # ymm6 = ymm1[2],mem[2],ymm1[3],mem[3],ymm1[6],mem[6],ymm1[7],mem[7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm2, %xmm1
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm2 = ymm6[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 1016(%rdi), %ymm0
-; AVX2-ONLY-NEXT: vpblendd $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm1
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm2 = ymm6[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 1016(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vblendps $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: # ymm0 = mem[0,1,2,3,4,5,6],ymm0[7]
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm3 = ymm1[2],ymm8[2],ymm1[3],ymm8[3],ymm1[6],ymm8[6],ymm1[7],ymm8[7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT: # ymm2 = ymm1[2],mem[2],ymm1[3],mem[3],ymm1[6],mem[6],ymm1[7],mem[7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm12 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm3 = ymm1[2],ymm8[2],ymm1[3],ymm8[3],ymm1[6],ymm8[6],ymm1[7],ymm8[7]
+; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm2 = ymm1[2],ymm15[2],ymm1[3],ymm15[3],ymm1[6],ymm15[6],ymm1[7],ymm15[7]
+; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm12 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: # ymm12 = ymm1[2],mem[2],ymm1[3],mem[3],ymm1[6],mem[6],ymm1[7],mem[7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm2, %xmm1
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm2 = ymm12[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 1272(%rdi), %ymm0
-; AVX2-ONLY-NEXT: vpblendd $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm1
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm2 = ymm12[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 1272(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vblendps $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: # ymm0 = mem[0,1,2,3,4,5,6],ymm0[7]
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm13 = ymm1[2],ymm9[2],ymm1[3],ymm9[3],ymm1[6],ymm9[6],ymm1[7],ymm9[7]
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm14 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm13 = ymm1[2],ymm7[2],ymm1[3],ymm7[3],ymm1[6],ymm7[6],ymm1[7],ymm7[7]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm14 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: # ymm14 = ymm1[2],mem[2],ymm1[3],mem[3],ymm1[6],mem[6],ymm1[7],mem[7]
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm9 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm9 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: # ymm9 = ymm1[2],mem[2],ymm1[3],mem[3],ymm1[6],mem[6],ymm1[7],mem[7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm14, %xmm1
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm2 = ymm9[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm13[0,1,2,3,4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 1528(%rdi), %ymm0
-; AVX2-ONLY-NEXT: vpblendd $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm14, %xmm1
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm2 = ymm9[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm13[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 1528(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vblendps $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: # ymm0 = mem[0,1,2,3,4,5,6],ymm0[7]
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm10 = ymm1[2],ymm7[2],ymm1[3],ymm7[3],ymm1[6],ymm7[6],ymm1[7],ymm7[7]
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm11 = ymm1[2],ymm5[2],ymm1[3],ymm5[3],ymm1[6],ymm5[6],ymm1[7],ymm5[7]
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm6 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm10 = ymm1[2],ymm5[2],ymm1[3],ymm5[3],ymm1[6],ymm5[6],ymm1[7],ymm5[7]
+; AVX2-ONLY-NEXT: vmovups (%rsp), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm11 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm11 = ymm1[2],mem[2],ymm1[3],mem[3],ymm1[6],mem[6],ymm1[7],mem[7]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm6 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: # ymm6 = ymm1[2],mem[2],ymm1[3],mem[3],ymm1[6],mem[6],ymm1[7],mem[7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm11, %xmm1
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm2 = ymm6[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm10[0,1,2,3,4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 1784(%rdi), %ymm0
-; AVX2-ONLY-NEXT: vpblendd $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm11, %xmm1
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm2 = ymm6[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm10[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 1784(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vblendps $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: # ymm0 = mem[0,1,2,3,4,5,6],ymm0[7]
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm7 = ymm1[2],ymm4[2],ymm1[3],ymm4[3],ymm1[6],ymm4[6],ymm1[7],ymm4[7]
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm8 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm7 = ymm1[2],ymm4[2],ymm1[3],ymm4[3],ymm1[6],ymm4[6],ymm1[7],ymm4[7]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm8 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: # ymm8 = ymm1[2],mem[2],ymm1[3],mem[3],ymm1[6],mem[6],ymm1[7],mem[7]
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm3 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm3 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: # ymm3 = ymm1[2],mem[2],ymm1[3],mem[3],ymm1[6],mem[6],ymm1[7],mem[7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm8, %xmm1
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm2 = ymm3[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm7[0,1,2,3,4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 2040(%rdi), %ymm0
-; AVX2-ONLY-NEXT: vpblendd $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm8, %xmm1
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm2 = ymm3[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm7[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 2040(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vblendps $127, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: # ymm1 = mem[0,1,2,3,4,5,6],ymm0[7]
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm4 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm4 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: # ymm4 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7]
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm5 = ymm0[2],ymm15[2],ymm0[3],ymm15[3],ymm0[6],ymm15[6],ymm0[7],ymm15[7]
-; AVX2-ONLY-NEXT: vmovdqu (%rsp), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm5 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: # ymm5 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7]
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm2 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: # ymm2 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm5, %xmm0
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm15 = ymm2[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm15[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm4[0,1,2,3,4,5],ymm1[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm5, %xmm0
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm15 = ymm2[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm15[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm4[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vbroadcastss 220(%rdi), %ymm0
; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: # ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7]
@@ -7937,53 +7969,53 @@ define void @load_i32_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vextractf128 $1, %ymm15, %xmm15
; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm15[0,1],xmm1[2,3]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm15 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT: vpbroadcastd 988(%rdi), %ymm0
-; AVX2-ONLY-NEXT: vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vbroadcastss 988(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: # ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7]
-; AVX2-ONLY-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm12, %xmm1
-; AVX2-ONLY-NEXT: vpshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm12, %xmm1
+; AVX2-ONLY-NEXT: vpermilps $238, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: # ymm12 = mem[2,3,2,3,6,7,6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm12, %xmm12
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm1 = xmm12[0,1],xmm1[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm12 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT: vpbroadcastd 1244(%rdi), %ymm0
-; AVX2-ONLY-NEXT: vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm12, %xmm12
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm12[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm12 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vbroadcastss 1244(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: # ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7]
-; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm0 = ymm13[1],ymm0[1],ymm13[3],ymm0[3]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm9, %xmm1
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm9 = ymm14[2,3,2,3,6,7,6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm9, %xmm9
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm1 = xmm9[0,1],xmm1[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm9 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT: vpbroadcastd 1500(%rdi), %ymm0
-; AVX2-ONLY-NEXT: vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm13[1],ymm0[1],ymm13[3],ymm0[3]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm9, %xmm1
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} ymm9 = ymm14[2,3,2,3,6,7,6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm9, %xmm9
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm9[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vbroadcastss 1500(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: # ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7]
-; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm0 = ymm10[1],ymm0[1],ymm10[3],ymm0[3]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm6, %xmm1
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm6 = ymm11[2,3,2,3,6,7,6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm6, %xmm6
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm1 = xmm6[0,1],xmm1[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm6 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT: vpbroadcastd 1756(%rdi), %ymm0
-; AVX2-ONLY-NEXT: vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm10[1],ymm0[1],ymm10[3],ymm0[3]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm6, %xmm1
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm11[2,3,2,3,6,7,6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm6, %xmm6
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm6[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vbroadcastss 1756(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: # ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7]
-; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm0 = ymm7[1],ymm0[1],ymm7[3],ymm0[3]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm3, %xmm1
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm3 = ymm8[2,3,2,3,6,7,6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm3, %xmm3
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm1 = xmm3[0,1],xmm1[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT: vpbroadcastd 2012(%rdi), %ymm0
-; AVX2-ONLY-NEXT: vpunpckhdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm7[1],ymm0[1],ymm7[3],ymm0[3]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm3, %xmm1
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm8[2,3,2,3,6,7,6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm3, %xmm3
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm3[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vbroadcastss 2012(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vunpckhps {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX2-ONLY-NEXT: # ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7]
-; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm0 = ymm4[1],ymm0[1],ymm4[3],ymm0[3]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm2, %xmm1
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm2 = ymm5[2,3,2,3,6,7,6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm2, %xmm2
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm4[1],ymm0[1],ymm4[3],ymm0[3]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm1
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm5[2,3,2,3,6,7,6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm2
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-ONLY-NEXT: vmovaps %ymm1, 192(%rsi)
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
@@ -8086,7 +8118,7 @@ define void @load_i32_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vmovaps %ymm1, 224(%rax)
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-ONLY-NEXT: vmovaps %ymm1, 192(%rax)
-; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT: vmovups (%rsp), %ymm1 # 32-byte Reload
; AVX2-ONLY-NEXT: vmovaps %ymm1, 160(%rax)
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-ONLY-NEXT: vmovaps %ymm1, 128(%rax)
@@ -8099,17 +8131,17 @@ define void @load_i32_stride8_vf64(ptr %in.vec, ptr %out.vec0, ptr %out.vec1, pt
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; AVX2-ONLY-NEXT: vmovaps %ymm1, (%rax)
; AVX2-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX2-ONLY-NEXT: vmovdqa %ymm0, 224(%rax)
-; AVX2-ONLY-NEXT: vmovdqa %ymm3, 192(%rax)
-; AVX2-ONLY-NEXT: vmovdqa %ymm6, 160(%rax)
-; AVX2-ONLY-NEXT: vmovdqa %ymm9, 128(%rax)
-; AVX2-ONLY-NEXT: vmovdqa %ymm12, 96(%rax)
+; AVX2-ONLY-NEXT: vmovaps %ymm0, 224(%rax)
+; AVX2-ONLY-NEXT: vmovaps %ymm3, 192(%rax)
+; AVX2-ONLY-NEXT: vmovaps %ymm6, 160(%rax)
+; AVX2-ONLY-NEXT: vmovaps %ymm9, 128(%rax)
+; AVX2-ONLY-NEXT: vmovaps %ymm12, 96(%rax)
; AVX2-ONLY-NEXT: vmovaps %ymm15, 64(%rax)
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-ONLY-NEXT: vmovaps %ymm0, 32(%rax)
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-ONLY-NEXT: vmovaps %ymm0, (%rax)
-; AVX2-ONLY-NEXT: addq $3272, %rsp # imm = 0xCC8
+; AVX2-ONLY-NEXT: addq $3464, %rsp # imm = 0xD88
; AVX2-ONLY-NEXT: vzeroupper
; AVX2-ONLY-NEXT: retq
;
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-5.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-5.ll
index 6341b2539ed08..fd46e1c63af6c 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-5.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-5.ll
@@ -165,7 +165,7 @@ define void @store_i32_stride5_vf4(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm2, %ymm5
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm3, %ymm6
; AVX1-ONLY-NEXT: vunpcklps {{.*#+}} ymm5 = ymm5[0],ymm6[0],ymm5[1],ymm6[1],ymm5[4],ymm6[4],ymm5[5],ymm6[5]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm5 = ymm5[0,0,2,3]
+; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm5 = ymm5[u,u,1,0,u,u,u,6]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm1
; AVX1-ONLY-NEXT: vunpcklps {{.*#+}} ymm7 = ymm4[0],ymm1[0],ymm4[1],ymm1[1],ymm4[4],ymm1[4],ymm4[5],ymm1[5]
; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm7 = ymm7[0,0,3,3]
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-8.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-8.ll
index fc8bd07e24844..0366e426342e0 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-8.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-8.ll
@@ -597,113 +597,110 @@ define void @store_i32_stride8_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX2-ONLY: # %bb.0:
; AVX2-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX2-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %r10
-; AVX2-ONLY-NEXT: vmovdqa (%rdi), %ymm1
-; AVX2-ONLY-NEXT: vmovdqa (%rsi), %ymm2
-; AVX2-ONLY-NEXT: vmovdqa (%rdx), %ymm3
-; AVX2-ONLY-NEXT: vmovdqa (%rcx), %ymm4
-; AVX2-ONLY-NEXT: vmovdqa (%r8), %ymm5
-; AVX2-ONLY-NEXT: vmovdqa (%r9), %ymm6
-; AVX2-ONLY-NEXT: vmovdqa (%r10), %ymm8
-; AVX2-ONLY-NEXT: vmovdqa (%rax), %ymm9
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm7 = ymm8[0],ymm9[0],ymm8[1],ymm9[1],ymm8[4],ymm9[4],ymm8[5],ymm9[5]
-; AVX2-ONLY-NEXT: vpbroadcastd 20(%r8), %ymm0
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm6[5],ymm0[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm7[6,7]
-; AVX2-ONLY-NEXT: vpmovzxdq {{.*#+}} xmm10 = mem[0],zero,mem[1],zero
-; AVX2-ONLY-NEXT: vpbroadcastd 20(%rcx), %ymm11
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm10 = xmm10[0,1,2],xmm11[3]
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm11 = ymm1[1,1,1,1,5,5,5,5]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm11 = ymm11[0],ymm2[1],ymm11[2,3,4],ymm2[5],ymm11[6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm11, %xmm11
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm10 = xmm11[0,1],xmm10[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm10[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 28(%r10), %ymm10
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm9 = ymm10[2],ymm9[2],ymm10[3],ymm9[3],ymm10[6],ymm9[6],ymm10[7],ymm9[7]
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm10 = ymm5[2],ymm6[2],ymm5[3],ymm6[3],ymm5[6],ymm6[6],ymm5[7],ymm6[7]
-; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm9 = ymm10[1],ymm9[1],ymm10[3],ymm9[3]
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm11 = ymm3[2],ymm4[2],ymm3[3],ymm4[3],ymm3[6],ymm4[6],ymm3[7],ymm4[7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm11, %xmm12
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm2 = ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[6],ymm2[6],ymm1[7],ymm2[7]
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm1 = ymm2[2,3,2,3,6,7,6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm1, %xmm1
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm12[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm9[4,5,6,7]
-; AVX2-ONLY-NEXT: vpbroadcastd 24(%rax), %ymm9
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5,6],ymm9[7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm8 = ymm10[0,1,2,3,4,5],ymm8[6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm2, %xmm2
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm9 = ymm11[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm2 = xmm2[0,1],xmm9[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm8[4,5,6,7]
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm5 = ymm5[0],ymm6[0],ymm5[1],ymm6[1],ymm5[4],ymm6[4],ymm5[5],ymm6[5]
-; AVX2-ONLY-NEXT: vpunpcklqdq {{.*#+}} ymm5 = ymm5[0],ymm7[0],ymm5[2],ymm7[2]
-; AVX2-ONLY-NEXT: vpbroadcastd 16(%rsi), %xmm6
-; AVX2-ONLY-NEXT: vpbroadcastd 16(%rdi), %ymm7
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} xmm6 = xmm7[0],xmm6[0],xmm7[1],xmm6[1]
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm3 = ymm3[0],ymm4[0],ymm3[1],ymm4[1],ymm3[4],ymm4[4],ymm3[5],ymm4[5]
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm3 = ymm3[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm3 = xmm6[0,1],xmm3[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm5[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqa (%rcx), %xmm4
-; AVX2-ONLY-NEXT: vmovdqa (%rdx), %xmm5
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} xmm6 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
-; AVX2-ONLY-NEXT: vmovdqa (%rsi), %xmm9
-; AVX2-ONLY-NEXT: vmovdqa (%rdi), %xmm10
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} xmm7 = xmm10[1,1,1,1]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm7 = xmm7[0],xmm9[1],xmm7[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm8 = xmm7[0,1],xmm6[2,3]
-; AVX2-ONLY-NEXT: vmovdqa (%rax), %xmm6
-; AVX2-ONLY-NEXT: vmovdqa (%r10), %xmm7
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} xmm11 = xmm7[0],xmm6[0],xmm7[1],xmm6[1]
-; AVX2-ONLY-NEXT: vinserti128 $1, %xmm11, %ymm0, %ymm11
-; AVX2-ONLY-NEXT: vmovdqa (%r9), %xmm12
-; AVX2-ONLY-NEXT: vmovdqa (%r8), %xmm13
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} xmm14 = xmm13[1,1,1,1]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm14 = xmm14[0],xmm12[1],xmm14[2,3]
-; AVX2-ONLY-NEXT: vinserti128 $1, %xmm14, %ymm0, %ymm14
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm11 = ymm14[0,1,2,3,4,5],ymm11[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm8 = ymm8[0,1,2,3],ymm11[4,5,6,7]
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} xmm11 = xmm7[2],xmm6[2],xmm7[3],xmm6[3]
-; AVX2-ONLY-NEXT: vinserti128 $1, %xmm11, %ymm0, %ymm11
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} xmm14 = xmm13[2],xmm12[2],xmm13[3],xmm12[3]
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} xmm15 = xmm14[2,3,2,3]
-; AVX2-ONLY-NEXT: vinserti128 $1, %xmm15, %ymm0, %ymm15
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm11 = ymm15[0,1,2,3,4,5],ymm11[6,7]
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} xmm15 = xmm5[2],xmm4[2],xmm5[3],xmm4[3]
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} xmm0 = xmm10[2],xmm9[2],xmm10[3],xmm9[3]
-; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} xmm15 = xmm0[1],xmm15[1]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm11 = ymm15[0,1,2,3],ymm11[4,5,6,7]
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} xmm9 = xmm10[0],xmm9[0],xmm10[1],xmm9[1]
-; AVX2-ONLY-NEXT: vpbroadcastd %xmm4, %xmm10
-; AVX2-ONLY-NEXT: vpbroadcastd %xmm5, %xmm15
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} xmm10 = xmm15[0],xmm10[0],xmm15[1],xmm10[1]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm9 = xmm9[0,1],xmm10[2,3]
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} xmm10 = xmm13[0],xmm12[0],xmm13[1],xmm12[1]
-; AVX2-ONLY-NEXT: vpbroadcastd %xmm6, %xmm12
-; AVX2-ONLY-NEXT: vpbroadcastd %xmm7, %xmm13
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} xmm12 = xmm13[0],xmm12[0],xmm13[1],xmm12[1]
-; AVX2-ONLY-NEXT: vinserti128 $1, %xmm10, %ymm0, %ymm10
-; AVX2-ONLY-NEXT: vinserti128 $1, %xmm12, %ymm0, %ymm12
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm10 = ymm10[0,1,2,3,4,5],ymm12[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm9 = ymm9[0,1,2,3],ymm10[4,5,6,7]
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} xmm4 = xmm4[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm4 = xmm5[0,1,2],xmm4[3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm4[2,3]
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} xmm4 = xmm6[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm4 = xmm7[0,1,2],xmm4[3]
-; AVX2-ONLY-NEXT: vinserti128 $1, %xmm14, %ymm0, %ymm5
-; AVX2-ONLY-NEXT: vinserti128 $1, %xmm4, %ymm0, %ymm4
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm4 = ymm5[0,1,2,3,4,5],ymm4[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm4[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovaps (%rdi), %ymm3
+; AVX2-ONLY-NEXT: vmovaps (%rsi), %ymm4
+; AVX2-ONLY-NEXT: vmovaps (%rdx), %ymm1
+; AVX2-ONLY-NEXT: vmovaps (%rcx), %ymm2
+; AVX2-ONLY-NEXT: vmovaps (%r8), %ymm5
+; AVX2-ONLY-NEXT: vmovaps (%r9), %ymm6
+; AVX2-ONLY-NEXT: vmovaps (%r10), %ymm9
+; AVX2-ONLY-NEXT: vmovaps (%rax), %ymm10
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm7 = ymm9[0],ymm10[0],ymm9[1],ymm10[1],ymm9[4],ymm10[4],ymm9[5],ymm10[5]
+; AVX2-ONLY-NEXT: vbroadcastss 20(%r8), %ymm0
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm6[5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm7[6,7]
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm8 = ymm1[0],ymm2[0],ymm1[1],ymm2[1],ymm1[4],ymm2[4],ymm1[5],ymm2[5]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm8, %xmm11
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} ymm12 = ymm3[1,1,1,1,5,5,5,5]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm12 = ymm12[0],ymm4[1],ymm12[2,3,4],ymm4[5],ymm12[6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm12, %xmm12
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm11 = xmm12[0,1],xmm11[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm11[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 28(%r10), %ymm11
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm10 = ymm11[2],ymm10[2],ymm11[3],ymm10[3],ymm11[6],ymm10[6],ymm11[7],ymm10[7]
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm11 = ymm5[2],ymm6[2],ymm5[3],ymm6[3],ymm5[6],ymm6[6],ymm5[7],ymm6[7]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm10 = ymm11[1],ymm10[1],ymm11[3],ymm10[3]
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm2 = ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[6],ymm2[6],ymm1[7],ymm2[7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm1
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm12 = ymm3[2],ymm4[2],ymm3[3],ymm4[3],ymm3[6],ymm4[6],ymm3[7],ymm4[7]
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} ymm13 = ymm12[2,3,2,3,6,7,6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm13, %xmm13
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm13[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm10[4,5,6,7]
+; AVX2-ONLY-NEXT: vbroadcastss 24(%rax), %ymm10
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3,4,5,6],ymm10[7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm11[0,1,2,3,4,5],ymm9[6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm12, %xmm10
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm10[0,1],xmm2[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm9[4,5,6,7]
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm5 = ymm5[0],ymm6[0],ymm5[1],ymm6[1],ymm5[4],ymm6[4],ymm5[5],ymm6[5]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm5 = ymm5[0],ymm7[0],ymm5[2],ymm7[2]
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm3 = ymm3[0],ymm4[0],ymm3[1],ymm4[1],ymm3[4],ymm4[4],ymm3[5],ymm4[5]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm3, %xmm3
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm4 = ymm8[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm4[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm5[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovaps (%rcx), %xmm4
+; AVX2-ONLY-NEXT: vmovaps (%rdx), %xmm5
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm6 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
+; AVX2-ONLY-NEXT: vmovaps (%rsi), %xmm9
+; AVX2-ONLY-NEXT: vmovaps (%rdi), %xmm10
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} xmm7 = xmm10[1,1,1,1]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm7 = xmm7[0],xmm9[1],xmm7[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm8 = xmm7[0,1],xmm6[2,3]
+; AVX2-ONLY-NEXT: vmovaps (%rax), %xmm6
+; AVX2-ONLY-NEXT: vmovaps (%r10), %xmm7
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm11 = xmm7[0],xmm6[0],xmm7[1],xmm6[1]
+; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm11, %ymm0, %ymm11
+; AVX2-ONLY-NEXT: vmovaps (%r9), %xmm12
+; AVX2-ONLY-NEXT: vmovaps (%r8), %xmm13
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} xmm14 = xmm13[1,1,1,1]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm14 = xmm14[0],xmm12[1],xmm14[2,3]
+; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm14, %ymm0, %ymm14
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm14[0,1,2,3,4,5],ymm11[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],ymm11[4,5,6,7]
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm11 = xmm7[2],xmm6[2],xmm7[3],xmm6[3]
+; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm11, %ymm0, %ymm11
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm14 = xmm13[2],xmm12[2],xmm13[3],xmm12[3]
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} xmm15 = xmm14[2,3,2,3]
+; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm15, %ymm0, %ymm15
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm15[0,1,2,3,4,5],ymm11[6,7]
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm15 = xmm5[2],xmm4[2],xmm5[3],xmm4[3]
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm0 = xmm10[2],xmm9[2],xmm10[3],xmm9[3]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm15 = xmm0[1],xmm15[1]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm15[0,1,2,3],ymm11[4,5,6,7]
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm9 = xmm10[0],xmm9[0],xmm10[1],xmm9[1]
+; AVX2-ONLY-NEXT: vbroadcastss %xmm4, %xmm10
+; AVX2-ONLY-NEXT: vbroadcastss %xmm5, %xmm15
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm10 = xmm15[0],xmm10[0],xmm15[1],xmm10[1]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm9 = xmm9[0,1],xmm10[2,3]
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm10 = xmm13[0],xmm12[0],xmm13[1],xmm12[1]
+; AVX2-ONLY-NEXT: vbroadcastss %xmm6, %xmm12
+; AVX2-ONLY-NEXT: vbroadcastss %xmm7, %xmm13
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm12 = xmm13[0],xmm12[0],xmm13[1],xmm12[1]
+; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm10, %ymm0, %ymm10
+; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm12, %ymm0, %ymm12
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3,4,5],ymm12[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3],ymm10[4,5,6,7]
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} xmm4 = xmm4[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm5[0,1,2],xmm4[3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm4[2,3]
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} xmm4 = xmm6[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm7[0,1,2],xmm4[3]
+; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm14, %ymm0, %ymm5
+; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm0, %ymm4
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3,4,5],ymm4[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm4[4,5,6,7]
; AVX2-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX2-ONLY-NEXT: vmovdqa %ymm0, 64(%rax)
-; AVX2-ONLY-NEXT: vmovdqa %ymm9, (%rax)
-; AVX2-ONLY-NEXT: vmovdqa %ymm11, 96(%rax)
-; AVX2-ONLY-NEXT: vmovdqa %ymm8, 32(%rax)
-; AVX2-ONLY-NEXT: vmovdqa %ymm3, 128(%rax)
-; AVX2-ONLY-NEXT: vmovdqa %ymm2, 192(%rax)
-; AVX2-ONLY-NEXT: vmovdqa %ymm1, 224(%rax)
+; AVX2-ONLY-NEXT: vmovaps %ymm0, 64(%rax)
+; AVX2-ONLY-NEXT: vmovaps %ymm9, (%rax)
+; AVX2-ONLY-NEXT: vmovaps %ymm11, 96(%rax)
+; AVX2-ONLY-NEXT: vmovaps %ymm8, 32(%rax)
+; AVX2-ONLY-NEXT: vmovaps %ymm3, 128(%rax)
+; AVX2-ONLY-NEXT: vmovaps %ymm2, 192(%rax)
+; AVX2-ONLY-NEXT: vmovaps %ymm1, 224(%rax)
; AVX2-ONLY-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; AVX2-ONLY-NEXT: vmovaps %ymm0, 160(%rax)
; AVX2-ONLY-NEXT: vzeroupper
@@ -1304,108 +1301,102 @@ define void @store_i32_stride8_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: subq $296, %rsp # imm = 0x128
; AVX2-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX2-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %r10
-; AVX2-ONLY-NEXT: vmovdqa (%rdi), %ymm7
-; AVX2-ONLY-NEXT: vmovdqa 32(%rdi), %ymm1
-; AVX2-ONLY-NEXT: vmovdqa (%rsi), %ymm8
-; AVX2-ONLY-NEXT: vmovdqa 32(%rsi), %ymm2
-; AVX2-ONLY-NEXT: vmovdqa (%rdx), %ymm6
-; AVX2-ONLY-NEXT: vmovdqa 32(%rdx), %ymm5
-; AVX2-ONLY-NEXT: vmovdqa 32(%rcx), %ymm10
-; AVX2-ONLY-NEXT: vmovdqa 32(%r8), %ymm11
-; AVX2-ONLY-NEXT: vmovdqa 32(%r9), %ymm4
-; AVX2-ONLY-NEXT: vmovdqa 32(%r10), %ymm9
-; AVX2-ONLY-NEXT: vmovdqa 32(%rax), %ymm12
-; AVX2-ONLY-NEXT: vpbroadcastd 56(%rax), %ymm0
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm9[0,1,2,3,4,5,6],ymm0[7]
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm3 = ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[6],ymm2[6],ymm1[7],ymm2[7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm3, %xmm13
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm14 = ymm5[2],ymm10[2],ymm5[3],ymm10[3],ymm5[6],ymm10[6],ymm5[7],ymm10[7]
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm15 = ymm14[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm13 = xmm13[0,1],xmm15[2,3]
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm15 = ymm11[2],ymm4[2],ymm11[3],ymm4[3],ymm11[6],ymm4[6],ymm11[7],ymm4[7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm15[0,1,2,3,4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm13[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 60(%r10), %ymm0
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm0 = ymm0[2],ymm12[2],ymm0[3],ymm12[3],ymm0[6],ymm12[6],ymm0[7],ymm12[7]
-; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm13 = ymm15[1],ymm0[1],ymm15[3],ymm0[3]
-; AVX2-ONLY-NEXT: vmovdqa (%rcx), %ymm0
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm14, %xmm14
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm3 = ymm3[2,3,2,3,6,7,6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm3, %xmm3
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm14 = xmm3[0,1],xmm14[2,3]
-; AVX2-ONLY-NEXT: vmovdqa (%r8), %ymm3
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm14[0,1,2,3],ymm13[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm12 = ymm9[0],ymm12[0],ymm9[1],ymm12[1],ymm9[4],ymm12[4],ymm9[5],ymm12[5]
-; AVX2-ONLY-NEXT: vpbroadcastd 48(%rsi), %xmm9
-; AVX2-ONLY-NEXT: vpbroadcastd 48(%rdi), %ymm13
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} xmm13 = xmm13[0],xmm9[0],xmm13[1],xmm9[1]
-; AVX2-ONLY-NEXT: vmovdqa (%r9), %ymm9
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm10 = ymm5[0],ymm10[0],ymm5[1],ymm10[1],ymm5[4],ymm10[4],ymm5[5],ymm10[5]
-; AVX2-ONLY-NEXT: vmovdqa (%r10), %ymm5
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm11 = ymm11[0],ymm4[0],ymm11[1],ymm4[1],ymm11[4],ymm4[4],ymm11[5],ymm4[5]
-; AVX2-ONLY-NEXT: vpunpcklqdq {{.*#+}} ymm11 = ymm11[0],ymm12[0],ymm11[2],ymm12[2]
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm10 = ymm10[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm10 = xmm13[0,1],xmm10[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm10 = ymm10[0,1,2,3],ymm11[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 52(%r8), %ymm10
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm4 = ymm10[0,1,2,3,4],ymm4[5],ymm10[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm12[6,7]
-; AVX2-ONLY-NEXT: vpmovzxdq {{.*#+}} xmm10 = mem[0],zero,mem[1],zero
-; AVX2-ONLY-NEXT: vpbroadcastd 52(%rcx), %ymm11
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm10 = xmm10[0,1,2],xmm11[3]
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[1,1,1,1,5,5,5,5]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0],ymm2[1],ymm1[2,3,4],ymm2[5],ymm1[6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm1, %xmm1
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm10[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm4[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 24(%rax), %ymm1
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm5[0,1,2,3,4,5,6],ymm1[7]
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm2 = ymm7[2],ymm8[2],ymm7[3],ymm8[3],ymm7[6],ymm8[6],ymm7[7],ymm8[7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm2, %xmm4
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm10 = ymm6[2],ymm0[2],ymm6[3],ymm0[3],ymm6[6],ymm0[6],ymm6[7],ymm0[7]
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm11 = ymm10[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm4 = xmm4[0,1],xmm11[2,3]
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm11 = ymm3[2],ymm9[2],ymm3[3],ymm9[3],ymm3[6],ymm9[6],ymm3[7],ymm9[7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm11[0,1,2,3,4,5],ymm1[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm4[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vmovdqa (%rax), %ymm1
-; AVX2-ONLY-NEXT: vpbroadcastd 28(%r10), %ymm4
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm4 = ymm4[2],ymm1[2],ymm4[3],ymm1[3],ymm4[6],ymm1[6],ymm4[7],ymm1[7]
-; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm4 = ymm11[1],ymm4[1],ymm11[3],ymm4[3]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm10, %xmm10
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm2 = ymm2[2,3,2,3,6,7,6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm2, %xmm2
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm2 = xmm2[0,1],xmm10[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm4[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm1 = ymm5[0],ymm1[0],ymm5[1],ymm1[1],ymm5[4],ymm1[4],ymm5[5],ymm1[5]
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm2 = ymm3[0],ymm9[0],ymm3[1],ymm9[1],ymm3[4],ymm9[4],ymm3[5],ymm9[5]
-; AVX2-ONLY-NEXT: vpunpcklqdq {{.*#+}} ymm2 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
-; AVX2-ONLY-NEXT: vpbroadcastd 16(%rsi), %xmm3
-; AVX2-ONLY-NEXT: vpbroadcastd 16(%rdi), %ymm4
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm0 = ymm6[0],ymm0[0],ymm6[1],ymm0[1],ymm6[4],ymm0[4],ymm6[5],ymm0[5]
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm0 = ymm0[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm0 = xmm3[0,1],xmm0[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 20(%r8), %ymm0
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm9[5],ymm0[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX2-ONLY-NEXT: vpmovzxdq {{.*#+}} xmm1 = mem[0],zero,mem[1],zero
-; AVX2-ONLY-NEXT: vpbroadcastd 20(%rcx), %ymm2
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0,1,2],xmm2[3]
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm2 = ymm7[1,1,1,1,5,5,5,5]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0],ymm8[1],ymm2[2,3,4],ymm8[5],ymm2[6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm2, %xmm2
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps (%rdi), %ymm7
+; AVX2-ONLY-NEXT: vmovaps 32(%rdi), %ymm2
+; AVX2-ONLY-NEXT: vmovaps (%rsi), %ymm8
+; AVX2-ONLY-NEXT: vmovaps 32(%rsi), %ymm3
+; AVX2-ONLY-NEXT: vmovaps (%rdx), %ymm6
+; AVX2-ONLY-NEXT: vmovaps 32(%rdx), %ymm5
+; AVX2-ONLY-NEXT: vmovaps 32(%rcx), %ymm11
+; AVX2-ONLY-NEXT: vmovaps 32(%r8), %ymm10
+; AVX2-ONLY-NEXT: vmovaps 32(%r9), %ymm4
+; AVX2-ONLY-NEXT: vmovaps 32(%r10), %ymm12
+; AVX2-ONLY-NEXT: vmovaps 32(%rax), %ymm13
+; AVX2-ONLY-NEXT: vbroadcastss 56(%rax), %ymm0
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm12[0,1,2,3,4,5,6],ymm0[7]
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm9 = ymm2[2],ymm3[2],ymm2[3],ymm3[3],ymm2[6],ymm3[6],ymm2[7],ymm3[7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm9, %xmm1
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm14 = ymm5[2],ymm11[2],ymm5[3],ymm11[3],ymm5[6],ymm11[6],ymm5[7],ymm11[7]
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm15 = ymm14[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm15[2,3]
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm15 = ymm10[2],ymm4[2],ymm10[3],ymm4[3],ymm10[6],ymm4[6],ymm10[7],ymm4[7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 60(%r10), %ymm0
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm0 = ymm0[2],ymm13[2],ymm0[3],ymm13[3],ymm0[6],ymm13[6],ymm0[7],ymm13[7]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm15[1],ymm0[1],ymm15[3],ymm0[3]
+; AVX2-ONLY-NEXT: vmovaps (%rcx), %ymm1
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm14, %xmm14
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} ymm9 = ymm9[2,3,2,3,6,7,6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm9, %xmm9
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm14 = xmm9[0,1],xmm14[2,3]
+; AVX2-ONLY-NEXT: vmovaps (%r8), %ymm9
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm14[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps (%r9), %ymm0
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm12 = ymm12[0],ymm13[0],ymm12[1],ymm13[1],ymm12[4],ymm13[4],ymm12[5],ymm13[5]
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm13 = ymm2[0],ymm3[0],ymm2[1],ymm3[1],ymm2[4],ymm3[4],ymm2[5],ymm3[5]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm13, %xmm13
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm11 = ymm5[0],ymm11[0],ymm5[1],ymm11[1],ymm5[4],ymm11[4],ymm5[5],ymm11[5]
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm5 = ymm11[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm13 = xmm13[0,1],xmm5[2,3]
+; AVX2-ONLY-NEXT: vmovaps (%r10), %ymm5
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm10 = ymm10[0],ymm4[0],ymm10[1],ymm4[1],ymm10[4],ymm4[4],ymm10[5],ymm4[5]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm10 = ymm10[0],ymm12[0],ymm10[2],ymm12[2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm13[0,1,2,3],ymm10[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 52(%r8), %ymm10
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm10[0,1,2,3,4],ymm4[5],ymm10[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm12[6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm11, %xmm10
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm2[1,1,1,1,5,5,5,5]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm3[1],ymm2[2,3,4],ymm3[5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm2
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0,1],xmm10[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm4[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 24(%rax), %ymm2
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm5[0,1,2,3,4,5,6],ymm2[7]
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm3 = ymm7[2],ymm8[2],ymm7[3],ymm8[3],ymm7[6],ymm8[6],ymm7[7],ymm8[7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm3, %xmm4
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm10 = ymm6[2],ymm1[2],ymm6[3],ymm1[3],ymm6[6],ymm1[6],ymm6[7],ymm1[7]
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm11 = ymm10[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm4[0,1],xmm11[2,3]
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm11 = ymm9[2],ymm0[2],ymm9[3],ymm0[3],ymm9[6],ymm0[6],ymm9[7],ymm0[7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm11[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm4[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps (%rax), %ymm2
+; AVX2-ONLY-NEXT: vbroadcastss 28(%r10), %ymm4
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm4 = ymm4[2],ymm2[2],ymm4[3],ymm2[3],ymm4[6],ymm2[6],ymm4[7],ymm2[7]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm4 = ymm11[1],ymm4[1],ymm11[3],ymm4[3]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm10, %xmm10
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm3[2,3,2,3,6,7,6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm3, %xmm3
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm10[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm2 = ymm5[0],ymm2[0],ymm5[1],ymm2[1],ymm5[4],ymm2[4],ymm5[5],ymm2[5]
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm3 = ymm9[0],ymm0[0],ymm9[1],ymm0[1],ymm9[4],ymm0[4],ymm9[5],ymm0[5]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm3 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm4 = ymm7[0],ymm8[0],ymm7[1],ymm8[1],ymm7[4],ymm8[4],ymm7[5],ymm8[5]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm4, %xmm4
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm1 = ymm6[0],ymm1[0],ymm6[1],ymm1[1],ymm6[4],ymm1[4],ymm6[5],ymm1[5]
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm5 = ymm1[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm4[0,1],xmm5[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 20(%r8), %ymm3
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4],ymm0[5],ymm3[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm7[1,1,1,1,5,5,5,5]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm8[1],ymm2[2,3,4],ymm8[5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm2
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps (%rcx), %xmm1
; AVX2-ONLY-NEXT: vmovaps (%rdx), %xmm15
; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm0 = xmm15[0],xmm1[0],xmm15[1],xmm1[1]
@@ -2784,210 +2775,198 @@ define void @store_i32_stride8_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: subq $648, %rsp # imm = 0x288
; AVX2-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %r10
; AVX2-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; AVX2-ONLY-NEXT: vmovdqa 32(%rdi), %ymm0
-; AVX2-ONLY-NEXT: vmovdqa (%rdi), %ymm4
-; AVX2-ONLY-NEXT: vmovdqa 32(%rsi), %ymm1
-; AVX2-ONLY-NEXT: vmovdqa (%rsi), %ymm5
-; AVX2-ONLY-NEXT: vmovdqa 32(%rdx), %ymm2
-; AVX2-ONLY-NEXT: vmovdqa (%rdx), %ymm6
-; AVX2-ONLY-NEXT: vmovdqa 32(%rcx), %ymm3
-; AVX2-ONLY-NEXT: vmovdqa (%rcx), %ymm8
-; AVX2-ONLY-NEXT: vmovdqa (%r8), %ymm10
-; AVX2-ONLY-NEXT: vmovdqa (%r9), %ymm12
-; AVX2-ONLY-NEXT: vmovdqa (%rax), %ymm9
-; AVX2-ONLY-NEXT: vmovdqa (%r10), %ymm11
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm7 = ymm9[0],ymm11[0],ymm9[1],ymm11[1],ymm9[4],ymm11[4],ymm9[5],ymm11[5]
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm13 = ymm10[0],ymm12[0],ymm10[1],ymm12[1],ymm10[4],ymm12[4],ymm10[5],ymm12[5]
-; AVX2-ONLY-NEXT: vpunpcklqdq {{.*#+}} ymm13 = ymm13[0],ymm7[0],ymm13[2],ymm7[2]
-; AVX2-ONLY-NEXT: vpbroadcastd 16(%rsi), %xmm14
-; AVX2-ONLY-NEXT: vpbroadcastd 16(%rdi), %ymm15
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} xmm14 = xmm15[0],xmm14[0],xmm15[1],xmm14[1]
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm15 = ymm6[0],ymm8[0],ymm6[1],ymm8[1],ymm6[4],ymm8[4],ymm6[5],ymm8[5]
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm15 = ymm15[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm14 = xmm14[0,1],xmm15[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm14[0,1,2,3],ymm13[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 20(%r8), %ymm13
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm13[0,1,2,3,4],ymm12[5],ymm13[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5],ymm7[6,7]
-; AVX2-ONLY-NEXT: vpmovzxdq {{.*#+}} xmm7 = mem[0],zero,mem[1],zero
-; AVX2-ONLY-NEXT: vpbroadcastd 20(%rcx), %ymm14
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm7 = xmm7[0,1,2],xmm14[3]
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm14 = ymm4[1,1,1,1,5,5,5,5]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm14 = ymm14[0],ymm5[1],ymm14[2,3,4],ymm5[5],ymm14[6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm14, %xmm14
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm14 = xmm14[0,1],xmm7[2,3]
-; AVX2-ONLY-NEXT: vmovdqa 32(%r8), %ymm7
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm14[0,1,2,3],ymm13[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 24(%r10), %ymm13
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm9[0,1,2,3,4,5,6],ymm13[7]
-; AVX2-ONLY-NEXT: vmovdqa 32(%r9), %ymm9
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm12 = ymm10[2],ymm12[2],ymm10[3],ymm12[3],ymm10[6],ymm12[6],ymm10[7],ymm12[7]
-; AVX2-ONLY-NEXT: vmovdqa 32(%rax), %ymm10
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm4 = ymm4[2],ymm5[2],ymm4[3],ymm5[3],ymm4[6],ymm5[6],ymm4[7],ymm5[7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm4, %xmm5
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm8 = ymm6[2],ymm8[2],ymm6[3],ymm8[3],ymm6[6],ymm8[6],ymm6[7],ymm8[7]
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm6 = ymm8[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm5 = xmm5[0,1],xmm6[2,3]
-; AVX2-ONLY-NEXT: vmovdqa 32(%r10), %ymm6
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm12[0,1,2,3,4,5],ymm13[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm13[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 28(%rax), %ymm5
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm5 = ymm5[2],ymm11[2],ymm5[3],ymm11[3],ymm5[6],ymm11[6],ymm5[7],ymm11[7]
-; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm5 = ymm12[1],ymm5[1],ymm12[3],ymm5[3]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm8, %xmm8
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm4 = ymm4[2,3,2,3,6,7,6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm4, %xmm4
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm4 = xmm4[0,1],xmm8[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm4 = ymm10[0],ymm6[0],ymm10[1],ymm6[1],ymm10[4],ymm6[4],ymm10[5],ymm6[5]
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm5 = ymm7[0],ymm9[0],ymm7[1],ymm9[1],ymm7[4],ymm9[4],ymm7[5],ymm9[5]
-; AVX2-ONLY-NEXT: vpunpcklqdq {{.*#+}} ymm5 = ymm5[0],ymm4[0],ymm5[2],ymm4[2]
-; AVX2-ONLY-NEXT: vpbroadcastd 48(%rsi), %xmm8
-; AVX2-ONLY-NEXT: vpbroadcastd 48(%rdi), %ymm11
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} xmm8 = xmm11[0],xmm8[0],xmm11[1],xmm8[1]
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm11 = ymm2[0],ymm3[0],ymm2[1],ymm3[1],ymm2[4],ymm3[4],ymm2[5],ymm3[5]
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm11 = ymm11[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm8 = xmm8[0,1],xmm11[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm5 = ymm8[0,1,2,3],ymm5[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 52(%r8), %ymm5
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3,4],ymm9[5],ymm5[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],ymm4[6,7]
-; AVX2-ONLY-NEXT: vpmovzxdq {{.*#+}} xmm4 = mem[0],zero,mem[1],zero
-; AVX2-ONLY-NEXT: vpbroadcastd 52(%rcx), %ymm8
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm4 = xmm4[0,1,2],xmm8[3]
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm8 = ymm0[1,1,1,1,5,5,5,5]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm8 = ymm8[0],ymm1[1],ymm8[2,3,4],ymm1[5],ymm8[6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm8, %xmm8
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm8 = xmm8[0,1],xmm4[2,3]
-; AVX2-ONLY-NEXT: vmovdqa 64(%rdx), %ymm4
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm5 = ymm8[0,1,2,3],ymm5[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 56(%r10), %ymm5
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm8 = ymm10[0,1,2,3,4,5,6],ymm5[7]
-; AVX2-ONLY-NEXT: vmovdqa 64(%rcx), %ymm5
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm9 = ymm7[2],ymm9[2],ymm7[3],ymm9[3],ymm7[6],ymm9[6],ymm7[7],ymm9[7]
-; AVX2-ONLY-NEXT: vmovdqa 64(%r8), %ymm7
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm10 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm10, %xmm0
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm3 = ymm2[2],ymm3[2],ymm2[3],ymm3[3],ymm2[6],ymm3[6],ymm2[7],ymm3[7]
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm1 = ymm3[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
-; AVX2-ONLY-NEXT: vmovdqa 64(%r9), %ymm1
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm9[0,1,2,3,4,5],ymm8[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 60(%rax), %ymm0
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm0 = ymm0[2],ymm6[2],ymm0[3],ymm6[3],ymm0[6],ymm6[6],ymm0[7],ymm6[7]
-; AVX2-ONLY-NEXT: vmovdqa 64(%rax), %ymm2
-; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm6 = ymm9[1],ymm0[1],ymm9[3],ymm0[3]
-; AVX2-ONLY-NEXT: vmovdqa 64(%r10), %ymm0
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm3, %xmm3
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm8 = ymm10[2,3,2,3,6,7,6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm8, %xmm8
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm3 = xmm8[0,1],xmm3[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm6[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 80(%rsi), %xmm3
-; AVX2-ONLY-NEXT: vpbroadcastd 80(%rdi), %ymm6
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm6[0],xmm3[0],xmm6[1],xmm3[1]
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm6 = ymm4[0],ymm5[0],ymm4[1],ymm5[1],ymm4[4],ymm5[4],ymm4[5],ymm5[5]
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm6 = ymm6[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm3 = xmm3[0,1],xmm6[2,3]
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm6 = ymm2[0],ymm0[0],ymm2[1],ymm0[1],ymm2[4],ymm0[4],ymm2[5],ymm0[5]
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm8 = ymm7[0],ymm1[0],ymm7[1],ymm1[1],ymm7[4],ymm1[4],ymm7[5],ymm1[5]
-; AVX2-ONLY-NEXT: vpunpcklqdq {{.*#+}} ymm8 = ymm8[0],ymm6[0],ymm8[2],ymm6[2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm8[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 84(%r8), %ymm3
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm1[5],ymm3[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm6[6,7]
-; AVX2-ONLY-NEXT: vpmovzxdq {{.*#+}} xmm6 = mem[0],zero,mem[1],zero
-; AVX2-ONLY-NEXT: vpbroadcastd 84(%rcx), %ymm8
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm6 = xmm6[0,1,2],xmm8[3]
-; AVX2-ONLY-NEXT: vmovdqa 64(%rdi), %ymm8
-; AVX2-ONLY-NEXT: vmovdqa 64(%rsi), %ymm9
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm10 = ymm8[1,1,1,1,5,5,5,5]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm10 = ymm10[0],ymm9[1],ymm10[2,3,4],ymm9[5],ymm10[6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm10, %xmm10
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm6 = xmm10[0,1],xmm6[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm6[0,1,2,3],ymm3[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 88(%r10), %ymm3
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm1 = ymm7[2],ymm1[2],ymm7[3],ymm1[3],ymm7[6],ymm1[6],ymm7[7],ymm1[7]
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm3 = ymm8[2],ymm9[2],ymm8[3],ymm9[3],ymm8[6],ymm9[6],ymm8[7],ymm9[7]
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm4 = ymm4[2],ymm5[2],ymm4[3],ymm5[3],ymm4[6],ymm5[6],ymm4[7],ymm5[7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm3, %xmm5
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm6 = ymm4[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm5 = xmm5[0,1],xmm6[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm1[0,1,2,3,4,5],ymm2[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm5[0,1,2,3],ymm2[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 92(%rax), %ymm2
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm2 = ymm2[2],ymm0[2],ymm2[3],ymm0[3],ymm2[6],ymm0[6],ymm2[7],ymm0[7]
-; AVX2-ONLY-NEXT: vmovdqa 96(%rdx), %ymm0
-; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm2 = ymm1[1],ymm2[1],ymm1[3],ymm2[3]
-; AVX2-ONLY-NEXT: vmovdqa 96(%rcx), %ymm1
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm4, %xmm4
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm3 = ymm3[2,3,2,3,6,7,6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm3, %xmm3
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm3 = xmm3[0,1],xmm4[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 112(%rsi), %xmm2
-; AVX2-ONLY-NEXT: vpbroadcastd 112(%rdi), %ymm3
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm3 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm3 = ymm3[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm2 = xmm2[0,1],xmm3[2,3]
-; AVX2-ONLY-NEXT: vmovdqa 96(%r8), %ymm3
-; AVX2-ONLY-NEXT: vmovdqa 96(%r9), %ymm4
-; AVX2-ONLY-NEXT: vmovdqa 96(%rax), %ymm5
-; AVX2-ONLY-NEXT: vmovdqa 96(%r10), %ymm6
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm7 = ymm5[0],ymm6[0],ymm5[1],ymm6[1],ymm5[4],ymm6[4],ymm5[5],ymm6[5]
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm8 = ymm3[0],ymm4[0],ymm3[1],ymm4[1],ymm3[4],ymm4[4],ymm3[5],ymm4[5]
-; AVX2-ONLY-NEXT: vpunpcklqdq {{.*#+}} ymm8 = ymm8[0],ymm7[0],ymm8[2],ymm7[2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm8[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 116(%r8), %ymm2
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm4[5],ymm2[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm7[6,7]
-; AVX2-ONLY-NEXT: vpmovzxdq {{.*#+}} xmm7 = mem[0],zero,mem[1],zero
-; AVX2-ONLY-NEXT: vpbroadcastd 116(%rcx), %ymm8
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm7 = xmm7[0,1,2],xmm8[3]
-; AVX2-ONLY-NEXT: vmovdqa 96(%rdi), %ymm8
-; AVX2-ONLY-NEXT: vmovdqa 96(%rsi), %ymm9
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm10 = ymm8[1,1,1,1,5,5,5,5]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm10 = ymm10[0],ymm9[1],ymm10[2,3,4],ymm9[5],ymm10[6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm10, %xmm10
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm7 = xmm10[0,1],xmm7[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm7[0,1,2,3],ymm2[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 120(%r10), %ymm2
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm5[0,1,2,3,4,5,6],ymm2[7]
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm3 = ymm3[2],ymm4[2],ymm3[3],ymm4[3],ymm3[6],ymm4[6],ymm3[7],ymm4[7]
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm4 = ymm8[2],ymm9[2],ymm8[3],ymm9[3],ymm8[6],ymm9[6],ymm8[7],ymm9[7]
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm0 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm4, %xmm1
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm5 = ymm0[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm5[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 124(%rax), %ymm1
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm1 = ymm1[2],ymm6[2],ymm1[3],ymm6[3],ymm1[6],ymm6[6],ymm1[7],ymm6[7]
-; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm1 = ymm3[1],ymm1[1],ymm3[3],ymm1[3]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm0, %xmm0
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm2 = ymm4[2,3,2,3,6,7,6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm2, %xmm2
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm0 = xmm2[0,1],xmm0[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 32(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovaps (%rdi), %ymm4
+; AVX2-ONLY-NEXT: vmovaps 32(%rsi), %ymm1
+; AVX2-ONLY-NEXT: vmovaps (%rsi), %ymm5
+; AVX2-ONLY-NEXT: vmovaps 32(%rdx), %ymm2
+; AVX2-ONLY-NEXT: vmovaps (%rdx), %ymm6
+; AVX2-ONLY-NEXT: vmovaps 32(%rcx), %ymm3
+; AVX2-ONLY-NEXT: vmovaps (%rcx), %ymm8
+; AVX2-ONLY-NEXT: vmovaps (%r8), %ymm10
+; AVX2-ONLY-NEXT: vmovaps (%r9), %ymm12
+; AVX2-ONLY-NEXT: vmovaps (%rax), %ymm9
+; AVX2-ONLY-NEXT: vmovaps (%r10), %ymm11
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm7 = ymm9[0],ymm11[0],ymm9[1],ymm11[1],ymm9[4],ymm11[4],ymm9[5],ymm11[5]
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm13 = ymm4[0],ymm5[0],ymm4[1],ymm5[1],ymm4[4],ymm5[4],ymm4[5],ymm5[5]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm13, %xmm13
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm14 = ymm6[0],ymm8[0],ymm6[1],ymm8[1],ymm6[4],ymm8[4],ymm6[5],ymm8[5]
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm15 = ymm14[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm13 = xmm13[0,1],xmm15[2,3]
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm15 = ymm10[0],ymm12[0],ymm10[1],ymm12[1],ymm10[4],ymm12[4],ymm10[5],ymm12[5]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm15 = ymm15[0],ymm7[0],ymm15[2],ymm7[2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3],ymm15[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 20(%r8), %ymm13
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4],ymm12[5],ymm13[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5],ymm7[6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm14, %xmm7
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm4[1,1,1,1,5,5,5,5]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0],ymm5[1],ymm14[2,3,4],ymm5[5],ymm14[6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm14, %xmm14
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm14 = xmm14[0,1],xmm7[2,3]
+; AVX2-ONLY-NEXT: vmovaps 32(%r8), %ymm7
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm14[0,1,2,3],ymm13[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 24(%r10), %ymm13
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm9[0,1,2,3,4,5,6],ymm13[7]
+; AVX2-ONLY-NEXT: vmovaps 32(%r9), %ymm9
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm12 = ymm10[2],ymm12[2],ymm10[3],ymm12[3],ymm10[6],ymm12[6],ymm10[7],ymm12[7]
+; AVX2-ONLY-NEXT: vmovaps 32(%rax), %ymm10
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm4 = ymm4[2],ymm5[2],ymm4[3],ymm5[3],ymm4[6],ymm5[6],ymm4[7],ymm5[7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm4, %xmm5
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm6 = ymm6[2],ymm8[2],ymm6[3],ymm8[3],ymm6[6],ymm8[6],ymm6[7],ymm8[7]
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm8 = ymm6[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm8 = xmm5[0,1],xmm8[2,3]
+; AVX2-ONLY-NEXT: vmovaps 32(%r10), %ymm5
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm12[0,1,2,3,4,5],ymm13[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],ymm13[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 28(%rax), %ymm8
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm8 = ymm8[2],ymm11[2],ymm8[3],ymm11[3],ymm8[6],ymm11[6],ymm8[7],ymm11[7]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm8 = ymm12[1],ymm8[1],ymm12[3],ymm8[3]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm6, %xmm6
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm4[2,3,2,3,6,7,6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm4, %xmm4
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm4[0,1],xmm6[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm8[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm4 = ymm10[0],ymm5[0],ymm10[1],ymm5[1],ymm10[4],ymm5[4],ymm10[5],ymm5[5]
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm6 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm6, %xmm6
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm8 = ymm2[0],ymm3[0],ymm2[1],ymm3[1],ymm2[4],ymm3[4],ymm2[5],ymm3[5]
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm11 = ymm8[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0,1],xmm11[2,3]
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm11 = ymm7[0],ymm9[0],ymm7[1],ymm9[1],ymm7[4],ymm9[4],ymm7[5],ymm9[5]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm11 = ymm11[0],ymm4[0],ymm11[2],ymm4[2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm11[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 52(%r8), %ymm6
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4],ymm9[5],ymm6[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm4[6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm8, %xmm4
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} ymm8 = ymm0[1,1,1,1,5,5,5,5]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0],ymm1[1],ymm8[2,3,4],ymm1[5],ymm8[6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm8, %xmm8
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm8 = xmm8[0,1],xmm4[2,3]
+; AVX2-ONLY-NEXT: vmovaps 64(%rdi), %ymm4
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm8[0,1,2,3],ymm6[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 56(%r10), %ymm6
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3,4,5,6],ymm6[7]
+; AVX2-ONLY-NEXT: vmovaps 64(%rsi), %ymm8
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm7 = ymm7[2],ymm9[2],ymm7[3],ymm9[3],ymm7[6],ymm9[6],ymm7[7],ymm9[7]
+; AVX2-ONLY-NEXT: vmovaps 64(%rdx), %ymm6
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm1 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm0
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm2 = ymm2[2],ymm3[2],ymm2[3],ymm3[3],ymm2[6],ymm3[6],ymm2[7],ymm3[7]
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm3 = ymm2[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm0[0,1],xmm3[2,3]
+; AVX2-ONLY-NEXT: vmovaps 64(%rcx), %ymm0
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm7[0,1,2,3,4,5],ymm10[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm9[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 60(%rax), %ymm3
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm3 = ymm3[2],ymm5[2],ymm3[3],ymm5[3],ymm3[6],ymm5[6],ymm3[7],ymm5[7]
+; AVX2-ONLY-NEXT: vmovaps 64(%r8), %ymm5
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm3 = ymm7[1],ymm3[1],ymm7[3],ymm3[3]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm2
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[2,3,2,3,6,7,6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm1 = ymm4[0],ymm8[0],ymm4[1],ymm8[1],ymm4[4],ymm8[4],ymm4[5],ymm8[5]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm2 = ymm6[0],ymm0[0],ymm6[1],ymm0[1],ymm6[4],ymm0[4],ymm6[5],ymm0[5]
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm3 = ymm2[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm3[2,3]
+; AVX2-ONLY-NEXT: vmovaps 64(%r9), %ymm3
+; AVX2-ONLY-NEXT: vmovaps 64(%rax), %ymm7
+; AVX2-ONLY-NEXT: vmovaps 64(%r10), %ymm9
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm10 = ymm7[0],ymm9[0],ymm7[1],ymm9[1],ymm7[4],ymm9[4],ymm7[5],ymm9[5]
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm11 = ymm5[0],ymm3[0],ymm5[1],ymm3[1],ymm5[4],ymm3[4],ymm5[5],ymm3[5]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm11 = ymm11[0],ymm10[0],ymm11[2],ymm10[2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm11[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 84(%r8), %ymm1
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm3[5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm10[6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm2
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm4[1,1,1,1,5,5,5,5]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0],ymm8[1],ymm10[2,3,4],ymm8[5],ymm10[6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm10, %xmm10
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm10[0,1],xmm2[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 88(%r10), %ymm1
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm7[0,1,2,3,4,5,6],ymm1[7]
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm2 = ymm5[2],ymm3[2],ymm5[3],ymm3[3],ymm5[6],ymm3[6],ymm5[7],ymm3[7]
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm3 = ymm4[2],ymm8[2],ymm4[3],ymm8[3],ymm4[6],ymm8[6],ymm4[7],ymm8[7]
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm0 = ymm6[2],ymm0[2],ymm6[3],ymm0[3],ymm6[6],ymm0[6],ymm6[7],ymm0[7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm3, %xmm4
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm5 = ymm0[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm4[0,1],xmm5[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm4[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 92(%rax), %ymm1
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm1 = ymm1[2],ymm9[2],ymm1[3],ymm9[3],ymm1[6],ymm9[6],ymm1[7],ymm9[7]
+; AVX2-ONLY-NEXT: vmovaps 96(%rdi), %ymm4
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm2[1],ymm1[1],ymm2[3],ymm1[3]
+; AVX2-ONLY-NEXT: vmovaps 96(%rsi), %ymm2
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm0, %xmm0
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm3[2,3,2,3,6,7,6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm3, %xmm3
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm3[0,1],xmm0[2,3]
+; AVX2-ONLY-NEXT: vmovaps 96(%rdx), %ymm3
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm0 = ymm4[0],ymm2[0],ymm4[1],ymm2[1],ymm4[4],ymm2[4],ymm4[5],ymm2[5]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm0, %xmm0
+; AVX2-ONLY-NEXT: vmovaps 96(%rcx), %ymm1
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm5 = ymm3[0],ymm1[0],ymm3[1],ymm1[1],ymm3[4],ymm1[4],ymm3[5],ymm1[5]
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm6 = ymm5[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm6[2,3]
+; AVX2-ONLY-NEXT: vmovaps 96(%r8), %ymm6
+; AVX2-ONLY-NEXT: vmovaps 96(%r9), %ymm7
+; AVX2-ONLY-NEXT: vmovaps 96(%rax), %ymm8
+; AVX2-ONLY-NEXT: vmovaps 96(%r10), %ymm9
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm10 = ymm8[0],ymm9[0],ymm8[1],ymm9[1],ymm8[4],ymm9[4],ymm8[5],ymm9[5]
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm11 = ymm6[0],ymm7[0],ymm6[1],ymm7[1],ymm6[4],ymm7[4],ymm6[5],ymm7[5]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm11 = ymm11[0],ymm10[0],ymm11[2],ymm10[2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm11[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 116(%r8), %ymm0
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm7[5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm10[6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm5, %xmm5
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm4[1,1,1,1,5,5,5,5]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0],ymm2[1],ymm10[2,3,4],ymm2[5],ymm10[6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm10, %xmm10
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm10[0,1],xmm5[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 120(%r10), %ymm0
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3,4,5,6],ymm0[7]
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm5 = ymm6[2],ymm7[2],ymm6[3],ymm7[3],ymm6[6],ymm7[6],ymm6[7],ymm7[7]
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm2 = ymm4[2],ymm2[2],ymm4[3],ymm2[3],ymm4[6],ymm2[6],ymm4[7],ymm2[7]
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm1 = ymm3[2],ymm1[2],ymm3[3],ymm1[3],ymm3[6],ymm1[6],ymm3[7],ymm1[7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm3
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm4 = ymm1[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm4[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 124(%rax), %ymm0
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm0 = ymm0[2],ymm9[2],ymm0[3],ymm9[3],ymm0[6],ymm9[6],ymm0[7],ymm9[7]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm5[1],ymm0[1],ymm5[3],ymm0[3]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm2[2,3,2,3,6,7,6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm2
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps (%rcx), %xmm0
; AVX2-ONLY-NEXT: vbroadcastss %xmm0, %xmm2
; AVX2-ONLY-NEXT: vmovaps (%rdx), %xmm1
@@ -5826,427 +5805,403 @@ define void @store_i32_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: subq $1672, %rsp # imm = 0x688
; AVX2-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %rax
; AVX2-ONLY-NEXT: movq {{[0-9]+}}(%rsp), %r10
-; AVX2-ONLY-NEXT: vmovdqa 32(%rdi), %ymm0
-; AVX2-ONLY-NEXT: vmovdqa (%rdi), %ymm4
-; AVX2-ONLY-NEXT: vmovdqa 32(%rsi), %ymm1
-; AVX2-ONLY-NEXT: vmovdqa (%rsi), %ymm5
-; AVX2-ONLY-NEXT: vmovdqa 32(%rdx), %ymm2
-; AVX2-ONLY-NEXT: vmovdqa (%rdx), %ymm6
-; AVX2-ONLY-NEXT: vmovdqa 32(%rcx), %ymm3
-; AVX2-ONLY-NEXT: vmovdqa (%rcx), %ymm7
-; AVX2-ONLY-NEXT: vmovdqa (%r8), %ymm8
-; AVX2-ONLY-NEXT: vmovdqa (%r9), %ymm11
-; AVX2-ONLY-NEXT: vmovdqa (%r10), %ymm10
-; AVX2-ONLY-NEXT: vmovdqa (%rax), %ymm12
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm9 = ymm10[0],ymm12[0],ymm10[1],ymm12[1],ymm10[4],ymm12[4],ymm10[5],ymm12[5]
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm13 = ymm8[0],ymm11[0],ymm8[1],ymm11[1],ymm8[4],ymm11[4],ymm8[5],ymm11[5]
-; AVX2-ONLY-NEXT: vpunpcklqdq {{.*#+}} ymm13 = ymm13[0],ymm9[0],ymm13[2],ymm9[2]
-; AVX2-ONLY-NEXT: vpbroadcastd 16(%rsi), %xmm14
-; AVX2-ONLY-NEXT: vpbroadcastd 16(%rdi), %ymm15
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} xmm14 = xmm15[0],xmm14[0],xmm15[1],xmm14[1]
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm15 = ymm6[0],ymm7[0],ymm6[1],ymm7[1],ymm6[4],ymm7[4],ymm6[5],ymm7[5]
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm15 = ymm15[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm14 = xmm14[0,1],xmm15[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm14[0,1,2,3],ymm13[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 20(%r8), %ymm13
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm13[0,1,2,3,4],ymm11[5],ymm13[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5],ymm9[6,7]
-; AVX2-ONLY-NEXT: vpmovzxdq {{.*#+}} xmm9 = mem[0],zero,mem[1],zero
-; AVX2-ONLY-NEXT: vpbroadcastd 20(%rcx), %ymm14
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm9 = xmm9[0,1,2],xmm14[3]
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm14 = ymm4[1,1,1,1,5,5,5,5]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm14 = ymm14[0],ymm5[1],ymm14[2,3,4],ymm5[5],ymm14[6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm14, %xmm14
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm14 = xmm14[0,1],xmm9[2,3]
-; AVX2-ONLY-NEXT: vmovdqa 32(%r8), %ymm9
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm14[0,1,2,3],ymm13[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 24(%rax), %ymm13
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm10[0,1,2,3,4,5,6],ymm13[7]
-; AVX2-ONLY-NEXT: vmovdqa 32(%r9), %ymm10
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm11 = ymm8[2],ymm11[2],ymm8[3],ymm11[3],ymm8[6],ymm11[6],ymm8[7],ymm11[7]
-; AVX2-ONLY-NEXT: vmovdqa 32(%r10), %ymm8
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm4 = ymm4[2],ymm5[2],ymm4[3],ymm5[3],ymm4[6],ymm5[6],ymm4[7],ymm5[7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm4, %xmm5
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm7 = ymm6[2],ymm7[2],ymm6[3],ymm7[3],ymm6[6],ymm7[6],ymm6[7],ymm7[7]
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm6 = ymm7[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm5 = xmm5[0,1],xmm6[2,3]
-; AVX2-ONLY-NEXT: vmovdqa 32(%rax), %ymm6
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm13 = ymm11[0,1,2,3,4,5],ymm13[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm13[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 28(%r10), %ymm5
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm5 = ymm5[2],ymm12[2],ymm5[3],ymm12[3],ymm5[6],ymm12[6],ymm5[7],ymm12[7]
-; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm5 = ymm11[1],ymm5[1],ymm11[3],ymm5[3]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm7, %xmm7
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm4 = ymm4[2,3,2,3,6,7,6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm4, %xmm4
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm4 = xmm4[0,1],xmm7[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm4 = ymm8[0],ymm6[0],ymm8[1],ymm6[1],ymm8[4],ymm6[4],ymm8[5],ymm6[5]
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm5 = ymm9[0],ymm10[0],ymm9[1],ymm10[1],ymm9[4],ymm10[4],ymm9[5],ymm10[5]
-; AVX2-ONLY-NEXT: vpunpcklqdq {{.*#+}} ymm5 = ymm5[0],ymm4[0],ymm5[2],ymm4[2]
-; AVX2-ONLY-NEXT: vpbroadcastd 48(%rsi), %xmm7
-; AVX2-ONLY-NEXT: vpbroadcastd 48(%rdi), %ymm11
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} xmm7 = xmm11[0],xmm7[0],xmm11[1],xmm7[1]
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm11 = ymm2[0],ymm3[0],ymm2[1],ymm3[1],ymm2[4],ymm3[4],ymm2[5],ymm3[5]
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm11 = ymm11[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm7 = xmm7[0,1],xmm11[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm5 = ymm7[0,1,2,3],ymm5[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 52(%r8), %ymm5
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3,4],ymm10[5],ymm5[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],ymm4[6,7]
-; AVX2-ONLY-NEXT: vpmovzxdq {{.*#+}} xmm4 = mem[0],zero,mem[1],zero
-; AVX2-ONLY-NEXT: vpbroadcastd 52(%rcx), %ymm7
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm4 = xmm4[0,1,2],xmm7[3]
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm7 = ymm0[1,1,1,1,5,5,5,5]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm7 = ymm7[0],ymm1[1],ymm7[2,3,4],ymm1[5],ymm7[6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm7, %xmm7
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm7 = xmm7[0,1],xmm4[2,3]
-; AVX2-ONLY-NEXT: vmovdqa 64(%rdx), %ymm4
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm5 = ymm7[0,1,2,3],ymm5[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 56(%rax), %ymm5
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5,6],ymm5[7]
-; AVX2-ONLY-NEXT: vmovdqa 64(%rcx), %ymm5
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm9 = ymm9[2],ymm10[2],ymm9[3],ymm10[3],ymm9[6],ymm10[6],ymm9[7],ymm10[7]
-; AVX2-ONLY-NEXT: vmovdqa 64(%r8), %ymm7
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm10 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm10, %xmm0
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm3 = ymm2[2],ymm3[2],ymm2[3],ymm3[3],ymm2[6],ymm3[6],ymm2[7],ymm3[7]
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm1 = ymm3[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
-; AVX2-ONLY-NEXT: vmovdqa 64(%r9), %ymm1
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm9[0,1,2,3,4,5],ymm8[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 60(%r10), %ymm0
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm0 = ymm0[2],ymm6[2],ymm0[3],ymm6[3],ymm0[6],ymm6[6],ymm0[7],ymm6[7]
-; AVX2-ONLY-NEXT: vmovdqa 64(%r10), %ymm2
-; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm6 = ymm9[1],ymm0[1],ymm9[3],ymm0[3]
-; AVX2-ONLY-NEXT: vmovdqa 64(%rax), %ymm0
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm3, %xmm3
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm8 = ymm10[2,3,2,3,6,7,6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm8, %xmm8
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm3 = xmm8[0,1],xmm3[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm6[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 80(%rsi), %xmm3
-; AVX2-ONLY-NEXT: vpbroadcastd 80(%rdi), %ymm6
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} xmm3 = xmm6[0],xmm3[0],xmm6[1],xmm3[1]
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm6 = ymm4[0],ymm5[0],ymm4[1],ymm5[1],ymm4[4],ymm5[4],ymm4[5],ymm5[5]
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm6 = ymm6[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm3 = xmm3[0,1],xmm6[2,3]
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm6 = ymm2[0],ymm0[0],ymm2[1],ymm0[1],ymm2[4],ymm0[4],ymm2[5],ymm0[5]
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm8 = ymm7[0],ymm1[0],ymm7[1],ymm1[1],ymm7[4],ymm1[4],ymm7[5],ymm1[5]
-; AVX2-ONLY-NEXT: vpunpcklqdq {{.*#+}} ymm8 = ymm8[0],ymm6[0],ymm8[2],ymm6[2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm8[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 84(%r8), %ymm3
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4],ymm1[5],ymm3[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm6[6,7]
-; AVX2-ONLY-NEXT: vpmovzxdq {{.*#+}} xmm6 = mem[0],zero,mem[1],zero
-; AVX2-ONLY-NEXT: vpbroadcastd 84(%rcx), %ymm8
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm6 = xmm6[0,1,2],xmm8[3]
-; AVX2-ONLY-NEXT: vmovdqa 64(%rdi), %ymm8
-; AVX2-ONLY-NEXT: vmovdqa 64(%rsi), %ymm9
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm10 = ymm8[1,1,1,1,5,5,5,5]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm10 = ymm10[0],ymm9[1],ymm10[2,3,4],ymm9[5],ymm10[6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm10, %xmm10
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm6 = xmm10[0,1],xmm6[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm3 = ymm6[0,1,2,3],ymm3[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 88(%rax), %ymm3
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5,6],ymm3[7]
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm1 = ymm7[2],ymm1[2],ymm7[3],ymm1[3],ymm7[6],ymm1[6],ymm7[7],ymm1[7]
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm3 = ymm8[2],ymm9[2],ymm8[3],ymm9[3],ymm8[6],ymm9[6],ymm8[7],ymm9[7]
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm4 = ymm4[2],ymm5[2],ymm4[3],ymm5[3],ymm4[6],ymm5[6],ymm4[7],ymm5[7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm3, %xmm5
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm6 = ymm4[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm5 = xmm5[0,1],xmm6[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm1[0,1,2,3,4,5],ymm2[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm5[0,1,2,3],ymm2[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 92(%r10), %ymm2
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm2 = ymm2[2],ymm0[2],ymm2[3],ymm0[3],ymm2[6],ymm0[6],ymm2[7],ymm0[7]
-; AVX2-ONLY-NEXT: vmovdqa 96(%rdx), %ymm0
-; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm2 = ymm1[1],ymm2[1],ymm1[3],ymm2[3]
-; AVX2-ONLY-NEXT: vmovdqa 96(%rcx), %ymm1
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm4, %xmm4
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm3 = ymm3[2,3,2,3,6,7,6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm3, %xmm3
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm3 = xmm3[0,1],xmm4[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 112(%rsi), %xmm2
-; AVX2-ONLY-NEXT: vpbroadcastd 112(%rdi), %ymm3
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm3 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm3 = ymm3[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm2 = xmm2[0,1],xmm3[2,3]
-; AVX2-ONLY-NEXT: vmovdqa 96(%r8), %ymm3
-; AVX2-ONLY-NEXT: vmovdqa 96(%r9), %ymm4
-; AVX2-ONLY-NEXT: vmovdqa 96(%r10), %ymm5
-; AVX2-ONLY-NEXT: vmovdqa 96(%rax), %ymm6
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm7 = ymm5[0],ymm6[0],ymm5[1],ymm6[1],ymm5[4],ymm6[4],ymm5[5],ymm6[5]
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm8 = ymm3[0],ymm4[0],ymm3[1],ymm4[1],ymm3[4],ymm4[4],ymm3[5],ymm4[5]
-; AVX2-ONLY-NEXT: vpunpcklqdq {{.*#+}} ymm8 = ymm8[0],ymm7[0],ymm8[2],ymm7[2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm8[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 116(%r8), %ymm2
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm4[5],ymm2[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm7[6,7]
-; AVX2-ONLY-NEXT: vpmovzxdq {{.*#+}} xmm7 = mem[0],zero,mem[1],zero
-; AVX2-ONLY-NEXT: vpbroadcastd 116(%rcx), %ymm8
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm7 = xmm7[0,1,2],xmm8[3]
-; AVX2-ONLY-NEXT: vmovdqa 96(%rdi), %ymm8
-; AVX2-ONLY-NEXT: vmovdqa 96(%rsi), %ymm9
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm10 = ymm8[1,1,1,1,5,5,5,5]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm10 = ymm10[0],ymm9[1],ymm10[2,3,4],ymm9[5],ymm10[6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm10, %xmm10
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm7 = xmm10[0,1],xmm7[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm7[0,1,2,3],ymm2[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 120(%rax), %ymm2
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm5[0,1,2,3,4,5,6],ymm2[7]
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm3 = ymm3[2],ymm4[2],ymm3[3],ymm4[3],ymm3[6],ymm4[6],ymm3[7],ymm4[7]
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm4 = ymm8[2],ymm9[2],ymm8[3],ymm9[3],ymm8[6],ymm9[6],ymm8[7],ymm9[7]
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm0 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm4, %xmm1
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm5 = ymm0[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0,1],xmm5[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 124(%r10), %ymm1
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm1 = ymm1[2],ymm6[2],ymm1[3],ymm6[3],ymm1[6],ymm6[6],ymm1[7],ymm6[7]
-; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm1 = ymm3[1],ymm1[1],ymm3[3],ymm1[3]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm0, %xmm0
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm2 = ymm4[2,3,2,3,6,7,6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm2, %xmm2
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm0 = xmm2[0,1],xmm0[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 144(%rsi), %xmm0
-; AVX2-ONLY-NEXT: vpbroadcastd 144(%rdi), %ymm1
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; AVX2-ONLY-NEXT: vmovdqa 128(%rdx), %ymm1
-; AVX2-ONLY-NEXT: vmovdqa 128(%rcx), %ymm2
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm3 = ymm1[0],ymm2[0],ymm1[1],ymm2[1],ymm1[4],ymm2[4],ymm1[5],ymm2[5]
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm3 = ymm3[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm3[2,3]
-; AVX2-ONLY-NEXT: vmovdqa 128(%r8), %ymm3
-; AVX2-ONLY-NEXT: vmovdqa 128(%r9), %ymm4
-; AVX2-ONLY-NEXT: vmovdqa 128(%r10), %ymm5
-; AVX2-ONLY-NEXT: vmovdqa 128(%rax), %ymm6
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm7 = ymm5[0],ymm6[0],ymm5[1],ymm6[1],ymm5[4],ymm6[4],ymm5[5],ymm6[5]
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm8 = ymm3[0],ymm4[0],ymm3[1],ymm4[1],ymm3[4],ymm4[4],ymm3[5],ymm4[5]
-; AVX2-ONLY-NEXT: vpunpcklqdq {{.*#+}} ymm8 = ymm8[0],ymm7[0],ymm8[2],ymm7[2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm8[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 148(%r8), %ymm0
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm4[5],ymm0[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm7[6,7]
-; AVX2-ONLY-NEXT: vpmovzxdq {{.*#+}} xmm7 = mem[0],zero,mem[1],zero
-; AVX2-ONLY-NEXT: vpbroadcastd 148(%rcx), %ymm8
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm7 = xmm7[0,1,2],xmm8[3]
-; AVX2-ONLY-NEXT: vmovdqa 128(%rdi), %ymm8
-; AVX2-ONLY-NEXT: vmovdqa 128(%rsi), %ymm9
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm10 = ymm8[1,1,1,1,5,5,5,5]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm10 = ymm10[0],ymm9[1],ymm10[2,3,4],ymm9[5],ymm10[6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm10, %xmm10
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm7 = xmm10[0,1],xmm7[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm7[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 152(%rax), %ymm0
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm5[0,1,2,3,4,5,6],ymm0[7]
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm3 = ymm3[2],ymm4[2],ymm3[3],ymm4[3],ymm3[6],ymm4[6],ymm3[7],ymm4[7]
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm4 = ymm8[2],ymm9[2],ymm8[3],ymm9[3],ymm8[6],ymm9[6],ymm8[7],ymm9[7]
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm1 = ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[6],ymm2[6],ymm1[7],ymm2[7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm4, %xmm2
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm5 = ymm1[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm2 = xmm2[0,1],xmm5[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 156(%r10), %ymm0
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm0 = ymm0[2],ymm6[2],ymm0[3],ymm6[3],ymm0[6],ymm6[6],ymm0[7],ymm6[7]
-; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm0 = ymm3[1],ymm0[1],ymm3[3],ymm0[3]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm1, %xmm1
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm2 = ymm4[2,3,2,3,6,7,6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm2, %xmm2
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 176(%rsi), %xmm0
-; AVX2-ONLY-NEXT: vpbroadcastd 176(%rdi), %ymm1
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; AVX2-ONLY-NEXT: vmovdqa 160(%rdx), %ymm1
-; AVX2-ONLY-NEXT: vmovdqa 160(%rcx), %ymm2
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm3 = ymm1[0],ymm2[0],ymm1[1],ymm2[1],ymm1[4],ymm2[4],ymm1[5],ymm2[5]
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm3 = ymm3[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm3[2,3]
-; AVX2-ONLY-NEXT: vmovdqa 160(%r8), %ymm3
-; AVX2-ONLY-NEXT: vmovdqa 160(%r9), %ymm4
-; AVX2-ONLY-NEXT: vmovdqa 160(%r10), %ymm5
-; AVX2-ONLY-NEXT: vmovdqa 160(%rax), %ymm6
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm7 = ymm5[0],ymm6[0],ymm5[1],ymm6[1],ymm5[4],ymm6[4],ymm5[5],ymm6[5]
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm8 = ymm3[0],ymm4[0],ymm3[1],ymm4[1],ymm3[4],ymm4[4],ymm3[5],ymm4[5]
-; AVX2-ONLY-NEXT: vpunpcklqdq {{.*#+}} ymm8 = ymm8[0],ymm7[0],ymm8[2],ymm7[2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm8[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 180(%r8), %ymm0
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm4[5],ymm0[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm7[6,7]
-; AVX2-ONLY-NEXT: vpmovzxdq {{.*#+}} xmm7 = mem[0],zero,mem[1],zero
-; AVX2-ONLY-NEXT: vpbroadcastd 180(%rcx), %ymm8
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm7 = xmm7[0,1,2],xmm8[3]
-; AVX2-ONLY-NEXT: vmovdqa 160(%rdi), %ymm8
-; AVX2-ONLY-NEXT: vmovdqa 160(%rsi), %ymm9
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm10 = ymm8[1,1,1,1,5,5,5,5]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm10 = ymm10[0],ymm9[1],ymm10[2,3,4],ymm9[5],ymm10[6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm10, %xmm10
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm7 = xmm10[0,1],xmm7[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm7[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 184(%rax), %ymm0
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm5[0,1,2,3,4,5,6],ymm0[7]
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm3 = ymm3[2],ymm4[2],ymm3[3],ymm4[3],ymm3[6],ymm4[6],ymm3[7],ymm4[7]
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm4 = ymm8[2],ymm9[2],ymm8[3],ymm9[3],ymm8[6],ymm9[6],ymm8[7],ymm9[7]
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm1 = ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[6],ymm2[6],ymm1[7],ymm2[7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm4, %xmm2
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm5 = ymm1[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm2 = xmm2[0,1],xmm5[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 188(%r10), %ymm0
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm0 = ymm0[2],ymm6[2],ymm0[3],ymm6[3],ymm0[6],ymm6[6],ymm0[7],ymm6[7]
-; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm0 = ymm3[1],ymm0[1],ymm3[3],ymm0[3]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm1, %xmm1
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm2 = ymm4[2,3,2,3,6,7,6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm2, %xmm2
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 208(%rsi), %xmm0
-; AVX2-ONLY-NEXT: vpbroadcastd 208(%rdi), %ymm1
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; AVX2-ONLY-NEXT: vmovdqa 192(%rdx), %ymm1
-; AVX2-ONLY-NEXT: vmovdqa 192(%rcx), %ymm2
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm3 = ymm1[0],ymm2[0],ymm1[1],ymm2[1],ymm1[4],ymm2[4],ymm1[5],ymm2[5]
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm3 = ymm3[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm3[2,3]
-; AVX2-ONLY-NEXT: vmovdqa 192(%r8), %ymm3
-; AVX2-ONLY-NEXT: vmovdqa 192(%r9), %ymm4
-; AVX2-ONLY-NEXT: vmovdqa 192(%r10), %ymm5
-; AVX2-ONLY-NEXT: vmovdqa 192(%rax), %ymm6
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm7 = ymm5[0],ymm6[0],ymm5[1],ymm6[1],ymm5[4],ymm6[4],ymm5[5],ymm6[5]
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm8 = ymm3[0],ymm4[0],ymm3[1],ymm4[1],ymm3[4],ymm4[4],ymm3[5],ymm4[5]
-; AVX2-ONLY-NEXT: vpunpcklqdq {{.*#+}} ymm8 = ymm8[0],ymm7[0],ymm8[2],ymm7[2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm8[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 212(%r8), %ymm0
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm4[5],ymm0[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm7[6,7]
-; AVX2-ONLY-NEXT: vpmovzxdq {{.*#+}} xmm7 = mem[0],zero,mem[1],zero
-; AVX2-ONLY-NEXT: vpbroadcastd 212(%rcx), %ymm8
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm7 = xmm7[0,1,2],xmm8[3]
-; AVX2-ONLY-NEXT: vmovdqa 192(%rdi), %ymm8
-; AVX2-ONLY-NEXT: vmovdqa 192(%rsi), %ymm9
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm10 = ymm8[1,1,1,1,5,5,5,5]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm10 = ymm10[0],ymm9[1],ymm10[2,3,4],ymm9[5],ymm10[6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm10, %xmm10
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm7 = xmm10[0,1],xmm7[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm7[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 216(%rax), %ymm0
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm5[0,1,2,3,4,5,6],ymm0[7]
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm3 = ymm3[2],ymm4[2],ymm3[3],ymm4[3],ymm3[6],ymm4[6],ymm3[7],ymm4[7]
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm4 = ymm8[2],ymm9[2],ymm8[3],ymm9[3],ymm8[6],ymm9[6],ymm8[7],ymm9[7]
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm1 = ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[6],ymm2[6],ymm1[7],ymm2[7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm4, %xmm2
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm5 = ymm1[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm2 = xmm2[0,1],xmm5[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 220(%r10), %ymm0
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm0 = ymm0[2],ymm6[2],ymm0[3],ymm6[3],ymm0[6],ymm6[6],ymm0[7],ymm6[7]
-; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm0 = ymm3[1],ymm0[1],ymm3[3],ymm0[3]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm1, %xmm1
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm2 = ymm4[2,3,2,3,6,7,6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm2, %xmm2
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 240(%rsi), %xmm0
-; AVX2-ONLY-NEXT: vpbroadcastd 240(%rdi), %ymm1
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; AVX2-ONLY-NEXT: vmovdqa 224(%rdx), %ymm1
-; AVX2-ONLY-NEXT: vmovdqa 224(%rcx), %ymm2
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm3 = ymm1[0],ymm2[0],ymm1[1],ymm2[1],ymm1[4],ymm2[4],ymm1[5],ymm2[5]
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm3 = ymm3[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm3[2,3]
-; AVX2-ONLY-NEXT: vmovdqa 224(%r8), %ymm3
-; AVX2-ONLY-NEXT: vmovdqa 224(%r9), %ymm4
-; AVX2-ONLY-NEXT: vmovdqa 224(%r10), %ymm5
-; AVX2-ONLY-NEXT: vmovdqa 224(%rax), %ymm6
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm7 = ymm5[0],ymm6[0],ymm5[1],ymm6[1],ymm5[4],ymm6[4],ymm5[5],ymm6[5]
-; AVX2-ONLY-NEXT: vpunpckldq {{.*#+}} ymm8 = ymm3[0],ymm4[0],ymm3[1],ymm4[1],ymm3[4],ymm4[4],ymm3[5],ymm4[5]
-; AVX2-ONLY-NEXT: vpunpcklqdq {{.*#+}} ymm8 = ymm8[0],ymm7[0],ymm8[2],ymm7[2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm8[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 244(%r8), %ymm0
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm4[5],ymm0[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm7[6,7]
-; AVX2-ONLY-NEXT: vpmovzxdq {{.*#+}} xmm7 = mem[0],zero,mem[1],zero
-; AVX2-ONLY-NEXT: vpbroadcastd 244(%rcx), %ymm8
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm7 = xmm7[0,1,2],xmm8[3]
-; AVX2-ONLY-NEXT: vmovdqa 224(%rdi), %ymm8
-; AVX2-ONLY-NEXT: vmovdqa 224(%rsi), %ymm9
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm10 = ymm8[1,1,1,1,5,5,5,5]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm10 = ymm10[0],ymm9[1],ymm10[2,3,4],ymm9[5],ymm10[6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm10, %xmm10
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm7 = xmm10[0,1],xmm7[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm7[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 248(%rax), %ymm0
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm5[0,1,2,3,4,5,6],ymm0[7]
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm3 = ymm3[2],ymm4[2],ymm3[3],ymm4[3],ymm3[6],ymm4[6],ymm3[7],ymm4[7]
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm4 = ymm8[2],ymm9[2],ymm8[3],ymm9[3],ymm8[6],ymm9[6],ymm8[7],ymm9[7]
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm1 = ymm1[2],ymm2[2],ymm1[3],ymm2[3],ymm1[6],ymm2[6],ymm1[7],ymm2[7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm4, %xmm2
-; AVX2-ONLY-NEXT: vpermq {{.*#+}} ymm5 = ymm1[2,2,2,2]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm2 = xmm2[0,1],xmm5[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vpbroadcastd 252(%r10), %ymm0
-; AVX2-ONLY-NEXT: vpunpckhdq {{.*#+}} ymm0 = ymm0[2],ymm6[2],ymm0[3],ymm6[3],ymm0[6],ymm6[6],ymm0[7],ymm6[7]
-; AVX2-ONLY-NEXT: vpunpckhqdq {{.*#+}} ymm0 = ymm3[1],ymm0[1],ymm3[3],ymm0[3]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm1, %xmm1
-; AVX2-ONLY-NEXT: vpshufd {{.*#+}} ymm2 = ymm4[2,3,2,3,6,7,6,7]
-; AVX2-ONLY-NEXT: vextracti128 $1, %ymm2, %xmm2
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
-; AVX2-ONLY-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 32(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovaps (%rdi), %ymm4
+; AVX2-ONLY-NEXT: vmovaps 32(%rsi), %ymm1
+; AVX2-ONLY-NEXT: vmovaps (%rsi), %ymm5
+; AVX2-ONLY-NEXT: vmovaps 32(%rdx), %ymm2
+; AVX2-ONLY-NEXT: vmovaps (%rdx), %ymm6
+; AVX2-ONLY-NEXT: vmovaps 32(%rcx), %ymm3
+; AVX2-ONLY-NEXT: vmovaps (%rcx), %ymm7
+; AVX2-ONLY-NEXT: vmovaps (%r8), %ymm8
+; AVX2-ONLY-NEXT: vmovaps (%r9), %ymm11
+; AVX2-ONLY-NEXT: vmovaps (%r10), %ymm10
+; AVX2-ONLY-NEXT: vmovaps (%rax), %ymm12
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm9 = ymm10[0],ymm12[0],ymm10[1],ymm12[1],ymm10[4],ymm12[4],ymm10[5],ymm12[5]
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm13 = ymm4[0],ymm5[0],ymm4[1],ymm5[1],ymm4[4],ymm5[4],ymm4[5],ymm5[5]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm13, %xmm13
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm14 = ymm6[0],ymm7[0],ymm6[1],ymm7[1],ymm6[4],ymm7[4],ymm6[5],ymm7[5]
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm15 = ymm14[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm13 = xmm13[0,1],xmm15[2,3]
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm15 = ymm8[0],ymm11[0],ymm8[1],ymm11[1],ymm8[4],ymm11[4],ymm8[5],ymm11[5]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm15 = ymm15[0],ymm9[0],ymm15[2],ymm9[2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3],ymm15[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 20(%r8), %ymm13
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4],ymm11[5],ymm13[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5],ymm9[6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm14, %xmm9
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} ymm14 = ymm4[1,1,1,1,5,5,5,5]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm14 = ymm14[0],ymm5[1],ymm14[2,3,4],ymm5[5],ymm14[6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm14, %xmm14
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm14 = xmm14[0,1],xmm9[2,3]
+; AVX2-ONLY-NEXT: vmovaps 32(%r8), %ymm9
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm14[0,1,2,3],ymm13[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 24(%rax), %ymm13
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm10[0,1,2,3,4,5,6],ymm13[7]
+; AVX2-ONLY-NEXT: vmovaps 32(%r9), %ymm10
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm11 = ymm8[2],ymm11[2],ymm8[3],ymm11[3],ymm8[6],ymm11[6],ymm8[7],ymm11[7]
+; AVX2-ONLY-NEXT: vmovaps 32(%r10), %ymm8
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm4 = ymm4[2],ymm5[2],ymm4[3],ymm5[3],ymm4[6],ymm5[6],ymm4[7],ymm5[7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm4, %xmm5
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm6 = ymm6[2],ymm7[2],ymm6[3],ymm7[3],ymm6[6],ymm7[6],ymm6[7],ymm7[7]
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm7 = ymm6[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm7 = xmm5[0,1],xmm7[2,3]
+; AVX2-ONLY-NEXT: vmovaps 32(%rax), %ymm5
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm11[0,1,2,3,4,5],ymm13[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm13[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 28(%r10), %ymm7
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm7 = ymm7[2],ymm12[2],ymm7[3],ymm12[3],ymm7[6],ymm12[6],ymm7[7],ymm12[7]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm7 = ymm11[1],ymm7[1],ymm11[3],ymm7[3]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm6, %xmm6
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} ymm4 = ymm4[2,3,2,3,6,7,6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm4, %xmm4
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm4[0,1],xmm6[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm7[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm4 = ymm8[0],ymm5[0],ymm8[1],ymm5[1],ymm8[4],ymm5[4],ymm8[5],ymm5[5]
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm6 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm6, %xmm6
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm7 = ymm2[0],ymm3[0],ymm2[1],ymm3[1],ymm2[4],ymm3[4],ymm2[5],ymm3[5]
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm11 = ymm7[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0,1],xmm11[2,3]
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm11 = ymm9[0],ymm10[0],ymm9[1],ymm10[1],ymm9[4],ymm10[4],ymm9[5],ymm10[5]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm11 = ymm11[0],ymm4[0],ymm11[2],ymm4[2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm11[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 52(%r8), %ymm6
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4],ymm10[5],ymm6[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm4[6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm7, %xmm4
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} ymm7 = ymm0[1,1,1,1,5,5,5,5]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0],ymm1[1],ymm7[2,3,4],ymm1[5],ymm7[6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm7, %xmm7
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm7 = xmm7[0,1],xmm4[2,3]
+; AVX2-ONLY-NEXT: vmovaps 64(%rdi), %ymm4
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3],ymm6[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 56(%rax), %ymm6
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5,6],ymm6[7]
+; AVX2-ONLY-NEXT: vmovaps 64(%rsi), %ymm7
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm9 = ymm9[2],ymm10[2],ymm9[3],ymm10[3],ymm9[6],ymm10[6],ymm9[7],ymm10[7]
+; AVX2-ONLY-NEXT: vmovaps 64(%rdx), %ymm6
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm1 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm0
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm2 = ymm2[2],ymm3[2],ymm2[3],ymm3[3],ymm2[6],ymm3[6],ymm2[7],ymm3[7]
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm3 = ymm2[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm0[0,1],xmm3[2,3]
+; AVX2-ONLY-NEXT: vmovaps 64(%rcx), %ymm0
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm9[0,1,2,3,4,5],ymm8[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm8[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 60(%r10), %ymm3
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm3 = ymm3[2],ymm5[2],ymm3[3],ymm5[3],ymm3[6],ymm5[6],ymm3[7],ymm5[7]
+; AVX2-ONLY-NEXT: vmovaps 64(%r8), %ymm5
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm3 = ymm9[1],ymm3[1],ymm9[3],ymm3[3]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm2
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[2,3,2,3,6,7,6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm1 = ymm4[0],ymm7[0],ymm4[1],ymm7[1],ymm4[4],ymm7[4],ymm4[5],ymm7[5]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm2 = ymm6[0],ymm0[0],ymm6[1],ymm0[1],ymm6[4],ymm0[4],ymm6[5],ymm0[5]
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm3 = ymm2[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],xmm3[2,3]
+; AVX2-ONLY-NEXT: vmovaps 64(%r9), %ymm3
+; AVX2-ONLY-NEXT: vmovaps 64(%r10), %ymm8
+; AVX2-ONLY-NEXT: vmovaps 64(%rax), %ymm9
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm10 = ymm8[0],ymm9[0],ymm8[1],ymm9[1],ymm8[4],ymm9[4],ymm8[5],ymm9[5]
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm11 = ymm5[0],ymm3[0],ymm5[1],ymm3[1],ymm5[4],ymm3[4],ymm5[5],ymm3[5]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm11 = ymm11[0],ymm10[0],ymm11[2],ymm10[2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm11[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 84(%r8), %ymm1
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4],ymm3[5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm10[6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm2
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm4[1,1,1,1,5,5,5,5]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0],ymm7[1],ymm10[2,3,4],ymm7[5],ymm10[6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm10, %xmm10
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm10[0,1],xmm2[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 88(%rax), %ymm1
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm8[0,1,2,3,4,5,6],ymm1[7]
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm2 = ymm5[2],ymm3[2],ymm5[3],ymm3[3],ymm5[6],ymm3[6],ymm5[7],ymm3[7]
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm3 = ymm4[2],ymm7[2],ymm4[3],ymm7[3],ymm4[6],ymm7[6],ymm4[7],ymm7[7]
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm0 = ymm6[2],ymm0[2],ymm6[3],ymm0[3],ymm6[6],ymm0[6],ymm6[7],ymm0[7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm3, %xmm4
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm5 = ymm0[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm4[0,1],xmm5[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm4[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 92(%r10), %ymm1
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm1 = ymm1[2],ymm9[2],ymm1[3],ymm9[3],ymm1[6],ymm9[6],ymm1[7],ymm9[7]
+; AVX2-ONLY-NEXT: vmovaps 96(%rdi), %ymm4
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm2[1],ymm1[1],ymm2[3],ymm1[3]
+; AVX2-ONLY-NEXT: vmovaps 96(%rsi), %ymm2
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm0, %xmm0
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm3[2,3,2,3,6,7,6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm3, %xmm3
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm3[0,1],xmm0[2,3]
+; AVX2-ONLY-NEXT: vmovaps 96(%rdx), %ymm3
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm0 = ymm4[0],ymm2[0],ymm4[1],ymm2[1],ymm4[4],ymm2[4],ymm4[5],ymm2[5]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm0, %xmm0
+; AVX2-ONLY-NEXT: vmovaps 96(%rcx), %ymm1
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm5 = ymm3[0],ymm1[0],ymm3[1],ymm1[1],ymm3[4],ymm1[4],ymm3[5],ymm1[5]
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm6 = ymm5[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm6[2,3]
+; AVX2-ONLY-NEXT: vmovaps 96(%r8), %ymm6
+; AVX2-ONLY-NEXT: vmovaps 96(%r9), %ymm7
+; AVX2-ONLY-NEXT: vmovaps 96(%r10), %ymm8
+; AVX2-ONLY-NEXT: vmovaps 96(%rax), %ymm9
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm10 = ymm8[0],ymm9[0],ymm8[1],ymm9[1],ymm8[4],ymm9[4],ymm8[5],ymm9[5]
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm11 = ymm6[0],ymm7[0],ymm6[1],ymm7[1],ymm6[4],ymm7[4],ymm6[5],ymm7[5]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm11 = ymm11[0],ymm10[0],ymm11[2],ymm10[2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm11[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 116(%r8), %ymm0
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm7[5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm10[6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm5, %xmm5
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm4[1,1,1,1,5,5,5,5]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0],ymm2[1],ymm10[2,3,4],ymm2[5],ymm10[6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm10, %xmm10
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm10[0,1],xmm5[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 120(%rax), %ymm0
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3,4,5,6],ymm0[7]
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm5 = ymm6[2],ymm7[2],ymm6[3],ymm7[3],ymm6[6],ymm7[6],ymm6[7],ymm7[7]
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm2 = ymm4[2],ymm2[2],ymm4[3],ymm2[3],ymm4[6],ymm2[6],ymm4[7],ymm2[7]
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm1 = ymm3[2],ymm1[2],ymm3[3],ymm1[3],ymm3[6],ymm1[6],ymm3[7],ymm1[7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm3
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm4 = ymm1[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm4[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 124(%r10), %ymm0
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm0 = ymm0[2],ymm9[2],ymm0[3],ymm9[3],ymm0[6],ymm9[6],ymm0[7],ymm9[7]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm5[1],ymm0[1],ymm5[3],ymm0[3]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm2[2,3,2,3,6,7,6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm2
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 128(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovaps 128(%rsi), %ymm1
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm2 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm2
+; AVX2-ONLY-NEXT: vmovaps 128(%rdx), %ymm3
+; AVX2-ONLY-NEXT: vmovaps 128(%rcx), %ymm4
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm5 = ymm3[0],ymm4[0],ymm3[1],ymm4[1],ymm3[4],ymm4[4],ymm3[5],ymm4[5]
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm6 = ymm5[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0,1],xmm6[2,3]
+; AVX2-ONLY-NEXT: vmovaps 128(%r8), %ymm6
+; AVX2-ONLY-NEXT: vmovaps 128(%r9), %ymm7
+; AVX2-ONLY-NEXT: vmovaps 128(%r10), %ymm8
+; AVX2-ONLY-NEXT: vmovaps 128(%rax), %ymm9
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm10 = ymm8[0],ymm9[0],ymm8[1],ymm9[1],ymm8[4],ymm9[4],ymm8[5],ymm9[5]
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm11 = ymm6[0],ymm7[0],ymm6[1],ymm7[1],ymm6[4],ymm7[4],ymm6[5],ymm7[5]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm11 = ymm11[0],ymm10[0],ymm11[2],ymm10[2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm11[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 148(%r8), %ymm2
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm7[5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm10[6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm5, %xmm5
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm0[1,1,1,1,5,5,5,5]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0],ymm1[1],ymm10[2,3,4],ymm1[5],ymm10[6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm10, %xmm10
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm10[0,1],xmm5[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm5[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 152(%rax), %ymm2
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm8[0,1,2,3,4,5,6],ymm2[7]
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm5 = ymm6[2],ymm7[2],ymm6[3],ymm7[3],ymm6[6],ymm7[6],ymm6[7],ymm7[7]
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm0 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm1 = ymm3[2],ymm4[2],ymm3[3],ymm4[3],ymm3[6],ymm4[6],ymm3[7],ymm4[7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm0, %xmm3
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm4 = ymm1[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm4[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm5[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 156(%r10), %ymm2
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm2 = ymm2[2],ymm9[2],ymm2[3],ymm9[3],ymm2[6],ymm9[6],ymm2[7],ymm9[7]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm2 = ymm5[1],ymm2[1],ymm5[3],ymm2[3]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,3,2,3,6,7,6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm0, %xmm0
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 160(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovaps 160(%rsi), %ymm1
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm2 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm2
+; AVX2-ONLY-NEXT: vmovaps 160(%rdx), %ymm3
+; AVX2-ONLY-NEXT: vmovaps 160(%rcx), %ymm4
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm5 = ymm3[0],ymm4[0],ymm3[1],ymm4[1],ymm3[4],ymm4[4],ymm3[5],ymm4[5]
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm6 = ymm5[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0,1],xmm6[2,3]
+; AVX2-ONLY-NEXT: vmovaps 160(%r8), %ymm6
+; AVX2-ONLY-NEXT: vmovaps 160(%r9), %ymm7
+; AVX2-ONLY-NEXT: vmovaps 160(%r10), %ymm8
+; AVX2-ONLY-NEXT: vmovaps 160(%rax), %ymm9
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm10 = ymm8[0],ymm9[0],ymm8[1],ymm9[1],ymm8[4],ymm9[4],ymm8[5],ymm9[5]
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm11 = ymm6[0],ymm7[0],ymm6[1],ymm7[1],ymm6[4],ymm7[4],ymm6[5],ymm7[5]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm11 = ymm11[0],ymm10[0],ymm11[2],ymm10[2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm11[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 180(%r8), %ymm2
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm7[5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm10[6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm5, %xmm5
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm0[1,1,1,1,5,5,5,5]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0],ymm1[1],ymm10[2,3,4],ymm1[5],ymm10[6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm10, %xmm10
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm10[0,1],xmm5[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm5[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 184(%rax), %ymm2
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm8[0,1,2,3,4,5,6],ymm2[7]
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm5 = ymm6[2],ymm7[2],ymm6[3],ymm7[3],ymm6[6],ymm7[6],ymm6[7],ymm7[7]
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm0 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm1 = ymm3[2],ymm4[2],ymm3[3],ymm4[3],ymm3[6],ymm4[6],ymm3[7],ymm4[7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm0, %xmm3
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm4 = ymm1[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm4[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm5[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 188(%r10), %ymm2
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm2 = ymm2[2],ymm9[2],ymm2[3],ymm9[3],ymm2[6],ymm9[6],ymm2[7],ymm9[7]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm2 = ymm5[1],ymm2[1],ymm5[3],ymm2[3]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,3,2,3,6,7,6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm0, %xmm0
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 192(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovaps 192(%rsi), %ymm1
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm2 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm2
+; AVX2-ONLY-NEXT: vmovaps 192(%rdx), %ymm3
+; AVX2-ONLY-NEXT: vmovaps 192(%rcx), %ymm4
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm5 = ymm3[0],ymm4[0],ymm3[1],ymm4[1],ymm3[4],ymm4[4],ymm3[5],ymm4[5]
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm6 = ymm5[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0,1],xmm6[2,3]
+; AVX2-ONLY-NEXT: vmovaps 192(%r8), %ymm6
+; AVX2-ONLY-NEXT: vmovaps 192(%r9), %ymm7
+; AVX2-ONLY-NEXT: vmovaps 192(%r10), %ymm8
+; AVX2-ONLY-NEXT: vmovaps 192(%rax), %ymm9
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm10 = ymm8[0],ymm9[0],ymm8[1],ymm9[1],ymm8[4],ymm9[4],ymm8[5],ymm9[5]
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm11 = ymm6[0],ymm7[0],ymm6[1],ymm7[1],ymm6[4],ymm7[4],ymm6[5],ymm7[5]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm11 = ymm11[0],ymm10[0],ymm11[2],ymm10[2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm11[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 212(%r8), %ymm2
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm7[5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm10[6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm5, %xmm5
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm0[1,1,1,1,5,5,5,5]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0],ymm1[1],ymm10[2,3,4],ymm1[5],ymm10[6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm10, %xmm10
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm10[0,1],xmm5[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm5[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 216(%rax), %ymm2
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm8[0,1,2,3,4,5,6],ymm2[7]
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm5 = ymm6[2],ymm7[2],ymm6[3],ymm7[3],ymm6[6],ymm7[6],ymm6[7],ymm7[7]
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm0 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm1 = ymm3[2],ymm4[2],ymm3[3],ymm4[3],ymm3[6],ymm4[6],ymm3[7],ymm4[7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm0, %xmm3
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm4 = ymm1[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm4[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm5[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 220(%r10), %ymm2
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm2 = ymm2[2],ymm9[2],ymm2[3],ymm9[3],ymm2[6],ymm9[6],ymm2[7],ymm9[7]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm2 = ymm5[1],ymm2[1],ymm5[3],ymm2[3]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,3,2,3,6,7,6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm0, %xmm0
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vmovaps 224(%rdi), %ymm0
+; AVX2-ONLY-NEXT: vmovaps 224(%rsi), %ymm1
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm2 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm2
+; AVX2-ONLY-NEXT: vmovaps 224(%rdx), %ymm3
+; AVX2-ONLY-NEXT: vmovaps 224(%rcx), %ymm4
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm5 = ymm3[0],ymm4[0],ymm3[1],ymm4[1],ymm3[4],ymm4[4],ymm3[5],ymm4[5]
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm6 = ymm5[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0,1],xmm6[2,3]
+; AVX2-ONLY-NEXT: vmovaps 224(%r8), %ymm6
+; AVX2-ONLY-NEXT: vmovaps 224(%r9), %ymm7
+; AVX2-ONLY-NEXT: vmovaps 224(%r10), %ymm8
+; AVX2-ONLY-NEXT: vmovaps 224(%rax), %ymm9
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm10 = ymm8[0],ymm9[0],ymm8[1],ymm9[1],ymm8[4],ymm9[4],ymm8[5],ymm9[5]
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} ymm11 = ymm6[0],ymm7[0],ymm6[1],ymm7[1],ymm6[4],ymm7[4],ymm6[5],ymm7[5]
+; AVX2-ONLY-NEXT: vunpcklpd {{.*#+}} ymm11 = ymm11[0],ymm10[0],ymm11[2],ymm10[2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm11[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 244(%r8), %ymm2
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4],ymm7[5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm10[6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm5, %xmm5
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} ymm10 = ymm0[1,1,1,1,5,5,5,5]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm10 = ymm10[0],ymm1[1],ymm10[2,3,4],ymm1[5],ymm10[6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm10, %xmm10
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm10[0,1],xmm5[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm5[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 248(%rax), %ymm2
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm8[0,1,2,3,4,5,6],ymm2[7]
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm5 = ymm6[2],ymm7[2],ymm6[3],ymm7[3],ymm6[6],ymm7[6],ymm6[7],ymm7[7]
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm0 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm1 = ymm3[2],ymm4[2],ymm3[3],ymm4[3],ymm3[6],ymm4[6],ymm3[7],ymm4[7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm0, %xmm3
+; AVX2-ONLY-NEXT: vpermpd {{.*#+}} ymm4 = ymm1[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1],xmm4[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm5[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT: vbroadcastss 252(%r10), %ymm2
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} ymm2 = ymm2[2],ymm9[2],ymm2[3],ymm9[3],ymm2[6],ymm9[6],ymm2[7],ymm9[7]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} ymm2 = ymm5[1],ymm2[1],ymm5[3],ymm2[3]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[2,3,2,3,6,7,6,7]
+; AVX2-ONLY-NEXT: vextractf128 $1, %ymm0, %xmm0
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vmovaps (%rcx), %xmm0
; AVX2-ONLY-NEXT: vbroadcastss %xmm0, %xmm2
; AVX2-ONLY-NEXT: vmovaps (%rdx), %xmm1
; AVX2-ONLY-NEXT: vbroadcastss %xmm1, %xmm3
; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
-; AVX2-ONLY-NEXT: vmovaps (%rsi), %xmm3
-; AVX2-ONLY-NEXT: vmovaps (%rdi), %xmm4
-; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm5 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm5[0,1],xmm2[2,3]
+; AVX2-ONLY-NEXT: vmovaps (%rsi), %xmm4
+; AVX2-ONLY-NEXT: vmovaps (%rdi), %xmm5
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm3 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm3[0,1],xmm2[2,3]
; AVX2-ONLY-NEXT: vmovaps (%rax), %xmm2
; AVX2-ONLY-NEXT: vbroadcastss %xmm2, %xmm7
-; AVX2-ONLY-NEXT: vmovaps (%r10), %xmm5
-; AVX2-ONLY-NEXT: vbroadcastss %xmm5, %xmm8
+; AVX2-ONLY-NEXT: vmovaps (%r10), %xmm3
+; AVX2-ONLY-NEXT: vbroadcastss %xmm3, %xmm8
; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm7 = xmm8[0],xmm7[0],xmm8[1],xmm7[1]
; AVX2-ONLY-NEXT: vmovaps (%r9), %xmm8
; AVX2-ONLY-NEXT: vmovaps (%r8), %xmm9
@@ -6256,11 +6211,11 @@ define void @store_i32_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm10[0,1,2,3,4,5],ymm7[6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vshufps {{.*#+}} xmm6 = xmm4[1,1,1,1]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0],xmm3[1],xmm6[2,3]
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} xmm6 = xmm5[1,1,1,1]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0],xmm4[1],xmm6[2,3]
; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm7 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0,1],xmm7[2,3]
-; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm7 = xmm5[0],xmm2[0],xmm5[1],xmm2[1]
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm7 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm7, %ymm0, %ymm7
; AVX2-ONLY-NEXT: vshufps {{.*#+}} xmm10 = xmm9[1,1,1,1]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm10 = xmm10[0],xmm8[1],xmm10[2,3]
@@ -6268,23 +6223,23 @@ define void @store_i32_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm10[0,1,2,3,4,5],ymm7[6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm3 = xmm4[2],xmm3[2],xmm4[3],xmm3[3]
-; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm4 = xmm9[2],xmm8[2],xmm9[3],xmm8[3]
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm4 = xmm5[2],xmm4[2],xmm5[3],xmm4[3]
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm5 = xmm9[2],xmm8[2],xmm9[3],xmm8[3]
; AVX2-ONLY-NEXT: vshufps {{.*#+}} xmm6 = xmm2[2,2,2,2]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm5[0,1,2],xmm6[3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm3[0,1,2],xmm6[3]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6
-; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm4, %ymm0, %ymm7
+; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm7
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3,4,5],ymm6[6,7]
; AVX2-ONLY-NEXT: vshufps {{.*#+}} xmm7 = xmm0[2,2,2,2]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm7 = xmm1[0,1,2],xmm7[3]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm7 = xmm3[0,1],xmm7[2,3]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm7 = xmm4[0,1],xmm7[2,3]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3],ymm6[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm0 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
-; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm3[1],xmm0[1]
-; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm1 = xmm5[2],xmm2[2],xmm5[3],xmm2[3]
+; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm4[1],xmm0[1]
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm1 = xmm3[2],xmm2[2],xmm3[3],xmm2[3]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX2-ONLY-NEXT: vshufps {{.*#+}} xmm2 = xmm4[2,3,2,3]
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} xmm2 = xmm5[2,3,2,3]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
@@ -6298,14 +6253,14 @@ define void @store_i32_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps 32(%rdi), %xmm5
; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm3 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm3[0,1],xmm2[2,3]
-; AVX2-ONLY-NEXT: vmovaps 32(%rax), %xmm2
-; AVX2-ONLY-NEXT: vbroadcastss %xmm2, %xmm7
-; AVX2-ONLY-NEXT: vmovaps 32(%r10), %xmm3
-; AVX2-ONLY-NEXT: vbroadcastss %xmm3, %xmm8
+; AVX2-ONLY-NEXT: vmovaps 32(%r10), %xmm2
+; AVX2-ONLY-NEXT: vmovaps 32(%rax), %xmm3
+; AVX2-ONLY-NEXT: vbroadcastss %xmm3, %xmm7
+; AVX2-ONLY-NEXT: vbroadcastss %xmm2, %xmm8
; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm7 = xmm8[0],xmm7[0],xmm8[1],xmm7[1]
-; AVX2-ONLY-NEXT: vmovaps 32(%r8), %xmm8
-; AVX2-ONLY-NEXT: vmovaps 32(%r9), %xmm9
-; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm10 = xmm8[0],xmm9[0],xmm8[1],xmm9[1]
+; AVX2-ONLY-NEXT: vmovaps 32(%r9), %xmm8
+; AVX2-ONLY-NEXT: vmovaps 32(%r8), %xmm9
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm10 = xmm9[0],xmm8[0],xmm9[1],xmm8[1]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm10, %ymm0, %ymm10
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm7, %ymm0, %ymm7
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm10[0,1,2,3,4,5],ymm7[6,7]
@@ -6315,18 +6270,18 @@ define void @store_i32_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0],xmm4[1],xmm6[2,3]
; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm7 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0,1],xmm7[2,3]
-; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm7 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm7 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm7, %ymm0, %ymm7
-; AVX2-ONLY-NEXT: vshufps {{.*#+}} xmm10 = xmm8[1,1,1,1]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm10 = xmm10[0],xmm9[1],xmm10[2,3]
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} xmm10 = xmm9[1,1,1,1]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm10 = xmm10[0],xmm8[1],xmm10[2,3]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm10, %ymm0, %ymm10
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm10[0,1,2,3,4,5],ymm7[6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm4 = xmm5[2],xmm4[2],xmm5[3],xmm4[3]
-; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm5 = xmm8[2],xmm9[2],xmm8[3],xmm9[3]
-; AVX2-ONLY-NEXT: vshufps {{.*#+}} xmm6 = xmm2[2,2,2,2]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm3[0,1,2],xmm6[3]
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm5 = xmm9[2],xmm8[2],xmm9[3],xmm8[3]
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} xmm6 = xmm3[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm2[0,1,2],xmm6[3]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm7
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3,4,5],ymm6[6,7]
@@ -6337,7 +6292,7 @@ define void @store_i32_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm0 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm4[1],xmm0[1]
-; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm1 = xmm3[2],xmm2[2],xmm3[3],xmm2[3]
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm1 = xmm2[2],xmm3[2],xmm2[3],xmm3[3]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX2-ONLY-NEXT: vshufps {{.*#+}} xmm2 = xmm5[2,3,2,3]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
@@ -6353,14 +6308,14 @@ define void @store_i32_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovaps 64(%rdi), %xmm5
; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm3 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm3[0,1],xmm2[2,3]
-; AVX2-ONLY-NEXT: vmovaps 64(%rax), %xmm2
-; AVX2-ONLY-NEXT: vmovaps 64(%r10), %xmm3
-; AVX2-ONLY-NEXT: vbroadcastss %xmm2, %xmm7
-; AVX2-ONLY-NEXT: vbroadcastss %xmm3, %xmm8
+; AVX2-ONLY-NEXT: vmovaps 64(%r10), %xmm2
+; AVX2-ONLY-NEXT: vmovaps 64(%rax), %xmm3
+; AVX2-ONLY-NEXT: vbroadcastss %xmm3, %xmm7
+; AVX2-ONLY-NEXT: vbroadcastss %xmm2, %xmm8
; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm7 = xmm8[0],xmm7[0],xmm8[1],xmm7[1]
-; AVX2-ONLY-NEXT: vmovaps 64(%r8), %xmm8
-; AVX2-ONLY-NEXT: vmovaps 64(%r9), %xmm9
-; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm10 = xmm8[0],xmm9[0],xmm8[1],xmm9[1]
+; AVX2-ONLY-NEXT: vmovaps 64(%r9), %xmm8
+; AVX2-ONLY-NEXT: vmovaps 64(%r8), %xmm9
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm10 = xmm9[0],xmm8[0],xmm9[1],xmm8[1]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm10, %ymm0, %ymm10
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm7, %ymm0, %ymm7
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm10[0,1,2,3,4,5],ymm7[6,7]
@@ -6370,18 +6325,18 @@ define void @store_i32_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0],xmm4[1],xmm6[2,3]
; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm7 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm6[0,1],xmm7[2,3]
-; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm7 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
+; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm7 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm7, %ymm0, %ymm7
-; AVX2-ONLY-NEXT: vshufps {{.*#+}} xmm10 = xmm8[1,1,1,1]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm10 = xmm10[0],xmm9[1],xmm10[2,3]
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} xmm10 = xmm9[1,1,1,1]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm10 = xmm10[0],xmm8[1],xmm10[2,3]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm10, %ymm0, %ymm10
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm10[0,1,2,3,4,5],ymm7[6,7]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5,6,7]
; AVX2-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm4 = xmm5[2],xmm4[2],xmm5[3],xmm4[3]
-; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm5 = xmm8[2],xmm9[2],xmm8[3],xmm9[3]
-; AVX2-ONLY-NEXT: vshufps {{.*#+}} xmm6 = xmm2[2,2,2,2]
-; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm3[0,1,2],xmm6[3]
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm5 = xmm9[2],xmm8[2],xmm9[3],xmm8[3]
+; AVX2-ONLY-NEXT: vshufps {{.*#+}} xmm6 = xmm3[2,2,2,2]
+; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm2[0,1,2],xmm6[3]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm7
; AVX2-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3,4,5],ymm6[6,7]
@@ -6392,7 +6347,7 @@ define void @store_i32_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm0 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
; AVX2-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm4[1],xmm0[1]
-; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm1 = xmm3[2],xmm2[2],xmm3[3],xmm2[3]
+; AVX2-ONLY-NEXT: vunpckhps {{.*#+}} xmm1 = xmm2[2],xmm3[2],xmm2[3],xmm3[3]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX2-ONLY-NEXT: vshufps {{.*#+}} xmm2 = xmm5[2,3,2,3]
; AVX2-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2
@@ -6519,8 +6474,8 @@ define void @store_i32_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm3 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm6 = xmm3[0,1],xmm2[2,3]
; AVX2-ONLY-NEXT: vmovaps 160(%rax), %xmm2
-; AVX2-ONLY-NEXT: vmovaps 160(%r10), %xmm3
; AVX2-ONLY-NEXT: vbroadcastss %xmm2, %xmm7
+; AVX2-ONLY-NEXT: vmovaps 160(%r10), %xmm3
; AVX2-ONLY-NEXT: vbroadcastss %xmm3, %xmm8
; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm7 = xmm8[0],xmm7[0],xmm8[1],xmm7[1]
; AVX2-ONLY-NEXT: vmovaps 160(%r9), %xmm8
@@ -6574,8 +6529,8 @@ define void @store_i32_stride8_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.ve
; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm3 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
; AVX2-ONLY-NEXT: vblendps {{.*#+}} xmm9 = xmm3[0,1],xmm2[2,3]
; AVX2-ONLY-NEXT: vmovaps 192(%rax), %xmm2
-; AVX2-ONLY-NEXT: vmovaps 192(%r10), %xmm3
; AVX2-ONLY-NEXT: vbroadcastss %xmm2, %xmm7
+; AVX2-ONLY-NEXT: vmovaps 192(%r10), %xmm3
; AVX2-ONLY-NEXT: vbroadcastss %xmm3, %xmm8
; AVX2-ONLY-NEXT: vunpcklps {{.*#+}} xmm8 = xmm8[0],xmm7[0],xmm8[1],xmm7[1]
; AVX2-ONLY-NEXT: vmovaps 192(%r9), %xmm0
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-5.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-5.ll
index 04bc4e5da890f..6f0919967a5d8 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-5.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-5.ll
@@ -205,12 +205,12 @@ define void @store_i8_stride5_vf4(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vecp
define void @store_i8_stride5_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vecptr2, ptr %in.vecptr3, ptr %in.vecptr4, ptr %out.vec) nounwind {
; SSE-LABEL: store_i8_stride5_vf8:
; SSE: # %bb.0:
-; SSE-NEXT: movq {{.*#+}} xmm1 = mem[0],zero
; SSE-NEXT: movq {{.*#+}} xmm2 = mem[0],zero
+; SSE-NEXT: movq {{.*#+}} xmm1 = mem[0],zero
; SSE-NEXT: movq {{.*#+}} xmm4 = mem[0],zero
; SSE-NEXT: movq {{.*#+}} xmm3 = mem[0],zero
; SSE-NEXT: movq {{.*#+}} xmm0 = mem[0],zero
-; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm1[1,1,1,1]
+; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm2[1,1,1,1]
; SSE-NEXT: movdqa {{.*#+}} xmm8 = [65535,65535,0,65535,0,65535,65535,0]
; SSE-NEXT: punpcklbw {{.*#+}} xmm3 = xmm3[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm3[2,1,2,3]
@@ -225,8 +225,8 @@ define void @store_i8_stride5_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vecp
; SSE-NEXT: movdqa {{.*#+}} xmm6 = [255,0,255,255,255,255,0,255,255,255,255,0,255,255,255,255]
; SSE-NEXT: movdqa %xmm6, %xmm10
; SSE-NEXT: pandn %xmm5, %xmm10
-; SSE-NEXT: movdqa %xmm2, %xmm7
-; SSE-NEXT: punpcklbw {{.*#+}} xmm7 = xmm7[0],xmm2[0],xmm7[1],xmm2[1],xmm7[2],xmm2[2],xmm7[3],xmm2[3],xmm7[4],xmm2[4],xmm7[5],xmm2[5],xmm7[6],xmm2[6],xmm7[7],xmm2[7]
+; SSE-NEXT: movdqa %xmm1, %xmm7
+; SSE-NEXT: punpcklbw {{.*#+}} xmm7 = xmm7[0],xmm1[0],xmm7[1],xmm1[1],xmm7[2],xmm1[2],xmm7[3],xmm1[3],xmm7[4],xmm1[4],xmm7[5],xmm1[5],xmm7[6],xmm1[6],xmm7[7],xmm1[7]
; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm7[2,1,2,3]
; SSE-NEXT: pshuflw {{.*#+}} xmm5 = xmm5[3,1,0,3,4,5,6,7]
; SSE-NEXT: pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,5,6,6]
@@ -249,7 +249,7 @@ define void @store_i8_stride5_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vecp
; SSE-NEXT: pandn %xmm10, %xmm8
; SSE-NEXT: por %xmm9, %xmm8
; SSE-NEXT: movdqa {{.*#+}} xmm9 = [255,255,0,0,255,255,255,0,0,255,255,255,0,0,255,255]
-; SSE-NEXT: pshufd {{.*#+}} xmm10 = xmm1[0,0,0,0]
+; SSE-NEXT: pshufd {{.*#+}} xmm10 = xmm2[0,0,0,0]
; SSE-NEXT: pand %xmm6, %xmm10
; SSE-NEXT: pshuflw {{.*#+}} xmm7 = xmm7[0,1,2,2,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm7[0,0,1,3]
@@ -271,9 +271,8 @@ define void @store_i8_stride5_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vecp
; SSE-NEXT: por %xmm4, %xmm7
; SSE-NEXT: movdqa {{.*#+}} xmm3 = [255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255]
; SSE-NEXT: pand %xmm3, %xmm7
-; SSE-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3],xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7]
-; SSE-NEXT: pshufhw {{.*#+}} xmm1 = xmm2[0,1,2,3,7,7,7,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm1[2,2,2,2]
+; SSE-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3],xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
+; SSE-NEXT: psrldq {{.*#+}} xmm1 = xmm1[11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
; SSE-NEXT: pandn %xmm1, %xmm3
; SSE-NEXT: por %xmm7, %xmm3
; SSE-NEXT: movdqa {{.*#+}} xmm1 = [255,255,0,255,255,255,255,0,255,255,255,255,255,255,255,255]
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-7.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-7.ll
index 844dc41240166..b7c514efff32f 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-7.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-7.ll
@@ -288,34 +288,35 @@ define void @store_i8_stride7_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vecp
; SSE-LABEL: store_i8_stride7_vf8:
; SSE: # %bb.0:
; SSE-NEXT: movq {{[0-9]+}}(%rsp), %rax
-; SSE-NEXT: movq {{.*#+}} xmm4 = mem[0],zero
+; SSE-NEXT: movq {{.*#+}} xmm0 = mem[0],zero
+; SSE-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movq {{.*#+}} xmm10 = mem[0],zero
-; SSE-NEXT: movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT: movq {{.*#+}} xmm3 = mem[0],zero
-; SSE-NEXT: movq {{.*#+}} xmm7 = mem[0],zero
-; SSE-NEXT: movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movq {{.*#+}} xmm1 = mem[0],zero
+; SSE-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movq {{.*#+}} xmm5 = mem[0],zero
; SSE-NEXT: movq {{.*#+}} xmm2 = mem[0],zero
; SSE-NEXT: movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; SSE-NEXT: movq {{.*#+}} xmm6 = mem[0],zero
-; SSE-NEXT: movq {{.*#+}} xmm5 = mem[0],zero
-; SSE-NEXT: punpcklqdq {{.*#+}} xmm5 = xmm5[0],xmm6[0]
-; SSE-NEXT: pshuflw {{.*#+}} xmm6 = xmm3[0,0,0,0,4,5,6,7]
-; SSE-NEXT: movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movq {{.*#+}} xmm4 = mem[0],zero
+; SSE-NEXT: punpcklqdq {{.*#+}} xmm4 = xmm4[0],xmm6[0]
+; SSE-NEXT: pshuflw {{.*#+}} xmm6 = xmm1[0,0,0,0,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm6[0,0,0,0]
; SSE-NEXT: movdqa {{.*#+}} xmm11 = [255,255,255,0,255,255,255,255,255,255,0,255,255,255,255,255]
; SSE-NEXT: pand %xmm11, %xmm6
-; SSE-NEXT: punpcklbw {{.*#+}} xmm7 = xmm7[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; SSE-NEXT: movdqa %xmm5, %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: pshuflw {{.*#+}} xmm8 = xmm7[0,0,2,1,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm8 = xmm8[0,1,1,3]
; SSE-NEXT: pandn %xmm8, %xmm11
; SSE-NEXT: por %xmm6, %xmm11
; SSE-NEXT: movdqa {{.*#+}} xmm9 = [255,255,0,0,255,255,255,255,255,0,0,255,255,255,255,255]
-; SSE-NEXT: pshuflw {{.*#+}} xmm6 = xmm4[0,0,2,1,4,5,6,7]
+; SSE-NEXT: pshuflw {{.*#+}} xmm6 = xmm0[0,0,2,1,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm6[0,0,2,1]
; SSE-NEXT: movdqa {{.*#+}} xmm8 = [255,0,255,255,255,255,255,255,0,255,255,255,255,255,255,0]
; SSE-NEXT: pand %xmm8, %xmm6
-; SSE-NEXT: punpcklbw {{.*#+}} xmm10 = xmm10[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE-NEXT: pshuflw {{.*#+}} xmm12 = xmm10[0,2,1,3,4,5,6,7]
+; SSE-NEXT: movdqa %xmm10, %xmm3
+; SSE-NEXT: punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm10[0],xmm3[1],xmm10[1],xmm3[2],xmm10[2],xmm3[3],xmm10[3],xmm3[4],xmm10[4],xmm3[5],xmm10[5],xmm3[6],xmm10[6],xmm3[7],xmm10[7]
+; SSE-NEXT: pshuflw {{.*#+}} xmm12 = xmm3[0,2,1,3,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm12 = xmm12[0,1,1,0]
; SSE-NEXT: movdqa %xmm8, %xmm13
; SSE-NEXT: pandn %xmm12, %xmm13
@@ -324,8 +325,9 @@ define void @store_i8_stride7_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vecp
; SSE-NEXT: pandn %xmm11, %xmm9
; SSE-NEXT: por %xmm13, %xmm9
; SSE-NEXT: pxor %xmm6, %xmm6
-; SSE-NEXT: movdqa %xmm5, %xmm12
-; SSE-NEXT: movdqa %xmm5, %xmm15
+; SSE-NEXT: movdqa %xmm4, %xmm15
+; SSE-NEXT: movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT: movdqa %xmm4, %xmm12
; SSE-NEXT: punpckhbw {{.*#+}} xmm15 = xmm15[8],xmm6[8],xmm15[9],xmm6[9],xmm15[10],xmm6[10],xmm15[11],xmm6[11],xmm15[12],xmm6[12],xmm15[13],xmm6[13],xmm15[14],xmm6[14],xmm15[15],xmm6[15]
; SSE-NEXT: punpcklbw {{.*#+}} xmm12 = xmm12[0],xmm6[0],xmm12[1],xmm6[1],xmm12[2],xmm6[2],xmm12[3],xmm6[3],xmm12[4],xmm6[4],xmm12[5],xmm6[5],xmm12[6],xmm6[6],xmm12[7],xmm6[7]
; SSE-NEXT: movdqa %xmm12, %xmm13
@@ -349,18 +351,20 @@ define void @store_i8_stride7_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vecp
; SSE-NEXT: pshufd {{.*#+}} xmm9 = xmm7[2,2,3,3]
; SSE-NEXT: movdqa %xmm1, %xmm11
; SSE-NEXT: pandn %xmm9, %xmm11
-; SSE-NEXT: pshuflw {{.*#+}} xmm9 = xmm3[2,1,3,3,4,5,6,7]
+; SSE-NEXT: pshuflw $246, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
+; SSE-NEXT: # xmm9 = mem[2,1,3,3,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm14 = xmm9[0,0,2,1]
; SSE-NEXT: pand %xmm1, %xmm14
; SSE-NEXT: por %xmm11, %xmm14
; SSE-NEXT: movdqa {{.*#+}} xmm9 = [255,255,255,255,255,0,0,255,255,255,255,255,0,0,255,255]
; SSE-NEXT: movdqa %xmm9, %xmm11
; SSE-NEXT: pandn %xmm14, %xmm11
-; SSE-NEXT: pshufhw {{.*#+}} xmm14 = xmm10[0,1,2,3,5,6,6,7]
+; SSE-NEXT: pshufhw {{.*#+}} xmm14 = xmm3[0,1,2,3,5,6,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm14[2,2,2,2]
; SSE-NEXT: movdqa {{.*#+}} xmm0 = [255,255,255,255,0,255,255,255,255,255,255,0,255,255,255,255]
; SSE-NEXT: movdqa %xmm0, %xmm14
; SSE-NEXT: pandn %xmm2, %xmm14
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm4[0,2,2,3,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,1,1,3]
; SSE-NEXT: pand %xmm0, %xmm2
@@ -395,19 +399,19 @@ define void @store_i8_stride7_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vecp
; SSE-NEXT: pshufd {{.*#+}} xmm11 = xmm11[0,0,0,0]
; SSE-NEXT: pandn %xmm11, %xmm0
; SSE-NEXT: por %xmm2, %xmm0
-; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm10[0,1,2,2]
-; SSE-NEXT: pshuflw {{.*#+}} xmm10 = xmm4[1,1,2,3,4,5,6,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm10 = xmm10[0,0,2,1]
-; SSE-NEXT: pand %xmm1, %xmm10
+; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm3[0,1,2,2]
+; SSE-NEXT: pshuflw {{.*#+}} xmm3 = xmm4[1,1,2,3,4,5,6,7]
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,0,2,1]
+; SSE-NEXT: pand %xmm1, %xmm3
; SSE-NEXT: pandn %xmm2, %xmm1
-; SSE-NEXT: por %xmm10, %xmm1
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm3[1,1,2,2,4,5,6,7]
+; SSE-NEXT: por %xmm3, %xmm1
+; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm11 # 16-byte Reload
+; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm11[1,1,2,2,4,5,6,7]
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,0,2,1]
; SSE-NEXT: pand %xmm8, %xmm2
-; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm7[1,1,2,1]
-; SSE-NEXT: pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,7,5,6,4]
-; SSE-NEXT: pandn %xmm7, %xmm8
+; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm7[1,1,2,1]
+; SSE-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,7,5,6,4]
+; SSE-NEXT: pandn %xmm3, %xmm8
; SSE-NEXT: por %xmm2, %xmm8
; SSE-NEXT: pand %xmm9, %xmm8
; SSE-NEXT: pandn %xmm1, %xmm9
@@ -416,31 +420,27 @@ define void @store_i8_stride7_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vecp
; SSE-NEXT: pand %xmm1, %xmm9
; SSE-NEXT: pandn %xmm0, %xmm1
; SSE-NEXT: por %xmm1, %xmm9
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-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]
-; SSE-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,7,7,7,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
-; SSE-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1],xmm2[2],xmm4[2],xmm2[3],xmm4[3],xmm2[4],xmm4[4],xmm2[5],xmm4[5],xmm2[6],xmm4[6],xmm2[7],xmm4[7]
-; SSE-NEXT: movdqa {{.*#+}} xmm1 = [255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255]
-; SSE-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,7,7,7]
-; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[2,2,2,2]
-; SSE-NEXT: pand %xmm1, %xmm2
-; SSE-NEXT: pandn %xmm0, %xmm1
-; SSE-NEXT: pshufd {{.*#+}} xmm0 = xmm5[3,1,2,3]
-; SSE-NEXT: por %xmm2, %xmm1
-; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm13[3,3,3,3,4,5,6,7]
-; SSE-NEXT: movdqa {{.*#+}} xmm3 = [255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255]
-; SSE-NEXT: pand %xmm3, %xmm2
-; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[1,3,2,3,4,5,6,7]
-; SSE-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,1,3,1,4,5,6,7]
-; SSE-NEXT: pandn %xmm0, %xmm3
-; SSE-NEXT: por %xmm2, %xmm3
+; SSE-NEXT: punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm11[0],xmm5[1],xmm11[1],xmm5[2],xmm11[2],xmm5[3],xmm11[3],xmm5[4],xmm11[4],xmm5[5],xmm11[5],xmm5[6],xmm11[6],xmm5[7],xmm11[7]
+; SSE-NEXT: psrldq {{.*#+}} xmm5 = xmm5[11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; SSE-NEXT: movdqa {{.*#+}} xmm0 = [255,255,255,0,0,255,255,255,255,255,255,255,255,255,255,255]
+; SSE-NEXT: pandn %xmm5, %xmm0
+; SSE-NEXT: punpcklbw {{.*#+}} xmm10 = xmm10[0],xmm4[0],xmm10[1],xmm4[1],xmm10[2],xmm4[2],xmm10[3],xmm4[3],xmm10[4],xmm4[4],xmm10[5],xmm4[5],xmm10[6],xmm4[6],xmm10[7],xmm4[7]
+; SSE-NEXT: pshufd $231, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; SSE-NEXT: # xmm1 = mem[3,1,2,3]
+; SSE-NEXT: psrldq {{.*#+}} xmm10 = xmm10[13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; SSE-NEXT: por %xmm0, %xmm10
+; SSE-NEXT: pshuflw {{.*#+}} xmm0 = xmm13[3,3,3,3,4,5,6,7]
+; SSE-NEXT: movdqa {{.*#+}} xmm2 = [255,255,255,255,255,0,0,255,255,255,255,255,255,255,255,255]
+; SSE-NEXT: pand %xmm2, %xmm0
+; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[1,3,2,3,4,5,6,7]
+; SSE-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[0,1,3,1,4,5,6,7]
+; SSE-NEXT: pandn %xmm1, %xmm2
+; SSE-NEXT: por %xmm0, %xmm2
; SSE-NEXT: movdqa {{.*#+}} xmm0 = [0,255,255,255,255,0,0,0,255,255,255,255,255,255,255,255]
-; SSE-NEXT: pand %xmm0, %xmm1
-; SSE-NEXT: pandn %xmm3, %xmm0
-; SSE-NEXT: por %xmm1, %xmm0
+; SSE-NEXT: pand %xmm0, %xmm10
+; SSE-NEXT: pandn %xmm2, %xmm0
+; SSE-NEXT: por %xmm10, %xmm0
; SSE-NEXT: movq {{[0-9]+}}(%rsp), %rax
; SSE-NEXT: movq %xmm0, 48(%rax)
; SSE-NEXT: movdqa %xmm9, 16(%rax)
diff --git a/llvm/test/CodeGen/X86/vector-shuffle-256-v16.ll b/llvm/test/CodeGen/X86/vector-shuffle-256-v16.ll
index 87103774ce0a7..74cd5dd8f86f1 100644
--- a/llvm/test/CodeGen/X86/vector-shuffle-256-v16.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-256-v16.ll
@@ -5155,9 +5155,10 @@ define <16 x i16> @shuffle_v16i16_00_16_01_17_02_18_03_27_08_24_09_25_10_26_11_2
;
; AVX2-LABEL: shuffle_v16i16_00_16_01_17_02_18_03_27_08_24_09_25_10_26_11_27:
; AVX2: # %bb.0:
-; AVX2-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; AVX2-NEXT: vpshufb {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,14,15,u,u,u,u,u,u,u,u,16,17,18,19,20,21,22,23,u,u,u,u,u,u,u,u]
; AVX2-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11]
+; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
+; AVX2-NEXT: vpblendw {{.*#+}} xmm1 = xmm0[0,1,2,3,4,5,6],xmm1[7]
+; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-NEXT: retq
;
; AVX512VL-LABEL: shuffle_v16i16_00_16_01_17_02_18_03_27_08_24_09_25_10_26_11_27:
@@ -5178,9 +5179,10 @@ define <16 x i16> @shuffle_v16i16_00_16_01_17_02_18_03_27_08_24_09_25_10_26_11_2
;
; XOPAVX2-LABEL: shuffle_v16i16_00_16_01_17_02_18_03_27_08_24_09_25_10_26_11_27:
; XOPAVX2: # %bb.0:
-; XOPAVX2-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
-; XOPAVX2-NEXT: vpshufb {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,14,15,u,u,u,u,u,u,u,u,16,17,18,19,20,21,22,23,u,u,u,u,u,u,u,u]
; XOPAVX2-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11]
+; XOPAVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
+; XOPAVX2-NEXT: vpblendw {{.*#+}} xmm1 = xmm0[0,1,2,3,4,5,6],xmm1[7]
+; XOPAVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; XOPAVX2-NEXT: retq
%shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 27, i32 8, i32 24, i32 9, i32 25, i32 10, i32 26, i32 11, i32 27>
ret <16 x i16> %shuffle
@@ -5248,10 +5250,10 @@ define <16 x i16> @shuffle_v16i16_04_20_05_21_06_22_07_31_12_28_13_29_14_30_15_3
;
; AVX2-LABEL: shuffle_v16i16_04_20_05_21_06_22_07_31_12_28_13_29_14_30_15_31:
; AVX2: # %bb.0:
-; AVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[2,3,2,3,6,7,6,7]
-; AVX2-NEXT: vpermq {{.*#+}} ymm1 = ymm1[1,3,3,3]
-; AVX2-NEXT: vpshufb {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,14,15,u,u,u,u,u,u,u,u,16,17,18,19,20,21,22,23,u,u,u,u,u,u,u,u]
-; AVX2-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11]
+; AVX2-NEXT: vpunpckhwd {{.*#+}} ymm0 = ymm0[4],ymm1[4],ymm0[5],ymm1[5],ymm0[6],ymm1[6],ymm0[7],ymm1[7],ymm0[12],ymm1[12],ymm0[13],ymm1[13],ymm0[14],ymm1[14],ymm0[15],ymm1[15]
+; AVX2-NEXT: vextracti128 $1, %ymm1, %xmm1
+; AVX2-NEXT: vpblendw {{.*#+}} xmm1 = xmm0[0,1,2,3,4,5,6],xmm1[7]
+; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-NEXT: retq
;
; AVX512VL-LABEL: shuffle_v16i16_04_20_05_21_06_22_07_31_12_28_13_29_14_30_15_31:
@@ -5272,10 +5274,10 @@ define <16 x i16> @shuffle_v16i16_04_20_05_21_06_22_07_31_12_28_13_29_14_30_15_3
;
; XOPAVX2-LABEL: shuffle_v16i16_04_20_05_21_06_22_07_31_12_28_13_29_14_30_15_31:
; XOPAVX2: # %bb.0:
-; XOPAVX2-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[2,3,2,3,6,7,6,7]
-; XOPAVX2-NEXT: vpermq {{.*#+}} ymm1 = ymm1[1,3,3,3]
-; XOPAVX2-NEXT: vpshufb {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5,14,15,u,u,u,u,u,u,u,u,16,17,18,19,20,21,22,23,u,u,u,u,u,u,u,u]
-; XOPAVX2-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11]
+; XOPAVX2-NEXT: vpunpckhwd {{.*#+}} ymm0 = ymm0[4],ymm1[4],ymm0[5],ymm1[5],ymm0[6],ymm1[6],ymm0[7],ymm1[7],ymm0[12],ymm1[12],ymm0[13],ymm1[13],ymm0[14],ymm1[14],ymm0[15],ymm1[15]
+; XOPAVX2-NEXT: vextracti128 $1, %ymm1, %xmm1
+; XOPAVX2-NEXT: vpblendw {{.*#+}} xmm1 = xmm0[0,1,2,3,4,5,6],xmm1[7]
+; XOPAVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; XOPAVX2-NEXT: retq
%shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 4, i32 20, i32 5, i32 21, i32 6, i32 22, i32 7, i32 31, i32 12, i32 28, i32 13, i32 29, i32 14, i32 30, i32 15, i32 31>
ret <16 x i16> %shuffle
@@ -5548,9 +5550,10 @@ define <16 x i16> @shuffle_v16i16_16_00_17_01_18_02_19_11_24_08_25_09_26_10_27_1
;
; AVX2-LABEL: shuffle_v16i16_16_00_17_01_18_02_19_11_24_08_25_09_26_10_27_11:
; AVX2: # %bb.0:
-; AVX2-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-NEXT: vpshufb {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,14,15,u,u,u,u,u,u,u,u,16,17,18,19,20,21,22,23,u,u,u,u,u,u,u,u]
; AVX2-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[1],ymm0[1],ymm1[2],ymm0[2],ymm1[3],ymm0[3],ymm1[8],ymm0[8],ymm1[9],ymm0[9],ymm1[10],ymm0[10],ymm1[11],ymm0[11]
+; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
+; AVX2-NEXT: vpblendw {{.*#+}} xmm1 = xmm0[0,1,2,3,4,5,6],xmm1[7]
+; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; AVX2-NEXT: retq
;
; AVX512VL-LABEL: shuffle_v16i16_16_00_17_01_18_02_19_11_24_08_25_09_26_10_27_11:
@@ -5572,9 +5575,10 @@ define <16 x i16> @shuffle_v16i16_16_00_17_01_18_02_19_11_24_08_25_09_26_10_27_1
;
; XOPAVX2-LABEL: shuffle_v16i16_16_00_17_01_18_02_19_11_24_08_25_09_26_10_27_11:
; XOPAVX2: # %bb.0:
-; XOPAVX2-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; XOPAVX2-NEXT: vpshufb {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,14,15,u,u,u,u,u,u,u,u,16,17,18,19,20,21,22,23,u,u,u,u,u,u,u,u]
; XOPAVX2-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[1],ymm0[1],ymm1[2],ymm0[2],ymm1[3],ymm0[3],ymm1[8],ymm0[8],ymm1[9],ymm0[9],ymm1[10],ymm0[10],ymm1[11],ymm0[11]
+; XOPAVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
+; XOPAVX2-NEXT: vpblendw {{.*#+}} xmm1 = xmm0[0,1,2,3,4,5,6],xmm1[7]
+; XOPAVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
; XOPAVX2-NEXT: retq
%shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 16, i32 0, i32 17, i32 1, i32 18, i32 2, i32 19, i32 11, i32 24, i32 8, i32 25, i32 9, i32 26, i32 10, i32 27, i32 11>
ret <16 x i16> %shuffle
@@ -5593,10 +5597,10 @@ define <16 x i16> @shuffle_v16i16_20_04_21_05_22_06_23_15_28_12_29_13_30_14_31_1
;
; AVX2-LABEL: shuffle_v16i16_20_04_21_05_22_06_23_15_28_12_29_13_30_14_31_15:
; AVX2: # %bb.0:
-; AVX2-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[2,3,2,3,6,7,6,7]
-; AVX2-NEXT: vpermq {{.*#+}} ymm0 = ymm0[1,3,3,3]
-; AVX2-NEXT: vpshufb {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,14,15,u,u,u,u,u,u,u,u,16,17,18,19,20,21,22,23,u,u,u,u,u,u,u,u]
-; AVX2-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[1],ymm0[1],ymm1[2],ymm0[2],ymm1[3],ymm0[3],ymm1[8],ymm0[8],ymm1[9],ymm0[9],ymm1[10],ymm0[10],ymm1[11],ymm0[11]
+; AVX2-NEXT: vpunpckhwd {{.*#+}} ymm1 = ymm1[4],ymm0[4],ymm1[5],ymm0[5],ymm1[6],ymm0[6],ymm1[7],ymm0[7],ymm1[12],ymm0[12],ymm1[13],ymm0[13],ymm1[14],ymm0[14],ymm1[15],ymm0[15]
+; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm0
+; AVX2-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3,4,5,6],xmm0[7]
+; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; AVX2-NEXT: retq
;
; AVX512VL-LABEL: shuffle_v16i16_20_04_21_05_22_06_23_15_28_12_29_13_30_14_31_15:
@@ -5618,10 +5622,10 @@ define <16 x i16> @shuffle_v16i16_20_04_21_05_22_06_23_15_28_12_29_13_30_14_31_1
;
; XOPAVX2-LABEL: shuffle_v16i16_20_04_21_05_22_06_23_15_28_12_29_13_30_14_31_15:
; XOPAVX2: # %bb.0:
-; XOPAVX2-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[2,3,2,3,6,7,6,7]
-; XOPAVX2-NEXT: vpermq {{.*#+}} ymm0 = ymm0[1,3,3,3]
-; XOPAVX2-NEXT: vpshufb {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5,14,15,u,u,u,u,u,u,u,u,16,17,18,19,20,21,22,23,u,u,u,u,u,u,u,u]
-; XOPAVX2-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[1],ymm0[1],ymm1[2],ymm0[2],ymm1[3],ymm0[3],ymm1[8],ymm0[8],ymm1[9],ymm0[9],ymm1[10],ymm0[10],ymm1[11],ymm0[11]
+; XOPAVX2-NEXT: vpunpckhwd {{.*#+}} ymm1 = ymm1[4],ymm0[4],ymm1[5],ymm0[5],ymm1[6],ymm0[6],ymm1[7],ymm0[7],ymm1[12],ymm0[12],ymm1[13],ymm0[13],ymm1[14],ymm0[14],ymm1[15],ymm0[15]
+; XOPAVX2-NEXT: vextracti128 $1, %ymm0, %xmm0
+; XOPAVX2-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3,4,5,6],xmm0[7]
+; XOPAVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; XOPAVX2-NEXT: retq
%shuffle = shufflevector <16 x i16> %a, <16 x i16> %b, <16 x i32> <i32 20, i32 4, i32 21, i32 5, i32 22, i32 6, i32 23, i32 15, i32 28, i32 12, i32 29, i32 13, i32 30, i32 14, i32 31, i32 15>
ret <16 x i16> %shuffle
diff --git a/llvm/test/CodeGen/X86/vector-shuffle-256-v8.ll b/llvm/test/CodeGen/X86/vector-shuffle-256-v8.ll
index c43cc6718d8e1..ec55616331ae4 100644
--- a/llvm/test/CodeGen/X86/vector-shuffle-256-v8.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-256-v8.ll
@@ -1335,15 +1335,10 @@ define <8 x float> @shuffle_v8f32_a2u3e6f7(<8 x float> %a, <8 x float> %b) {
define <8 x float> @shuffle_v8f32_084c195d(<8 x float> %a, <8 x float> %b) {
; AVX1-LABEL: shuffle_v8f32_084c195d:
; AVX1: # %bb.0:
-; AVX1-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm1[2,3,0,1]
-; AVX1-NEXT: vshufps {{.*#+}} ymm2 = ymm2[0,1,2,0,4,5,6,4]
-; AVX1-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,0,2,1,4,4,6,5]
-; AVX1-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3,4,5],ymm1[6,7]
-; AVX1-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm0[2,3,0,1]
-; AVX1-NEXT: vshufps {{.*#+}} ymm2 = ymm2[1,1,0,3,5,5,4,7]
-; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,1,1,3,4,5,5,7]
-; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3,4,5],ymm0[6,7]
-; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
+; AVX1-NEXT: vunpcklps {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm1
+; AVX1-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
+; AVX1-NEXT: vshufpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[3],ymm0[3]
; AVX1-NEXT: retq
;
; AVX2-LABEL: shuffle_v8f32_084c195d:
More information about the llvm-commits
mailing list