[DAGCombiner][X86] `visitVECTOR_SHUFFLE()`: splats with a single non-undef element are not splats

[lebedev.ri]

Summary

This addresses a significant portion of regressions that would otherwise appear in D140677.

@RKSimon this seems obviously good in general overall,
but there are some dubious changes here, at mostly for SSE2:
we fail to simplify some and/andn masks,
and pull identical target-specific shuffles out of commutative opcodes.

Please can you indicate which of the test changes should be dealt with?

[RKSimon]

Cheers - I'll review the x86 diffs and get back to you

llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp

23792

Instead of the count() - it might be worth adding a bool arg to isSplat()/isSplatMask() to only match splats with more than a single matching element?

[lebedev.ri]

In D140811#4021032, @RKSimon wrote:

Cheers - I'll review the x86 diffs and get back to you

Thank you for taking a look!
I've de-shedded a number of regressions off of D140677,
but the last few remaining are a last few for a reason :S

llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp

23792

Sure, and i think that should even be default.

[RKSimon]

Sorry for the delay - I'm slowly returning from Christmas break.

I think if we can address the horizontal-op regressions then the other cases are pretty trivial to address - tweaking canonicalizeShuffleMaskWithHorizOp might help us.

llvm/test/CodeGen/X86/haddsub-3.ll

7

(trivial) Looks like you can add back the baseline common AVX prefix to cover both AVX1/AVX2 checks

[lebedev.ri]

In D140811#4023270, @RKSimon wrote:

Sorry for the delay - I'm slowly returning from Christmas break.

I think if we can address the horizontal-op regressions

Err. As far as i was concerned, the horisontal-op changes *weren't* regressions,
because they all happen for the run lines where horisontal math is slow.
Do we actually want those to remain HOps?

then the other cases are pretty trivial to address - tweaking canonicalizeShuffleMaskWithHorizOp might help us.

[lebedev.ri]

llvm/test/CodeGen/X86/haddsub-undef.ll

471

@RKSimon these no longer match because previously we had shuffle-fadd-shuffle,
so isHorizontalBinOp() called shouldUseHorizontalOp() with IsSingleSource=false,
and now we have shuffle-fadd, and shouldUseHorizontalOp() is called with IsSingleSource=false.
So this really doesn't look like a regression to me.

[lebedev.ri]

llvm/test/CodeGen/X86/haddsub-undef.ll

471

and now we have shuffle-fadd, and shouldUseHorizontalOp() is called with IsSingleSource=true, of course that is

[lebedev.ri]

FYI, these are the regressions i was aware of.

llvm/test/CodeGen/X86/vector-narrow-binop.ll

112

This is all-ones mask

170

Failure to pull identical shuffle out of binop

llvm/test/CodeGen/X86/vector-shuffle-128-v4.ll

1540–1549

Missing lowering strategy?

llvm/test/CodeGen/X86/vector-shuffle-combining-avx2.ll

843

We intentionally lower it this way,
but given that we end with an extra shuffle, it may be a regression?

[lebedev.ri]

Rebased, no changes, NFC.

[RKSimon]

llvm/test/CodeGen/X86/haddsub-3.ll

7

why not just call the common prefix AVX? AVX1/AVX1-ONLY just seems to be confusing tbh.

[RKSimon]

llvm/test/CodeGen/X86/haddsub-undef.ll

471

yes - for slow hops, 2shuffles+alu is the threshold for using hop (unless building for size)

[RKSimon]

llvm/test/CodeGen/X86/horizontal-sum.ll

66

regression - we've gone from 3hops to 4hops + extra shuffles

[lebedev.ri]

llvm/test/CodeGen/X86/haddsub-undef.ll

471

So most of these hop changes are not regressions, correct?

[lebedev.ri]

llvm/test/CodeGen/X86/horizontal-sum.ll

66

Ok, i'll start with this one i guess.
It also seems reasonably straight-forward,
at least as the first step we need to go from

    t23: v2f32 = vector_shuffle<1,u> t21, undef:v2f32
  t24: v2f32 = fadd t21, t23
    t33: v2f32 = vector_shuffle<1,u> t32, undef:v2f32
  t34: v2f32 = fadd t32, t33
t75: v4f32 = concat_vectors t24, t34

to (pseudocode)

i0: v4f32 = concat_vectors t21, t32
i1: v4f32 = vector_shuffle<1,u,3,u> i0, undef:v4f32
i2: v4f32 = fadd i1, i0

I'm guessing *just* folding concat_vectors of identical opcodes
to a single opcode of multiple concat_vectors may not be a win though,
and shuffles must be matched too. Not sure yet.

But, i'm getting mixed signals here.
@RKSimon Should this kind of straight-forward yak shaving be just committed, or submitted to phab first?

[RKSimon]

@lebedev.ri I'm sorry about the slow response - but its taking forever to catch up after the Christmas break

[lebedev.ri]

Rebased, NFC.
Same regression remains, and given that my attempts to improve other things aren't being well-received, i'm not sure if i should bothed.

[lebedev.ri]

llvm/test/CodeGen/X86/vector-narrow-binop.ll

112

Err, no, not all-ones mask. It masks away high half/byte of each i16 element.

[lebedev.ri]

llvm/test/CodeGen/X86/vector-narrow-binop.ll

170

X86's narrowShuffle() intentionally does this, we can't treat this as a general shuffle combining issue.
Do we have an inverse of canonicalizeShuffleWithBinOps()?

[RKSimon]

llvm/test/CodeGen/X86/haddsub-undef.ll

471

Agreed - all the haddsub-undef.ll changes are improvements or neutral.

llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp

   // i.e. Merge SVN(OtherSVN, N1) -> shuffle(SV0, SV1, Mask) iff Commute = false
   //      Merge SVN(N1, OtherSVN) -> shuffle(SV0, SV1, Mask') iff Commute = true
   auto MergeInnerShuffle =
       [NumElts, &VT](bool Commute, ShuffleVectorSDNode *SVN,
                      ShuffleVectorSDNode *OtherSVN, SDValue N1,
                      const TargetLowering &TLI, SDValue &SV0, SDValue &SV1,
                      SmallVectorImpl<int> &Mask) -> bool {
     // Don't try to fold splats; they're likely to simplify somehow, or they
-    // might be free.
-    if (OtherSVN->isSplat())
+    // might be free. But only if there are at least two non-undef indices.
+    if (OtherSVN->isSplat() &&
+        count(OtherSVN->getMask(), OtherSVN->getSplatIndex()) > 1)
       return false;
 
     SV0 = SV1 = SDValue();
     Mask.clear();
 
     for (unsigned i = 0; i != NumElts; ++i) {
       int Idx = SVN->getMaskElt(i);
       if (Idx < 0) {

llvm/test/CodeGen/AArch64/shuffles.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s -mtriple=aarch64--linux-gnu | FileCheck %s
 
 define <16 x i32> @test_shuf1(<16 x i32> %x, <16 x i32> %y) {
 ; CHECK-LABEL: test_shuf1:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ext v16.16b, v6.16b, v1.16b, #4
-; CHECK-NEXT:    dup v5.4s, v4.s[0]
-; CHECK-NEXT:    uzp1 v17.4s, v1.4s, v0.4s
-; CHECK-NEXT:    uzp2 v18.4s, v2.4s, v4.4s
+; CHECK-NEXT:    ext v5.16b, v6.16b, v1.16b, #4
+; CHECK-NEXT:    uzp1 v16.4s, v1.4s, v0.4s
+; CHECK-NEXT:    uzp2 v17.4s, v2.4s, v4.4s
+; CHECK-NEXT:    ext v6.16b, v6.16b, v4.16b, #12
+; CHECK-NEXT:    trn2 v4.4s, v1.4s, v5.4s
 ; CHECK-NEXT:    rev64 v3.4s, v7.4s
-; CHECK-NEXT:    trn2 v4.4s, v1.4s, v16.4s
-; CHECK-NEXT:    mov v5.s[0], v6.s[3]
-; CHECK-NEXT:    trn2 v1.4s, v17.4s, v1.4s
-; CHECK-NEXT:    trn1 v2.4s, v18.4s, v2.4s
+; CHECK-NEXT:    trn2 v1.4s, v16.4s, v1.4s
+; CHECK-NEXT:    trn1 v2.4s, v17.4s, v2.4s
 ; CHECK-NEXT:    mov v4.s[0], v7.s[1]
-; CHECK-NEXT:    mov v3.d[0], v5.d[0]
+; CHECK-NEXT:    mov v3.d[0], v6.d[0]
 ; CHECK-NEXT:    ext v1.16b, v0.16b, v1.16b, #12
 ; CHECK-NEXT:    mov v2.s[3], v7.s[0]
 ; CHECK-NEXT:    mov v0.16b, v4.16b
 ; CHECK-NEXT:    ret
   %s3 = shufflevector <16 x i32> %x, <16 x i32> %y, <16 x i32> <i32 29, i32 26, i32 7, i32 4, i32 3, i32 6, i32 5, i32 2, i32 9, i32 8, i32 17, i32 28, i32 27, i32 16, i32 31, i32 30>
   ret <16 x i32> %s3
 }
 

llvm/test/CodeGen/PowerPC/v8i16_scalar_to_vector_shuffle.ll

entry:
   store <8 x i16> %shuff, ptr undef
   ret void
 }
 
 define void @test_none_v4i32(ptr %ptr, ptr %ptr2, i8 %v3) local_unnamed_addr #0 {
 ; CHECK-LE-P8-LABEL: test_none_v4i32:
 ; CHECK-LE-P8:       # %bb.0: # %entry
 ; CHECK-LE-P8-NEXT:    addis r4, r2, .LCPI2_0@toc@ha
-; CHECK-LE-P8-NEXT:    mtvsrd v3, r5
+; CHECK-LE-P8-NEXT:    lxsiwzx v2, 0, r3
+; CHECK-LE-P8-NEXT:    mtvsrd v4, r5
 ; CHECK-LE-P8-NEXT:    addi r4, r4, .LCPI2_0@toc@l
 ; CHECK-LE-P8-NEXT:    lxvd2x vs0, 0, r4
-; CHECK-LE-P8-NEXT:    addis r4, r2, .LCPI2_1@toc@ha
-; CHECK-LE-P8-NEXT:    addi r4, r4, .LCPI2_1@toc@l
-; CHECK-LE-P8-NEXT:    xxswapd v2, vs0
-; CHECK-LE-P8-NEXT:    lxvd2x vs0, 0, r4
-; CHECK-LE-P8-NEXT:    vperm v2, v3, v3, v2
-; CHECK-LE-P8-NEXT:    lxsiwzx v3, 0, r3
-; CHECK-LE-P8-NEXT:    xxswapd v4, vs0
-; CHECK-LE-P8-NEXT:    vperm v2, v2, v3, v4
+; CHECK-LE-P8-NEXT:    xxswapd v3, vs0
+; CHECK-LE-P8-NEXT:    vperm v2, v4, v2, v3
 ; CHECK-LE-P8-NEXT:    xxswapd vs0, v2
 ; CHECK-LE-P8-NEXT:    stfdx f0, 0, r3
 ; CHECK-LE-P8-NEXT:    blr
 ;
 ; CHECK-LE-P9-LABEL: test_none_v4i32:
 ; CHECK-LE-P9:       # %bb.0: # %entry
-; CHECK-LE-P9-NEXT:    lxsiwzx v2, 0, r3
+; CHECK-LE-P9-NEXT:    lfiwzx f0, 0, r3
 ; CHECK-LE-P9-NEXT:    addis r3, r2, .LCPI2_0@toc@ha
-; CHECK-LE-P9-NEXT:    mtvsrd v3, r5
+; CHECK-LE-P9-NEXT:    mtfprd f1, r5
 ; CHECK-LE-P9-NEXT:    addi r3, r3, .LCPI2_0@toc@l
-; CHECK-LE-P9-NEXT:    lxv vs0, 0(r3)
-; CHECK-LE-P9-NEXT:    addis r3, r2, .LCPI2_1@toc@ha
-; CHECK-LE-P9-NEXT:    addi r3, r3, .LCPI2_1@toc@l
-; CHECK-LE-P9-NEXT:    lxv v4, 0(r3)
-; CHECK-LE-P9-NEXT:    xxperm v3, v3, vs0
-; CHECK-LE-P9-NEXT:    vperm v2, v3, v2, v4
-; CHECK-LE-P9-NEXT:    xxswapd vs0, v2
+; CHECK-LE-P9-NEXT:    lxv vs2, 0(r3)
+; CHECK-LE-P9-NEXT:    xxperm vs0, vs1, vs2
+; CHECK-LE-P9-NEXT:    xxswapd vs0, vs0
 ; CHECK-LE-P9-NEXT:    stfd f0, 0(r3)
 ; CHECK-LE-P9-NEXT:    blr
 ;
 ; CHECK-BE-P8-LABEL: test_none_v4i32:
 ; CHECK-BE-P8:       # %bb.0: # %entry
 ; CHECK-BE-P8-NEXT:    addis r4, r2, .LCPI2_0@toc@ha
-; CHECK-BE-P8-NEXT:    mtvsrwz v3, r5
+; CHECK-BE-P8-NEXT:    lxsiwzx v2, 0, r3
+; CHECK-BE-P8-NEXT:    mtvsrwz v4, r5
 ; CHECK-BE-P8-NEXT:    addi r4, r4, .LCPI2_0@toc@l
-; CHECK-BE-P8-NEXT:    lxvw4x v2, 0, r4
-; CHECK-BE-P8-NEXT:    addis r4, r2, .LCPI2_1@toc@ha
-; CHECK-BE-P8-NEXT:    addi r4, r4, .LCPI2_1@toc@l
-; CHECK-BE-P8-NEXT:    lxvw4x v4, 0, r4
-; CHECK-BE-P8-NEXT:    vperm v2, v3, v3, v2
-; CHECK-BE-P8-NEXT:    lxsiwzx v3, 0, r3
-; CHECK-BE-P8-NEXT:    vperm v2, v3, v2, v4
+; CHECK-BE-P8-NEXT:    lxvw4x v3, 0, r4
+; CHECK-BE-P8-NEXT:    vperm v2, v2, v4, v3
 ; CHECK-BE-P8-NEXT:    stxsdx v2, 0, r3
 ; CHECK-BE-P8-NEXT:    blr
 ;
 ; CHECK-BE-P9-LABEL: test_none_v4i32:
 ; CHECK-BE-P9:       # %bb.0: # %entry
-; CHECK-BE-P9-NEXT:    lxsiwzx v2, 0, r3
+; CHECK-BE-P9-NEXT:    lfiwzx f0, 0, r3
 ; CHECK-BE-P9-NEXT:    addis r3, r2, .LCPI2_0@toc@ha
-; CHECK-BE-P9-NEXT:    mtvsrwz v3, r5
+; CHECK-BE-P9-NEXT:    mtfprwz f1, r5
 ; CHECK-BE-P9-NEXT:    addi r3, r3, .LCPI2_0@toc@l
-; CHECK-BE-P9-NEXT:    lxv vs0, 0(r3)
-; CHECK-BE-P9-NEXT:    addis r3, r2, .LCPI2_1@toc@ha
-; CHECK-BE-P9-NEXT:    addi r3, r3, .LCPI2_1@toc@l
-; CHECK-BE-P9-NEXT:    lxv v4, 0(r3)
-; CHECK-BE-P9-NEXT:    xxperm v3, v3, vs0
-; CHECK-BE-P9-NEXT:    vperm v2, v2, v3, v4
-; CHECK-BE-P9-NEXT:    stxsd v2, 0(r3)
+; CHECK-BE-P9-NEXT:    lxv vs2, 0(r3)
+; CHECK-BE-P9-NEXT:    xxperm vs0, vs1, vs2
+; CHECK-BE-P9-NEXT:    stfd f0, 0(r3)
 ; CHECK-BE-P9-NEXT:    blr
 ;
 ; CHECK-AIX-64-P8-LABEL: test_none_v4i32:
 ; CHECK-AIX-64-P8:       # %bb.0: # %entry
 ; CHECK-AIX-64-P8-NEXT:    ld r4, L..C3(r2) # %const.0
-; CHECK-AIX-64-P8-NEXT:    mtvsrwz v3, r5
-; CHECK-AIX-64-P8-NEXT:    lxvw4x v2, 0, r4
-; CHECK-AIX-64-P8-NEXT:    ld r4, L..C4(r2) # %const.1
-; CHECK-AIX-64-P8-NEXT:    vperm v2, v3, v3, v2
-; CHECK-AIX-64-P8-NEXT:    lxsiwzx v3, 0, r3
-; CHECK-AIX-64-P8-NEXT:    lxvw4x v4, 0, r4
-; CHECK-AIX-64-P8-NEXT:    vperm v2, v3, v2, v4
+; CHECK-AIX-64-P8-NEXT:    lxsiwzx v2, 0, r3
+; CHECK-AIX-64-P8-NEXT:    mtvsrwz v4, r5
+; CHECK-AIX-64-P8-NEXT:    lxvw4x v3, 0, r4
+; CHECK-AIX-64-P8-NEXT:    vperm v2, v2, v4, v3
 ; CHECK-AIX-64-P8-NEXT:    stxsdx v2, 0, r3
 ; CHECK-AIX-64-P8-NEXT:    blr
 ;
 ; CHECK-AIX-64-P9-LABEL: test_none_v4i32:
 ; CHECK-AIX-64-P9:       # %bb.0: # %entry
-; CHECK-AIX-64-P9-NEXT:    lxsiwzx v2, 0, r3
+; CHECK-AIX-64-P9-NEXT:    lfiwzx f0, 0, r3
 ; CHECK-AIX-64-P9-NEXT:    ld r3, L..C2(r2) # %const.0
-; CHECK-AIX-64-P9-NEXT:    mtvsrwz v3, r5
-; CHECK-AIX-64-P9-NEXT:    lxv vs0, 0(r3)
-; CHECK-AIX-64-P9-NEXT:    ld r3, L..C3(r2) # %const.1
-; CHECK-AIX-64-P9-NEXT:    lxv v4, 0(r3)
-; CHECK-AIX-64-P9-NEXT:    xxperm v3, v3, vs0
-; CHECK-AIX-64-P9-NEXT:    vperm v2, v2, v3, v4
-; CHECK-AIX-64-P9-NEXT:    stxsd v2, 0(r3)
+; CHECK-AIX-64-P9-NEXT:    mtfprwz f1, r5
+; CHECK-AIX-64-P9-NEXT:    lxv vs2, 0(r3)
+; CHECK-AIX-64-P9-NEXT:    xxperm vs0, vs1, vs2
+; CHECK-AIX-64-P9-NEXT:    stfd f0, 0(r3)
 ; CHECK-AIX-64-P9-NEXT:    blr
 ;
 ; CHECK-AIX-32-P8-LABEL: test_none_v4i32:
 ; CHECK-AIX-32-P8:       # %bb.0: # %entry
-; CHECK-AIX-32-P8-NEXT:    addi r4, r1, -32
+; CHECK-AIX-32-P8-NEXT:    lwz r4, L..C3(r2) # %const.0
 ; CHECK-AIX-32-P8-NEXT:    lxsiwzx v2, 0, r3
 ; CHECK-AIX-32-P8-NEXT:    stb r5, -32(r1)
-; CHECK-AIX-32-P8-NEXT:    lwz r3, L..C3(r2) # %const.0
-; CHECK-AIX-32-P8-NEXT:    lxvw4x v3, 0, r4
-; CHECK-AIX-32-P8-NEXT:    lxvw4x v4, 0, r3
+; CHECK-AIX-32-P8-NEXT:    addi r3, r1, -32
+; CHECK-AIX-32-P8-NEXT:    lxvw4x v3, 0, r3
 ; CHECK-AIX-32-P8-NEXT:    addi r3, r1, -16
-; CHECK-AIX-32-P8-NEXT:    vmrghh v3, v3, v3
+; CHECK-AIX-32-P8-NEXT:    lxvw4x v4, 0, r4
 ; CHECK-AIX-32-P8-NEXT:    vperm v2, v2, v3, v4
 ; CHECK-AIX-32-P8-NEXT:    stxvw4x v2, 0, r3
 ; CHECK-AIX-32-P8-NEXT:    lwz r3, -12(r1)
 ; CHECK-AIX-32-P8-NEXT:    stw r3, 0(r3)
 ; CHECK-AIX-32-P8-NEXT:    lwz r3, -16(r1)
 ; CHECK-AIX-32-P8-NEXT:    stw r3, 0(r3)
 ; CHECK-AIX-32-P8-NEXT:    blr
 ;
 ; CHECK-AIX-32-P9-LABEL: test_none_v4i32:
 ; CHECK-AIX-32-P9:       # %bb.0: # %entry
-; CHECK-AIX-32-P9-NEXT:    lxsiwzx v2, 0, r3
+; CHECK-AIX-32-P9-NEXT:    lfiwzx f0, 0, r3
 ; CHECK-AIX-32-P9-NEXT:    lwz r3, L..C2(r2) # %const.0
 ; CHECK-AIX-32-P9-NEXT:    stb r5, -32(r1)
-; CHECK-AIX-32-P9-NEXT:    lxv v3, -32(r1)
-; CHECK-AIX-32-P9-NEXT:    lxv v4, 0(r3)
-; CHECK-AIX-32-P9-NEXT:    vmrghh v3, v3, v3
-; CHECK-AIX-32-P9-NEXT:    vperm v2, v2, v3, v4
-; CHECK-AIX-32-P9-NEXT:    stxv v2, -16(r1)
+; CHECK-AIX-32-P9-NEXT:    lxv vs1, -32(r1)
+; CHECK-AIX-32-P9-NEXT:    lxv vs2, 0(r3)
+; CHECK-AIX-32-P9-NEXT:    xxperm vs1, vs0, vs2
+; CHECK-AIX-32-P9-NEXT:    stxv vs1, -16(r1)
 ; CHECK-AIX-32-P9-NEXT:    lwz r3, -12(r1)
 ; CHECK-AIX-32-P9-NEXT:    stw r3, 0(r3)
 ; CHECK-AIX-32-P9-NEXT:    lwz r3, -16(r1)
 ; CHECK-AIX-32-P9-NEXT:    stw r3, 0(r3)
 ; CHECK-AIX-32-P9-NEXT:    blr
 entry:
   %0 = load <2 x i16>, ptr %ptr, align 4
   %tmp = insertelement <4 x i8> undef, i8 %v3, i32 0

 ; CHECK-BE-P9-NEXT:    addi r3, r3, .LCPI3_0@toc@l
 ; CHECK-BE-P9-NEXT:    lxv vs1, 0(r3)
 ; CHECK-BE-P9-NEXT:    xxperm vs0, vs2, vs1
 ; CHECK-BE-P9-NEXT:    stxv vs0, 0(r3)
 ; CHECK-BE-P9-NEXT:    blr
 ;
 ; CHECK-AIX-64-P8-LABEL: test_v4i32_none:
 ; CHECK-AIX-64-P8:       # %bb.0: # %entry
-; CHECK-AIX-64-P8-NEXT:    ld r4, L..C5(r2) # %const.0
+; CHECK-AIX-64-P8-NEXT:    ld r4, L..C4(r2) # %const.0
 ; CHECK-AIX-64-P8-NEXT:    lxsiwzx v2, 0, r3
 ; CHECK-AIX-64-P8-NEXT:    xxlxor v4, v4, v4
 ; CHECK-AIX-64-P8-NEXT:    lxvw4x v3, 0, r4
 ; CHECK-AIX-64-P8-NEXT:    vperm v2, v4, v2, v3
 ; CHECK-AIX-64-P8-NEXT:    stxvw4x v2, 0, r3
 ; CHECK-AIX-64-P8-NEXT:    blr
 ;
 ; CHECK-AIX-64-P9-LABEL: test_v4i32_none:
 ; CHECK-AIX-64-P9:       # %bb.0: # %entry
 ; CHECK-AIX-64-P9-NEXT:    lfiwzx f0, 0, r3
-; CHECK-AIX-64-P9-NEXT:    ld r3, L..C4(r2) # %const.0
+; CHECK-AIX-64-P9-NEXT:    ld r3, L..C3(r2) # %const.0
 ; CHECK-AIX-64-P9-NEXT:    xxlxor vs2, vs2, vs2
 ; CHECK-AIX-64-P9-NEXT:    lxv vs1, 0(r3)
 ; CHECK-AIX-64-P9-NEXT:    xxperm vs0, vs2, vs1
 ; CHECK-AIX-64-P9-NEXT:    stxv vs0, 0(r3)
 ; CHECK-AIX-64-P9-NEXT:    blr
 ;
 ; CHECK-AIX-32-P8-LABEL: test_v4i32_none:
 ; CHECK-AIX-32-P8:       # %bb.0: # %entry

 ; CHECK-BE-P9-NEXT:    lxv vs1, 0(r3)
 ; CHECK-BE-P9-NEXT:    xxperm v2, vs0, vs1
 ; CHECK-BE-P9-NEXT:    vmrghh v2, v3, v2
 ; CHECK-BE-P9-NEXT:    stxv v2, 0(r3)
 ; CHECK-BE-P9-NEXT:    blr
 ;
 ; CHECK-AIX-64-P8-LABEL: test_none_v2i64:
 ; CHECK-AIX-64-P8:       # %bb.0: # %entry
-; CHECK-AIX-64-P8-NEXT:    ld r5, L..C6(r2) # %const.0
+; CHECK-AIX-64-P8-NEXT:    ld r5, L..C5(r2) # %const.0
 ; CHECK-AIX-64-P8-NEXT:    lxsdx v2, 0, r3
 ; CHECK-AIX-64-P8-NEXT:    lxvw4x v3, 0, r4
 ; CHECK-AIX-64-P8-NEXT:    lxvw4x v4, 0, r5
 ; CHECK-AIX-64-P8-NEXT:    vperm v2, v3, v2, v4
 ; CHECK-AIX-64-P8-NEXT:    xxlxor v3, v3, v3
 ; CHECK-AIX-64-P8-NEXT:    vmrghh v2, v3, v2
 ; CHECK-AIX-64-P8-NEXT:    stxvw4x v2, 0, r3
 ; CHECK-AIX-64-P8-NEXT:    blr
 ;
 ; CHECK-AIX-64-P9-LABEL: test_none_v2i64:
 ; CHECK-AIX-64-P9:       # %bb.0: # %entry
 ; CHECK-AIX-64-P9-NEXT:    lxsd v2, 0(r3)
-; CHECK-AIX-64-P9-NEXT:    ld r3, L..C5(r2) # %const.0
+; CHECK-AIX-64-P9-NEXT:    ld r3, L..C4(r2) # %const.0
 ; CHECK-AIX-64-P9-NEXT:    lxv vs0, 0(r4)
 ; CHECK-AIX-64-P9-NEXT:    xxlxor v3, v3, v3
 ; CHECK-AIX-64-P9-NEXT:    lxv vs1, 0(r3)
 ; CHECK-AIX-64-P9-NEXT:    xxperm v2, vs0, vs1
 ; CHECK-AIX-64-P9-NEXT:    vmrghh v2, v3, v2
 ; CHECK-AIX-64-P9-NEXT:    stxv v2, 0(r3)
 ; CHECK-AIX-64-P9-NEXT:    blr
 ;

 ; CHECK-BE-P9-NEXT:    addi r3, r3, .LCPI5_0@toc@l
 ; CHECK-BE-P9-NEXT:    lxv vs1, 0(r3)
 ; CHECK-BE-P9-NEXT:    xxperm vs0, vs2, vs1
 ; CHECK-BE-P9-NEXT:    stxv vs0, 0(r3)
 ; CHECK-BE-P9-NEXT:    blr
 ;
 ; CHECK-AIX-64-P8-LABEL: test_v2i64_none:
 ; CHECK-AIX-64-P8:       # %bb.0: # %entry
-; CHECK-AIX-64-P8-NEXT:    ld r4, L..C7(r2) # %const.0
+; CHECK-AIX-64-P8-NEXT:    ld r4, L..C6(r2) # %const.0
 ; CHECK-AIX-64-P8-NEXT:    lxsdx v2, 0, r3
 ; CHECK-AIX-64-P8-NEXT:    xxlxor v4, v4, v4
 ; CHECK-AIX-64-P8-NEXT:    lxvw4x v3, 0, r4
 ; CHECK-AIX-64-P8-NEXT:    vperm v2, v4, v2, v3
 ; CHECK-AIX-64-P8-NEXT:    stxvw4x v2, 0, r3
 ; CHECK-AIX-64-P8-NEXT:    blr
 ;
 ; CHECK-AIX-64-P9-LABEL: test_v2i64_none:
 ; CHECK-AIX-64-P9:       # %bb.0: # %entry
 ; CHECK-AIX-64-P9-NEXT:    lfd f0, 0(r3)
-; CHECK-AIX-64-P9-NEXT:    ld r3, L..C6(r2) # %const.0
+; CHECK-AIX-64-P9-NEXT:    ld r3, L..C5(r2) # %const.0
 ; CHECK-AIX-64-P9-NEXT:    xxlxor vs2, vs2, vs2
 ; CHECK-AIX-64-P9-NEXT:    lxv vs1, 0(r3)
 ; CHECK-AIX-64-P9-NEXT:    xxperm vs0, vs2, vs1
 ; CHECK-AIX-64-P9-NEXT:    stxv vs0, 0(r3)
 ; CHECK-AIX-64-P9-NEXT:    blr
 ;
 ; CHECK-AIX-32-P8-LABEL: test_v2i64_none:
 ; CHECK-AIX-32-P8:       # %bb.0: # %entry

 ; CHECK-BE-P9-NEXT:    lxsihzx v2, 0, r4
 ; CHECK-BE-P9-NEXT:    addi r3, r3, .LCPI6_0@toc@l
 ; CHECK-BE-P9-NEXT:    lxv vs1, 0(r3)
 ; CHECK-BE-P9-NEXT:    xxperm v2, vs0, vs1
 ; CHECK-BE-P9-NEXT:    blr
 ;
 ; CHECK-AIX-64-P8-LABEL: test_v8i16_v8i16:
 ; CHECK-AIX-64-P8:       # %bb.0: # %entry
-; CHECK-AIX-64-P8-NEXT:    ld r5, L..C8(r2) # %const.0
+; CHECK-AIX-64-P8-NEXT:    ld r5, L..C7(r2) # %const.0
 ; CHECK-AIX-64-P8-NEXT:    lhz r3, 0(r3)
 ; CHECK-AIX-64-P8-NEXT:    lhz r4, 0(r4)
 ; CHECK-AIX-64-P8-NEXT:    mtvsrwz v2, r3
 ; CHECK-AIX-64-P8-NEXT:    lxvw4x v3, 0, r5
 ; CHECK-AIX-64-P8-NEXT:    mtvsrwz v4, r4
 ; CHECK-AIX-64-P8-NEXT:    vperm v2, v2, v4, v3
 ; CHECK-AIX-64-P8-NEXT:    blr
 ;
 ; CHECK-AIX-64-P9-LABEL: test_v8i16_v8i16:
 ; CHECK-AIX-64-P9:       # %bb.0: # %entry
 ; CHECK-AIX-64-P9-NEXT:    lxsihzx f0, 0, r3
-; CHECK-AIX-64-P9-NEXT:    ld r3, L..C7(r2) # %const.0
+; CHECK-AIX-64-P9-NEXT:    ld r3, L..C6(r2) # %const.0
 ; CHECK-AIX-64-P9-NEXT:    lxsihzx v2, 0, r4
 ; CHECK-AIX-64-P9-NEXT:    lxv vs1, 0(r3)
 ; CHECK-AIX-64-P9-NEXT:    xxperm v2, vs0, vs1
 ; CHECK-AIX-64-P9-NEXT:    blr
 ;
 ; CHECK-AIX-32-P8-LABEL: test_v8i16_v8i16:
 ; CHECK-AIX-32-P8:       # %bb.0: # %entry
 ; CHECK-AIX-32-P8-NEXT:    lwz r5, L..C7(r2) # %const.0

 ; CHECK-BE-P9-NEXT:    lxv vs1, 0(r3)
 ; CHECK-BE-P9-NEXT:    xxperm v2, vs0, vs1
 ; CHECK-BE-P9-NEXT:    vmrghh v2, v3, v2
 ; CHECK-BE-P9-NEXT:    stxv v2, 0(r3)
 ; CHECK-BE-P9-NEXT:    blr
 ;
 ; CHECK-AIX-64-P8-LABEL: test_v4i32_v4i32:
 ; CHECK-AIX-64-P8:       # %bb.0: # %entry
-; CHECK-AIX-64-P8-NEXT:    ld r5, L..C9(r2) # %const.0
+; CHECK-AIX-64-P8-NEXT:    ld r5, L..C8(r2) # %const.0
 ; CHECK-AIX-64-P8-NEXT:    lxsiwzx v2, 0, r3
 ; CHECK-AIX-64-P8-NEXT:    lxsiwzx v3, 0, r4
 ; CHECK-AIX-64-P8-NEXT:    lxvw4x v4, 0, r5
 ; CHECK-AIX-64-P8-NEXT:    vperm v2, v3, v2, v4
 ; CHECK-AIX-64-P8-NEXT:    xxlxor v3, v3, v3
 ; CHECK-AIX-64-P8-NEXT:    vmrghh v2, v3, v2
 ; CHECK-AIX-64-P8-NEXT:    stxvw4x v2, 0, r3
 ; CHECK-AIX-64-P8-NEXT:    blr
 ;
 ; CHECK-AIX-64-P9-LABEL: test_v4i32_v4i32:
 ; CHECK-AIX-64-P9:       # %bb.0: # %entry
 ; CHECK-AIX-64-P9-NEXT:    lxsiwzx v2, 0, r3
-; CHECK-AIX-64-P9-NEXT:    ld r3, L..C8(r2) # %const.0
+; CHECK-AIX-64-P9-NEXT:    ld r3, L..C7(r2) # %const.0
 ; CHECK-AIX-64-P9-NEXT:    lfiwzx f0, 0, r4
 ; CHECK-AIX-64-P9-NEXT:    xxlxor v3, v3, v3
 ; CHECK-AIX-64-P9-NEXT:    lxv vs1, 0(r3)
 ; CHECK-AIX-64-P9-NEXT:    xxperm v2, vs0, vs1
 ; CHECK-AIX-64-P9-NEXT:    vmrghh v2, v3, v2
 ; CHECK-AIX-64-P9-NEXT:    stxv v2, 0(r3)
 ; CHECK-AIX-64-P9-NEXT:    blr
 ;

 ; CHECK-BE-P9-NEXT:    lxv vs1, 0(r3)
 ; CHECK-BE-P9-NEXT:    xxperm v2, vs0, vs1
 ; CHECK-BE-P9-NEXT:    vmrghh v2, v3, v2
 ; CHECK-BE-P9-NEXT:    stxv v2, 0(r3)
 ; CHECK-BE-P9-NEXT:    blr
 ;
 ; CHECK-AIX-64-P8-LABEL: test_v2i64_v2i64:
 ; CHECK-AIX-64-P8:       # %bb.0: # %entry
-; CHECK-AIX-64-P8-NEXT:    ld r5, L..C10(r2) # %const.0
+; CHECK-AIX-64-P8-NEXT:    ld r5, L..C9(r2) # %const.0
 ; CHECK-AIX-64-P8-NEXT:    lxsdx v2, 0, r3
 ; CHECK-AIX-64-P8-NEXT:    lxsdx v3, 0, r4
 ; CHECK-AIX-64-P8-NEXT:    lxvw4x v4, 0, r5
 ; CHECK-AIX-64-P8-NEXT:    vperm v2, v3, v2, v4
 ; CHECK-AIX-64-P8-NEXT:    xxlxor v3, v3, v3
 ; CHECK-AIX-64-P8-NEXT:    vmrghh v2, v3, v2
 ; CHECK-AIX-64-P8-NEXT:    stxvw4x v2, 0, r3
 ; CHECK-AIX-64-P8-NEXT:    blr
 ;
 ; CHECK-AIX-64-P9-LABEL: test_v2i64_v2i64:
 ; CHECK-AIX-64-P9:       # %bb.0: # %entry
 ; CHECK-AIX-64-P9-NEXT:    lxsd v2, 0(r3)
-; CHECK-AIX-64-P9-NEXT:    ld r3, L..C9(r2) # %const.0
+; CHECK-AIX-64-P9-NEXT:    ld r3, L..C8(r2) # %const.0
 ; CHECK-AIX-64-P9-NEXT:    lfd f0, 0(r4)
 ; CHECK-AIX-64-P9-NEXT:    xxlxor v3, v3, v3
 ; CHECK-AIX-64-P9-NEXT:    lxv vs1, 0(r3)
 ; CHECK-AIX-64-P9-NEXT:    xxperm v2, vs0, vs1
 ; CHECK-AIX-64-P9-NEXT:    vmrghh v2, v3, v2
 ; CHECK-AIX-64-P9-NEXT:    stxv v2, 0(r3)
 ; CHECK-AIX-64-P9-NEXT:    blr
 ;

llvm/test/CodeGen/X86/haddsub-3.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+sse2            | FileCheck %s --check-prefix=SSE2
 ; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+ssse3           | FileCheck %s --check-prefixes=SSSE3,SSSE3-SLOW
 ; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+ssse3,fast-hops | FileCheck %s --check-prefixes=SSSE3,SSSE3-FAST
 ; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+avx             | FileCheck %s --check-prefixes=AVX1,AVX1-SLOW
 ; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+avx,fast-hops   | FileCheck %s --check-prefixes=AVX1,AVX1-FAST
 ; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+avx2            | FileCheck %s --check-prefix=AVX2
 
 define float @pr26491(<4 x float> %a0) {
 ; SSE2-LABEL: pr26491:
 ; SSE2:       # %bb.0:
 ; SSE2-NEXT:    movaps %xmm0, %xmm1
 ; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,1],xmm0[3,3]
 ; SSE2-NEXT:    addps %xmm1, %xmm0
 ; SSE2-NEXT:    movaps %xmm0, %xmm1

 ; SSE2-LABEL: PR48823:
 ; SSE2:       # %bb.0:
 ; SSE2-NEXT:    movaps %xmm0, %xmm2
 ; SSE2-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,1],xmm1[2,3]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,2]
 ; SSE2-NEXT:    subps %xmm2, %xmm0
 ; SSE2-NEXT:    retq
 ;
-; SSSE3-SLOW-LABEL: PR48823:
-; SSSE3-SLOW:       # %bb.0:
-; SSSE3-SLOW-NEXT:    movaps %xmm0, %xmm2
-; SSSE3-SLOW-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,1],xmm1[2,3]
-; SSSE3-SLOW-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,2]
-; SSSE3-SLOW-NEXT:    subps %xmm2, %xmm0
-; SSSE3-SLOW-NEXT:    retq
-;
-; SSSE3-FAST-LABEL: PR48823:
-; SSSE3-FAST:       # %bb.0:
-; SSSE3-FAST-NEXT:    hsubps %xmm1, %xmm0
-; SSSE3-FAST-NEXT:    retq
-;
-; AVX1-SLOW-LABEL: PR48823:
-; AVX1-SLOW:       # %bb.0:
-; AVX1-SLOW-NEXT:    vshufps {{.*#+}} xmm2 = xmm0[1,1],xmm1[2,3]
-; AVX1-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,2]
-; AVX1-SLOW-NEXT:    vsubps %xmm2, %xmm0, %xmm0
-; AVX1-SLOW-NEXT:    retq
+; SSSE3-LABEL: PR48823:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    hsubps %xmm1, %xmm0
+; SSSE3-NEXT:    retq
 ;
-; AVX1-FAST-LABEL: PR48823:
-; AVX1-FAST:       # %bb.0:
-; AVX1-FAST-NEXT:    vhsubps %xmm1, %xmm0, %xmm0
-; AVX1-FAST-NEXT:    retq
+; AVX1-LABEL: PR48823:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vhsubps %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: PR48823:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vshufps {{.*#+}} xmm2 = xmm0[1,1],xmm1[2,3]
-; AVX2-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,2]
-; AVX2-NEXT:    vsubps %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vhsubps %xmm1, %xmm0, %xmm0
 ; AVX2-NEXT:    retq
   %3 = shufflevector <4 x float> %0, <4 x float> poison, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef>
   %4 = fsub <4 x float> %0, %3
   %5 = shufflevector <4 x float> %1, <4 x float> poison, <4 x i32> <i32 undef, i32 undef, i32 undef, i32 2>
   %6 = fsub <4 x float> %5, %1
   %7 = shufflevector <4 x float> %4, <4 x float> %6, <4 x i32> <i32 0, i32 undef, i32 undef, i32 7>
   ret <4 x float> %7
 }

llvm/test/CodeGen/X86/haddsub-undef.ll

 ; AVX-FAST-NEXT:    vhaddpd %xmm0, %xmm0, %xmm0
 ; AVX-FAST-NEXT:    retq
   %l = shufflevector <2 x double> %x, <2 x double> undef, <2 x i32> <i32 1, i32 0>
   %add = fadd <2 x double> %l, %x
   ret <2 x double> %add
 }
 
 define <2 x double> @add_pd_010(<2 x double> %x) {
-; SSE-LABEL: add_pd_010:
-; SSE:       # %bb.0:
-; SSE-NEXT:    haddpd %xmm0, %xmm0
-; SSE-NEXT:    retq
+; SSE-SLOW-LABEL: add_pd_010:
+; SSE-SLOW:       # %bb.0:
+; SSE-SLOW-NEXT:    movapd %xmm0, %xmm1
+; SSE-SLOW-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-SLOW-NEXT:    addpd %xmm1, %xmm0
+; SSE-SLOW-NEXT:    retq
+;
+; SSE-FAST-LABEL: add_pd_010:
+; SSE-FAST:       # %bb.0:
+; SSE-FAST-NEXT:    haddpd %xmm0, %xmm0
+; SSE-FAST-NEXT:    retq
 ;
 ; AVX-SLOW-LABEL: add_pd_010:
 ; AVX-SLOW:       # %bb.0:
-; AVX-SLOW-NEXT:    vmovddup {{.*#+}} xmm1 = xmm0[0,0]
-; AVX-SLOW-NEXT:    vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
-; AVX-SLOW-NEXT:    vaddpd %xmm0, %xmm1, %xmm0
+; AVX-SLOW-NEXT:    vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-SLOW-NEXT:    vaddpd %xmm1, %xmm0, %xmm0
 ; AVX-SLOW-NEXT:    retq
 ;
 ; AVX-FAST-LABEL: add_pd_010:
 ; AVX-FAST:       # %bb.0:
 ; AVX-FAST-NEXT:    vhaddpd %xmm0, %xmm0, %xmm0
 ; AVX-FAST-NEXT:    retq
   %l = shufflevector <2 x double> %x, <2 x double> undef, <2 x i32> <i32 undef, i32 0>
   %add = fadd <2 x double> %l, %x
; AVX-NEXT:    retq
   ret <4 x float> %9
 }
 
 define <4 x float> @add_ps_017(<4 x float> %x) {
 ; SSE-SLOW-LABEL: add_ps_017:
 ; SSE-SLOW:       # %bb.0:
 ; SSE-SLOW-NEXT:    movaps %xmm0, %xmm1
 ; SSE-SLOW-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,3],xmm0[3,3]
-; SSE-SLOW-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,2,2,2]
+; SSE-SLOW-NEXT:    movhlps {{.*#+}} xmm0 = xmm0[1,1]
 ; SSE-SLOW-NEXT:    addps %xmm1, %xmm0
 ; SSE-SLOW-NEXT:    retq
 ;
 ; SSE-FAST-LABEL: add_ps_017:
 ; SSE-FAST:       # %bb.0:
 ; SSE-FAST-NEXT:    haddps %xmm0, %xmm0
 ; SSE-FAST-NEXT:    movshdup {{.*#+}} xmm0 = xmm0[1,1,3,3]
 ; SSE-FAST-NEXT:    retq
 ;
 ; AVX-SLOW-LABEL: add_ps_017:
 ; AVX-SLOW:       # %bb.0:
 ; AVX-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[3,3,3,3]
-; AVX-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,2,2,2]
+; AVX-SLOW-NEXT:    vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
 ; AVX-SLOW-NEXT:    vaddps %xmm1, %xmm0, %xmm0
 ; AVX-SLOW-NEXT:    retq
 ;
 ; AVX-FAST-LABEL: add_ps_017:
 ; AVX-FAST:       # %bb.0:
 ; AVX-FAST-NEXT:    vhaddps %xmm0, %xmm0, %xmm0
 ; AVX-FAST-NEXT:    vmovshdup {{.*#+}} xmm0 = xmm0[1,1,3,3]
 ; AVX-FAST-NEXT:    retq
   %l = shufflevector <4 x float> %x, <4 x float> undef, <4 x i32> <i32 undef, i32 undef, i32 undef, i32 2>
   %add = fadd <4 x float> %l, %x
   %shuffle2 = shufflevector <4 x float> %add, <4 x float> undef, <4 x i32> <i32 3, i32 undef, i32 undef, i32 undef>
   ret <4 x float> %shuffle2
 }
 
 define <4 x float> @add_ps_018(<4 x float> %x) {
-; SSE-LABEL: add_ps_018:
-; SSE:       # %bb.0:
-; SSE-NEXT:    haddps %xmm0, %xmm0
-; SSE-NEXT:    movsldup {{.*#+}} xmm0 = xmm0[0,0,2,2]
-; SSE-NEXT:    retq
+; SSE-SLOW-LABEL: add_ps_018:
+; SSE-SLOW:       # %bb.0:
+; SSE-SLOW-NEXT:    movsldup {{.*#+}} xmm1 = xmm0[0,0,2,2]
+; SSE-SLOW-NEXT:    addps %xmm1, %xmm0
+; SSE-SLOW-NEXT:    retq
+;
+; SSE-FAST-LABEL: add_ps_018:
+; SSE-FAST:       # %bb.0:
+; SSE-FAST-NEXT:    haddps %xmm0, %xmm0
+; SSE-FAST-NEXT:    movsldup {{.*#+}} xmm0 = xmm0[0,0,2,2]
+; SSE-FAST-NEXT:    retq
 ;
 ; AVX1-SLOW-LABEL: add_ps_018:
 ; AVX1-SLOW:       # %bb.0:
-; AVX1-SLOW-NEXT:    vhaddps %xmm0, %xmm0, %xmm0
-; AVX1-SLOW-NEXT:    vmovsldup {{.*#+}} xmm0 = xmm0[0,0,2,2]
+; AVX1-SLOW-NEXT:    vmovsldup {{.*#+}} xmm1 = xmm0[0,0,2,2]
+; AVX1-SLOW-NEXT:    vaddps %xmm0, %xmm1, %xmm0
 ; AVX1-SLOW-NEXT:    retq
 ;
 ; AVX1-FAST-LABEL: add_ps_018:
 ; AVX1-FAST:       # %bb.0:
 ; AVX1-FAST-NEXT:    vhaddps %xmm0, %xmm0, %xmm0
 ; AVX1-FAST-NEXT:    vmovsldup {{.*#+}} xmm0 = xmm0[0,0,2,2]
 ; AVX1-FAST-NEXT:    retq
 ;
-; AVX512-LABEL: add_ps_018:
-; AVX512:       # %bb.0:
-; AVX512-NEXT:    vhaddps %xmm0, %xmm0, %xmm0
-; AVX512-NEXT:    vbroadcastss %xmm0, %xmm0
-; AVX512-NEXT:    retq
+; AVX512-SLOW-LABEL: add_ps_018:
+; AVX512-SLOW:       # %bb.0:
+; AVX512-SLOW-NEXT:    vbroadcastss %xmm0, %xmm1
+; AVX512-SLOW-NEXT:    vaddps %xmm0, %xmm1, %xmm0
+; AVX512-SLOW-NEXT:    retq
+;
+; AVX512-FAST-LABEL: add_ps_018:
+; AVX512-FAST:       # %bb.0:
+; AVX512-FAST-NEXT:    vhaddps %xmm0, %xmm0, %xmm0
+; AVX512-FAST-NEXT:    vbroadcastss %xmm0, %xmm0
+; AVX512-FAST-NEXT:    retq
   %l = shufflevector <4 x float> %x, <4 x float> undef, <4 x i32> <i32 undef, i32 undef, i32 0, i32 undef>
   %r = shufflevector <4 x float> %x, <4 x float> undef, <4 x i32> <i32 undef, i32 undef, i32 1, i32 undef>
   %add = fadd <4 x float> %l, %r
   %shuffle2 = shufflevector <4 x float> %add, <4 x float> undef, <4 x i32> <i32 undef, i32 2, i32 undef, i32 undef>
   ret <4 x float> %shuffle2
 }
 
 define <4 x double> @add_pd_011(<4 x double> %0, <4 x double> %1) {

 ; AVX512-NEXT:    retq
   %3 = shufflevector <4 x double> %0, <4 x double> %1, <4 x i32> <i32 undef, i32 2, i32 4, i32 6>
   %4 = shufflevector <4 x double> %0, <4 x double> %1, <4 x i32> <i32 undef, i32 3, i32 5, i32 7>
   %5 = fadd <4 x double> %3, %4
   ret <4 x double> %5
 }
 
 define <4 x float> @PR45747_1(<4 x float> %a, <4 x float> %b) nounwind {
-; SSE-SLOW-LABEL: PR45747_1:
-; SSE-SLOW:       # %bb.0:
-; SSE-SLOW-NEXT:    movaps %xmm0, %xmm1
-; SSE-SLOW-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,2],xmm0[2,2]
-; SSE-SLOW-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
-; SSE-SLOW-NEXT:    addps %xmm1, %xmm0
-; SSE-SLOW-NEXT:    retq
-;
-; SSE-FAST-LABEL: PR45747_1:
-; SSE-FAST:       # %bb.0:
-; SSE-FAST-NEXT:    haddps %xmm0, %xmm0
-; SSE-FAST-NEXT:    retq
-;
-; AVX-SLOW-LABEL: PR45747_1:
-; AVX-SLOW:       # %bb.0:
-; AVX-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,2,2,2]
-; AVX-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,3,3,3]
-; AVX-SLOW-NEXT:    vaddps %xmm1, %xmm0, %xmm0
-; AVX-SLOW-NEXT:    retq
+; SSE-LABEL: PR45747_1:
+; SSE:       # %bb.0:
+; SSE-NEXT:    haddps %xmm0, %xmm0
+; SSE-NEXT:    retq
 ;
-; AVX-FAST-LABEL: PR45747_1:
-; AVX-FAST:       # %bb.0:
-; AVX-FAST-NEXT:    vhaddps %xmm0, %xmm0, %xmm0
-; AVX-FAST-NEXT:    retq
+; AVX-LABEL: PR45747_1:
+; AVX:       # %bb.0:
+; AVX-NEXT:    vhaddps %xmm0, %xmm0, %xmm0
+; AVX-NEXT:    retq
   %t0 = shufflevector <4 x float> %a, <4 x float> undef, <4 x i32> <i32 undef, i32 undef, i32 3, i32 undef>
   %t1 = fadd <4 x float> %t0, %a
   %shuffle = shufflevector <4 x float> %t1, <4 x float> undef, <4 x i32> <i32 undef, i32 2, i32 undef, i32 undef>
   ret <4 x float> %shuffle
 }
 
 define <4 x float> @PR45747_2(<4 x float> %a, <4 x float> %b) nounwind {
 ; SSE-SLOW-LABEL: PR45747_2:
 ; SSE-SLOW:       # %bb.0:
 ; SSE-SLOW-NEXT:    movaps %xmm1, %xmm0
 ; SSE-SLOW-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
 ; SSE-SLOW-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
 ; SSE-SLOW-NEXT:    addps %xmm1, %xmm0
 ; SSE-SLOW-NEXT:    retq
 ;
 ; SSE-FAST-LABEL: PR45747_2:
 ; SSE-FAST:       # %bb.0:
 ; SSE-FAST-NEXT:    haddps %xmm1, %xmm1
 ; SSE-FAST-NEXT:    movshdup {{.*#+}} xmm0 = xmm1[1,1,3,3]
 ; SSE-FAST-NEXT:    retq
 ;
 ; AVX-SLOW-LABEL: PR45747_2:
 ; AVX-SLOW:       # %bb.0:
 ; AVX-SLOW-NEXT:    vpermilpd {{.*#+}} xmm0 = xmm1[1,0]
-; AVX-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,3,1,1]
+; AVX-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[3,3,3,3]
 ; AVX-SLOW-NEXT:    vaddps %xmm0, %xmm1, %xmm0
 ; AVX-SLOW-NEXT:    retq
 ;
 ; AVX-FAST-LABEL: PR45747_2:
 ; AVX-FAST:       # %bb.0:
 ; AVX-FAST-NEXT:    vhaddps %xmm1, %xmm1, %xmm0
 ; AVX-FAST-NEXT:    vmovshdup {{.*#+}} xmm0 = xmm0[1,1,3,3]
 ; AVX-FAST-NEXT:    retq
; AVX-NEXT:    retq
   %6 = shufflevector <2 x float> %5, <2 x float> undef, <4 x i32> <i32 undef, i32 0, i32 1, i32 undef>
   %7 = shufflevector <4 x float> %1, <4 x float> undef, <4 x i32> <i32 undef, i32 undef, i32 undef, i32 2>
   %8 = fadd <4 x float> %7, %1
   %9 = shufflevector <4 x float> %6, <4 x float> %8, <4 x i32> <i32 undef, i32 1, i32 2, i32 7>
   ret <4 x float> %9
 }
 
 define <4 x float> @PR34724_add_v4f32_0u23(<4 x float> %0, <4 x float> %1) {
-; SSE-SLOW-LABEL: PR34724_add_v4f32_0u23:
-; SSE-SLOW:       # %bb.0:
-; SSE-SLOW-NEXT:    movaps %xmm0, %xmm2
-; SSE-SLOW-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm1[0]
-; SSE-SLOW-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm1[1,1]
-; SSE-SLOW-NEXT:    addps %xmm2, %xmm0
-; SSE-SLOW-NEXT:    movsldup {{.*#+}} xmm2 = xmm1[0,0,2,2]
-; SSE-SLOW-NEXT:    addps %xmm1, %xmm2
-; SSE-SLOW-NEXT:    shufps {{.*#+}} xmm2 = xmm2[3,0],xmm0[2,0]
-; SSE-SLOW-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm2[2,0]
-; SSE-SLOW-NEXT:    retq
-;
-; SSE-FAST-LABEL: PR34724_add_v4f32_0u23:
-; SSE-FAST:       # %bb.0:
-; SSE-FAST-NEXT:    haddps %xmm1, %xmm0
-; SSE-FAST-NEXT:    retq
-;
-; AVX-SLOW-LABEL: PR34724_add_v4f32_0u23:
-; AVX-SLOW:       # %bb.0:
-; AVX-SLOW-NEXT:    vshufps {{.*#+}} xmm2 = xmm0[0,1],xmm1[0,3]
-; AVX-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,1],xmm1[1,2]
-; AVX-SLOW-NEXT:    vaddps %xmm2, %xmm0, %xmm0
-; AVX-SLOW-NEXT:    retq
+; SSE-LABEL: PR34724_add_v4f32_0u23:
+; SSE:       # %bb.0:
+; SSE-NEXT:    haddps %xmm1, %xmm0
+; SSE-NEXT:    retq
 ;
-; AVX-FAST-LABEL: PR34724_add_v4f32_0u23:
-; AVX-FAST:       # %bb.0:
-; AVX-FAST-NEXT:    vhaddps %xmm1, %xmm0, %xmm0
-; AVX-FAST-NEXT:    retq
+; AVX-LABEL: PR34724_add_v4f32_0u23:
+; AVX:       # %bb.0:
+; AVX-NEXT:    vhaddps %xmm1, %xmm0, %xmm0
+; AVX-NEXT:    retq
   %3 = shufflevector <4 x float> %0, <4 x float> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef>
   %4 = fadd <4 x float> %3, %0
   %5 = shufflevector <4 x float> %1, <4 x float> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef>
   %6 = fadd <4 x float> %5, %1
   %7 = shufflevector <4 x float> %4, <4 x float> %6, <4 x i32> <i32 0, i32 undef, i32 4, i32 undef>
   %8 = shufflevector <4 x float> %1, <4 x float> undef, <4 x i32> <i32 undef, i32 undef, i32 undef, i32 2>
   %9 = fadd <4 x float> %8, %1
   %10 = shufflevector <4 x float> %7, <4 x float> %9, <4 x i32> <i32 0, i32 undef, i32 2, i32 7>

llvm/test/CodeGen/X86/horizontal-sum.ll

 ; SSSE3-FAST-NEXT:    haddps %xmm2, %xmm0
 ; SSSE3-FAST-NEXT:    retq
 ;
 ; AVX1-SLOW-LABEL: pair_sum_v4f32_v4f32:
 ; AVX1-SLOW:       # %bb.0:
 ; AVX1-SLOW-NEXT:    vhaddps %xmm1, %xmm0, %xmm0
 ; AVX1-SLOW-NEXT:    vhaddps %xmm2, %xmm2, %xmm1
 ; AVX1-SLOW-NEXT:    vshufps {{.*#+}} xmm2 = xmm0[0,2],xmm1[0,1]
-; AVX1-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,3],xmm1[1,1]
+; AVX1-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,3],xmm1[1,3]
 ; AVX1-SLOW-NEXT:    vhaddps %xmm3, %xmm3, %xmm1
 ; AVX1-SLOW-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1,2],xmm1[3]
 ; AVX1-SLOW-NEXT:    vinsertps {{.*#+}} xmm1 = xmm2[0,1,2],xmm1[0]
 ; AVX1-SLOW-NEXT:    vaddps %xmm0, %xmm1, %xmm0
 ; AVX1-SLOW-NEXT:    retq
 ;
 ; AVX-FAST-LABEL: pair_sum_v4f32_v4f32:
 ; AVX-FAST:       # %bb.0:
 ; AVX-FAST-NEXT:    vhaddps %xmm1, %xmm0, %xmm0
 ; AVX-FAST-NEXT:    vhaddps %xmm3, %xmm2, %xmm1
 ; AVX-FAST-NEXT:    vhaddps %xmm1, %xmm0, %xmm0
 ; AVX-FAST-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: pair_sum_v4f32_v4f32:
 ; AVX2-SLOW:       # %bb.0:
 ; AVX2-SLOW-NEXT:    vhaddps %xmm1, %xmm0, %xmm0
+; AVX2-SLOW-NEXT:    vhaddps %xmm0, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vhaddps %xmm2, %xmm2, %xmm1
-; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm2 = xmm0[0,2],xmm1[0,3]
-; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,3],xmm1[1,3]
+; AVX2-SLOW-NEXT:    vmovshdup {{.*#+}} xmm2 = xmm1[1,1,3,3]
+; AVX2-SLOW-NEXT:    vaddps %xmm2, %xmm1, %xmm1
+; AVX2-SLOW-NEXT:    vmovlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
 ; AVX2-SLOW-NEXT:    vhaddps %xmm3, %xmm3, %xmm1
-; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1,2],xmm1[3]
-; AVX2-SLOW-NEXT:    vinsertps {{.*#+}} xmm1 = xmm2[0,1,2],xmm1[0]
-; AVX2-SLOW-NEXT:    vaddps %xmm0, %xmm1, %xmm0
+; AVX2-SLOW-NEXT:    vmovshdup {{.*#+}} xmm2 = xmm1[1,1,3,3]
+; AVX2-SLOW-NEXT:    vaddps %xmm2, %xmm1, %xmm1
+; AVX2-SLOW-NEXT:    vinsertps {{.*#+}} xmm0 = xmm0[0,1,2],xmm1[0]
 ; AVX2-SLOW-NEXT:    retq
   %5 = shufflevector <4 x float> %0, <4 x float> poison, <2 x i32> <i32 0, i32 2>
   %6 = shufflevector <4 x float> %0, <4 x float> poison, <2 x i32> <i32 1, i32 3>
   %7 = fadd <2 x float> %5, %6
   %8 = shufflevector <2 x float> %7, <2 x float> poison, <2 x i32> <i32 1, i32 undef>
   %9 = fadd <2 x float> %7, %8
   %10 = shufflevector <4 x float> %1, <4 x float> poison, <2 x i32> <i32 0, i32 2>
   %11 = shufflevector <4 x float> %1, <4 x float> poison, <2 x i32> <i32 1, i32 3>
   %12 = fadd <2 x float> %10, %11
; AVX2-SLOW-NEXT:    retq
   %29 = shufflevector <4 x float> %22, <4 x float> %28, <4 x i32> <i32 0, i32 1, i32 2, i32 4>
   ret <4 x float> %29
 }
 
 define <4 x i32> @pair_sum_v4i32_v4i32(<4 x i32> %0, <4 x i32> %1, <4 x i32> %2, <4 x i32> %3) {
 ; SSSE3-SLOW-LABEL: pair_sum_v4i32_v4i32:
 ; SSSE3-SLOW:       # %bb.0:
 ; SSSE3-SLOW-NEXT:    phaddd %xmm1, %xmm0
-; SSSE3-SLOW-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,3,1,3]
-; SSSE3-SLOW-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,1,3]
-; SSSE3-SLOW-NEXT:    paddd %xmm1, %xmm0
 ; SSSE3-SLOW-NEXT:    phaddd %xmm2, %xmm3
-; SSSE3-SLOW-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[1,1,3,3]
-; SSSE3-SLOW-NEXT:    paddd %xmm3, %xmm1
-; SSSE3-SLOW-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,0]
+; SSSE3-SLOW-NEXT:    phaddd %xmm3, %xmm0
+; SSSE3-SLOW-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,3,2]
 ; SSSE3-SLOW-NEXT:    retq
 ;
 ; SSSE3-FAST-LABEL: pair_sum_v4i32_v4i32:
 ; SSSE3-FAST:       # %bb.0:
 ; SSSE3-FAST-NEXT:    phaddd %xmm1, %xmm0
 ; SSSE3-FAST-NEXT:    phaddd %xmm3, %xmm2
 ; SSSE3-FAST-NEXT:    phaddd %xmm2, %xmm0
 ; SSSE3-FAST-NEXT:    retq
 ;
 ; AVX1-SLOW-LABEL: pair_sum_v4i32_v4i32:
 ; AVX1-SLOW:       # %bb.0:
 ; AVX1-SLOW-NEXT:    vphaddd %xmm1, %xmm0, %xmm0
-; AVX1-SLOW-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,3,1,3]
-; AVX1-SLOW-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,2,1,3]
-; AVX1-SLOW-NEXT:    vpaddd %xmm1, %xmm0, %xmm0
-; AVX1-SLOW-NEXT:    vphaddd %xmm2, %xmm2, %xmm1
-; AVX1-SLOW-NEXT:    vpshufd {{.*#+}} xmm2 = xmm1[1,1,1,1]
-; AVX1-SLOW-NEXT:    vpaddd %xmm2, %xmm1, %xmm1
-; AVX1-SLOW-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
-; AVX1-SLOW-NEXT:    vphaddd %xmm3, %xmm3, %xmm1
-; AVX1-SLOW-NEXT:    vpshufd {{.*#+}} xmm2 = xmm1[0,0,0,0]
-; AVX1-SLOW-NEXT:    vpaddd %xmm1, %xmm2, %xmm1
-; AVX1-SLOW-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5],xmm1[6,7]
+; AVX1-SLOW-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[0,2,2,3]
+; AVX1-SLOW-NEXT:    vphaddd %xmm2, %xmm2, %xmm2
+; AVX1-SLOW-NEXT:    vpunpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX1-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[1,3],xmm2[1,3]
+; AVX1-SLOW-NEXT:    vphaddd %xmm3, %xmm3, %xmm2
+; AVX1-SLOW-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1,2],xmm2[3]
+; AVX1-SLOW-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[0,0,0,0]
+; AVX1-SLOW-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,5],xmm2[6,7]
+; AVX1-SLOW-NEXT:    vpaddd %xmm0, %xmm1, %xmm0
 ; AVX1-SLOW-NEXT:    retq
 ;
-; AVX1-FAST-LABEL: pair_sum_v4i32_v4i32:
-; AVX1-FAST:       # %bb.0:
-; AVX1-FAST-NEXT:    vphaddd %xmm1, %xmm0, %xmm0
-; AVX1-FAST-NEXT:    vphaddd %xmm3, %xmm2, %xmm1
-; AVX1-FAST-NEXT:    vphaddd %xmm1, %xmm0, %xmm0
-; AVX1-FAST-NEXT:    retq
+; AVX-FAST-LABEL: pair_sum_v4i32_v4i32:
+; AVX-FAST:       # %bb.0:
+; AVX-FAST-NEXT:    vphaddd %xmm1, %xmm0, %xmm0
+; AVX-FAST-NEXT:    vphaddd %xmm3, %xmm2, %xmm1
+; AVX-FAST-NEXT:    vphaddd %xmm1, %xmm0, %xmm0
+; AVX-FAST-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: pair_sum_v4i32_v4i32:
 ; AVX2-SLOW:       # %bb.0:
 ; AVX2-SLOW-NEXT:    vphaddd %xmm1, %xmm0, %xmm0
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
-; AVX2-SLOW-NEXT:    vpaddd %xmm1, %xmm0, %xmm0
+; AVX2-SLOW-NEXT:    vphaddd %xmm0, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vphaddd %xmm2, %xmm2, %xmm1
 ; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm2 = xmm1[1,1,1,1]
 ; AVX2-SLOW-NEXT:    vpaddd %xmm2, %xmm1, %xmm1
-; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,3]
+; AVX2-SLOW-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
 ; AVX2-SLOW-NEXT:    vphaddd %xmm3, %xmm3, %xmm1
 ; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm2 = xmm1[1,1,1,1]
+; AVX2-SLOW-NEXT:    vpbroadcastd %xmm1, %xmm1
 ; AVX2-SLOW-NEXT:    vpaddd %xmm2, %xmm1, %xmm1
-; AVX2-SLOW-NEXT:    vinsertps {{.*#+}} xmm0 = xmm0[0,1,2],xmm1[0]
+; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1,2],xmm1[3]
 ; AVX2-SLOW-NEXT:    retq
-;
-; AVX2-FAST-LABEL: pair_sum_v4i32_v4i32:
-; AVX2-FAST:       # %bb.0:
-; AVX2-FAST-NEXT:    vphaddd %xmm3, %xmm2, %xmm2
-; AVX2-FAST-NEXT:    vphaddd %xmm1, %xmm0, %xmm0
-; AVX2-FAST-NEXT:    vphaddd %xmm2, %xmm0, %xmm0
-; AVX2-FAST-NEXT:    retq
   %5 = shufflevector <4 x i32> %0, <4 x i32> poison, <2 x i32> <i32 0, i32 2>
   %6 = shufflevector <4 x i32> %0, <4 x i32> poison, <2 x i32> <i32 1, i32 3>
   %7 = add <2 x i32> %5, %6
   %8 = shufflevector <2 x i32> %7, <2 x i32> poison, <2 x i32> <i32 1, i32 undef>
   %9 = add <2 x i32> %7, %8
   %10 = shufflevector <4 x i32> %1, <4 x i32> poison, <2 x i32> <i32 0, i32 2>
   %11 = shufflevector <4 x i32> %1, <4 x i32> poison, <2 x i32> <i32 1, i32 3>
   %12 = add <2 x i32> %10, %11
; AVX2-SLOW-NEXT:    retq
   %29 = shufflevector <4 x i32> %22, <4 x i32> %28, <4 x i32> <i32 0, i32 1, i32 2, i32 4>
   ret <4 x i32> %29
 }
 
 define <8 x float> @pair_sum_v8f32_v4f32(<4 x float> %0, <4 x float> %1, <4 x float> %2, <4 x float> %3, <4 x float> %4, <4 x float> %5, <4 x float> %6, <4 x float> %7) {
 ; SSSE3-SLOW-LABEL: pair_sum_v8f32_v4f32:
 ; SSSE3-SLOW:       # %bb.0:
 ; SSSE3-SLOW-NEXT:    haddps %xmm1, %xmm0
-; SSSE3-SLOW-NEXT:    movaps %xmm0, %xmm1
-; SSSE3-SLOW-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,3],xmm0[1,3]
-; SSSE3-SLOW-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,2,1,3]
-; SSSE3-SLOW-NEXT:    addps %xmm1, %xmm0
+; SSSE3-SLOW-NEXT:    haddps %xmm0, %xmm0
 ; SSSE3-SLOW-NEXT:    haddps %xmm3, %xmm2
 ; SSSE3-SLOW-NEXT:    movaps %xmm5, %xmm1
 ; SSSE3-SLOW-NEXT:    haddps %xmm4, %xmm1
 ; SSSE3-SLOW-NEXT:    haddps %xmm1, %xmm2
 ; SSSE3-SLOW-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1,3,2]
 ; SSSE3-SLOW-NEXT:    movlhps {{.*#+}} xmm0 = xmm0[0],xmm2[0]
 ; SSSE3-SLOW-NEXT:    haddps %xmm7, %xmm6
 ; SSSE3-SLOW-NEXT:    haddps %xmm5, %xmm4

 ; SSSE3-FAST-NEXT:    haddps %xmm7, %xmm6
 ; SSSE3-FAST-NEXT:    haddps %xmm6, %xmm4
 ; SSSE3-FAST-NEXT:    movaps %xmm4, %xmm1
 ; SSSE3-FAST-NEXT:    retq
 ;
 ; AVX1-SLOW-LABEL: pair_sum_v8f32_v4f32:
 ; AVX1-SLOW:       # %bb.0:
 ; AVX1-SLOW-NEXT:    vhaddps %xmm1, %xmm0, %xmm0
-; AVX1-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[1,3,1,3]
-; AVX1-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,1,3]
-; AVX1-SLOW-NEXT:    vaddps %xmm1, %xmm0, %xmm0
+; AVX1-SLOW-NEXT:    vhaddps %xmm0, %xmm0, %xmm0
 ; AVX1-SLOW-NEXT:    vhaddps %xmm4, %xmm4, %xmm1
 ; AVX1-SLOW-NEXT:    vhaddps %xmm5, %xmm5, %xmm4
 ; AVX1-SLOW-NEXT:    vhaddps %xmm3, %xmm2, %xmm2
 ; AVX1-SLOW-NEXT:    vshufps {{.*#+}} xmm3 = xmm2[0,2],xmm1[0,1]
 ; AVX1-SLOW-NEXT:    vinsertps {{.*#+}} xmm3 = xmm3[0,1,2],xmm4[0]
 ; AVX1-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm2[1,3],xmm1[1,3]
 ; AVX1-SLOW-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1,2],xmm4[3]
 ; AVX1-SLOW-NEXT:    vaddps %xmm1, %xmm3, %xmm1

 ; AVX1-FAST-NEXT:    vhaddps %xmm0, %xmm2, %xmm2
 ; AVX1-FAST-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
 ; AVX1-FAST-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[2]
 ; AVX1-FAST-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: pair_sum_v8f32_v4f32:
 ; AVX2-SLOW:       # %bb.0:
 ; AVX2-SLOW-NEXT:    vhaddps %xmm1, %xmm0, %xmm0
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[1,3,1,3]
-; AVX2-SLOW-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,2,1,3]
-; AVX2-SLOW-NEXT:    vaddps %xmm1, %xmm0, %xmm0
+; AVX2-SLOW-NEXT:    vhaddps %xmm0, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vhaddps %xmm4, %xmm4, %xmm1
 ; AVX2-SLOW-NEXT:    vhaddps %xmm5, %xmm5, %xmm4
 ; AVX2-SLOW-NEXT:    vhaddps %xmm3, %xmm2, %xmm2
 ; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm3 = xmm2[0,2],xmm1[0,3]
 ; AVX2-SLOW-NEXT:    vinsertps {{.*#+}} xmm3 = xmm3[0,1,2],xmm4[0]
 ; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm2[1,3],xmm1[1,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1,2],xmm4[3]
 ; AVX2-SLOW-NEXT:    vaddps %xmm1, %xmm3, %xmm1
; AVX2-FAST-NEXT:    retq
   %53 = shufflevector <8 x float> %42, <8 x float> %52, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 8, i32 9>
   ret <8 x float> %53
 }
 
 define <8 x i32> @pair_sum_v8i32_v4i32(<4 x i32> %0, <4 x i32> %1, <4 x i32> %2, <4 x i32> %3, <4 x i32> %4, <4 x i32> %5, <4 x i32> %6, <4 x i32> %7) {
 ; SSSE3-SLOW-LABEL: pair_sum_v8i32_v4i32:
 ; SSSE3-SLOW:       # %bb.0:
 ; SSSE3-SLOW-NEXT:    phaddd %xmm1, %xmm0
-; SSSE3-SLOW-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,3,1,3]
-; SSSE3-SLOW-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,1,3]
-; SSSE3-SLOW-NEXT:    paddd %xmm1, %xmm0
+; SSSE3-SLOW-NEXT:    phaddd %xmm0, %xmm0
 ; SSSE3-SLOW-NEXT:    phaddd %xmm3, %xmm2
 ; SSSE3-SLOW-NEXT:    phaddd %xmm4, %xmm5
 ; SSSE3-SLOW-NEXT:    phaddd %xmm5, %xmm2
 ; SSSE3-SLOW-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,1,3,2]
 ; SSSE3-SLOW-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
 ; SSSE3-SLOW-NEXT:    phaddd %xmm7, %xmm6
 ; SSSE3-SLOW-NEXT:    phaddd %xmm6, %xmm6
 ; SSSE3-SLOW-NEXT:    palignr {{.*#+}} xmm6 = xmm1[8,9,10,11,12,13,14,15],xmm6[0,1,2,3,4,5,6,7]

 ; SSSE3-FAST-NEXT:    phaddd %xmm7, %xmm6
 ; SSSE3-FAST-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,3],xmm6[0,2]
 ; SSSE3-FAST-NEXT:    movaps %xmm2, %xmm1
 ; SSSE3-FAST-NEXT:    retq
 ;
 ; AVX1-SLOW-LABEL: pair_sum_v8i32_v4i32:
 ; AVX1-SLOW:       # %bb.0:
 ; AVX1-SLOW-NEXT:    vphaddd %xmm1, %xmm0, %xmm0
-; AVX1-SLOW-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,3,1,3]
-; AVX1-SLOW-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,2,1,3]
-; AVX1-SLOW-NEXT:    vpaddd %xmm1, %xmm0, %xmm0
+; AVX1-SLOW-NEXT:    vphaddd %xmm0, %xmm0, %xmm0
 ; AVX1-SLOW-NEXT:    vphaddd %xmm4, %xmm4, %xmm1
 ; AVX1-SLOW-NEXT:    vphaddd %xmm5, %xmm5, %xmm4
 ; AVX1-SLOW-NEXT:    vphaddd %xmm3, %xmm2, %xmm2
 ; AVX1-SLOW-NEXT:    vpshufd {{.*#+}} xmm3 = xmm2[0,2,2,3]
 ; AVX1-SLOW-NEXT:    vpunpcklqdq {{.*#+}} xmm3 = xmm3[0],xmm1[0]
 ; AVX1-SLOW-NEXT:    vpshufd {{.*#+}} xmm5 = xmm4[0,0,0,0]
 ; AVX1-SLOW-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5],xmm5[6,7]
 ; AVX1-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm2[1,3],xmm1[1,3]

 ; AVX1-FAST-NEXT:    vphaddd %xmm0, %xmm2, %xmm2
 ; AVX1-FAST-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
 ; AVX1-FAST-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[2]
 ; AVX1-FAST-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: pair_sum_v8i32_v4i32:
 ; AVX2-SLOW:       # %bb.0:
 ; AVX2-SLOW-NEXT:    vphaddd %xmm1, %xmm0, %xmm0
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[1,3,1,3]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,2,1,3]
-; AVX2-SLOW-NEXT:    vpaddd %xmm1, %xmm0, %xmm0
+; AVX2-SLOW-NEXT:    vphaddd %xmm0, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vphaddd %xmm4, %xmm4, %xmm1
 ; AVX2-SLOW-NEXT:    vphaddd %xmm5, %xmm5, %xmm4
 ; AVX2-SLOW-NEXT:    vphaddd %xmm3, %xmm2, %xmm2
 ; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm3 = xmm2[0,2],xmm1[0,3]
 ; AVX2-SLOW-NEXT:    vinsertps {{.*#+}} xmm3 = xmm3[0,1,2],xmm4[0]
 ; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm2[1,3],xmm1[1,3]
 ; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1,2],xmm4[3]
 ; AVX2-SLOW-NEXT:    vpaddd %xmm1, %xmm3, %xmm1

 ;
 ; VTYPE sum4(VTYPE A0, VTYPE A1, VTYPE A2, VTYPE A3) {
 ;   return (VTYPE) { sum( A0 ), sum( A1 ), sum( A2 ), sum( A3 ) };
 ; }
 
 define <4 x float> @sequential_sum_v4f32_v4f32(<4 x float> %0, <4 x float> %1, <4 x float> %2, <4 x float> %3) {
 ; SSSE3-SLOW-LABEL: sequential_sum_v4f32_v4f32:
 ; SSSE3-SLOW:       # %bb.0:
-; SSSE3-SLOW-NEXT:    movaps %xmm0, %xmm5
-; SSSE3-SLOW-NEXT:    haddps %xmm1, %xmm5
 ; SSSE3-SLOW-NEXT:    movaps %xmm0, %xmm4
-; SSSE3-SLOW-NEXT:    unpckhps {{.*#+}} xmm4 = xmm4[2],xmm1[2],xmm4[3],xmm1[3]
+; SSSE3-SLOW-NEXT:    haddps %xmm1, %xmm4
+; SSSE3-SLOW-NEXT:    movaps %xmm0, %xmm5
+; SSSE3-SLOW-NEXT:    unpckhps {{.*#+}} xmm5 = xmm5[2],xmm1[2],xmm5[3],xmm1[3]
 ; SSSE3-SLOW-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,3],xmm0[3,3]
-; SSSE3-SLOW-NEXT:    movshdup {{.*#+}} xmm0 = xmm2[1,1,3,3]
-; SSSE3-SLOW-NEXT:    addps %xmm2, %xmm0
-; SSSE3-SLOW-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,2],xmm0[0,1]
-; SSSE3-SLOW-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm2[2,3]
+; SSSE3-SLOW-NEXT:    movaps %xmm3, %xmm0
+; SSSE3-SLOW-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm2[1]
+; SSSE3-SLOW-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,1],xmm0[2,0]
+; SSSE3-SLOW-NEXT:    movaps %xmm3, %xmm0
+; SSSE3-SLOW-NEXT:    movlhps {{.*#+}} xmm0 = xmm0[0],xmm2[0]
+; SSSE3-SLOW-NEXT:    movaps %xmm3, %xmm6
+; SSSE3-SLOW-NEXT:    shufps {{.*#+}} xmm6 = xmm6[1,1],xmm2[1,1]
+; SSSE3-SLOW-NEXT:    addps %xmm0, %xmm6
+; SSSE3-SLOW-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,2],xmm6[2,0]
 ; SSSE3-SLOW-NEXT:    addps %xmm5, %xmm4
-; SSSE3-SLOW-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm2[3,3]
+; SSSE3-SLOW-NEXT:    shufps {{.*#+}} xmm3 = xmm3[3,3],xmm2[3,3]
+; SSSE3-SLOW-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm3[2,0]
 ; SSSE3-SLOW-NEXT:    addps %xmm1, %xmm4
-; SSSE3-SLOW-NEXT:    movshdup {{.*#+}} xmm0 = xmm3[1,1,3,3]
-; SSSE3-SLOW-NEXT:    addps %xmm3, %xmm0
-; SSSE3-SLOW-NEXT:    movaps %xmm3, %xmm1
-; SSSE3-SLOW-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm3[1]
-; SSSE3-SLOW-NEXT:    addps %xmm0, %xmm1
-; SSSE3-SLOW-NEXT:    shufps {{.*#+}} xmm3 = xmm3[3,3,3,3]
-; SSSE3-SLOW-NEXT:    addps %xmm1, %xmm3
-; SSSE3-SLOW-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm4[2,3]
-; SSSE3-SLOW-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm3[2,0]
 ; SSSE3-SLOW-NEXT:    movaps %xmm4, %xmm0
 ; SSSE3-SLOW-NEXT:    retq
 ;
 ; SSSE3-FAST-LABEL: sequential_sum_v4f32_v4f32:
 ; SSSE3-FAST:       # %bb.0:
 ; SSSE3-FAST-NEXT:    movaps %xmm0, %xmm5
 ; SSSE3-FAST-NEXT:    haddps %xmm1, %xmm5
 ; SSSE3-FAST-NEXT:    movaps %xmm0, %xmm4
 ; SSSE3-FAST-NEXT:    unpckhps {{.*#+}} xmm4 = xmm4[2],xmm1[2],xmm4[3],xmm1[3]
 ; SSSE3-FAST-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,3],xmm0[3,3]
-; SSSE3-FAST-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm2[2,3]
-; SSSE3-FAST-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm2[3,3]
-; SSSE3-FAST-NEXT:    haddps %xmm2, %xmm2
-; SSSE3-FAST-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,2],xmm2[0,1]
+; SSSE3-FAST-NEXT:    movaps %xmm3, %xmm0
+; SSSE3-FAST-NEXT:    haddps %xmm2, %xmm0
+; SSSE3-FAST-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,2],xmm0[2,0]
+; SSSE3-FAST-NEXT:    movaps %xmm3, %xmm0
+; SSSE3-FAST-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm2[1]
+; SSSE3-FAST-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm0[2,0]
 ; SSSE3-FAST-NEXT:    addps %xmm5, %xmm4
+; SSSE3-FAST-NEXT:    shufps {{.*#+}} xmm3 = xmm3[3,3],xmm2[3,3]
+; SSSE3-FAST-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm3[2,0]
 ; SSSE3-FAST-NEXT:    addps %xmm1, %xmm4
-; SSSE3-FAST-NEXT:    movaps %xmm3, %xmm0
-; SSSE3-FAST-NEXT:    haddps %xmm3, %xmm0
-; SSSE3-FAST-NEXT:    movaps %xmm3, %xmm1
-; SSSE3-FAST-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm3[1]
-; SSSE3-FAST-NEXT:    addps %xmm0, %xmm1
-; SSSE3-FAST-NEXT:    shufps {{.*#+}} xmm3 = xmm3[3,3,3,3]
-; SSSE3-FAST-NEXT:    addps %xmm1, %xmm3
-; SSSE3-FAST-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm4[2,3]
-; SSSE3-FAST-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm3[2,0]
 ; SSSE3-FAST-NEXT:    movaps %xmm4, %xmm0
 ; SSSE3-FAST-NEXT:    retq
 ;
 ; AVX-SLOW-LABEL: sequential_sum_v4f32_v4f32:
 ; AVX-SLOW:       # %bb.0:
 ; AVX-SLOW-NEXT:    vhaddps %xmm1, %xmm0, %xmm4
 ; AVX-SLOW-NEXT:    vunpckhps {{.*#+}} xmm5 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
 ; AVX-SLOW-NEXT:    vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
 ; AVX-SLOW-NEXT:    vinsertps {{.*#+}} xmm0 = xmm0[3],xmm1[1],zero,zero
 ; AVX-SLOW-NEXT:    vmovshdup {{.*#+}} xmm1 = xmm2[1,1,3,3]
 ; AVX-SLOW-NEXT:    vaddps %xmm2, %xmm1, %xmm1
-; AVX-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm4[0,2],xmm1[0,1]
+; AVX-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm4[0,2],xmm1[0,3]
 ; AVX-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,1],xmm2[3,3]
 ; AVX-SLOW-NEXT:    vblendps {{.*#+}} xmm2 = xmm5[0,1],xmm2[2,3]
 ; AVX-SLOW-NEXT:    vmovshdup {{.*#+}} xmm4 = xmm3[1,1,3,3]
 ; AVX-SLOW-NEXT:    vaddps %xmm3, %xmm4, %xmm4
 ; AVX-SLOW-NEXT:    vinsertps {{.*#+}} xmm1 = xmm1[0,1,2],xmm4[0]
 ; AVX-SLOW-NEXT:    vinsertps {{.*#+}} xmm2 = xmm2[0,1,2],xmm3[2]
 ; AVX-SLOW-NEXT:    vaddps %xmm1, %xmm2, %xmm1
 ; AVX-SLOW-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1,2],xmm3[3]
 ; AVX-SLOW-NEXT:    vaddps %xmm1, %xmm0, %xmm0
 ; AVX-SLOW-NEXT:    retq
 ;
 ; AVX-FAST-LABEL: sequential_sum_v4f32_v4f32:
 ; AVX-FAST:       # %bb.0:
 ; AVX-FAST-NEXT:    vhaddps %xmm1, %xmm0, %xmm4
 ; AVX-FAST-NEXT:    vunpckhps {{.*#+}} xmm5 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
 ; AVX-FAST-NEXT:    vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
 ; AVX-FAST-NEXT:    vinsertps {{.*#+}} xmm0 = xmm0[3],xmm1[1],zero,zero
 ; AVX-FAST-NEXT:    vhaddps %xmm2, %xmm2, %xmm1
-; AVX-FAST-NEXT:    vshufps {{.*#+}} xmm1 = xmm4[0,2],xmm1[0,1]
+; AVX-FAST-NEXT:    vshufps {{.*#+}} xmm1 = xmm4[0,2],xmm1[0,3]
 ; AVX-FAST-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,1],xmm2[3,3]
 ; AVX-FAST-NEXT:    vblendps {{.*#+}} xmm2 = xmm5[0,1],xmm2[2,3]
 ; AVX-FAST-NEXT:    vhaddps %xmm3, %xmm3, %xmm4
 ; AVX-FAST-NEXT:    vinsertps {{.*#+}} xmm1 = xmm1[0,1,2],xmm4[0]
 ; AVX-FAST-NEXT:    vinsertps {{.*#+}} xmm2 = xmm2[0,1,2],xmm3[2]
 ; AVX-FAST-NEXT:    vaddps %xmm1, %xmm2, %xmm1
 ; AVX-FAST-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1,2],xmm3[3]
 ; AVX-FAST-NEXT:    vaddps %xmm1, %xmm0, %xmm0
; AVX-FAST-NEXT:    retq
   ret <4 x float> %26
 }
 
 define <4 x i32> @sequential_sum_v4i32_v4i32(<4 x i32> %0, <4 x i32> %1, <4 x i32> %2, <4 x i32> %3) {
 ; SSSE3-SLOW-LABEL: sequential_sum_v4i32_v4i32:
 ; SSSE3-SLOW:       # %bb.0:
 ; SSSE3-SLOW-NEXT:    movdqa %xmm0, %xmm4
 ; SSSE3-SLOW-NEXT:    phaddd %xmm1, %xmm4
-; SSSE3-SLOW-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,2,2,3]
 ; SSSE3-SLOW-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
-; SSSE3-SLOW-NEXT:    paddd %xmm0, %xmm4
-; SSSE3-SLOW-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[1,1,1,1]
-; SSSE3-SLOW-NEXT:    pshufd {{.*#+}} xmm5 = xmm2[0,1,0,1]
-; SSSE3-SLOW-NEXT:    paddd %xmm2, %xmm5
-; SSSE3-SLOW-NEXT:    paddd %xmm1, %xmm5
-; SSSE3-SLOW-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[1,1,1,1]
-; SSSE3-SLOW-NEXT:    pshufd {{.*#+}} xmm6 = xmm3[2,3,2,3]
-; SSSE3-SLOW-NEXT:    paddd %xmm3, %xmm6
+; SSSE3-SLOW-NEXT:    movaps %xmm3, %xmm1
+; SSSE3-SLOW-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1]
+; SSSE3-SLOW-NEXT:    movdqa %xmm0, %xmm5
+; SSSE3-SLOW-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,1],xmm1[2,0]
+; SSSE3-SLOW-NEXT:    movaps %xmm3, %xmm1
+; SSSE3-SLOW-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; SSSE3-SLOW-NEXT:    movaps %xmm3, %xmm6
+; SSSE3-SLOW-NEXT:    shufps {{.*#+}} xmm6 = xmm6[1,1],xmm2[1,1]
 ; SSSE3-SLOW-NEXT:    paddd %xmm1, %xmm6
-; SSSE3-SLOW-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,1],xmm5[2,3]
-; SSSE3-SLOW-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm6[2,0]
+; SSSE3-SLOW-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,2],xmm6[2,0]
 ; SSSE3-SLOW-NEXT:    shufps {{.*#+}} xmm3 = xmm3[3,3],xmm2[3,3]
 ; SSSE3-SLOW-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,3],xmm3[2,0]
 ; SSSE3-SLOW-NEXT:    paddd %xmm4, %xmm0
+; SSSE3-SLOW-NEXT:    paddd %xmm5, %xmm0
 ; SSSE3-SLOW-NEXT:    retq
 ;
 ; SSSE3-FAST-LABEL: sequential_sum_v4i32_v4i32:
 ; SSSE3-FAST:       # %bb.0:
 ; SSSE3-FAST-NEXT:    movdqa %xmm0, %xmm4
 ; SSSE3-FAST-NEXT:    phaddd %xmm1, %xmm4
-; SSSE3-FAST-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[0,2,2,3]
 ; SSSE3-FAST-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
-; SSSE3-FAST-NEXT:    paddd %xmm0, %xmm4
-; SSSE3-FAST-NEXT:    movdqa %xmm2, %xmm1
+; SSSE3-FAST-NEXT:    movdqa %xmm3, %xmm1
 ; SSSE3-FAST-NEXT:    phaddd %xmm2, %xmm1
-; SSSE3-FAST-NEXT:    paddd %xmm2, %xmm1
-; SSSE3-FAST-NEXT:    movdqa %xmm3, %xmm5
-; SSSE3-FAST-NEXT:    phaddd %xmm3, %xmm5
-; SSSE3-FAST-NEXT:    pshufd {{.*#+}} xmm6 = xmm3[2,3,2,3]
-; SSSE3-FAST-NEXT:    paddd %xmm5, %xmm6
-; SSSE3-FAST-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,1],xmm1[2,3]
-; SSSE3-FAST-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm6[2,0]
+; SSSE3-FAST-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,2],xmm1[2,0]
+; SSSE3-FAST-NEXT:    movdqa %xmm3, %xmm1
+; SSSE3-FAST-NEXT:    punpckhqdq {{.*#+}} xmm1 = xmm1[1],xmm2[1]
+; SSSE3-FAST-NEXT:    movdqa %xmm0, %xmm5
+; SSSE3-FAST-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,1],xmm1[2,0]
 ; SSSE3-FAST-NEXT:    shufps {{.*#+}} xmm3 = xmm3[3,3],xmm2[3,3]
 ; SSSE3-FAST-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,3],xmm3[2,0]
 ; SSSE3-FAST-NEXT:    paddd %xmm4, %xmm0
+; SSSE3-FAST-NEXT:    paddd %xmm5, %xmm0
 ; SSSE3-FAST-NEXT:    retq
 ;
 ; AVX1-SLOW-LABEL: sequential_sum_v4i32_v4i32:
 ; AVX1-SLOW:       # %bb.0:
 ; AVX1-SLOW-NEXT:    vphaddd %xmm1, %xmm0, %xmm4
-; AVX1-SLOW-NEXT:    vpshufd {{.*#+}} xmm4 = xmm4[0,2,2,3]
 ; AVX1-SLOW-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
 ; AVX1-SLOW-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; AVX1-SLOW-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX1-SLOW-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3],xmm0[4,5,6,7]
 ; AVX1-SLOW-NEXT:    vpshufd {{.*#+}} xmm1 = xmm2[1,1,1,1]
 ; AVX1-SLOW-NEXT:    vpaddd %xmm2, %xmm1, %xmm1
-; AVX1-SLOW-NEXT:    vpunpcklqdq {{.*#+}} xmm1 = xmm4[0],xmm1[0]
-; AVX1-SLOW-NEXT:    vpshufd {{.*#+}} xmm4 = xmm2[3,3,3,3]
-; AVX1-SLOW-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm4[0]
-; AVX1-SLOW-NEXT:    vpaddd %xmm1, %xmm0, %xmm0
-; AVX1-SLOW-NEXT:    vpblendw {{.*#+}} xmm1 = xmm5[0,1,2,3],xmm2[4,5,6,7]
-; AVX1-SLOW-NEXT:    vpaddd %xmm1, %xmm0, %xmm0
-; AVX1-SLOW-NEXT:    vpshufd {{.*#+}} xmm1 = xmm3[1,1,1,1]
-; AVX1-SLOW-NEXT:    vpshufd {{.*#+}} xmm2 = xmm3[0,0,0,0]
+; AVX1-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm4[0,2],xmm1[0,3]
+; AVX1-SLOW-NEXT:    vinsertps {{.*#+}} xmm0 = xmm0[0,1],xmm2[3],zero
+; AVX1-SLOW-NEXT:    vpblendw {{.*#+}} xmm2 = xmm5[0,1,2,3],xmm2[4,5,6,7]
+; AVX1-SLOW-NEXT:    vpshufd {{.*#+}} xmm4 = xmm3[1,1,1,1]
+; AVX1-SLOW-NEXT:    vpshufd {{.*#+}} xmm5 = xmm3[0,0,0,0]
+; AVX1-SLOW-NEXT:    vpaddd %xmm5, %xmm4, %xmm4
+; AVX1-SLOW-NEXT:    vblendps {{.*#+}} xmm1 = xmm1[0,1,2],xmm4[3]
 ; AVX1-SLOW-NEXT:    vpshufd {{.*#+}} xmm4 = xmm3[2,2,2,2]
-; AVX1-SLOW-NEXT:    vpaddd %xmm2, %xmm4, %xmm2
-; AVX1-SLOW-NEXT:    vpaddd %xmm1, %xmm3, %xmm1
-; AVX1-SLOW-NEXT:    vpaddd %xmm2, %xmm1, %xmm1
-; AVX1-SLOW-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5],xmm1[6,7]
+; AVX1-SLOW-NEXT:    vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5],xmm4[6,7]
+; AVX1-SLOW-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1,2],xmm3[3]
+; AVX1-SLOW-NEXT:    vpaddd %xmm1, %xmm0, %xmm0
+; AVX1-SLOW-NEXT:    vpaddd %xmm2, %xmm0, %xmm0
 ; AVX1-SLOW-NEXT:    retq
 ;
 ; AVX1-FAST-LABEL: sequential_sum_v4i32_v4i32:
 ; AVX1-FAST:       # %bb.0:
 ; AVX1-FAST-NEXT:    vphaddd %xmm1, %xmm0, %xmm4
-; AVX1-FAST-NEXT:    vpshufd {{.*#+}} xmm4 = xmm4[0,2,2,3]
 ; AVX1-FAST-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
 ; AVX1-FAST-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; AVX1-FAST-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX1-FAST-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3],xmm0[4,5,6,7]
 ; AVX1-FAST-NEXT:    vphaddd %xmm2, %xmm2, %xmm1
-; AVX1-FAST-NEXT:    vpunpcklqdq {{.*#+}} xmm1 = xmm4[0],xmm1[0]
-; AVX1-FAST-NEXT:    vpshufd {{.*#+}} xmm4 = xmm2[3,3,3,3]
-; AVX1-FAST-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm4[0]
+; AVX1-FAST-NEXT:    vshufps {{.*#+}} xmm1 = xmm4[0,2],xmm1[0,3]
+; AVX1-FAST-NEXT:    vinsertps {{.*#+}} xmm0 = xmm0[0,1],xmm2[3],zero
+; AVX1-FAST-NEXT:    vpblendw {{.*#+}} xmm2 = xmm5[0,1,2,3],xmm2[4,5,6,7]
+; AVX1-FAST-NEXT:    vphaddd %xmm3, %xmm3, %xmm4
+; AVX1-FAST-NEXT:    vinsertps {{.*#+}} xmm1 = xmm1[0,1,2],xmm4[0]
+; AVX1-FAST-NEXT:    vpshufd {{.*#+}} xmm4 = xmm3[2,2,2,2]
+; AVX1-FAST-NEXT:    vpblendw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5],xmm4[6,7]
+; AVX1-FAST-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1,2],xmm3[3]
 ; AVX1-FAST-NEXT:    vpaddd %xmm1, %xmm0, %xmm0
-; AVX1-FAST-NEXT:    vpblendw {{.*#+}} xmm1 = xmm5[0,1,2,3],xmm2[4,5,6,7]
-; AVX1-FAST-NEXT:    vpaddd %xmm1, %xmm0, %xmm0
-; AVX1-FAST-NEXT:    vphaddd %xmm3, %xmm3, %xmm1
-; AVX1-FAST-NEXT:    vpshufd {{.*#+}} xmm2 = xmm3[2,2,2,2]
-; AVX1-FAST-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,0,0,0]
-; AVX1-FAST-NEXT:    vpaddd %xmm1, %xmm3, %xmm1
-; AVX1-FAST-NEXT:    vpaddd %xmm2, %xmm1, %xmm1
-; AVX1-FAST-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5],xmm1[6,7]
+; AVX1-FAST-NEXT:    vpaddd %xmm2, %xmm0, %xmm0
 ; AVX1-FAST-NEXT:    retq
 ;
 ; AVX2-SLOW-LABEL: sequential_sum_v4i32_v4i32:
 ; AVX2-SLOW:       # %bb.0:
 ; AVX2-SLOW-NEXT:    vphaddd %xmm1, %xmm0, %xmm4
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm4 = xmm4[0,2,2,3]
 ; AVX2-SLOW-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
 ; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3]
 ; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm1 = xmm2[1,1,1,1]
 ; AVX2-SLOW-NEXT:    vpaddd %xmm2, %xmm1, %xmm1
-; AVX2-SLOW-NEXT:    vpunpcklqdq {{.*#+}} xmm1 = xmm4[0],xmm1[0]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm4 = xmm2[3,3,3,3]
-; AVX2-SLOW-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm4[0]
+; AVX2-SLOW-NEXT:    vshufps {{.*#+}} xmm1 = xmm4[0,2],xmm1[0,3]
+; AVX2-SLOW-NEXT:    vinsertps {{.*#+}} xmm0 = xmm0[0,1],xmm2[3],zero
 ; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} xmm2 = xmm5[0,1],xmm2[2,3]
 ; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm4 = xmm3[1,1,1,1]
-; AVX2-SLOW-NEXT:    vpbroadcastd %xmm3, %xmm5
-; AVX2-SLOW-NEXT:    vpaddd %xmm5, %xmm4, %xmm4
-; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0,1,2],xmm4[3]
+; AVX2-SLOW-NEXT:    vpaddd %xmm3, %xmm4, %xmm4
+; AVX2-SLOW-NEXT:    vinsertps {{.*#+}} xmm1 = xmm1[0,1,2],xmm4[0]
 ; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm4 = xmm3[2,2,2,2]
 ; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} xmm2 = xmm2[0,1,2],xmm4[3]
-; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1,2],xmm3[3]
+; AVX2-SLOW-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1,2],xmm3[3]
 ; AVX2-SLOW-NEXT:    vpaddd %xmm1, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    vpaddd %xmm2, %xmm0, %xmm0
 ; AVX2-SLOW-NEXT:    retq
 ;
 ; AVX2-FAST-LABEL: sequential_sum_v4i32_v4i32:
 ; AVX2-FAST:       # %bb.0:
 ; AVX2-FAST-NEXT:    vphaddd %xmm1, %xmm0, %xmm4
-; AVX2-FAST-NEXT:    vpshufd {{.*#+}} xmm4 = xmm4[0,2,2,3]
 ; AVX2-FAST-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
 ; AVX2-FAST-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; AVX2-FAST-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
 ; AVX2-FAST-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3]
 ; AVX2-FAST-NEXT:    vphaddd %xmm2, %xmm2, %xmm1
-; AVX2-FAST-NEXT:    vpunpcklqdq {{.*#+}} xmm1 = xmm4[0],xmm1[0]
-; AVX2-FAST-NEXT:    vpshufd {{.*#+}} xmm4 = xmm2[3,3,3,3]
-; AVX2-FAST-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm4[0]
+; AVX2-FAST-NEXT:    vshufps {{.*#+}} xmm1 = xmm4[0,2],xmm1[0,3]
+; AVX2-FAST-NEXT:    vinsertps {{.*#+}} xmm0 = xmm0[0,1],xmm2[3],zero
 ; AVX2-FAST-NEXT:    vpblendd {{.*#+}} xmm2 = xmm5[0,1],xmm2[2,3]
 ; AVX2-FAST-NEXT:    vphaddd %xmm3, %xmm3, %xmm4
-; AVX2-FAST-NEXT:    vpshufd {{.*#+}} xmm5 = xmm3[2,2,2,2]
-; AVX2-FAST-NEXT:    vpblendd {{.*#+}} xmm2 = xmm2[0,1,2],xmm5[3]
-; AVX2-FAST-NEXT:    vpbroadcastd %xmm4, %xmm4
-; AVX2-FAST-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0,1,2],xmm4[3]
-; AVX2-FAST-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1,2],xmm3[3]
+; AVX2-FAST-NEXT:    vinsertps {{.*#+}} xmm1 = xmm1[0,1,2],xmm4[0]
+; AVX2-FAST-NEXT:    vpshufd {{.*#+}} xmm4 = xmm3[2,2,2,2]
+; AVX2-FAST-NEXT:    vpblendd {{.*#+}} xmm2 = xmm2[0,1,2],xmm4[3]
+; AVX2-FAST-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1,2],xmm3[3]
 ; AVX2-FAST-NEXT:    vpaddd %xmm1, %xmm0, %xmm0
 ; AVX2-FAST-NEXT:    vpaddd %xmm2, %xmm0, %xmm0
 ; AVX2-FAST-NEXT:    retq
   %5 = shufflevector <4 x i32> %0, <4 x i32> %1, <2 x i32> <i32 0, i32 4>
   %6 = shufflevector <4 x i32> %0, <4 x i32> %1, <2 x i32> <i32 1, i32 5>
   %7 = add <2 x i32> %5, %6
   %8 = shufflevector <4 x i32> %0, <4 x i32> %1, <2 x i32> <i32 2, i32 6>
   %9 = add <2 x i32> %8, %7

 declare float @llvm.vector.reduce.fadd.f32.v4f32(float, <4 x float>)
 
 define <4 x float> @reduction_sum_v4f32_v4f32_reassoc(<4 x float> %0, <4 x float> %1, <4 x float> %2, <4 x float> %3) {
 ; SSSE3-SLOW-LABEL: reduction_sum_v4f32_v4f32_reassoc:
 ; SSSE3-SLOW:       # %bb.0:
 ; SSSE3-SLOW-NEXT:    movaps %xmm0, %xmm4
 ; SSSE3-SLOW-NEXT:    unpckhpd {{.*#+}} xmm4 = xmm4[1],xmm0[1]
 ; SSSE3-SLOW-NEXT:    addps %xmm4, %xmm0
-; SSSE3-SLOW-NEXT:    movshdup {{.*#+}} xmm4 = xmm0[1,1,3,3]
-; SSSE3-SLOW-NEXT:    movaps %xmm1, %xmm5
-; SSSE3-SLOW-NEXT:    unpckhpd {{.*#+}} xmm5 = xmm5[1],xmm1[1]
-; SSSE3-SLOW-NEXT:    addps %xmm1, %xmm5
-; SSSE3-SLOW-NEXT:    movshdup {{.*#+}} xmm1 = xmm5[1,1,3,3]
-; SSSE3-SLOW-NEXT:    unpcklps {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
+; SSSE3-SLOW-NEXT:    movaps %xmm1, %xmm4
+; SSSE3-SLOW-NEXT:    unpckhpd {{.*#+}} xmm4 = xmm4[1],xmm1[1]
+; SSSE3-SLOW-NEXT:    addps %xmm1, %xmm4
+; SSSE3-SLOW-NEXT:    haddps %xmm4, %xmm0
 ; SSSE3-SLOW-NEXT:    movaps %xmm2, %xmm1
 ; SSSE3-SLOW-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1]
 ; SSSE3-SLOW-NEXT:    addps %xmm2, %xmm1
 ; SSSE3-SLOW-NEXT:    movaps %xmm3, %xmm2
 ; SSSE3-SLOW-NEXT:    unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm3[1]
 ; SSSE3-SLOW-NEXT:    addps %xmm3, %xmm2
-; SSSE3-SLOW-NEXT:    movaps %xmm2, %xmm3
-; SSSE3-SLOW-NEXT:    movlhps {{.*#+}} xmm3 = xmm3[0],xmm1[0]
-; SSSE3-SLOW-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1]
-; SSSE3-SLOW-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm3[2,0]
-; SSSE3-SLOW-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,1],xmm1[1,1]
-; SSSE3-SLOW-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm2[2,0]
-; SSSE3-SLOW-NEXT:    addps %xmm4, %xmm0
+; SSSE3-SLOW-NEXT:    haddps %xmm2, %xmm1
+; SSSE3-SLOW-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; SSSE3-SLOW-NEXT:    retq
 ;
 ; SSSE3-FAST-LABEL: reduction_sum_v4f32_v4f32_reassoc:
 ; SSSE3-FAST:       # %bb.0:
 ; SSSE3-FAST-NEXT:    movaps %xmm0, %xmm4
 ; SSSE3-FAST-NEXT:    unpckhpd {{.*#+}} xmm4 = xmm4[1],xmm0[1]
 ; SSSE3-FAST-NEXT:    addps %xmm4, %xmm0
 ; SSSE3-FAST-NEXT:    movaps %xmm1, %xmm4

 ; SSSE3-FAST-NEXT:    retq
 ;
 ; AVX-SLOW-LABEL: reduction_sum_v4f32_v4f32_reassoc:
 ; AVX-SLOW:       # %bb.0:
 ; AVX-SLOW-NEXT:    vpermilpd {{.*#+}} xmm4 = xmm0[1,0]
 ; AVX-SLOW-NEXT:    vaddps %xmm4, %xmm0, %xmm0
 ; AVX-SLOW-NEXT:    vpermilpd {{.*#+}} xmm4 = xmm1[1,0]
 ; AVX-SLOW-NEXT:    vaddps %xmm4, %xmm1, %xmm1
-; AVX-SLOW-NEXT:    vpermilpd {{.*#+}} xmm4 = xmm2[1,0]
-; AVX-SLOW-NEXT:    vaddps %xmm4, %xmm2, %xmm2
-; AVX-SLOW-NEXT:    vpermilpd {{.*#+}} xmm4 = xmm3[1,0]
-; AVX-SLOW-NEXT:    vaddps %xmm4, %xmm3, %xmm3
-; AVX-SLOW-NEXT:    vshufps {{.*#+}} xmm4 = xmm3[1,1],xmm2[1,1]
-; AVX-SLOW-NEXT:    vinsertps {{.*#+}} xmm5 = xmm0[1],xmm1[1],zero,zero
-; AVX-SLOW-NEXT:    vshufps {{.*#+}} xmm4 = xmm5[0,1],xmm4[2,0]
-; AVX-SLOW-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm3[0],xmm2[0]
-; AVX-SLOW-NEXT:    vunpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; AVX-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,1],xmm2[2,0]
-; AVX-SLOW-NEXT:    vaddps %xmm4, %xmm0, %xmm0
+; AVX-SLOW-NEXT:    vhaddps %xmm1, %xmm0, %xmm0
+; AVX-SLOW-NEXT:    vpermilpd {{.*#+}} xmm1 = xmm2[1,0]
+; AVX-SLOW-NEXT:    vaddps %xmm1, %xmm2, %xmm1
+; AVX-SLOW-NEXT:    vpermilpd {{.*#+}} xmm2 = xmm3[1,0]
+; AVX-SLOW-NEXT:    vaddps %xmm2, %xmm3, %xmm2
+; AVX-SLOW-NEXT:    vhaddps %xmm2, %xmm1, %xmm1
+; AVX-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX-SLOW-NEXT:    retq
 ;
 ; AVX-FAST-LABEL: reduction_sum_v4f32_v4f32_reassoc:
 ; AVX-FAST:       # %bb.0:
 ; AVX-FAST-NEXT:    vpermilpd {{.*#+}} xmm4 = xmm0[1,0]
 ; AVX-FAST-NEXT:    vaddps %xmm4, %xmm0, %xmm0
 ; AVX-FAST-NEXT:    vpermilpd {{.*#+}} xmm4 = xmm1[1,0]
 ; AVX-FAST-NEXT:    vaddps %xmm4, %xmm1, %xmm1
; AVX-FAST-NEXT:    retq
   ret <4 x float> %12
 }
 
 define <4 x i32> @reduction_sum_v4i32_v4i32(<4 x i32> %0, <4 x i32> %1, <4 x i32> %2, <4 x i32> %3) {
 ; SSSE3-SLOW-LABEL: reduction_sum_v4i32_v4i32:
 ; SSSE3-SLOW:       # %bb.0:
 ; SSSE3-SLOW-NEXT:    pshufd {{.*#+}} xmm4 = xmm0[2,3,2,3]
 ; SSSE3-SLOW-NEXT:    paddd %xmm4, %xmm0
-; SSSE3-SLOW-NEXT:    pshufd {{.*#+}} xmm4 = xmm0[1,1,1,1]
-; SSSE3-SLOW-NEXT:    pshufd {{.*#+}} xmm5 = xmm1[2,3,2,3]
-; SSSE3-SLOW-NEXT:    paddd %xmm1, %xmm5
-; SSSE3-SLOW-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[1,1,1,1]
-; SSSE3-SLOW-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
+; SSSE3-SLOW-NEXT:    pshufd {{.*#+}} xmm4 = xmm1[2,3,2,3]
+; SSSE3-SLOW-NEXT:    paddd %xmm1, %xmm4
+; SSSE3-SLOW-NEXT:    phaddd %xmm4, %xmm0
 ; SSSE3-SLOW-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[2,3,2,3]
 ; SSSE3-SLOW-NEXT:    paddd %xmm2, %xmm1
-; SSSE3-SLOW-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[1,1,1,1]
-; SSSE3-SLOW-NEXT:    pshufd {{.*#+}} xmm6 = xmm3[2,3,2,3]
-; SSSE3-SLOW-NEXT:    paddd %xmm3, %xmm6
-; SSSE3-SLOW-NEXT:    pshufd {{.*#+}} xmm3 = xmm6[1,1,1,1]
-; SSSE3-SLOW-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
-; SSSE3-SLOW-NEXT:    punpcklqdq {{.*#+}} xmm4 = xmm4[0],xmm2[0]
-; SSSE3-SLOW-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm6[0],xmm1[1],xmm6[1]
-; SSSE3-SLOW-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1]
-; SSSE3-SLOW-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
-; SSSE3-SLOW-NEXT:    paddd %xmm4, %xmm0
+; SSSE3-SLOW-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[2,3,2,3]
+; SSSE3-SLOW-NEXT:    paddd %xmm3, %xmm2
+; SSSE3-SLOW-NEXT:    phaddd %xmm2, %xmm1
+; SSSE3-SLOW-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; SSSE3-SLOW-NEXT:    retq
 ;
 ; SSSE3-FAST-LABEL: reduction_sum_v4i32_v4i32:
 ; SSSE3-FAST:       # %bb.0:
 ; SSSE3-FAST-NEXT:    pshufd {{.*#+}} xmm4 = xmm0[2,3,2,3]
 ; SSSE3-FAST-NEXT:    paddd %xmm4, %xmm0
 ; SSSE3-FAST-NEXT:    pshufd {{.*#+}} xmm4 = xmm1[2,3,2,3]
 ; SSSE3-FAST-NEXT:    paddd %xmm1, %xmm4
 ; SSSE3-FAST-NEXT:    phaddd %xmm4, %xmm0
 ; SSSE3-FAST-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[2,3,2,3]
 ; SSSE3-FAST-NEXT:    paddd %xmm2, %xmm1
 ; SSSE3-FAST-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[2,3,2,3]
 ; SSSE3-FAST-NEXT:    paddd %xmm3, %xmm2
 ; SSSE3-FAST-NEXT:    phaddd %xmm2, %xmm1
 ; SSSE3-FAST-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; SSSE3-FAST-NEXT:    retq
 ;
-; AVX1-SLOW-LABEL: reduction_sum_v4i32_v4i32:
-; AVX1-SLOW:       # %bb.0:
-; AVX1-SLOW-NEXT:    vpshufd {{.*#+}} xmm4 = xmm0[2,3,2,3]
-; AVX1-SLOW-NEXT:    vpaddd %xmm4, %xmm0, %xmm0
-; AVX1-SLOW-NEXT:    vpshufd {{.*#+}} xmm4 = xmm0[1,1,1,1]
-; AVX1-SLOW-NEXT:    vpshufd {{.*#+}} xmm5 = xmm1[2,3,2,3]
-; AVX1-SLOW-NEXT:    vpaddd %xmm5, %xmm1, %xmm1
-; AVX1-SLOW-NEXT:    vpshufd {{.*#+}} xmm5 = xmm1[1,1,1,1]
-; AVX1-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm4 = xmm4[0],xmm5[0],xmm4[1],xmm5[1]
-; AVX1-SLOW-NEXT:    vpshufd {{.*#+}} xmm5 = xmm2[2,3,2,3]
-; AVX1-SLOW-NEXT:    vpaddd %xmm5, %xmm2, %xmm2
-; AVX1-SLOW-NEXT:    vpshufd {{.*#+}} xmm5 = xmm2[1,1,1,1]
-; AVX1-SLOW-NEXT:    vpshufd {{.*#+}} xmm6 = xmm3[2,3,2,3]
-; AVX1-SLOW-NEXT:    vpaddd %xmm6, %xmm3, %xmm3
-; AVX1-SLOW-NEXT:    vpshufd {{.*#+}} xmm6 = xmm3[1,1,1,1]
-; AVX1-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1]
-; AVX1-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
-; AVX1-SLOW-NEXT:    vpaddd %xmm5, %xmm2, %xmm2
-; AVX1-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; AVX1-SLOW-NEXT:    vpaddd %xmm4, %xmm0, %xmm0
-; AVX1-SLOW-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm2[0]
-; AVX1-SLOW-NEXT:    retq
+; AVX-SLOW-LABEL: reduction_sum_v4i32_v4i32:
+; AVX-SLOW:       # %bb.0:
+; AVX-SLOW-NEXT:    vpshufd {{.*#+}} xmm4 = xmm0[2,3,2,3]
+; AVX-SLOW-NEXT:    vpaddd %xmm4, %xmm0, %xmm0
+; AVX-SLOW-NEXT:    vpshufd {{.*#+}} xmm4 = xmm1[2,3,2,3]
+; AVX-SLOW-NEXT:    vpaddd %xmm4, %xmm1, %xmm1
+; AVX-SLOW-NEXT:    vphaddd %xmm1, %xmm0, %xmm0
+; AVX-SLOW-NEXT:    vpshufd {{.*#+}} xmm1 = xmm2[2,3,2,3]
+; AVX-SLOW-NEXT:    vpaddd %xmm1, %xmm2, %xmm1
+; AVX-SLOW-NEXT:    vpshufd {{.*#+}} xmm2 = xmm3[2,3,2,3]
+; AVX-SLOW-NEXT:    vpaddd %xmm2, %xmm3, %xmm2
+; AVX-SLOW-NEXT:    vphaddd %xmm2, %xmm1, %xmm1
+; AVX-SLOW-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
+; AVX-SLOW-NEXT:    retq
 ;
 ; AVX-FAST-LABEL: reduction_sum_v4i32_v4i32:
 ; AVX-FAST:       # %bb.0:
 ; AVX-FAST-NEXT:    vpshufd {{.*#+}} xmm4 = xmm0[2,3,2,3]
 ; AVX-FAST-NEXT:    vpaddd %xmm4, %xmm0, %xmm0
 ; AVX-FAST-NEXT:    vpshufd {{.*#+}} xmm4 = xmm1[2,3,2,3]
 ; AVX-FAST-NEXT:    vpaddd %xmm4, %xmm1, %xmm1
 ; AVX-FAST-NEXT:    vphaddd %xmm1, %xmm0, %xmm0
 ; AVX-FAST-NEXT:    vpshufd {{.*#+}} xmm1 = xmm2[2,3,2,3]
 ; AVX-FAST-NEXT:    vpaddd %xmm1, %xmm2, %xmm1
 ; AVX-FAST-NEXT:    vpshufd {{.*#+}} xmm2 = xmm3[2,3,2,3]
 ; AVX-FAST-NEXT:    vpaddd %xmm2, %xmm3, %xmm2
 ; AVX-FAST-NEXT:    vphaddd %xmm2, %xmm1, %xmm1
 ; AVX-FAST-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX-FAST-NEXT:    retq
-;
-; AVX2-SLOW-LABEL: reduction_sum_v4i32_v4i32:
-; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm4 = xmm0[2,3,2,3]
-; AVX2-SLOW-NEXT:    vpaddd %xmm4, %xmm0, %xmm0
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm4 = xmm0[1,1,1,1]
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm5 = xmm1[2,3,2,3]
-; AVX2-SLOW-NEXT:    vpaddd %xmm5, %xmm1, %xmm1
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm5 = xmm2[2,3,2,3]
-; AVX2-SLOW-NEXT:    vpaddd %xmm5, %xmm2, %xmm2
-; AVX2-SLOW-NEXT:    vpshufd {{.*#+}} xmm5 = xmm3[2,3,2,3]
-; AVX2-SLOW-NEXT:    vpaddd %xmm5, %xmm3, %xmm3
-; AVX2-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm5 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
-; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} xmm4 = xmm4[0],xmm1[1],xmm4[2,3]
-; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} xmm4 = xmm4[0,1],xmm5[2,3]
-; AVX2-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; AVX2-SLOW-NEXT:    vpbroadcastd %xmm3, %xmm1
-; AVX2-SLOW-NEXT:    vpbroadcastd %xmm2, %xmm2
-; AVX2-SLOW-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; AVX2-SLOW-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
-; AVX2-SLOW-NEXT:    vpaddd %xmm4, %xmm0, %xmm0
-; AVX2-SLOW-NEXT:    retq
   %5 = call i32 @llvm.vector.reduce.add.i32.v4i32(<4 x i32> %0)
   %6 = call i32 @llvm.vector.reduce.add.i32.v4i32(<4 x i32> %1)
   %7 = call i32 @llvm.vector.reduce.add.i32.v4i32(<4 x i32> %2)
   %8 = call i32 @llvm.vector.reduce.add.i32.v4i32(<4 x i32> %3)
   %9 = insertelement <4 x i32> undef, i32 %5, i32 0
   %10 = insertelement <4 x i32> %9,   i32 %6, i32 1
   %11 = insertelement <4 x i32> %10,  i32 %7, i32 2
   %12 = insertelement <4 x i32> %11,  i32 %8, i32 3
   ret <4 x i32> %12
 }
 declare i32 @llvm.vector.reduce.add.i32.v4i32(<4 x i32>)

llvm/test/CodeGen/X86/phaddsub.ll

; AVX2-SHUF-NEXT:    retq
   %l = shufflevector <4 x i32> %x, <4 x i32> undef, <4 x i32> <i32 undef, i32 undef, i32 undef, i32 2>
   %add = add <4 x i32> %l, %x
   ret <4 x i32> %add
 }
 
 define <4 x i32> @phaddd_single_source5(<4 x i32> %x) {
 ; SSSE3-SLOW-LABEL: phaddd_single_source5:
 ; SSSE3-SLOW:       # %bb.0:
-; SSSE3-SLOW-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,2,2,2]
-; SSSE3-SLOW-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; SSSE3-SLOW-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[3,3,3,3]
+; SSSE3-SLOW-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
 ; SSSE3-SLOW-NEXT:    paddd %xmm1, %xmm0
 ; SSSE3-SLOW-NEXT:    retq
 ;
 ; SSSE3-FAST-LABEL: phaddd_single_source5:
 ; SSSE3-FAST:       # %bb.0:
 ; SSSE3-FAST-NEXT:    phaddd %xmm0, %xmm0
-; SSSE3-FAST-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; SSSE3-FAST-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
 ; SSSE3-FAST-NEXT:    retq
 ;
 ; AVX-SLOW-LABEL: phaddd_single_source5:
 ; AVX-SLOW:       # %bb.0:
-; AVX-SLOW-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,2,2,2]
-; AVX-SLOW-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
-; AVX-SLOW-NEXT:    vpaddd %xmm0, %xmm1, %xmm0
+; AVX-SLOW-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[3,3,3,3]
+; AVX-SLOW-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX-SLOW-NEXT:    vpaddd %xmm1, %xmm0, %xmm0
 ; AVX-SLOW-NEXT:    retq
 ;
 ; AVX-FAST-LABEL: phaddd_single_source5:
 ; AVX-FAST:       # %bb.0:
 ; AVX-FAST-NEXT:    vphaddd %xmm0, %xmm0, %xmm0
-; AVX-FAST-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; AVX-FAST-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
 ; AVX-FAST-NEXT:    retq
 ;
 ; AVX2-SHUF-LABEL: phaddd_single_source5:
 ; AVX2-SHUF:       # %bb.0:
-; AVX2-SHUF-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,2,2,2]
-; AVX2-SHUF-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
-; AVX2-SHUF-NEXT:    vpaddd %xmm0, %xmm1, %xmm0
+; AVX2-SHUF-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[3,3,3,3]
+; AVX2-SHUF-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX2-SHUF-NEXT:    vpaddd %xmm1, %xmm0, %xmm0
 ; AVX2-SHUF-NEXT:    retq
   %l = shufflevector <4 x i32> %x, <4 x i32> undef, <4 x i32> <i32 undef, i32 undef, i32 undef, i32 2>
   %add = add <4 x i32> %l, %x
   %shuffle2 = shufflevector <4 x i32> %add, <4 x i32> undef, <4 x i32> <i32 3, i32 undef, i32 undef, i32 undef>
   ret <4 x i32> %shuffle2
 }
 
 define <4 x i32> @phaddd_single_source6(<4 x i32> %x) {
-; SSSE3-LABEL: phaddd_single_source6:
-; SSSE3:       # %bb.0:
-; SSSE3-NEXT:    phaddd %xmm0, %xmm0
-; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
-; SSSE3-NEXT:    retq
+; SSSE3-SLOW-LABEL: phaddd_single_source6:
+; SSSE3-SLOW:       # %bb.0:
+; SSSE3-SLOW-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[0,0,1,1]
+; SSSE3-SLOW-NEXT:    paddd %xmm1, %xmm0
+; SSSE3-SLOW-NEXT:    retq
 ;
-; AVX-LABEL: phaddd_single_source6:
-; AVX:       # %bb.0:
-; AVX-NEXT:    vphaddd %xmm0, %xmm0, %xmm0
-; AVX-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
-; AVX-NEXT:    retq
+; SSSE3-FAST-LABEL: phaddd_single_source6:
+; SSSE3-FAST:       # %bb.0:
+; SSSE3-FAST-NEXT:    phaddd %xmm0, %xmm0
+; SSSE3-FAST-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
+; SSSE3-FAST-NEXT:    retq
+;
+; AVX-SLOW-LABEL: phaddd_single_source6:
+; AVX-SLOW:       # %bb.0:
+; AVX-SLOW-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[0,0,1,1]
+; AVX-SLOW-NEXT:    vpaddd %xmm0, %xmm1, %xmm0
+; AVX-SLOW-NEXT:    retq
+;
+; AVX-FAST-LABEL: phaddd_single_source6:
+; AVX-FAST:       # %bb.0:
+; AVX-FAST-NEXT:    vphaddd %xmm0, %xmm0, %xmm0
+; AVX-FAST-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,1,1]
+; AVX-FAST-NEXT:    retq
+;
+; AVX2-SHUF-LABEL: phaddd_single_source6:
+; AVX2-SHUF:       # %bb.0:
+; AVX2-SHUF-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[0,0,1,1]
+; AVX2-SHUF-NEXT:    vpaddd %xmm0, %xmm1, %xmm0
+; AVX2-SHUF-NEXT:    retq
   %l = shufflevector <4 x i32> %x, <4 x i32> undef, <4 x i32> <i32 undef, i32 undef, i32 0, i32 undef>
   %r = shufflevector <4 x i32> %x, <4 x i32> undef, <4 x i32> <i32 undef, i32 undef, i32 1, i32 undef>
   %add = add <4 x i32> %l, %r
   %shuffle2 = shufflevector <4 x i32> %add, <4 x i32> undef, <4 x i32> <i32 undef, i32 2, i32 undef, i32 undef>
   ret <4 x i32> %shuffle2
 }
 
 define <8 x i16> @phaddw_single_source1(<8 x i16> %x) {

 ; AVX2-SHUF-NEXT:    vpaddw %xmm0, %xmm1, %xmm0
 ; AVX2-SHUF-NEXT:    retq
   %l = shufflevector <8 x i16> %x, <8 x i16> undef, <8 x i32> <i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 6>
   %add = add <8 x i16> %l, %x
   ret <8 x i16> %add
 }
 
 define <8 x i16> @phaddw_single_source6(<8 x i16> %x) {
-; SSSE3-LABEL: phaddw_single_source6:
-; SSSE3:       # %bb.0:
-; SSSE3-NEXT:    phaddw %xmm0, %xmm0
-; SSSE3-NEXT:    psrldq {{.*#+}} xmm0 = xmm0[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
-; SSSE3-NEXT:    retq
+; SSSE3-SLOW-LABEL: phaddw_single_source6:
+; SSSE3-SLOW:       # %bb.0:
+; SSSE3-SLOW-NEXT:    movdqa %xmm0, %xmm1
+; SSSE3-SLOW-NEXT:    pslld $16, %xmm1
+; SSSE3-SLOW-NEXT:    paddw %xmm1, %xmm0
+; SSSE3-SLOW-NEXT:    retq
 ;
-; AVX-LABEL: phaddw_single_source6:
-; AVX:       # %bb.0:
-; AVX-NEXT:    vphaddw %xmm0, %xmm0, %xmm0
-; AVX-NEXT:    vpsrldq {{.*#+}} xmm0 = xmm0[6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero,zero,zero
-; AVX-NEXT:    retq
+; SSSE3-FAST-LABEL: phaddw_single_source6:
+; SSSE3-FAST:       # %bb.0:
+; SSSE3-FAST-NEXT:    phaddw %xmm0, %xmm0
+; SSSE3-FAST-NEXT:    pslld $16, %xmm0
+; SSSE3-FAST-NEXT:    retq
+;
+; AVX-SLOW-LABEL: phaddw_single_source6:
+; AVX-SLOW:       # %bb.0:
+; AVX-SLOW-NEXT:    vpslld $16, %xmm0, %xmm1
+; AVX-SLOW-NEXT:    vpaddw %xmm0, %xmm1, %xmm0
+; AVX-SLOW-NEXT:    retq
+;
+; AVX-FAST-LABEL: phaddw_single_source6:
+; AVX-FAST:       # %bb.0:
+; AVX-FAST-NEXT:    vphaddw %xmm0, %xmm0, %xmm0
+; AVX-FAST-NEXT:    vpslld $16, %xmm0, %xmm0
+; AVX-FAST-NEXT:    retq
+;
+; AVX2-SHUF-LABEL: phaddw_single_source6:
+; AVX2-SHUF:       # %bb.0:
+; AVX2-SHUF-NEXT:    vpslld $16, %xmm0, %xmm1
+; AVX2-SHUF-NEXT:    vpaddw %xmm0, %xmm1, %xmm0
+; AVX2-SHUF-NEXT:    retq
   %l = shufflevector <8 x i16> %x, <8 x i16> undef, <8 x i32> <i32 undef, i32 undef, i32 undef, i32 undef, i32 0, i32 undef, i32 undef, i32 undef>
   %r = shufflevector <8 x i16> %x, <8 x i16> undef, <8 x i32> <i32 undef, i32 undef, i32 undef, i32 undef, i32 1, i32 undef, i32 undef, i32 undef>
   %add = add <8 x i16> %l, %r
   %shuffle2 = shufflevector <8 x i16> %add, <8 x i16> undef, <8 x i32> <i32 undef, i32 4, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
   ret <8 x i16> %shuffle2
 }
 
 ; PR39921 + PR39936

llvm/test/CodeGen/X86/sse3.ll

entry:
   %tmp9 = shufflevector <16 x i8> %tmp8, <16 x i8> %T0,  <16 x i32> < i32 0, i32 1, i32 2, i32 17,  i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef , i32 undef >
   ret <16 x i8> %tmp9
 }
 
 ; rdar://8520311
 define <4 x i32> @t17() nounwind {
 ; X86-LABEL: t17:
 ; X86:       # %bb.0: # %entry
-; X86-NEXT:    pshufd {{.*#+}} xmm0 = mem[0,1,0,1]
-; X86-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-NEXT:    movd {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,0,1]
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: t17:
 ; X64:       # %bb.0: # %entry
-; X64-NEXT:    pshufd {{.*#+}} xmm0 = mem[0,1,0,1]
-; X64-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-NEXT:    movd {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X64-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,0,1]
 ; X64-NEXT:    retq
 entry:
   %tmp1 = load <4 x float>, ptr undef, align 16
   %tmp2 = shufflevector <4 x float> %tmp1, <4 x float> undef, <4 x i32> <i32 4, i32 1, i32 2, i32 3>
   %tmp3 = load <4 x float>, ptr undef, align 16
   %tmp4 = shufflevector <4 x float> %tmp2, <4 x float> undef, <4 x i32> <i32 undef, i32 undef, i32 0, i32 1>
   %tmp5 = bitcast <4 x float> %tmp3 to <4 x i32>
   %tmp6 = shufflevector <4 x i32> %tmp5, <4 x i32> undef, <4 x i32> <i32 undef, i32 undef, i32 0, i32 1>
   %tmp7 = and <4 x i32> %tmp6, <i32 undef, i32 undef, i32 -1, i32 0>
   ret <4 x i32> %tmp7
 }

llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-7.ll

 ; SSE-NEXT:    movdqa {{.*#+}} xmm6 = [65535,65535,0,0,0,65535,65535,65535]
 ; SSE-NEXT:    pand %xmm6, %xmm5
 ; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm3[3,1,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm9 = xmm8[0,1,2,0,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm1[1,1,1,1]
 ; SSE-NEXT:    punpcklqdq {{.*#+}} xmm9 = xmm9[0],xmm10[0]
 ; SSE-NEXT:    pandn %xmm9, %xmm6
 ; SSE-NEXT:    por %xmm5, %xmm6
-; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [65535,65535,65535,0,65535,65535,65535,65535]
-; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm8[0,3,1,3,4,5,6,7]
-; SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm8
 ; SSE-NEXT:    psrlq $48, %xmm4
-; SSE-NEXT:    por %xmm8, %xmm4
-; SSE-NEXT:    pand %xmm5, %xmm4
-; SSE-NEXT:    pandn %xmm1, %xmm5
-; SSE-NEXT:    por %xmm4, %xmm5
+; SSE-NEXT:    movdqa {{.*#+}} xmm5 = [0,65535,65535,65535,65535,65535,65535,65535]
+; SSE-NEXT:    pshuflw {{.*#+}} xmm8 = xmm8[0,3,1,3,4,5,6,7]
+; SSE-NEXT:    pand %xmm5, %xmm8
+; SSE-NEXT:    pandn %xmm4, %xmm5
+; SSE-NEXT:    por %xmm8, %xmm5
+; SSE-NEXT:    movdqa {{.*#+}} xmm4 = [65535,65535,65535,0,65535,65535,65535,65535]
+; SSE-NEXT:    pand %xmm4, %xmm5
+; SSE-NEXT:    pandn %xmm1, %xmm4
+; SSE-NEXT:    por %xmm5, %xmm4
 ; SSE-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm2[0]
 ; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm11[2,1]
 ; SSE-NEXT:    movaps {{.*#+}} xmm2 = [65535,65535,65535,65535,0,0,0,65535]
 ; SSE-NEXT:    andps %xmm2, %xmm0
 ; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[0,2,2,3]
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm3 = xmm3[0,1,0,2,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
 ; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
 ; SSE-NEXT:    andnps %xmm3, %xmm2
 ; SSE-NEXT:    orps %xmm0, %xmm2
 ; SSE-NEXT:    movaps %xmm2, (%rax)
-; SSE-NEXT:    movq %xmm5, 48(%rax)
+; SSE-NEXT:    movq %xmm4, 48(%rax)
 ; SSE-NEXT:    movdqa %xmm6, 32(%rax)
 ; SSE-NEXT:    movdqa %xmm7, 16(%rax)
 ; SSE-NEXT:    retq
 ;
 ; AVX1-ONLY-LABEL: store_i16_stride7_vf4:
 ; AVX1-ONLY:       # %bb.0:
 ; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
 ; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %r10

llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-5.ll

 ; SSE-NEXT:    movaps %xmm2, %xmm5
 ; SSE-NEXT:    unpcklps {{.*#+}} xmm5 = xmm5[0],xmm3[0],xmm5[1],xmm3[1]
 ; SSE-NEXT:    movaps %xmm0, %xmm6
 ; SSE-NEXT:    unpcklps {{.*#+}} xmm6 = xmm6[0],xmm1[0],xmm6[1],xmm1[1]
 ; SSE-NEXT:    movlhps {{.*#+}} xmm6 = xmm6[0],xmm5[0]
 ; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,1],xmm1[1,1]
 ; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm2[2,0]
 ; SSE-NEXT:    movss {{.*#+}} xmm0 = xmm4[0],xmm0[1,2,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[1,1,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; SSE-NEXT:    movq %xmm2, 32(%r9)
+; SSE-NEXT:    unpcklps {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[2,3,2,3]
+; SSE-NEXT:    movq %xmm1, 32(%r9)
 ; SSE-NEXT:    movaps %xmm6, (%r9)
 ; SSE-NEXT:    movaps %xmm0, 16(%r9)
 ; SSE-NEXT:    retq
 ;
 ; AVX1-ONLY-LABEL: store_i32_stride5_vf2:
 ; AVX1-ONLY:       # %bb.0:
 ; AVX1-ONLY-NEXT:    vmovsd {{.*#+}} xmm0 = mem[0],zero
 ; AVX1-ONLY-NEXT:    vmovsd {{.*#+}} xmm1 = mem[0],zero
; AVX512-NEXT:    retq
   store <20 x i32> %interleaved.vec, ptr %out.vec, align 64
   ret void
 }
 
 define void @store_i32_stride5_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vecptr2, ptr %in.vecptr3, ptr %in.vecptr4, ptr %out.vec) nounwind {
 ; SSE-LABEL: store_i32_stride5_vf8:
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    movaps (%rdi), %xmm1
-; SSE-NEXT:    movaps 16(%rdi), %xmm5
-; SSE-NEXT:    movdqa (%rsi), %xmm6
-; SSE-NEXT:    movdqa 16(%rsi), %xmm3
-; SSE-NEXT:    movdqa (%rdx), %xmm7
-; SSE-NEXT:    movdqa 16(%rdx), %xmm10
+; SSE-NEXT:    movaps 16(%rdi), %xmm0
+; SSE-NEXT:    movaps (%rsi), %xmm6
+; SSE-NEXT:    movaps 16(%rsi), %xmm7
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps (%rdx), %xmm3
+; SSE-NEXT:    movaps 16(%rdx), %xmm10
 ; SSE-NEXT:    movaps (%rcx), %xmm4
 ; SSE-NEXT:    movaps 16(%rcx), %xmm2
-; SSE-NEXT:    movaps (%r8), %xmm0
+; SSE-NEXT:    movaps (%r8), %xmm5
 ; SSE-NEXT:    movaps 16(%r8), %xmm11
-; SSE-NEXT:    movaps %xmm2, %xmm9
-; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[3,3],xmm11[3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm12 = xmm10[3,3,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm8 = xmm3[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm8 = xmm8[0],xmm12[0],xmm8[1],xmm12[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,1],xmm9[0,2]
-; SSE-NEXT:    movaps %xmm4, %xmm12
-; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[3,3],xmm0[3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm13 = xmm6[3,3,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm9 = xmm7[2,3,2,3]
-; SSE-NEXT:    movss {{.*#+}} xmm9 = xmm13[0],xmm9[1,2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[0,1],xmm12[0,2]
-; SSE-NEXT:    movaps %xmm5, %xmm13
-; SSE-NEXT:    shufps {{.*#+}} xmm13 = xmm13[3,1],xmm11[2,3]
-; SSE-NEXT:    movaps %xmm2, %xmm12
-; SSE-NEXT:    unpckhpd {{.*#+}} xmm12 = xmm12[1],xmm10[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[2,0],xmm13[2,0]
-; SSE-NEXT:    movaps %xmm5, %xmm13
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    unpckhps {{.*#+}} xmm13 = xmm13[2],xmm3[2],xmm13[3],xmm3[3]
-; SSE-NEXT:    movaps %xmm10, %xmm14
-; SSE-NEXT:    unpcklps {{.*#+}} xmm14 = xmm14[0],xmm2[0],xmm14[1],xmm2[1]
+; SSE-NEXT:    movaps %xmm2, %xmm8
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[3,3],xmm11[3,3]
+; SSE-NEXT:    movaps %xmm7, %xmm9
+; SSE-NEXT:    unpckhps {{.*#+}} xmm9 = xmm9[2],xmm10[2],xmm9[3],xmm10[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[2,3],xmm8[0,2]
+; SSE-NEXT:    movaps %xmm0, %xmm12
+; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[3,1],xmm11[2,3]
+; SSE-NEXT:    movaps %xmm2, %xmm8
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm8 = xmm8[1],xmm10[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[2,0],xmm12[2,0]
+; SSE-NEXT:    movaps %xmm0, %xmm12
+; SSE-NEXT:    unpckhps {{.*#+}} xmm12 = xmm12[2],xmm7[2],xmm12[3],xmm7[3]
+; SSE-NEXT:    movaps %xmm10, %xmm13
+; SSE-NEXT:    unpcklps {{.*#+}} xmm13 = xmm13[0],xmm2[0],xmm13[1],xmm2[1]
 ; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,1],xmm11[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,2],xmm13[0,1]
-; SSE-NEXT:    movaps %xmm5, %xmm13
-; SSE-NEXT:    unpcklps {{.*#+}} xmm13 = xmm13[0],xmm3[0],xmm13[1],xmm3[1]
-; SSE-NEXT:    movlhps {{.*#+}} xmm13 = xmm13[0],xmm14[0]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,2],xmm12[0,1]
+; SSE-NEXT:    movaps %xmm0, %xmm12
+; SSE-NEXT:    unpcklps {{.*#+}} xmm12 = xmm12[0],xmm7[0],xmm12[1],xmm7[1]
+; SSE-NEXT:    movlhps {{.*#+}} xmm12 = xmm12[0],xmm13[0]
+; SSE-NEXT:    movaps %xmm4, %xmm14
+; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[3,3],xmm5[3,3]
+; SSE-NEXT:    movaps %xmm6, %xmm13
+; SSE-NEXT:    shufps {{.*#+}} xmm13 = xmm13[3,3],xmm3[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm13 = xmm13[0,2],xmm14[0,2]
 ; SSE-NEXT:    movaps %xmm1, %xmm14
-; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[3,0],xmm0[2,0]
-; SSE-NEXT:    movaps %xmm0, %xmm3
-; SSE-NEXT:    movdqa %xmm7, %xmm15
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm15 = xmm15[2],xmm4[2],xmm15[3],xmm4[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[3,0],xmm5[2,0]
+; SSE-NEXT:    movaps %xmm3, %xmm15
+; SSE-NEXT:    unpckhps {{.*#+}} xmm15 = xmm15[2],xmm4[2],xmm15[3],xmm4[3]
 ; SSE-NEXT:    shufps {{.*#+}} xmm15 = xmm15[0,1],xmm14[2,0]
 ; SSE-NEXT:    movaps %xmm1, %xmm14
-; SSE-NEXT:    movdqa %xmm6, %xmm0
 ; SSE-NEXT:    unpckhps {{.*#+}} xmm14 = xmm14[2],xmm6[2],xmm14[3],xmm6[3]
-; SSE-NEXT:    movdqa %xmm7, %xmm6
-; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm4[0],xmm6[1],xmm4[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[1,1],xmm3[1,1]
+; SSE-NEXT:    movaps %xmm6, %xmm7
+; SSE-NEXT:    movaps %xmm3, %xmm6
+; SSE-NEXT:    unpcklps {{.*#+}} xmm6 = xmm6[0],xmm4[0],xmm6[1],xmm4[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[1,1],xmm5[1,1]
 ; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,2],xmm14[0,1]
 ; SSE-NEXT:    movaps %xmm1, %xmm14
-; SSE-NEXT:    unpcklps {{.*#+}} xmm14 = xmm14[0],xmm0[0],xmm14[1],xmm0[1]
+; SSE-NEXT:    unpcklps {{.*#+}} xmm14 = xmm14[0],xmm7[0],xmm14[1],xmm7[1]
 ; SSE-NEXT:    movlhps {{.*#+}} xmm14 = xmm14[0],xmm6[0]
 ; SSE-NEXT:    shufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Folded Reload
 ; SSE-NEXT:    # xmm10 = xmm10[1,1],mem[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,1],xmm10[2,0]
-; SSE-NEXT:    movss {{.*#+}} xmm5 = xmm11[0],xmm5[1,2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[1,1],xmm0[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm7[2,0]
-; SSE-NEXT:    movss {{.*#+}} xmm1 = xmm3[0],xmm1[1,2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm10[2,0]
+; SSE-NEXT:    movss {{.*#+}} xmm0 = xmm11[0],xmm0[1,2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,1],xmm7[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm3[2,0]
+; SSE-NEXT:    movss {{.*#+}} xmm1 = xmm5[0],xmm1[1,2,3]
 ; SSE-NEXT:    movaps %xmm14, (%r9)
 ; SSE-NEXT:    movaps %xmm4, 32(%r9)
 ; SSE-NEXT:    movaps %xmm15, 48(%r9)
-; SSE-NEXT:    movaps %xmm13, 80(%r9)
+; SSE-NEXT:    movaps %xmm13, 64(%r9)
+; SSE-NEXT:    movaps %xmm12, 80(%r9)
 ; SSE-NEXT:    movaps %xmm2, 112(%r9)
-; SSE-NEXT:    movaps %xmm12, 128(%r9)
+; SSE-NEXT:    movaps %xmm8, 128(%r9)
+; SSE-NEXT:    movaps %xmm9, 144(%r9)
 ; SSE-NEXT:    movaps %xmm1, 16(%r9)
-; SSE-NEXT:    movaps %xmm9, 64(%r9)
-; SSE-NEXT:    movaps %xmm5, 96(%r9)
-; SSE-NEXT:    movaps %xmm8, 144(%r9)
+; SSE-NEXT:    movaps %xmm0, 96(%r9)
 ; SSE-NEXT:    retq
 ;
 ; AVX1-ONLY-LABEL: store_i32_stride5_vf8:
 ; AVX1-ONLY:       # %bb.0:
 ; AVX1-ONLY-NEXT:    vmovaps (%rdi), %ymm1
 ; AVX1-ONLY-NEXT:    vmovaps (%rsi), %ymm3
 ; AVX1-ONLY-NEXT:    vmovaps (%rdx), %ymm2
 ; AVX1-ONLY-NEXT:    vmovaps (%rcx), %ymm4
; AVX512-NEXT:    retq
   %interleaved.vec = shufflevector <40 x i32> %5, <40 x i32> poison, <40 x i32> <i32 0, i32 8, i32 16, i32 24, i32 32, i32 1, i32 9, i32 17, i32 25, i32 33, i32 2, i32 10, i32 18, i32 26, i32 34, i32 3, i32 11, i32 19, i32 27, i32 35, i32 4, i32 12, i32 20, i32 28, i32 36, i32 5, i32 13, i32 21, i32 29, i32 37, i32 6, i32 14, i32 22, i32 30, i32 38, i32 7, i32 15, i32 23, i32 31, i32 39>
   store <40 x i32> %interleaved.vec, ptr %out.vec, align 64
   ret void
 }
 
 define void @store_i32_stride5_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vecptr2, ptr %in.vecptr3, ptr %in.vecptr4, ptr %out.vec) nounwind {
 ; SSE-LABEL: store_i32_stride5_vf16:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    subq $168, %rsp
-; SSE-NEXT:    movdqa (%rsi), %xmm10
-; SSE-NEXT:    movdqa 16(%rsi), %xmm6
-; SSE-NEXT:    movdqa 32(%rsi), %xmm7
-; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa (%rdx), %xmm5
-; SSE-NEXT:    movdqa 16(%rdx), %xmm9
-; SSE-NEXT:    movdqa 32(%rdx), %xmm4
-; SSE-NEXT:    movaps (%rcx), %xmm12
-; SSE-NEXT:    movaps 16(%rcx), %xmm13
-; SSE-NEXT:    movaps 32(%rcx), %xmm11
-; SSE-NEXT:    movaps (%r8), %xmm3
-; SSE-NEXT:    movaps 16(%r8), %xmm15
-; SSE-NEXT:    movaps 32(%r8), %xmm8
+; SSE-NEXT:    subq $152, %rsp
+; SSE-NEXT:    movaps (%rdi), %xmm5
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 16(%rdi), %xmm3
+; SSE-NEXT:    movaps (%rsi), %xmm8
 ; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm12, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm3[3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm10[3,3,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm5[2,3,2,3]
-; SSE-NEXT:    movss {{.*#+}} xmm2 = xmm1[0],xmm2[1,2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm13, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm15[3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm9[3,3,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm6[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[3,3,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[2,3,2,3]
-; SSE-NEXT:    movss {{.*#+}} xmm1 = xmm0[0],xmm1[1,2,3]
-; SSE-NEXT:    movaps %xmm11, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm8[3,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 48(%rsi), %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 48(%rdx), %xmm8
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[3,3,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movaps 48(%rcx), %xmm7
-; SSE-NEXT:    movaps 48(%r8), %xmm2
-; SSE-NEXT:    movaps %xmm2, (%rsp) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm7, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm2[3,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm5, %xmm0
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm12[0],xmm0[1],xmm12[1]
-; SSE-NEXT:    movdqa (%rdi), %xmm14
-; SSE-NEXT:    movdqa %xmm14, %xmm2
-; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm10[0],xmm2[1],xmm10[1]
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm0[0]
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm14, %xmm0
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm10[2],xmm0[3],xmm10[3]
-; SSE-NEXT:    movdqa %xmm5, %xmm2
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm12[2],xmm2[3],xmm12[3]
+; SSE-NEXT:    movaps 16(%rsi), %xmm6
+; SSE-NEXT:    movaps (%rdx), %xmm14
+; SSE-NEXT:    movaps 16(%rdx), %xmm7
+; SSE-NEXT:    movaps (%rcx), %xmm1
+; SSE-NEXT:    movaps 16(%rcx), %xmm0
+; SSE-NEXT:    movaps (%r8), %xmm11
+; SSE-NEXT:    movaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 16(%r8), %xmm10
+; SSE-NEXT:    movaps %xmm14, %xmm2
+; SSE-NEXT:    unpcklps {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; SSE-NEXT:    movaps %xmm5, %xmm4
+; SSE-NEXT:    unpcklps {{.*#+}} xmm4 = xmm4[0],xmm8[0],xmm4[1],xmm8[1]
+; SSE-NEXT:    movlhps {{.*#+}} xmm4 = xmm4[0],xmm2[0]
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm5, %xmm2
+; SSE-NEXT:    unpckhps {{.*#+}} xmm2 = xmm2[2],xmm8[2],xmm2[3],xmm8[3]
+; SSE-NEXT:    movaps %xmm1, %xmm4
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[1,1],xmm11[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,2],xmm2[0,1]
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm5, %xmm2
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[3,0],xmm11[2,0]
+; SSE-NEXT:    movaps %xmm14, %xmm4
+; SSE-NEXT:    unpckhps {{.*#+}} xmm4 = xmm4[2],xmm1[2],xmm4[3],xmm1[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm2[2,0]
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,3],xmm11[3,3]
+; SSE-NEXT:    movaps %xmm8, %xmm2
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[3,3],xmm14[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,2],xmm1[0,2]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm7, %xmm1
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
 ; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[1,1],xmm3[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[0,2],xmm0[0,1]
-; SSE-NEXT:    movaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm14, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,0],xmm3[2,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm0[2,0]
+; SSE-NEXT:    movaps %xmm3, %xmm2
+; SSE-NEXT:    unpcklps {{.*#+}} xmm2 = xmm2[0],xmm6[0],xmm2[1],xmm6[1]
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm1[0]
 ; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm9, %xmm1
-; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm9, %xmm0
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm13[0],xmm0[1],xmm13[1]
-; SSE-NEXT:    movdqa 16(%rdi), %xmm10
-; SSE-NEXT:    movdqa %xmm10, %xmm12
-; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckldq {{.*#+}} xmm12 = xmm12[0],xmm6[0],xmm12[1],xmm6[1]
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm12 = xmm12[0],xmm0[0]
-; SSE-NEXT:    movdqa %xmm10, %xmm0
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm6[2],xmm0[3],xmm6[3]
-; SSE-NEXT:    movaps %xmm13, %xmm9
-; SSE-NEXT:    movaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[1,1],xmm15[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[0,2],xmm0[0,1]
-; SSE-NEXT:    movdqa %xmm10, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,1],xmm15[2,3]
-; SSE-NEXT:    punpckhqdq {{.*#+}} xmm13 = xmm13[1],xmm1[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm13 = xmm13[2,0],xmm0[2,0]
-; SSE-NEXT:    movdqa %xmm4, %xmm15
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm4, %xmm0
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm11[0],xmm0[1],xmm11[1]
-; SSE-NEXT:    movdqa 32(%rdi), %xmm2
-; SSE-NEXT:    movdqa %xmm2, %xmm6
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm6 = xmm6[0],xmm3[0],xmm6[1],xmm3[1]
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm6 = xmm6[0],xmm0[0]
-; SSE-NEXT:    movdqa %xmm2, %xmm0
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm3[2],xmm0[3],xmm3[3]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm15 = xmm15[2],xmm11[2],xmm15[3],xmm11[3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[1,1],xmm1[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[0,2],xmm0[0,1]
-; SSE-NEXT:    movdqa %xmm2, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,0],xmm1[2,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm15 = xmm15[0,1],xmm0[2,0]
-; SSE-NEXT:    movdqa %xmm8, %xmm3
-; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm7[0],xmm3[1],xmm7[1]
-; SSE-NEXT:    movdqa 48(%rdi), %xmm4
-; SSE-NEXT:    movdqa %xmm4, %xmm5
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm1[0],xmm5[1],xmm1[1]
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm5 = xmm5[0],xmm3[0]
-; SSE-NEXT:    movdqa %xmm4, %xmm0
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
-; SSE-NEXT:    movaps %xmm7, %xmm3
-; SSE-NEXT:    movaps (%rsp), %xmm1 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,1],xmm1[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,2],xmm0[0,1]
-; SSE-NEXT:    movdqa %xmm4, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,1],xmm1[2,3]
-; SSE-NEXT:    punpckhqdq {{.*#+}} xmm7 = xmm7[1],xmm8[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[2,0],xmm0[2,0]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[1,1],mem[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movss {{.*#+}} xmm14 = xmm0[0],xmm14[1,2,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[1,1],mem[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm10 = xmm10[0,1],xmm0[2,0]
+; SSE-NEXT:    movaps %xmm3, %xmm1
+; SSE-NEXT:    unpckhps {{.*#+}} xmm1 = xmm1[2],xmm6[2],xmm1[3],xmm6[3]
+; SSE-NEXT:    movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm0, %xmm2
+; SSE-NEXT:    movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,1],xmm10[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,2],xmm1[0,1]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm3, %xmm1
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,1],xmm10[2,3]
+; SSE-NEXT:    movaps %xmm0, %xmm2
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm7[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm1[2,0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm10[3,3]
+; SSE-NEXT:    movaps %xmm6, %xmm1
+; SSE-NEXT:    unpckhps {{.*#+}} xmm1 = xmm1[2],xmm7[2],xmm1[3],xmm7[3]
+; SSE-NEXT:    movaps %xmm7, %xmm12
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,3],xmm0[0,2]
+; SSE-NEXT:    movaps %xmm1, (%rsp) # 16-byte Spill
+; SSE-NEXT:    movaps 32(%rdx), %xmm7
+; SSE-NEXT:    movaps 32(%rcx), %xmm0
+; SSE-NEXT:    movaps %xmm7, %xmm1
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps 32(%rdi), %xmm15
+; SSE-NEXT:    movaps 32(%rsi), %xmm13
+; SSE-NEXT:    movaps %xmm15, %xmm2
+; SSE-NEXT:    unpcklps {{.*#+}} xmm2 = xmm2[0],xmm13[0],xmm2[1],xmm13[1]
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm1[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm15, %xmm1
+; SSE-NEXT:    unpckhps {{.*#+}} xmm1 = xmm1[2],xmm13[2],xmm1[3],xmm13[3]
+; SSE-NEXT:    movaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 32(%r8), %xmm2
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm0, %xmm11
+; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[1,1],xmm2[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[0,2],xmm1[0,1]
+; SSE-NEXT:    movaps %xmm15, %xmm1
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,0],xmm2[2,0]
+; SSE-NEXT:    movaps %xmm7, %xmm10
+; SSE-NEXT:    unpckhps {{.*#+}} xmm10 = xmm10[2],xmm0[2],xmm10[3],xmm0[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm10 = xmm10[0,1],xmm1[2,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm2[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm13 = xmm13[3,3],xmm7[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm13 = xmm13[0,2],xmm0[0,2]
+; SSE-NEXT:    movaps 48(%rdx), %xmm2
+; SSE-NEXT:    movaps 48(%rcx), %xmm3
+; SSE-NEXT:    movaps %xmm2, %xmm1
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
+; SSE-NEXT:    movaps 48(%rdi), %xmm5
+; SSE-NEXT:    movaps 48(%rsi), %xmm9
+; SSE-NEXT:    movaps %xmm5, %xmm6
+; SSE-NEXT:    unpcklps {{.*#+}} xmm6 = xmm6[0],xmm9[0],xmm6[1],xmm9[1]
+; SSE-NEXT:    movlhps {{.*#+}} xmm6 = xmm6[0],xmm1[0]
+; SSE-NEXT:    movaps %xmm5, %xmm1
+; SSE-NEXT:    unpckhps {{.*#+}} xmm1 = xmm1[2],xmm9[2],xmm1[3],xmm9[3]
+; SSE-NEXT:    movaps 48(%r8), %xmm8
+; SSE-NEXT:    movaps %xmm3, %xmm4
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[1,1],xmm8[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,2],xmm1[0,1]
+; SSE-NEXT:    movaps %xmm5, %xmm0
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,1],xmm8[2,3]
+; SSE-NEXT:    movaps %xmm3, %xmm1
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm0[2,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[3,3],xmm8[3,3]
+; SSE-NEXT:    movaps %xmm9, %xmm0
+; SSE-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,3],xmm3[0,2]
+; SSE-NEXT:    shufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm14 = xmm14[1,1],mem[1,1]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm14[2,0]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT:    movss {{.*#+}} xmm3 = xmm14[0],xmm3[1,2,3]
+; SSE-NEXT:    shufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm12 = xmm12[1,1],mem[1,1]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[0,1],xmm12[2,0]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT:    movss {{.*#+}} xmm14 = xmm12[0],xmm14[1,2,3]
+; SSE-NEXT:    shufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm7 = xmm7[1,1],mem[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm15 = xmm15[0,1],xmm7[2,0]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    movss {{.*#+}} xmm15 = xmm7[0],xmm15[1,2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,1],xmm9[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,1],xmm2[2,0]
+; SSE-NEXT:    movss {{.*#+}} xmm5 = xmm8[0],xmm5[1,2,3]
+; SSE-NEXT:    movaps %xmm0, 304(%r9)
+; SSE-NEXT:    movaps %xmm1, 288(%r9)
+; SSE-NEXT:    movaps %xmm4, 272(%r9)
+; SSE-NEXT:    movaps %xmm6, 240(%r9)
+; SSE-NEXT:    movaps %xmm13, 224(%r9)
+; SSE-NEXT:    movaps %xmm10, 208(%r9)
+; SSE-NEXT:    movaps %xmm11, 192(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movss {{.*#+}} xmm10 = xmm0[0],xmm10[1,2,3]
+; SSE-NEXT:    movaps %xmm0, 160(%r9)
+; SSE-NEXT:    movaps (%rsp), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 144(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[1,1],mem[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm0[2,0]
+; SSE-NEXT:    movaps %xmm0, 128(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movss {{.*#+}} xmm2 = xmm0[0],xmm2[1,2,3]
-; SSE-NEXT:    shufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm8 = xmm8[1,1],mem[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm8[2,0]
-; SSE-NEXT:    movss {{.*#+}} xmm4 = xmm1[0],xmm4[1,2,3]
-; SSE-NEXT:    movaps %xmm7, 288(%r9)
-; SSE-NEXT:    movaps %xmm3, 272(%r9)
-; SSE-NEXT:    movdqa %xmm5, 240(%r9)
-; SSE-NEXT:    movaps %xmm15, 208(%r9)
-; SSE-NEXT:    movaps %xmm11, 192(%r9)
-; SSE-NEXT:    movdqa %xmm6, 160(%r9)
-; SSE-NEXT:    movaps %xmm13, 128(%r9)
-; SSE-NEXT:    movaps %xmm9, 112(%r9)
-; SSE-NEXT:    movdqa %xmm12, 80(%r9)
+; SSE-NEXT:    movaps %xmm0, 112(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 80(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 48(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 32(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, (%r9)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 304(%r9)
-; SSE-NEXT:    movaps %xmm4, 256(%r9)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 224(%r9)
-; SSE-NEXT:    movaps %xmm2, 176(%r9)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 144(%r9)
-; SSE-NEXT:    movaps %xmm10, 96(%r9)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 64(%r9)
-; SSE-NEXT:    movaps %xmm14, 16(%r9)
-; SSE-NEXT:    addq $168, %rsp
+; SSE-NEXT:    movaps %xmm5, 256(%r9)
+; SSE-NEXT:    movaps %xmm15, 176(%r9)
+; SSE-NEXT:    movaps %xmm14, 96(%r9)
+; SSE-NEXT:    movaps %xmm3, 16(%r9)
+; SSE-NEXT:    addq $152, %rsp
 ; SSE-NEXT:    retq
 ;
 ; AVX1-ONLY-LABEL: store_i32_stride5_vf16:
 ; AVX1-ONLY:       # %bb.0:
 ; AVX1-ONLY-NEXT:    pushq %rax
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %ymm2
 ; AVX1-ONLY-NEXT:    vmovaps (%rdi), %ymm1
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512BW-NEXT:    retq
   %interleaved.vec = shufflevector <80 x i32> %5, <80 x i32> poison, <80 x i32> <i32 0, i32 16, i32 32, i32 48, i32 64, i32 1, i32 17, i32 33, i32 49, i32 65, i32 2, i32 18, i32 34, i32 50, i32 66, i32 3, i32 19, i32 35, i32 51, i32 67, i32 4, i32 20, i32 36, i32 52, i32 68, i32 5, i32 21, i32 37, i32 53, i32 69, i32 6, i32 22, i32 38, i32 54, i32 70, i32 7, i32 23, i32 39, i32 55, i32 71, i32 8, i32 24, i32 40, i32 56, i32 72, i32 9, i32 25, i32 41, i32 57, i32 73, i32 10, i32 26, i32 42, i32 58, i32 74, i32 11, i32 27, i32 43, i32 59, i32 75, i32 12, i32 28, i32 44, i32 60, i32 76, i32 13, i32 29, i32 45, i32 61, i32 77, i32 14, i32 30, i32 46, i32 62, i32 78, i32 15, i32 31, i32 47, i32 63, i32 79>
   store <80 x i32> %interleaved.vec, ptr %out.vec, align 64
   ret void
 }
 
 define void @store_i32_stride5_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vecptr2, ptr %in.vecptr3, ptr %in.vecptr4, ptr %out.vec) nounwind {
 ; SSE-LABEL: store_i32_stride5_vf32:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    subq $728, %rsp # imm = 0x2D8
-; SSE-NEXT:    movdqa (%rsi), %xmm9
-; SSE-NEXT:    movdqa 16(%rsi), %xmm7
-; SSE-NEXT:    movdqa 32(%rsi), %xmm8
-; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa (%rdx), %xmm11
-; SSE-NEXT:    movdqa 16(%rdx), %xmm10
-; SSE-NEXT:    movdqa 32(%rdx), %xmm12
-; SSE-NEXT:    movdqa %xmm12, (%rsp) # 16-byte Spill
-; SSE-NEXT:    movaps (%rcx), %xmm3
-; SSE-NEXT:    movaps 16(%rcx), %xmm5
-; SSE-NEXT:    movaps 32(%rcx), %xmm6
-; SSE-NEXT:    movaps (%r8), %xmm4
-; SSE-NEXT:    movaps 16(%r8), %xmm15
-; SSE-NEXT:    movaps 32(%r8), %xmm13
-; SSE-NEXT:    movaps %xmm3, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm4[3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm9[3,3,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm11[2,3,2,3]
-; SSE-NEXT:    movss {{.*#+}} xmm2 = xmm1[0],xmm2[1,2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm0[0,2]
+; SSE-NEXT:    subq $664, %rsp # imm = 0x298
+; SSE-NEXT:    movaps (%rdi), %xmm5
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 16(%rdi), %xmm3
+; SSE-NEXT:    movaps (%rsi), %xmm7
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 16(%rsi), %xmm6
+; SSE-NEXT:    movaps (%rdx), %xmm9
+; SSE-NEXT:    movaps 16(%rdx), %xmm8
+; SSE-NEXT:    movaps (%rcx), %xmm1
+; SSE-NEXT:    movaps 16(%rcx), %xmm0
+; SSE-NEXT:    movaps (%r8), %xmm11
+; SSE-NEXT:    movaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 16(%r8), %xmm10
+; SSE-NEXT:    movaps %xmm9, %xmm2
+; SSE-NEXT:    unpcklps {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; SSE-NEXT:    movaps %xmm5, %xmm4
+; SSE-NEXT:    unpcklps {{.*#+}} xmm4 = xmm4[0],xmm7[0],xmm4[1],xmm7[1]
+; SSE-NEXT:    movlhps {{.*#+}} xmm4 = xmm4[0],xmm2[0]
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm5, %xmm2
+; SSE-NEXT:    unpckhps {{.*#+}} xmm2 = xmm2[2],xmm7[2],xmm2[3],xmm7[3]
+; SSE-NEXT:    movaps %xmm1, %xmm4
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[1,1],xmm11[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,2],xmm2[0,1]
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm5, %xmm2
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[3,0],xmm11[2,0]
+; SSE-NEXT:    movaps %xmm9, %xmm4
+; SSE-NEXT:    unpckhps {{.*#+}} xmm4 = xmm4[2],xmm1[2],xmm4[3],xmm1[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm2[2,0]
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,3],xmm11[3,3]
+; SSE-NEXT:    movaps %xmm7, %xmm2
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[3,3],xmm9[3,3]
+; SSE-NEXT:    movaps %xmm9, %xmm15
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,2],xmm1[0,2]
 ; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm5, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm15[3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm10[3,3,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm7[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[3,3,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm12[2,3,2,3]
-; SSE-NEXT:    movss {{.*#+}} xmm1 = xmm0[0],xmm1[1,2,3]
-; SSE-NEXT:    movaps %xmm6, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm13[3,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 48(%rsi), %xmm12
-; SSE-NEXT:    movdqa 48(%rdx), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm12[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movaps 48(%rcx), %xmm8
-; SSE-NEXT:    movaps 48(%r8), %xmm2
+; SSE-NEXT:    movaps %xmm8, %xmm1
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm3, %xmm2
+; SSE-NEXT:    unpcklps {{.*#+}} xmm2 = xmm2[0],xmm6[0],xmm2[1],xmm6[1]
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm1[0]
 ; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm8, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm2[3,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 64(%rsi), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 64(%rdx), %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
-; SSE-NEXT:    movss {{.*#+}} xmm1 = xmm0[0],xmm1[1,2,3]
-; SSE-NEXT:    movaps 64(%rcx), %xmm0
-; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 64(%r8), %xmm2
+; SSE-NEXT:    movaps %xmm3, %xmm1
+; SSE-NEXT:    unpckhps {{.*#+}} xmm1 = xmm1[2],xmm6[2],xmm1[3],xmm6[3]
+; SSE-NEXT:    movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm0, %xmm2
+; SSE-NEXT:    movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,1],xmm10[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,2],xmm1[0,1]
 ; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm2[3,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 80(%rsi), %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 80(%rdx), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movaps 80(%rcx), %xmm0
-; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 80(%r8), %xmm2
+; SSE-NEXT:    movaps %xmm3, %xmm1
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,1],xmm10[2,3]
+; SSE-NEXT:    movaps %xmm0, %xmm2
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm8[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm1[2,0]
 ; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm2[3,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[0,2]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm10[3,3]
+; SSE-NEXT:    movaps %xmm6, %xmm1
+; SSE-NEXT:    unpckhps {{.*#+}} xmm1 = xmm1[2],xmm8[2],xmm1[3],xmm8[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,3],xmm0[0,2]
 ; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 96(%rsi), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 96(%rdx), %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
-; SSE-NEXT:    movss {{.*#+}} xmm1 = xmm0[0],xmm1[1,2,3]
-; SSE-NEXT:    movaps 96(%rcx), %xmm0
-; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 96(%r8), %xmm2
+; SSE-NEXT:    movaps 32(%rdx), %xmm4
+; SSE-NEXT:    movaps 32(%rcx), %xmm0
+; SSE-NEXT:    movaps %xmm4, %xmm1
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps 32(%rdi), %xmm2
 ; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm2[3,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 112(%rsi), %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 112(%rdx), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movaps 112(%rcx), %xmm0
-; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 112(%r8), %xmm14
-; SSE-NEXT:    movaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm14[3,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm11, %xmm1
-; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm11, %xmm0
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
-; SSE-NEXT:    movdqa (%rdi), %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm2, %xmm11
-; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckldq {{.*#+}} xmm11 = xmm11[0],xmm9[0],xmm11[1],xmm9[1]
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm11 = xmm11[0],xmm0[0]
-; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm2, %xmm0
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm9[2],xmm0[3],xmm9[3]
-; SSE-NEXT:    movdqa %xmm1, %xmm9
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm9 = xmm9[2],xmm3[2],xmm9[3],xmm3[3]
-; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,1],xmm4[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,2],xmm0[0,1]
+; SSE-NEXT:    movaps 32(%rsi), %xmm5
+; SSE-NEXT:    movaps %xmm2, %xmm3
+; SSE-NEXT:    unpcklps {{.*#+}} xmm3 = xmm3[0],xmm5[0],xmm3[1],xmm5[1]
+; SSE-NEXT:    movlhps {{.*#+}} xmm3 = xmm3[0],xmm1[0]
 ; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm2, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,0],xmm4[2,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm10, %xmm0
-; SSE-NEXT:    movaps %xmm5, %xmm4
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1]
-; SSE-NEXT:    movdqa 16(%rdi), %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm2, %xmm5
-; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm7[0],xmm5[1],xmm7[1]
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm5 = xmm5[0],xmm0[0]
-; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm2, %xmm0
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm7[2],xmm0[3],xmm7[3]
-; SSE-NEXT:    movaps %xmm4, %xmm5
-; SSE-NEXT:    movaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[1,1],xmm15[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,2],xmm0[0,1]
+; SSE-NEXT:    movaps %xmm2, %xmm1
+; SSE-NEXT:    unpckhps {{.*#+}} xmm1 = xmm1[2],xmm5[2],xmm1[3],xmm5[3]
+; SSE-NEXT:    movaps %xmm5, %xmm6
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 32(%r8), %xmm5
 ; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm2, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,1],xmm15[2,3]
-; SSE-NEXT:    unpckhpd {{.*#+}} xmm4 = xmm4[1],xmm10[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[2,0],xmm0[2,0]
+; SSE-NEXT:    movaps %xmm0, %xmm3
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,1],xmm5[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,2],xmm1[0,1]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm2, %xmm1
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,0],xmm5[2,0]
 ; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps (%rsp), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, %xmm0
-; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm6[0],xmm0[1],xmm6[1]
-; SSE-NEXT:    movaps 32(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm4, %xmm2
+; SSE-NEXT:    unpckhps {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,0]
 ; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm5[3,3]
+; SSE-NEXT:    movaps %xmm6, %xmm1
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,3],xmm4[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,2],xmm0[0,2]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 48(%rdx), %xmm4
+; SSE-NEXT:    movaps 48(%rcx), %xmm0
+; SSE-NEXT:    movaps %xmm4, %xmm1
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps 48(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 48(%rsi), %xmm5
 ; SSE-NEXT:    movaps %xmm2, %xmm3
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
 ; SSE-NEXT:    unpcklps {{.*#+}} xmm3 = xmm3[0],xmm5[0],xmm3[1],xmm5[1]
-; SSE-NEXT:    movlhps {{.*#+}} xmm3 = xmm3[0],xmm0[0]
+; SSE-NEXT:    movlhps {{.*#+}} xmm3 = xmm3[0],xmm1[0]
 ; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm2, %xmm0
-; SSE-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2],xmm5[2],xmm0[3],xmm5[3]
-; SSE-NEXT:    movaps %xmm1, %xmm3
-; SSE-NEXT:    unpckhps {{.*#+}} xmm3 = xmm3[2],xmm6[2],xmm3[3],xmm6[3]
-; SSE-NEXT:    movaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[1,1],xmm13[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,2],xmm0[0,1]
-; SSE-NEXT:    movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm2, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,0],xmm13[2,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm0[2,0]
+; SSE-NEXT:    movaps %xmm2, %xmm1
+; SSE-NEXT:    unpckhps {{.*#+}} xmm1 = xmm1[2],xmm5[2],xmm1[3],xmm5[3]
+; SSE-NEXT:    movaps %xmm5, %xmm6
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 48(%r8), %xmm5
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm0, %xmm3
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,1],xmm5[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,2],xmm1[0,1]
 ; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, %xmm0
-; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1]
-; SSE-NEXT:    movdqa 48(%rdi), %xmm10
-; SSE-NEXT:    movdqa %xmm10, %xmm3
-; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm12[0],xmm3[1],xmm12[1]
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm3 = xmm3[0],xmm0[0]
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm10, %xmm0
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm12[2],xmm0[3],xmm12[3]
-; SSE-NEXT:    movaps %xmm8, %xmm3
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,1],xmm4[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,2],xmm0[0,1]
+; SSE-NEXT:    movaps %xmm2, %xmm1
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,1],xmm5[2,3]
+; SSE-NEXT:    movaps %xmm0, %xmm2
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm4[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm1[2,0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm5[3,3]
+; SSE-NEXT:    movaps %xmm6, %xmm1
+; SSE-NEXT:    unpckhps {{.*#+}} xmm1 = xmm1[2],xmm4[2],xmm1[3],xmm4[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,3],xmm0[0,2]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 64(%rdx), %xmm4
+; SSE-NEXT:    movaps 64(%rcx), %xmm0
+; SSE-NEXT:    movaps %xmm4, %xmm1
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps 64(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 64(%rsi), %xmm5
+; SSE-NEXT:    movaps %xmm2, %xmm3
+; SSE-NEXT:    unpcklps {{.*#+}} xmm3 = xmm3[0],xmm5[0],xmm3[1],xmm5[1]
+; SSE-NEXT:    movlhps {{.*#+}} xmm3 = xmm3[0],xmm1[0]
 ; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm10, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,1],xmm4[2,3]
-; SSE-NEXT:    unpckhpd {{.*#+}} xmm8 = xmm8[1],xmm1[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[2,0],xmm0[2,0]
-; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, %xmm0
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1]
-; SSE-NEXT:    movaps 64(%rdi), %xmm11
-; SSE-NEXT:    movaps %xmm11, %xmm3
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
-; SSE-NEXT:    movlhps {{.*#+}} xmm3 = xmm3[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm2, %xmm1
+; SSE-NEXT:    unpckhps {{.*#+}} xmm1 = xmm1[2],xmm5[2],xmm1[3],xmm5[3]
+; SSE-NEXT:    movaps %xmm5, %xmm6
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 64(%r8), %xmm5
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm0, %xmm3
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,1],xmm5[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,2],xmm1[0,1]
 ; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm11, %xmm0
-; SSE-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
-; SSE-NEXT:    movaps %xmm1, %xmm2
-; SSE-NEXT:    unpckhps {{.*#+}} xmm2 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[1,1],xmm1[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,2],xmm0[0,1]
-; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm11, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,0],xmm1[2,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm0[2,0]
+; SSE-NEXT:    movaps %xmm2, %xmm1
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,0],xmm5[2,0]
+; SSE-NEXT:    movaps %xmm4, %xmm2
+; SSE-NEXT:    unpckhps {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,0]
 ; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm3, %xmm0
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT:    movaps 80(%rdi), %xmm6
-; SSE-NEXT:    movaps %xmm6, %xmm13
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm13 = xmm13[0],xmm2[0],xmm13[1],xmm2[1]
-; SSE-NEXT:    movlhps {{.*#+}} xmm13 = xmm13[0],xmm0[0]
-; SSE-NEXT:    movaps %xmm6, %xmm0
-; SSE-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
-; SSE-NEXT:    movaps %xmm1, %xmm9
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[1,1],xmm2[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[0,2],xmm0[0,1]
-; SSE-NEXT:    movaps %xmm6, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,1],xmm2[2,3]
-; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm3[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm0[2,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm5[3,3]
+; SSE-NEXT:    movaps %xmm6, %xmm1
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,3],xmm4[3,3]
+; SSE-NEXT:    movaps %xmm4, %xmm14
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,2],xmm0[0,2]
 ; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm14, %xmm0
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT:    movaps 96(%rdi), %xmm5
-; SSE-NEXT:    movaps %xmm5, %xmm8
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm8 = xmm8[0],xmm2[0],xmm8[1],xmm2[1]
-; SSE-NEXT:    movlhps {{.*#+}} xmm8 = xmm8[0],xmm0[0]
-; SSE-NEXT:    movaps %xmm5, %xmm0
-; SSE-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
-; SSE-NEXT:    movaps %xmm14, %xmm7
-; SSE-NEXT:    unpckhps {{.*#+}} xmm7 = xmm7[2],xmm1[2],xmm7[3],xmm1[3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,1],xmm2[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,2],xmm0[0,1]
+; SSE-NEXT:    movaps 80(%rdx), %xmm12
+; SSE-NEXT:    movaps 80(%rcx), %xmm0
+; SSE-NEXT:    movaps %xmm12, %xmm1
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps 80(%rdi), %xmm3
+; SSE-NEXT:    movaps %xmm3, (%rsp) # 16-byte Spill
+; SSE-NEXT:    movaps 80(%rsi), %xmm4
+; SSE-NEXT:    movaps %xmm3, %xmm2
+; SSE-NEXT:    unpcklps {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1]
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm1[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm3, %xmm1
+; SSE-NEXT:    unpckhps {{.*#+}} xmm1 = xmm1[2],xmm4[2],xmm1[3],xmm4[3]
+; SSE-NEXT:    movaps %xmm4, %xmm5
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 80(%r8), %xmm4
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm0, %xmm2
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,1],xmm4[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,2],xmm1[0,1]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm3, %xmm1
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,1],xmm4[2,3]
+; SSE-NEXT:    movaps %xmm0, %xmm2
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm12[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm1[2,0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm4[3,3]
+; SSE-NEXT:    movaps %xmm5, %xmm1
+; SSE-NEXT:    unpckhps {{.*#+}} xmm1 = xmm1[2],xmm12[2],xmm1[3],xmm12[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,3],xmm0[0,2]
 ; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm5, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,0],xmm2[2,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm2, %xmm12
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm12 = xmm12[0],xmm1[0],xmm12[1],xmm1[1]
-; SSE-NEXT:    movaps 112(%rdi), %xmm3
+; SSE-NEXT:    movaps 96(%rdx), %xmm9
+; SSE-NEXT:    movaps 96(%rcx), %xmm0
+; SSE-NEXT:    movaps %xmm9, %xmm1
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps 96(%rdi), %xmm13
+; SSE-NEXT:    movaps 96(%rsi), %xmm10
+; SSE-NEXT:    movaps %xmm13, %xmm2
+; SSE-NEXT:    unpcklps {{.*#+}} xmm2 = xmm2[0],xmm10[0],xmm2[1],xmm10[1]
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm1[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm13, %xmm1
+; SSE-NEXT:    unpckhps {{.*#+}} xmm1 = xmm1[2],xmm10[2],xmm1[3],xmm10[3]
+; SSE-NEXT:    movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 96(%r8), %xmm3
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm0, %xmm2
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,1],xmm3[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,2],xmm1[0,1]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm13, %xmm1
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,0],xmm3[2,0]
+; SSE-NEXT:    movaps %xmm9, %xmm11
+; SSE-NEXT:    unpckhps {{.*#+}} xmm11 = xmm11[2],xmm0[2],xmm11[3],xmm0[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[0,1],xmm1[2,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm3[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm10 = xmm10[3,3],xmm9[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm10 = xmm10[0,2],xmm0[0,2]
+; SSE-NEXT:    movaps 112(%rdx), %xmm0
+; SSE-NEXT:    movaps 112(%rcx), %xmm3
+; SSE-NEXT:    movaps %xmm0, %xmm1
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
+; SSE-NEXT:    movaps 112(%rdi), %xmm2
+; SSE-NEXT:    movaps 112(%rsi), %xmm8
+; SSE-NEXT:    movaps %xmm2, %xmm5
+; SSE-NEXT:    unpcklps {{.*#+}} xmm5 = xmm5[0],xmm8[0],xmm5[1],xmm8[1]
+; SSE-NEXT:    movlhps {{.*#+}} xmm5 = xmm5[0],xmm1[0]
+; SSE-NEXT:    movaps %xmm2, %xmm1
+; SSE-NEXT:    unpckhps {{.*#+}} xmm1 = xmm1[2],xmm8[2],xmm1[3],xmm8[3]
+; SSE-NEXT:    movaps 112(%r8), %xmm6
 ; SSE-NEXT:    movaps %xmm3, %xmm4
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[1,1],xmm6[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,2],xmm1[0,1]
+; SSE-NEXT:    movaps %xmm2, %xmm7
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[3,1],xmm6[2,3]
+; SSE-NEXT:    movaps %xmm3, %xmm1
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm7[2,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[3,3],xmm6[3,3]
+; SSE-NEXT:    movaps %xmm8, %xmm7
+; SSE-NEXT:    unpckhps {{.*#+}} xmm7 = xmm7[2],xmm0[2],xmm7[3],xmm0[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[2,3],xmm3[0,2]
+; SSE-NEXT:    shufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = xmm15[1,1],mem[1,1]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm15[2,0]
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm4 = xmm4[0],xmm15[0],xmm4[1],xmm15[1]
-; SSE-NEXT:    movlhps {{.*#+}} xmm4 = xmm4[0],xmm12[0]
-; SSE-NEXT:    movaps %xmm3, %xmm0
-; SSE-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2],xmm15[2],xmm0[3],xmm15[3]
-; SSE-NEXT:    movaps %xmm1, %xmm15
-; SSE-NEXT:    movaps %xmm1, %xmm12
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[1,1],xmm1[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[0,2],xmm0[0,1]
-; SSE-NEXT:    movaps %xmm3, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,1],xmm1[2,3]
-; SSE-NEXT:    unpckhpd {{.*#+}} xmm15 = xmm15[1],xmm2[1]
-; SSE-NEXT:    movaps %xmm2, %xmm1
-; SSE-NEXT:    shufps {{.*#+}} xmm15 = xmm15[2,0],xmm0[2,0]
-; SSE-NEXT:    movaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    shufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm2 = xmm2[1,1],mem[1,1]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm2[2,0]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    movss {{.*#+}} xmm0 = xmm2[0],xmm0[1,2,3]
-; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[1,1],mem[1,1]
+; SSE-NEXT:    movss {{.*#+}} xmm3 = xmm15[0],xmm3[1,2,3]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm15 = xmm15[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movss {{.*#+}} xmm15 = xmm0[0],xmm15[1,2,3]
-; SSE-NEXT:    movaps (%rsp), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[1,1],mem[1,1]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    shufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = xmm15[1,1],mem[1,1]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm15[2,0]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movss {{.*#+}} xmm3 = xmm15[0],xmm3[1,2,3]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    shufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = xmm15[1,1],mem[1,1]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm15[2,0]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movss {{.*#+}} xmm3 = xmm15[0],xmm3[1,2,3]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    shufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = xmm15[1,1],mem[1,1]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm15[2,0]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movss {{.*#+}} xmm3 = xmm15[0],xmm3[1,2,3]
+; SSE-NEXT:    shufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm14 = xmm14[1,1],mem[1,1]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    shufps {{.*#+}} xmm15 = xmm15[0,1],xmm14[2,0]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT:    movss {{.*#+}} xmm15 = xmm14[0],xmm15[1,2,3]
+; SSE-NEXT:    shufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm12 = xmm12[1,1],mem[1,1]
+; SSE-NEXT:    movaps (%rsp), %xmm14 # 16-byte Reload
+; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[0,1],xmm12[2,0]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; SSE-NEXT:    movss {{.*#+}} xmm14 = xmm12[0],xmm14[1,2,3]
+; SSE-NEXT:    shufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm9 = xmm9[1,1],mem[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm13 = xmm13[0,1],xmm9[2,0]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    movss {{.*#+}} xmm13 = xmm9[0],xmm13[1,2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm8[1,1]
 ; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm0[2,0]
+; SSE-NEXT:    movss {{.*#+}} xmm2 = xmm6[0],xmm2[1,2,3]
+; SSE-NEXT:    movaps %xmm7, 624(%r9)
+; SSE-NEXT:    movaps %xmm1, 608(%r9)
+; SSE-NEXT:    movaps %xmm4, 592(%r9)
+; SSE-NEXT:    movaps %xmm5, 560(%r9)
+; SSE-NEXT:    movaps %xmm10, 544(%r9)
+; SSE-NEXT:    movaps %xmm11, 528(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movss {{.*#+}} xmm2 = xmm0[0],xmm2[1,2,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[1,1],mem[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm10 = xmm10[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movss {{.*#+}} xmm10 = xmm0[0],xmm10[1,2,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[1,1],mem[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movss {{.*#+}} xmm11 = xmm0[0],xmm11[1,2,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[1,1],mem[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,1],xmm0[2,0]
+; SSE-NEXT:    movaps %xmm0, 512(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movss {{.*#+}} xmm6 = xmm0[0],xmm6[1,2,3]
-; SSE-NEXT:    shufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm14 = xmm14[1,1],mem[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,1],xmm14[2,0]
+; SSE-NEXT:    movaps %xmm0, 480(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movss {{.*#+}} xmm5 = xmm0[0],xmm5[1,2,3]
-; SSE-NEXT:    movaps %xmm1, %xmm0
-; SSE-NEXT:    shufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[1,1],mem[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm0[2,0]
+; SSE-NEXT:    movaps %xmm0, 464(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movss {{.*#+}} xmm3 = xmm0[0],xmm3[1,2,3]
+; SSE-NEXT:    movaps %xmm0, 448(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 608(%r9)
-; SSE-NEXT:    movaps %xmm12, 592(%r9)
-; SSE-NEXT:    movaps %xmm4, 560(%r9)
-; SSE-NEXT:    movaps %xmm7, 528(%r9)
+; SSE-NEXT:    movaps %xmm0, 432(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 512(%r9)
-; SSE-NEXT:    movaps %xmm8, 480(%r9)
+; SSE-NEXT:    movaps %xmm0, 400(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 448(%r9)
-; SSE-NEXT:    movaps %xmm9, 432(%r9)
-; SSE-NEXT:    movaps %xmm13, 400(%r9)
+; SSE-NEXT:    movaps %xmm0, 384(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 368(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 352(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 320(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 304(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 288(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 272(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 240(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 224(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 208(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 192(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 160(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 144(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 128(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 112(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 80(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 48(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 32(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, (%r9)
+; SSE-NEXT:    movaps %xmm2, 576(%r9)
+; SSE-NEXT:    movaps %xmm13, 496(%r9)
+; SSE-NEXT:    movaps %xmm14, 416(%r9)
+; SSE-NEXT:    movaps %xmm15, 336(%r9)
+; SSE-NEXT:    movaps %xmm3, 256(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 624(%r9)
-; SSE-NEXT:    movaps %xmm3, 576(%r9)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 544(%r9)
-; SSE-NEXT:    movaps %xmm5, 496(%r9)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 464(%r9)
-; SSE-NEXT:    movaps %xmm6, 416(%r9)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 384(%r9)
-; SSE-NEXT:    movaps %xmm11, 336(%r9)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 304(%r9)
-; SSE-NEXT:    movaps %xmm10, 256(%r9)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 224(%r9)
-; SSE-NEXT:    movaps %xmm2, 176(%r9)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 144(%r9)
-; SSE-NEXT:    movaps %xmm15, 96(%r9)
+; SSE-NEXT:    movaps %xmm0, 176(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 64(%r9)
+; SSE-NEXT:    movaps %xmm0, 96(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 16(%r9)
-; SSE-NEXT:    addq $728, %rsp # imm = 0x2D8
+; SSE-NEXT:    addq $664, %rsp # imm = 0x298
 ; SSE-NEXT:    retq
 ;
 ; AVX1-ONLY-LABEL: store_i32_stride5_vf32:
 ; AVX1-ONLY:       # %bb.0:
 ; AVX1-ONLY-NEXT:    subq $728, %rsp # imm = 0x2D8
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %ymm6
 ; AVX1-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps (%rdi), %ymm4
; AVX512BW-NEXT:    retq
   %interleaved.vec = shufflevector <160 x i32> %5, <160 x i32> poison, <160 x i32> <i32 0, i32 32, i32 64, i32 96, i32 128, i32 1, i32 33, i32 65, i32 97, i32 129, i32 2, i32 34, i32 66, i32 98, i32 130, i32 3, i32 35, i32 67, i32 99, i32 131, i32 4, i32 36, i32 68, i32 100, i32 132, i32 5, i32 37, i32 69, i32 101, i32 133, i32 6, i32 38, i32 70, i32 102, i32 134, i32 7, i32 39, i32 71, i32 103, i32 135, i32 8, i32 40, i32 72, i32 104, i32 136, i32 9, i32 41, i32 73, i32 105, i32 137, i32 10, i32 42, i32 74, i32 106, i32 138, i32 11, i32 43, i32 75, i32 107, i32 139, i32 12, i32 44, i32 76, i32 108, i32 140, i32 13, i32 45, i32 77, i32 109, i32 141, i32 14, i32 46, i32 78, i32 110, i32 142, i32 15, i32 47, i32 79, i32 111, i32 143, i32 16, i32 48, i32 80, i32 112, i32 144, i32 17, i32 49, i32 81, i32 113, i32 145, i32 18, i32 50, i32 82, i32 114, i32 146, i32 19, i32 51, i32 83, i32 115, i32 147, i32 20, i32 52, i32 84, i32 116, i32 148, i32 21, i32 53, i32 85, i32 117, i32 149, i32 22, i32 54, i32 86, i32 118, i32 150, i32 23, i32 55, i32 87, i32 119, i32 151, i32 24, i32 56, i32 88, i32 120, i32 152, i32 25, i32 57, i32 89, i32 121, i32 153, i32 26, i32 58, i32 90, i32 122, i32 154, i32 27, i32 59, i32 91, i32 123, i32 155, i32 28, i32 60, i32 92, i32 124, i32 156, i32 29, i32 61, i32 93, i32 125, i32 157, i32 30, i32 62, i32 94, i32 126, i32 158, i32 31, i32 63, i32 95, i32 127, i32 159>
   store <160 x i32> %interleaved.vec, ptr %out.vec, align 64
   ret void
 }
 
 define void @store_i32_stride5_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vecptr2, ptr %in.vecptr3, ptr %in.vecptr4, ptr %out.vec) nounwind {
 ; SSE-LABEL: store_i32_stride5_vf64:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    subq $1736, %rsp # imm = 0x6C8
-; SSE-NEXT:    movdqa (%rsi), %xmm12
-; SSE-NEXT:    movdqa 16(%rsi), %xmm6
-; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 32(%rsi), %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa (%rdx), %xmm8
-; SSE-NEXT:    movdqa 16(%rdx), %xmm11
-; SSE-NEXT:    movdqa 32(%rdx), %xmm10
-; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps (%rcx), %xmm4
-; SSE-NEXT:    movaps 16(%rcx), %xmm9
-; SSE-NEXT:    movaps 32(%rcx), %xmm13
-; SSE-NEXT:    movaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps (%r8), %xmm5
-; SSE-NEXT:    movaps 16(%r8), %xmm7
-; SSE-NEXT:    movaps 32(%r8), %xmm14
-; SSE-NEXT:    movaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm4, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm5[3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm12[3,3,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm8[2,3,2,3]
-; SSE-NEXT:    movss {{.*#+}} xmm2 = xmm1[0],xmm2[1,2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm0[0,2]
+; SSE-NEXT:    subq $1688, %rsp # imm = 0x698
+; SSE-NEXT:    movaps (%rdi), %xmm5
+; SSE-NEXT:    movaps 16(%rdi), %xmm3
+; SSE-NEXT:    movaps (%rsi), %xmm7
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 16(%rsi), %xmm6
+; SSE-NEXT:    movaps (%rdx), %xmm9
+; SSE-NEXT:    movaps 16(%rdx), %xmm8
+; SSE-NEXT:    movaps (%rcx), %xmm1
+; SSE-NEXT:    movaps 16(%rcx), %xmm0
+; SSE-NEXT:    movaps (%r8), %xmm11
+; SSE-NEXT:    movaps 16(%r8), %xmm10
+; SSE-NEXT:    movaps %xmm9, %xmm2
+; SSE-NEXT:    unpcklps {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm5, %xmm4
+; SSE-NEXT:    unpcklps {{.*#+}} xmm4 = xmm4[0],xmm7[0],xmm4[1],xmm7[1]
+; SSE-NEXT:    movlhps {{.*#+}} xmm4 = xmm4[0],xmm2[0]
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm5, %xmm2
+; SSE-NEXT:    unpckhps {{.*#+}} xmm2 = xmm2[2],xmm7[2],xmm2[3],xmm7[3]
+; SSE-NEXT:    movaps %xmm1, %xmm4
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[1,1],xmm11[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,2],xmm2[0,1]
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm5, %xmm2
+; SSE-NEXT:    movaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[3,0],xmm11[2,0]
+; SSE-NEXT:    movaps %xmm9, %xmm4
+; SSE-NEXT:    unpckhps {{.*#+}} xmm4 = xmm4[2],xmm1[2],xmm4[3],xmm1[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm2[2,0]
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,3],xmm11[3,3]
+; SSE-NEXT:    movaps %xmm7, %xmm2
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[3,3],xmm9[3,3]
+; SSE-NEXT:    movaps %xmm9, %xmm15
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,2],xmm1[0,2]
 ; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm9, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm7[3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm11[3,3,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm6[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[3,3,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm10[2,3,2,3]
-; SSE-NEXT:    movss {{.*#+}} xmm1 = xmm0[0],xmm1[1,2,3]
-; SSE-NEXT:    movaps %xmm13, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm14[3,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 48(%rsi), %xmm10
-; SSE-NEXT:    movdqa 48(%rdx), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm10[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movaps 48(%rcx), %xmm6
-; SSE-NEXT:    movaps 48(%r8), %xmm13
-; SSE-NEXT:    movaps %xmm6, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm13[3,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 64(%rsi), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 64(%rdx), %xmm15
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm15[2,3,2,3]
-; SSE-NEXT:    movss {{.*#+}} xmm1 = xmm0[0],xmm1[1,2,3]
-; SSE-NEXT:    movaps 64(%rcx), %xmm14
-; SSE-NEXT:    movaps 64(%r8), %xmm2
+; SSE-NEXT:    movaps %xmm8, %xmm1
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm3, %xmm2
+; SSE-NEXT:    unpcklps {{.*#+}} xmm2 = xmm2[0],xmm6[0],xmm2[1],xmm6[1]
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm1[0]
 ; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm14, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm2[3,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 80(%rsi), %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 80(%rdx), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movaps 80(%rcx), %xmm0
-; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 80(%r8), %xmm2
+; SSE-NEXT:    movaps %xmm3, %xmm1
+; SSE-NEXT:    unpckhps {{.*#+}} xmm1 = xmm1[2],xmm6[2],xmm1[3],xmm6[3]
+; SSE-NEXT:    movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm0, %xmm2
+; SSE-NEXT:    movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,1],xmm10[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,2],xmm1[0,1]
 ; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm2[3,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 96(%rsi), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 96(%rdx), %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
-; SSE-NEXT:    movss {{.*#+}} xmm1 = xmm0[0],xmm1[1,2,3]
-; SSE-NEXT:    movaps 96(%rcx), %xmm0
-; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 96(%r8), %xmm2
+; SSE-NEXT:    movaps %xmm3, %xmm1
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,1],xmm10[2,3]
+; SSE-NEXT:    movaps %xmm0, %xmm2
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm8[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm1[2,0]
 ; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm2[3,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[0,2]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm10[3,3]
+; SSE-NEXT:    movaps %xmm6, %xmm1
+; SSE-NEXT:    unpckhps {{.*#+}} xmm1 = xmm1[2],xmm8[2],xmm1[3],xmm8[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,3],xmm0[0,2]
 ; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 112(%rsi), %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 112(%rdx), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movaps 112(%rcx), %xmm0
-; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 112(%r8), %xmm2
+; SSE-NEXT:    movaps 32(%rdx), %xmm4
+; SSE-NEXT:    movaps 32(%rcx), %xmm0
+; SSE-NEXT:    movaps %xmm4, %xmm1
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps 32(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, (%rsp) # 16-byte Spill
+; SSE-NEXT:    movaps 32(%rsi), %xmm5
+; SSE-NEXT:    movaps %xmm2, %xmm3
+; SSE-NEXT:    unpcklps {{.*#+}} xmm3 = xmm3[0],xmm5[0],xmm3[1],xmm5[1]
+; SSE-NEXT:    movlhps {{.*#+}} xmm3 = xmm3[0],xmm1[0]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm2, %xmm1
+; SSE-NEXT:    unpckhps {{.*#+}} xmm1 = xmm1[2],xmm5[2],xmm1[3],xmm5[3]
+; SSE-NEXT:    movaps %xmm5, %xmm6
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 32(%r8), %xmm5
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm0, %xmm3
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,1],xmm5[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,2],xmm1[0,1]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm2, %xmm1
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,0],xmm5[2,0]
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm4, %xmm2
+; SSE-NEXT:    unpckhps {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,0]
 ; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm2[3,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[0,2]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm5[3,3]
+; SSE-NEXT:    movaps %xmm6, %xmm1
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,3],xmm4[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,2],xmm0[0,2]
 ; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 128(%rsi), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 128(%rdx), %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
-; SSE-NEXT:    movss {{.*#+}} xmm1 = xmm0[0],xmm1[1,2,3]
-; SSE-NEXT:    movaps 128(%rcx), %xmm0
-; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 128(%r8), %xmm2
+; SSE-NEXT:    movaps 48(%rdx), %xmm4
+; SSE-NEXT:    movaps 48(%rcx), %xmm0
+; SSE-NEXT:    movaps %xmm4, %xmm1
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps 48(%rdi), %xmm2
 ; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm2[3,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 144(%rsi), %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 144(%rdx), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movaps 144(%rcx), %xmm0
-; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 144(%r8), %xmm2
+; SSE-NEXT:    movaps 48(%rsi), %xmm5
+; SSE-NEXT:    movaps %xmm2, %xmm3
+; SSE-NEXT:    unpcklps {{.*#+}} xmm3 = xmm3[0],xmm5[0],xmm3[1],xmm5[1]
+; SSE-NEXT:    movlhps {{.*#+}} xmm3 = xmm3[0],xmm1[0]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm2, %xmm1
+; SSE-NEXT:    unpckhps {{.*#+}} xmm1 = xmm1[2],xmm5[2],xmm1[3],xmm5[3]
+; SSE-NEXT:    movaps %xmm5, %xmm6
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 48(%r8), %xmm5
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm0, %xmm3
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,1],xmm5[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,2],xmm1[0,1]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm2, %xmm1
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,1],xmm5[2,3]
+; SSE-NEXT:    movaps %xmm0, %xmm2
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm4[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm1[2,0]
 ; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm2[3,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[0,2]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm5[3,3]
+; SSE-NEXT:    movaps %xmm6, %xmm1
+; SSE-NEXT:    unpckhps {{.*#+}} xmm1 = xmm1[2],xmm4[2],xmm1[3],xmm4[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,3],xmm0[0,2]
 ; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 160(%rsi), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 160(%rdx), %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
-; SSE-NEXT:    movss {{.*#+}} xmm1 = xmm0[0],xmm1[1,2,3]
-; SSE-NEXT:    movaps 160(%rcx), %xmm0
-; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 160(%r8), %xmm2
+; SSE-NEXT:    movaps 64(%rdx), %xmm4
+; SSE-NEXT:    movaps 64(%rcx), %xmm0
+; SSE-NEXT:    movaps %xmm4, %xmm1
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps 64(%rdi), %xmm2
 ; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm2[3,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 176(%rsi), %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 176(%rdx), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movaps 176(%rcx), %xmm0
-; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 176(%r8), %xmm2
+; SSE-NEXT:    movaps 64(%rsi), %xmm5
+; SSE-NEXT:    movaps %xmm2, %xmm3
+; SSE-NEXT:    unpcklps {{.*#+}} xmm3 = xmm3[0],xmm5[0],xmm3[1],xmm5[1]
+; SSE-NEXT:    movlhps {{.*#+}} xmm3 = xmm3[0],xmm1[0]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm2, %xmm1
+; SSE-NEXT:    unpckhps {{.*#+}} xmm1 = xmm1[2],xmm5[2],xmm1[3],xmm5[3]
+; SSE-NEXT:    movaps %xmm5, %xmm6
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 64(%r8), %xmm5
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm0, %xmm3
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,1],xmm5[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,2],xmm1[0,1]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm2, %xmm1
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,0],xmm5[2,0]
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm4, %xmm2
+; SSE-NEXT:    unpckhps {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,0]
 ; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm2[3,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[0,2]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm5[3,3]
+; SSE-NEXT:    movaps %xmm6, %xmm1
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,3],xmm4[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,2],xmm0[0,2]
 ; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 192(%rsi), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 192(%rdx), %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
-; SSE-NEXT:    movss {{.*#+}} xmm1 = xmm0[0],xmm1[1,2,3]
-; SSE-NEXT:    movaps 192(%rcx), %xmm0
-; SSE-NEXT:    movaps %xmm0, (%rsp) # 16-byte Spill
-; SSE-NEXT:    movaps 192(%r8), %xmm2
+; SSE-NEXT:    movaps 80(%rdx), %xmm4
+; SSE-NEXT:    movaps 80(%rcx), %xmm0
+; SSE-NEXT:    movaps %xmm4, %xmm1
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps 80(%rdi), %xmm2
 ; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm2[3,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 208(%rsi), %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 208(%rdx), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movaps 208(%rcx), %xmm0
-; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 208(%r8), %xmm2
+; SSE-NEXT:    movaps 80(%rsi), %xmm5
+; SSE-NEXT:    movaps %xmm2, %xmm3
+; SSE-NEXT:    unpcklps {{.*#+}} xmm3 = xmm3[0],xmm5[0],xmm3[1],xmm5[1]
+; SSE-NEXT:    movlhps {{.*#+}} xmm3 = xmm3[0],xmm1[0]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm2, %xmm1
+; SSE-NEXT:    unpckhps {{.*#+}} xmm1 = xmm1[2],xmm5[2],xmm1[3],xmm5[3]
+; SSE-NEXT:    movaps %xmm5, %xmm6
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 80(%r8), %xmm5
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm0, %xmm3
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,1],xmm5[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,2],xmm1[0,1]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm2, %xmm1
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,1],xmm5[2,3]
+; SSE-NEXT:    movaps %xmm0, %xmm2
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm4[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm1[2,0]
 ; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm2[3,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[0,2]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm5[3,3]
+; SSE-NEXT:    movaps %xmm6, %xmm1
+; SSE-NEXT:    unpckhps {{.*#+}} xmm1 = xmm1[2],xmm4[2],xmm1[3],xmm4[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,3],xmm0[0,2]
 ; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 224(%rsi), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 224(%rdx), %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
-; SSE-NEXT:    movss {{.*#+}} xmm1 = xmm0[0],xmm1[1,2,3]
-; SSE-NEXT:    movaps 224(%rcx), %xmm0
-; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 224(%r8), %xmm2
+; SSE-NEXT:    movaps 96(%rdx), %xmm4
+; SSE-NEXT:    movaps 96(%rcx), %xmm0
+; SSE-NEXT:    movaps %xmm4, %xmm1
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps 96(%rdi), %xmm2
 ; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm2[3,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 240(%rsi), %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 240(%rdx), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movaps 240(%rcx), %xmm0
-; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 240(%r8), %xmm3
+; SSE-NEXT:    movaps 96(%rsi), %xmm5
+; SSE-NEXT:    movaps %xmm2, %xmm3
+; SSE-NEXT:    unpcklps {{.*#+}} xmm3 = xmm3[0],xmm5[0],xmm3[1],xmm5[1]
+; SSE-NEXT:    movlhps {{.*#+}} xmm3 = xmm3[0],xmm1[0]
 ; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm3[3,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm8, %xmm0
-; SSE-NEXT:    movaps %xmm4, %xmm3
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1]
-; SSE-NEXT:    movdqa (%rdi), %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm1, %xmm4
-; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm12[0],xmm4[1],xmm12[1]
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm4 = xmm4[0],xmm0[0]
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm1, %xmm0
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm12[2],xmm0[3],xmm12[3]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm8 = xmm8[2],xmm3[2],xmm8[3],xmm3[3]
+; SSE-NEXT:    movaps %xmm2, %xmm1
+; SSE-NEXT:    unpckhps {{.*#+}} xmm1 = xmm1[2],xmm5[2],xmm1[3],xmm5[3]
+; SSE-NEXT:    movaps %xmm5, %xmm6
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 96(%r8), %xmm5
 ; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm0, %xmm3
 ; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,1],xmm5[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,2],xmm0[0,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,2],xmm1[0,1]
 ; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm2, %xmm1
 ; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,0],xmm5[2,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,1],xmm1[2,0]
-; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm11, %xmm0
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1]
-; SSE-NEXT:    movdqa 16(%rdi), %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm3, %xmm5
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm2[0],xmm5[1],xmm2[1]
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm5 = xmm5[0],xmm0[0]
-; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm3, %xmm0
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
-; SSE-NEXT:    movaps %xmm9, %xmm8
-; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[1,1],xmm7[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,2],xmm0[0,1]
-; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm3, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,1],xmm7[2,3]
-; SSE-NEXT:    punpckhqdq {{.*#+}} xmm9 = xmm9[1],xmm11[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[2,0],xmm0[2,0]
-; SSE-NEXT:    movaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, %xmm0
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1]
-; SSE-NEXT:    movaps 32(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm4, %xmm2
+; SSE-NEXT:    unpckhps {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm5[3,3]
+; SSE-NEXT:    movaps %xmm6, %xmm1
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,3],xmm4[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,2],xmm0[0,2]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 112(%rdx), %xmm4
+; SSE-NEXT:    movaps 112(%rcx), %xmm0
+; SSE-NEXT:    movaps %xmm4, %xmm1
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps 112(%rdi), %xmm2
 ; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 112(%rsi), %xmm5
 ; SSE-NEXT:    movaps %xmm2, %xmm3
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
 ; SSE-NEXT:    unpcklps {{.*#+}} xmm3 = xmm3[0],xmm5[0],xmm3[1],xmm5[1]
-; SSE-NEXT:    movlhps {{.*#+}} xmm3 = xmm3[0],xmm0[0]
+; SSE-NEXT:    movlhps {{.*#+}} xmm3 = xmm3[0],xmm1[0]
 ; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm2, %xmm0
-; SSE-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2],xmm5[2],xmm0[3],xmm5[3]
-; SSE-NEXT:    movaps %xmm1, %xmm3
-; SSE-NEXT:    unpckhps {{.*#+}} xmm3 = xmm3[2],xmm4[2],xmm3[3],xmm4[3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[1,1],xmm1[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,2],xmm0[0,1]
-; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm2, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,0],xmm1[2,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm0[2,0]
+; SSE-NEXT:    movaps %xmm2, %xmm1
+; SSE-NEXT:    unpckhps {{.*#+}} xmm1 = xmm1[2],xmm5[2],xmm1[3],xmm5[3]
+; SSE-NEXT:    movaps %xmm5, %xmm6
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 112(%r8), %xmm5
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm0, %xmm3
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,1],xmm5[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,2],xmm1[0,1]
 ; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, %xmm0
-; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm6[0],xmm0[1],xmm6[1]
-; SSE-NEXT:    movdqa 48(%rdi), %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm3, %xmm4
-; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm10[0],xmm4[1],xmm10[1]
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm4 = xmm4[0],xmm0[0]
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm3, %xmm0
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm10[2],xmm0[3],xmm10[3]
-; SSE-NEXT:    movaps %xmm6, %xmm4
-; SSE-NEXT:    movaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[1,1],xmm13[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,2],xmm0[0,1]
+; SSE-NEXT:    movaps %xmm2, %xmm1
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,1],xmm5[2,3]
+; SSE-NEXT:    movaps %xmm0, %xmm2
 ; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm3, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,1],xmm13[2,3]
-; SSE-NEXT:    unpckhpd {{.*#+}} xmm6 = xmm6[1],xmm1[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[2,0],xmm0[2,0]
-; SSE-NEXT:    movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm15, %xmm0
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm14[0],xmm0[1],xmm14[1]
-; SSE-NEXT:    movdqa 64(%rdi), %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm3, %xmm2
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1]
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm0[0]
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm3, %xmm0
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm15 = xmm15[2],xmm14[2],xmm15[3],xmm14[3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[1,1],xmm1[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[0,2],xmm0[0,1]
-; SSE-NEXT:    movaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm3, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,0],xmm1[2,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm15 = xmm15[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm2, %xmm0
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1]
-; SSE-NEXT:    movaps 80(%rdi), %xmm3
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm4[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm1[2,0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm5[3,3]
+; SSE-NEXT:    movaps %xmm6, %xmm1
+; SSE-NEXT:    unpckhps {{.*#+}} xmm1 = xmm1[2],xmm4[2],xmm1[3],xmm4[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,3],xmm0[0,2]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 128(%rdx), %xmm2
+; SSE-NEXT:    movaps 128(%rcx), %xmm0
+; SSE-NEXT:    movaps %xmm2, %xmm1
+; SSE-NEXT:    movaps %xmm2, %xmm5
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps 128(%rdi), %xmm3
 ; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm3, %xmm4
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
-; SSE-NEXT:    movlhps {{.*#+}} xmm4 = xmm4[0],xmm0[0]
+; SSE-NEXT:    movaps 128(%rsi), %xmm4
+; SSE-NEXT:    movaps %xmm3, %xmm2
+; SSE-NEXT:    unpcklps {{.*#+}} xmm2 = xmm2[0],xmm4[0],xmm2[1],xmm4[1]
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm1[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm3, %xmm1
+; SSE-NEXT:    unpckhps {{.*#+}} xmm1 = xmm1[2],xmm4[2],xmm1[3],xmm4[3]
+; SSE-NEXT:    movaps %xmm4, %xmm6
 ; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm3, %xmm0
-; SSE-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
-; SSE-NEXT:    movaps %xmm5, %xmm4
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[1,1],xmm1[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,2],xmm0[0,1]
+; SSE-NEXT:    movaps 128(%r8), %xmm4
 ; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm3, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,1],xmm1[2,3]
-; SSE-NEXT:    unpckhpd {{.*#+}} xmm5 = xmm5[1],xmm2[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,0],xmm0[2,0]
+; SSE-NEXT:    movaps %xmm0, %xmm2
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,1],xmm4[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,2],xmm1[0,1]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm3, %xmm1
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,0],xmm4[2,0]
 ; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, %xmm0
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1]
-; SSE-NEXT:    movaps 96(%rdi), %xmm5
+; SSE-NEXT:    movaps %xmm5, %xmm2
+; SSE-NEXT:    unpckhps {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm4[3,3]
+; SSE-NEXT:    movaps %xmm6, %xmm1
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,3],xmm5[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,2],xmm0[0,2]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 144(%rdx), %xmm4
+; SSE-NEXT:    movaps 144(%rcx), %xmm0
+; SSE-NEXT:    movaps %xmm4, %xmm1
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps 144(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 144(%rsi), %xmm5
+; SSE-NEXT:    movaps %xmm2, %xmm3
+; SSE-NEXT:    unpcklps {{.*#+}} xmm3 = xmm3[0],xmm5[0],xmm3[1],xmm5[1]
+; SSE-NEXT:    movlhps {{.*#+}} xmm3 = xmm3[0],xmm1[0]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm2, %xmm1
+; SSE-NEXT:    unpckhps {{.*#+}} xmm1 = xmm1[2],xmm5[2],xmm1[3],xmm5[3]
 ; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm5, %xmm3
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
-; SSE-NEXT:    movlhps {{.*#+}} xmm3 = xmm3[0],xmm0[0]
+; SSE-NEXT:    movaps 144(%r8), %xmm6
+; SSE-NEXT:    movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm0, %xmm3
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,1],xmm6[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,2],xmm1[0,1]
 ; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm5, %xmm0
-; SSE-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
-; SSE-NEXT:    movaps %xmm1, %xmm2
-; SSE-NEXT:    unpckhps {{.*#+}} xmm2 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[1,1],xmm1[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,2],xmm0[0,1]
+; SSE-NEXT:    movaps %xmm2, %xmm1
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,1],xmm6[2,3]
+; SSE-NEXT:    movaps %xmm0, %xmm2
 ; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm5, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,0],xmm1[2,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm0[2,0]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm4[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm1[2,0]
 ; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm2, %xmm0
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1]
-; SSE-NEXT:    movaps 112(%rdi), %xmm3
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm6[3,3]
+; SSE-NEXT:    movaps %xmm5, %xmm1
+; SSE-NEXT:    unpckhps {{.*#+}} xmm1 = xmm1[2],xmm4[2],xmm1[3],xmm4[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,3],xmm0[0,2]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 160(%rdx), %xmm4
+; SSE-NEXT:    movaps 160(%rcx), %xmm0
+; SSE-NEXT:    movaps %xmm4, %xmm1
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps 160(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 160(%rsi), %xmm5
+; SSE-NEXT:    movaps %xmm2, %xmm3
+; SSE-NEXT:    unpcklps {{.*#+}} xmm3 = xmm3[0],xmm5[0],xmm3[1],xmm5[1]
+; SSE-NEXT:    movlhps {{.*#+}} xmm3 = xmm3[0],xmm1[0]
 ; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm3, %xmm4
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
-; SSE-NEXT:    movlhps {{.*#+}} xmm4 = xmm4[0],xmm0[0]
-; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm3, %xmm0
-; SSE-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
-; SSE-NEXT:    movaps %xmm5, %xmm4
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[1,1],xmm1[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,2],xmm0[0,1]
-; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm3, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,1],xmm1[2,3]
-; SSE-NEXT:    unpckhpd {{.*#+}} xmm5 = xmm5[1],xmm2[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,0],xmm0[2,0]
+; SSE-NEXT:    movaps %xmm2, %xmm1
+; SSE-NEXT:    unpckhps {{.*#+}} xmm1 = xmm1[2],xmm5[2],xmm1[3],xmm5[3]
+; SSE-NEXT:    movaps %xmm5, %xmm6
 ; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, %xmm0
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1]
-; SSE-NEXT:    movaps 128(%rdi), %xmm5
+; SSE-NEXT:    movaps 160(%r8), %xmm5
 ; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm5, %xmm3
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
-; SSE-NEXT:    movlhps {{.*#+}} xmm3 = xmm3[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm0, %xmm3
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,1],xmm5[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,2],xmm1[0,1]
 ; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm5, %xmm0
-; SSE-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
-; SSE-NEXT:    movaps %xmm1, %xmm2
-; SSE-NEXT:    unpckhps {{.*#+}} xmm2 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[1,1],xmm1[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,2],xmm0[0,1]
+; SSE-NEXT:    movaps %xmm2, %xmm1
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,0],xmm5[2,0]
 ; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm5, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,0],xmm1[2,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm0[2,0]
+; SSE-NEXT:    movaps %xmm4, %xmm2
+; SSE-NEXT:    unpckhps {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,0]
 ; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm3, %xmm0
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT:    movaps 144(%rdi), %xmm4
-; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm4, %xmm5
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm5 = xmm5[0],xmm2[0],xmm5[1],xmm2[1]
-; SSE-NEXT:    movlhps {{.*#+}} xmm5 = xmm5[0],xmm0[0]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm5[3,3]
+; SSE-NEXT:    movaps %xmm6, %xmm1
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,3],xmm4[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,2],xmm0[0,2]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 176(%rdx), %xmm4
+; SSE-NEXT:    movaps 176(%rcx), %xmm0
+; SSE-NEXT:    movaps %xmm4, %xmm1
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps 176(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 176(%rsi), %xmm5
+; SSE-NEXT:    movaps %xmm2, %xmm3
+; SSE-NEXT:    unpcklps {{.*#+}} xmm3 = xmm3[0],xmm5[0],xmm3[1],xmm5[1]
+; SSE-NEXT:    movlhps {{.*#+}} xmm3 = xmm3[0],xmm1[0]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm2, %xmm1
+; SSE-NEXT:    unpckhps {{.*#+}} xmm1 = xmm1[2],xmm5[2],xmm1[3],xmm5[3]
+; SSE-NEXT:    movaps %xmm5, %xmm6
 ; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm4, %xmm0
-; SSE-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
-; SSE-NEXT:    movaps %xmm1, %xmm5
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[1,1],xmm2[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,2],xmm0[0,1]
+; SSE-NEXT:    movaps 176(%r8), %xmm5
 ; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm4, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,1],xmm2[2,3]
-; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm3[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm0[2,0]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm2, %xmm0
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT:    movaps 160(%rdi), %xmm15
-; SSE-NEXT:    movaps %xmm15, %xmm4
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
-; SSE-NEXT:    movlhps {{.*#+}} xmm4 = xmm4[0],xmm0[0]
+; SSE-NEXT:    movaps %xmm0, %xmm3
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,1],xmm5[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,2],xmm1[0,1]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm2, %xmm1
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,1],xmm5[2,3]
+; SSE-NEXT:    movaps %xmm0, %xmm2
 ; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm15, %xmm0
-; SSE-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2],xmm3[2],xmm0[3],xmm3[3]
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm4[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm1[2,0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm5[3,3]
+; SSE-NEXT:    movaps %xmm6, %xmm1
+; SSE-NEXT:    unpckhps {{.*#+}} xmm1 = xmm1[2],xmm4[2],xmm1[3],xmm4[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,3],xmm0[0,2]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 192(%rdx), %xmm7
+; SSE-NEXT:    movaps 192(%rcx), %xmm0
+; SSE-NEXT:    movaps %xmm7, %xmm1
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps 192(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 192(%rsi), %xmm5
 ; SSE-NEXT:    movaps %xmm2, %xmm3
-; SSE-NEXT:    unpckhps {{.*#+}} xmm3 = xmm3[2],xmm1[2],xmm3[3],xmm1[3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,1],xmm2[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,2],xmm0[0,1]
+; SSE-NEXT:    unpcklps {{.*#+}} xmm3 = xmm3[0],xmm5[0],xmm3[1],xmm5[1]
+; SSE-NEXT:    movlhps {{.*#+}} xmm3 = xmm3[0],xmm1[0]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm2, %xmm1
+; SSE-NEXT:    unpckhps {{.*#+}} xmm1 = xmm1[2],xmm5[2],xmm1[3],xmm5[3]
+; SSE-NEXT:    movaps %xmm5, %xmm6
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 192(%r8), %xmm5
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm0, %xmm3
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,1],xmm5[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,2],xmm1[0,1]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm2, %xmm1
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,0],xmm5[2,0]
+; SSE-NEXT:    movaps %xmm7, %xmm2
+; SSE-NEXT:    unpckhps {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm5[3,3]
+; SSE-NEXT:    movaps %xmm6, %xmm1
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,3],xmm7[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,2],xmm0[0,2]
 ; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm15, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,0],xmm2[2,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm0[2,0]
+; SSE-NEXT:    movaps 208(%rdx), %xmm8
+; SSE-NEXT:    movaps 208(%rcx), %xmm0
+; SSE-NEXT:    movaps %xmm8, %xmm1
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps 208(%rdi), %xmm2
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 208(%rsi), %xmm4
+; SSE-NEXT:    movaps %xmm2, %xmm3
+; SSE-NEXT:    unpcklps {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
+; SSE-NEXT:    movlhps {{.*#+}} xmm3 = xmm3[0],xmm1[0]
 ; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm3, %xmm0
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT:    movaps 176(%rdi), %xmm13
-; SSE-NEXT:    movaps %xmm13, %xmm4
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm4 = xmm4[0],xmm2[0],xmm4[1],xmm2[1]
-; SSE-NEXT:    movlhps {{.*#+}} xmm4 = xmm4[0],xmm0[0]
-; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm13, %xmm0
-; SSE-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
-; SSE-NEXT:    movaps %xmm1, %xmm4
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[1,1],xmm2[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,2],xmm0[0,1]
+; SSE-NEXT:    movaps %xmm2, %xmm1
+; SSE-NEXT:    unpckhps {{.*#+}} xmm1 = xmm1[2],xmm4[2],xmm1[3],xmm4[3]
+; SSE-NEXT:    movaps %xmm4, %xmm5
 ; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm13, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,1],xmm2[2,3]
-; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm3[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm0[2,0]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm2, %xmm0
-; SSE-NEXT:    movaps (%rsp), %xmm1 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT:    movaps 192(%rdi), %xmm11
-; SSE-NEXT:    movaps %xmm11, %xmm4
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
-; SSE-NEXT:    movlhps {{.*#+}} xmm4 = xmm4[0],xmm0[0]
+; SSE-NEXT:    movaps 208(%r8), %xmm4
 ; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm11, %xmm0
-; SSE-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2],xmm3[2],xmm0[3],xmm3[3]
-; SSE-NEXT:    movaps %xmm2, %xmm3
-; SSE-NEXT:    unpckhps {{.*#+}} xmm3 = xmm3[2],xmm1[2],xmm3[3],xmm1[3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,1],xmm2[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,2],xmm0[0,1]
-; SSE-NEXT:    movaps %xmm1, (%rsp) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm11, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,0],xmm2[2,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm0[2,0]
+; SSE-NEXT:    movaps %xmm0, %xmm3
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,1],xmm4[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,2],xmm1[0,1]
 ; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm3, %xmm0
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT:    movaps 208(%rdi), %xmm7
-; SSE-NEXT:    movaps %xmm7, %xmm12
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm12 = xmm12[0],xmm2[0],xmm12[1],xmm2[1]
-; SSE-NEXT:    movlhps {{.*#+}} xmm12 = xmm12[0],xmm0[0]
-; SSE-NEXT:    movaps %xmm7, %xmm0
-; SSE-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
-; SSE-NEXT:    movaps %xmm1, %xmm10
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm10 = xmm10[1,1],xmm2[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm10 = xmm10[0,2],xmm0[0,1]
-; SSE-NEXT:    movaps %xmm7, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,1],xmm2[2,3]
-; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm3[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm0[2,0]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm8, %xmm0
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT:    movaps 224(%rdi), %xmm4
-; SSE-NEXT:    movaps %xmm4, %xmm9
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm9 = xmm9[0],xmm3[0],xmm9[1],xmm3[1]
-; SSE-NEXT:    movlhps {{.*#+}} xmm9 = xmm9[0],xmm0[0]
-; SSE-NEXT:    movaps %xmm4, %xmm0
-; SSE-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2],xmm3[2],xmm0[3],xmm3[3]
-; SSE-NEXT:    unpckhps {{.*#+}} xmm8 = xmm8[2],xmm1[2],xmm8[3],xmm1[3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,1],xmm2[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,2],xmm0[0,1]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm4, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,0],xmm2[2,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm2, %xmm1
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
-; SSE-NEXT:    movaps 240(%rdi), %xmm5
-; SSE-NEXT:    movaps %xmm5, %xmm6
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,1],xmm4[2,3]
+; SSE-NEXT:    movaps %xmm0, %xmm2
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm8[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,0],xmm1[2,0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm4[3,3]
+; SSE-NEXT:    movaps %xmm5, %xmm1
+; SSE-NEXT:    unpckhps {{.*#+}} xmm1 = xmm1[2],xmm8[2],xmm1[3],xmm8[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,3],xmm0[0,2]
+; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 224(%rdx), %xmm5
+; SSE-NEXT:    movaps 224(%rcx), %xmm0
+; SSE-NEXT:    movaps %xmm5, %xmm1
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE-NEXT:    movaps 224(%rdi), %xmm9
+; SSE-NEXT:    movaps 224(%rsi), %xmm11
+; SSE-NEXT:    movaps %xmm9, %xmm2
+; SSE-NEXT:    unpcklps {{.*#+}} xmm2 = xmm2[0],xmm11[0],xmm2[1],xmm11[1]
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm1[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm9, %xmm1
+; SSE-NEXT:    unpckhps {{.*#+}} xmm1 = xmm1[2],xmm11[2],xmm1[3],xmm11[3]
+; SSE-NEXT:    movaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 224(%r8), %xmm3
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm0, %xmm2
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,1],xmm3[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,2],xmm1[0,1]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm9, %xmm1
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,0],xmm3[2,0]
+; SSE-NEXT:    movaps %xmm5, %xmm12
+; SSE-NEXT:    unpckhps {{.*#+}} xmm12 = xmm12[2],xmm0[2],xmm12[3],xmm0[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[0,1],xmm1[2,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm3[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[3,3],xmm5[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[0,2],xmm0[0,2]
+; SSE-NEXT:    movaps 240(%rdx), %xmm0
+; SSE-NEXT:    movaps 240(%rcx), %xmm2
+; SSE-NEXT:    movaps %xmm0, %xmm1
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; SSE-NEXT:    movaps 240(%rdi), %xmm3
+; SSE-NEXT:    movaps 240(%rsi), %xmm14
+; SSE-NEXT:    movaps %xmm3, %xmm6
 ; SSE-NEXT:    unpcklps {{.*#+}} xmm6 = xmm6[0],xmm14[0],xmm6[1],xmm14[1]
 ; SSE-NEXT:    movlhps {{.*#+}} xmm6 = xmm6[0],xmm1[0]
-; SSE-NEXT:    movaps %xmm5, %xmm0
-; SSE-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2],xmm14[2],xmm0[3],xmm14[3]
-; SSE-NEXT:    movaps %xmm3, %xmm14
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[1,1],xmm1[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,2],xmm0[0,1]
-; SSE-NEXT:    movaps %xmm5, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,1],xmm1[2,3]
-; SSE-NEXT:    unpckhpd {{.*#+}} xmm14 = xmm14[1],xmm2[1]
+; SSE-NEXT:    movaps %xmm3, %xmm1
+; SSE-NEXT:    unpckhps {{.*#+}} xmm1 = xmm1[2],xmm14[2],xmm1[3],xmm14[3]
+; SSE-NEXT:    movaps 240(%r8), %xmm13
+; SSE-NEXT:    movaps %xmm2, %xmm4
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[1,1],xmm13[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,2],xmm1[0,1]
+; SSE-NEXT:    movaps %xmm3, %xmm10
+; SSE-NEXT:    shufps {{.*#+}} xmm10 = xmm10[3,1],xmm13[2,3]
 ; SSE-NEXT:    movaps %xmm2, %xmm1
-; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[2,0],xmm0[2,0]
-; SSE-NEXT:    movaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm10[2,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[3,3],xmm13[3,3]
+; SSE-NEXT:    movaps %xmm14, %xmm10
+; SSE-NEXT:    unpckhps {{.*#+}} xmm10 = xmm10[2],xmm0[2],xmm10[3],xmm0[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm10 = xmm10[2,3],xmm2[0,2]
+; SSE-NEXT:    shufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = xmm15[1,1],mem[1,1]
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    shufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm2 = xmm2[1,1],mem[1,1]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm2[2,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm15[2,0]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movss {{.*#+}} xmm2 = xmm15[0],xmm2[1,2,3]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    shufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = xmm15[1,1],mem[1,1]
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    movss {{.*#+}} xmm0 = xmm2[0],xmm0[1,2,3]
-; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[1,1],mem[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm15[2,0]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movss {{.*#+}} xmm2 = xmm15[0],xmm2[1,2,3]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    shufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = xmm15[1,1],mem[1,1]
+; SSE-NEXT:    movaps (%rsp), %xmm2 # 16-byte Reload
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm15[2,0]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movss {{.*#+}} xmm2 = xmm15[0],xmm2[1,2,3]
+; SSE-NEXT:    movaps %xmm2, (%rsp) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    shufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = xmm15[1,1],mem[1,1]
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movss {{.*#+}} xmm2 = xmm0[0],xmm2[1,2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm15[2,0]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movss {{.*#+}} xmm2 = xmm15[0],xmm2[1,2,3]
 ; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[1,1],mem[1,1]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    shufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = xmm15[1,1],mem[1,1]
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movss {{.*#+}} xmm2 = xmm0[0],xmm2[1,2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm15[2,0]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movss {{.*#+}} xmm2 = xmm15[0],xmm2[1,2,3]
 ; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[1,1],mem[1,1]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    shufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = xmm15[1,1],mem[1,1]
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movss {{.*#+}} xmm2 = xmm0[0],xmm2[1,2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm15[2,0]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movss {{.*#+}} xmm2 = xmm15[0],xmm2[1,2,3]
 ; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[1,1],mem[1,1]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    shufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = xmm15[1,1],mem[1,1]
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movss {{.*#+}} xmm2 = xmm0[0],xmm2[1,2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm15[2,0]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movss {{.*#+}} xmm2 = xmm15[0],xmm2[1,2,3]
 ; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[1,1],mem[1,1]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    shufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = xmm15[1,1],mem[1,1]
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movss {{.*#+}} xmm2 = xmm0[0],xmm2[1,2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm15[2,0]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movss {{.*#+}} xmm2 = xmm15[0],xmm2[1,2,3]
 ; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[1,1],mem[1,1]
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movss {{.*#+}} xmm2 = xmm0[0],xmm2[1,2,3]
+; SSE-NEXT:    shufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm2 = xmm2[1,1],mem[1,1]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    shufps {{.*#+}} xmm15 = xmm15[0,1],xmm2[2,0]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    movss {{.*#+}} xmm15 = xmm2[0],xmm15[1,2,3]
+; SSE-NEXT:    movaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    shufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = xmm15[1,1],mem[1,1]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm15[2,0]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movss {{.*#+}} xmm2 = xmm15[0],xmm2[1,2,3]
 ; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[1,1],mem[1,1]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    shufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = xmm15[1,1],mem[1,1]
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movss {{.*#+}} xmm2 = xmm0[0],xmm2[1,2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm15[2,0]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movss {{.*#+}} xmm2 = xmm15[0],xmm2[1,2,3]
 ; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[1,1],mem[1,1]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movss {{.*#+}} xmm14 = xmm0[0],xmm14[1,2,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[1,1],mem[1,1]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    shufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm15 = xmm15[1,1],mem[1,1]
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movss {{.*#+}} xmm2 = xmm0[0],xmm2[1,2,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[1,1],mem[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm15 = xmm15[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movss {{.*#+}} xmm15 = xmm0[0],xmm15[1,2,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[1,1],mem[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm13 = xmm13[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movss {{.*#+}} xmm13 = xmm0[0],xmm13[1,2,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[1,1],mem[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movss {{.*#+}} xmm11 = xmm0[0],xmm11[1,2,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[1,1],mem[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,1],xmm0[2,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm15[2,0]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
+; SSE-NEXT:    movss {{.*#+}} xmm2 = xmm15[0],xmm2[1,2,3]
+; SSE-NEXT:    movaps %xmm2, %xmm15
+; SSE-NEXT:    shufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm7 = xmm7[1,1],mem[1,1]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm7[2,0]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    movss {{.*#+}} xmm2 = xmm7[0],xmm2[1,2,3]
+; SSE-NEXT:    shufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm8 = xmm8[1,1],mem[1,1]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,1],xmm8[2,0]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; SSE-NEXT:    movss {{.*#+}} xmm7 = xmm8[0],xmm7[1,2,3]
+; SSE-NEXT:    shufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm5 = xmm5[1,1],mem[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[0,1],xmm5[2,0]
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    movss {{.*#+}} xmm9 = xmm5[0],xmm9[1,2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm14[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm0[2,0]
+; SSE-NEXT:    movss {{.*#+}} xmm3 = xmm13[0],xmm3[1,2,3]
+; SSE-NEXT:    movaps %xmm10, 1264(%r9)
+; SSE-NEXT:    movaps %xmm1, 1248(%r9)
+; SSE-NEXT:    movaps %xmm4, 1232(%r9)
+; SSE-NEXT:    movaps %xmm6, 1200(%r9)
+; SSE-NEXT:    movaps %xmm11, 1184(%r9)
+; SSE-NEXT:    movaps %xmm12, 1168(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movss {{.*#+}} xmm7 = xmm0[0],xmm7[1,2,3]
+; SSE-NEXT:    movaps %xmm0, 1152(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[1,1],mem[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm0[2,0]
+; SSE-NEXT:    movaps %xmm0, 1120(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movss {{.*#+}} xmm4 = xmm0[0],xmm4[1,2,3]
-; SSE-NEXT:    movaps %xmm1, %xmm0
-; SSE-NEXT:    shufps $85, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[1,1],mem[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,1],xmm0[2,0]
+; SSE-NEXT:    movaps %xmm0, 1104(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movss {{.*#+}} xmm5 = xmm0[0],xmm5[1,2,3]
+; SSE-NEXT:    movaps %xmm0, 1088(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 1248(%r9)
-; SSE-NEXT:    movaps %xmm3, 1232(%r9)
-; SSE-NEXT:    movaps %xmm6, 1200(%r9)
-; SSE-NEXT:    movaps %xmm8, 1168(%r9)
+; SSE-NEXT:    movaps %xmm0, 1072(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 1152(%r9)
-; SSE-NEXT:    movaps %xmm9, 1120(%r9)
+; SSE-NEXT:    movaps %xmm0, 1040(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 1088(%r9)
-; SSE-NEXT:    movaps %xmm10, 1072(%r9)
-; SSE-NEXT:    movaps %xmm12, 1040(%r9)
+; SSE-NEXT:    movaps %xmm0, 1024(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 1008(%r9)
-; SSE-NEXT:    movaps (%rsp), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 992(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 960(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 944(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 928(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 912(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 880(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 864(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 848(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 832(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 800(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 784(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 768(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 752(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 720(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 704(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 688(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 672(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 640(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 624(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 608(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 592(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 560(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 544(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 528(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 512(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 480(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 464(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 448(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 432(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 400(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 384(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 368(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 352(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 320(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 304(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 288(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 272(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 240(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 224(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 208(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 192(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 160(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 144(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 128(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 112(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 80(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%r9)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 48(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 32(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, (%r9)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 1264(%r9)
-; SSE-NEXT:    movaps %xmm5, 1216(%r9)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 1184(%r9)
-; SSE-NEXT:    movaps %xmm4, 1136(%r9)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 1104(%r9)
+; SSE-NEXT:    movaps %xmm3, 1216(%r9)
+; SSE-NEXT:    movaps %xmm9, 1136(%r9)
 ; SSE-NEXT:    movaps %xmm7, 1056(%r9)
+; SSE-NEXT:    movaps %xmm2, 976(%r9)
+; SSE-NEXT:    movaps %xmm15, 896(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 1024(%r9)
-; SSE-NEXT:    movaps %xmm11, 976(%r9)
+; SSE-NEXT:    movaps %xmm0, 816(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 944(%r9)
-; SSE-NEXT:    movaps %xmm13, 896(%r9)
+; SSE-NEXT:    movaps %xmm0, 736(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 864(%r9)
-; SSE-NEXT:    movaps %xmm15, 816(%r9)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 784(%r9)
-; SSE-NEXT:    movaps %xmm2, 736(%r9)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 704(%r9)
-; SSE-NEXT:    movaps %xmm14, 656(%r9)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 624(%r9)
+; SSE-NEXT:    movaps %xmm0, 656(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 576(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 544(%r9)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 496(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 464(%r9)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 416(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 384(%r9)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 336(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 304(%r9)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 256(%r9)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 224(%r9)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps (%rsp), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 176(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 144(%r9)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 96(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 64(%r9)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 16(%r9)
-; SSE-NEXT:    addq $1736, %rsp # imm = 0x6C8
+; SSE-NEXT:    addq $1688, %rsp # imm = 0x698
 ; SSE-NEXT:    retq
 ;
 ; AVX1-ONLY-LABEL: store_i32_stride5_vf64:
 ; AVX1-ONLY:       # %bb.0:
 ; AVX1-ONLY-NEXT:    subq $1944, %rsp # imm = 0x798
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %ymm9
 ; AVX1-ONLY-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps (%rdi), %ymm5

llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-7.ll

; AVX512BW-FAST-NEXT:    retq
   ret void
 }
 
 define void @store_i32_stride7_vf4(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vecptr2, ptr %in.vecptr3, ptr %in.vecptr4, ptr %in.vecptr5, ptr %in.vecptr6, ptr %out.vec) nounwind {
 ; SSE-LABEL: store_i32_stride7_vf4:
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
 ; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %r10
-; SSE-NEXT:    movaps (%rdi), %xmm0
-; SSE-NEXT:    movaps (%rsi), %xmm6
-; SSE-NEXT:    movaps (%rdx), %xmm5
+; SSE-NEXT:    movaps (%rdi), %xmm2
+; SSE-NEXT:    movaps (%rsi), %xmm0
+; SSE-NEXT:    movaps (%rdx), %xmm8
 ; SSE-NEXT:    movaps (%rcx), %xmm1
 ; SSE-NEXT:    movaps (%r8), %xmm4
-; SSE-NEXT:    movaps (%r9), %xmm2
-; SSE-NEXT:    movaps (%r10), %xmm8
-; SSE-NEXT:    movaps %xmm5, %xmm7
-; SSE-NEXT:    unpcklps {{.*#+}} xmm7 = xmm7[0],xmm1[0],xmm7[1],xmm1[1]
-; SSE-NEXT:    movaps %xmm0, %xmm3
-; SSE-NEXT:    unpcklps {{.*#+}} xmm3 = xmm3[0],xmm6[0],xmm3[1],xmm6[1]
-; SSE-NEXT:    movlhps {{.*#+}} xmm3 = xmm3[0],xmm7[0]
-; SSE-NEXT:    movaps %xmm4, %xmm9
-; SSE-NEXT:    unpckhps {{.*#+}} xmm9 = xmm9[2],xmm2[2],xmm9[3],xmm2[3]
-; SSE-NEXT:    movaps %xmm5, %xmm7
-; SSE-NEXT:    unpckhps {{.*#+}} xmm7 = xmm7[2],xmm1[2],xmm7[3],xmm1[3]
-; SSE-NEXT:    movlhps {{.*#+}} xmm7 = xmm7[0],xmm9[0]
+; SSE-NEXT:    movaps (%r9), %xmm3
+; SSE-NEXT:    movaps (%r10), %xmm6
+; SSE-NEXT:    movaps %xmm2, %xmm5
+; SSE-NEXT:    unpcklps {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm8, %xmm7
 ; SSE-NEXT:    movaps %xmm8, %xmm9
-; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[3,3],xmm2[3,3]
-; SSE-NEXT:    movaps %xmm4, %xmm10
-; SSE-NEXT:    shufps {{.*#+}} xmm10 = xmm10[1,1],xmm1[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[3,3],xmm0[3,3]
+; SSE-NEXT:    movaps %xmm2, %xmm10
+; SSE-NEXT:    unpckhps {{.*#+}} xmm10 = xmm10[2],xmm0[2],xmm10[3],xmm0[3]
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm8[0],xmm0[1],xmm8[1]
+; SSE-NEXT:    unpcklps {{.*#+}} xmm8 = xmm8[0],xmm1[0],xmm8[1],xmm1[1]
+; SSE-NEXT:    movlhps {{.*#+}} xmm5 = xmm5[0],xmm8[0]
+; SSE-NEXT:    movaps %xmm4, %xmm8
+; SSE-NEXT:    unpckhps {{.*#+}} xmm8 = xmm8[2],xmm3[2],xmm8[3],xmm3[3]
+; SSE-NEXT:    unpckhps {{.*#+}} xmm7 = xmm7[2],xmm1[2],xmm7[3],xmm1[3]
+; SSE-NEXT:    movlhps {{.*#+}} xmm7 = xmm7[0],xmm8[0]
+; SSE-NEXT:    movaps %xmm6, %xmm8
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[3,3],xmm3[3,3]
+; SSE-NEXT:    movaps %xmm4, %xmm11
+; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[1,1],xmm1[1,1]
 ; SSE-NEXT:    unpckhps {{.*#+}} xmm1 = xmm1[2],xmm4[2],xmm1[3],xmm4[3]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,3],xmm9[2,0]
-; SSE-NEXT:    movaps %xmm0, %xmm9
-; SSE-NEXT:    unpckhps {{.*#+}} xmm9 = xmm9[2],xmm6[2],xmm9[3],xmm6[3]
-; SSE-NEXT:    unpcklps {{.*#+}} xmm4 = xmm4[0],xmm2[0],xmm4[1],xmm2[1]
-; SSE-NEXT:    unpcklps {{.*#+}} xmm2 = xmm2[0],xmm8[0],xmm2[1],xmm8[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,3],xmm9[0,1]
-; SSE-NEXT:    movaps %xmm6, %xmm9
-; SSE-NEXT:    unpcklps {{.*#+}} xmm9 = xmm9[0],xmm5[0],xmm9[1],xmm5[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[2,3],xmm10[2,0]
-; SSE-NEXT:    pshufd {{.*#+}} xmm10 = xmm8[2,3,2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,1],xmm0[1,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm8[0,2]
-; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[3,3],xmm6[3,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,3],xmm5[2,0]
-; SSE-NEXT:    movss {{.*#+}} xmm0 = xmm10[0],xmm0[1,2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,3],xmm8[2,0]
+; SSE-NEXT:    movaps %xmm6, %xmm8
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[2,1],xmm2[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,2],xmm9[2,0]
+; SSE-NEXT:    unpcklps {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
+; SSE-NEXT:    unpcklps {{.*#+}} xmm3 = xmm3[0],xmm6[0],xmm3[1],xmm6[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[2,3],xmm10[0,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,3],xmm11[2,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,1],xmm2[1,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm6[0,2]
 ; SSE-NEXT:    movaps %xmm4, 16(%rax)
-; SSE-NEXT:    movaps %xmm9, 32(%rax)
-; SSE-NEXT:    movaps %xmm2, 48(%rax)
+; SSE-NEXT:    movaps %xmm0, 32(%rax)
+; SSE-NEXT:    movaps %xmm3, 48(%rax)
+; SSE-NEXT:    movaps %xmm8, 80(%rax)
 ; SSE-NEXT:    movaps %xmm1, 96(%rax)
 ; SSE-NEXT:    movaps %xmm7, 64(%rax)
-; SSE-NEXT:    movaps %xmm3, (%rax)
-; SSE-NEXT:    movaps %xmm0, 80(%rax)
+; SSE-NEXT:    movaps %xmm5, (%rax)
 ; SSE-NEXT:    retq
 ;
 ; AVX1-ONLY-LABEL: store_i32_stride7_vf4:
 ; AVX1-ONLY:       # %bb.0:
 ; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
 ; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %r10
 ; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm5
 ; AVX1-ONLY-NEXT:    vmovaps (%rsi), %xmm6
; AVX512-NEXT:    retq
   %interleaved.vec = shufflevector <28 x i32> %8, <28 x i32> poison, <28 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27>
   store <28 x i32> %interleaved.vec, ptr %out.vec, align 64
   ret void
 }
 
 define void @store_i32_stride7_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vecptr2, ptr %in.vecptr3, ptr %in.vecptr4, ptr %in.vecptr5, ptr %in.vecptr6, ptr %out.vec) nounwind {
 ; SSE-LABEL: store_i32_stride7_vf8:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    subq $104, %rsp
 ; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; SSE-NEXT:    movdqa (%rdi), %xmm8
-; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 16(%rdi), %xmm14
-; SSE-NEXT:    movdqa (%rsi), %xmm3
-; SSE-NEXT:    movdqa 16(%rsi), %xmm5
-; SSE-NEXT:    movdqa 16(%rdx), %xmm13
-; SSE-NEXT:    movdqa 16(%rcx), %xmm9
-; SSE-NEXT:    movdqa 16(%r8), %xmm11
-; SSE-NEXT:    movdqa %xmm11, (%rsp) # 16-byte Spill
-; SSE-NEXT:    movdqa (%r9), %xmm4
-; SSE-NEXT:    movaps 16(%r9), %xmm1
-; SSE-NEXT:    movdqa (%rax), %xmm7
-; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 16(%rax), %xmm12
-; SSE-NEXT:    movaps %xmm12, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm1[3,3]
-; SSE-NEXT:    movaps %xmm1, %xmm6
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm11[3,3,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm9[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm11, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm9[1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm13[1,1,1,1]
-; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[1,1,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm14, %xmm0
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm5[2],xmm0[3],xmm5[3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm12[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm6[1,1,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm0[0]
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[1,1,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movdqa %xmm8, %xmm0
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm3[2],xmm0[3],xmm3[3]
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm0[0]
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps (%rdx), %xmm11
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm3[1,1,1,1]
-; SSE-NEXT:    movaps %xmm11, %xmm1
-; SSE-NEXT:    movss {{.*#+}} xmm1 = xmm5[0],xmm1[1,2,3]
-; SSE-NEXT:    movaps (%rcx), %xmm3
-; SSE-NEXT:    movaps (%r8), %xmm7
-; SSE-NEXT:    movaps %xmm7, %xmm15
-; SSE-NEXT:    shufps {{.*#+}} xmm15 = xmm15[1,1],xmm3[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm15[2,0]
+; SSE-NEXT:    movaps (%rdi), %xmm0
+; SSE-NEXT:    movaps 16(%rdi), %xmm7
+; SSE-NEXT:    movaps (%rsi), %xmm6
+; SSE-NEXT:    movaps 16(%rsi), %xmm1
+; SSE-NEXT:    movaps (%rdx), %xmm4
+; SSE-NEXT:    movaps 16(%rdx), %xmm10
+; SSE-NEXT:    movaps 16(%rcx), %xmm3
+; SSE-NEXT:    movaps 16(%r8), %xmm8
+; SSE-NEXT:    movaps 16(%r9), %xmm5
+; SSE-NEXT:    movaps 16(%rax), %xmm11
+; SSE-NEXT:    movaps %xmm11, %xmm9
+; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[3,3],xmm5[3,3]
+; SSE-NEXT:    movaps %xmm3, %xmm2
+; SSE-NEXT:    unpckhps {{.*#+}} xmm2 = xmm2[2],xmm8[2],xmm2[3],xmm8[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,3],xmm9[2,0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm8, %xmm12
+; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[1,1],xmm3[1,1]
+; SSE-NEXT:    movaps %xmm1, %xmm2
 ; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm13, %xmm15
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm15 = xmm15[2],xmm9[2],xmm15[3],xmm9[3]
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm9 = xmm9[0],xmm13[0]
-; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm13 = xmm13[3,3],xmm10[3,3]
-; SSE-NEXT:    movdqa %xmm14, %xmm1
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,1],xmm12[0,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[2,1],xmm14[3,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[0,2],xmm13[2,0]
-; SSE-NEXT:    movaps (%rsp), %xmm8 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm8, %xmm13
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm13 = xmm13[0],xmm6[0],xmm13[1],xmm6[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm13 = xmm13[0,1],xmm1[2,0]
-; SSE-NEXT:    movaps %xmm7, %xmm1
-; SSE-NEXT:    movdqa %xmm4, %xmm9
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    unpckhps {{.*#+}} xmm1 = xmm1[2],xmm4[2],xmm1[3],xmm4[3]
-; SSE-NEXT:    movaps %xmm11, %xmm2
+; SSE-NEXT:    unpcklps {{.*#+}} xmm2 = xmm2[0],xmm10[0],xmm2[1],xmm10[1]
+; SSE-NEXT:    movaps %xmm10, %xmm13
+; SSE-NEXT:    unpckhps {{.*#+}} xmm13 = xmm13[2],xmm3[2],xmm13[3],xmm3[3]
+; SSE-NEXT:    movlhps {{.*#+}} xmm3 = xmm3[0],xmm10[0]
 ; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm10, %xmm14
+; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[3,3],xmm1[3,3]
+; SSE-NEXT:    movaps %xmm7, %xmm15
+; SSE-NEXT:    shufps {{.*#+}} xmm15 = xmm15[1,1],xmm11[0,3]
+; SSE-NEXT:    movaps %xmm8, %xmm10
+; SSE-NEXT:    unpcklps {{.*#+}} xmm10 = xmm10[0],xmm5[0],xmm10[1],xmm5[1]
+; SSE-NEXT:    unpckhps {{.*#+}} xmm8 = xmm8[2],xmm5[2],xmm8[3],xmm5[3]
+; SSE-NEXT:    unpcklps {{.*#+}} xmm5 = xmm5[0],xmm11[0],xmm5[1],xmm11[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[2,1],xmm7[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[0,2],xmm14[2,0]
+; SSE-NEXT:    movaps (%rax), %xmm14
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,3],xmm12[2,0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm10 = xmm10[0,1],xmm15[2,0]
+; SSE-NEXT:    movaps %xmm4, %xmm2
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm4, %xmm12
+; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[3,3],xmm6[3,3]
+; SSE-NEXT:    movaps %xmm14, %xmm15
+; SSE-NEXT:    movaps %xmm0, %xmm9
+; SSE-NEXT:    shufps {{.*#+}} xmm15 = xmm15[2,1],xmm0[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm15 = xmm15[0,2],xmm12[2,0]
+; SSE-NEXT:    movaps (%r8), %xmm1
+; SSE-NEXT:    movaps (%r9), %xmm0
+; SSE-NEXT:    movaps %xmm1, %xmm12
+; SSE-NEXT:    unpckhps {{.*#+}} xmm12 = xmm12[2],xmm0[2],xmm12[3],xmm0[3]
+; SSE-NEXT:    movaps (%rcx), %xmm3
 ; SSE-NEXT:    unpckhps {{.*#+}} xmm2 = xmm2[2],xmm3[2],xmm2[3],xmm3[3]
-; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm1[0]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm4, %xmm1
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm5[1,3]
-; SSE-NEXT:    movaps %xmm7, %xmm0
-; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[0,2]
-; SSE-NEXT:    movaps %xmm11, %xmm1
-; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1]
-; SSE-NEXT:    movaps %xmm5, %xmm3
-; SSE-NEXT:    movaps %xmm5, %xmm9
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm9 = xmm9[0],xmm5[0],xmm9[1],xmm5[1]
-; SSE-NEXT:    movlhps {{.*#+}} xmm9 = xmm9[0],xmm1[0]
-; SSE-NEXT:    unpckhps {{.*#+}} xmm8 = xmm8[2],xmm6[2],xmm8[3],xmm6[3]
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm15 = xmm15[0],xmm8[0]
-; SSE-NEXT:    unpcklps {{.*#+}} xmm14 = xmm14[0],xmm10[0],xmm14[1],xmm10[1]
-; SSE-NEXT:    shufps $36, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm14 = xmm14[0,1],mem[2,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[3,3],xmm5[3,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,3],xmm11[2,0]
-; SSE-NEXT:    movaps %xmm3, %xmm6
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[2,3,2,3]
-; SSE-NEXT:    movss {{.*#+}} xmm6 = xmm1[0],xmm6[1,2,3]
-; SSE-NEXT:    shufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm4 = xmm4[3,3],mem[3,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,3],xmm4[2,0]
-; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = mem[3,3,3,3]
-; SSE-NEXT:    movss {{.*#+}} xmm7 = xmm1[0],xmm7[1,2,3]
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm12[0]
+; SSE-NEXT:    movaps %xmm1, %xmm12
+; SSE-NEXT:    unpcklps {{.*#+}} xmm12 = xmm12[0],xmm0[0],xmm12[1],xmm0[1]
+; SSE-NEXT:    movaps %xmm14, %xmm4
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[3,3],xmm0[3,3]
+; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm14[0],xmm0[1],xmm14[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[0,1],xmm9[1,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[0,1],xmm14[0,2]
+; SSE-NEXT:    movaps %xmm9, %xmm14
+; SSE-NEXT:    unpcklps {{.*#+}} xmm14 = xmm14[0],xmm6[0],xmm14[1],xmm6[1]
+; SSE-NEXT:    unpckhps {{.*#+}} xmm9 = xmm9[2],xmm6[2],xmm9[3],xmm6[3]
+; SSE-NEXT:    movaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[1,1],xmm9[1,1]
+; SSE-NEXT:    unpcklps {{.*#+}} xmm9 = xmm9[0],xmm3[0],xmm9[1],xmm3[1]
+; SSE-NEXT:    movlhps {{.*#+}} xmm14 = xmm14[0],xmm9[0]
+; SSE-NEXT:    movlhps {{.*#+}} xmm13 = xmm13[0],xmm8[0]
+; SSE-NEXT:    movaps %xmm7, %xmm8
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; SSE-NEXT:    unpckhps {{.*#+}} xmm8 = xmm8[2],xmm9[2],xmm8[3],xmm9[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,3],xmm8[0,1]
+; SSE-NEXT:    unpcklps {{.*#+}} xmm7 = xmm7[0],xmm9[0],xmm7[1],xmm9[1]
+; SSE-NEXT:    shufps $36, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm7 = xmm7[0,1],mem[2,0]
+; SSE-NEXT:    movaps %xmm3, %xmm8
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[3,3],xmm1[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,2],xmm4[2,0]
+; SSE-NEXT:    shufpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
+; SSE-NEXT:    # xmm0 = xmm0[1],mem[0]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,1],xmm3[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,2],xmm1[2,0]
 ; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; SSE-NEXT:    movaps %xmm14, 112(%rax)
-; SSE-NEXT:    movdqa %xmm15, 176(%rax)
-; SSE-NEXT:    movaps %xmm9, (%rax)
-; SSE-NEXT:    movaps %xmm0, 16(%rax)
+; SSE-NEXT:    movaps %xmm6, 32(%rax)
+; SSE-NEXT:    movapd %xmm0, 48(%rax)
+; SSE-NEXT:    movaps %xmm8, 96(%rax)
+; SSE-NEXT:    movaps %xmm7, 112(%rax)
+; SSE-NEXT:    movaps %xmm5, 160(%rax)
+; SSE-NEXT:    movaps %xmm13, 176(%rax)
+; SSE-NEXT:    movaps %xmm14, (%rax)
+; SSE-NEXT:    movaps %xmm12, 16(%rax)
 ; SSE-NEXT:    movaps %xmm2, 64(%rax)
-; SSE-NEXT:    movaps %xmm13, 128(%rax)
-; SSE-NEXT:    movaps %xmm12, 192(%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 32(%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 48(%rax)
-; SSE-NEXT:    movaps %xmm7, 96(%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 160(%rax)
-; SSE-NEXT:    movaps %xmm6, 80(%rax)
+; SSE-NEXT:    movaps %xmm15, 80(%rax)
+; SSE-NEXT:    movaps %xmm10, 128(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 144(%rax)
+; SSE-NEXT:    movaps %xmm11, 192(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 208(%rax)
-; SSE-NEXT:    addq $104, %rsp
 ; SSE-NEXT:    retq
 ;
 ; AVX1-ONLY-LABEL: store_i32_stride7_vf8:
 ; AVX1-ONLY:       # %bb.0:
 ; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
 ; AVX1-ONLY-NEXT:    vmovaps (%rdi), %ymm0
 ; AVX1-ONLY-NEXT:    vmovaps (%rsi), %ymm1
 ; AVX1-ONLY-NEXT:    vmovaps (%rdx), %ymm2
; AVX512BW-NEXT:    retq
   %interleaved.vec = shufflevector <56 x i32> %8, <56 x i32> poison, <56 x i32> <i32 0, i32 8, i32 16, i32 24, i32 32, i32 40, i32 48, i32 1, i32 9, i32 17, i32 25, i32 33, i32 41, i32 49, i32 2, i32 10, i32 18, i32 26, i32 34, i32 42, i32 50, i32 3, i32 11, i32 19, i32 27, i32 35, i32 43, i32 51, i32 4, i32 12, i32 20, i32 28, i32 36, i32 44, i32 52, i32 5, i32 13, i32 21, i32 29, i32 37, i32 45, i32 53, i32 6, i32 14, i32 22, i32 30, i32 38, i32 46, i32 54, i32 7, i32 15, i32 23, i32 31, i32 39, i32 47, i32 55>
   store <56 x i32> %interleaved.vec, ptr %out.vec, align 64
   ret void
 }
 
 define void @store_i32_stride7_vf16(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vecptr2, ptr %in.vecptr3, ptr %in.vecptr4, ptr %in.vecptr5, ptr %in.vecptr6, ptr %out.vec) nounwind {
 ; SSE-LABEL: store_i32_stride7_vf16:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    subq $536, %rsp # imm = 0x218
+; SSE-NEXT:    subq $168, %rsp
 ; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; SSE-NEXT:    movdqa (%rdi), %xmm3
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa (%rsi), %xmm7
-; SSE-NEXT:    movdqa 16(%rsi), %xmm9
-; SSE-NEXT:    movaps (%rdx), %xmm5
-; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 16(%rdx), %xmm10
-; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps (%rcx), %xmm14
-; SSE-NEXT:    movaps 16(%rcx), %xmm2
-; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps (%r8), %xmm15
-; SSE-NEXT:    movaps 16(%r8), %xmm13
-; SSE-NEXT:    movaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa (%r9), %xmm11
-; SSE-NEXT:    movdqa 16(%r9), %xmm12
-; SSE-NEXT:    movdqa %xmm12, (%rsp) # 16-byte Spill
-; SSE-NEXT:    movdqa (%rax), %xmm4
-; SSE-NEXT:    movaps %xmm15, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm14[1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm7[1,1,1,1]
-; SSE-NEXT:    movss {{.*#+}} xmm5 = xmm1[0],xmm5[1,2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm3, %xmm0
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm7[2],xmm0[3],xmm7[3]
-; SSE-NEXT:    movdqa %xmm7, %xmm3
-; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[1,1,1,1]
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm11[1,1,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm1[0],xmm5[1],xmm1[1]
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm5 = xmm5[0],xmm0[0]
-; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm9[1,1,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movaps %xmm13, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm2[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 16(%rax), %xmm10
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm10[1,1,1,1]
-; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm12[1,1,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movdqa 16(%rdi), %xmm13
-; SSE-NEXT:    movdqa %xmm13, %xmm0
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm9[2],xmm0[3],xmm9[3]
-; SSE-NEXT:    movdqa %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm0[0]
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 32(%rsi), %xmm8
-; SSE-NEXT:    movaps 32(%rdx), %xmm5
-; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[1,1,1,1]
-; SSE-NEXT:    movaps %xmm5, %xmm1
-; SSE-NEXT:    movss {{.*#+}} xmm1 = xmm0[0],xmm1[1,2,3]
-; SSE-NEXT:    movaps 32(%rcx), %xmm5
-; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 32(%r8), %xmm0
-; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm5[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 32(%r9), %xmm5
-; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 32(%rax), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[1,1,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movdqa 32(%rdi), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm8[2],xmm0[3],xmm8[3]
-; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm0[0]
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 48(%rsi), %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 48(%rdx), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm1[1,1,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm0[0],xmm5[1],xmm0[1]
-; SSE-NEXT:    movaps 48(%rcx), %xmm2
-; SSE-NEXT:    movaps 48(%r8), %xmm1
-; SSE-NEXT:    movaps %xmm1, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm2[1,1]
-; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 48(%rdi), %xmm0
-; SSE-NEXT:    movaps 48(%rax), %xmm7
-; SSE-NEXT:    movaps %xmm0, %xmm5
-; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[1,1],xmm7[0,3]
-; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 48(%r9), %xmm6
-; SSE-NEXT:    movaps %xmm6, %xmm5
-; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[1,1],xmm7[1,1]
-; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,1],xmm7[2,3]
-; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[3,3],xmm6[3,3]
+; SSE-NEXT:    movaps (%rdi), %xmm0
+; SSE-NEXT:    movaps 16(%rdi), %xmm13
+; SSE-NEXT:    movaps (%rsi), %xmm1
+; SSE-NEXT:    movaps 16(%rsi), %xmm9
+; SSE-NEXT:    movaps (%rdx), %xmm2
+; SSE-NEXT:    movaps 16(%rdx), %xmm15
+; SSE-NEXT:    movaps (%rcx), %xmm10
+; SSE-NEXT:    movaps (%r8), %xmm6
+; SSE-NEXT:    movaps (%r9), %xmm3
+; SSE-NEXT:    movaps (%rax), %xmm4
+; SSE-NEXT:    movaps %xmm2, %xmm5
+; SSE-NEXT:    unpcklps {{.*#+}} xmm5 = xmm5[0],xmm10[0],xmm5[1],xmm10[1]
+; SSE-NEXT:    movaps %xmm0, %xmm7
+; SSE-NEXT:    unpcklps {{.*#+}} xmm7 = xmm7[0],xmm1[0],xmm7[1],xmm1[1]
+; SSE-NEXT:    movlhps {{.*#+}} xmm7 = xmm7[0],xmm5[0]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm4, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,1],xmm0[1,3]
+; SSE-NEXT:    movaps %xmm6, %xmm7
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[1,1],xmm10[1,1]
+; SSE-NEXT:    movaps %xmm6, %xmm8
+; SSE-NEXT:    movaps %xmm2, %xmm11
+; SSE-NEXT:    unpckhps {{.*#+}} xmm11 = xmm11[2],xmm10[2],xmm11[3],xmm10[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm10 = xmm10[3,3],xmm6[3,3]
+; SSE-NEXT:    unpcklps {{.*#+}} xmm6 = xmm6[0],xmm3[0],xmm6[1],xmm3[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,1],xmm5[0,2]
+; SSE-NEXT:    movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movaps %xmm1, %xmm5
-; SSE-NEXT:    unpcklps {{.*#+}} xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[1,1],xmm2[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,2],xmm7[2,0]
 ; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    unpckhpd {{.*#+}} xmm6 = xmm6[1],xmm1[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[3,3,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm5 = xmm2[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm5 = xmm5[0],xmm1[0],xmm5[1],xmm1[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,1],xmm7[2,0]
+; SSE-NEXT:    movaps %xmm4, %xmm6
+; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[2,1],xmm0[3,3]
+; SSE-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; SSE-NEXT:    movaps %xmm3, %xmm5
+; SSE-NEXT:    unpcklps {{.*#+}} xmm5 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,3],xmm0[0,1]
 ; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm5, %xmm0
-; SSE-NEXT:    movaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm14[0],xmm0[1],xmm14[1]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm2, %xmm7
-; SSE-NEXT:    punpckldq {{.*#+}} xmm7 = xmm7[0],xmm3[0],xmm7[1],xmm3[1]
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm7 = xmm7[0],xmm0[0]
-; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm2[1,3]
-; SSE-NEXT:    movaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm15, %xmm2
-; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    unpcklps {{.*#+}} xmm2 = xmm2[0],xmm11[0],xmm2[1],xmm11[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm4[0,2]
-; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    unpckhps {{.*#+}} xmm15 = xmm15[2],xmm11[2],xmm15[3],xmm11[3]
-; SSE-NEXT:    movaps %xmm5, %xmm2
-; SSE-NEXT:    movaps %xmm5, %xmm7
-; SSE-NEXT:    unpckhps {{.*#+}} xmm2 = xmm2[2],xmm14[2],xmm2[3],xmm14[3]
+; SSE-NEXT:    movaps 16(%rcx), %xmm7
+; SSE-NEXT:    unpckhps {{.*#+}} xmm8 = xmm8[2],xmm3[2],xmm8[3],xmm3[3]
+; SSE-NEXT:    movlhps {{.*#+}} xmm11 = xmm11[0],xmm8[0]
+; SSE-NEXT:    movaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 16(%r8), %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[3,3],xmm1[3,3]
+; SSE-NEXT:    movaps 16(%r9), %xmm0
+; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,2],xmm2[2,0]
+; SSE-NEXT:    movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 16(%rax), %xmm1
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[3,3],xmm3[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm10 = xmm10[0,2],xmm4[2,0]
+; SSE-NEXT:    movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm7, %xmm2
 ; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm15[0]
+; SSE-NEXT:    movaps %xmm13, %xmm4
+; SSE-NEXT:    unpcklps {{.*#+}} xmm4 = xmm4[0],xmm9[0],xmm4[1],xmm9[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm2[2,0]
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm13, %xmm2
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,1],xmm1[0,3]
+; SSE-NEXT:    movaps %xmm5, %xmm4
+; SSE-NEXT:    unpcklps {{.*#+}} xmm4 = xmm4[0],xmm0[0],xmm4[1],xmm0[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm2[2,0]
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm5, %xmm2
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,1],xmm7[1,1]
+; SSE-NEXT:    movaps %xmm9, %xmm4
+; SSE-NEXT:    unpcklps {{.*#+}} xmm4 = xmm4[0],xmm15[0],xmm4[1],xmm15[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[2,3],xmm2[2,0]
+; SSE-NEXT:    movaps %xmm4, (%rsp) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,1],xmm13[3,3]
+; SSE-NEXT:    movaps %xmm2, %xmm4
+; SSE-NEXT:    unpckhps {{.*#+}} xmm13 = xmm13[2],xmm9[2],xmm13[3],xmm9[3]
+; SSE-NEXT:    movaps %xmm0, %xmm2
+; SSE-NEXT:    unpcklps {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[2,3],xmm13[0,1]
 ; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm2, %xmm0
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    movlhps {{.*#+}} xmm0 = xmm0[0],xmm4[0]
-; SSE-NEXT:    movdqa %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm13, %xmm15
-; SSE-NEXT:    punpckldq {{.*#+}} xmm15 = xmm15[0],xmm9[0],xmm15[1],xmm9[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm15 = xmm15[0,1],xmm0[2,0]
-; SSE-NEXT:    movdqa %xmm13, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm10[0,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm14, %xmm12
-; SSE-NEXT:    movaps (%rsp), %xmm3 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm12 = xmm12[0],xmm3[0],xmm12[1],xmm3[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps %xmm4, %xmm10
-; SSE-NEXT:    unpckhps {{.*#+}} xmm10 = xmm10[2],xmm2[2],xmm10[3],xmm2[3]
-; SSE-NEXT:    movaps %xmm14, %xmm0
-; SSE-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2],xmm3[2],xmm0[3],xmm3[3]
-; SSE-NEXT:    movlhps {{.*#+}} xmm10 = xmm10[0],xmm0[0]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm5, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm13, %xmm11
-; SSE-NEXT:    punpckldq {{.*#+}} xmm11 = xmm11[0],xmm8[0],xmm11[1],xmm8[1]
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm11 = xmm11[0],xmm0[0]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm13[1,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm9, %xmm8
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm8 = xmm8[0],xmm4[0],xmm8[1],xmm4[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,1],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm9, %xmm0
-; SSE-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm5 = xmm5[2],xmm2[2],xmm5[3],xmm2[3]
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm5 = xmm5[0],xmm0[0]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, %xmm4
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    unpckhps {{.*#+}} xmm4 = xmm4[2],xmm3[2],xmm4[3],xmm3[3]
-; SSE-NEXT:    movlhps {{.*#+}} xmm3 = xmm3[0],xmm1[0]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm3[2,0]
-; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    shufps $36, {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm3 = xmm3[0,1],mem[2,0]
+; SSE-NEXT:    movaps %xmm15, %xmm2
+; SSE-NEXT:    unpckhps {{.*#+}} xmm2 = xmm2[2],xmm7[2],xmm2[3],xmm7[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[3,3],xmm5[3,3]
+; SSE-NEXT:    unpckhps {{.*#+}} xmm5 = xmm5[2],xmm0[2],xmm5[3],xmm0[3]
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm5[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm15 = xmm15[3,3],xmm9[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,2],xmm15[2,0]
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,3],xmm0[3,3]
+; SSE-NEXT:    movaps 32(%rdx), %xmm0
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,2],xmm1[2,0]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 32(%rcx), %xmm15
+; SSE-NEXT:    movaps %xmm0, %xmm2
+; SSE-NEXT:    unpcklps {{.*#+}} xmm2 = xmm2[0],xmm15[0],xmm2[1],xmm15[1]
+; SSE-NEXT:    movaps 32(%rdi), %xmm4
+; SSE-NEXT:    movaps 32(%rsi), %xmm1
+; SSE-NEXT:    movaps %xmm4, %xmm3
+; SSE-NEXT:    unpcklps {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; SSE-NEXT:    movlhps {{.*#+}} xmm3 = xmm3[0],xmm2[0]
 ; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,2],xmm0[0,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm6[2,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,3],xmm2[3,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[2,0]
-; SSE-NEXT:    shufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm7 = xmm7[3,3],mem[3,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm7[2,0]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm7[2,3,2,3]
-; SSE-NEXT:    movss {{.*#+}} xmm1 = xmm6[0],xmm1[1,2,3]
-; SSE-NEXT:    movaps %xmm1, %xmm2
-; SSE-NEXT:    shufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm7 = xmm7[3,3],mem[3,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm7[2,0]
-; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm6 = mem[3,3,3,3]
-; SSE-NEXT:    movss {{.*#+}} xmm1 = xmm6[0],xmm1[1,2,3]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT:    shufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm6 = xmm6[3,3],mem[3,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,3],xmm6[2,0]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm1[2,3,2,3]
-; SSE-NEXT:    movss {{.*#+}} xmm7 = xmm6[0],xmm7[1,2,3]
-; SSE-NEXT:    shufps $255, (%rsp), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = xmm1[3,3],mem[3,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[0,3],xmm1[2,0]
-; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm6 = mem[3,3,3,3]
-; SSE-NEXT:    movss {{.*#+}} xmm14 = xmm6[0],xmm14[1,2,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    shufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = xmm1[3,3],mem[3,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm13 = xmm13[0,3],xmm1[2,0]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm6 = xmm1[2,3,2,3]
-; SSE-NEXT:    movss {{.*#+}} xmm13 = xmm6[0],xmm13[1,2,3]
-; SSE-NEXT:    shufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = xmm1[3,3],mem[3,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[0,3],xmm1[2,0]
-; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm6 = mem[3,3,3,3]
-; SSE-NEXT:    movss {{.*#+}} xmm9 = xmm6[0],xmm9[1,2,3]
+; SSE-NEXT:    movaps 32(%rax), %xmm2
+; SSE-NEXT:    movaps %xmm2, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,1],xmm4[1,3]
+; SSE-NEXT:    movaps 32(%r8), %xmm6
+; SSE-NEXT:    movaps 32(%r9), %xmm3
+; SSE-NEXT:    movaps %xmm6, %xmm7
+; SSE-NEXT:    unpcklps {{.*#+}} xmm7 = xmm7[0],xmm3[0],xmm7[1],xmm3[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,1],xmm5[0,2]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm6, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[1,1],xmm15[1,1]
+; SSE-NEXT:    movaps %xmm1, %xmm7
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[1,1],xmm0[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,2],xmm5[2,0]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm2, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,1],xmm4[3,3]
+; SSE-NEXT:    unpckhps {{.*#+}} xmm4 = xmm4[2],xmm1[2],xmm4[3],xmm1[3]
+; SSE-NEXT:    movaps %xmm3, %xmm14
+; SSE-NEXT:    unpcklps {{.*#+}} xmm14 = xmm14[0],xmm2[0],xmm14[1],xmm2[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[2,3],xmm4[0,1]
+; SSE-NEXT:    movaps %xmm0, %xmm13
+; SSE-NEXT:    unpckhps {{.*#+}} xmm13 = xmm13[2],xmm15[2],xmm13[3],xmm15[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm15 = xmm15[3,3],xmm6[3,3]
+; SSE-NEXT:    unpckhps {{.*#+}} xmm6 = xmm6[2],xmm3[2],xmm6[3],xmm3[3]
+; SSE-NEXT:    movlhps {{.*#+}} xmm13 = xmm13[0],xmm6[0]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm1[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,2],xmm0[2,0]
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[3,3],xmm3[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm15 = xmm15[0,2],xmm2[2,0]
+; SSE-NEXT:    movaps 48(%rdx), %xmm4
+; SSE-NEXT:    movaps 48(%rcx), %xmm0
+; SSE-NEXT:    movaps %xmm0, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm4[0]
+; SSE-NEXT:    movaps 48(%rdi), %xmm1
+; SSE-NEXT:    movaps 48(%rsi), %xmm12
+; SSE-NEXT:    movaps %xmm1, %xmm7
+; SSE-NEXT:    unpcklps {{.*#+}} xmm7 = xmm7[0],xmm12[0],xmm7[1],xmm12[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,1],xmm2[2,0]
+; SSE-NEXT:    movaps 48(%rax), %xmm8
+; SSE-NEXT:    movaps %xmm1, %xmm11
+; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[1,1],xmm8[0,3]
+; SSE-NEXT:    movaps 48(%r8), %xmm9
+; SSE-NEXT:    movaps %xmm9, %xmm10
+; SSE-NEXT:    shufps {{.*#+}} xmm10 = xmm10[1,1],xmm0[1,1]
+; SSE-NEXT:    movaps 48(%r9), %xmm5
+; SSE-NEXT:    movaps %xmm5, %xmm6
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm6 = xmm6[1],xmm9[1]
+; SSE-NEXT:    movaps %xmm4, %xmm3
+; SSE-NEXT:    unpckhps {{.*#+}} xmm3 = xmm3[2],xmm0[2],xmm3[3],xmm0[3]
+; SSE-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2],xmm9[2],xmm0[3],xmm9[3]
+; SSE-NEXT:    unpcklps {{.*#+}} xmm9 = xmm9[0],xmm5[0],xmm9[1],xmm5[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[0,1],xmm11[2,0]
+; SSE-NEXT:    movaps %xmm12, %xmm11
+; SSE-NEXT:    unpcklps {{.*#+}} xmm11 = xmm11[0],xmm4[0],xmm11[1],xmm4[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[2,3],xmm10[2,0]
+; SSE-NEXT:    movaps %xmm1, %xmm10
+; SSE-NEXT:    unpckhps {{.*#+}} xmm10 = xmm10[2],xmm12[2],xmm10[3],xmm12[3]
+; SSE-NEXT:    movaps %xmm5, %xmm2
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,1],xmm8[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,2],xmm10[0,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm6[2,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[3,3],xmm12[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,1],xmm8[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm4[2,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[3,3],xmm8[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,3],xmm5[0,2]
 ; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; SSE-NEXT:    movaps %xmm0, 416(%rax)
-; SSE-NEXT:    movaps %xmm4, 400(%rax)
-; SSE-NEXT:    movaps %xmm3, 384(%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 352(%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 336(%rax)
-; SSE-NEXT:    movdqa %xmm5, 288(%rax)
-; SSE-NEXT:    movaps %xmm8, 240(%rax)
-; SSE-NEXT:    movdqa %xmm11, 224(%rax)
-; SSE-NEXT:    movaps %xmm10, 176(%rax)
-; SSE-NEXT:    movaps %xmm12, 128(%rax)
-; SSE-NEXT:    movaps %xmm15, 112(%rax)
+; SSE-NEXT:    movaps %xmm0, 432(%rax)
+; SSE-NEXT:    movaps %xmm1, 416(%rax)
+; SSE-NEXT:    movaps %xmm3, 400(%rax)
+; SSE-NEXT:    movaps %xmm2, 384(%rax)
+; SSE-NEXT:    movaps %xmm11, 368(%rax)
+; SSE-NEXT:    movaps %xmm9, 352(%rax)
+; SSE-NEXT:    movaps %xmm7, 336(%rax)
+; SSE-NEXT:    movaps %xmm15, 320(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 64(%rax)
+; SSE-NEXT:    movaps %xmm0, 304(%rax)
+; SSE-NEXT:    movaps %xmm13, 288(%rax)
+; SSE-NEXT:    movaps %xmm14, 272(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 16(%rax)
+; SSE-NEXT:    movaps %xmm0, 256(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, (%rax)
+; SSE-NEXT:    movaps %xmm0, 240(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 432(%rax)
+; SSE-NEXT:    movaps %xmm0, 224(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 368(%rax)
-; SSE-NEXT:    movaps %xmm9, 320(%rax)
-; SSE-NEXT:    movaps %xmm13, 304(%rax)
+; SSE-NEXT:    movaps %xmm0, 208(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 272(%rax)
+; SSE-NEXT:    movaps %xmm0, 192(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 256(%rax)
-; SSE-NEXT:    movaps %xmm14, 208(%rax)
-; SSE-NEXT:    movaps %xmm7, 192(%rax)
+; SSE-NEXT:    movaps %xmm0, 176(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 160(%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps (%rsp), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 144(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 128(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 112(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 96(%rax)
-; SSE-NEXT:    movaps %xmm2, 80(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 80(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 48(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 32(%rax)
-; SSE-NEXT:    addq $536, %rsp # imm = 0x218
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%rax)
+; SSE-NEXT:    addq $168, %rsp
 ; SSE-NEXT:    retq
 ;
 ; AVX1-ONLY-LABEL: store_i32_stride7_vf16:
 ; AVX1-ONLY:       # %bb.0:
 ; AVX1-ONLY-NEXT:    subq $488, %rsp # imm = 0x1E8
 ; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %ymm5
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rsi), %ymm6
; AVX512BW-NEXT:    retq
   %interleaved.vec = shufflevector <112 x i32> %8, <112 x i32> poison, <112 x i32> <i32 0, i32 16, i32 32, i32 48, i32 64, i32 80, i32 96, i32 1, i32 17, i32 33, i32 49, i32 65, i32 81, i32 97, i32 2, i32 18, i32 34, i32 50, i32 66, i32 82, i32 98, i32 3, i32 19, i32 35, i32 51, i32 67, i32 83, i32 99, i32 4, i32 20, i32 36, i32 52, i32 68, i32 84, i32 100, i32 5, i32 21, i32 37, i32 53, i32 69, i32 85, i32 101, i32 6, i32 22, i32 38, i32 54, i32 70, i32 86, i32 102, i32 7, i32 23, i32 39, i32 55, i32 71, i32 87, i32 103, i32 8, i32 24, i32 40, i32 56, i32 72, i32 88, i32 104, i32 9, i32 25, i32 41, i32 57, i32 73, i32 89, i32 105, i32 10, i32 26, i32 42, i32 58, i32 74, i32 90, i32 106, i32 11, i32 27, i32 43, i32 59, i32 75, i32 91, i32 107, i32 12, i32 28, i32 44, i32 60, i32 76, i32 92, i32 108, i32 13, i32 29, i32 45, i32 61, i32 77, i32 93, i32 109, i32 14, i32 30, i32 46, i32 62, i32 78, i32 94, i32 110, i32 15, i32 31, i32 47, i32 63, i32 79, i32 95, i32 111>
   store <112 x i32> %interleaved.vec, ptr %out.vec, align 64
   ret void
 }
 
 define void @store_i32_stride7_vf32(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vecptr2, ptr %in.vecptr3, ptr %in.vecptr4, ptr %in.vecptr5, ptr %in.vecptr6, ptr %out.vec) nounwind {
 ; SSE-LABEL: store_i32_stride7_vf32:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    subq $1256, %rsp # imm = 0x4E8
+; SSE-NEXT:    subq $616, %rsp # imm = 0x268
 ; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; SSE-NEXT:    movdqa (%rdi), %xmm4
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa (%rsi), %xmm7
-; SSE-NEXT:    movdqa 16(%rsi), %xmm5
-; SSE-NEXT:    movaps (%rdx), %xmm9
-; SSE-NEXT:    movdqa 16(%rdx), %xmm12
-; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps (%rcx), %xmm10
-; SSE-NEXT:    movaps 16(%rcx), %xmm13
+; SSE-NEXT:    movaps (%rdi), %xmm3
+; SSE-NEXT:    movaps 16(%rdi), %xmm2
+; SSE-NEXT:    movaps (%rsi), %xmm4
+; SSE-NEXT:    movaps 16(%rsi), %xmm0
+; SSE-NEXT:    movaps (%rdx), %xmm5
+; SSE-NEXT:    movaps 16(%rdx), %xmm1
+; SSE-NEXT:    movaps (%rcx), %xmm12
+; SSE-NEXT:    movaps (%r8), %xmm7
+; SSE-NEXT:    movaps (%r9), %xmm6
+; SSE-NEXT:    movaps (%rax), %xmm8
+; SSE-NEXT:    movaps %xmm5, %xmm9
+; SSE-NEXT:    unpcklps {{.*#+}} xmm9 = xmm9[0],xmm12[0],xmm9[1],xmm12[1]
+; SSE-NEXT:    movaps %xmm3, %xmm10
+; SSE-NEXT:    unpcklps {{.*#+}} xmm10 = xmm10[0],xmm4[0],xmm10[1],xmm4[1]
+; SSE-NEXT:    movlhps {{.*#+}} xmm10 = xmm10[0],xmm9[0]
+; SSE-NEXT:    movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm8, %xmm9
+; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[0,1],xmm3[1,3]
+; SSE-NEXT:    movaps %xmm7, %xmm10
+; SSE-NEXT:    shufps {{.*#+}} xmm10 = xmm10[1,1],xmm12[1,1]
+; SSE-NEXT:    movaps %xmm7, %xmm11
+; SSE-NEXT:    movaps %xmm5, %xmm13
+; SSE-NEXT:    unpckhps {{.*#+}} xmm13 = xmm13[2],xmm12[2],xmm13[3],xmm12[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[3,3],xmm7[3,3]
+; SSE-NEXT:    unpcklps {{.*#+}} xmm7 = xmm7[0],xmm6[0],xmm7[1],xmm6[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,1],xmm9[0,2]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm4, %xmm7
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[1,1],xmm5[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,2],xmm10[2,0]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm8, %xmm9
+; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[2,1],xmm3[3,3]
+; SSE-NEXT:    unpckhps {{.*#+}} xmm3 = xmm3[2],xmm4[2],xmm3[3],xmm4[3]
+; SSE-NEXT:    movaps %xmm6, %xmm7
+; SSE-NEXT:    unpcklps {{.*#+}} xmm7 = xmm7[0],xmm8[0],xmm7[1],xmm8[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[2,3],xmm3[0,1]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 16(%rcx), %xmm10
+; SSE-NEXT:    unpckhps {{.*#+}} xmm11 = xmm11[2],xmm6[2],xmm11[3],xmm6[3]
+; SSE-NEXT:    movlhps {{.*#+}} xmm13 = xmm13[0],xmm11[0]
 ; SSE-NEXT:    movaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps (%r8), %xmm0
-; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 16(%r8), %xmm14
-; SSE-NEXT:    movaps %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa (%r9), %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 16(%r9), %xmm15
-; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa (%rax), %xmm11
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm10[1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm7[1,1,1,1]
-; SSE-NEXT:    movaps %xmm9, %xmm3
-; SSE-NEXT:    movss {{.*#+}} xmm3 = xmm1[0],xmm3[1,2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm4, %xmm0
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm7[2],xmm0[3],xmm7[3]
-; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm11[1,1,1,1]
-; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[1,1,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm3 = xmm3[0],xmm0[0]
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[1,1,1,1]
-; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[1,1,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movaps %xmm14, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm13[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 16(%rax), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm15[1,1,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movdqa 16(%rdi), %xmm14
-; SSE-NEXT:    movdqa %xmm14, %xmm0
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm5[2],xmm0[3],xmm5[3]
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm0[0]
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 32(%rsi), %xmm1
-; SSE-NEXT:    movaps 32(%rdx), %xmm2
-; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[1,1,1,1]
-; SSE-NEXT:    movdqa %xmm1, %xmm8
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm2, %xmm1
-; SSE-NEXT:    movss {{.*#+}} xmm1 = xmm0[0],xmm1[1,2,3]
-; SSE-NEXT:    movaps 32(%rcx), %xmm2
-; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 32(%r8), %xmm0
-; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm2[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 32(%r9), %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 32(%rax), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movdqa 32(%rdi), %xmm6
-; SSE-NEXT:    movdqa %xmm6, %xmm0
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm8[2],xmm0[3],xmm8[3]
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm0[0]
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 48(%rsi), %xmm8
-; SSE-NEXT:    movdqa 48(%rdx), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm8[1,1,1,1]
-; SSE-NEXT:    movdqa %xmm8, %xmm12
-; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movaps 48(%rcx), %xmm13
-; SSE-NEXT:    movaps 48(%r8), %xmm0
-; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm13[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 48(%r9), %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 48(%rax), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movdqa 48(%rdi), %xmm8
-; SSE-NEXT:    movdqa %xmm8, %xmm0
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm12[2],xmm0[3],xmm12[3]
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm0[0]
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 64(%rsi), %xmm1
-; SSE-NEXT:    movaps 64(%rdx), %xmm2
-; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[1,1,1,1]
-; SSE-NEXT:    movdqa %xmm1, %xmm12
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm2, %xmm1
-; SSE-NEXT:    movss {{.*#+}} xmm1 = xmm0[0],xmm1[1,2,3]
-; SSE-NEXT:    movaps 64(%rcx), %xmm2
-; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 64(%r8), %xmm0
-; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm2[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 64(%r9), %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 64(%rax), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movdqa 64(%rdi), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm12[2],xmm0[3],xmm12[3]
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm0[0]
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 80(%rsi), %xmm12
-; SSE-NEXT:    movdqa 80(%rdx), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm12[1,1,1,1]
-; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movaps 80(%rcx), %xmm2
-; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 80(%r8), %xmm0
-; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm2[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 80(%r9), %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 80(%rax), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movdqa 80(%rdi), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm12[2],xmm0[3],xmm12[3]
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm0[0]
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 96(%rsi), %xmm4
-; SSE-NEXT:    movaps 96(%rdx), %xmm12
+; SSE-NEXT:    movaps 16(%r8), %xmm7
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[3,3],xmm4[3,3]
+; SSE-NEXT:    movaps 16(%r9), %xmm4
+; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[0,2],xmm5[2,0]
+; SSE-NEXT:    movaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 16(%rax), %xmm3
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[3,3],xmm6[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[0,2],xmm8[2,0]
 ; SSE-NEXT:    movaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[1,1,1,1]
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm12, %xmm1
-; SSE-NEXT:    movss {{.*#+}} xmm1 = xmm0[0],xmm1[1,2,3]
-; SSE-NEXT:    movaps 96(%rcx), %xmm3
-; SSE-NEXT:    movaps 96(%r8), %xmm15
-; SSE-NEXT:    movaps %xmm15, %xmm0
-; SSE-NEXT:    movaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm3[1,1]
-; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 96(%r9), %xmm2
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 96(%rax), %xmm12
-; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm12[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[1,1,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movdqa 96(%rdi), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm0[0]
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm15[3,3,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movdqa %xmm12, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm2[3,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 112(%rsi), %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 112(%rdx), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[1,1,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
-; SSE-NEXT:    movaps 112(%rcx), %xmm3
-; SSE-NEXT:    movaps 112(%r8), %xmm0
-; SSE-NEXT:    movaps %xmm0, %xmm1
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,1],xmm3[1,1]
-; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,0]
-; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm0, %xmm1
-; SSE-NEXT:    movaps 112(%r9), %xmm15
-; SSE-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm15[0]
-; SSE-NEXT:    movaps %xmm1, (%rsp) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm15, %xmm1
-; SSE-NEXT:    movaps %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    unpckhpd {{.*#+}} xmm15 = xmm15[1],xmm0[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
-; SSE-NEXT:    movaps 112(%rax), %xmm12
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,3],xmm12[3,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[0,2]
+; SSE-NEXT:    movaps %xmm10, %xmm5
+; SSE-NEXT:    movlhps {{.*#+}} xmm5 = xmm5[0],xmm1[0]
+; SSE-NEXT:    movaps %xmm2, %xmm8
+; SSE-NEXT:    unpcklps {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,1],xmm5[2,0]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm2, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[1,1],xmm3[0,3]
+; SSE-NEXT:    movaps %xmm7, %xmm8
+; SSE-NEXT:    unpcklps {{.*#+}} xmm8 = xmm8[0],xmm4[0],xmm8[1],xmm4[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,1],xmm5[2,0]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm7, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[1,1],xmm10[1,1]
+; SSE-NEXT:    movaps %xmm0, %xmm8
+; SSE-NEXT:    unpcklps {{.*#+}} xmm8 = xmm8[0],xmm1[0],xmm8[1],xmm1[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[2,3],xmm5[2,0]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm3, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,1],xmm2[3,3]
+; SSE-NEXT:    movaps %xmm5, %xmm8
+; SSE-NEXT:    unpckhps {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE-NEXT:    movaps %xmm4, %xmm5
+; SSE-NEXT:    unpcklps {{.*#+}} xmm5 = xmm5[0],xmm3[0],xmm5[1],xmm3[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,3],xmm2[0,1]
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm1, %xmm2
+; SSE-NEXT:    unpckhps {{.*#+}} xmm2 = xmm2[2],xmm10[2],xmm2[3],xmm10[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm10 = xmm10[3,3],xmm7[3,3]
+; SSE-NEXT:    unpckhps {{.*#+}} xmm7 = xmm7[2],xmm4[2],xmm7[3],xmm4[3]
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm7[0]
 ; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm9, %xmm0
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,3],xmm0[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,2],xmm1[2,0]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[3,3],xmm4[3,3]
+; SSE-NEXT:    movaps 32(%rdx), %xmm0
+; SSE-NEXT:    shufps {{.*#+}} xmm10 = xmm10[0,2],xmm3[2,0]
 ; SSE-NEXT:    movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm10[0],xmm0[1],xmm10[1]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm2, %xmm1
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm7[0],xmm1[1],xmm7[1]
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm0[0]
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[0,1],xmm2[1,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm3, %xmm1
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm11[0,2]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm3, %xmm0
-; SSE-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
-; SSE-NEXT:    movaps %xmm9, %xmm1
-; SSE-NEXT:    unpckhps {{.*#+}} xmm1 = xmm1[2],xmm10[2],xmm1[3],xmm10[3]
-; SSE-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, %xmm0
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT:    movlhps {{.*#+}} xmm0 = xmm0[0],xmm5[0]
-; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm14, %xmm2
-; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm2 = xmm2[0],mem[0],xmm2[1],mem[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps $197, {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm14 = xmm14[1,1],mem[0,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm2, %xmm3
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm14[2,0]
-; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm5, %xmm3
-; SSE-NEXT:    unpckhps {{.*#+}} xmm3 = xmm3[2],xmm1[2],xmm3[3],xmm1[3]
-; SSE-NEXT:    movaps %xmm2, %xmm0
-; SSE-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
-; SSE-NEXT:    movlhps {{.*#+}} xmm3 = xmm3[0],xmm0[0]
+; SSE-NEXT:    movaps 32(%rcx), %xmm7
+; SSE-NEXT:    movaps %xmm0, %xmm2
+; SSE-NEXT:    unpcklps {{.*#+}} xmm2 = xmm2[0],xmm7[0],xmm2[1],xmm7[1]
+; SSE-NEXT:    movaps 32(%rdi), %xmm4
+; SSE-NEXT:    movaps 32(%rsi), %xmm1
+; SSE-NEXT:    movaps %xmm4, %xmm3
+; SSE-NEXT:    unpcklps {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; SSE-NEXT:    movlhps {{.*#+}} xmm3 = xmm3[0],xmm2[0]
 ; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm2, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm6, %xmm3
-; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm3 = xmm3[0],mem[0],xmm3[1],mem[1]
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm3 = xmm3[0],xmm0[0]
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm6[1,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm5, %xmm4
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm0[0,2]
+; SSE-NEXT:    movaps 32(%rax), %xmm2
+; SSE-NEXT:    movaps %xmm2, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,1],xmm4[1,3]
+; SSE-NEXT:    movaps 32(%r8), %xmm6
+; SSE-NEXT:    movaps 32(%r9), %xmm3
+; SSE-NEXT:    movaps %xmm6, %xmm8
+; SSE-NEXT:    unpcklps {{.*#+}} xmm8 = xmm8[0],xmm3[0],xmm8[1],xmm3[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,1],xmm5[0,2]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm6, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[1,1],xmm7[1,1]
+; SSE-NEXT:    movaps %xmm1, %xmm8
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[1,1],xmm0[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,2],xmm5[2,0]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm2, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,1],xmm4[3,3]
+; SSE-NEXT:    unpckhps {{.*#+}} xmm4 = xmm4[2],xmm1[2],xmm4[3],xmm1[3]
+; SSE-NEXT:    movaps %xmm3, %xmm8
+; SSE-NEXT:    unpcklps {{.*#+}} xmm8 = xmm8[0],xmm2[0],xmm8[1],xmm2[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[2,3],xmm4[0,1]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm0, %xmm4
+; SSE-NEXT:    unpckhps {{.*#+}} xmm4 = xmm4[2],xmm7[2],xmm4[3],xmm7[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[3,3],xmm6[3,3]
+; SSE-NEXT:    unpckhps {{.*#+}} xmm6 = xmm6[2],xmm3[2],xmm6[3],xmm3[3]
+; SSE-NEXT:    movlhps {{.*#+}} xmm4 = xmm4[0],xmm6[0]
 ; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm5, %xmm0
-; SSE-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2],xmm3[2],xmm0[3],xmm3[3]
-; SSE-NEXT:    movdqa %xmm2, %xmm3
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm1[2],xmm3[3],xmm1[3]
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm3 = xmm3[0],xmm0[0]
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm13, %xmm0
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
-; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm8, %xmm3
-; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm3 = xmm3[0],mem[0],xmm3[1],mem[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm0[2,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm1[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,2],xmm0[2,0]
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[3,3],xmm3[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,2],xmm2[2,0]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 48(%rdx), %xmm0
+; SSE-NEXT:    movaps 48(%rcx), %xmm7
+; SSE-NEXT:    movaps %xmm7, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps 48(%rdi), %xmm4
+; SSE-NEXT:    movaps 48(%rsi), %xmm1
+; SSE-NEXT:    movaps %xmm4, %xmm3
+; SSE-NEXT:    unpcklps {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm2[2,0]
 ; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps $197, {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm8 = xmm8[1,1],mem[0,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm5, %xmm4
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm4 = xmm4[0],xmm2[0],xmm4[1],xmm2[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm8[2,0]
+; SSE-NEXT:    movaps 48(%rax), %xmm2
+; SSE-NEXT:    movaps %xmm4, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[1,1],xmm2[0,3]
+; SSE-NEXT:    movaps 48(%r8), %xmm6
+; SSE-NEXT:    movaps 48(%r9), %xmm3
+; SSE-NEXT:    movaps %xmm6, %xmm8
+; SSE-NEXT:    unpcklps {{.*#+}} xmm8 = xmm8[0],xmm3[0],xmm8[1],xmm3[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,1],xmm5[2,0]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm6, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[1,1],xmm7[1,1]
+; SSE-NEXT:    movaps %xmm1, %xmm8
+; SSE-NEXT:    unpcklps {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[2,3],xmm5[2,0]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm2, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,1],xmm4[3,3]
+; SSE-NEXT:    unpckhps {{.*#+}} xmm4 = xmm4[2],xmm1[2],xmm4[3],xmm1[3]
+; SSE-NEXT:    movaps %xmm3, %xmm8
+; SSE-NEXT:    unpcklps {{.*#+}} xmm8 = xmm8[0],xmm2[0],xmm8[1],xmm2[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[2,3],xmm4[0,1]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm0, %xmm4
+; SSE-NEXT:    unpckhps {{.*#+}} xmm4 = xmm4[2],xmm7[2],xmm4[3],xmm7[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[3,3],xmm6[3,3]
+; SSE-NEXT:    unpckhps {{.*#+}} xmm6 = xmm6[2],xmm3[2],xmm6[3],xmm3[3]
+; SSE-NEXT:    movlhps {{.*#+}} xmm4 = xmm4[0],xmm6[0]
 ; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm1, %xmm4
-; SSE-NEXT:    unpckhps {{.*#+}} xmm4 = xmm4[2],xmm13[2],xmm4[3],xmm13[3]
-; SSE-NEXT:    movaps %xmm5, %xmm0
-; SSE-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
-; SSE-NEXT:    movlhps {{.*#+}} xmm4 = xmm4[0],xmm0[0]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm1[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,2],xmm0[2,0]
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[3,3],xmm3[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,2],xmm2[2,0]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 64(%rdx), %xmm0
+; SSE-NEXT:    movaps 64(%rcx), %xmm7
+; SSE-NEXT:    movaps %xmm0, %xmm2
+; SSE-NEXT:    unpcklps {{.*#+}} xmm2 = xmm2[0],xmm7[0],xmm2[1],xmm7[1]
+; SSE-NEXT:    movaps 64(%rdi), %xmm4
+; SSE-NEXT:    movaps 64(%rsi), %xmm1
+; SSE-NEXT:    movaps %xmm4, %xmm3
+; SSE-NEXT:    unpcklps {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; SSE-NEXT:    movlhps {{.*#+}} xmm3 = xmm3[0],xmm2[0]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 64(%rax), %xmm2
+; SSE-NEXT:    movaps %xmm2, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,1],xmm4[1,3]
+; SSE-NEXT:    movaps 64(%r8), %xmm6
+; SSE-NEXT:    movaps 64(%r9), %xmm3
+; SSE-NEXT:    movaps %xmm6, %xmm8
+; SSE-NEXT:    unpcklps {{.*#+}} xmm8 = xmm8[0],xmm3[0],xmm8[1],xmm3[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,1],xmm5[0,2]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm6, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[1,1],xmm7[1,1]
+; SSE-NEXT:    movaps %xmm1, %xmm8
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[1,1],xmm0[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,2],xmm5[2,0]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm2, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,1],xmm4[3,3]
+; SSE-NEXT:    unpckhps {{.*#+}} xmm4 = xmm4[2],xmm1[2],xmm4[3],xmm1[3]
+; SSE-NEXT:    movaps %xmm3, %xmm8
+; SSE-NEXT:    unpcklps {{.*#+}} xmm8 = xmm8[0],xmm2[0],xmm8[1],xmm2[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[2,3],xmm4[0,1]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm0, %xmm4
+; SSE-NEXT:    unpckhps {{.*#+}} xmm4 = xmm4[2],xmm7[2],xmm4[3],xmm7[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[3,3],xmm6[3,3]
+; SSE-NEXT:    unpckhps {{.*#+}} xmm6 = xmm6[2],xmm3[2],xmm6[3],xmm3[3]
+; SSE-NEXT:    movlhps {{.*#+}} xmm4 = xmm4[0],xmm6[0]
 ; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm3, %xmm0
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, %xmm4
-; SSE-NEXT:    unpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm4 = xmm4[0],mem[0],xmm4[1],mem[1]
-; SSE-NEXT:    movlhps {{.*#+}} xmm4 = xmm4[0],xmm0[0]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm1[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,2],xmm0[2,0]
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[3,3],xmm3[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,2],xmm2[2,0]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 80(%rdx), %xmm0
+; SSE-NEXT:    movaps 80(%rcx), %xmm7
+; SSE-NEXT:    movaps %xmm7, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps 80(%rdi), %xmm4
+; SSE-NEXT:    movaps 80(%rsi), %xmm1
+; SSE-NEXT:    movaps %xmm4, %xmm3
+; SSE-NEXT:    unpcklps {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm2[2,0]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 80(%rax), %xmm2
+; SSE-NEXT:    movaps %xmm4, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[1,1],xmm2[0,3]
+; SSE-NEXT:    movaps 80(%r8), %xmm6
+; SSE-NEXT:    movaps 80(%r9), %xmm3
+; SSE-NEXT:    movaps %xmm6, %xmm8
+; SSE-NEXT:    unpcklps {{.*#+}} xmm8 = xmm8[0],xmm3[0],xmm8[1],xmm3[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,1],xmm5[2,0]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm6, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[1,1],xmm7[1,1]
+; SSE-NEXT:    movaps %xmm1, %xmm8
+; SSE-NEXT:    unpcklps {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[2,3],xmm5[2,0]
+; SSE-NEXT:    movaps %xmm8, (%rsp) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm2, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,1],xmm4[3,3]
+; SSE-NEXT:    unpckhps {{.*#+}} xmm4 = xmm4[2],xmm1[2],xmm4[3],xmm1[3]
+; SSE-NEXT:    movaps %xmm3, %xmm8
+; SSE-NEXT:    unpcklps {{.*#+}} xmm8 = xmm8[0],xmm2[0],xmm8[1],xmm2[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[2,3],xmm4[0,1]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm0, %xmm4
+; SSE-NEXT:    unpckhps {{.*#+}} xmm4 = xmm4[2],xmm7[2],xmm4[3],xmm7[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[3,3],xmm6[3,3]
+; SSE-NEXT:    unpckhps {{.*#+}} xmm6 = xmm6[2],xmm3[2],xmm6[3],xmm3[3]
+; SSE-NEXT:    movlhps {{.*#+}} xmm4 = xmm4[0],xmm6[0]
 ; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[1,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm1[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,2],xmm0[2,0]
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[3,3],xmm3[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,2],xmm2[2,0]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 96(%rdx), %xmm0
+; SSE-NEXT:    movaps 96(%rcx), %xmm8
+; SSE-NEXT:    movaps %xmm0, %xmm2
+; SSE-NEXT:    unpcklps {{.*#+}} xmm2 = xmm2[0],xmm8[0],xmm2[1],xmm8[1]
+; SSE-NEXT:    movaps 96(%rdi), %xmm4
+; SSE-NEXT:    movaps 96(%rsi), %xmm1
+; SSE-NEXT:    movaps %xmm4, %xmm3
+; SSE-NEXT:    unpcklps {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; SSE-NEXT:    movlhps {{.*#+}} xmm3 = xmm3[0],xmm2[0]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 96(%rax), %xmm2
+; SSE-NEXT:    movaps %xmm2, %xmm6
+; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,1],xmm4[1,3]
+; SSE-NEXT:    movaps 96(%r8), %xmm7
+; SSE-NEXT:    movaps 96(%r9), %xmm3
+; SSE-NEXT:    movaps %xmm7, %xmm5
+; SSE-NEXT:    unpcklps {{.*#+}} xmm5 = xmm5[0],xmm3[0],xmm5[1],xmm3[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,1],xmm6[0,2]
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm7, %xmm6
+; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[1,1],xmm8[1,1]
 ; SSE-NEXT:    movaps %xmm1, %xmm5
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm5 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,1],xmm0[0,2]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[1,1],xmm0[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,2],xmm6[2,0]
 ; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm1, %xmm0
-; SSE-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
-; SSE-NEXT:    movaps %xmm3, %xmm1
-; SSE-NEXT:    unpckhps {{.*#+}} xmm1 = xmm1[2],xmm2[2],xmm1[3],xmm2[3]
-; SSE-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm3, %xmm0
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; SSE-NEXT:    movlhps {{.*#+}} xmm0 = xmm0[0],xmm10[0]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, %xmm13
-; SSE-NEXT:    unpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm13 = xmm13[0],mem[0],xmm13[1],mem[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm13 = xmm13[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps %xmm1, %xmm0
-; SSE-NEXT:    shufps $197, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[1,1],mem[0,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, %xmm11
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm11 = xmm11[0],xmm4[0],xmm11[1],xmm4[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[0,1],xmm0[2,0]
-; SSE-NEXT:    unpckhps {{.*#+}} xmm10 = xmm10[2],xmm3[2],xmm10[3],xmm3[3]
-; SSE-NEXT:    movaps %xmm1, %xmm0
-; SSE-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
-; SSE-NEXT:    movlhps {{.*#+}} xmm10 = xmm10[0],xmm0[0]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm7, %xmm0
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm14, %xmm9
-; SSE-NEXT:    unpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm9 = xmm9[0],mem[0],xmm9[1],mem[1]
-; SSE-NEXT:    movlhps {{.*#+}} xmm9 = xmm9[0],xmm0[0]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm14[1,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, %xmm8
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm8 = xmm8[0],xmm4[0],xmm8[1],xmm4[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,1],xmm0[0,2]
-; SSE-NEXT:    unpckhps {{.*#+}} xmm1 = xmm1[2],xmm4[2],xmm1[3],xmm4[3]
-; SSE-NEXT:    unpckhps {{.*#+}} xmm7 = xmm7[2],xmm3[2],xmm7[3],xmm3[3]
-; SSE-NEXT:    movlhps {{.*#+}} xmm7 = xmm7[0],xmm1[0]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm2, %xmm5
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT:    unpckhps {{.*#+}} xmm5 = xmm5[2],xmm6[2],xmm5[3],xmm6[3]
-; SSE-NEXT:    movlhps {{.*#+}} xmm6 = xmm6[0],xmm2[0]
-; SSE-NEXT:    movaps 112(%rdi), %xmm3
-; SSE-NEXT:    movaps %xmm3, %xmm4
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm6[2,0]
-; SSE-NEXT:    movaps %xmm3, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm12[0,3]
-; SSE-NEXT:    movaps (%rsp), %xmm6 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,2],xmm0[2,0]
-; SSE-NEXT:    movaps %xmm6, (%rsp) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm3, %xmm0
-; SSE-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[1,1],xmm12[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,2],xmm0[0,1]
-; SSE-NEXT:    movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,1],xmm15[2,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[3,3],xmm1[3,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[3,1],xmm12[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[2,0],xmm2[2,0]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[3,3],mem[3,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm0[2,0]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm15[2,3,2,3]
-; SSE-NEXT:    movss {{.*#+}} xmm1 = xmm0[0],xmm1[1,2,3]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm15 = xmm15[3,3],mem[3,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm15[2,0]
-; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
-; SSE-NEXT:    movss {{.*#+}} xmm1 = xmm0[0],xmm1[1,2,3]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[3,3],mem[3,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,3],xmm0[2,0]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[2,3,2,3]
-; SSE-NEXT:    movss {{.*#+}} xmm2 = xmm0[0],xmm2[1,2,3]
-; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = xmm1[3,3],mem[3,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,3],xmm1[2,0]
-; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
-; SSE-NEXT:    movss {{.*#+}} xmm2 = xmm0[0],xmm2[1,2,3]
-; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[3,3],mem[3,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,3],xmm0[2,0]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[2,3,2,3]
-; SSE-NEXT:    movss {{.*#+}} xmm2 = xmm0[0],xmm2[1,2,3]
-; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = xmm1[3,3],mem[3,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,3],xmm1[2,0]
-; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
-; SSE-NEXT:    movss {{.*#+}} xmm2 = xmm0[0],xmm2[1,2,3]
-; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[3,3],mem[3,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,3],xmm0[2,0]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[2,3,2,3]
-; SSE-NEXT:    movss {{.*#+}} xmm2 = xmm0[0],xmm2[1,2,3]
-; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = xmm1[3,3],mem[3,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm15 = xmm15[0,3],xmm1[2,0]
-; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
-; SSE-NEXT:    movss {{.*#+}} xmm15 = xmm0[0],xmm15[1,2,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[3,3],mem[3,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[0,3],xmm0[2,0]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[2,3,2,3]
-; SSE-NEXT:    movss {{.*#+}} xmm12 = xmm0[0],xmm12[1,2,3]
-; SSE-NEXT:    shufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = xmm1[3,3],mem[3,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,3],xmm1[2,0]
-; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
-; SSE-NEXT:    movss {{.*#+}} xmm6 = xmm0[0],xmm6[1,2,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[3,3],mem[3,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,3],xmm0[2,0]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[2,3,2,3]
-; SSE-NEXT:    movss {{.*#+}} xmm2 = xmm0[0],xmm2[1,2,3]
-; SSE-NEXT:    shufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = xmm1[3,3],mem[3,3]
-; SSE-NEXT:    movaps %xmm1, %xmm0
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm0[2,0]
-; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[3,3,3,3]
-; SSE-NEXT:    movss {{.*#+}} xmm1 = xmm0[0],xmm1[1,2,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[3,3],mem[3,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[0,3],xmm0[2,0]
-; SSE-NEXT:    pshufd $238, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = mem[2,3,2,3]
-; SSE-NEXT:    movss {{.*#+}} xmm14 = xmm0[0],xmm14[1,2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,1],xmm4[3,3]
+; SSE-NEXT:    unpckhps {{.*#+}} xmm4 = xmm4[2],xmm1[2],xmm4[3],xmm1[3]
+; SSE-NEXT:    movaps %xmm3, %xmm15
+; SSE-NEXT:    unpcklps {{.*#+}} xmm15 = xmm15[0],xmm2[0],xmm15[1],xmm2[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm15 = xmm15[2,3],xmm4[0,1]
+; SSE-NEXT:    movaps %xmm0, %xmm14
+; SSE-NEXT:    unpckhps {{.*#+}} xmm14 = xmm14[2],xmm8[2],xmm14[3],xmm8[3]
+; SSE-NEXT:    unpckhps {{.*#+}} xmm8 = xmm8[2],xmm7[2],xmm8[3],xmm7[3]
+; SSE-NEXT:    unpckhps {{.*#+}} xmm7 = xmm7[2],xmm3[2],xmm7[3],xmm3[3]
+; SSE-NEXT:    movlhps {{.*#+}} xmm14 = xmm14[0],xmm7[0]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm1[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,2],xmm0[2,0]
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[3,3],xmm3[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[2,3],xmm2[2,0]
+; SSE-NEXT:    movaps 112(%rdx), %xmm4
+; SSE-NEXT:    movaps 112(%rcx), %xmm0
+; SSE-NEXT:    movaps %xmm0, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm4[0]
+; SSE-NEXT:    movaps 112(%rdi), %xmm1
+; SSE-NEXT:    movaps 112(%rsi), %xmm13
+; SSE-NEXT:    movaps %xmm1, %xmm7
+; SSE-NEXT:    unpcklps {{.*#+}} xmm7 = xmm7[0],xmm13[0],xmm7[1],xmm13[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,1],xmm2[2,0]
+; SSE-NEXT:    movaps 112(%rax), %xmm9
+; SSE-NEXT:    movaps %xmm1, %xmm12
+; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[1,1],xmm9[0,3]
+; SSE-NEXT:    movaps 112(%r8), %xmm10
+; SSE-NEXT:    movaps %xmm10, %xmm11
+; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[1,1],xmm0[1,1]
+; SSE-NEXT:    movaps 112(%r9), %xmm5
+; SSE-NEXT:    movaps %xmm5, %xmm6
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm6 = xmm6[1],xmm10[1]
+; SSE-NEXT:    movaps %xmm4, %xmm3
+; SSE-NEXT:    unpckhps {{.*#+}} xmm3 = xmm3[2],xmm0[2],xmm3[3],xmm0[3]
+; SSE-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2],xmm10[2],xmm0[3],xmm10[3]
+; SSE-NEXT:    movlhps {{.*#+}} xmm10 = xmm10[0],xmm5[0]
+; SSE-NEXT:    shufps {{.*#+}} xmm10 = xmm10[0,2],xmm12[2,0]
+; SSE-NEXT:    movaps %xmm13, %xmm12
+; SSE-NEXT:    unpcklps {{.*#+}} xmm12 = xmm12[0],xmm4[0],xmm12[1],xmm4[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[2,3],xmm11[2,0]
+; SSE-NEXT:    movaps %xmm1, %xmm11
+; SSE-NEXT:    unpckhps {{.*#+}} xmm11 = xmm11[2],xmm13[2],xmm11[3],xmm13[3]
+; SSE-NEXT:    movaps %xmm5, %xmm2
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[1,1],xmm9[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,2],xmm11[0,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm6[2,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[3,3],xmm13[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,1],xmm9[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm4[2,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[3,3],xmm9[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,3],xmm5[0,2]
 ; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; SSE-NEXT:    movaps %xmm3, 864(%rax)
-; SSE-NEXT:    movaps %xmm5, 848(%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 832(%rax)
-; SSE-NEXT:    movaps (%rsp), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 800(%rax)
-; SSE-NEXT:    movaps %xmm4, 784(%rax)
-; SSE-NEXT:    movaps %xmm7, 736(%rax)
-; SSE-NEXT:    movaps %xmm8, 688(%rax)
-; SSE-NEXT:    movaps %xmm9, 672(%rax)
-; SSE-NEXT:    movaps %xmm10, 624(%rax)
-; SSE-NEXT:    movaps %xmm11, 576(%rax)
-; SSE-NEXT:    movaps %xmm13, 560(%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 512(%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 464(%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 448(%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 400(%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 352(%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 336(%rax)
+; SSE-NEXT:    movaps %xmm0, 880(%rax)
+; SSE-NEXT:    movaps %xmm1, 864(%rax)
+; SSE-NEXT:    movaps %xmm3, 848(%rax)
+; SSE-NEXT:    movaps %xmm2, 832(%rax)
+; SSE-NEXT:    movaps %xmm12, 816(%rax)
+; SSE-NEXT:    movaps %xmm10, 800(%rax)
+; SSE-NEXT:    movaps %xmm7, 784(%rax)
+; SSE-NEXT:    movaps %xmm8, 768(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 288(%rax)
+; SSE-NEXT:    movaps %xmm0, 752(%rax)
+; SSE-NEXT:    movaps %xmm14, 736(%rax)
+; SSE-NEXT:    movaps %xmm15, 720(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 240(%rax)
+; SSE-NEXT:    movaps %xmm0, 704(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 224(%rax)
+; SSE-NEXT:    movaps %xmm0, 688(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 176(%rax)
+; SSE-NEXT:    movaps %xmm0, 672(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 128(%rax)
+; SSE-NEXT:    movaps %xmm0, 656(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 112(%rax)
+; SSE-NEXT:    movaps %xmm0, 640(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 64(%rax)
+; SSE-NEXT:    movaps %xmm0, 624(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 16(%rax)
+; SSE-NEXT:    movaps %xmm0, 608(%rax)
+; SSE-NEXT:    movaps (%rsp), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 592(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, (%rax)
+; SSE-NEXT:    movaps %xmm0, 576(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 880(%rax)
+; SSE-NEXT:    movaps %xmm0, 560(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 816(%rax)
+; SSE-NEXT:    movaps %xmm0, 544(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 768(%rax)
-; SSE-NEXT:    movaps %xmm14, 752(%rax)
+; SSE-NEXT:    movaps %xmm0, 528(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 720(%rax)
+; SSE-NEXT:    movaps %xmm0, 512(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 704(%rax)
-; SSE-NEXT:    movaps %xmm1, 656(%rax)
-; SSE-NEXT:    movaps %xmm2, 640(%rax)
+; SSE-NEXT:    movaps %xmm0, 496(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 608(%rax)
+; SSE-NEXT:    movaps %xmm0, 480(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 592(%rax)
-; SSE-NEXT:    movaps %xmm6, 544(%rax)
-; SSE-NEXT:    movaps %xmm12, 528(%rax)
+; SSE-NEXT:    movaps %xmm0, 464(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 496(%rax)
+; SSE-NEXT:    movaps %xmm0, 448(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 480(%rax)
-; SSE-NEXT:    movaps %xmm15, 432(%rax)
+; SSE-NEXT:    movaps %xmm0, 432(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 416(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 400(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 384(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 368(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 352(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 336(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 320(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 304(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 288(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 272(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 256(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 240(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 224(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 208(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 192(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 176(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 160(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 144(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 128(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 112(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 96(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 80(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 48(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 32(%rax)
-; SSE-NEXT:    addq $1256, %rsp # imm = 0x4E8
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%rax)
+; SSE-NEXT:    addq $616, %rsp # imm = 0x268
 ; SSE-NEXT:    retq
 ;
 ; AVX1-ONLY-LABEL: store_i32_stride7_vf32:
 ; AVX1-ONLY:       # %bb.0:
 ; AVX1-ONLY-NEXT:    subq $1624, %rsp # imm = 0x658
 ; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %ymm2
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX512BW-NEXT:    retq
   %interleaved.vec = shufflevector <224 x i32> %8, <224 x i32> poison, <224 x i32> <i32 0, i32 32, i32 64, i32 96, i32 128, i32 160, i32 192, i32 1, i32 33, i32 65, i32 97, i32 129, i32 161, i32 193, i32 2, i32 34, i32 66, i32 98, i32 130, i32 162, i32 194, i32 3, i32 35, i32 67, i32 99, i32 131, i32 163, i32 195, i32 4, i32 36, i32 68, i32 100, i32 132, i32 164, i32 196, i32 5, i32 37, i32 69, i32 101, i32 133, i32 165, i32 197, i32 6, i32 38, i32 70, i32 102, i32 134, i32 166, i32 198, i32 7, i32 39, i32 71, i32 103, i32 135, i32 167, i32 199, i32 8, i32 40, i32 72, i32 104, i32 136, i32 168, i32 200, i32 9, i32 41, i32 73, i32 105, i32 137, i32 169, i32 201, i32 10, i32 42, i32 74, i32 106, i32 138, i32 170, i32 202, i32 11, i32 43, i32 75, i32 107, i32 139, i32 171, i32 203, i32 12, i32 44, i32 76, i32 108, i32 140, i32 172, i32 204, i32 13, i32 45, i32 77, i32 109, i32 141, i32 173, i32 205, i32 14, i32 46, i32 78, i32 110, i32 142, i32 174, i32 206, i32 15, i32 47, i32 79, i32 111, i32 143, i32 175, i32 207, i32 16, i32 48, i32 80, i32 112, i32 144, i32 176, i32 208, i32 17, i32 49, i32 81, i32 113, i32 145, i32 177, i32 209, i32 18, i32 50, i32 82, i32 114, i32 146, i32 178, i32 210, i32 19, i32 51, i32 83, i32 115, i32 147, i32 179, i32 211, i32 20, i32 52, i32 84, i32 116, i32 148, i32 180, i32 212, i32 21, i32 53, i32 85, i32 117, i32 149, i32 181, i32 213, i32 22, i32 54, i32 86, i32 118, i32 150, i32 182, i32 214, i32 23, i32 55, i32 87, i32 119, i32 151, i32 183, i32 215, i32 24, i32 56, i32 88, i32 120, i32 152, i32 184, i32 216, i32 25, i32 57, i32 89, i32 121, i32 153, i32 185, i32 217, i32 26, i32 58, i32 90, i32 122, i32 154, i32 186, i32 218, i32 27, i32 59, i32 91, i32 123, i32 155, i32 187, i32 219, i32 28, i32 60, i32 92, i32 124, i32 156, i32 188, i32 220, i32 29, i32 61, i32 93, i32 125, i32 157, i32 189, i32 221, i32 30, i32 62, i32 94, i32 126, i32 158, i32 190, i32 222, i32 31, i32 63, i32 95, i32 127, i32 159, i32 191, i32 223>
   store <224 x i32> %interleaved.vec, ptr %out.vec, align 64
   ret void
 }
 
 define void @store_i32_stride7_vf64(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vecptr2, ptr %in.vecptr3, ptr %in.vecptr4, ptr %in.vecptr5, ptr %in.vecptr6, ptr %out.vec) nounwind {
 ; SSE-LABEL: store_i32_stride7_vf64:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    subq $2760, %rsp # imm = 0xAC8
+; SSE-NEXT:    subq $1512, %rsp # imm = 0x5E8
 ; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; SSE-NEXT:    movdqa (%rdi), %xmm6
-; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa (%rsi), %xmm5
-; SSE-NEXT:    movdqa 16(%rsi), %xmm4
-; SSE-NEXT:    movaps (%rdx), %xmm3
-; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 16(%rdx), %xmm7
-; SSE-NEXT:    movdqa %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps (%rcx), %xmm9
-; SSE-NEXT:    movaps 16(%rcx), %xmm10
+; SSE-NEXT:    movaps (%rdi), %xmm3
+; SSE-NEXT:    movaps 16(%rdi), %xmm1
+; SSE-NEXT:    movaps (%rsi), %xmm4
+; SSE-NEXT:    movaps 16(%rsi), %xmm0
+; SSE-NEXT:    movaps (%rdx), %xmm5
+; SSE-NEXT:    movaps 16(%rdx), %xmm2
+; SSE-NEXT:    movaps (%rcx), %xmm11
+; SSE-NEXT:    movaps (%r8), %xmm6
+; SSE-NEXT:    movaps (%r9), %xmm7
+; SSE-NEXT:    movaps (%rax), %xmm8
+; SSE-NEXT:    movaps %xmm5, %xmm9
+; SSE-NEXT:    unpcklps {{.*#+}} xmm9 = xmm9[0],xmm11[0],xmm9[1],xmm11[1]
+; SSE-NEXT:    movaps %xmm3, %xmm10
+; SSE-NEXT:    unpcklps {{.*#+}} xmm10 = xmm10[0],xmm4[0],xmm10[1],xmm4[1]
+; SSE-NEXT:    movlhps {{.*#+}} xmm10 = xmm10[0],xmm9[0]
 ; SSE-NEXT:    movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps (%r8), %xmm0
-; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 16(%r8), %xmm8
+; SSE-NEXT:    movaps %xmm8, %xmm9
+; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[0,1],xmm3[1,3]
+; SSE-NEXT:    movaps %xmm6, %xmm10
+; SSE-NEXT:    shufps {{.*#+}} xmm10 = xmm10[1,1],xmm11[1,1]
+; SSE-NEXT:    movaps %xmm11, %xmm12
+; SSE-NEXT:    movaps %xmm6, %xmm11
+; SSE-NEXT:    movaps %xmm5, %xmm13
+; SSE-NEXT:    unpckhps {{.*#+}} xmm13 = xmm13[2],xmm12[2],xmm13[3],xmm12[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[3,3],xmm6[3,3]
+; SSE-NEXT:    unpcklps {{.*#+}} xmm6 = xmm6[0],xmm7[0],xmm6[1],xmm7[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,1],xmm9[0,2]
+; SSE-NEXT:    movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm4, %xmm6
+; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[1,1],xmm5[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,2],xmm10[2,0]
+; SSE-NEXT:    movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm8, %xmm9
+; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[2,1],xmm3[3,3]
+; SSE-NEXT:    unpckhps {{.*#+}} xmm3 = xmm3[2],xmm4[2],xmm3[3],xmm4[3]
+; SSE-NEXT:    movaps %xmm7, %xmm6
+; SSE-NEXT:    unpcklps {{.*#+}} xmm6 = xmm6[0],xmm8[0],xmm6[1],xmm8[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[2,3],xmm3[0,1]
+; SSE-NEXT:    movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 16(%rcx), %xmm10
+; SSE-NEXT:    unpckhps {{.*#+}} xmm11 = xmm11[2],xmm7[2],xmm11[3],xmm7[3]
+; SSE-NEXT:    movlhps {{.*#+}} xmm13 = xmm13[0],xmm11[0]
+; SSE-NEXT:    movaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 16(%r8), %xmm6
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[3,3],xmm4[3,3]
+; SSE-NEXT:    movaps 16(%r9), %xmm4
+; SSE-NEXT:    shufps {{.*#+}} xmm9 = xmm9[0,2],xmm5[2,0]
+; SSE-NEXT:    movaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 16(%rax), %xmm3
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[3,3],xmm7[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[0,2],xmm8[2,0]
+; SSE-NEXT:    movaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm10, %xmm5
+; SSE-NEXT:    movlhps {{.*#+}} xmm5 = xmm5[0],xmm2[0]
+; SSE-NEXT:    movaps %xmm1, %xmm8
+; SSE-NEXT:    unpcklps {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,1],xmm5[2,0]
 ; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa (%r9), %xmm11
-; SSE-NEXT:    movdqa %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 16(%r9), %xmm12
-; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa (%rax), %xmm15
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm9[1,1]
-; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm5[1,1,1,1]
-; SSE-NEXT:    movss {{.*#+}} xmm3 = xmm1[0],xmm3[1,2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm6, %xmm0
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm5[2],xmm0[3],xmm5[3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm15[1,1,1,1]
-; SSE-NEXT:    movdqa %xmm15, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm11[1,1,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm3 = xmm3[0],xmm0[0]
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm7[1,1,1,1]
-; SSE-NEXT:    movdqa %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm4[1,1,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movaps %xmm8, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm10[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 16(%rax), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm12[1,1,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movdqa 16(%rdi), %xmm5
-; SSE-NEXT:    movdqa %xmm5, %xmm0
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm0[0]
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 32(%rsi), %xmm1
-; SSE-NEXT:    movaps 32(%rdx), %xmm4
-; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[1,1,1,1]
-; SSE-NEXT:    movdqa %xmm1, %xmm2
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm4, %xmm1
-; SSE-NEXT:    movss {{.*#+}} xmm1 = xmm0[0],xmm1[1,2,3]
-; SSE-NEXT:    movaps 32(%rcx), %xmm4
-; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 32(%r8), %xmm0
-; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm4[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 32(%r9), %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 32(%rax), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movdqa 32(%rdi), %xmm6
-; SSE-NEXT:    movdqa %xmm6, %xmm0
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm0[0]
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 48(%rsi), %xmm2
-; SSE-NEXT:    movdqa 48(%rdx), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[1,1,1,1]
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movaps 48(%rcx), %xmm4
-; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 48(%r8), %xmm7
-; SSE-NEXT:    movaps %xmm7, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm4[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 48(%r9), %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 48(%rax), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movdqa 48(%rdi), %xmm10
-; SSE-NEXT:    movdqa %xmm10, %xmm0
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm0[0]
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 64(%rsi), %xmm1
-; SSE-NEXT:    movaps 64(%rdx), %xmm4
-; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[1,1,1,1]
-; SSE-NEXT:    movdqa %xmm1, %xmm2
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm4, %xmm1
-; SSE-NEXT:    movss {{.*#+}} xmm1 = xmm0[0],xmm1[1,2,3]
-; SSE-NEXT:    movaps 64(%rcx), %xmm4
-; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 64(%r8), %xmm13
-; SSE-NEXT:    movaps %xmm13, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm4[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 64(%r9), %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 64(%rax), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movdqa 64(%rdi), %xmm14
-; SSE-NEXT:    movdqa %xmm14, %xmm0
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm0[0]
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 80(%rsi), %xmm2
-; SSE-NEXT:    movdqa 80(%rdx), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[1,1,1,1]
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movaps 80(%rcx), %xmm4
-; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 80(%r8), %xmm0
-; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm4[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 80(%r9), %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 80(%rax), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movdqa 80(%rdi), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm0[0]
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 96(%rsi), %xmm1
-; SSE-NEXT:    movaps 96(%rdx), %xmm4
-; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[1,1,1,1]
-; SSE-NEXT:    movdqa %xmm1, %xmm2
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm4, %xmm1
-; SSE-NEXT:    movss {{.*#+}} xmm1 = xmm0[0],xmm1[1,2,3]
-; SSE-NEXT:    movaps 96(%rcx), %xmm4
-; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 96(%r8), %xmm0
-; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm4[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 96(%r9), %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 96(%rax), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movdqa 96(%rdi), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm0[0]
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 112(%rsi), %xmm2
-; SSE-NEXT:    movdqa 112(%rdx), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[1,1,1,1]
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movaps 112(%rcx), %xmm4
-; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 112(%r8), %xmm0
-; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm4[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 112(%r9), %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 112(%rax), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movdqa 112(%rdi), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm0[0]
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 128(%rsi), %xmm1
-; SSE-NEXT:    movaps 128(%rdx), %xmm4
-; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[1,1,1,1]
-; SSE-NEXT:    movdqa %xmm1, %xmm2
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm4, %xmm1
-; SSE-NEXT:    movss {{.*#+}} xmm1 = xmm0[0],xmm1[1,2,3]
-; SSE-NEXT:    movaps 128(%rcx), %xmm4
-; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 128(%r8), %xmm0
-; SSE-NEXT:    movaps %xmm0, (%rsp) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm4[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 128(%r9), %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 128(%rax), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movdqa 128(%rdi), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm0[0]
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 144(%rsi), %xmm2
-; SSE-NEXT:    movdqa 144(%rdx), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[1,1,1,1]
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movaps 144(%rcx), %xmm4
-; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 144(%r8), %xmm0
-; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm4[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 144(%r9), %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 144(%rax), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movdqa 144(%rdi), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm0[0]
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 160(%rsi), %xmm1
-; SSE-NEXT:    movaps 160(%rdx), %xmm4
-; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[1,1,1,1]
-; SSE-NEXT:    movdqa %xmm1, %xmm2
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm4, %xmm1
-; SSE-NEXT:    movss {{.*#+}} xmm1 = xmm0[0],xmm1[1,2,3]
-; SSE-NEXT:    movaps 160(%rcx), %xmm4
-; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 160(%r8), %xmm0
-; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm4[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 160(%r9), %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 160(%rax), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movdqa 160(%rdi), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm0[0]
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 176(%rsi), %xmm2
-; SSE-NEXT:    movdqa 176(%rdx), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[1,1,1,1]
-; SSE-NEXT:    movdqa %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movaps 176(%rcx), %xmm4
-; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 176(%r8), %xmm0
-; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm4[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 176(%r9), %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 176(%rax), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movdqa 176(%rdi), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm0[0]
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 192(%rsi), %xmm1
-; SSE-NEXT:    movaps 192(%rdx), %xmm4
-; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[1,1,1,1]
-; SSE-NEXT:    movdqa %xmm1, %xmm2
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm4, %xmm1
-; SSE-NEXT:    movss {{.*#+}} xmm1 = xmm0[0],xmm1[1,2,3]
-; SSE-NEXT:    movaps 192(%rcx), %xmm4
-; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 192(%r8), %xmm0
-; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm4[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 192(%r9), %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 192(%rax), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movdqa 192(%rdi), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm2[2],xmm0[3],xmm2[3]
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm0[0]
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 208(%rsi), %xmm3
-; SSE-NEXT:    movdqa 208(%rdx), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[1,1,1,1]
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movaps 208(%rcx), %xmm8
-; SSE-NEXT:    movaps 208(%r8), %xmm4
-; SSE-NEXT:    movaps %xmm4, %xmm0
-; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm8[1,1]
-; SSE-NEXT:    movaps %xmm8, %xmm11
+; SSE-NEXT:    movaps %xmm1, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[1,1],xmm3[0,3]
+; SSE-NEXT:    movaps %xmm6, %xmm8
+; SSE-NEXT:    unpcklps {{.*#+}} xmm8 = xmm8[0],xmm4[0],xmm8[1],xmm4[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,1],xmm5[2,0]
 ; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 208(%r9), %xmm12
-; SSE-NEXT:    movdqa %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 208(%rax), %xmm8
-; SSE-NEXT:    movdqa %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm8[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm12[1,1,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movdqa 208(%rdi), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm3[2],xmm0[3],xmm3[3]
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm0[0]
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm4[3,3,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm11[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
-; SSE-NEXT:    movdqa %xmm8, %xmm0
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm12[3,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 224(%rsi), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps 224(%rdx), %xmm2
-; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
-; SSE-NEXT:    movss {{.*#+}} xmm2 = xmm0[0],xmm2[1,2,3]
-; SSE-NEXT:    movaps 224(%rcx), %xmm3
-; SSE-NEXT:    movaps 224(%r8), %xmm0
-; SSE-NEXT:    movaps %xmm0, %xmm1
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,1],xmm3[1,1]
-; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,0]
-; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm0, %xmm1
-; SSE-NEXT:    movaps 224(%r9), %xmm11
-; SSE-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm11[0]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm11, %xmm1
-; SSE-NEXT:    movaps %xmm11, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    unpckhpd {{.*#+}} xmm11 = xmm11[1],xmm0[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
-; SSE-NEXT:    movaps 224(%rax), %xmm0
-; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,3],xmm0[3,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[0,2]
-; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 240(%rsi), %xmm1
-; SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa 240(%rdx), %xmm0
-; SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[1,1,1,1]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
-; SSE-NEXT:    movaps 240(%rcx), %xmm8
-; SSE-NEXT:    movaps 240(%r8), %xmm0
-; SSE-NEXT:    movaps %xmm0, %xmm1
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,1],xmm8[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,0]
-; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm0, %xmm1
-; SSE-NEXT:    movaps 240(%r9), %xmm12
-; SSE-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm12[0]
+; SSE-NEXT:    movaps %xmm6, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[1,1],xmm10[1,1]
+; SSE-NEXT:    movaps %xmm0, %xmm8
+; SSE-NEXT:    unpcklps {{.*#+}} xmm8 = xmm8[0],xmm2[0],xmm8[1],xmm2[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[2,3],xmm5[2,0]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm3, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,1],xmm1[3,3]
+; SSE-NEXT:    movaps %xmm5, %xmm8
+; SSE-NEXT:    unpckhps {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSE-NEXT:    movaps %xmm4, %xmm5
+; SSE-NEXT:    unpcklps {{.*#+}} xmm5 = xmm5[0],xmm3[0],xmm5[1],xmm3[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,3],xmm1[0,1]
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm2, %xmm1
+; SSE-NEXT:    unpckhps {{.*#+}} xmm1 = xmm1[2],xmm10[2],xmm1[3],xmm10[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm10 = xmm10[3,3],xmm6[3,3]
+; SSE-NEXT:    unpckhps {{.*#+}} xmm6 = xmm6[2],xmm4[2],xmm6[3],xmm4[3]
+; SSE-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm6[0]
 ; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm12, %xmm1
-; SSE-NEXT:    movaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    unpckhpd {{.*#+}} xmm12 = xmm12[1],xmm0[1]
-; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
-; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm8[3,3,3,3]
-; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
-; SSE-NEXT:    movaps 240(%rax), %xmm0
-; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[3,3],xmm0[3,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[0,2]
-; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, %xmm0
-; SSE-NEXT:    movaps %xmm9, %xmm2
-; SSE-NEXT:    movaps %xmm9, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm9[0],xmm0[1],xmm9[1]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[3,3],xmm0[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,2],xmm2[2,0]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[3,3],xmm4[3,3]
+; SSE-NEXT:    movaps 32(%rdx), %xmm0
+; SSE-NEXT:    shufps {{.*#+}} xmm10 = xmm10[0,2],xmm3[2,0]
+; SSE-NEXT:    movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 32(%rcx), %xmm7
+; SSE-NEXT:    movaps %xmm0, %xmm2
+; SSE-NEXT:    unpcklps {{.*#+}} xmm2 = xmm2[0],xmm7[0],xmm2[1],xmm7[1]
+; SSE-NEXT:    movaps 32(%rdi), %xmm4
+; SSE-NEXT:    movaps 32(%rsi), %xmm1
 ; SSE-NEXT:    movaps %xmm4, %xmm3
-; SSE-NEXT:    unpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm3 = xmm3[0],mem[0],xmm3[1],mem[1]
-; SSE-NEXT:    movlhps {{.*#+}} xmm3 = xmm3[0],xmm0[0]
+; SSE-NEXT:    unpcklps {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; SSE-NEXT:    movlhps {{.*#+}} xmm3 = xmm3[0],xmm2[0]
 ; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm15 = xmm15[0,1],xmm4[1,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; SSE-NEXT:    movaps 32(%rax), %xmm2
+; SSE-NEXT:    movaps %xmm2, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,1],xmm4[1,3]
+; SSE-NEXT:    movaps 32(%r8), %xmm6
+; SSE-NEXT:    movaps 32(%r9), %xmm3
+; SSE-NEXT:    movaps %xmm6, %xmm8
+; SSE-NEXT:    unpcklps {{.*#+}} xmm8 = xmm8[0],xmm3[0],xmm8[1],xmm3[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,1],xmm5[0,2]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm6, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[1,1],xmm7[1,1]
+; SSE-NEXT:    movaps %xmm1, %xmm8
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[1,1],xmm0[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,2],xmm5[2,0]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm2, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,1],xmm4[3,3]
+; SSE-NEXT:    unpckhps {{.*#+}} xmm4 = xmm4[2],xmm1[2],xmm4[3],xmm1[3]
+; SSE-NEXT:    movaps %xmm3, %xmm8
+; SSE-NEXT:    unpcklps {{.*#+}} xmm8 = xmm8[0],xmm2[0],xmm8[1],xmm2[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[2,3],xmm4[0,1]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm0, %xmm4
+; SSE-NEXT:    unpckhps {{.*#+}} xmm4 = xmm4[2],xmm7[2],xmm4[3],xmm7[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[3,3],xmm6[3,3]
+; SSE-NEXT:    unpckhps {{.*#+}} xmm6 = xmm6[2],xmm3[2],xmm6[3],xmm3[3]
+; SSE-NEXT:    movlhps {{.*#+}} xmm4 = xmm4[0],xmm6[0]
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm1[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,2],xmm0[2,0]
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[3,3],xmm3[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,2],xmm2[2,0]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 48(%rdx), %xmm0
+; SSE-NEXT:    movaps 48(%rcx), %xmm7
+; SSE-NEXT:    movaps %xmm7, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps 48(%rdi), %xmm4
+; SSE-NEXT:    movaps 48(%rsi), %xmm1
 ; SSE-NEXT:    movaps %xmm4, %xmm3
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm3 = xmm3[0],xmm9[0],xmm3[1],xmm9[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm15[0,2]
+; SSE-NEXT:    unpcklps {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm2[2,0]
 ; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm4, %xmm0
-; SSE-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2],xmm9[2],xmm0[3],xmm9[3]
-; SSE-NEXT:    movaps %xmm1, %xmm3
-; SSE-NEXT:    unpckhps {{.*#+}} xmm3 = xmm3[2],xmm2[2],xmm3[3],xmm2[3]
-; SSE-NEXT:    movlhps {{.*#+}} xmm3 = xmm3[0],xmm0[0]
+; SSE-NEXT:    movaps 48(%rax), %xmm2
+; SSE-NEXT:    movaps %xmm4, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[1,1],xmm2[0,3]
+; SSE-NEXT:    movaps 48(%r8), %xmm6
+; SSE-NEXT:    movaps 48(%r9), %xmm3
+; SSE-NEXT:    movaps %xmm6, %xmm8
+; SSE-NEXT:    unpcklps {{.*#+}} xmm8 = xmm8[0],xmm3[0],xmm8[1],xmm3[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,1],xmm5[2,0]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm6, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[1,1],xmm7[1,1]
+; SSE-NEXT:    movaps %xmm1, %xmm8
+; SSE-NEXT:    unpcklps {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[2,3],xmm5[2,0]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm2, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,1],xmm4[3,3]
+; SSE-NEXT:    unpckhps {{.*#+}} xmm4 = xmm4[2],xmm1[2],xmm4[3],xmm1[3]
+; SSE-NEXT:    movaps %xmm3, %xmm8
+; SSE-NEXT:    unpcklps {{.*#+}} xmm8 = xmm8[0],xmm2[0],xmm8[1],xmm2[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[2,3],xmm4[0,1]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm0, %xmm4
+; SSE-NEXT:    unpckhps {{.*#+}} xmm4 = xmm4[2],xmm7[2],xmm4[3],xmm7[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[3,3],xmm6[3,3]
+; SSE-NEXT:    unpckhps {{.*#+}} xmm6 = xmm6[2],xmm3[2],xmm6[3],xmm3[3]
+; SSE-NEXT:    movlhps {{.*#+}} xmm4 = xmm4[0],xmm6[0]
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm1[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,2],xmm0[2,0]
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[3,3],xmm3[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,2],xmm2[2,0]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 64(%rdx), %xmm0
+; SSE-NEXT:    movaps 64(%rcx), %xmm7
+; SSE-NEXT:    movaps %xmm0, %xmm2
+; SSE-NEXT:    unpcklps {{.*#+}} xmm2 = xmm2[0],xmm7[0],xmm2[1],xmm7[1]
+; SSE-NEXT:    movaps 64(%rdi), %xmm4
+; SSE-NEXT:    movaps 64(%rsi), %xmm1
+; SSE-NEXT:    movaps %xmm4, %xmm3
+; SSE-NEXT:    unpcklps {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; SSE-NEXT:    movlhps {{.*#+}} xmm3 = xmm3[0],xmm2[0]
 ; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, %xmm0
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    movlhps {{.*#+}} xmm0 = xmm0[0],xmm4[0]
-; SSE-NEXT:    movdqa %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm5, %xmm2
-; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm2 = xmm2[0],mem[0],xmm2[1],mem[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm5, %xmm0
-; SSE-NEXT:    shufps $197, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[1,1],mem[0,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; SSE-NEXT:    movaps 64(%rax), %xmm2
 ; SSE-NEXT:    movaps %xmm2, %xmm5
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm5 = xmm5[0],xmm3[0],xmm5[1],xmm3[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,1],xmm0[2,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,1],xmm4[1,3]
+; SSE-NEXT:    movaps 64(%r8), %xmm6
+; SSE-NEXT:    movaps 64(%r9), %xmm3
+; SSE-NEXT:    movaps %xmm6, %xmm8
+; SSE-NEXT:    unpcklps {{.*#+}} xmm8 = xmm8[0],xmm3[0],xmm8[1],xmm3[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,1],xmm5[0,2]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm6, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[1,1],xmm7[1,1]
+; SSE-NEXT:    movaps %xmm1, %xmm8
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[1,1],xmm0[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,2],xmm5[2,0]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm2, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,1],xmm4[3,3]
+; SSE-NEXT:    unpckhps {{.*#+}} xmm4 = xmm4[2],xmm1[2],xmm4[3],xmm1[3]
+; SSE-NEXT:    movaps %xmm3, %xmm8
+; SSE-NEXT:    unpcklps {{.*#+}} xmm8 = xmm8[0],xmm2[0],xmm8[1],xmm2[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[2,3],xmm4[0,1]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm0, %xmm4
+; SSE-NEXT:    unpckhps {{.*#+}} xmm4 = xmm4[2],xmm7[2],xmm4[3],xmm7[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[3,3],xmm6[3,3]
+; SSE-NEXT:    unpckhps {{.*#+}} xmm6 = xmm6[2],xmm3[2],xmm6[3],xmm3[3]
+; SSE-NEXT:    movlhps {{.*#+}} xmm4 = xmm4[0],xmm6[0]
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm1[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,2],xmm0[2,0]
 ; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[3,3],xmm3[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,2],xmm2[2,0]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 80(%rdx), %xmm0
+; SSE-NEXT:    movaps 80(%rcx), %xmm7
+; SSE-NEXT:    movaps %xmm7, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps 80(%rdi), %xmm4
+; SSE-NEXT:    movaps 80(%rsi), %xmm1
+; SSE-NEXT:    movaps %xmm4, %xmm3
+; SSE-NEXT:    unpcklps {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm2[2,0]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 80(%rax), %xmm2
 ; SSE-NEXT:    movaps %xmm4, %xmm5
-; SSE-NEXT:    unpckhps {{.*#+}} xmm5 = xmm5[2],xmm1[2],xmm5[3],xmm1[3]
-; SSE-NEXT:    movaps %xmm2, %xmm0
-; SSE-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2],xmm3[2],xmm0[3],xmm3[3]
-; SSE-NEXT:    movlhps {{.*#+}} xmm5 = xmm5[0],xmm0[0]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[1,1],xmm2[0,3]
+; SSE-NEXT:    movaps 80(%r8), %xmm6
+; SSE-NEXT:    movaps 80(%r9), %xmm3
+; SSE-NEXT:    movaps %xmm6, %xmm8
+; SSE-NEXT:    unpcklps {{.*#+}} xmm8 = xmm8[0],xmm3[0],xmm8[1],xmm3[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,1],xmm5[2,0]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm6, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[1,1],xmm7[1,1]
+; SSE-NEXT:    movaps %xmm1, %xmm8
+; SSE-NEXT:    unpcklps {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[2,3],xmm5[2,0]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm2, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,1],xmm4[3,3]
+; SSE-NEXT:    unpckhps {{.*#+}} xmm4 = xmm4[2],xmm1[2],xmm4[3],xmm1[3]
+; SSE-NEXT:    movaps %xmm3, %xmm8
+; SSE-NEXT:    unpcklps {{.*#+}} xmm8 = xmm8[0],xmm2[0],xmm8[1],xmm2[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[2,3],xmm4[0,1]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm0, %xmm4
+; SSE-NEXT:    unpckhps {{.*#+}} xmm4 = xmm4[2],xmm7[2],xmm4[3],xmm7[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[3,3],xmm6[3,3]
+; SSE-NEXT:    unpckhps {{.*#+}} xmm6 = xmm6[2],xmm3[2],xmm6[3],xmm3[3]
+; SSE-NEXT:    movlhps {{.*#+}} xmm4 = xmm4[0],xmm6[0]
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm1[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,2],xmm0[2,0]
 ; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm2, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT:    movdqa %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm6, %xmm3
-; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm3 = xmm3[0],mem[0],xmm3[1],mem[1]
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm3 = xmm3[0],xmm0[0]
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm6[1,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm4, %xmm6
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm6 = xmm6[0],xmm3[0],xmm6[1],xmm3[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,1],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm4, %xmm0
-; SSE-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2],xmm3[2],xmm0[3],xmm3[3]
-; SSE-NEXT:    movdqa %xmm2, %xmm3
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm1[2],xmm3[3],xmm1[3]
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm3 = xmm3[0],xmm0[0]
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm2, %xmm0
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
-; SSE-NEXT:    movdqa %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm10, %xmm3
-; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm3 = xmm3[0],mem[0],xmm3[1],mem[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm10, %xmm0
-; SSE-NEXT:    shufps $197, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[1,1],mem[0,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[3,3],xmm3[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,2],xmm2[2,0]
 ; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm7, %xmm5
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm5 = xmm5[0],xmm3[0],xmm5[1],xmm3[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm1, %xmm5
-; SSE-NEXT:    unpckhps {{.*#+}} xmm5 = xmm5[2],xmm2[2],xmm5[3],xmm2[3]
-; SSE-NEXT:    movaps %xmm7, %xmm0
-; SSE-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2],xmm3[2],xmm0[3],xmm3[3]
-; SSE-NEXT:    movlhps {{.*#+}} xmm5 = xmm5[0],xmm0[0]
-; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    movdqa %xmm2, %xmm0
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; SSE-NEXT:    movdqa %xmm14, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movdqa %xmm14, %xmm3
-; SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm3 = xmm3[0],mem[0],xmm3[1],mem[1]
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm3 = xmm3[0],xmm0[0]
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm14[1,3]
-; SSE-NEXT:    movaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm13, %xmm5
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm5 = xmm5[0],xmm3[0],xmm5[1],xmm3[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,1],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm13, %xmm0
-; SSE-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2],xmm3[2],xmm0[3],xmm3[3]
-; SSE-NEXT:    movdqa %xmm2, %xmm3
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm1[2],xmm3[3],xmm1[3]
-; SSE-NEXT:    punpcklqdq {{.*#+}} xmm3 = xmm3[0],xmm0[0]
-; SSE-NEXT:    movdqa %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm3, %xmm0
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    movlhps {{.*#+}} xmm0 = xmm0[0],xmm2[0]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, %xmm5
-; SSE-NEXT:    unpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm5 = xmm5[0],mem[0],xmm5[1],mem[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm1, %xmm0
-; SSE-NEXT:    shufps $197, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[1,1],mem[0,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, %xmm6
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm6 = xmm6[0],xmm4[0],xmm6[1],xmm4[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm2, %xmm6
+; SSE-NEXT:    movaps 96(%rdx), %xmm0
+; SSE-NEXT:    movaps 96(%rcx), %xmm7
+; SSE-NEXT:    movaps %xmm0, %xmm2
+; SSE-NEXT:    unpcklps {{.*#+}} xmm2 = xmm2[0],xmm7[0],xmm2[1],xmm7[1]
+; SSE-NEXT:    movaps 96(%rdi), %xmm4
+; SSE-NEXT:    movaps 96(%rsi), %xmm1
+; SSE-NEXT:    movaps %xmm4, %xmm3
+; SSE-NEXT:    unpcklps {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; SSE-NEXT:    movlhps {{.*#+}} xmm3 = xmm3[0],xmm2[0]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 96(%rax), %xmm2
+; SSE-NEXT:    movaps %xmm2, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,1],xmm4[1,3]
+; SSE-NEXT:    movaps 96(%r8), %xmm6
+; SSE-NEXT:    movaps 96(%r9), %xmm3
+; SSE-NEXT:    movaps %xmm6, %xmm8
+; SSE-NEXT:    unpcklps {{.*#+}} xmm8 = xmm8[0],xmm3[0],xmm8[1],xmm3[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,1],xmm5[0,2]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm6, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[1,1],xmm7[1,1]
+; SSE-NEXT:    movaps %xmm1, %xmm8
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[1,1],xmm0[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,2],xmm5[2,0]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm2, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,1],xmm4[3,3]
+; SSE-NEXT:    unpckhps {{.*#+}} xmm4 = xmm4[2],xmm1[2],xmm4[3],xmm1[3]
+; SSE-NEXT:    movaps %xmm3, %xmm8
+; SSE-NEXT:    unpcklps {{.*#+}} xmm8 = xmm8[0],xmm2[0],xmm8[1],xmm2[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[2,3],xmm4[0,1]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm0, %xmm4
+; SSE-NEXT:    unpckhps {{.*#+}} xmm4 = xmm4[2],xmm7[2],xmm4[3],xmm7[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[3,3],xmm6[3,3]
 ; SSE-NEXT:    unpckhps {{.*#+}} xmm6 = xmm6[2],xmm3[2],xmm6[3],xmm3[3]
-; SSE-NEXT:    movaps %xmm1, %xmm0
-; SSE-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
-; SSE-NEXT:    movlhps {{.*#+}} xmm6 = xmm6[0],xmm0[0]
-; SSE-NEXT:    movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm3, %xmm0
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, %xmm5
-; SSE-NEXT:    unpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm5 = xmm5[0],mem[0],xmm5[1],mem[1]
-; SSE-NEXT:    movlhps {{.*#+}} xmm5 = xmm5[0],xmm0[0]
-; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[1,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, %xmm6
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm6 = xmm6[0],xmm4[0],xmm6[1],xmm4[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,1],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm1, %xmm0
-; SSE-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
-; SSE-NEXT:    movaps %xmm3, %xmm1
-; SSE-NEXT:    unpckhps {{.*#+}} xmm1 = xmm1[2],xmm2[2],xmm1[3],xmm2[3]
-; SSE-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm3, %xmm0
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    movlhps {{.*#+}} xmm0 = xmm0[0],xmm2[0]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, %xmm5
-; SSE-NEXT:    unpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm5 = xmm5[0],mem[0],xmm5[1],mem[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,1],xmm0[2,0]
+; SSE-NEXT:    movlhps {{.*#+}} xmm4 = xmm4[0],xmm6[0]
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm1[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,2],xmm0[2,0]
 ; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm1, %xmm0
-; SSE-NEXT:    shufps $197, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[1,1],mem[0,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, %xmm6
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm6 = xmm6[0],xmm4[0],xmm6[1],xmm4[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm2, %xmm6
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[3,3],xmm3[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,2],xmm2[2,0]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 112(%rdx), %xmm0
+; SSE-NEXT:    movaps 112(%rcx), %xmm7
+; SSE-NEXT:    movaps %xmm7, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps 112(%rdi), %xmm4
+; SSE-NEXT:    movaps 112(%rsi), %xmm1
+; SSE-NEXT:    movaps %xmm4, %xmm3
+; SSE-NEXT:    unpcklps {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm2[2,0]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 112(%rax), %xmm2
+; SSE-NEXT:    movaps %xmm4, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[1,1],xmm2[0,3]
+; SSE-NEXT:    movaps 112(%r8), %xmm6
+; SSE-NEXT:    movaps 112(%r9), %xmm3
+; SSE-NEXT:    movaps %xmm6, %xmm8
+; SSE-NEXT:    unpcklps {{.*#+}} xmm8 = xmm8[0],xmm3[0],xmm8[1],xmm3[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,1],xmm5[2,0]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm6, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[1,1],xmm7[1,1]
+; SSE-NEXT:    movaps %xmm1, %xmm8
+; SSE-NEXT:    unpcklps {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[2,3],xmm5[2,0]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm2, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,1],xmm4[3,3]
+; SSE-NEXT:    unpckhps {{.*#+}} xmm4 = xmm4[2],xmm1[2],xmm4[3],xmm1[3]
+; SSE-NEXT:    movaps %xmm3, %xmm8
+; SSE-NEXT:    unpcklps {{.*#+}} xmm8 = xmm8[0],xmm2[0],xmm8[1],xmm2[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[2,3],xmm4[0,1]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm0, %xmm4
+; SSE-NEXT:    unpckhps {{.*#+}} xmm4 = xmm4[2],xmm7[2],xmm4[3],xmm7[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[3,3],xmm6[3,3]
 ; SSE-NEXT:    unpckhps {{.*#+}} xmm6 = xmm6[2],xmm3[2],xmm6[3],xmm3[3]
-; SSE-NEXT:    movaps %xmm1, %xmm0
-; SSE-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
-; SSE-NEXT:    movlhps {{.*#+}} xmm6 = xmm6[0],xmm0[0]
-; SSE-NEXT:    movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm3, %xmm0
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, %xmm5
-; SSE-NEXT:    unpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm5 = xmm5[0],mem[0],xmm5[1],mem[1]
-; SSE-NEXT:    movlhps {{.*#+}} xmm5 = xmm5[0],xmm0[0]
-; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[1,3]
-; SSE-NEXT:    movaps (%rsp), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, %xmm6
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm6 = xmm6[0],xmm4[0],xmm6[1],xmm4[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,1],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm1, %xmm0
-; SSE-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
-; SSE-NEXT:    movaps %xmm3, %xmm1
-; SSE-NEXT:    unpckhps {{.*#+}} xmm1 = xmm1[2],xmm2[2],xmm1[3],xmm2[3]
-; SSE-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm3, %xmm0
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    movlhps {{.*#+}} xmm0 = xmm0[0],xmm2[0]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, %xmm5
-; SSE-NEXT:    unpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm5 = xmm5[0],mem[0],xmm5[1],mem[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,1],xmm0[2,0]
+; SSE-NEXT:    movlhps {{.*#+}} xmm4 = xmm4[0],xmm6[0]
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm1[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,2],xmm0[2,0]
 ; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm1, %xmm0
-; SSE-NEXT:    shufps $197, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[1,1],mem[0,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, %xmm6
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm6 = xmm6[0],xmm4[0],xmm6[1],xmm4[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm2, %xmm6
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[3,3],xmm3[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,2],xmm2[2,0]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 128(%rdx), %xmm0
+; SSE-NEXT:    movaps 128(%rcx), %xmm7
+; SSE-NEXT:    movaps %xmm0, %xmm2
+; SSE-NEXT:    unpcklps {{.*#+}} xmm2 = xmm2[0],xmm7[0],xmm2[1],xmm7[1]
+; SSE-NEXT:    movaps 128(%rdi), %xmm4
+; SSE-NEXT:    movaps 128(%rsi), %xmm1
+; SSE-NEXT:    movaps %xmm4, %xmm3
+; SSE-NEXT:    unpcklps {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; SSE-NEXT:    movlhps {{.*#+}} xmm3 = xmm3[0],xmm2[0]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 128(%rax), %xmm2
+; SSE-NEXT:    movaps %xmm2, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,1],xmm4[1,3]
+; SSE-NEXT:    movaps 128(%r8), %xmm6
+; SSE-NEXT:    movaps 128(%r9), %xmm3
+; SSE-NEXT:    movaps %xmm6, %xmm8
+; SSE-NEXT:    unpcklps {{.*#+}} xmm8 = xmm8[0],xmm3[0],xmm8[1],xmm3[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,1],xmm5[0,2]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm6, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[1,1],xmm7[1,1]
+; SSE-NEXT:    movaps %xmm1, %xmm8
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[1,1],xmm0[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,2],xmm5[2,0]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm2, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,1],xmm4[3,3]
+; SSE-NEXT:    unpckhps {{.*#+}} xmm4 = xmm4[2],xmm1[2],xmm4[3],xmm1[3]
+; SSE-NEXT:    movaps %xmm3, %xmm8
+; SSE-NEXT:    unpcklps {{.*#+}} xmm8 = xmm8[0],xmm2[0],xmm8[1],xmm2[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[2,3],xmm4[0,1]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm0, %xmm4
+; SSE-NEXT:    unpckhps {{.*#+}} xmm4 = xmm4[2],xmm7[2],xmm4[3],xmm7[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[3,3],xmm6[3,3]
 ; SSE-NEXT:    unpckhps {{.*#+}} xmm6 = xmm6[2],xmm3[2],xmm6[3],xmm3[3]
-; SSE-NEXT:    movaps %xmm1, %xmm0
-; SSE-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
-; SSE-NEXT:    movlhps {{.*#+}} xmm6 = xmm6[0],xmm0[0]
-; SSE-NEXT:    movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm3, %xmm0
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, %xmm5
-; SSE-NEXT:    unpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm5 = xmm5[0],mem[0],xmm5[1],mem[1]
-; SSE-NEXT:    movlhps {{.*#+}} xmm5 = xmm5[0],xmm0[0]
-; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[1,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, %xmm6
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm6 = xmm6[0],xmm4[0],xmm6[1],xmm4[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,1],xmm0[0,2]
-; SSE-NEXT:    movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm1, %xmm0
-; SSE-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
-; SSE-NEXT:    movaps %xmm3, %xmm1
-; SSE-NEXT:    unpckhps {{.*#+}} xmm1 = xmm1[2],xmm2[2],xmm1[3],xmm2[3]
-; SSE-NEXT:    movlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm3, %xmm0
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    movlhps {{.*#+}} xmm0 = xmm0[0],xmm2[0]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, %xmm4
-; SSE-NEXT:    unpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm4 = xmm4[0],mem[0],xmm4[1],mem[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm0[2,0]
+; SSE-NEXT:    movlhps {{.*#+}} xmm4 = xmm4[0],xmm6[0]
 ; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm1, %xmm0
-; SSE-NEXT:    shufps $197, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[1,1],mem[0,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, %xmm4
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm4 = xmm4[0],xmm5[0],xmm4[1],xmm5[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,1],xmm0[2,0]
-; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm2, %xmm4
-; SSE-NEXT:    unpckhps {{.*#+}} xmm4 = xmm4[2],xmm3[2],xmm4[3],xmm3[3]
-; SSE-NEXT:    movaps %xmm1, %xmm0
-; SSE-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2],xmm5[2],xmm0[3],xmm5[3]
-; SSE-NEXT:    movlhps {{.*#+}} xmm4 = xmm4[0],xmm0[0]
-; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm15 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm15, %xmm0
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, %xmm4
-; SSE-NEXT:    unpcklps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm4 = xmm4[0],mem[0],xmm4[1],mem[1]
-; SSE-NEXT:    movlhps {{.*#+}} xmm4 = xmm4[0],xmm0[0]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm1[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,2],xmm0[2,0]
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[3,3],xmm3[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,2],xmm2[2,0]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 144(%rdx), %xmm0
+; SSE-NEXT:    movaps 144(%rcx), %xmm7
+; SSE-NEXT:    movaps %xmm7, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps 144(%rdi), %xmm4
+; SSE-NEXT:    movaps 144(%rsi), %xmm1
+; SSE-NEXT:    movaps %xmm4, %xmm3
+; SSE-NEXT:    unpcklps {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm2[2,0]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 144(%rax), %xmm2
+; SSE-NEXT:    movaps %xmm4, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[1,1],xmm2[0,3]
+; SSE-NEXT:    movaps 144(%r8), %xmm6
+; SSE-NEXT:    movaps 144(%r9), %xmm3
+; SSE-NEXT:    movaps %xmm6, %xmm8
+; SSE-NEXT:    unpcklps {{.*#+}} xmm8 = xmm8[0],xmm3[0],xmm8[1],xmm3[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,1],xmm5[2,0]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm6, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[1,1],xmm7[1,1]
+; SSE-NEXT:    movaps %xmm1, %xmm8
+; SSE-NEXT:    unpcklps {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[2,3],xmm5[2,0]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm2, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,1],xmm4[3,3]
+; SSE-NEXT:    unpckhps {{.*#+}} xmm4 = xmm4[2],xmm1[2],xmm4[3],xmm1[3]
+; SSE-NEXT:    movaps %xmm3, %xmm8
+; SSE-NEXT:    unpcklps {{.*#+}} xmm8 = xmm8[0],xmm2[0],xmm8[1],xmm2[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[2,3],xmm4[0,1]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm0, %xmm4
+; SSE-NEXT:    unpckhps {{.*#+}} xmm4 = xmm4[2],xmm7[2],xmm4[3],xmm7[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[3,3],xmm6[3,3]
+; SSE-NEXT:    unpckhps {{.*#+}} xmm6 = xmm6[2],xmm3[2],xmm6[3],xmm3[3]
+; SSE-NEXT:    movlhps {{.*#+}} xmm4 = xmm4[0],xmm6[0]
 ; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[1,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, %xmm5
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm5 = xmm5[0],xmm4[0],xmm5[1],xmm4[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,1],xmm0[0,2]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm1[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,2],xmm0[2,0]
 ; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm1, %xmm0
-; SSE-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
-; SSE-NEXT:    unpckhps {{.*#+}} xmm15 = xmm15[2],xmm2[2],xmm15[3],xmm2[3]
-; SSE-NEXT:    movlhps {{.*#+}} xmm15 = xmm15[0],xmm0[0]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm14 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm14, %xmm0
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm1[0,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,1],xmm14[3,3]
-; SSE-NEXT:    movaps %xmm1, %xmm4
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm2, %xmm13
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    unpckhps {{.*#+}} xmm13 = xmm13[2],xmm3[2],xmm13[3],xmm3[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[3,3],xmm3[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,2],xmm2[2,0]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 160(%rdx), %xmm0
+; SSE-NEXT:    movaps 160(%rcx), %xmm7
+; SSE-NEXT:    movaps %xmm0, %xmm2
+; SSE-NEXT:    unpcklps {{.*#+}} xmm2 = xmm2[0],xmm7[0],xmm2[1],xmm7[1]
+; SSE-NEXT:    movaps 160(%rdi), %xmm4
+; SSE-NEXT:    movaps 160(%rsi), %xmm1
+; SSE-NEXT:    movaps %xmm4, %xmm3
+; SSE-NEXT:    unpcklps {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
 ; SSE-NEXT:    movlhps {{.*#+}} xmm3 = xmm3[0],xmm2[0]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm14 = xmm14[0],xmm1[0],xmm14[1],xmm1[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[0,1],xmm3[2,0]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm3, %xmm10
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm10 = xmm10[0],xmm5[0],xmm10[1],xmm5[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm10 = xmm10[0,1],xmm0[2,0]
-; SSE-NEXT:    unpckhps {{.*#+}} xmm3 = xmm3[2],xmm5[2],xmm3[3],xmm5[3]
-; SSE-NEXT:    movlhps {{.*#+}} xmm13 = xmm13[0],xmm3[0]
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[3,3],xmm1[3,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[0,2],xmm2[2,0]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 160(%rax), %xmm2
+; SSE-NEXT:    movaps %xmm2, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,1],xmm4[1,3]
+; SSE-NEXT:    movaps 160(%r8), %xmm6
+; SSE-NEXT:    movaps 160(%r9), %xmm3
+; SSE-NEXT:    movaps %xmm6, %xmm8
+; SSE-NEXT:    unpcklps {{.*#+}} xmm8 = xmm8[0],xmm3[0],xmm8[1],xmm3[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,1],xmm5[0,2]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm6, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[1,1],xmm7[1,1]
+; SSE-NEXT:    movaps %xmm1, %xmm8
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[1,1],xmm0[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,2],xmm5[2,0]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm2, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,1],xmm4[3,3]
+; SSE-NEXT:    unpckhps {{.*#+}} xmm4 = xmm4[2],xmm1[2],xmm4[3],xmm1[3]
+; SSE-NEXT:    movaps %xmm3, %xmm8
+; SSE-NEXT:    unpcklps {{.*#+}} xmm8 = xmm8[0],xmm2[0],xmm8[1],xmm2[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[2,3],xmm4[0,1]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm0, %xmm4
+; SSE-NEXT:    unpckhps {{.*#+}} xmm4 = xmm4[2],xmm7[2],xmm4[3],xmm7[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[3,3],xmm6[3,3]
+; SSE-NEXT:    unpckhps {{.*#+}} xmm6 = xmm6[2],xmm3[2],xmm6[3],xmm3[3]
+; SSE-NEXT:    movlhps {{.*#+}} xmm4 = xmm4[0],xmm6[0]
 ; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm2, %xmm0
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1]
-; SSE-NEXT:    movaps 224(%rdi), %xmm6
-; SSE-NEXT:    movaps %xmm6, %xmm9
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm9 = xmm9[0],xmm3[0],xmm9[1],xmm3[1]
-; SSE-NEXT:    movlhps {{.*#+}} xmm9 = xmm9[0],xmm0[0]
-; SSE-NEXT:    movaps %xmm6, %xmm0
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm1[0,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm1[3,3]
 ; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,2],xmm0[2,0]
 ; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm6, %xmm0
-; SSE-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2],xmm3[2],xmm0[3],xmm3[3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[1,1],xmm1[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,2],xmm0[0,1]
-; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps %xmm2, %xmm7
-; SSE-NEXT:    unpckhps {{.*#+}} xmm7 = xmm7[2],xmm4[2],xmm7[3],xmm4[3]
-; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,1],xmm11[2,0]
 ; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[3,3],xmm3[3,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[3,1],xmm1[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[2,0],xmm2[2,0]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, %xmm11
-; SSE-NEXT:    unpckhps {{.*#+}} xmm11 = xmm11[2],xmm8[2],xmm11[3],xmm8[3]
-; SSE-NEXT:    movlhps {{.*#+}} xmm8 = xmm8[0],xmm0[0]
-; SSE-NEXT:    movaps %xmm0, %xmm4
-; SSE-NEXT:    movaps 240(%rdi), %xmm3
-; SSE-NEXT:    movaps %xmm3, %xmm5
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    unpcklps {{.*#+}} xmm5 = xmm5[0],xmm1[0],xmm5[1],xmm1[1]
-; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,1],xmm8[2,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,2],xmm2[2,0]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 176(%rdx), %xmm0
+; SSE-NEXT:    movaps 176(%rcx), %xmm7
+; SSE-NEXT:    movaps %xmm7, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps 176(%rdi), %xmm4
+; SSE-NEXT:    movaps 176(%rsi), %xmm1
+; SSE-NEXT:    movaps %xmm4, %xmm3
+; SSE-NEXT:    unpcklps {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm2[2,0]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 176(%rax), %xmm2
+; SSE-NEXT:    movaps %xmm4, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[1,1],xmm2[0,3]
+; SSE-NEXT:    movaps 176(%r8), %xmm6
+; SSE-NEXT:    movaps 176(%r9), %xmm3
+; SSE-NEXT:    movaps %xmm6, %xmm8
+; SSE-NEXT:    unpcklps {{.*#+}} xmm8 = xmm8[0],xmm3[0],xmm8[1],xmm3[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,1],xmm5[2,0]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm6, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[1,1],xmm7[1,1]
+; SSE-NEXT:    movaps %xmm1, %xmm8
+; SSE-NEXT:    unpcklps {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[2,3],xmm5[2,0]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm2, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,1],xmm4[3,3]
+; SSE-NEXT:    unpckhps {{.*#+}} xmm4 = xmm4[2],xmm1[2],xmm4[3],xmm1[3]
 ; SSE-NEXT:    movaps %xmm3, %xmm8
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[1,1],xmm2[0,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,2],xmm8[2,0]
-; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    unpcklps {{.*#+}} xmm8 = xmm8[0],xmm2[0],xmm8[1],xmm2[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[2,3],xmm4[0,1]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm0, %xmm4
+; SSE-NEXT:    unpckhps {{.*#+}} xmm4 = xmm4[2],xmm7[2],xmm4[3],xmm7[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[3,3],xmm6[3,3]
+; SSE-NEXT:    unpckhps {{.*#+}} xmm6 = xmm6[2],xmm3[2],xmm6[3],xmm3[3]
+; SSE-NEXT:    movlhps {{.*#+}} xmm4 = xmm4[0],xmm6[0]
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm1[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,2],xmm0[2,0]
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[3,3],xmm3[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,2],xmm2[2,0]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 192(%rdx), %xmm0
+; SSE-NEXT:    movaps 192(%rcx), %xmm7
+; SSE-NEXT:    movaps %xmm0, %xmm2
+; SSE-NEXT:    unpcklps {{.*#+}} xmm2 = xmm2[0],xmm7[0],xmm2[1],xmm7[1]
+; SSE-NEXT:    movaps 192(%rdi), %xmm4
+; SSE-NEXT:    movaps 192(%rsi), %xmm1
+; SSE-NEXT:    movaps %xmm4, %xmm3
+; SSE-NEXT:    unpcklps {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; SSE-NEXT:    movlhps {{.*#+}} xmm3 = xmm3[0],xmm2[0]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 192(%rax), %xmm2
+; SSE-NEXT:    movaps %xmm2, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,1],xmm4[1,3]
+; SSE-NEXT:    movaps 192(%r8), %xmm6
+; SSE-NEXT:    movaps 192(%r9), %xmm3
+; SSE-NEXT:    movaps %xmm6, %xmm8
+; SSE-NEXT:    unpcklps {{.*#+}} xmm8 = xmm8[0],xmm3[0],xmm8[1],xmm3[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,1],xmm5[0,2]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm6, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[1,1],xmm7[1,1]
+; SSE-NEXT:    movaps %xmm1, %xmm8
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[1,1],xmm0[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,2],xmm5[2,0]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm2, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,1],xmm4[3,3]
+; SSE-NEXT:    unpckhps {{.*#+}} xmm4 = xmm4[2],xmm1[2],xmm4[3],xmm1[3]
 ; SSE-NEXT:    movaps %xmm3, %xmm8
-; SSE-NEXT:    unpckhps {{.*#+}} xmm8 = xmm8[2],xmm1[2],xmm8[3],xmm1[3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1],xmm2[1,1]
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,2],xmm8[0,1]
-; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps {{.*#+}} xmm11 = xmm11[0,1],xmm12[2,0]
-; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[3,3],xmm1[3,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[3,1],xmm2[2,3]
-; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[2,0],xmm4[2,0]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    shufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = xmm1[3,3],mem[3,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,3],xmm1[2,0]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm12[2,3,2,3]
-; SSE-NEXT:    movss {{.*#+}} xmm0 = xmm4[0],xmm0[1,2,3]
-; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm12 = xmm12[3,3],mem[3,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,3],xmm12[2,0]
-; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm4 = mem[3,3,3,3]
-; SSE-NEXT:    movss {{.*#+}} xmm0 = xmm4[0],xmm0[1,2,3]
-; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[3,3],mem[3,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm0[2,0]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm0[2,3,2,3]
-; SSE-NEXT:    movss {{.*#+}} xmm1 = xmm4[0],xmm1[1,2,3]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[3,3],mem[3,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm0[2,0]
-; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm4 = mem[3,3,3,3]
-; SSE-NEXT:    movss {{.*#+}} xmm1 = xmm4[0],xmm1[1,2,3]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[3,3],mem[3,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm0[2,0]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm0[2,3,2,3]
-; SSE-NEXT:    movss {{.*#+}} xmm1 = xmm4[0],xmm1[1,2,3]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[3,3],mem[3,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm0[2,0]
-; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm4 = mem[3,3,3,3]
-; SSE-NEXT:    movss {{.*#+}} xmm1 = xmm4[0],xmm1[1,2,3]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[3,3],mem[3,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm0[2,0]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm0[2,3,2,3]
-; SSE-NEXT:    movss {{.*#+}} xmm1 = xmm4[0],xmm1[1,2,3]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[3,3],mem[3,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm0[2,0]
-; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm4 = mem[3,3,3,3]
-; SSE-NEXT:    movss {{.*#+}} xmm1 = xmm4[0],xmm1[1,2,3]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[3,3],mem[3,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,3],xmm0[2,0]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm0[2,3,2,3]
-; SSE-NEXT:    movss {{.*#+}} xmm2 = xmm4[0],xmm2[1,2,3]
-; SSE-NEXT:    shufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[3,3],mem[3,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm0[2,0]
-; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm4 = mem[3,3,3,3]
-; SSE-NEXT:    movss {{.*#+}} xmm1 = xmm4[0],xmm1[1,2,3]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[3,3],mem[3,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm0[2,0]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm0[2,3,2,3]
-; SSE-NEXT:    movss {{.*#+}} xmm1 = xmm4[0],xmm1[1,2,3]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[3,3],mem[3,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm0[2,0]
-; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm4 = mem[3,3,3,3]
-; SSE-NEXT:    movss {{.*#+}} xmm1 = xmm4[0],xmm1[1,2,3]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[3,3],mem[3,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm0[2,0]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm0[2,3,2,3]
-; SSE-NEXT:    movss {{.*#+}} xmm1 = xmm4[0],xmm1[1,2,3]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[3,3],mem[3,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm0[2,0]
-; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm4 = mem[3,3,3,3]
-; SSE-NEXT:    movss {{.*#+}} xmm1 = xmm4[0],xmm1[1,2,3]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[3,3],mem[3,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm0[2,0]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm0[2,3,2,3]
-; SSE-NEXT:    movss {{.*#+}} xmm1 = xmm4[0],xmm1[1,2,3]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[3,3],mem[3,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm0[2,0]
-; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm4 = mem[3,3,3,3]
-; SSE-NEXT:    movss {{.*#+}} xmm1 = xmm4[0],xmm1[1,2,3]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[3,3],mem[3,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm0[2,0]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm0[2,3,2,3]
-; SSE-NEXT:    movss {{.*#+}} xmm1 = xmm4[0],xmm1[1,2,3]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[3,3],mem[3,3]
-; SSE-NEXT:    movaps (%rsp), %xmm1 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm0[2,0]
-; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm4 = mem[3,3,3,3]
-; SSE-NEXT:    movss {{.*#+}} xmm1 = xmm4[0],xmm1[1,2,3]
-; SSE-NEXT:    movaps %xmm1, (%rsp) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[3,3],mem[3,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm0[2,0]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm0[2,3,2,3]
-; SSE-NEXT:    movss {{.*#+}} xmm1 = xmm4[0],xmm1[1,2,3]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[3,3],mem[3,3]
+; SSE-NEXT:    unpcklps {{.*#+}} xmm8 = xmm8[0],xmm2[0],xmm8[1],xmm2[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[2,3],xmm4[0,1]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movaps %xmm0, %xmm4
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,3],xmm4[2,0]
-; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm4 = mem[3,3,3,3]
-; SSE-NEXT:    movss {{.*#+}} xmm0 = xmm4[0],xmm0[1,2,3]
-; SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[3,3],mem[3,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm0[2,0]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm0[2,3,2,3]
-; SSE-NEXT:    movss {{.*#+}} xmm1 = xmm4[0],xmm1[1,2,3]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[3,3],mem[3,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm0[2,0]
-; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm4 = mem[3,3,3,3]
-; SSE-NEXT:    movss {{.*#+}} xmm1 = xmm4[0],xmm1[1,2,3]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[3,3],mem[3,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,3],xmm0[2,0]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm0[2,3,2,3]
-; SSE-NEXT:    movss {{.*#+}} xmm1 = xmm4[0],xmm1[1,2,3]
-; SSE-NEXT:    movaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; SSE-NEXT:    shufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[3,3],mem[3,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[0,3],xmm0[2,0]
-; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm4 = mem[3,3,3,3]
-; SSE-NEXT:    movss {{.*#+}} xmm12 = xmm4[0],xmm12[1,2,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm0 = xmm0[3,3],mem[3,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,3],xmm0[2,0]
-; SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm1[2,3,2,3]
-; SSE-NEXT:    movss {{.*#+}} xmm8 = xmm4[0],xmm8[1,2,3]
-; SSE-NEXT:    shufps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm1 = xmm1[3,3],mem[3,3]
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,3],xmm1[2,0]
-; SSE-NEXT:    pshufd $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Folded Reload
-; SSE-NEXT:    # xmm4 = mem[3,3,3,3]
-; SSE-NEXT:    movss {{.*#+}} xmm0 = xmm4[0],xmm0[1,2,3]
+; SSE-NEXT:    unpckhps {{.*#+}} xmm4 = xmm4[2],xmm7[2],xmm4[3],xmm7[3]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[3,3],xmm6[3,3]
+; SSE-NEXT:    unpckhps {{.*#+}} xmm6 = xmm6[2],xmm3[2],xmm6[3],xmm3[3]
+; SSE-NEXT:    movlhps {{.*#+}} xmm4 = xmm4[0],xmm6[0]
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm1[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,2],xmm0[2,0]
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[3,3],xmm3[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,2],xmm2[2,0]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 208(%rdx), %xmm0
+; SSE-NEXT:    movaps 208(%rcx), %xmm7
+; SSE-NEXT:    movaps %xmm7, %xmm2
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm0[0]
+; SSE-NEXT:    movaps 208(%rdi), %xmm4
+; SSE-NEXT:    movaps 208(%rsi), %xmm1
+; SSE-NEXT:    movaps %xmm4, %xmm3
+; SSE-NEXT:    unpcklps {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[0,1],xmm2[2,0]
+; SSE-NEXT:    movaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 208(%rax), %xmm2
+; SSE-NEXT:    movaps %xmm4, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[1,1],xmm2[0,3]
+; SSE-NEXT:    movaps 208(%r8), %xmm6
+; SSE-NEXT:    movaps 208(%r9), %xmm3
+; SSE-NEXT:    movaps %xmm6, %xmm8
+; SSE-NEXT:    unpcklps {{.*#+}} xmm8 = xmm8[0],xmm3[0],xmm8[1],xmm3[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[0,1],xmm5[2,0]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm6, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[1,1],xmm7[1,1]
+; SSE-NEXT:    movaps %xmm1, %xmm8
+; SSE-NEXT:    unpcklps {{.*#+}} xmm8 = xmm8[0],xmm0[0],xmm8[1],xmm0[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[2,3],xmm5[2,0]
+; SSE-NEXT:    movaps %xmm8, (%rsp) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm2, %xmm5
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,1],xmm4[3,3]
+; SSE-NEXT:    unpckhps {{.*#+}} xmm4 = xmm4[2],xmm1[2],xmm4[3],xmm1[3]
+; SSE-NEXT:    movaps %xmm3, %xmm8
+; SSE-NEXT:    unpcklps {{.*#+}} xmm8 = xmm8[0],xmm2[0],xmm8[1],xmm2[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm8 = xmm8[2,3],xmm4[0,1]
+; SSE-NEXT:    movaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; SSE-NEXT:    movaps %xmm0, %xmm4
+; SSE-NEXT:    unpckhps {{.*#+}} xmm4 = xmm4[2],xmm7[2],xmm4[3],xmm7[3]
+; SSE-NEXT:    unpckhps {{.*#+}} xmm7 = xmm7[2],xmm6[2],xmm7[3],xmm6[3]
+; SSE-NEXT:    unpckhps {{.*#+}} xmm6 = xmm6[2],xmm3[2],xmm6[3],xmm3[3]
+; SSE-NEXT:    movlhps {{.*#+}} xmm4 = xmm4[0],xmm6[0]
+; SSE-NEXT:    movaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3],xmm1[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[0,2],xmm0[2,0]
+; SSE-NEXT:    movaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[3,3],xmm3[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[2,3],xmm2[2,0]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 224(%rdx), %xmm5
+; SSE-NEXT:    movaps 224(%rcx), %xmm15
+; SSE-NEXT:    movaps %xmm5, %xmm1
+; SSE-NEXT:    unpcklps {{.*#+}} xmm1 = xmm1[0],xmm15[0],xmm1[1],xmm15[1]
+; SSE-NEXT:    movaps 224(%rdi), %xmm13
+; SSE-NEXT:    movaps 224(%rsi), %xmm0
+; SSE-NEXT:    movaps %xmm13, %xmm2
+; SSE-NEXT:    unpcklps {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; SSE-NEXT:    movlhps {{.*#+}} xmm2 = xmm2[0],xmm1[0]
+; SSE-NEXT:    movaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps 224(%rax), %xmm1
+; SSE-NEXT:    movaps %xmm13, %xmm6
+; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[1,1],xmm1[0,3]
+; SSE-NEXT:    movaps 224(%r8), %xmm10
+; SSE-NEXT:    movaps %xmm10, %xmm4
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[1,1],xmm15[1,1]
+; SSE-NEXT:    movaps 224(%r9), %xmm2
+; SSE-NEXT:    movaps %xmm2, %xmm3
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm3 = xmm3[1],xmm10[1]
+; SSE-NEXT:    movaps %xmm5, %xmm7
+; SSE-NEXT:    unpckhps {{.*#+}} xmm7 = xmm7[2],xmm15[2],xmm7[3],xmm15[3]
+; SSE-NEXT:    unpckhps {{.*#+}} xmm15 = xmm15[2],xmm10[2],xmm15[3],xmm10[3]
+; SSE-NEXT:    movlhps {{.*#+}} xmm10 = xmm10[0],xmm2[0]
+; SSE-NEXT:    shufps {{.*#+}} xmm10 = xmm10[0,2],xmm6[2,0]
+; SSE-NEXT:    movaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm0, %xmm6
+; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[1,1],xmm5[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,2],xmm4[2,0]
+; SSE-NEXT:    movaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    movaps %xmm13, %xmm4
+; SSE-NEXT:    unpckhps {{.*#+}} xmm4 = xmm4[2],xmm0[2],xmm4[3],xmm0[3]
+; SSE-NEXT:    movaps %xmm2, %xmm12
+; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[1,1],xmm1[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm12 = xmm12[0,2],xmm4[0,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,1],xmm3[2,0]
+; SSE-NEXT:    movaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[3,3],xmm0[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm13 = xmm13[3,1],xmm1[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm13 = xmm13[2,0],xmm5[2,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[3,3],xmm1[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm15 = xmm15[2,3],xmm2[0,2]
+; SSE-NEXT:    movaps 240(%rdx), %xmm4
+; SSE-NEXT:    movaps 240(%rcx), %xmm0
+; SSE-NEXT:    movaps %xmm0, %xmm3
+; SSE-NEXT:    movlhps {{.*#+}} xmm3 = xmm3[0],xmm4[0]
+; SSE-NEXT:    movaps 240(%rdi), %xmm5
+; SSE-NEXT:    movaps 240(%rsi), %xmm11
+; SSE-NEXT:    movaps %xmm5, %xmm6
+; SSE-NEXT:    unpcklps {{.*#+}} xmm6 = xmm6[0],xmm11[0],xmm6[1],xmm11[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,1],xmm3[2,0]
+; SSE-NEXT:    movaps 240(%rax), %xmm9
+; SSE-NEXT:    movaps %xmm5, %xmm14
+; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[1,1],xmm9[0,3]
+; SSE-NEXT:    movaps 240(%r8), %xmm7
+; SSE-NEXT:    movaps %xmm7, %xmm10
+; SSE-NEXT:    shufps {{.*#+}} xmm10 = xmm10[1,1],xmm0[1,1]
+; SSE-NEXT:    movaps 240(%r9), %xmm3
+; SSE-NEXT:    movaps %xmm3, %xmm8
+; SSE-NEXT:    unpckhpd {{.*#+}} xmm8 = xmm8[1],xmm7[1]
+; SSE-NEXT:    movaps %xmm4, %xmm2
+; SSE-NEXT:    unpckhps {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE-NEXT:    unpckhps {{.*#+}} xmm0 = xmm0[2],xmm7[2],xmm0[3],xmm7[3]
+; SSE-NEXT:    movlhps {{.*#+}} xmm7 = xmm7[0],xmm3[0]
+; SSE-NEXT:    shufps {{.*#+}} xmm7 = xmm7[0,2],xmm14[2,0]
+; SSE-NEXT:    movaps %xmm11, %xmm14
+; SSE-NEXT:    unpcklps {{.*#+}} xmm14 = xmm14[0],xmm4[0],xmm14[1],xmm4[1]
+; SSE-NEXT:    shufps {{.*#+}} xmm14 = xmm14[2,3],xmm10[2,0]
+; SSE-NEXT:    movaps %xmm5, %xmm10
+; SSE-NEXT:    unpckhps {{.*#+}} xmm10 = xmm10[2],xmm11[2],xmm10[3],xmm11[3]
+; SSE-NEXT:    movaps %xmm3, %xmm1
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,1],xmm9[1,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,2],xmm10[0,1]
+; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,1],xmm8[2,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm4 = xmm4[3,3],xmm11[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[3,1],xmm9[2,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm5 = xmm5[2,0],xmm4[2,0]
+; SSE-NEXT:    shufps {{.*#+}} xmm3 = xmm3[3,3],xmm9[3,3]
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,3],xmm3[0,2]
 ; SSE-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; SSE-NEXT:    movaps %xmm3, 1760(%rax)
-; SSE-NEXT:    movaps %xmm11, 1744(%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 1728(%rax)
+; SSE-NEXT:    movaps %xmm0, 1776(%rax)
+; SSE-NEXT:    movaps %xmm5, 1760(%rax)
+; SSE-NEXT:    movaps %xmm2, 1744(%rax)
+; SSE-NEXT:    movaps %xmm1, 1728(%rax)
+; SSE-NEXT:    movaps %xmm14, 1712(%rax)
+; SSE-NEXT:    movaps %xmm7, 1696(%rax)
+; SSE-NEXT:    movaps %xmm6, 1680(%rax)
+; SSE-NEXT:    movaps %xmm15, 1664(%rax)
+; SSE-NEXT:    movaps %xmm13, 1648(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 1696(%rax)
-; SSE-NEXT:    movaps %xmm5, 1680(%rax)
-; SSE-NEXT:    movaps %xmm6, 1648(%rax)
-; SSE-NEXT:    movaps %xmm7, 1632(%rax)
+; SSE-NEXT:    movaps %xmm0, 1632(%rax)
+; SSE-NEXT:    movaps %xmm12, 1616(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 1616(%rax)
+; SSE-NEXT:    movaps %xmm0, 1600(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 1584(%rax)
-; SSE-NEXT:    movaps %xmm9, 1568(%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 1536(%rax)
-; SSE-NEXT:    movaps %xmm13, 1520(%rax)
-; SSE-NEXT:    movaps %xmm10, 1472(%rax)
-; SSE-NEXT:    movaps %xmm14, 1456(%rax)
-; SSE-NEXT:    movaps %xmm15, 1408(%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 1360(%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 1344(%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 1296(%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 1248(%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 1232(%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 1184(%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 1136(%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 1120(%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 1072(%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 1024(%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 1008(%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 960(%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 912(%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 896(%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 848(%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 800(%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 784(%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 736(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 688(%rax)
+; SSE-NEXT:    movaps %xmm0, 1568(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 672(%rax)
+; SSE-NEXT:    movaps %xmm0, 1552(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 624(%rax)
+; SSE-NEXT:    movaps %xmm0, 1536(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 576(%rax)
+; SSE-NEXT:    movaps %xmm0, 1520(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 560(%rax)
+; SSE-NEXT:    movaps %xmm0, 1504(%rax)
+; SSE-NEXT:    movaps (%rsp), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 1488(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 512(%rax)
+; SSE-NEXT:    movaps %xmm0, 1472(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 464(%rax)
+; SSE-NEXT:    movaps %xmm0, 1456(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 448(%rax)
+; SSE-NEXT:    movaps %xmm0, 1440(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 400(%rax)
+; SSE-NEXT:    movaps %xmm0, 1424(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 352(%rax)
+; SSE-NEXT:    movaps %xmm0, 1408(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 336(%rax)
+; SSE-NEXT:    movaps %xmm0, 1392(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 288(%rax)
+; SSE-NEXT:    movaps %xmm0, 1376(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 240(%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 224(%rax)
+; SSE-NEXT:    movaps %xmm0, 1360(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 176(%rax)
+; SSE-NEXT:    movaps %xmm0, 1344(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 128(%rax)
+; SSE-NEXT:    movaps %xmm0, 1328(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 112(%rax)
+; SSE-NEXT:    movaps %xmm0, 1312(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 64(%rax)
+; SSE-NEXT:    movaps %xmm0, 1296(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 16(%rax)
+; SSE-NEXT:    movaps %xmm0, 1280(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, (%rax)
+; SSE-NEXT:    movaps %xmm0, 1264(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 1776(%rax)
+; SSE-NEXT:    movaps %xmm0, 1248(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 1712(%rax)
+; SSE-NEXT:    movaps %xmm0, 1232(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 1664(%rax)
+; SSE-NEXT:    movaps %xmm0, 1216(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 1600(%rax)
+; SSE-NEXT:    movaps %xmm0, 1200(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 1552(%rax)
+; SSE-NEXT:    movaps %xmm0, 1184(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 1504(%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, 1488(%rax)
-; SSE-NEXT:    movaps %xmm4, 1440(%rax)
-; SSE-NEXT:    movaps %xmm8, 1424(%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, 1392(%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, 1376(%rax)
-; SSE-NEXT:    movaps %xmm12, 1328(%rax)
+; SSE-NEXT:    movaps %xmm0, 1168(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 1312(%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, 1280(%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, 1264(%rax)
+; SSE-NEXT:    movaps %xmm0, 1152(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 1216(%rax)
+; SSE-NEXT:    movaps %xmm0, 1136(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm0, 1200(%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, 1168(%rax)
-; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; SSE-NEXT:    movaps %xmm1, 1152(%rax)
+; SSE-NEXT:    movaps %xmm0, 1120(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 1104(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 1088(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 1072(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 1056(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 1040(%rax)
-; SSE-NEXT:    movaps (%rsp), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 1024(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 1008(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 992(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 976(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 960(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 944(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 928(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 912(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 896(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 880(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 864(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 848(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 832(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 816(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 800(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 784(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 768(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 752(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 736(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 720(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 704(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 688(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 672(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 656(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 640(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 624(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 608(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 592(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 576(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 560(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 544(%rax)
-; SSE-NEXT:    movaps %xmm2, 528(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 528(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 512(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 496(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 480(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 464(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 448(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 432(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 416(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 400(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 384(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 368(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 352(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 336(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 320(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 304(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 288(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 272(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 256(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 240(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 224(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 208(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 192(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 176(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 160(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 144(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 128(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 112(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 96(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 80(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 64(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 48(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 32(%rax)
-; SSE-NEXT:    addq $2760, %rsp # imm = 0xAC8
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, 16(%rax)
+; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; SSE-NEXT:    movaps %xmm0, (%rax)
+; SSE-NEXT:    addq $1512, %rsp # imm = 0x5E8
 ; SSE-NEXT:    retq
 ;
 ; AVX1-ONLY-LABEL: store_i32_stride7_vf64:
 ; AVX1-ONLY:       # %bb.0:
 ; AVX1-ONLY-NEXT:    subq $3384, %rsp # imm = 0xD38
 ; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
 ; AVX1-ONLY-NEXT:    vmovaps 224(%rdi), %ymm3
 ; AVX1-ONLY-NEXT:    vmovaps 224(%rsi), %ymm2

llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-5.ll

 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, 16(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, (%r9)
 ; SSE-NEXT:    retq
 ;
 ; AVX1-ONLY-LABEL: store_i64_stride5_vf8:
 ; AVX1-ONLY:       # %bb.0:
-; AVX1-ONLY-NEXT:    vmovapd 32(%rdi), %ymm10
-; AVX1-ONLY-NEXT:    vmovaps (%rdi), %ymm1
-; AVX1-ONLY-NEXT:    vmovaps (%rcx), %ymm0
-; AVX1-ONLY-NEXT:    vmovaps (%r8), %ymm3
-; AVX1-ONLY-NEXT:    vmovapd 32(%r8), %ymm5
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vmovaps (%rdx), %xmm2
-; AVX1-ONLY-NEXT:    vmovaps 16(%rdx), %xmm8
-; AVX1-ONLY-NEXT:    vmovaps 32(%rdx), %xmm6
+; AVX1-ONLY-NEXT:    vmovapd (%rdi), %ymm5
+; AVX1-ONLY-NEXT:    vmovapd 32(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovapd (%rsi), %ymm9
+; AVX1-ONLY-NEXT:    vmovapd (%r8), %ymm6
+; AVX1-ONLY-NEXT:    vmovapd 32(%r8), %ymm3
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = mem[2,3],ymm2[2,3]
+; AVX1-ONLY-NEXT:    vmovaps (%rdx), %xmm1
+; AVX1-ONLY-NEXT:    vmovapd 16(%rdx), %xmm10
+; AVX1-ONLY-NEXT:    vmovaps 32(%rdx), %xmm7
 ; AVX1-ONLY-NEXT:    vmovapd 48(%rdx), %xmm11
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm8[1],ymm0[3],ymm8[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1],ymm0[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[6,7]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm4
-; AVX1-ONLY-NEXT:    vbroadcastsd 8(%rsi), %ymm7
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm1[0,1,2,3],ymm7[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm7[0,1,2,3,4,5],ymm4[6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm3[0,1],ymm4[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm7 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm9 = ymm10[0],mem[0],ymm10[2],mem[2]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm7 = ymm7[0,1],ymm9[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm7 = ymm7[0],ymm5[1],ymm7[2,3]
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm9 = mem[2,3],ymm10[2,3]
-; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm9 = ymm11[0],ymm9[0],ymm11[2],ymm9[3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm9 = ymm9[0,1],ymm5[2],ymm9[3]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm12
-; AVX1-ONLY-NEXT:    vbroadcastsd 40(%rsi), %ymm13
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm10 = ymm10[0,1],ymm13[2],ymm10[3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm10 = ymm10[0,1,2],ymm12[3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm10 = ymm5[0],ymm10[1,2,3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm12 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm13 = ymm1[0],mem[0],ymm1[2],mem[2]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3],ymm13[4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1],ymm3[2,3],ymm12[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovapd 48(%rsi), %xmm13
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm11 = xmm13[1],xmm11[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 56(%rcx), %ymm13
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm11 = ymm11[0,1],ymm13[2],ymm11[3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm5 = ymm11[0,1,2],ymm5[3]
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm8 = xmm8[0],mem[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm8[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm3 = xmm6[0],mem[0]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm11[0],ymm0[0],ymm11[2],ymm0[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm3[2],ymm0[3]
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm4 = mem[0,0]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm8 = ymm2[0],mem[0],ymm2[2],mem[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm4 = ymm4[0],ymm8[1,2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm4 = ymm4[0],ymm3[1],ymm4[2,3]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm8 = mem[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm8 = ymm9[2,3],ymm8[2,3]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm8 = ymm8[1],ymm10[1],ymm8[2],ymm10[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm8 = ymm8[0,1,2],ymm6[3]
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm12 = mem[0,0]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm9 = ymm5[0],ymm9[0],ymm5[2],ymm9[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm9 = ymm12[0],ymm9[1,2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm9 = ymm9[0],ymm6[1],ymm9[2,3]
+; AVX1-ONLY-NEXT:    vmovlpd {{.*#+}} xmm11 = mem[0],xmm11[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 56(%rcx), %ymm12
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm11 = ymm11[0,1],ymm12[2],ymm11[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm11 = ymm11[0,1,2],ymm3[3]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm10 = xmm10[0],mem[0]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm10 = ymm10[0,1,2],ymm5[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm10 = ymm10[0,1],ymm6[2],ymm10[3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm5, %ymm5
+; AVX1-ONLY-NEXT:    vbroadcastsd 8(%rsi), %ymm12
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm5 = ymm5[0,1],ymm12[2],ymm5[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm5 = ymm6[0],ymm5[1,2,3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm7, %ymm2, %ymm2
+; AVX1-ONLY-NEXT:    vbroadcastsd 40(%rsi), %ymm6
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm2[0,1],ymm6[2],ymm2[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm3[0],ymm2[1,2,3]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm3 = xmm7[0],mem[0]
 ; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm6
-; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm8
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm8 = xmm8[0],mem[0]
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm2 = xmm2[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm7
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm7 = xmm7[0],mem[0]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm1[0],mem[0]
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm6 = xmm6[0],mem[0]
 ; AVX1-ONLY-NEXT:    vmovaps %xmm6, (%r9)
-; AVX1-ONLY-NEXT:    vmovaps %xmm2, 16(%r9)
-; AVX1-ONLY-NEXT:    vmovaps %xmm8, 160(%r9)
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, 16(%r9)
+; AVX1-ONLY-NEXT:    vmovaps %xmm7, 160(%r9)
 ; AVX1-ONLY-NEXT:    vmovaps %xmm3, 176(%r9)
-; AVX1-ONLY-NEXT:    vmovaps %ymm12, 64(%r9)
-; AVX1-ONLY-NEXT:    vmovapd %ymm10, 192(%r9)
-; AVX1-ONLY-NEXT:    vmovapd %ymm9, 256(%r9)
-; AVX1-ONLY-NEXT:    vmovapd %ymm7, 224(%r9)
-; AVX1-ONLY-NEXT:    vmovaps %ymm4, 32(%r9)
-; AVX1-ONLY-NEXT:    vmovaps %ymm1, 96(%r9)
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 128(%r9)
-; AVX1-ONLY-NEXT:    vmovapd %ymm5, 288(%r9)
+; AVX1-ONLY-NEXT:    vmovapd %ymm9, 64(%r9)
+; AVX1-ONLY-NEXT:    vmovapd %ymm8, 128(%r9)
+; AVX1-ONLY-NEXT:    vmovapd %ymm4, 224(%r9)
+; AVX1-ONLY-NEXT:    vmovapd %ymm2, 192(%r9)
+; AVX1-ONLY-NEXT:    vmovapd %ymm0, 256(%r9)
+; AVX1-ONLY-NEXT:    vmovapd %ymm5, 32(%r9)
+; AVX1-ONLY-NEXT:    vmovapd %ymm10, 96(%r9)
+; AVX1-ONLY-NEXT:    vmovapd %ymm11, 288(%r9)
 ; AVX1-ONLY-NEXT:    vzeroupper
 ; AVX1-ONLY-NEXT:    retq
 ;
 ; AVX2-ONLY-LABEL: store_i64_stride5_vf8:
 ; AVX2-ONLY:       # %bb.0:
-; AVX2-ONLY-NEXT:    vmovaps 32(%rdi), %ymm7
-; AVX2-ONLY-NEXT:    vmovaps (%rdi), %ymm2
-; AVX2-ONLY-NEXT:    vmovaps (%rdx), %ymm4
-; AVX2-ONLY-NEXT:    vmovaps (%rcx), %ymm6
-; AVX2-ONLY-NEXT:    vmovaps (%r8), %ymm5
-; AVX2-ONLY-NEXT:    vmovaps 32(%r8), %ymm9
-; AVX2-ONLY-NEXT:    vmovaps (%rdx), %xmm1
-; AVX2-ONLY-NEXT:    vmovaps 32(%rdx), %xmm8
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX2-ONLY-NEXT:    vbroadcastsd 8(%rsi), %ymm3
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm2[0,1,2,3],ymm3[4,5],ymm2[6,7]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm3 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm10 = ymm7[0],mem[0],ymm7[2],mem[2]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm10[4,5,6,7]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm9[2,3],ymm3[4,5,6,7]
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm0, %ymm10
-; AVX2-ONLY-NEXT:    vbroadcastsd 40(%rsi), %ymm11
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm7[0,1,2,3],ymm11[4,5],ymm7[6,7]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm11[0,1,2,3,4,5],ymm10[6,7]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm9[0,1],ymm10[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm11 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm12 = ymm2[0],mem[0],ymm2[2],mem[2]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3],ymm12[4,5,6,7]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1],ymm5[2,3],ymm11[4,5,6,7]
-; AVX2-ONLY-NEXT:    vbroadcastsd 56(%rsi), %ymm12
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm12 = xmm12[0,1],mem[2,3]
-; AVX2-ONLY-NEXT:    vbroadcastsd 56(%rcx), %ymm13
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3],ymm13[4,5],ymm12[6,7]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3,4,5],ymm9[6,7]
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm13 = ymm4[1],ymm6[1],ymm4[3],ymm6[3]
-; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm13 = ymm13[0,2,3,3]
-; AVX2-ONLY-NEXT:    vbroadcastsd 24(%rsi), %ymm14
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm14[0,1],ymm13[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5],ymm5[6,7]
-; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm7 = mem[2,3],ymm7[2,3]
-; AVX2-ONLY-NEXT:    vbroadcastsd 48(%rcx), %ymm14
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm14[2,3],ymm7[4,5,6,7]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm9[4,5],ymm7[6,7]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm4[0],ymm6[0],ymm4[2],ymm6[2]
-; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm4[2,3],ymm2[2,3]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm5[4,5],ymm2[6,7]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm4 = xmm8[0],mem[0]
+; AVX2-ONLY-NEXT:    vmovaps 32(%rdi), %ymm10
+; AVX2-ONLY-NEXT:    vmovaps (%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovaps 32(%rsi), %ymm8
+; AVX2-ONLY-NEXT:    vmovaps 32(%rdx), %ymm9
+; AVX2-ONLY-NEXT:    vmovaps (%rdx), %ymm5
+; AVX2-ONLY-NEXT:    vmovaps (%rcx), %ymm7
+; AVX2-ONLY-NEXT:    vmovaps (%r8), %ymm6
+; AVX2-ONLY-NEXT:    vmovaps 32(%r8), %ymm11
+; AVX2-ONLY-NEXT:    vmovaps (%rdx), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps 32(%rdx), %xmm3
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX2-ONLY-NEXT:    vbroadcastsd 8(%rsi), %ymm4
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm1[0,1,2,3],ymm4[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm6[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm4
+; AVX2-ONLY-NEXT:    vbroadcastsd 40(%rsi), %ymm12
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm12 = ymm10[0,1,2,3],ymm12[4,5],ymm10[6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm12[0,1,2,3,4,5],ymm4[6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm11[0,1],ymm4[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm12 = mem[0,0]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm13 = ymm10[0],ymm8[0],ymm10[2],ymm8[2]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1],ymm13[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1],ymm11[2,3],ymm12[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm13 = mem[0,0]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm14 = ymm1[0],mem[0],ymm1[2],mem[2]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1],ymm14[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1],ymm6[2,3],ymm13[4,5,6,7]
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm14 = ymm5[1],ymm7[1],ymm5[3],ymm7[3]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm14 = ymm14[0,2,3,3]
+; AVX2-ONLY-NEXT:    vbroadcastsd 24(%rsi), %ymm15
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm15[0,1],ymm14[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3,4,5],ymm6[6,7]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm10 = ymm9[2,3],ymm10[2,3]
+; AVX2-ONLY-NEXT:    vbroadcastsd 48(%rcx), %ymm15
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1],ymm15[2,3],ymm10[4,5,6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3],ymm11[4,5],ymm10[6,7]
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm8 = ymm8[1],ymm9[1],ymm8[3],ymm9[3]
+; AVX2-ONLY-NEXT:    vbroadcastsd 56(%rcx), %ymm9
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm8 = ymm8[2,3],ymm9[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm11[6,7]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm5 = ymm5[0],ymm7[0],ymm5[2],ymm7[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm5[2,3],ymm1[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm6[4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vmovaps (%rdi), %xmm5
 ; AVX2-ONLY-NEXT:    vmovaps 32(%rdi), %xmm6
 ; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm6 = xmm6[0],mem[0]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm3 = xmm3[0],mem[0]
 ; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm5 = xmm5[0],mem[0]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm1[0],mem[0]
-; AVX2-ONLY-NEXT:    vmovaps %xmm1, 16(%r9)
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm2 = xmm2[0],mem[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, 16(%r9)
 ; AVX2-ONLY-NEXT:    vmovaps %xmm5, (%r9)
+; AVX2-ONLY-NEXT:    vmovaps %xmm3, 176(%r9)
 ; AVX2-ONLY-NEXT:    vmovaps %xmm6, 160(%r9)
-; AVX2-ONLY-NEXT:    vmovaps %xmm4, 176(%r9)
-; AVX2-ONLY-NEXT:    vmovaps %ymm11, 64(%r9)
-; AVX2-ONLY-NEXT:    vmovaps %ymm2, 96(%r9)
-; AVX2-ONLY-NEXT:    vmovaps %ymm10, 192(%r9)
-; AVX2-ONLY-NEXT:    vmovaps %ymm7, 256(%r9)
-; AVX2-ONLY-NEXT:    vmovaps %ymm3, 224(%r9)
+; AVX2-ONLY-NEXT:    vmovaps %ymm13, 64(%r9)
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 96(%r9)
+; AVX2-ONLY-NEXT:    vmovaps %ymm12, 224(%r9)
+; AVX2-ONLY-NEXT:    vmovaps %ymm4, 192(%r9)
+; AVX2-ONLY-NEXT:    vmovaps %ymm8, 288(%r9)
+; AVX2-ONLY-NEXT:    vmovaps %ymm10, 256(%r9)
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 32(%r9)
-; AVX2-ONLY-NEXT:    vmovaps %ymm13, 128(%r9)
-; AVX2-ONLY-NEXT:    vmovaps %ymm12, 288(%r9)
+; AVX2-ONLY-NEXT:    vmovaps %ymm14, 128(%r9)
 ; AVX2-ONLY-NEXT:    vzeroupper
 ; AVX2-ONLY-NEXT:    retq
 ;
 ; AVX512F-LABEL: store_i64_stride5_vf8:
 ; AVX512F:       # %bb.0:
 ; AVX512F-NEXT:    vmovdqa64 (%rdi), %zmm2
 ; AVX512F-NEXT:    vmovdqa64 (%rsi), %zmm3
 ; AVX512F-NEXT:    vmovdqa64 (%rdx), %zmm4

 ; SSE-NEXT:    movaps %xmm0, 16(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, (%r9)
 ; SSE-NEXT:    addq $280, %rsp # imm = 0x118
 ; SSE-NEXT:    retq
 ;
 ; AVX1-ONLY-LABEL: store_i64_stride5_vf16:
 ; AVX1-ONLY:       # %bb.0:
-; AVX1-ONLY-NEXT:    subq $216, %rsp
-; AVX1-ONLY-NEXT:    vmovapd 32(%rdi), %ymm6
-; AVX1-ONLY-NEXT:    vmovaps (%rdi), %ymm5
-; AVX1-ONLY-NEXT:    vmovapd 96(%rdi), %ymm8
-; AVX1-ONLY-NEXT:    vmovaps (%rcx), %ymm0
-; AVX1-ONLY-NEXT:    vmovaps 64(%rcx), %ymm1
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vmovaps (%rdx), %xmm2
-; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 16(%rdx), %xmm9
-; AVX1-ONLY-NEXT:    vmovaps 32(%rdx), %xmm3
-; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm9[1],ymm0[3],ymm9[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1],ymm0[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vmovaps 80(%rdx), %xmm13
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm13[1],ymm1[3],ymm13[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, (%rsp) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm8[0],mem[0],ymm8[2],mem[2]
-; AVX1-ONLY-NEXT:    vmovapd 96(%rcx), %xmm1
-; AVX1-ONLY-NEXT:    vmovapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vbroadcastsd 8(%rsi), %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm5[0,1,2,3],ymm1[4,5],ymm5[6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm5[0],mem[0],ymm5[2],mem[2]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vbroadcastsd 40(%rsi), %ymm1
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm6[0,1],ymm1[2],ymm6[3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm6[0],mem[0],ymm6[2],mem[2]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm14 = ymm0[0,1],ymm1[2,3]
-; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %ymm15
-; AVX1-ONLY-NEXT:    vbroadcastsd 72(%rsi), %ymm0
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3],ymm0[4,5],ymm15[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 64(%rdx), %xmm11
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm11, %ymm0, %ymm7
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm0[0,1,2,3,4,5],ymm7[6,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm10 = ymm15[0],mem[0],ymm15[2],mem[2]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm0[0,1,2,3],ymm10[4,5,6,7]
-; AVX1-ONLY-NEXT:    vbroadcastsd 104(%rsi), %ymm0
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm10 = ymm8[0,1],ymm0[2],ymm8[3]
-; AVX1-ONLY-NEXT:    vmovaps 96(%rdx), %xmm2
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm12
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm10[0,1,2],ymm12[3]
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm9 = xmm9[0],mem[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3,4,5],ymm5[6,7]
-; AVX1-ONLY-NEXT:    vmovaps (%r8), %ymm0
-; AVX1-ONLY-NEXT:    vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm1 = mem[0,1,2,3,4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendps $252, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm1 = ymm0[0,1],mem[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm1 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm9[0,1,2,3],ymm0[4,5],ymm9[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = mem[2,3],ymm6[2,3]
-; AVX1-ONLY-NEXT:    vmovapd 48(%rdx), %xmm5
-; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm5[0],ymm0[0],ymm5[2],ymm0[3]
-; AVX1-ONLY-NEXT:    vmovapd 48(%rsi), %xmm6
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm5 = xmm6[1],xmm5[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 56(%rcx), %ymm6
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm6 = ymm5[0,1],ymm6[2],ymm5[3]
-; AVX1-ONLY-NEXT:    vmovapd 32(%r8), %ymm0
-; AVX1-ONLY-NEXT:    vblendpd $14, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm5 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm5 = ymm0[0],mem[1,2,3]
+; AVX1-ONLY-NEXT:    subq $152, %rsp
+; AVX1-ONLY-NEXT:    vmovapd 96(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovapd 64(%rdi), %ymm9
+; AVX1-ONLY-NEXT:    vmovapd 32(%rdi), %ymm3
+; AVX1-ONLY-NEXT:    vmovapd (%rdi), %ymm4
+; AVX1-ONLY-NEXT:    vmovapd 64(%rsi), %ymm11
+; AVX1-ONLY-NEXT:    vmovapd (%rsi), %ymm6
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm5 = ymm4[0],ymm6[0],ymm4[2],ymm6[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm1[0],ymm5[1,2,3]
+; AVX1-ONLY-NEXT:    vmovaps (%rdx), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 48(%rdx), %xmm10
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm7 = mem[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm7 = ymm6[2,3],ymm7[2,3]
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm6 = mem[0,0]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm8 = ymm3[0],mem[0],ymm3[2],mem[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm5 = ymm6[0],ymm8[1,2,3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm9 = ymm14[0],ymm0[1],ymm14[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm14 = ymm1[0,1],ymm0[2],ymm1[3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm6[0,1,2],ymm0[3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm13[0],mem[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm15[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 64(%r8), %ymm13
-; AVX1-ONLY-NEXT:    vblendps $63, (%rsp), %ymm13, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm0 = mem[0,1,2,3,4,5],ymm13[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, (%rsp) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm15 = ymm13[0,1],ymm7[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm13[2,3],ymm4[4,5,6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm13[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = mem[2,3],ymm8[2,3]
-; AVX1-ONLY-NEXT:    vmovapd 112(%rdx), %xmm8
-; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm8[0],ymm1[0],ymm8[2],ymm1[3]
-; AVX1-ONLY-NEXT:    vmovapd 112(%rsi), %xmm13
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm8 = xmm13[1],xmm8[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 120(%rcx), %ymm13
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm8 = ymm8[0,1],ymm13[2],ymm8[3]
-; AVX1-ONLY-NEXT:    vmovapd 96(%r8), %ymm13
-; AVX1-ONLY-NEXT:    vblendpd $13, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm5 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm5 = mem[0],ymm13[1],mem[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm13[0],ymm3[1,2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm13[2],ymm1[3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm8 = ymm8[0,1,2],ymm13[3]
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm2 = xmm2[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %xmm13
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm13 = xmm13[0],mem[0]
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm11 = xmm11[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm12
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm12 = xmm12[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm8 = mem[2,3],ymm3[2,3]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm5 = ymm10[0],ymm8[0],ymm10[2],ymm8[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm8 = mem[0,0]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm12 = ymm9[0],ymm11[0],ymm9[2],ymm11[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm5 = ymm8[0],ymm12[1,2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm12 = mem[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm12 = ymm11[2,3],ymm12[2,3]
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm11 = mem[0,0]
+; AVX1-ONLY-NEXT:    vmovapd %ymm0, %ymm6
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm13 = ymm0[0],mem[0],ymm0[2],mem[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm11 = ymm11[0],ymm13[1,2,3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm13 = mem[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 112(%rdx), %xmm14
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm13 = ymm14[0],ymm13[0],ymm14[2],ymm13[3]
+; AVX1-ONLY-NEXT:    vmovlpd {{.*#+}} xmm10 = mem[0],xmm10[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 56(%rcx), %ymm15
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm15 = ymm10[0,1],ymm15[2],ymm10[3]
+; AVX1-ONLY-NEXT:    vmovlpd {{.*#+}} xmm10 = mem[0],xmm14[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 120(%rcx), %ymm14
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm5 = ymm10[0,1],ymm14[2],ymm10[3]
+; AVX1-ONLY-NEXT:    vmovapd 16(%rdx), %xmm10
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm7 = ymm7[1],ymm10[1],ymm7[2],ymm10[2]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm10 = xmm10[0],mem[0]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm14 = ymm10[0,1,2],ymm4[3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm4, %ymm4
+; AVX1-ONLY-NEXT:    vbroadcastsd 8(%rsi), %ymm10
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm4[0,1],ymm10[2],ymm4[3]
+; AVX1-ONLY-NEXT:    vmovapd (%r8), %ymm0
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm2[0],ymm0[1],ymm2[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm7[0,1,2],ymm0[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm14[0,1],ymm0[2],ymm14[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 32(%rdx), %xmm2
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm3, %ymm1
+; AVX1-ONLY-NEXT:    vbroadcastsd 40(%rsi), %ymm3
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm3[2],ymm1[3]
+; AVX1-ONLY-NEXT:    vmovapd 32(%r8), %ymm0
+; AVX1-ONLY-NEXT:    vblendpd $13, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm3 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm3 = mem[0],ymm0[1],mem[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd $11, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm14 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm14 = mem[0,1],ymm0[2],mem[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm15[0,1,2],ymm0[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 80(%rdx), %xmm0
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm12[1],ymm0[1],ymm12[2],ymm0[2]
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %xmm6
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm6 = xmm6[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm7
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm9[3]
+; AVX1-ONLY-NEXT:    vmovaps 64(%rdx), %xmm12
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm12, %ymm9, %ymm9
+; AVX1-ONLY-NEXT:    vbroadcastsd 72(%rsi), %ymm8
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm8 = ymm9[0,1],ymm8[2],ymm9[3]
+; AVX1-ONLY-NEXT:    vmovapd 64(%r8), %ymm9
+; AVX1-ONLY-NEXT:    vblendpd $13, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm3 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm3 = mem[0],ymm9[1],mem[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1,2],ymm9[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm9[2],ymm0[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm8 = ymm9[0],ymm8[1,2,3]
+; AVX1-ONLY-NEXT:    vmovaps 96(%rdx), %xmm9
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm9, %ymm6, %ymm0
+; AVX1-ONLY-NEXT:    vbroadcastsd 104(%rsi), %ymm7
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm7[2],ymm0[3]
+; AVX1-ONLY-NEXT:    vmovapd 96(%r8), %ymm7
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm11 = ymm11[0],ymm7[1],ymm11[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm13 = ymm13[0,1],ymm7[2],ymm13[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm5 = ymm5[0,1,2],ymm7[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm7[0],ymm0[1,2,3]
+; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %xmm7
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm7 = xmm7[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm12 = xmm12[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm10
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm10 = xmm10[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps %xmm10, 16(%r9)
-; AVX1-ONLY-NEXT:    vmovaps %xmm7, (%r9)
-; AVX1-ONLY-NEXT:    vmovaps %xmm2, 496(%r9)
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm2 = xmm2[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %xmm6
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm6 = xmm6[0],mem[0]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm9 = xmm9[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm15
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm15 = xmm15[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm4 = xmm4[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm4, 16(%r9)
+; AVX1-ONLY-NEXT:    vmovaps %xmm15, (%r9)
+; AVX1-ONLY-NEXT:    vmovaps %xmm9, 496(%r9)
 ; AVX1-ONLY-NEXT:    vmovaps %xmm6, 480(%r9)
-; AVX1-ONLY-NEXT:    vmovaps %xmm0, 176(%r9)
-; AVX1-ONLY-NEXT:    vmovaps %xmm12, 160(%r9)
-; AVX1-ONLY-NEXT:    vmovaps %xmm11, 336(%r9)
-; AVX1-ONLY-NEXT:    vmovaps %xmm13, 320(%r9)
-; AVX1-ONLY-NEXT:    vmovapd %ymm1, 576(%r9)
-; AVX1-ONLY-NEXT:    vmovapd %ymm3, 512(%r9)
-; AVX1-ONLY-NEXT:    vmovaps %ymm4, 384(%r9)
-; AVX1-ONLY-NEXT:    vmovaps %ymm15, 352(%r9)
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, 176(%r9)
+; AVX1-ONLY-NEXT:    vmovaps %xmm10, 160(%r9)
+; AVX1-ONLY-NEXT:    vmovaps %xmm12, 336(%r9)
+; AVX1-ONLY-NEXT:    vmovaps %xmm7, 320(%r9)
+; AVX1-ONLY-NEXT:    vmovapd %ymm13, 576(%r9)
+; AVX1-ONLY-NEXT:    vmovapd %ymm11, 544(%r9)
+; AVX1-ONLY-NEXT:    vmovapd %ymm0, 512(%r9)
+; AVX1-ONLY-NEXT:    vmovapd %ymm1, 448(%r9)
+; AVX1-ONLY-NEXT:    vmovapd %ymm3, 384(%r9)
+; AVX1-ONLY-NEXT:    vmovapd %ymm8, 352(%r9)
 ; AVX1-ONLY-NEXT:    vmovapd %ymm14, 256(%r9)
-; AVX1-ONLY-NEXT:    vmovapd %ymm9, 224(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 224(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 192(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 128(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 64(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 32(%r9)
-; AVX1-ONLY-NEXT:    vmovapd %ymm8, 608(%r9)
-; AVX1-ONLY-NEXT:    vmovapd %ymm5, 544(%r9)
-; AVX1-ONLY-NEXT:    vmovups (%rsp), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 448(%r9)
+; AVX1-ONLY-NEXT:    vmovapd %ymm5, 608(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 416(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 288(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 128(%r9)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 96(%r9)
-; AVX1-ONLY-NEXT:    addq $216, %rsp
+; AVX1-ONLY-NEXT:    addq $152, %rsp
 ; AVX1-ONLY-NEXT:    vzeroupper
 ; AVX1-ONLY-NEXT:    retq
 ;
 ; AVX2-ONLY-LABEL: store_i64_stride5_vf16:
 ; AVX2-ONLY:       # %bb.0:
-; AVX2-ONLY-NEXT:    subq $312, %rsp # imm = 0x138
-; AVX2-ONLY-NEXT:    vmovaps 96(%rdi), %ymm14
-; AVX2-ONLY-NEXT:    vmovaps 64(%rdi), %ymm11
-; AVX2-ONLY-NEXT:    vmovaps 32(%rdi), %ymm5
-; AVX2-ONLY-NEXT:    vmovaps (%rdi), %ymm6
+; AVX2-ONLY-NEXT:    subq $216, %rsp
+; AVX2-ONLY-NEXT:    vmovaps 96(%rdi), %ymm5
+; AVX2-ONLY-NEXT:    vmovaps 64(%rdi), %ymm4
+; AVX2-ONLY-NEXT:    vmovaps 32(%rdi), %ymm6
+; AVX2-ONLY-NEXT:    vmovaps (%rdi), %ymm7
+; AVX2-ONLY-NEXT:    vmovaps 32(%rsi), %ymm8
+; AVX2-ONLY-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovaps (%rdx), %xmm0
 ; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-ONLY-NEXT:    vmovaps 32(%rdx), %xmm1
 ; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-ONLY-NEXT:    vmovaps 64(%rdx), %xmm2
 ; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX2-ONLY-NEXT:    vbroadcastsd 8(%rsi), %ymm3
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm6[0,1,2,3],ymm3[4,5],ymm6[6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm7[0,1,2,3],ymm3[4,5],ymm7[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups %ymm0, (%rsp) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm6[0],mem[0],ymm6[2],mem[2]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm7[0],mem[0],ymm7[2],mem[2]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm3[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-ONLY-NEXT:    vbroadcastsd 40(%rsi), %ymm3
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0,1,2,3],ymm3[4,5],ymm5[6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm6[0,1,2,3],ymm3[4,5],ymm6[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm5[0],mem[0],ymm5[2],mem[2]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm6[0],ymm8[0],ymm6[2],ymm8[2]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm3[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
 ; AVX2-ONLY-NEXT:    vbroadcastsd 72(%rsi), %ymm3
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm11[0,1,2,3],ymm3[4,5],ymm11[6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3],ymm3[4,5],ymm4[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm11[0],mem[0],ymm11[2],mem[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 96(%rdx), %xmm9
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm9, %ymm0, %ymm0
-; AVX2-ONLY-NEXT:    vbroadcastsd 104(%rsi), %ymm3
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm14[0,1,2,3],ymm3[4,5],ymm14[6,7]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm14[0],mem[0],ymm14[2],mem[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:    vbroadcastsd 56(%rsi), %ymm0
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
-; AVX2-ONLY-NEXT:    vbroadcastsd 56(%rcx), %ymm3
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm0[0,1,2,3],ymm3[4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT:    vbroadcastsd 120(%rsi), %ymm3
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm3 = xmm3[0,1],mem[2,3]
-; AVX2-ONLY-NEXT:    vbroadcastsd 120(%rcx), %ymm15
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm3[0,1,2,3],ymm15[4,5],ymm3[6,7]
-; AVX2-ONLY-NEXT:    vmovaps (%rdx), %ymm3
-; AVX2-ONLY-NEXT:    vmovaps (%rcx), %ymm10
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm15 = ymm3[1],ymm10[1],ymm3[3],ymm10[3]
-; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[0,2,3,3]
-; AVX2-ONLY-NEXT:    vbroadcastsd 24(%rsi), %ymm8
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm15 = ymm8[0,1],ymm15[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovaps 64(%rdx), %ymm1
-; AVX2-ONLY-NEXT:    vmovaps 64(%rcx), %ymm2
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm8 = ymm1[1],ymm2[1],ymm1[3],ymm2[3]
-; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm8 = ymm8[0,2,3,3]
-; AVX2-ONLY-NEXT:    vbroadcastsd 88(%rsi), %ymm12
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm12[0,1],ymm8[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm3[0],ymm10[0],ymm3[2],ymm10[2]
-; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm3[2,3],ymm6[2,3]
-; AVX2-ONLY-NEXT:    vmovaps (%r8), %ymm6
-; AVX2-ONLY-NEXT:    vblendps $252, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm0 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm0 = ymm6[0,1],mem[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm0 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm0 = mem[0,1],ymm6[2,3],mem[4,5,6,7]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm4[0],mem[0],ymm4[2],mem[2]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm0[0,1],ymm3[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovaps 96(%rdx), %xmm10
+; AVX2-ONLY-NEXT:    vbroadcastsd 104(%rsi), %ymm0
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm0[4,5],ymm5[6,7]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm0, %ymm3
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm15 = ymm0[0,1,2,3,4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 96(%rsi), %ymm2
+; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm3 = mem[0,0]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm12 = ymm5[0],ymm2[0],ymm5[2],ymm2[2]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm12[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovaps (%rdx), %ymm12
+; AVX2-ONLY-NEXT:    vmovaps (%rcx), %ymm0
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm11 = ymm12[1],ymm0[1],ymm12[3],ymm0[3]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm11 = ymm11[0,2,3,3]
+; AVX2-ONLY-NEXT:    vbroadcastsd 24(%rsi), %ymm9
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1],ymm11[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovaps 64(%rdx), %ymm11
+; AVX2-ONLY-NEXT:    vmovaps 64(%rcx), %ymm1
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm14 = ymm11[1],ymm1[1],ymm11[3],ymm1[3]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm14 = ymm14[0,2,3,3]
+; AVX2-ONLY-NEXT:    vbroadcastsd 88(%rsi), %ymm13
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm13[0,1],ymm14[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm12[0],ymm0[0],ymm12[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm7 = ymm0[2,3],ymm7[2,3]
+; AVX2-ONLY-NEXT:    vmovaps (%r8), %ymm12
+; AVX2-ONLY-NEXT:    vblendps $252, (%rsp), %ymm12, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm12[0,1],mem[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3,4,5],ymm6[6,7]
+; AVX2-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = mem[0,1],ymm12[2,3],mem[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm6[4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm9[0,1,2,3,4,5],ymm12[6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, (%rsp) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm7[0,1,2,3],ymm12[4,5],ymm7[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = mem[2,3],ymm5[2,3]
-; AVX2-ONLY-NEXT:    vbroadcastsd 48(%rcx), %ymm5
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm5[2,3],ymm3[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovaps 32(%rdx), %ymm9
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm6 = ymm9[2,3],ymm6[2,3]
+; AVX2-ONLY-NEXT:    vbroadcastsd 48(%rcx), %ymm12
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm12 = ymm6[0,1],ymm12[2,3],ymm6[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm6 = ymm0[1],ymm9[1],ymm0[3],ymm9[3]
+; AVX2-ONLY-NEXT:    vbroadcastsd 56(%rcx), %ymm9
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm13 = ymm6[2,3],ymm9[2,3]
 ; AVX2-ONLY-NEXT:    vmovaps 32(%r8), %ymm0
-; AVX2-ONLY-NEXT:    vblendps $252, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm5 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm5 = ymm0[0,1],mem[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm5 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm5 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm13[0,1,2,3,4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm3[0,1,2,3],ymm0[4,5],ymm3[6,7]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
-; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm11[2,3]
+; AVX2-ONLY-NEXT:    vblendps $252, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm6 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm6 = ymm0[0,1],mem[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm6 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm6 = mem[0,1],ymm0[2,3],mem[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm12[0,1,2,3],ymm0[4,5],ymm12[6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm11[0],ymm1[0],ymm11[2],ymm1[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm4[2,3]
 ; AVX2-ONLY-NEXT:    vmovaps 64(%r8), %ymm1
 ; AVX2-ONLY-NEXT:    vblendps $252, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm11 # 32-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # ymm11 = ymm1[0,1],mem[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm6 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm6 = mem[0,1],ymm1[2,3],mem[4,5,6,7]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm4[0,1,2,3,4,5],ymm1[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = mem[2,3],ymm14[2,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm8[0,1],ymm1[2,3],ymm8[4,5,6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm14[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 96(%rdx), %ymm1
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm5 = ymm1[2,3],ymm5[2,3]
 ; AVX2-ONLY-NEXT:    vbroadcastsd 112(%rcx), %ymm14
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm1[0,1],ymm14[2,3],ymm1[4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovaps 96(%r8), %ymm1
-; AVX2-ONLY-NEXT:    vblendps $252, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm4 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm4 = ymm1[0,1],mem[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm3 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm3 = mem[0,1],ymm1[2,3],mem[4,5,6,7]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm1[6,7]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm14[0,1,2,3],ymm1[4,5],ymm14[6,7]
-; AVX2-ONLY-NEXT:    vmovaps 64(%rdi), %xmm14
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm14 = xmm14[0],mem[0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm14[2,3],ymm5[4,5,6,7]
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm2[1],ymm1[1],ymm2[3],ymm1[3]
+; AVX2-ONLY-NEXT:    vbroadcastsd 120(%rcx), %ymm2
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3],ymm2[2,3]
+; AVX2-ONLY-NEXT:    vmovaps 96(%r8), %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm2[0,1],ymm15[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm2[2,3],ymm3[4,5,6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm2[4,5],ymm5[6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 64(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm2 = xmm2[0],mem[0]
 ; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
-; AVX2-ONLY-NEXT:    vmovaps 32(%rdi), %xmm13
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm13 = xmm13[0],mem[0]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm15 = xmm0[0],mem[0]
+; AVX2-ONLY-NEXT:    vmovaps 32(%rdi), %xmm7
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm7 = xmm7[0],mem[0]
 ; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
-; AVX2-ONLY-NEXT:    vmovaps 96(%rdi), %xmm12
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm12 = xmm12[0],mem[0]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm9 = xmm9[0],mem[0]
-; AVX2-ONLY-NEXT:    vmovaps (%rdi), %xmm15
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm15 = xmm15[0],mem[0]
-; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps 96(%rdi), %xmm8
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm8 = xmm8[0],mem[0]
 ; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm10 = xmm10[0],mem[0]
-; AVX2-ONLY-NEXT:    vmovaps %xmm10, 16(%r9)
-; AVX2-ONLY-NEXT:    vmovaps %xmm15, (%r9)
-; AVX2-ONLY-NEXT:    vmovaps %xmm9, 496(%r9)
-; AVX2-ONLY-NEXT:    vmovaps %xmm12, 480(%r9)
+; AVX2-ONLY-NEXT:    vmovaps (%rdi), %xmm9
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm9 = xmm9[0],mem[0]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm12 = xmm12[0],mem[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm12, 16(%r9)
+; AVX2-ONLY-NEXT:    vmovaps %xmm9, (%r9)
+; AVX2-ONLY-NEXT:    vmovaps %xmm10, 496(%r9)
+; AVX2-ONLY-NEXT:    vmovaps %xmm8, 480(%r9)
 ; AVX2-ONLY-NEXT:    vmovaps %xmm0, 176(%r9)
-; AVX2-ONLY-NEXT:    vmovaps %xmm13, 160(%r9)
-; AVX2-ONLY-NEXT:    vmovaps %xmm1, 336(%r9)
-; AVX2-ONLY-NEXT:    vmovaps %xmm14, 320(%r9)
-; AVX2-ONLY-NEXT:    vmovaps %ymm2, 576(%r9)
+; AVX2-ONLY-NEXT:    vmovaps %xmm7, 160(%r9)
+; AVX2-ONLY-NEXT:    vmovaps %xmm15, 336(%r9)
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, 320(%r9)
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 608(%r9)
+; AVX2-ONLY-NEXT:    vmovaps %ymm5, 576(%r9)
 ; AVX2-ONLY-NEXT:    vmovaps %ymm3, 544(%r9)
-; AVX2-ONLY-NEXT:    vmovaps %ymm4, 512(%r9)
-; AVX2-ONLY-NEXT:    vmovaps %ymm5, 416(%r9)
+; AVX2-ONLY-NEXT:    vmovaps %ymm14, 512(%r9)
+; AVX2-ONLY-NEXT:    vmovaps %ymm4, 416(%r9)
 ; AVX2-ONLY-NEXT:    vmovaps %ymm6, 384(%r9)
 ; AVX2-ONLY-NEXT:    vmovaps %ymm11, 352(%r9)
-; AVX2-ONLY-NEXT:    vmovaps %ymm8, 256(%r9)
+; AVX2-ONLY-NEXT:    vmovaps %ymm13, 288(%r9)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 256(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 224(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 192(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 96(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 64(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 32(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 448(%r9)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovups (%rsp), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 128(%r9)
-; AVX2-ONLY-NEXT:    vmovaps %ymm7, 608(%r9)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 288(%r9)
-; AVX2-ONLY-NEXT:    addq $312, %rsp # imm = 0x138
+; AVX2-ONLY-NEXT:    addq $216, %rsp
 ; AVX2-ONLY-NEXT:    vzeroupper
 ; AVX2-ONLY-NEXT:    retq
 ;
 ; AVX512F-LABEL: store_i64_stride5_vf16:
 ; AVX512F:       # %bb.0:
 ; AVX512F-NEXT:    vmovdqa64 (%rdi), %zmm1
 ; AVX512F-NEXT:    vmovdqa64 64(%rdi), %zmm0
 ; AVX512F-NEXT:    vmovdqa64 (%rsi), %zmm2

 ; SSE-NEXT:    movaps %xmm0, 16(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, (%r9)
 ; SSE-NEXT:    addq $920, %rsp # imm = 0x398
 ; SSE-NEXT:    retq
 ;
 ; AVX1-ONLY-LABEL: store_i64_stride5_vf32:
 ; AVX1-ONLY:       # %bb.0:
-; AVX1-ONLY-NEXT:    subq $1032, %rsp # imm = 0x408
-; AVX1-ONLY-NEXT:    vmovaps 192(%rdi), %ymm4
-; AVX1-ONLY-NEXT:    vmovapd 160(%rdi), %ymm11
-; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %ymm6
-; AVX1-ONLY-NEXT:    vmovaps 128(%rcx), %ymm0
-; AVX1-ONLY-NEXT:    vmovaps (%rcx), %ymm1
-; AVX1-ONLY-NEXT:    vmovaps 64(%rcx), %ymm2
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vmovaps 16(%rdx), %xmm7
-; AVX1-ONLY-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm7[1],ymm1[3],ymm7[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vmovaps 80(%rdx), %xmm3
-; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm2 = ymm2[1],ymm3[1],ymm2[3],ymm3[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm6[0],mem[0],ymm6[2],mem[2]
-; AVX1-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 96(%rcx), %xmm2
-; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vmovaps 144(%rdx), %xmm2
-; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm2[1],ymm0[3],ymm2[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm11[0],mem[0],ymm11[2],mem[2]
-; AVX1-ONLY-NEXT:    vmovapd 160(%rcx), %xmm1
-; AVX1-ONLY-NEXT:    vmovapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm4[0],mem[0],ymm4[2],mem[2]
-; AVX1-ONLY-NEXT:    vmovaps 192(%rcx), %xmm1
-; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 192(%rcx), %ymm0
-; AVX1-ONLY-NEXT:    vmovaps 208(%rdx), %xmm1
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps (%rdi), %ymm9
-; AVX1-ONLY-NEXT:    vbroadcastsd 8(%rsi), %ymm0
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm9[0,1,2,3],ymm0[4,5],ymm9[6,7]
-; AVX1-ONLY-NEXT:    vmovaps (%rdx), %xmm1
-; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm9[0],mem[0],ymm9[2],mem[2]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    subq $936, %rsp # imm = 0x3A8
+; AVX1-ONLY-NEXT:    vmovapd 160(%rdi), %ymm5
+; AVX1-ONLY-NEXT:    vmovapd 128(%rdi), %ymm6
+; AVX1-ONLY-NEXT:    vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 96(%rdi), %ymm7
+; AVX1-ONLY-NEXT:    vmovapd 64(%rdi), %ymm8
+; AVX1-ONLY-NEXT:    vmovupd %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovapd 32(%rdi), %ymm14
-; AVX1-ONLY-NEXT:    vbroadcastsd 40(%rsi), %ymm0
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm14[0,1],ymm0[2],ymm14[3]
-; AVX1-ONLY-NEXT:    vmovaps 32(%rdx), %xmm1
-; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3]
+; AVX1-ONLY-NEXT:    vmovapd (%rdi), %ymm9
+; AVX1-ONLY-NEXT:    vmovapd 128(%rsi), %ymm0
+; AVX1-ONLY-NEXT:    vmovapd 64(%rsi), %ymm1
+; AVX1-ONLY-NEXT:    vmovapd (%rsi), %ymm2
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm3 = mem[0,0]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm9[0],ymm2[0],ymm9[2],ymm2[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm3[0],ymm4[1,2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm3 = mem[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm2[2,3],ymm3[2,3]
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm3 = mem[0,0]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm14[0],mem[0],ymm14[2],mem[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm3[0],ymm4[1,2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm3 = mem[0,0]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm8[0],ymm1[0],ymm8[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm3[0],ymm4[1,2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm3 = mem[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3],ymm3[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm7[0],mem[0],ymm7[2],mem[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0],ymm3[1,2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm6[0],ymm0[0],ymm6[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0],ymm3[1,2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm1 = mem[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm14[0],mem[0],ymm14[2],mem[2]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
+; AVX1-ONLY-NEXT:    vmovapd %ymm5, %ymm8
+; AVX1-ONLY-NEXT:    vmovupd %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm5[0],mem[0],ymm5[2],mem[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %ymm2
-; AVX1-ONLY-NEXT:    vbroadcastsd 72(%rsi), %ymm0
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 64(%rdx), %xmm1
-; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],mem[0],ymm2[2],mem[2]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, (%rsp) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vbroadcastsd 104(%rsi), %ymm0
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3],ymm0[4,5],ymm6[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 96(%rdx), %xmm1
-; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 128(%rdi), %ymm1
-; AVX1-ONLY-NEXT:    vbroadcastsd 136(%rsi), %ymm0
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 128(%rdx), %xmm3
-; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm1[0],mem[0],ymm1[2],mem[2]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vbroadcastsd 168(%rsi), %ymm0
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm11[0,1],ymm0[2],ymm11[3]
-; AVX1-ONLY-NEXT:    vmovaps 160(%rdx), %xmm6
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm3
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm12 = ymm0[0,1,2],ymm3[3]
-; AVX1-ONLY-NEXT:    vbroadcastsd 200(%rsi), %ymm0
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm0[4,5],ymm4[6,7]
-; AVX1-ONLY-NEXT:    vmovaps %ymm4, %ymm5
-; AVX1-ONLY-NEXT:    vmovaps 192(%rdx), %xmm3
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm4
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm0[0,1,2,3,4,5],ymm4[6,7]
-; AVX1-ONLY-NEXT:    vmovapd 224(%rdi), %ymm0
-; AVX1-ONLY-NEXT:    vbroadcastsd 232(%rsi), %ymm4
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm4 = ymm0[0,1],ymm4[2],ymm0[3]
-; AVX1-ONLY-NEXT:    vmovaps 224(%rdx), %xmm7
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm7, %ymm0, %ymm8
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm8 = ymm4[0,1,2],ymm8[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm15 = ymm0[0],mem[0],ymm0[2],mem[2]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm4 = ymm4[0,1],ymm15[2,3]
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm15 = xmm15[0],mem[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm15[0,1,2,3,4,5],ymm9[6,7]
-; AVX1-ONLY-NEXT:    vmovaps (%r8), %ymm15
-; AVX1-ONLY-NEXT:    vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm13 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm13 = mem[0,1,2,3,4,5],ymm15[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendps $252, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm13 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm13 = ymm15[0,1],mem[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm13 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm13 = mem[0,1],ymm15[2,3],mem[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3],ymm15[4,5],ymm9[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm9 = mem[2,3],ymm14[2,3]
-; AVX1-ONLY-NEXT:    vmovapd 48(%rdx), %xmm14
-; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm9 = ymm14[0],ymm9[0],ymm14[2],ymm9[3]
-; AVX1-ONLY-NEXT:    vmovapd 48(%rsi), %xmm15
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm14 = xmm15[1],xmm14[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 56(%rcx), %ymm15
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm14 = ymm14[0,1],ymm15[2],ymm14[3]
-; AVX1-ONLY-NEXT:    vmovapd 32(%r8), %ymm15
-; AVX1-ONLY-NEXT:    vblendpd $14, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm13 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm13 = ymm15[0],mem[1,2,3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendpd $13, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm13 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm13 = mem[0],ymm15[1],mem[2,3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 192(%rdi), %ymm12
+; AVX1-ONLY-NEXT:    vmovapd 192(%rsi), %ymm0
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm12[0],ymm0[0],ymm12[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0],ymm3[1,2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm1 = mem[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, (%rsp) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 224(%rdi), %ymm10
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm10[0],mem[0],ymm10[2],mem[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 48(%rdx), %xmm0
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = mem[2,3],ymm14[2,3]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm6 = ymm0[0],ymm1[0],ymm0[2],ymm1[3]
+; AVX1-ONLY-NEXT:    vmovlpd {{.*#+}} xmm0 = mem[0],xmm0[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 56(%rcx), %ymm4
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm0[0,1],ymm4[2],ymm0[3]
+; AVX1-ONLY-NEXT:    vmovapd %ymm7, %ymm3
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm4 = mem[2,3],ymm7[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 112(%rdx), %xmm5
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm7 = ymm5[0],ymm4[0],ymm5[2],ymm4[3]
+; AVX1-ONLY-NEXT:    vmovlpd {{.*#+}} xmm4 = mem[0],xmm5[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 120(%rcx), %ymm5
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm11 = ymm4[0,1],ymm5[2],ymm4[3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm4 = mem[2,3],ymm8[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 176(%rdx), %xmm5
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm5[0],ymm4[0],ymm5[2],ymm4[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovlpd {{.*#+}} xmm4 = mem[0],xmm5[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 184(%rcx), %ymm5
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm4[0,1],ymm5[2],ymm4[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm4 = mem[2,3],ymm10[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 240(%rdx), %xmm8
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm8[0],ymm4[0],ymm8[2],ymm4[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovlpd {{.*#+}} xmm4 = mem[0],xmm8[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 248(%rcx), %ymm8
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm4[0,1],ymm8[2],ymm4[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 16(%rdx), %xmm4
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm2 = ymm2[1],ymm4[1],ymm2[2],ymm4[2]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm4 = xmm4[0],mem[0]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm4[0,1,2],ymm9[3]
+; AVX1-ONLY-NEXT:    vmovaps (%rdx), %xmm13
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm13, %ymm9, %ymm9
+; AVX1-ONLY-NEXT:    vbroadcastsd 8(%rsi), %ymm15
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm9 = ymm9[0,1],ymm15[2],ymm9[3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm9 = ymm14[0,1,2],ymm15[3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm9 = xmm9[0],mem[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm9[0,1,2,3,4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 64(%r8), %ymm9
-; AVX1-ONLY-NEXT:    vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm13 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm13 = mem[0,1,2,3,4,5],ymm9[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendps $252, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm13 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm13 = ymm9[0,1],mem[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendps $243, (%rsp), %ymm9, %ymm13 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm13 = mem[0,1],ymm9[2,3],mem[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm13, (%rsp) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm9[4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm2 = mem[2,3],ymm2[2,3]
-; AVX1-ONLY-NEXT:    vmovapd 112(%rdx), %xmm9
-; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm2 = ymm9[0],ymm2[0],ymm9[2],ymm2[3]
-; AVX1-ONLY-NEXT:    vmovapd 112(%rsi), %xmm14
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm9 = xmm14[1],xmm9[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 120(%rcx), %ymm14
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm9 = ymm9[0,1],ymm14[2],ymm9[3]
-; AVX1-ONLY-NEXT:    vmovapd 96(%r8), %ymm14
-; AVX1-ONLY-NEXT:    vblendpd $13, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm13 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm13 = mem[0],ymm14[1],mem[2,3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendpd $14, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm13 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm13 = ymm14[0],mem[1,2,3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm2[0,1],ymm14[2],ymm2[3]
+; AVX1-ONLY-NEXT:    vmovapd (%r8), %ymm15
+; AVX1-ONLY-NEXT:    vblendpd $13, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm4 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm4 = mem[0],ymm15[1],mem[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm2[0,1,2],ymm15[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm9[0,1,2],ymm14[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm15[2],ymm0[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm15[0],ymm9[1,2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 32(%rdx), %xmm8
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm14, %ymm0
+; AVX1-ONLY-NEXT:    vbroadcastsd 40(%rsi), %ymm9
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm9[2],ymm0[3]
+; AVX1-ONLY-NEXT:    vmovapd 32(%r8), %ymm9
+; AVX1-ONLY-NEXT:    vblendpd $13, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm2 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm2 = mem[0],ymm9[1],mem[2,3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm2 = xmm2[0],mem[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 128(%r8), %ymm2
-; AVX1-ONLY-NEXT:    vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm9 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm9 = mem[0,1,2,3,4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendps $252, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm9 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm9 = ymm2[0,1],mem[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm9 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm9 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = mem[2,3],ymm11[2,3]
-; AVX1-ONLY-NEXT:    vmovapd 176(%rdx), %xmm2
-; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[3]
-; AVX1-ONLY-NEXT:    vmovapd 176(%rsi), %xmm14
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm2 = xmm14[1],xmm2[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 184(%rcx), %ymm14
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm2[0,1],ymm14[2],ymm2[3]
-; AVX1-ONLY-NEXT:    vmovapd 160(%r8), %ymm14
-; AVX1-ONLY-NEXT:    vblendpd $13, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm9 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm9 = mem[0],ymm14[1],mem[2,3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm9 = ymm14[0],ymm12[1,2,3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm14[2],ymm1[3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm2[0,1,2],ymm14[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm6[0,1],ymm9[2],ymm6[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1,2],ymm9[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm2 = xmm1[0],mem[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm5[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 192(%r8), %ymm14
-; AVX1-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm1 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm1 = mem[0,1],ymm14[2,3],mem[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm1 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm1 = mem[0,1,2,3,4,5],ymm14[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm14[0,1],ymm10[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm14[4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = mem[2,3],ymm0[2,3]
-; AVX1-ONLY-NEXT:    vmovapd 240(%rdx), %xmm2
-; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm2[0],ymm0[0],ymm2[2],ymm0[3]
-; AVX1-ONLY-NEXT:    vmovapd 240(%rsi), %xmm14
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm2 = xmm14[1],xmm2[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 248(%rcx), %ymm14
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm2[0,1],ymm14[2],ymm2[3]
-; AVX1-ONLY-NEXT:    vmovapd 224(%r8), %ymm14
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm14[0],ymm8[1,2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm9[0],ymm0[1,2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 80(%rdx), %xmm0
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm6 = ymm1[1],ymm0[1],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3]
+; AVX1-ONLY-NEXT:    vmovaps 64(%rdx), %xmm5
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm1, %ymm9
+; AVX1-ONLY-NEXT:    vbroadcastsd 72(%rsi), %ymm14
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm9 = ymm9[0,1],ymm14[2],ymm9[3]
+; AVX1-ONLY-NEXT:    vmovapd 64(%r8), %ymm14
+; AVX1-ONLY-NEXT:    vblendpd $13, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm1 = mem[0],ymm14[1],mem[2,3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm4[0],ymm14[1],ymm4[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm6[0,1,2],ymm14[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm14[2],ymm0[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm2[0,1,2],ymm14[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm14[0],ymm9[1,2,3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm13 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm13 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm6 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm6 = xmm6[0],mem[0]
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm3 = xmm3[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps 128(%rdi), %xmm14
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm14 = xmm14[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm12 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps 96(%rdx), %xmm4
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm3, %ymm6
+; AVX1-ONLY-NEXT:    vbroadcastsd 104(%rsi), %ymm9
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm6 = ymm6[0,1],ymm9[2],ymm6[3]
+; AVX1-ONLY-NEXT:    vmovapd 96(%r8), %ymm9
+; AVX1-ONLY-NEXT:    vblendpd $13, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = mem[0],ymm9[1],mem[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm7[0,1],ymm9[2],ymm7[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm11[0,1,2],ymm9[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm9[0],ymm6[1,2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 144(%rdx), %xmm6
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm7 = ymm0[1],ymm6[1],ymm0[2],ymm6[2]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm6 = xmm6[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm6 = ymm6[0,1,2],ymm0[3]
+; AVX1-ONLY-NEXT:    vmovaps 128(%rdx), %xmm1
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm9
+; AVX1-ONLY-NEXT:    vbroadcastsd 136(%rsi), %ymm11
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm9 = ymm9[0,1],ymm11[2],ymm9[3]
+; AVX1-ONLY-NEXT:    vmovapd 128(%r8), %ymm11
+; AVX1-ONLY-NEXT:    vblendpd $13, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = mem[0],ymm11[1],mem[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm7[0,1,2],ymm11[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm6[0,1],ymm11[2],ymm6[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm11[0],ymm9[1,2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 160(%rdx), %xmm3
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm6
+; AVX1-ONLY-NEXT:    vbroadcastsd 168(%rsi), %ymm7
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5],ymm6[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 160(%r8), %ymm7
+; AVX1-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = mem[0,1],ymm7[2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = mem[0,1,2,3],ymm7[4,5],mem[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = mem[0,1,2,3,4,5],ymm7[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm7[0,1],ymm6[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 208(%rdx), %xmm6
+; AVX1-ONLY-NEXT:    vmovupd (%rsp), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm7 = ymm0[1],ymm6[1],ymm0[2],ymm6[2]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm6 = xmm6[0],mem[0]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm6 = ymm6[0,1,2],ymm12[3]
+; AVX1-ONLY-NEXT:    vmovaps 192(%rdx), %xmm2
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm12, %ymm9
+; AVX1-ONLY-NEXT:    vbroadcastsd 200(%rsi), %ymm11
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm9 = ymm9[0,1],ymm11[2],ymm9[3]
+; AVX1-ONLY-NEXT:    vmovapd 192(%r8), %ymm12
+; AVX1-ONLY-NEXT:    vblendpd $13, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = mem[0],ymm12[1],mem[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm7[0,1,2],ymm12[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm6[0,1],ymm12[2],ymm6[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm12[0],ymm9[1,2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, (%rsp) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 224(%rdx), %xmm0
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm10, %ymm6
+; AVX1-ONLY-NEXT:    vbroadcastsd 232(%rsi), %ymm7
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm6[0,1,2,3],ymm7[4,5],ymm6[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 224(%r8), %ymm10
+; AVX1-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm6 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm6 = mem[0,1],ymm10[2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm6 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm6 = mem[0,1,2,3],ymm10[4,5],mem[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm6 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm6 = mem[0,1,2,3,4,5],ymm10[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm10[0,1],ymm7[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 128(%rdi), %xmm10
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm10 = xmm10[0],mem[0]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm1[0],mem[0]
 ; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %xmm15
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm15 = xmm15[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm11
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm5 = xmm5[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm14
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm14 = xmm14[0],mem[0]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm8 = xmm8[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %xmm12
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm12 = xmm12[0],mem[0]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm4 = xmm4[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps 160(%rdi), %xmm11
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm11 = xmm11[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %xmm2
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm2 = xmm2[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps 160(%rdi), %xmm10
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm10 = xmm10[0],mem[0]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm3 = xmm3[0],mem[0]
 ; AVX1-ONLY-NEXT:    vmovaps 224(%rdi), %xmm9
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm9 = xmm9[0],mem[0]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps 192(%rdi), %xmm7
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm7 = xmm7[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps 192(%rdi), %xmm8
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm8 = xmm8[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm5
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm5 = xmm5[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm4 = xmm4[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps %xmm4, 16(%r9)
-; AVX1-ONLY-NEXT:    vmovaps %xmm5, (%r9)
-; AVX1-ONLY-NEXT:    vmovaps %xmm3, 976(%r9)
-; AVX1-ONLY-NEXT:    vmovaps %xmm8, 960(%r9)
-; AVX1-ONLY-NEXT:    vmovaps %xmm7, 1136(%r9)
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm2 = xmm2[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm6
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm6 = xmm6[0],mem[0]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm13 = xmm13[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm13, 16(%r9)
+; AVX1-ONLY-NEXT:    vmovaps %xmm6, (%r9)
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, 976(%r9)
+; AVX1-ONLY-NEXT:    vmovaps %xmm7, 960(%r9)
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, 1136(%r9)
 ; AVX1-ONLY-NEXT:    vmovaps %xmm9, 1120(%r9)
-; AVX1-ONLY-NEXT:    vmovaps %xmm6, 816(%r9)
-; AVX1-ONLY-NEXT:    vmovaps %xmm10, 800(%r9)
-; AVX1-ONLY-NEXT:    vmovaps %xmm13, 496(%r9)
-; AVX1-ONLY-NEXT:    vmovaps %xmm2, 480(%r9)
-; AVX1-ONLY-NEXT:    vmovaps %xmm0, 176(%r9)
-; AVX1-ONLY-NEXT:    vmovaps %xmm11, 160(%r9)
-; AVX1-ONLY-NEXT:    vmovaps %xmm1, 336(%r9)
+; AVX1-ONLY-NEXT:    vmovaps %xmm3, 816(%r9)
+; AVX1-ONLY-NEXT:    vmovaps %xmm11, 800(%r9)
+; AVX1-ONLY-NEXT:    vmovaps %xmm4, 496(%r9)
+; AVX1-ONLY-NEXT:    vmovaps %xmm12, 480(%r9)
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 176(%r9)
+; AVX1-ONLY-NEXT:    vmovaps %xmm14, 160(%r9)
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 336(%r9)
 ; AVX1-ONLY-NEXT:    vmovaps %xmm15, 320(%r9)
-; AVX1-ONLY-NEXT:    vmovaps %xmm12, 656(%r9)
-; AVX1-ONLY-NEXT:    vmovaps %xmm14, 640(%r9)
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, 656(%r9)
+; AVX1-ONLY-NEXT:    vmovaps %xmm10, 640(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1216(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1184(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1152(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1088(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1024(%r9)
+; AVX1-ONLY-NEXT:    vmovups (%rsp), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 992(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 896(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 864(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 832(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 768(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 704(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 672(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 576(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 544(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 512(%r9)
-; AVX1-ONLY-NEXT:    vmovups (%rsp), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 448(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 384(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 352(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 256(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 224(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 192(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 128(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 64(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 32(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1248(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1088(%r9)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1056(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1024(%r9)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 928(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 864(%r9)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 768(%r9)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 736(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 608(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 544(%r9)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 448(%r9)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 416(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 288(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 128(%r9)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 96(%r9)
-; AVX1-ONLY-NEXT:    addq $1032, %rsp # imm = 0x408
+; AVX1-ONLY-NEXT:    addq $936, %rsp # imm = 0x3A8
 ; AVX1-ONLY-NEXT:    vzeroupper
 ; AVX1-ONLY-NEXT:    retq
 ;
 ; AVX2-ONLY-LABEL: store_i64_stride5_vf32:
 ; AVX2-ONLY:       # %bb.0:
 ; AVX2-ONLY-NEXT:    subq $1032, %rsp # imm = 0x408
 ; AVX2-ONLY-NEXT:    vmovaps 96(%rdi), %ymm4
 ; AVX2-ONLY-NEXT:    vmovaps 64(%rdi), %ymm5
-; AVX2-ONLY-NEXT:    vmovaps 32(%rdi), %ymm6
-; AVX2-ONLY-NEXT:    vmovaps (%rdi), %ymm7
-; AVX2-ONLY-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 32(%rdi), %ymm7
+; AVX2-ONLY-NEXT:    vmovaps (%rdi), %ymm8
+; AVX2-ONLY-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 32(%rsi), %ymm6
+; AVX2-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovaps (%rdx), %xmm0
 ; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-ONLY-NEXT:    vmovaps 32(%rdx), %xmm2
 ; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-ONLY-NEXT:    vmovaps 64(%rdx), %xmm3
 ; AVX2-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX2-ONLY-NEXT:    vbroadcastsd 8(%rsi), %ymm1
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm7[0,1,2,3],ymm1[4,5],ymm7[6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm8[0,1,2,3],ymm1[4,5],ymm8[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm7[0],mem[0],ymm7[2],mem[2]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm8[0],mem[0],ymm8[2],mem[2]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
 ; AVX2-ONLY-NEXT:    vbroadcastsd 40(%rsi), %ymm1
-; AVX2-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm6[0,1,2,3],ymm1[4,5],ymm6[6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm7[0,1,2,3],ymm1[4,5],ymm7[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm6[0],mem[0],ymm6[2],mem[2]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm7[0],ymm6[0],ymm7[2],ymm6[2]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm0
 ; AVX2-ONLY-NEXT:    vbroadcastsd 72(%rsi), %ymm1
 ; AVX2-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm5[0,1,2,3],ymm1[4,5],ymm5[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
 ; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm5[0],mem[0],ymm5[2],mem[2]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 96(%rdx), %xmm0
-; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-ONLY-NEXT:    vbroadcastsd 104(%rsi), %ymm1
+; AVX2-ONLY-NEXT:    vmovaps 96(%rdx), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastsd 104(%rsi), %ymm0
 ; AVX2-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm4[0,1,2,3],ymm1[4,5],ymm4[6,7]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm0[4,5],ymm4[6,7]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 96(%rsi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm4[0],mem[0],ymm4[2],mem[2]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm4[0],ymm1[0],ymm4[2],ymm1[2]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-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:    vbroadcastsd 136(%rsi), %ymm0
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5],ymm2[6,7]
 ; AVX2-ONLY-NEXT:    vmovaps 128(%rdx), %xmm1
 ; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
 ; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],mem[0],ymm2[2],mem[2]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 160(%rdi), %ymm9
+; AVX2-ONLY-NEXT:    vmovaps 160(%rdi), %ymm12
 ; AVX2-ONLY-NEXT:    vbroadcastsd 168(%rsi), %ymm0
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm9[0,1,2,3],ymm0[4,5],ymm9[6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm12[0,1,2,3],ymm0[4,5],ymm12[6,7]
 ; AVX2-ONLY-NEXT:    vmovaps 160(%rdx), %xmm1
 ; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 160(%rsi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm9[0],mem[0],ymm9[2],mem[2]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm12[0],ymm1[0],ymm12[2],ymm1[2]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 192(%rdi), %ymm5
+; AVX2-ONLY-NEXT:    vmovaps 192(%rdi), %ymm11
 ; AVX2-ONLY-NEXT:    vbroadcastsd 200(%rsi), %ymm0
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm0[4,5],ymm5[6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm11[0,1,2,3],ymm0[4,5],ymm11[6,7]
 ; AVX2-ONLY-NEXT:    vmovaps 192(%rdx), %xmm1
 ; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm4
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm4[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, (%rsp) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm5[0],mem[0],ymm5[2],mem[2]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm4[4,5,6,7]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm11[0],mem[0],ymm11[2],mem[2]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 224(%rdi), %ymm3
-; AVX2-ONLY-NEXT:    vbroadcastsd 232(%rsi), %ymm4
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm3[0,1,2,3],ymm4[4,5],ymm3[6,7]
-; AVX2-ONLY-NEXT:    vmovaps 224(%rdx), %xmm0
-; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm6
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3,4,5],ymm6[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 224(%rdi), %ymm9
+; AVX2-ONLY-NEXT:    vbroadcastsd 232(%rsi), %ymm0
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm9[0,1,2,3],ymm0[4,5],ymm9[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 224(%rdx), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm4 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm6 = ymm3[0],mem[0],ymm3[2],mem[2]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[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:    vbroadcastsd 56(%rsi), %ymm4
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm4[0,1],mem[2,3]
-; AVX2-ONLY-NEXT:    vbroadcastsd 56(%rcx), %ymm6
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm6[4,5],ymm4[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm0, (%rsp) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vbroadcastsd 120(%rsi), %ymm4
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm4[0,1],mem[2,3]
-; AVX2-ONLY-NEXT:    vbroadcastsd 120(%rcx), %ymm6
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm6[4,5],ymm4[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vbroadcastsd 184(%rsi), %ymm4
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm4 = xmm4[0,1],mem[2,3]
-; AVX2-ONLY-NEXT:    vbroadcastsd 184(%rcx), %ymm6
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm6[4,5],ymm4[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vbroadcastsd 248(%rsi), %ymm6
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = xmm6[0,1],mem[2,3]
-; AVX2-ONLY-NEXT:    vbroadcastsd 248(%rcx), %ymm13
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3],ymm13[4,5],ymm6[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps (%rdx), %ymm13
-; AVX2-ONLY-NEXT:    vmovaps (%rcx), %ymm14
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm15 = ymm13[1],ymm14[1],ymm13[3],ymm14[3]
-; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[0,2,3,3]
-; AVX2-ONLY-NEXT:    vbroadcastsd 24(%rsi), %ymm7
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm15[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovaps 64(%rdx), %ymm15
-; AVX2-ONLY-NEXT:    vmovaps 64(%rcx), %ymm0
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm8 = ymm15[1],ymm0[1],ymm15[3],ymm0[3]
-; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm8 = ymm8[0,2,3,3]
-; AVX2-ONLY-NEXT:    vbroadcastsd 88(%rsi), %ymm12
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm12[0,1],ymm8[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovaps 128(%rdx), %ymm12
-; AVX2-ONLY-NEXT:    vmovaps 128(%rcx), %ymm1
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm4 = ymm12[1],ymm1[1],ymm12[3],ymm1[3]
+; AVX2-ONLY-NEXT:    vmovaps 224(%rsi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm9[0],ymm1[0],ymm9[2],ymm1[2]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps (%rdx), %ymm1
+; AVX2-ONLY-NEXT:    vmovaps (%rcx), %ymm0
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm2 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,2,3,3]
+; AVX2-ONLY-NEXT:    vbroadcastsd 24(%rsi), %ymm3
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm3[0,1],ymm2[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovaps 64(%rdx), %ymm3
+; AVX2-ONLY-NEXT:    vmovaps 64(%rcx), %ymm2
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm4 = ymm3[1],ymm2[1],ymm3[3],ymm2[3]
 ; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[0,2,3,3]
-; AVX2-ONLY-NEXT:    vbroadcastsd 152(%rsi), %ymm6
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm6[0,1],ymm4[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovaps 192(%rdx), %ymm6
-; AVX2-ONLY-NEXT:    vmovaps 192(%rcx), %ymm2
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm11 = ymm6[1],ymm2[1],ymm6[3],ymm2[3]
-; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm11 = ymm11[0,2,3,3]
-; AVX2-ONLY-NEXT:    vbroadcastsd 216(%rsi), %ymm10
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1],ymm11[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm11 = ymm13[0],ymm14[0],ymm13[2],ymm14[2]
-; AVX2-ONLY-NEXT:    vperm2f128 $49, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm11 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm11 = ymm11[2,3],mem[2,3]
-; AVX2-ONLY-NEXT:    vmovaps (%r8), %ymm13
-; AVX2-ONLY-NEXT:    vblendps $252, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm14 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm14 = ymm13[0,1],mem[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm14 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm14 = mem[0,1],ymm13[2,3],mem[4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm13[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm11[0,1,2,3],ymm13[4,5],ymm11[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm7 = mem[2,3],ymm7[2,3]
-; AVX2-ONLY-NEXT:    vbroadcastsd 48(%rcx), %ymm11
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm11[2,3],ymm7[4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovaps 32(%r8), %ymm11
-; AVX2-ONLY-NEXT:    vblendps $252, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm13 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm13 = ymm11[0,1],mem[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm13 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm13 = mem[0,1],ymm11[2,3],mem[4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vblendps $63, (%rsp), %ymm11, %ymm13 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm13 = mem[0,1,2,3,4,5],ymm11[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm11[4,5],ymm7[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm7, (%rsp) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm15[0],ymm0[0],ymm15[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vbroadcastsd 88(%rsi), %ymm5
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm5[0,1],ymm4[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovaps 128(%rdx), %ymm5
+; AVX2-ONLY-NEXT:    vmovaps 128(%rcx), %ymm4
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm6 = ymm5[1],ymm4[1],ymm5[3],ymm4[3]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm6 = ymm6[0,2,3,3]
+; AVX2-ONLY-NEXT:    vbroadcastsd 152(%rsi), %ymm7
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm7[0,1],ymm6[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovaps 192(%rdx), %ymm7
+; AVX2-ONLY-NEXT:    vmovaps 192(%rcx), %ymm6
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm15 = ymm7[1],ymm6[1],ymm7[3],ymm6[3]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[0,2,3,3]
+; AVX2-ONLY-NEXT:    vbroadcastsd 216(%rsi), %ymm14
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm15 = ymm14[0,1],ymm15[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
 ; AVX2-ONLY-NEXT:    vperm2f128 $49, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # ymm0 = ymm0[2,3],mem[2,3]
-; AVX2-ONLY-NEXT:    vmovaps 64(%r8), %ymm7
-; AVX2-ONLY-NEXT:    vblendps $252, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm11 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm11 = ymm7[0,1],mem[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm11 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm11 = mem[0,1],ymm7[2,3],mem[4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm7[6,7]
+; AVX2-ONLY-NEXT:    vmovaps (%r8), %ymm1
+; AVX2-ONLY-NEXT:    vblendps $252, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm14 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm14 = ymm1[0,1],mem[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm14 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm14 = mem[0,1],ymm1[2,3],mem[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm7[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = mem[2,3],ymm0[2,3]
-; AVX2-ONLY-NEXT:    vbroadcastsd 112(%rcx), %ymm7
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm7[2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovaps 96(%r8), %ymm7
-; AVX2-ONLY-NEXT:    vblendps $252, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm8 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm8 = ymm7[0,1],mem[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm8 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm8 = mem[0,1],ymm7[2,3],mem[4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm8 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm8 = mem[0,1,2,3,4,5],ymm7[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm7[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 32(%rdx), %ymm0
+; AVX2-ONLY-NEXT:    vperm2f128 $49, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = ymm0[2,3],mem[2,3]
+; AVX2-ONLY-NEXT:    vbroadcastsd 48(%rcx), %ymm8
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm8[2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm8 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm8[1],ymm0[1],ymm8[3],ymm0[3]
+; AVX2-ONLY-NEXT:    vbroadcastsd 56(%rcx), %ymm8
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm8[2,3]
+; AVX2-ONLY-NEXT:    vmovaps 32(%r8), %ymm8
+; AVX2-ONLY-NEXT:    vblendps $252, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm14 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm14 = ymm8[0,1],mem[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm14 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm14 = mem[0,1],ymm8[2,3],mem[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm8[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm8[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm12[0],ymm1[0],ymm12[2],ymm1[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
 ; AVX2-ONLY-NEXT:    vperm2f128 $49, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # ymm0 = ymm0[2,3],mem[2,3]
-; AVX2-ONLY-NEXT:    vmovaps 128(%r8), %ymm1
-; AVX2-ONLY-NEXT:    vblendps $252, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm7 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm7 = ymm1[0,1],mem[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm7 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm7 = mem[0,1],ymm1[2,3],mem[4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm1[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = mem[2,3],ymm9[2,3]
-; AVX2-ONLY-NEXT:    vbroadcastsd 176(%rcx), %ymm1
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovaps 160(%r8), %ymm1
-; AVX2-ONLY-NEXT:    vblendps $252, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm4 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm4 = ymm1[0,1],mem[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm4 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm4 = mem[0,1],ymm1[2,3],mem[4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm4 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm4 = mem[0,1,2,3,4,5],ymm1[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm6[0],ymm2[0],ymm6[2],ymm2[2]
-; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm5[2,3]
-; AVX2-ONLY-NEXT:    vmovaps 192(%r8), %ymm1
+; AVX2-ONLY-NEXT:    vmovaps 64(%r8), %ymm1
 ; AVX2-ONLY-NEXT:    vblendps $252, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # ymm2 = ymm1[0,1],mem[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # ymm2 = mem[0,1],ymm1[2,3],mem[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm10[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = mem[2,3],ymm3[2,3]
-; AVX2-ONLY-NEXT:    vbroadcastsd 240(%rcx), %ymm1
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovaps 224(%r8), %ymm1
+; AVX2-ONLY-NEXT:    vmovaps 96(%rdx), %ymm0
+; AVX2-ONLY-NEXT:    vperm2f128 $49, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = ymm0[2,3],mem[2,3]
+; AVX2-ONLY-NEXT:    vbroadcastsd 112(%rcx), %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm2[1],ymm0[1],ymm2[3],ymm0[3]
+; AVX2-ONLY-NEXT:    vbroadcastsd 120(%rcx), %ymm2
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm2[2,3]
+; AVX2-ONLY-NEXT:    vmovaps 96(%r8), %ymm2
+; AVX2-ONLY-NEXT:    vblendps $252, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm3 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm3 = ymm2[0,1],mem[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm3 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm3 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm5[0],ymm4[0],ymm5[2],ymm4[2]
+; AVX2-ONLY-NEXT:    vperm2f128 $49, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[2,3],mem[2,3]
+; AVX2-ONLY-NEXT:    vmovaps 128(%r8), %ymm1
 ; AVX2-ONLY-NEXT:    vblendps $252, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # ymm2 = ymm1[0,1],mem[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # ymm2 = mem[0,1],ymm1[2,3],mem[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm2 = mem[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm13[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 128(%rdi), %xmm0
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm7 = xmm0[0],mem[0]
+; AVX2-ONLY-NEXT:    vmovaps 160(%rdx), %ymm0
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm0[2,3],ymm12[2,3]
+; AVX2-ONLY-NEXT:    vbroadcastsd 176(%rcx), %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm2[1],ymm0[1],ymm2[3],ymm0[3]
+; AVX2-ONLY-NEXT:    vbroadcastsd 184(%rcx), %ymm2
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm2[2,3]
+; AVX2-ONLY-NEXT:    vmovaps 160(%r8), %ymm2
+; AVX2-ONLY-NEXT:    vblendps $252, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm3 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm3 = ymm2[0,1],mem[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm3 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm3 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm7[0],ymm6[0],ymm7[2],ymm6[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm11[2,3]
+; AVX2-ONLY-NEXT:    vmovaps 192(%r8), %ymm2
+; AVX2-ONLY-NEXT:    vblendps $252, (%rsp), %ymm2, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = ymm2[0,1],mem[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, (%rsp) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm15[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 224(%rdx), %ymm0
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm0[2,3],ymm9[2,3]
+; AVX2-ONLY-NEXT:    vbroadcastsd 240(%rcx), %ymm4
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm4[2,3],ymm2[4,5,6,7]
+; 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:    vbroadcastsd 248(%rcx), %ymm4
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm6 = ymm0[2,3],ymm4[2,3]
+; AVX2-ONLY-NEXT:    vmovaps 224(%r8), %ymm7
+; AVX2-ONLY-NEXT:    vblendps $252, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm7[0,1],mem[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = mem[0,1],ymm7[2,3],mem[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm7[4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3,4,5],ymm7[6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 128(%rdi), %xmm6
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm6 = xmm6[0],mem[0]
 ; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm6 = xmm0[0],mem[0]
-; AVX2-ONLY-NEXT:    vmovaps 64(%rdi), %xmm10
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm10 = xmm10[0],mem[0]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm7 = xmm0[0],mem[0]
+; AVX2-ONLY-NEXT:    vmovaps 64(%rdi), %xmm9
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm9 = xmm9[0],mem[0]
 ; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm5 = xmm0[0],mem[0]
-; AVX2-ONLY-NEXT:    vmovaps 32(%rdi), %xmm15
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm15 = xmm15[0],mem[0]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm11 = xmm0[0],mem[0]
+; AVX2-ONLY-NEXT:    vmovaps 32(%rdi), %xmm13
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm13 = xmm13[0],mem[0]
 ; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm4 = xmm0[0],mem[0]
-; AVX2-ONLY-NEXT:    vmovaps 96(%rdi), %xmm14
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm14 = xmm14[0],mem[0]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm14 = xmm0[0],mem[0]
+; AVX2-ONLY-NEXT:    vmovaps 96(%rdi), %xmm8
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm8 = xmm8[0],mem[0]
 ; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm3 = xmm0[0],mem[0]
-; AVX2-ONLY-NEXT:    vmovaps 160(%rdi), %xmm13
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm13 = xmm13[0],mem[0]
+; AVX2-ONLY-NEXT:    vmovaps 160(%rdi), %xmm12
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm12 = xmm12[0],mem[0]
 ; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm2 = xmm0[0],mem[0]
-; AVX2-ONLY-NEXT:    vmovaps 224(%rdi), %xmm12
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm12 = xmm12[0],mem[0]
+; AVX2-ONLY-NEXT:    vmovaps 224(%rdi), %xmm10
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm10 = xmm10[0],mem[0]
 ; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
-; AVX2-ONLY-NEXT:    vmovaps 192(%rdi), %xmm11
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm11 = xmm11[0],mem[0]
+; AVX2-ONLY-NEXT:    vmovaps 192(%rdi), %xmm5
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm5 = xmm5[0],mem[0]
 ; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
-; AVX2-ONLY-NEXT:    vmovaps (%rdi), %xmm9
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm9 = xmm9[0],mem[0]
-; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm8 = xmm8[0],mem[0]
-; AVX2-ONLY-NEXT:    vmovaps %xmm8, 16(%r9)
-; AVX2-ONLY-NEXT:    vmovaps %xmm9, (%r9)
+; AVX2-ONLY-NEXT:    vmovaps (%rdi), %xmm15
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm15 = xmm15[0],mem[0]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm4 = xmm4[0],mem[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 16(%r9)
+; AVX2-ONLY-NEXT:    vmovaps %xmm15, (%r9)
 ; AVX2-ONLY-NEXT:    vmovaps %xmm0, 976(%r9)
-; AVX2-ONLY-NEXT:    vmovaps %xmm11, 960(%r9)
+; AVX2-ONLY-NEXT:    vmovaps %xmm5, 960(%r9)
 ; AVX2-ONLY-NEXT:    vmovaps %xmm1, 1136(%r9)
-; AVX2-ONLY-NEXT:    vmovaps %xmm12, 1120(%r9)
+; AVX2-ONLY-NEXT:    vmovaps %xmm10, 1120(%r9)
 ; AVX2-ONLY-NEXT:    vmovaps %xmm2, 816(%r9)
-; AVX2-ONLY-NEXT:    vmovaps %xmm13, 800(%r9)
+; AVX2-ONLY-NEXT:    vmovaps %xmm12, 800(%r9)
 ; AVX2-ONLY-NEXT:    vmovaps %xmm3, 496(%r9)
-; AVX2-ONLY-NEXT:    vmovaps %xmm14, 480(%r9)
-; AVX2-ONLY-NEXT:    vmovaps %xmm4, 176(%r9)
-; AVX2-ONLY-NEXT:    vmovaps %xmm15, 160(%r9)
-; AVX2-ONLY-NEXT:    vmovaps %xmm5, 336(%r9)
-; AVX2-ONLY-NEXT:    vmovaps %xmm10, 320(%r9)
-; AVX2-ONLY-NEXT:    vmovaps %xmm6, 656(%r9)
-; AVX2-ONLY-NEXT:    vmovaps %xmm7, 640(%r9)
+; AVX2-ONLY-NEXT:    vmovaps %xmm8, 480(%r9)
+; AVX2-ONLY-NEXT:    vmovaps %xmm14, 176(%r9)
+; AVX2-ONLY-NEXT:    vmovaps %xmm13, 160(%r9)
+; AVX2-ONLY-NEXT:    vmovaps %xmm11, 336(%r9)
+; AVX2-ONLY-NEXT:    vmovaps %xmm9, 320(%r9)
+; AVX2-ONLY-NEXT:    vmovaps %xmm7, 656(%r9)
+; AVX2-ONLY-NEXT:    vmovaps %xmm6, 640(%r9)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1248(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1216(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1184(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1152(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1056(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1024(%r9)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovups (%rsp), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 992(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 928(%r9)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 896(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 864(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 832(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 736(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 704(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 672(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 608(%r9)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 576(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 544(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 512(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 416(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 384(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 352(%r9)
-; AVX2-ONLY-NEXT:    vmovups (%rsp), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 288(%r9)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 256(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 224(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 192(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 96(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 64(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 32(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1088(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 768(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 448(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 128(%r9)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1248(%r9)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 928(%r9)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 608(%r9)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 288(%r9)
 ; AVX2-ONLY-NEXT:    addq $1032, %rsp # imm = 0x408
 ; AVX2-ONLY-NEXT:    vzeroupper
 ; AVX2-ONLY-NEXT:    retq
 ;
 ; AVX512F-LABEL: store_i64_stride5_vf32:
 ; AVX512F:       # %bb.0:
 ; AVX512F-NEXT:    subq $648, %rsp # imm = 0x288
 ; AVX512F-NEXT:    vmovdqa64 128(%rdi), %zmm16

 ; SSE-NEXT:    movaps %xmm0, 16(%r9)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, (%r9)
 ; SSE-NEXT:    addq $2200, %rsp # imm = 0x898
 ; SSE-NEXT:    retq
 ;
 ; AVX1-ONLY-LABEL: store_i64_stride5_vf64:
 ; AVX1-ONLY:       # %bb.0:
-; AVX1-ONLY-NEXT:    subq $2280, %rsp # imm = 0x8E8
-; AVX1-ONLY-NEXT:    vmovaps 192(%rdi), %ymm4
-; AVX1-ONLY-NEXT:    vmovaps 160(%rdi), %ymm5
-; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %ymm6
-; AVX1-ONLY-NEXT:    vmovaps 64(%rcx), %ymm1
-; AVX1-ONLY-NEXT:    vmovaps 128(%rcx), %ymm0
-; AVX1-ONLY-NEXT:    vmovaps (%rcx), %ymm2
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vmovaps 16(%rdx), %xmm7
-; AVX1-ONLY-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm2 = ymm2[1],ymm7[1],ymm2[3],ymm7[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vmovaps 80(%rdx), %xmm3
-; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm3[1],ymm1[3],ymm3[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm6[0],mem[0],ymm6[2],mem[2]
-; AVX1-ONLY-NEXT:    vmovaps %ymm6, %ymm3
-; AVX1-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 96(%rcx), %xmm2
-; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vmovaps 144(%rdx), %xmm2
-; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm2[1],ymm0[3],ymm2[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm5[0],mem[0],ymm5[2],mem[2]
-; AVX1-ONLY-NEXT:    vmovaps %ymm5, %ymm2
-; AVX1-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 160(%rcx), %xmm1
-; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm4[0],mem[0],ymm4[2],mem[2]
-; AVX1-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 192(%rcx), %xmm1
-; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 192(%rcx), %ymm0
-; AVX1-ONLY-NEXT:    vmovaps 208(%rdx), %xmm1
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 256(%rcx), %ymm0
-; AVX1-ONLY-NEXT:    vmovaps 272(%rdx), %xmm1
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 288(%rdi), %ymm1
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],mem[0],ymm1[2],mem[2]
-; AVX1-ONLY-NEXT:    vmovaps %ymm1, %ymm10
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 288(%rcx), %xmm1
-; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 320(%rdi), %ymm1
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],mem[0],ymm1[2],mem[2]
-; AVX1-ONLY-NEXT:    vmovaps %ymm1, %ymm8
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 320(%rcx), %xmm1
-; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 320(%rcx), %ymm0
-; AVX1-ONLY-NEXT:    vmovaps 336(%rdx), %xmm1
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 384(%rdi), %ymm1
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],mem[0],ymm1[2],mem[2]
-; AVX1-ONLY-NEXT:    vmovaps %ymm1, %ymm11
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 384(%rcx), %xmm1
-; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 384(%rcx), %ymm0
-; AVX1-ONLY-NEXT:    vmovaps 400(%rdx), %xmm1
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 448(%rcx), %ymm0
-; AVX1-ONLY-NEXT:    vmovaps 464(%rdx), %xmm1
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 480(%rdi), %ymm1
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm1[0],mem[0],ymm1[2],mem[2]
-; AVX1-ONLY-NEXT:    vmovapd %ymm1, %ymm15
-; AVX1-ONLY-NEXT:    vmovapd 480(%rcx), %xmm1
-; AVX1-ONLY-NEXT:    vmovapd %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps (%rdi), %ymm9
-; AVX1-ONLY-NEXT:    vbroadcastsd 8(%rsi), %ymm0
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm9[0,1,2,3],ymm0[4,5],ymm9[6,7]
-; AVX1-ONLY-NEXT:    vmovaps (%rdx), %xmm1
-; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm9[0],mem[0],ymm9[2],mem[2]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 32(%rdi), %ymm7
-; AVX1-ONLY-NEXT:    vbroadcastsd 40(%rsi), %ymm0
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm7[0,1],ymm0[2],ymm7[3]
-; AVX1-ONLY-NEXT:    vmovaps 32(%rdx), %xmm1
-; AVX1-ONLY-NEXT:    vmovaps %xmm1, (%rsp) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3]
+; AVX1-ONLY-NEXT:    subq $2248, %rsp # imm = 0x8C8
+; AVX1-ONLY-NEXT:    vmovapd 160(%rdi), %ymm12
+; AVX1-ONLY-NEXT:    vmovapd 128(%rdi), %ymm7
+; AVX1-ONLY-NEXT:    vmovupd %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 96(%rdi), %ymm6
+; AVX1-ONLY-NEXT:    vmovapd 64(%rdi), %ymm9
+; AVX1-ONLY-NEXT:    vmovupd %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 32(%rdi), %ymm11
+; AVX1-ONLY-NEXT:    vmovapd (%rdi), %ymm8
+; AVX1-ONLY-NEXT:    vmovapd 128(%rsi), %ymm0
+; AVX1-ONLY-NEXT:    vmovapd 64(%rsi), %ymm1
+; AVX1-ONLY-NEXT:    vmovapd (%rsi), %ymm2
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm3 = mem[0,0]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm5 = ymm8[0],ymm2[0],ymm8[2],ymm2[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm3[0],ymm5[1,2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm3 = mem[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm2[2,3],ymm3[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm2 = mem[0,0]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm11[0],mem[0],ymm11[2],mem[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm2[0],ymm3[1,2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm2 = mem[0,0]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm9[0],ymm1[0],ymm9[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm2[0],ymm3[1,2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm2 = mem[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3],ymm2[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm6[0],mem[0],ymm6[2],mem[2]
+; AVX1-ONLY-NEXT:    vmovapd %ymm6, %ymm9
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm7[0],ymm0[0],ymm7[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm1 = mem[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm7[0],mem[0],ymm7[2],mem[2]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %ymm6
-; AVX1-ONLY-NEXT:    vbroadcastsd 72(%rsi), %ymm0
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3],ymm0[4,5],ymm6[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 64(%rdx), %xmm1
-; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm6[0],mem[0],ymm6[2],mem[2]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vbroadcastsd 104(%rsi), %ymm0
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5],ymm3[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 96(%rdx), %xmm1
-; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 128(%rdi), %ymm5
-; AVX1-ONLY-NEXT:    vbroadcastsd 136(%rsi), %ymm0
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm0[4,5],ymm5[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 128(%rdx), %xmm1
-; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm5[0],mem[0],ymm5[2],mem[2]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vbroadcastsd 168(%rsi), %ymm0
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 160(%rdx), %xmm1
-; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vbroadcastsd 200(%rsi), %ymm0
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm0[4,5],ymm4[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 192(%rdx), %xmm1
-; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 224(%rdi), %ymm4
-; AVX1-ONLY-NEXT:    vbroadcastsd 232(%rsi), %ymm0
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm4[0,1],ymm0[2],ymm4[3]
-; AVX1-ONLY-NEXT:    vmovaps 224(%rdx), %xmm1
-; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3]
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm12[0],mem[0],ymm12[2],mem[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm4[0],mem[0],ymm4[2],mem[2]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 192(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 192(%rsi), %ymm0
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm2[0],ymm0[0],ymm2[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm1 = mem[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 256(%rdi), %ymm3
-; AVX1-ONLY-NEXT:    vbroadcastsd 264(%rsi), %ymm0
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm0[4,5],ymm3[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 256(%rdx), %xmm1
-; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm3[0],mem[0],ymm3[2],mem[2]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vbroadcastsd 296(%rsi), %ymm0
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm10[0,1,2,3],ymm0[4,5],ymm10[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 288(%rdx), %xmm1
-; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vbroadcastsd 328(%rsi), %ymm0
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm8[0,1,2,3],ymm0[4,5],ymm8[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 320(%rdx), %xmm1
-; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 352(%rdi), %ymm2
-; AVX1-ONLY-NEXT:    vbroadcastsd 360(%rsi), %ymm0
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm2[0,1],ymm0[2],ymm2[3]
-; AVX1-ONLY-NEXT:    vmovaps 352(%rdx), %xmm1
-; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3]
+; AVX1-ONLY-NEXT:    vmovapd 224(%rdi), %ymm14
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm14[0],mem[0],ymm14[2],mem[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 256(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 256(%rsi), %ymm0
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm2[0],ymm0[0],ymm2[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm1 = mem[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 288(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],mem[0],ymm2[2],mem[2]
+; AVX1-ONLY-NEXT:    vmovapd %ymm2, %ymm7
+; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 320(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 320(%rsi), %ymm0
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm2[0],ymm0[0],ymm2[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm1 = mem[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 352(%rdi), %ymm13
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm13[0],mem[0],ymm13[2],mem[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 384(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 384(%rsi), %ymm0
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm2[0],ymm0[0],ymm2[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm1 = mem[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = mem[2,3,2,3]
+; AVX1-ONLY-NEXT:    vmovapd 416(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],mem[0],ymm2[2],mem[2]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vmovapd %ymm2, %ymm10
+; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vbroadcastsd 392(%rsi), %ymm0
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm11[0,1,2,3],ymm0[4,5],ymm11[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 384(%rdx), %xmm1
-; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 416(%rdi), %ymm1
-; AVX1-ONLY-NEXT:    vbroadcastsd 424(%rsi), %ymm0
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2],ymm1[3]
-; AVX1-ONLY-NEXT:    vmovaps 416(%rdx), %xmm8
-; AVX1-ONLY-NEXT:    vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm0, %ymm8
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm8[3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm8 = ymm1[0],mem[0],ymm1[2],mem[2]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm8[2,3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 448(%rdi), %ymm0
-; AVX1-ONLY-NEXT:    vbroadcastsd 456(%rsi), %ymm8
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm0[0,1,2,3],ymm8[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 448(%rdx), %xmm10
-; AVX1-ONLY-NEXT:    vmovaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm0, %ymm10
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm8[0,1,2,3,4,5],ymm10[6,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm8 = mem[2,3,2,3]
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm10 = ymm0[0],mem[0],ymm0[2],mem[2]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm12 = ymm8[0,1,2,3],ymm10[4,5,6,7]
-; AVX1-ONLY-NEXT:    vbroadcastsd 488(%rsi), %ymm8
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm10 = ymm15[0,1],ymm8[2],ymm15[3]
-; AVX1-ONLY-NEXT:    vmovapd %ymm15, %ymm11
-; AVX1-ONLY-NEXT:    vmovaps 480(%rdx), %xmm8
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm0, %ymm15
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm10 = ymm10[0,1,2],ymm15[3]
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm15 = xmm15[0],mem[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm15[0,1,2,3,4,5],ymm9[6,7]
-; AVX1-ONLY-NEXT:    vmovaps (%r8), %ymm15
-; AVX1-ONLY-NEXT:    vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm14 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm14 = mem[0,1,2,3,4,5],ymm15[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendps $252, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm14 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm14 = ymm15[0,1],mem[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm14 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm14 = mem[0,1],ymm15[2,3],mem[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3],ymm15[4,5],ymm9[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm7 = mem[2,3],ymm7[2,3]
-; AVX1-ONLY-NEXT:    vmovapd 48(%rdx), %xmm9
-; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm7 = ymm9[0],ymm7[0],ymm9[2],ymm7[3]
-; AVX1-ONLY-NEXT:    vmovapd 48(%rsi), %xmm15
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm9 = xmm15[1],xmm9[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 56(%rcx), %ymm15
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm9 = ymm9[0,1],ymm15[2],ymm9[3]
-; AVX1-ONLY-NEXT:    vmovapd 32(%r8), %ymm15
-; AVX1-ONLY-NEXT:    vblendpd $14, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm14 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm14 = ymm15[0],mem[1,2,3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendpd $13, {{[-0-9]+}}(%r{{[sb]}}p), %ymm15, %ymm14 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm14 = mem[0],ymm15[1],mem[2,3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm7 = ymm7[0,1],ymm15[2],ymm7[3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm7 = ymm9[0,1,2],ymm15[3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm7 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm7 = xmm7[0],mem[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1,2,3,4,5],ymm6[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 64(%r8), %ymm7
-; AVX1-ONLY-NEXT:    vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm9 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm9 = mem[0,1,2,3,4,5],ymm7[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendps $252, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm9 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm9 = ymm7[0,1],mem[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm9 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm9 = mem[0,1],ymm7[2,3],mem[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4,5],ymm6[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm6 = mem[2,3],ymm6[2,3]
-; AVX1-ONLY-NEXT:    vmovapd 112(%rdx), %xmm7
+; AVX1-ONLY-NEXT:    vmovapd 448(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 448(%rsi), %ymm0
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm2[0],ymm0[0],ymm2[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm1 = mem[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 480(%rdi), %ymm4
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm4[0],mem[0],ymm4[2],mem[2]
+; AVX1-ONLY-NEXT:    vmovupd %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 48(%rdx), %xmm0
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = mem[2,3],ymm11[2,3]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm5 = ymm0[0],ymm1[0],ymm0[2],ymm1[3]
+; AVX1-ONLY-NEXT:    vmovlpd {{.*#+}} xmm0 = mem[0],xmm0[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 56(%rcx), %ymm1
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = mem[2,3],ymm6[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 112(%rdx), %xmm2
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm15 = ymm2[0],ymm1[0],ymm2[2],ymm1[3]
+; AVX1-ONLY-NEXT:    vmovlpd {{.*#+}} xmm1 = mem[0],xmm2[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 120(%rcx), %ymm2
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm2 = mem[2,3],ymm12[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 176(%rdx), %xmm3
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm2 = ymm3[0],ymm2[0],ymm3[2],ymm2[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovlpd {{.*#+}} xmm2 = mem[0],xmm3[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 184(%rcx), %ymm3
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm2[0,1],ymm3[2],ymm2[3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm2 = mem[2,3],ymm14[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 240(%rdx), %xmm6
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm2 = ymm6[0],ymm2[0],ymm6[2],ymm2[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovlpd {{.*#+}} xmm2 = mem[0],xmm6[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 248(%rcx), %ymm6
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm2[0,1],ymm6[2],ymm2[3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm6 = mem[2,3],ymm7[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 304(%rdx), %xmm7
 ; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm6 = ymm7[0],ymm6[0],ymm7[2],ymm6[3]
-; AVX1-ONLY-NEXT:    vmovapd 112(%rsi), %xmm9
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm7 = xmm9[1],xmm7[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 120(%rcx), %ymm9
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm7 = ymm7[0,1],ymm9[2],ymm7[3]
-; AVX1-ONLY-NEXT:    vmovapd 96(%r8), %ymm9
-; AVX1-ONLY-NEXT:    vblendpd $13, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm14 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm14 = mem[0],ymm9[1],mem[2,3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendpd $14, {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm14 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm14 = ymm9[0],mem[1,2,3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm6 = ymm6[0,1],ymm9[2],ymm6[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm6 = ymm7[0,1,2],ymm9[3]
+; AVX1-ONLY-NEXT:    vmovlpd {{.*#+}} xmm6 = mem[0],xmm7[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 312(%rcx), %ymm7
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm6 = ymm6[0,1],ymm7[2],ymm6[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm6 = xmm6[0],mem[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4,5],ymm5[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 128(%r8), %ymm6
-; AVX1-ONLY-NEXT:    vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm7 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm7 = mem[0,1,2,3,4,5],ymm6[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendps $252, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm7 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm7 = ymm6[0,1],mem[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm7 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm7 = mem[0,1],ymm6[2,3],mem[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5],ymm5[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm5 = mem[2,3],ymm5[2,3]
-; AVX1-ONLY-NEXT:    vmovapd 176(%rdx), %xmm6
-; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm5 = ymm6[0],ymm5[0],ymm6[2],ymm5[3]
-; AVX1-ONLY-NEXT:    vmovapd 176(%rsi), %xmm7
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm6 = xmm7[1],xmm6[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 184(%rcx), %ymm7
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm6 = mem[2,3],ymm13[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 368(%rdx), %xmm7
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm6 = ymm7[0],ymm6[0],ymm7[2],ymm6[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovlpd {{.*#+}} xmm6 = mem[0],xmm7[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 376(%rcx), %ymm7
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm6 = ymm6[0,1],ymm7[2],ymm6[3]
-; AVX1-ONLY-NEXT:    vmovapd 160(%r8), %ymm7
-; AVX1-ONLY-NEXT:    vblendpd $13, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm9 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm9 = mem[0],ymm7[1],mem[2,3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendpd $14, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm9 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm9 = ymm7[0],mem[1,2,3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm5 = ymm5[0,1],ymm7[2],ymm5[3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm5 = ymm6[0,1,2],ymm7[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm6 = mem[2,3],ymm10[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 432(%rdx), %xmm7
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm6 = ymm7[0],ymm6[0],ymm7[2],ymm6[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovlpd {{.*#+}} xmm6 = mem[0],xmm7[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 440(%rcx), %ymm7
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm6 = ymm6[0,1],ymm7[2],ymm6[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm6 = mem[2,3],ymm4[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 496(%rdx), %xmm7
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm4 = ymm7[0],ymm6[0],ymm7[2],ymm6[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovlpd {{.*#+}} xmm6 = mem[0],xmm7[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 504(%rcx), %ymm7
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm4 = ymm6[0,1],ymm7[2],ymm6[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 16(%rdx), %xmm6
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm7 = ymm4[1],ymm6[1],ymm4[2],ymm6[2]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm6 = xmm6[0],mem[0]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm6 = ymm6[0,1,2],ymm8[3]
+; AVX1-ONLY-NEXT:    vmovaps (%rdx), %xmm4
+; AVX1-ONLY-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm8, %ymm4
+; AVX1-ONLY-NEXT:    vbroadcastsd 8(%rsi), %ymm8
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm4 = ymm4[0,1],ymm8[2],ymm4[3]
+; AVX1-ONLY-NEXT:    vmovapd (%r8), %ymm8
+; AVX1-ONLY-NEXT:    vblendpd $13, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm10 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm10 = mem[0],ymm8[1],mem[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm7 = ymm7[0,1,2],ymm8[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm6 = ymm6[0,1],ymm8[2],ymm6[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm4 = ymm8[0],ymm4[1,2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 32(%rdx), %xmm4
+; AVX1-ONLY-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm11, %ymm4
+; AVX1-ONLY-NEXT:    vbroadcastsd 40(%rsi), %ymm6
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm4 = ymm4[0,1],ymm6[2],ymm4[3]
+; AVX1-ONLY-NEXT:    vmovapd 32(%r8), %ymm6
+; AVX1-ONLY-NEXT:    vblendpd $13, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm7 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm7 = mem[0],ymm6[1],mem[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm5 = ymm5[0,1],ymm6[2],ymm5[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm5 = xmm5[0],mem[0]
-; AVX1-ONLY-NEXT:    vblendps $192, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm5 = ymm5[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 192(%r8), %ymm6
-; AVX1-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm7 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm7 = mem[0,1],ymm6[2,3],mem[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm7 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm7 = mem[0,1,2,3,4,5],ymm6[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendps $252, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm7 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm7 = ymm6[0,1],mem[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm6[4,5],ymm5[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm4 = mem[2,3],ymm4[2,3]
-; AVX1-ONLY-NEXT:    vmovapd 240(%rdx), %xmm5
-; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm4 = ymm5[0],ymm4[0],ymm5[2],ymm4[3]
-; AVX1-ONLY-NEXT:    vmovapd 240(%rsi), %xmm6
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm5 = xmm6[1],xmm5[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 248(%rcx), %ymm6
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm6[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm6[0],ymm4[1,2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 80(%rdx), %xmm0
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm4 = ymm4[1],ymm0[1],ymm4[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm5[3]
+; AVX1-ONLY-NEXT:    vmovaps 64(%rdx), %xmm6
+; AVX1-ONLY-NEXT:    vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm5, %ymm5
+; AVX1-ONLY-NEXT:    vbroadcastsd 72(%rsi), %ymm6
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm5 = ymm5[0,1],ymm6[2],ymm5[3]
-; AVX1-ONLY-NEXT:    vmovapd 224(%r8), %ymm6
-; AVX1-ONLY-NEXT:    vblendpd $14, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm7 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm7 = ymm6[0],mem[1,2,3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 64(%r8), %ymm6
 ; AVX1-ONLY-NEXT:    vblendpd $13, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm7 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm7 = mem[0],ymm6[1],mem[2,3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm4 = ymm4[0,1],ymm6[2],ymm4[3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm4 = ymm5[0,1,2],ymm6[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm4 = ymm4[0,1,2],ymm6[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm4 = xmm4[0],mem[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3,4,5],ymm3[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 256(%r8), %ymm4
-; AVX1-ONLY-NEXT:    vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm5 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm5 = mem[0,1,2,3,4,5],ymm4[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendps $252, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm5 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm5 = ymm4[0,1],mem[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm5 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm5 = mem[0,1],ymm4[2,3],mem[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5],ymm3[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = mem[2,3],ymm3[2,3]
-; AVX1-ONLY-NEXT:    vmovapd 304(%rdx), %xmm4
-; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm3 = ymm4[0],ymm3[0],ymm4[2],ymm3[3]
-; AVX1-ONLY-NEXT:    vmovapd 304(%rsi), %xmm5
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm4 = xmm5[1],xmm4[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 312(%rcx), %ymm5
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm6[2],ymm0[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm6[0],ymm5[1,2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 96(%rdx), %xmm10
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm9, %ymm0
+; AVX1-ONLY-NEXT:    vbroadcastsd 104(%rsi), %ymm4
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm4[2],ymm0[3]
+; AVX1-ONLY-NEXT:    vmovapd 96(%r8), %ymm4
+; AVX1-ONLY-NEXT:    vblendpd $13, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm5 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm5 = mem[0],ymm4[1],mem[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm5 = ymm15[0,1],ymm4[2],ymm15[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1,2],ymm4[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm4[0],ymm0[1,2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 144(%rdx), %xmm0
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm1[1],ymm0[1],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm4[3]
+; AVX1-ONLY-NEXT:    vmovaps 128(%rdx), %xmm15
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm4, %ymm4
+; AVX1-ONLY-NEXT:    vbroadcastsd 136(%rsi), %ymm5
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm4 = ymm4[0,1],ymm5[2],ymm4[3]
-; AVX1-ONLY-NEXT:    vmovapd 288(%r8), %ymm5
+; AVX1-ONLY-NEXT:    vmovapd 128(%r8), %ymm5
 ; AVX1-ONLY-NEXT:    vblendpd $13, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm6 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm6 = mem[0],ymm5[1],mem[2,3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendpd $14, {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm6 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm6 = ymm5[0],mem[1,2,3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm3[0,1],ymm5[2],ymm3[3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm4[0,1,2],ymm5[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1,2],ymm5[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm5[2],ymm0[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm5[0],ymm4[1,2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 160(%rdx), %xmm9
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm9, %ymm12, %ymm0
+; AVX1-ONLY-NEXT:    vbroadcastsd 168(%rsi), %ymm1
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3]
+; AVX1-ONLY-NEXT:    vmovapd 160(%r8), %ymm1
+; AVX1-ONLY-NEXT:    vblendpd $13, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm4 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm4 = mem[0],ymm1[1],mem[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd $11, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm4 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm4 = mem[0,1],ymm1[2],mem[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm3[0,1,2],ymm1[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm3 = xmm3[0],mem[0]
-; AVX1-ONLY-NEXT:    vblendps $192, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm3 = ymm3[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 320(%r8), %ymm4
-; AVX1-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm5 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm5 = mem[0,1],ymm4[2,3],mem[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm5 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm5 = mem[0,1,2,3,4,5],ymm4[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendps $252, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm5 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm5 = ymm4[0,1],mem[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5],ymm3[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm2 = mem[2,3],ymm2[2,3]
-; AVX1-ONLY-NEXT:    vmovapd 368(%rdx), %xmm3
-; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm2 = ymm3[0],ymm2[0],ymm3[2],ymm2[3]
-; AVX1-ONLY-NEXT:    vmovapd 368(%rsi), %xmm4
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm3 = xmm4[1],xmm3[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 376(%rcx), %ymm4
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 208(%rdx), %xmm0
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm1[1],ymm0[1],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm12[3]
+; AVX1-ONLY-NEXT:    vmovaps 192(%rdx), %xmm3
+; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm12, %ymm3
+; AVX1-ONLY-NEXT:    vbroadcastsd 200(%rsi), %ymm4
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm3[0,1],ymm4[2],ymm3[3]
-; AVX1-ONLY-NEXT:    vmovapd 352(%r8), %ymm4
-; AVX1-ONLY-NEXT:    vblendpd $14, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm5 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm5 = ymm4[0],mem[1,2,3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 192(%r8), %ymm4
 ; AVX1-ONLY-NEXT:    vblendpd $13, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm5 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm5 = mem[0],ymm4[1],mem[2,3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm2[0,1],ymm4[2],ymm2[3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm3[0,1,2],ymm4[3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm2 = xmm2[0],mem[0]
-; AVX1-ONLY-NEXT:    vblendps $192, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm2 = ymm2[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 384(%r8), %ymm3
-; AVX1-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm4 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm4 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm4 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm4 = mem[0,1,2,3,4,5],ymm3[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendps $252, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm4 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm4 = ymm3[0,1],mem[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = mem[2,3],ymm1[2,3]
-; AVX1-ONLY-NEXT:    vmovapd 432(%rdx), %xmm2
-; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[3]
-; AVX1-ONLY-NEXT:    vmovapd 432(%rsi), %xmm3
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm2 = xmm3[1],xmm2[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 440(%rcx), %ymm3
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm2[0,1],ymm3[2],ymm2[3]
-; AVX1-ONLY-NEXT:    vmovapd 416(%r8), %ymm3
-; AVX1-ONLY-NEXT:    vblendpd $14, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm4 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm4 = ymm3[0],mem[1,2,3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendpd $13, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm4 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm4 = mem[0],ymm3[1],mem[2,3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm3[2],ymm1[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1,2],ymm4[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm2[0,1,2],ymm3[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm4[2],ymm0[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm4[0],ymm3[1,2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 224(%rdx), %xmm8
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm14, %ymm0
+; AVX1-ONLY-NEXT:    vbroadcastsd 232(%rsi), %ymm1
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3]
+; AVX1-ONLY-NEXT:    vmovapd 224(%r8), %ymm1
+; AVX1-ONLY-NEXT:    vblendpd $13, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm3 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm3 = mem[0],ymm1[1],mem[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd $11, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm3 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm3 = mem[0,1],ymm1[2],mem[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm2[0,1,2],ymm1[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 272(%rdx), %xmm0
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm1[1],ymm0[1],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm14[3]
+; AVX1-ONLY-NEXT:    vmovaps 256(%rdx), %xmm5
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm14, %ymm2
+; AVX1-ONLY-NEXT:    vbroadcastsd 264(%rsi), %ymm4
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm2[0,1],ymm4[2],ymm2[3]
+; AVX1-ONLY-NEXT:    vmovapd 256(%r8), %ymm4
+; AVX1-ONLY-NEXT:    vblendpd $13, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm3 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm3 = mem[0],ymm4[1],mem[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1,2],ymm4[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm1[0],mem[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 448(%r8), %ymm1
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm4[2],ymm0[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm4[0],ymm2[1,2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 288(%rdx), %xmm7
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm7, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vbroadcastsd 296(%rsi), %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 288(%r8), %ymm1
+; AVX1-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm2 = mem[0,1],ymm1[2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm2 = mem[0,1,2,3],ymm1[4,5],mem[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm2 = mem[0,1,2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm1[0,1],ymm13[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 336(%rdx), %xmm0
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm1[1],ymm0[1],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3]
+; AVX1-ONLY-NEXT:    vmovaps 320(%rdx), %xmm6
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm2, %ymm2
+; AVX1-ONLY-NEXT:    vbroadcastsd 328(%rsi), %ymm4
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm2[0,1],ymm4[2],ymm2[3]
+; AVX1-ONLY-NEXT:    vmovapd 320(%r8), %ymm4
+; AVX1-ONLY-NEXT:    vblendpd $13, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm3 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm3 = mem[0],ymm4[1],mem[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1,2],ymm4[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm4[2],ymm0[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm4[0],ymm2[1,2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 352(%rdx), %xmm4
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm13, %ymm0
+; AVX1-ONLY-NEXT:    vbroadcastsd 360(%rsi), %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 352(%r8), %ymm1
+; AVX1-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm2 = mem[0,1],ymm1[2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm2 = mem[0,1,2,3],ymm1[4,5],mem[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm12[0,1],ymm1[2,3],ymm12[4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm2 = mem[0,1,2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = mem[2,3],ymm11[2,3]
-; AVX1-ONLY-NEXT:    vmovapd 496(%rdx), %xmm1
-; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[3]
-; AVX1-ONLY-NEXT:    vmovapd 496(%rsi), %xmm2
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm2[1],xmm1[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 504(%rcx), %ymm2
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3]
-; AVX1-ONLY-NEXT:    vmovapd 480(%r8), %ymm2
-; AVX1-ONLY-NEXT:    vblendpd $13, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm3 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm3 = mem[0],ymm2[1],mem[2,3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm2[0],ymm10[1,2,3]
+; AVX1-ONLY-NEXT:    vmovapd 400(%rdx), %xmm0
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm1[1],ymm0[1],ymm1[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3]
+; AVX1-ONLY-NEXT:    vmovaps 384(%rdx), %xmm13
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm13, %ymm2, %ymm2
+; AVX1-ONLY-NEXT:    vbroadcastsd 392(%rsi), %ymm14
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm2[0,1],ymm14[2],ymm2[3]
+; AVX1-ONLY-NEXT:    vmovapd 384(%r8), %ymm14
+; AVX1-ONLY-NEXT:    vblendpd $13, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14, %ymm3 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm3 = mem[0],ymm14[1],mem[2,3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm2[2],ymm0[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1,2],ymm14[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm14[2],ymm0[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1,2],ymm2[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm14[0],ymm2[1,2,3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm0 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm0 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm0 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm0 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm0 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm0 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm0 = xmm8[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm4 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm4 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps 256(%rdi), %xmm2
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm2[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 128(%rdi), %xmm2
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm2[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %xmm15
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm15[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm14
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm14[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps (%rsp), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps 416(%rdx), %xmm3
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vbroadcastsd 424(%rsi), %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 416(%r8), %ymm1
+; AVX1-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm2 = mem[0,1],ymm1[2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm2 = mem[0,1,2,3],ymm1[4,5],mem[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm2 = mem[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 464(%rdx), %xmm0
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm14 = ymm1[1],ymm0[1],ymm1[2],ymm0[2]
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps %xmm0, (%rsp) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %xmm13
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm13[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 160(%rdi), %xmm12
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm12[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 224(%rdi), %xmm11
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm11 = xmm11[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm3 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps 192(%rdi), %xmm10
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm10 = xmm10[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps 288(%rdi), %xmm9
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm9 = xmm9[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps 352(%rdi), %xmm8
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm8 = xmm8[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm2 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps 320(%rdi), %xmm7
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm0[0,1,2],ymm12[3]
+; AVX1-ONLY-NEXT:    vmovaps 448(%rdx), %xmm2
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm12, %ymm0
+; AVX1-ONLY-NEXT:    vbroadcastsd 456(%rsi), %ymm12
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm12[2],ymm0[3]
+; AVX1-ONLY-NEXT:    vmovapd 448(%r8), %ymm12
+; AVX1-ONLY-NEXT:    vblendpd $13, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm11 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm11 = mem[0],ymm12[1],mem[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm11 = ymm14[0,1,2],ymm12[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm12[2],ymm1[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm12[0],ymm0[1,2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 480(%rdx), %xmm0
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vbroadcastsd 488(%rsi), %ymm12
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm12[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 480(%r8), %ymm12
+; AVX1-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm11 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm11 = mem[0,1],ymm12[2,3],mem[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps $207, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm11 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm11 = mem[0,1,2,3],ymm12[4,5],mem[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm14 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm14 = mem[0,1,2,3,4,5],ymm12[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm12[0,1],ymm1[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 256(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm1[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm5[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 128(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm1[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm15[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm1[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm1[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm1[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm1[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm1[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm10[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, (%rsp) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 160(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm1[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm9[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 224(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm1[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm8[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 192(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm1[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm1[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 288(%rdi), %xmm5
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm5 = xmm5[0],mem[0]
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm7 = xmm7[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps 416(%rdi), %xmm6
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm6 = xmm6[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps 480(%rdi), %xmm15
+; AVX1-ONLY-NEXT:    vmovaps 352(%rdi), %xmm15
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm15 = xmm15[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps 448(%rdi), %xmm5
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm5 = xmm5[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps 384(%rdi), %xmm14
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm4 = xmm4[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps 320(%rdi), %xmm14
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm14 = xmm14[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm13
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm13 = xmm13[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm6 = xmm6[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps 416(%rdi), %xmm12
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm12 = xmm12[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps %xmm12, 16(%r9)
-; AVX1-ONLY-NEXT:    vmovaps %xmm13, (%r9)
-; AVX1-ONLY-NEXT:    vmovaps %xmm4, 1936(%r9)
-; AVX1-ONLY-NEXT:    vmovaps %xmm14, 1920(%r9)
-; AVX1-ONLY-NEXT:    vmovaps %xmm0, 2256(%r9)
-; AVX1-ONLY-NEXT:    vmovaps %xmm5, 2240(%r9)
-; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm3 = xmm3[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps 480(%rdi), %xmm11
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm11 = xmm11[0],mem[0]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps 448(%rdi), %xmm10
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm10 = xmm10[0],mem[0]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm2 = xmm2[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps 384(%rdi), %xmm9
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm9 = xmm9[0],mem[0]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm13 = xmm13[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm8
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm8 = xmm8[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm1[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, 16(%r9)
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, (%r9)
+; AVX1-ONLY-NEXT:    vmovaps %xmm13, 1936(%r9)
+; AVX1-ONLY-NEXT:    vmovaps %xmm9, 1920(%r9)
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, 2256(%r9)
+; AVX1-ONLY-NEXT:    vmovaps %xmm10, 2240(%r9)
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, 2416(%r9)
-; AVX1-ONLY-NEXT:    vmovaps %xmm15, 2400(%r9)
-; AVX1-ONLY-NEXT:    vmovaps %xmm1, 2096(%r9)
-; AVX1-ONLY-NEXT:    vmovaps %xmm6, 2080(%r9)
+; AVX1-ONLY-NEXT:    vmovaps %xmm11, 2400(%r9)
+; AVX1-ONLY-NEXT:    vmovaps %xmm3, 2096(%r9)
+; AVX1-ONLY-NEXT:    vmovaps %xmm12, 2080(%r9)
+; AVX1-ONLY-NEXT:    vmovaps %xmm6, 1616(%r9)
+; AVX1-ONLY-NEXT:    vmovaps %xmm14, 1600(%r9)
+; AVX1-ONLY-NEXT:    vmovaps %xmm4, 1776(%r9)
+; AVX1-ONLY-NEXT:    vmovaps %xmm15, 1760(%r9)
+; AVX1-ONLY-NEXT:    vmovaps %xmm7, 1456(%r9)
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 1440(%r9)
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %xmm0, 1616(%r9)
-; AVX1-ONLY-NEXT:    vmovaps %xmm7, 1600(%r9)
-; AVX1-ONLY-NEXT:    vmovaps %xmm2, 1776(%r9)
-; AVX1-ONLY-NEXT:    vmovaps %xmm8, 1760(%r9)
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, 976(%r9)
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %xmm0, 1456(%r9)
-; AVX1-ONLY-NEXT:    vmovaps %xmm9, 1440(%r9)
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, 960(%r9)
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %xmm0, 976(%r9)
-; AVX1-ONLY-NEXT:    vmovaps %xmm10, 960(%r9)
-; AVX1-ONLY-NEXT:    vmovaps %xmm3, 1136(%r9)
-; AVX1-ONLY-NEXT:    vmovaps %xmm11, 1120(%r9)
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, 1136(%r9)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, 1120(%r9)
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, 816(%r9)
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, 800(%r9)
-; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps (%rsp), %xmm0 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, 496(%r9)
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, 480(%r9)
-; AVX1-ONLY-NEXT:    vmovaps (%rsp), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, 176(%r9)
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, 160(%r9)
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, 336(%r9)
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, 320(%r9)
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, 656(%r9)
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, 640(%r9)
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, 1296(%r9)
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, 1280(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2496(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2464(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2432(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2368(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2304(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2272(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2176(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2144(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2112(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2048(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1984(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1952(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1856(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1824(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1792(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1728(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1664(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1632(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1536(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1504(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1472(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1408(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1344(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1312(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1216(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1184(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1152(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1088(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1024(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 992(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 896(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 864(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 832(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 768(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 704(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 672(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 576(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 544(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 512(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 448(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 384(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 352(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 256(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 224(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 192(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 128(%r9)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 64(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 32(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2528(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2464(%r9)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2368(%r9)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2336(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2208(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2048(%r9)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2016(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1984(%r9)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1888(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1728(%r9)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1696(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1664(%r9)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1568(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1504(%r9)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1408(%r9)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1376(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1248(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1088(%r9)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1056(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1024(%r9)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 928(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 864(%r9)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 768(%r9)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 736(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 608(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 544(%r9)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 448(%r9)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 416(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 288(%r9)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 128(%r9)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 96(%r9)
-; AVX1-ONLY-NEXT:    addq $2280, %rsp # imm = 0x8E8
+; AVX1-ONLY-NEXT:    addq $2248, %rsp # imm = 0x8C8
 ; AVX1-ONLY-NEXT:    vzeroupper
 ; AVX1-ONLY-NEXT:    retq
 ;
 ; AVX2-ONLY-LABEL: store_i64_stride5_vf64:
 ; AVX2-ONLY:       # %bb.0:
-; AVX2-ONLY-NEXT:    subq $2440, %rsp # imm = 0x988
+; AVX2-ONLY-NEXT:    subq $2392, %rsp # imm = 0x958
 ; AVX2-ONLY-NEXT:    vmovaps 96(%rdi), %ymm4
 ; AVX2-ONLY-NEXT:    vmovaps 64(%rdi), %ymm5
 ; AVX2-ONLY-NEXT:    vmovaps 32(%rdi), %ymm6
-; AVX2-ONLY-NEXT:    vmovaps (%rdi), %ymm7
+; AVX2-ONLY-NEXT:    vmovaps (%rdi), %ymm8
+; AVX2-ONLY-NEXT:    vmovaps 32(%rsi), %ymm7
+; AVX2-ONLY-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovaps (%rdx), %xmm0
 ; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-ONLY-NEXT:    vmovaps 32(%rdx), %xmm2
 ; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-ONLY-NEXT:    vmovaps 64(%rdx), %xmm3
 ; AVX2-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX2-ONLY-NEXT:    vbroadcastsd 8(%rsi), %ymm1
-; AVX2-ONLY-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm7[0,1,2,3],ymm1[4,5],ymm7[6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm8[0,1,2,3],ymm1[4,5],ymm8[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm7[0],mem[0],ymm7[2],mem[2]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm8[0],mem[0],ymm8[2],mem[2]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
 ; AVX2-ONLY-NEXT:    vbroadcastsd 40(%rsi), %ymm1
 ; AVX2-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm6[0,1,2,3],ymm1[4,5],ymm6[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm6[0],mem[0],ymm6[2],mem[2]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm6[0],ymm7[0],ymm6[2],ymm7[2]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm0
 ; AVX2-ONLY-NEXT:    vbroadcastsd 72(%rsi), %ymm1
 ; AVX2-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm5[0,1,2,3],ymm1[4,5],ymm5[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
 ; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm5[0],mem[0],ymm5[2],mem[2]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 96(%rdx), %xmm0
-; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX2-ONLY-NEXT:    vbroadcastsd 104(%rsi), %ymm1
+; AVX2-ONLY-NEXT:    vmovaps 96(%rdx), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastsd 104(%rsi), %ymm0
 ; AVX2-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm4[0,1,2,3],ymm1[4,5],ymm4[6,7]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm0[4,5],ymm4[6,7]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 96(%rsi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm4[0],mem[0],ymm4[2],mem[2]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm4[0],ymm1[0],ymm4[2],ymm1[2]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-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:    vbroadcastsd 136(%rsi), %ymm0
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5],ymm2[6,7]
 ; AVX2-ONLY-NEXT:    vmovaps 128(%rdx), %xmm1
 ; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
 ; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],mem[0],ymm2[2],mem[2]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovaps 160(%rdi), %ymm2
 ; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vbroadcastsd 168(%rsi), %ymm0
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5],ymm2[6,7]
 ; AVX2-ONLY-NEXT:    vmovaps 160(%rdx), %xmm1
 ; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 160(%rsi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],mem[0],ymm2[2],mem[2]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovaps 192(%rdi), %ymm2
 ; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vbroadcastsd 200(%rsi), %ymm0
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5],ymm2[6,7]
 ; AVX2-ONLY-NEXT:    vmovaps 192(%rdx), %xmm1
 ; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
 ; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],mem[0],ymm2[2],mem[2]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovaps 224(%rdi), %ymm2
 ; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vbroadcastsd 232(%rsi), %ymm0
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5],ymm2[6,7]
 ; AVX2-ONLY-NEXT:    vmovaps 224(%rdx), %xmm1
 ; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 224(%rsi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],mem[0],ymm2[2],mem[2]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovaps 256(%rdi), %ymm2
 ; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vbroadcastsd 264(%rsi), %ymm0
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5],ymm2[6,7]
 ; AVX2-ONLY-NEXT:    vmovaps 256(%rdx), %xmm1
 ; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
 ; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],mem[0],ymm2[2],mem[2]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovaps 288(%rdi), %ymm2
 ; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vbroadcastsd 296(%rsi), %ymm0
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5],ymm2[6,7]
 ; AVX2-ONLY-NEXT:    vmovaps 288(%rdx), %xmm1
 ; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 288(%rsi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],mem[0],ymm2[2],mem[2]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-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:    vbroadcastsd 328(%rsi), %ymm0
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5],ymm2[6,7]
 ; AVX2-ONLY-NEXT:    vmovaps 320(%rdx), %xmm1
 ; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
 ; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],mem[0],ymm2[2],mem[2]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovaps 352(%rdi), %ymm2
 ; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vbroadcastsd 360(%rsi), %ymm0
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5],ymm2[6,7]
 ; AVX2-ONLY-NEXT:    vmovaps 352(%rdx), %xmm1
 ; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 352(%rsi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],mem[0],ymm2[2],mem[2]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovaps 384(%rdi), %ymm2
 ; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vbroadcastsd 392(%rsi), %ymm0
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5],ymm2[6,7]
 ; AVX2-ONLY-NEXT:    vmovaps 384(%rdx), %xmm1
 ; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
 ; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],mem[0],ymm2[2],mem[2]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovaps 416(%rdi), %ymm2
 ; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vbroadcastsd 424(%rsi), %ymm0
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5],ymm2[6,7]
 ; AVX2-ONLY-NEXT:    vmovaps 416(%rdx), %xmm1
 ; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 416(%rsi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],mem[0],ymm2[2],mem[2]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovaps 448(%rdi), %ymm2
 ; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vbroadcastsd 456(%rsi), %ymm0
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5],ymm2[6,7]
 ; AVX2-ONLY-NEXT:    vmovaps 448(%rdx), %xmm1
 ; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
 ; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],mem[0],ymm2[2],mem[2]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovaps 480(%rdi), %ymm2
 ; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vbroadcastsd 488(%rsi), %ymm0
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5],ymm2[6,7]
 ; AVX2-ONLY-NEXT:    vmovaps 480(%rdx), %xmm1
 ; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 480(%rsi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],mem[0],ymm2[2],mem[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:    vbroadcastsd 56(%rsi), %ymm0
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
-; AVX2-ONLY-NEXT:    vbroadcastsd 56(%rcx), %ymm1
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vbroadcastsd 120(%rsi), %ymm0
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
-; AVX2-ONLY-NEXT:    vbroadcastsd 120(%rcx), %ymm1
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vbroadcastsd 184(%rsi), %ymm0
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
-; AVX2-ONLY-NEXT:    vbroadcastsd 184(%rcx), %ymm1
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vbroadcastsd 248(%rsi), %ymm0
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
-; AVX2-ONLY-NEXT:    vbroadcastsd 248(%rcx), %ymm1
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vbroadcastsd 312(%rsi), %ymm0
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
-; AVX2-ONLY-NEXT:    vbroadcastsd 312(%rcx), %ymm1
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vbroadcastsd 376(%rsi), %ymm0
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
-; AVX2-ONLY-NEXT:    vbroadcastsd 376(%rcx), %ymm1
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vbroadcastsd 440(%rsi), %ymm0
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
-; AVX2-ONLY-NEXT:    vbroadcastsd 440(%rcx), %ymm1
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vbroadcastsd 504(%rsi), %ymm0
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0,1],mem[2,3]
-; AVX2-ONLY-NEXT:    vbroadcastsd 504(%rcx), %ymm1
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps (%rdx), %ymm0
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps (%rdx), %ymm10
 ; AVX2-ONLY-NEXT:    vmovaps (%rcx), %ymm1
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm2 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm2 = ymm10[1],ymm1[1],ymm10[3],ymm1[3]
 ; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,2,3,3]
 ; AVX2-ONLY-NEXT:    vbroadcastsd 24(%rsi), %ymm3
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1],ymm2[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovaps 64(%rdx), %ymm3
 ; AVX2-ONLY-NEXT:    vmovaps 64(%rcx), %ymm2
 ; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm4 = ymm3[1],ymm2[1],ymm3[3],ymm2[3]
 ; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[0,2,3,3]

 ; AVX2-ONLY-NEXT:    vmovaps 192(%rcx), %ymm6
 ; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm8 = ymm7[1],ymm6[1],ymm7[3],ymm6[3]
 ; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm8 = ymm8[0,2,3,3]
 ; AVX2-ONLY-NEXT:    vbroadcastsd 216(%rsi), %ymm9
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm9[0,1],ymm8[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovaps 256(%rdx), %ymm9
 ; AVX2-ONLY-NEXT:    vmovaps 256(%rcx), %ymm8
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm10 = ymm9[1],ymm8[1],ymm9[3],ymm8[3]
-; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm10 = ymm10[0,2,3,3]
-; AVX2-ONLY-NEXT:    vbroadcastsd 280(%rsi), %ymm11
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm11[0,1],ymm10[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 320(%rdx), %ymm11
-; AVX2-ONLY-NEXT:    vmovaps 320(%rcx), %ymm10
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm12 = ymm11[1],ymm10[1],ymm11[3],ymm10[3]
-; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm12 = ymm12[0,2,3,3]
-; AVX2-ONLY-NEXT:    vbroadcastsd 344(%rsi), %ymm13
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm13[0,1],ymm12[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm0, (%rsp) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 384(%rdx), %ymm13
-; AVX2-ONLY-NEXT:    vmovaps 384(%rcx), %ymm12
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm14 = ymm13[1],ymm12[1],ymm13[3],ymm12[3]
-; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm14 = ymm14[0,2,3,3]
-; AVX2-ONLY-NEXT:    vbroadcastsd 408(%rsi), %ymm15
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1],ymm14[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm11 = ymm9[1],ymm8[1],ymm9[3],ymm8[3]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm11 = ymm11[0,2,3,3]
+; AVX2-ONLY-NEXT:    vbroadcastsd 280(%rsi), %ymm12
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm12[0,1],ymm11[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 320(%rdx), %ymm12
+; AVX2-ONLY-NEXT:    vmovaps 320(%rcx), %ymm11
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm13 = ymm12[1],ymm11[1],ymm12[3],ymm11[3]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm13 = ymm13[0,2,3,3]
+; AVX2-ONLY-NEXT:    vbroadcastsd 344(%rsi), %ymm14
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm14[0,1],ymm13[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 384(%rdx), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 384(%rcx), %ymm13
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm15 = ymm0[1],ymm13[1],ymm0[3],ymm13[3]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[0,2,3,3]
+; AVX2-ONLY-NEXT:    vbroadcastsd 408(%rsi), %ymm14
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm14[0,1],ymm15[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovaps 448(%rdx), %ymm14
 ; AVX2-ONLY-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovaps 448(%rcx), %ymm0
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm15 = ymm14[1],ymm0[1],ymm14[3],ymm0[3]
-; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[0,2,3,3]
-; AVX2-ONLY-NEXT:    vbroadcastsd 472(%rsi), %ymm14
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm15 = ymm14[0,1],ymm15[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm14 = ymm14[1],ymm0[1],ymm14[3],ymm0[3]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm14 = ymm14[0,2,3,3]
+; AVX2-ONLY-NEXT:    vbroadcastsd 472(%rsi), %ymm15
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm15 = ymm15[0,1],ymm14[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm10[0],ymm1[0],ymm10[2],ymm1[2]
 ; AVX2-ONLY-NEXT:    vperm2f128 $49, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # ymm0 = ymm0[2,3],mem[2,3]
 ; AVX2-ONLY-NEXT:    vmovaps (%r8), %ymm1
 ; AVX2-ONLY-NEXT:    vblendps $252, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm14 # 32-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # ymm14 = ymm1[0,1],mem[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm14 # 32-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # ymm14 = mem[0,1],ymm1[2,3],mem[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm14 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm14 = mem[0,1,2,3,4,5],ymm1[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm10 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm10 = mem[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = mem[2,3],ymm0[2,3]
-; AVX2-ONLY-NEXT:    vbroadcastsd 48(%rcx), %ymm1
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovaps 32(%r8), %ymm1
-; AVX2-ONLY-NEXT:    vblendps $252, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm14 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm14 = ymm1[0,1],mem[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovaps 32(%rdx), %ymm0
+; AVX2-ONLY-NEXT:    vperm2f128 $49, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = ymm0[2,3],mem[2,3]
+; AVX2-ONLY-NEXT:    vbroadcastsd 48(%rcx), %ymm10
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm10[2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm10 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm10[1],ymm0[1],ymm10[3],ymm0[3]
+; AVX2-ONLY-NEXT:    vbroadcastsd 56(%rcx), %ymm10
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm10[2,3]
+; AVX2-ONLY-NEXT:    vmovaps 32(%r8), %ymm10
+; AVX2-ONLY-NEXT:    vblendps $252, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm14 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm14 = ymm10[0,1],mem[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm14 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm14 = mem[0,1],ymm1[2,3],mem[4,5,6,7]
+; AVX2-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm14 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm14 = mem[0,1],ymm10[2,3],mem[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm14 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm14 = mem[0,1,2,3,4,5],ymm1[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm10[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm10[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
 ; AVX2-ONLY-NEXT:    vperm2f128 $49, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # ymm0 = ymm0[2,3],mem[2,3]
 ; AVX2-ONLY-NEXT:    vmovaps 64(%r8), %ymm1
 ; AVX2-ONLY-NEXT:    vblendps $252, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # ymm2 = ymm1[0,1],mem[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # ymm2 = mem[0,1],ymm1[2,3],mem[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # ymm2 = mem[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = mem[2,3],ymm0[2,3]
-; AVX2-ONLY-NEXT:    vbroadcastsd 112(%rcx), %ymm1
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovaps 96(%r8), %ymm1
-; AVX2-ONLY-NEXT:    vblendps $252, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm2 = ymm1[0,1],mem[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm2 = mem[0,1],ymm1[2,3],mem[4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm2 = mem[0,1,2,3,4,5],ymm1[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 96(%rdx), %ymm0
+; AVX2-ONLY-NEXT:    vperm2f128 $49, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = ymm0[2,3],mem[2,3]
+; AVX2-ONLY-NEXT:    vbroadcastsd 112(%rcx), %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm2[1],ymm0[1],ymm2[3],ymm0[3]
+; AVX2-ONLY-NEXT:    vbroadcastsd 120(%rcx), %ymm2
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm2[2,3]
+; AVX2-ONLY-NEXT:    vmovaps 96(%r8), %ymm2
+; AVX2-ONLY-NEXT:    vblendps $252, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm3 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm3 = ymm2[0,1],mem[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm3 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm3 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm5[0],ymm4[0],ymm5[2],ymm4[2]
 ; AVX2-ONLY-NEXT:    vperm2f128 $49, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # ymm0 = ymm0[2,3],mem[2,3]
 ; AVX2-ONLY-NEXT:    vmovaps 128(%r8), %ymm1
 ; AVX2-ONLY-NEXT:    vblendps $252, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # ymm2 = ymm1[0,1],mem[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # ymm2 = mem[0,1],ymm1[2,3],mem[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # ymm2 = mem[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = mem[2,3],ymm0[2,3]
-; AVX2-ONLY-NEXT:    vbroadcastsd 176(%rcx), %ymm1
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovaps 160(%r8), %ymm1
-; AVX2-ONLY-NEXT:    vblendps $252, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm2 = ymm1[0,1],mem[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm2 = mem[0,1],ymm1[2,3],mem[4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm2 = mem[0,1,2,3,4,5],ymm1[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 160(%rdx), %ymm0
+; AVX2-ONLY-NEXT:    vperm2f128 $49, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = ymm0[2,3],mem[2,3]
+; AVX2-ONLY-NEXT:    vbroadcastsd 176(%rcx), %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm2[1],ymm0[1],ymm2[3],ymm0[3]
+; AVX2-ONLY-NEXT:    vbroadcastsd 184(%rcx), %ymm2
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm2[2,3]
+; AVX2-ONLY-NEXT:    vmovaps 160(%r8), %ymm2
+; AVX2-ONLY-NEXT:    vblendps $252, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm3 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm3 = ymm2[0,1],mem[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm3 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm3 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm7[0],ymm6[0],ymm7[2],ymm6[2]
 ; AVX2-ONLY-NEXT:    vperm2f128 $49, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # ymm0 = ymm0[2,3],mem[2,3]
 ; AVX2-ONLY-NEXT:    vmovaps 192(%r8), %ymm1
 ; AVX2-ONLY-NEXT:    vblendps $252, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # ymm2 = ymm1[0,1],mem[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # ymm2 = mem[0,1],ymm1[2,3],mem[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # ymm2 = mem[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = mem[2,3],ymm0[2,3]
-; AVX2-ONLY-NEXT:    vbroadcastsd 240(%rcx), %ymm1
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovaps 224(%r8), %ymm1
-; AVX2-ONLY-NEXT:    vblendps $252, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm2 = ymm1[0,1],mem[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm2 = mem[0,1],ymm1[2,3],mem[4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm2 = mem[0,1,2,3,4,5],ymm1[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 224(%rdx), %ymm0
+; AVX2-ONLY-NEXT:    vperm2f128 $49, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = ymm0[2,3],mem[2,3]
+; AVX2-ONLY-NEXT:    vbroadcastsd 240(%rcx), %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm2[1],ymm0[1],ymm2[3],ymm0[3]
+; AVX2-ONLY-NEXT:    vbroadcastsd 248(%rcx), %ymm2
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm2[2,3]
+; AVX2-ONLY-NEXT:    vmovaps 224(%r8), %ymm2
+; AVX2-ONLY-NEXT:    vblendps $252, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm3 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm3 = ymm2[0,1],mem[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm3 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm3 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm9[0],ymm8[0],ymm9[2],ymm8[2]
 ; AVX2-ONLY-NEXT:    vperm2f128 $49, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # ymm0 = ymm0[2,3],mem[2,3]
 ; AVX2-ONLY-NEXT:    vmovaps 256(%r8), %ymm1
 ; AVX2-ONLY-NEXT:    vblendps $252, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # ymm2 = ymm1[0,1],mem[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # ymm2 = mem[0,1],ymm1[2,3],mem[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # ymm2 = mem[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = mem[2,3],ymm0[2,3]
-; AVX2-ONLY-NEXT:    vbroadcastsd 304(%rcx), %ymm1
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovaps 288(%r8), %ymm1
-; AVX2-ONLY-NEXT:    vblendps $252, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm2 = ymm1[0,1],mem[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm2 = mem[0,1],ymm1[2,3],mem[4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm2 = mem[0,1,2,3,4,5],ymm1[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 288(%rdx), %ymm0
+; AVX2-ONLY-NEXT:    vperm2f128 $49, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = ymm0[2,3],mem[2,3]
+; AVX2-ONLY-NEXT:    vbroadcastsd 304(%rcx), %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm2[1],ymm0[1],ymm2[3],ymm0[3]
+; AVX2-ONLY-NEXT:    vbroadcastsd 312(%rcx), %ymm2
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm2[2,3]
+; AVX2-ONLY-NEXT:    vmovaps 288(%r8), %ymm2
+; AVX2-ONLY-NEXT:    vblendps $252, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm3 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm3 = ymm2[0,1],mem[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm3 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm3 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm11[0],ymm10[0],ymm11[2],ymm10[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm12[0],ymm11[0],ymm12[2],ymm11[2]
 ; AVX2-ONLY-NEXT:    vperm2f128 $49, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # ymm0 = ymm0[2,3],mem[2,3]
 ; AVX2-ONLY-NEXT:    vmovaps 320(%r8), %ymm1
 ; AVX2-ONLY-NEXT:    vblendps $252, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # ymm2 = ymm1[0,1],mem[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # ymm2 = mem[0,1],ymm1[2,3],mem[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vblendps $63, (%rsp), %ymm1, %ymm2 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm2 = mem[0,1,2,3,4,5],ymm1[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = mem[2,3],ymm0[2,3]
-; AVX2-ONLY-NEXT:    vbroadcastsd 368(%rcx), %ymm1
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovaps 352(%r8), %ymm1
-; AVX2-ONLY-NEXT:    vblendps $252, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm2 = ymm1[0,1],mem[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm2 = mem[0,1],ymm1[2,3],mem[4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # ymm2 = mem[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm13[0],ymm12[0],ymm13[2],ymm12[2]
+; AVX2-ONLY-NEXT:    vmovaps 352(%rdx), %ymm0
+; AVX2-ONLY-NEXT:    vperm2f128 $49, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = ymm0[2,3],mem[2,3]
+; AVX2-ONLY-NEXT:    vbroadcastsd 368(%rcx), %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm2[1],ymm0[1],ymm2[3],ymm0[3]
+; AVX2-ONLY-NEXT:    vbroadcastsd 376(%rcx), %ymm2
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm2[2,3]
+; AVX2-ONLY-NEXT:    vmovaps 352(%r8), %ymm2
+; AVX2-ONLY-NEXT:    vblendps $252, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm3 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm3 = ymm2[0,1],mem[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm3 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm3 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm13[0],ymm0[2],ymm13[2]
 ; AVX2-ONLY-NEXT:    vperm2f128 $49, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # ymm0 = ymm0[2,3],mem[2,3]
 ; AVX2-ONLY-NEXT:    vmovaps 384(%r8), %ymm1
 ; AVX2-ONLY-NEXT:    vblendps $252, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # ymm2 = ymm1[0,1],mem[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # ymm2 = mem[0,1],ymm1[2,3],mem[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # ymm2 = mem[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = mem[2,3],ymm0[2,3]
-; AVX2-ONLY-NEXT:    vbroadcastsd 432(%rcx), %ymm1
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovaps 416(%r8), %ymm1
-; AVX2-ONLY-NEXT:    vblendps $252, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm2 = ymm1[0,1],mem[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm2 = mem[0,1],ymm1[2,3],mem[4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm2 = mem[0,1,2,3,4,5],ymm1[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 416(%rdx), %ymm0
+; AVX2-ONLY-NEXT:    vperm2f128 $49, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = ymm0[2,3],mem[2,3]
+; AVX2-ONLY-NEXT:    vbroadcastsd 432(%rcx), %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm2[1],ymm0[1],ymm2[3],ymm0[3]
+; AVX2-ONLY-NEXT:    vbroadcastsd 440(%rcx), %ymm2
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm2[2,3]
+; AVX2-ONLY-NEXT:    vmovaps 416(%r8), %ymm2
+; AVX2-ONLY-NEXT:    vblendps $252, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm3 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm3 = ymm2[0,1],mem[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm3 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm3 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # ymm0 = ymm0[0],mem[0],ymm0[2],mem[2]
 ; AVX2-ONLY-NEXT:    vperm2f128 $49, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # ymm0 = ymm0[2,3],mem[2,3]
 ; AVX2-ONLY-NEXT:    vmovaps 448(%r8), %ymm1
 ; AVX2-ONLY-NEXT:    vblendps $252, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # ymm2 = ymm1[0,1],mem[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # ymm2 = mem[0,1],ymm1[2,3],mem[4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm15[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = mem[2,3],ymm0[2,3]
-; AVX2-ONLY-NEXT:    vbroadcastsd 496(%rcx), %ymm1
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovaps 480(%r8), %ymm1
-; AVX2-ONLY-NEXT:    vblendps $252, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm2 = ymm1[0,1],mem[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm2 = mem[0,1],ymm1[2,3],mem[4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vblendps $63, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm2 = mem[0,1,2,3,4,5],ymm1[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 480(%rdx), %ymm0
+; AVX2-ONLY-NEXT:    vperm2f128 $49, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = ymm0[2,3],mem[2,3]
+; AVX2-ONLY-NEXT:    vbroadcastsd 496(%rcx), %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm2[1],ymm0[1],ymm2[3],ymm0[3]
+; AVX2-ONLY-NEXT:    vbroadcastsd 504(%rcx), %ymm2
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm2[2,3]
+; AVX2-ONLY-NEXT:    vmovaps 480(%r8), %ymm2
+; AVX2-ONLY-NEXT:    vblendps $252, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm3 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm3 = ymm2[0,1],mem[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm3 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm3 = mem[0,1],ymm2[2,3],mem[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovaps 256(%rdi), %xmm0
 ; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
 ; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
 ; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-ONLY-NEXT:    vmovaps 128(%rdi), %xmm0

 ; AVX2-ONLY-NEXT:    vmovaps 320(%rdi), %xmm14
 ; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm14 = xmm14[0],mem[0]
 ; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm4 = xmm0[0],mem[0]
 ; AVX2-ONLY-NEXT:    vmovaps 416(%rdi), %xmm13
 ; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm13 = xmm13[0],mem[0]
 ; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm3 = xmm0[0],mem[0]
-; AVX2-ONLY-NEXT:    vmovaps 480(%rdi), %xmm12
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm12 = xmm12[0],mem[0]
+; AVX2-ONLY-NEXT:    vmovaps 480(%rdi), %xmm11
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm11 = xmm11[0],mem[0]
 ; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm2 = xmm0[0],mem[0]
-; AVX2-ONLY-NEXT:    vmovaps 448(%rdi), %xmm11
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm11 = xmm11[0],mem[0]
+; AVX2-ONLY-NEXT:    vmovaps 448(%rdi), %xmm12
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm12 = xmm12[0],mem[0]
 ; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
 ; AVX2-ONLY-NEXT:    vmovaps 384(%rdi), %xmm10
 ; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm10 = xmm10[0],mem[0]
 ; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
 ; AVX2-ONLY-NEXT:    vmovaps (%rdi), %xmm9
 ; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm9 = xmm9[0],mem[0]
 ; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
 ; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm8 = xmm8[0],mem[0]
 ; AVX2-ONLY-NEXT:    vmovaps %xmm8, 16(%r9)
 ; AVX2-ONLY-NEXT:    vmovaps %xmm9, (%r9)
 ; AVX2-ONLY-NEXT:    vmovaps %xmm0, 1936(%r9)
 ; AVX2-ONLY-NEXT:    vmovaps %xmm10, 1920(%r9)
 ; AVX2-ONLY-NEXT:    vmovaps %xmm1, 2256(%r9)
-; AVX2-ONLY-NEXT:    vmovaps %xmm11, 2240(%r9)
+; AVX2-ONLY-NEXT:    vmovaps %xmm12, 2240(%r9)
 ; AVX2-ONLY-NEXT:    vmovaps %xmm2, 2416(%r9)
-; AVX2-ONLY-NEXT:    vmovaps %xmm12, 2400(%r9)
+; AVX2-ONLY-NEXT:    vmovaps %xmm11, 2400(%r9)
 ; AVX2-ONLY-NEXT:    vmovaps %xmm3, 2096(%r9)
 ; AVX2-ONLY-NEXT:    vmovaps %xmm13, 2080(%r9)
 ; AVX2-ONLY-NEXT:    vmovaps %xmm4, 1616(%r9)
 ; AVX2-ONLY-NEXT:    vmovaps %xmm14, 1600(%r9)
 ; AVX2-ONLY-NEXT:    vmovaps %xmm5, 1776(%r9)
 ; AVX2-ONLY-NEXT:    vmovaps %xmm15, 1760(%r9)
 ; AVX2-ONLY-NEXT:    vmovaps %xmm6, 1456(%r9)
 ; AVX2-ONLY-NEXT:    vmovaps %xmm7, 1440(%r9)

 ; AVX2-ONLY-NEXT:    vmovaps %xmm0, 656(%r9)
 ; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %xmm0, 640(%r9)
 ; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %xmm0, 1296(%r9)
 ; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %xmm0, 1280(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 2528(%r9)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 2496(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 2464(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 2432(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 2336(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 2304(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 2272(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 2208(%r9)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 2176(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 2144(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 2112(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 2016(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1984(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1952(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1888(%r9)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1856(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1824(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1792(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1696(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1664(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1632(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1568(%r9)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1536(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1504(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1472(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1376(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1344(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1312(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1248(%r9)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1216(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1184(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1152(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1056(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1024(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 992(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 928(%r9)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 896(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 864(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 832(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 736(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 704(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 672(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 608(%r9)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 576(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 544(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 512(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 416(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 384(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 352(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 288(%r9)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 256(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 224(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 192(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 96(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload

 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1088(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 768(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 448(%r9)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 128(%r9)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 2528(%r9)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 2208(%r9)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1888(%r9)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1568(%r9)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1248(%r9)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 928(%r9)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 608(%r9)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 288(%r9)
-; AVX2-ONLY-NEXT:    addq $2440, %rsp # imm = 0x988
+; AVX2-ONLY-NEXT:    addq $2392, %rsp # imm = 0x958
 ; AVX2-ONLY-NEXT:    vzeroupper
 ; AVX2-ONLY-NEXT:    retq
 ;
 ; AVX512F-LABEL: store_i64_stride5_vf64:
 ; AVX512F:       # %bb.0:
 ; AVX512F-NEXT:    subq $3080, %rsp # imm = 0xC08
 ; AVX512F-NEXT:    vmovdqa64 320(%rdi), %zmm5
 ; AVX512F-NEXT:    vmovdqa64 256(%rdi), %zmm6

llvm/test/CodeGen/X86/vector-interleaved-store-i64-stride-7.ll

 ; SSE-NEXT:    movaps %xmm1, 192(%rax)
 ; SSE-NEXT:    movaps %xmm5, 208(%rax)
 ; SSE-NEXT:    retq
 ;
 ; AVX1-ONLY-LABEL: store_i64_stride7_vf4:
 ; AVX1-ONLY:       # %bb.0:
 ; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
 ; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %r10
-; AVX1-ONLY-NEXT:    vmovaps (%rdi), %ymm2
-; AVX1-ONLY-NEXT:    vmovaps (%rsi), %ymm3
-; AVX1-ONLY-NEXT:    vmovaps (%r8), %ymm4
-; AVX1-ONLY-NEXT:    vmovaps (%r10), %xmm5
-; AVX1-ONLY-NEXT:    vmovaps 16(%r10), %xmm0
-; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm1
-; AVX1-ONLY-NEXT:    vmovaps 16(%rdi), %xmm6
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm6 = ymm3[1],ymm6[1],ymm3[3],ymm6[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm6[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovapd (%rdi), %ymm3
+; AVX1-ONLY-NEXT:    vmovapd (%rsi), %ymm4
+; AVX1-ONLY-NEXT:    vmovapd (%r8), %ymm5
+; AVX1-ONLY-NEXT:    vmovapd (%r10), %ymm0
 ; AVX1-ONLY-NEXT:    vmovaps (%r8), %xmm6
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm6, %ymm7
-; AVX1-ONLY-NEXT:    vmovaps (%r9), %xmm8
+; AVX1-ONLY-NEXT:    vmovapd (%r10), %xmm7
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm7, %ymm6, %ymm1
+; AVX1-ONLY-NEXT:    vmovapd (%r9), %xmm8
 ; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm9 = xmm8[0,1,0,1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm9, %ymm9
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm9[2,3],ymm7[4,5],ymm9[6,7]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm6
+; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovapd 16(%rdi), %xmm10
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm9, %ymm9
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm9[2,3],ymm1[4,5],ymm9[6,7]
 ; AVX1-ONLY-NEXT:    vmovaps (%rdx), %xmm9
-; AVX1-ONLY-NEXT:    vmovaps (%rsi), %xmm10
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm11 = xmm10[1],xmm9[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 8(%rcx), %ymm12
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3],ymm12[4,5],ymm11[6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm11[0,1,2,3,4,5],ymm6[6,7]
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm4[0],mem[0],ymm4[2],mem[2]
-; AVX1-ONLY-NEXT:    vmovaps 16(%rcx), %xmm11
-; AVX1-ONLY-NEXT:    vmovaps 16(%rdx), %xmm12
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm12 = xmm12[0],xmm11[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm12[0,1,2,3],ymm4[4,5,6,7]
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm2[0],ymm3[0],ymm2[2],ymm3[2]
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm3 = xmm8[1],xmm5[1]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm3 = xmm11[1],mem[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 24(%r9), %ymm5
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm5[4,5],ymm3[6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm10[0]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm9, %ymm6
+; AVX1-ONLY-NEXT:    vmovaps (%rsi), %xmm11
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm12 = xmm11[2,3,2,3]
+; AVX1-ONLY-NEXT:    vbroadcastsd 8(%rcx), %ymm13
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1,2,3],ymm13[4,5],ymm12[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm12[0,1],ymm6[2,3],ymm12[4,5],ymm6[6,7]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm10 = ymm10[0,1],mem[2,3]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm12 = ymm4[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm12 = ymm0[2,3],ymm12[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm10 = ymm12[0],ymm10[1],ymm12[2],ymm10[3]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm3[0],ymm4[0],ymm3[2],ymm4[2]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm4 = xmm8[1],xmm7[1]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm4[0,1],ymm3[2,3]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm5[0],mem[0],ymm5[2],mem[2]
+; AVX1-ONLY-NEXT:    vmovapd 16(%rcx), %xmm5
+; AVX1-ONLY-NEXT:    vmovapd 16(%rdx), %xmm7
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm7 = xmm7[0],xmm5[0]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm4 = ymm7[0,1],ymm4[2,3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm5 = xmm5[1],mem[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 24(%r9), %ymm7
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm5 = ymm5[0,1],ymm7[2],ymm5[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm5[0,1,2],ymm0[3]
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm2[0],xmm11[0]
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm5 = xmm9[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps %ymm2, 96(%rax)
 ; AVX1-ONLY-NEXT:    vmovaps %xmm5, 16(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm1, (%rax)
-; AVX1-ONLY-NEXT:    vmovaps %ymm4, 128(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %ymm3, 192(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, (%rax)
+; AVX1-ONLY-NEXT:    vmovapd %ymm4, 128(%rax)
+; AVX1-ONLY-NEXT:    vmovapd %ymm3, 96(%rax)
+; AVX1-ONLY-NEXT:    vmovapd %ymm10, 160(%rax)
+; AVX1-ONLY-NEXT:    vmovapd %ymm0, 192(%rax)
 ; AVX1-ONLY-NEXT:    vmovaps %ymm6, 64(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %ymm7, 32(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 160(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 32(%rax)
 ; AVX1-ONLY-NEXT:    vzeroupper
 ; AVX1-ONLY-NEXT:    retq
 ;
 ; AVX2-ONLY-LABEL: store_i64_stride7_vf4:
 ; AVX2-ONLY:       # %bb.0:
 ; AVX2-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
 ; AVX2-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %r10
 ; AVX2-ONLY-NEXT:    vmovaps (%rdi), %ymm1

 ; SSE-NEXT:    movaps %xmm0, 16(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, (%rax)
 ; SSE-NEXT:    addq $88, %rsp
 ; SSE-NEXT:    retq
 ;
 ; AVX1-ONLY-LABEL: store_i64_stride7_vf8:
 ; AVX1-ONLY:       # %bb.0:
-; AVX1-ONLY-NEXT:    pushq %rax
 ; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; AVX1-ONLY-NEXT:    vmovaps (%rsi), %ymm3
-; AVX1-ONLY-NEXT:    vmovaps (%rdx), %ymm8
-; AVX1-ONLY-NEXT:    vmovaps (%r9), %ymm4
-; AVX1-ONLY-NEXT:    vmovaps (%rax), %xmm5
-; AVX1-ONLY-NEXT:    vmovaps 16(%rax), %xmm0
-; AVX1-ONLY-NEXT:    vmovapd 32(%rax), %xmm9
-; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm1
-; AVX1-ONLY-NEXT:    vmovaps 16(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovapd (%rdi), %ymm10
+; AVX1-ONLY-NEXT:    vmovapd (%rsi), %ymm1
+; AVX1-ONLY-NEXT:    vmovapd (%r9), %ymm2
+; AVX1-ONLY-NEXT:    vmovapd (%rax), %ymm4
+; AVX1-ONLY-NEXT:    vmovaps (%r8), %xmm6
+; AVX1-ONLY-NEXT:    vmovapd 16(%r8), %xmm9
+; AVX1-ONLY-NEXT:    vmovaps 32(%r8), %xmm3
+; AVX1-ONLY-NEXT:    vmovapd (%rax), %xmm13
+; AVX1-ONLY-NEXT:    vmovapd 32(%rax), %xmm11
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm11, %ymm3, %ymm0
+; AVX1-ONLY-NEXT:    vmovapd (%r9), %xmm14
+; AVX1-ONLY-NEXT:    vmovapd 32(%r9), %xmm12
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm5 = xmm12[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm7
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm2 = ymm3[1],ymm2[1],ymm3[3],ymm2[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm8[6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm7, %ymm5, %ymm5
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm5[2,3],ymm0[4,5],ymm5[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps (%r8), %xmm12
-; AVX1-ONLY-NEXT:    vmovaps 16(%r8), %xmm11
-; AVX1-ONLY-NEXT:    vmovapd 32(%r8), %xmm6
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm2
-; AVX1-ONLY-NEXT:    vmovaps 32(%rdx), %xmm10
-; AVX1-ONLY-NEXT:    vmovaps 32(%rsi), %xmm13
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm14 = xmm13[1],xmm10[1]
+; AVX1-ONLY-NEXT:    vmovaps 32(%rsi), %xmm5
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm8 = xmm5[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vbroadcastsd 40(%rcx), %ymm15
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3],ymm15[4,5],ymm14[6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm14[0,1,2,3,4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],ymm15[4,5],ymm8[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 32(%rdx), %xmm15
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm15, %ymm3
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm8[0,1],ymm3[2,3],ymm8[4,5],ymm3[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm7, %ymm0, %ymm15
-; AVX1-ONLY-NEXT:    vmovapd 32(%r9), %xmm14
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm6 = xmm6[0],xmm14[0]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm9, %ymm6, %ymm6
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3,4,5],ymm15[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm7 = xmm7[0],xmm13[0]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 32(%rcx), %ymm7, %ymm13
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm7, %ymm7
-; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm7[0],ymm13[1],ymm7[2],ymm13[2]
-; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vbroadcastsd 8(%rcx), %ymm10
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],ymm10[4,5],ymm8[6,7]
-; AVX1-ONLY-NEXT:    vmovaps (%rsi), %xmm10
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm13 = xmm10[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm13[0,1],ymm8[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm12, %ymm0, %ymm13
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],ymm13[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 16(%rcx), %xmm13
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm15 = xmm13[2,3,2,3]
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm11 = ymm4[1],ymm11[1],ymm4[3],ymm11[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm15[0,1],ymm11[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps (%r9), %xmm15
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm12, %ymm12
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm15[0,1,0,1]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vmovaps %xmm1, %xmm7
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1],ymm0[2,3],ymm12[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm14[1],xmm9[1]
-; AVX1-ONLY-NEXT:    vmovapd 32(%rdi), %ymm9
-; AVX1-ONLY-NEXT:    vmovapd 32(%rsi), %ymm14
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm9 = ymm9[0],ymm14[0],ymm9[2],ymm14[2]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm9 = ymm0[0,1],ymm9[2,3]
-; AVX1-ONLY-NEXT:    vmovapd 48(%rdi), %xmm0
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],mem[2,3]
-; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm14 = ymm14[0,0,3,2]
-; AVX1-ONLY-NEXT:    vmovapd 32(%rax), %ymm1
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm14 = ymm1[2,3],ymm14[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm14[0],ymm0[1],ymm14[2],ymm0[3]
-; AVX1-ONLY-NEXT:    vmovapd 32(%r8), %ymm14
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm14 = ymm14[0],mem[0],ymm14[2],mem[2]
-; AVX1-ONLY-NEXT:    vmovapd 48(%rcx), %xmm0
-; AVX1-ONLY-NEXT:    vmovapd 48(%rdx), %xmm6
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm6 = xmm6[0],xmm0[0]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm6 = ymm6[0,1],ymm14[2,3]
-; AVX1-ONLY-NEXT:    vmovaps (%rdi), %ymm14
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm14[0],ymm3[0],ymm14[2],ymm3[2]
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm5 = xmm15[1],xmm5[1]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0,1,2,3],ymm3[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps (%r8), %ymm5
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm5[0],ymm4[0],ymm5[2],ymm4[2]
-; AVX1-ONLY-NEXT:    vmovaps 16(%rdx), %xmm5
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm5 = xmm5[0],xmm13[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3],ymm4[4,5,6,7]
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],mem[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 56(%r9), %ymm5
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm5[2],ymm0[3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3]
+; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm8
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm5 = xmm7[0],xmm5[0]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 32(%rcx), %ymm5, %ymm7
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm5, %ymm5
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm5 = ymm5[0],ymm7[1],ymm5[2],ymm7[2]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm13, %ymm6, %ymm7
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm15 = xmm14[0,1,0,1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm15, %ymm15
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1],ymm15[2,3],ymm7[4,5],ymm15[6,7]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm9 = ymm2[0],ymm9[1],ymm2[3],ymm9[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm15 = ymm9[0,1,2],ymm4[3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm9 = mem[2,3],ymm9[2,3]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm9 = ymm9[1],ymm15[1],ymm9[2],ymm15[3]
+; AVX1-ONLY-NEXT:    vmovapd 32(%rsi), %ymm15
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm13 = xmm14[1],xmm13[1]
+; AVX1-ONLY-NEXT:    vmovapd 32(%rax), %ymm14
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm10 = ymm10[0],ymm1[0],ymm10[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm10 = ymm13[0,1],ymm10[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 48(%rdi), %xmm13
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm13 = ymm13[0,1],mem[2,3]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm0 = ymm15[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm14[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm13 = ymm0[0],ymm13[1],ymm0[2],ymm13[3]
+; AVX1-ONLY-NEXT:    vmovapd 32(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm15[0],ymm0[2],ymm15[2]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm11 = xmm12[1],xmm11[1]
+; AVX1-ONLY-NEXT:    vmovapd 32(%r8), %ymm12
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm11 = ymm11[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm12[0],mem[0],ymm12[2],mem[2]
+; AVX1-ONLY-NEXT:    vmovapd 48(%rcx), %xmm12
+; AVX1-ONLY-NEXT:    vmovapd 48(%rdx), %xmm15
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm15 = xmm15[0],xmm12[0]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm15[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm15 = mem[0,0]
+; AVX1-ONLY-NEXT:    vbroadcastsd 8(%rcx), %ymm3
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm15[0,1],ymm3[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovapd (%rdx), %ymm15
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm15, %ymm6
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm6[2,3],ymm3[4,5],ymm6[6,7]
+; AVX1-ONLY-NEXT:    vmovapd 16(%rdi), %xmm6
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm6 = ymm6[0,1],ymm15[2,3]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm1[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm4[2,3],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0],ymm6[1],ymm1[2],ymm6[3]
+; AVX1-ONLY-NEXT:    vmovapd (%r8), %ymm4
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm4[0],ymm2[0],ymm4[2],ymm2[2]
+; AVX1-ONLY-NEXT:    vmovapd 16(%rdx), %xmm4
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm4 = xmm4[0],mem[0]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm4[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm4 = xmm12[1],mem[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 56(%r9), %ymm6
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm4 = ymm4[0,1],ymm6[2],ymm4[3]
 ; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm7[0],xmm10[0]
-; AVX1-ONLY-NEXT:    vmovaps (%rdx), %xmm5
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm5 = xmm5[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps %xmm5, 16(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm1, (%rax)
-; AVX1-ONLY-NEXT:    vmovaps %ymm4, 128(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %ymm3, 96(%rax)
-; AVX1-ONLY-NEXT:    vmovapd %ymm6, 352(%rax)
-; AVX1-ONLY-NEXT:    vmovapd %ymm2, 384(%rax)
-; AVX1-ONLY-NEXT:    vmovapd %ymm9, 320(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %ymm12, 32(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %ymm11, 192(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %ymm8, 64(%rax)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm1, 224(%rax)
-; AVX1-ONLY-NEXT:    vmovapd %ymm0, 416(%rax)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 256(%rax)
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm4 = ymm4[0,1,2],ymm14[3]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm6 = xmm8[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps (%rdx), %xmm8
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm8 = xmm8[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 16(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm6, (%rax)
+; AVX1-ONLY-NEXT:    vmovapd %ymm2, 128(%rax)
+; AVX1-ONLY-NEXT:    vmovapd %ymm1, 160(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm3, 64(%rax)
+; AVX1-ONLY-NEXT:    vmovapd %ymm0, 352(%rax)
+; AVX1-ONLY-NEXT:    vmovapd %ymm11, 320(%rax)
+; AVX1-ONLY-NEXT:    vmovapd %ymm13, 384(%rax)
+; AVX1-ONLY-NEXT:    vmovapd %ymm10, 96(%rax)
+; AVX1-ONLY-NEXT:    vmovapd %ymm9, 192(%rax)
+; AVX1-ONLY-NEXT:    vmovapd %ymm4, 416(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm7, 32(%rax)
+; AVX1-ONLY-NEXT:    vmovapd %ymm5, 224(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 288(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 160(%rax)
-; AVX1-ONLY-NEXT:    popq %rax
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 256(%rax)
 ; AVX1-ONLY-NEXT:    vzeroupper
 ; AVX1-ONLY-NEXT:    retq
 ;
 ; AVX2-ONLY-LABEL: store_i64_stride7_vf8:
 ; AVX2-ONLY:       # %bb.0:
+; AVX2-ONLY-NEXT:    pushq %rax
 ; AVX2-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; AVX2-ONLY-NEXT:    vmovaps 32(%rdi), %ymm10
-; AVX2-ONLY-NEXT:    vmovaps (%rdi), %ymm6
-; AVX2-ONLY-NEXT:    vmovaps (%rsi), %ymm8
-; AVX2-ONLY-NEXT:    vmovaps 32(%rsi), %ymm11
-; AVX2-ONLY-NEXT:    vmovaps 32(%rdx), %ymm12
-; AVX2-ONLY-NEXT:    vmovaps (%rdx), %ymm7
-; AVX2-ONLY-NEXT:    vmovaps (%r8), %ymm1
-; AVX2-ONLY-NEXT:    vmovaps 32(%r8), %ymm5
-; AVX2-ONLY-NEXT:    vmovaps (%r9), %ymm2
-; AVX2-ONLY-NEXT:    vmovaps (%rax), %xmm4
-; AVX2-ONLY-NEXT:    vmovaps 16(%rax), %xmm0
-; AVX2-ONLY-NEXT:    vmovaps 32(%rax), %xmm13
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm3 = ymm6[1],ymm8[1],ymm6[3],ymm8[3]
+; AVX2-ONLY-NEXT:    vmovaps (%rdi), %ymm8
+; AVX2-ONLY-NEXT:    vmovaps 32(%rdi), %ymm7
+; AVX2-ONLY-NEXT:    vmovaps (%rsi), %ymm10
+; AVX2-ONLY-NEXT:    vmovaps 32(%rsi), %ymm9
+; AVX2-ONLY-NEXT:    vmovaps (%rdx), %ymm2
+; AVX2-ONLY-NEXT:    vmovaps 32(%rdx), %ymm6
+; AVX2-ONLY-NEXT:    vmovaps 32(%rdi), %xmm4
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm0, %ymm0
+; AVX2-ONLY-NEXT:    vmovaps 32(%r9), %xmm13
+; AVX2-ONLY-NEXT:    vmovaps 32(%r8), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm3 = xmm1[0],xmm13[0]
+; AVX2-ONLY-NEXT:    vmovaps 16(%rax), %xmm5
+; AVX2-ONLY-NEXT:    vmovaps 32(%rax), %xmm14
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm14, %ymm3, %ymm3
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 32(%rdx), %xmm11
+; AVX2-ONLY-NEXT:    vmovaps 32(%rsi), %xmm12
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm3 = xmm12[1],xmm11[1]
+; AVX2-ONLY-NEXT:    vbroadcastsd 40(%rcx), %ymm15
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm15[4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 48(%rax), %xmm15
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm3 = ymm7[1],ymm9[1],ymm7[3],ymm9[3]
 ; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,2,3,3]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm7[6,7]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm3[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm6[6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1],ymm3[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 32(%r8), %xmm15
-; AVX2-ONLY-NEXT:    vmovaps 32(%rdx), %xmm9
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm3 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm3 = xmm3[1],xmm9[1]
-; AVX2-ONLY-NEXT:    vbroadcastsd 40(%rcx), %ymm14
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm14[4,5],ymm3[6,7]
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm0, %ymm14
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm14[6,7]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm5 = ymm5[0],mem[0],ymm5[2],mem[2]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm14 = ymm12[0],mem[0],ymm12[2],mem[2]
-; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm5 = ymm14[2,3],ymm5[2,3]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm6 = ymm6[0],ymm8[0],ymm6[2],ymm8[2]
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm8 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm8 = xmm8[1],xmm4[1]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm8[0,1,2,3],ymm6[4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovaps 32(%rdi), %xmm14
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm14, %ymm0, %ymm8
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm15 = xmm15[0],mem[0]
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm13, %ymm15, %ymm15
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm15[0,1,2,3,4,5],ymm8[6,7]
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm15 = ymm10[1],ymm11[1],ymm10[3],ymm11[3]
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm15 = ymm8[1],ymm10[1],ymm8[3],ymm10[3]
 ; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm15 = ymm15[0,2,3,3]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm12 = ymm15[0,1,2,3,4,5],ymm12[6,7]
-; AVX2-ONLY-NEXT:    vmovaps 48(%rax), %xmm15
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm12 = ymm15[0,1],ymm12[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovaps (%r8), %xmm15
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm10 = ymm10[0],ymm11[0],ymm10[2],ymm11[2]
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm11 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm11 = xmm11[1],xmm13[1]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm11[0,1,2,3],ymm10[4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm11 = mem[0,0]
-; AVX2-ONLY-NEXT:    vbroadcastsd 8(%rcx), %ymm13
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm7[0,1,2,3],ymm13[4,5],ymm7[6,7]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1],ymm13[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm0, %ymm13
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3,4,5],ymm13[6,7]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm13 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm7 = ymm7[0],mem[0],ymm7[2],mem[2]
-; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm7 = ymm7[2,3],ymm13[2,3]
-; AVX2-ONLY-NEXT:    vbroadcastsd 56(%rcx), %ymm13
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm13 = xmm13[0,1],mem[2,3]
-; AVX2-ONLY-NEXT:    vbroadcastsd 56(%r9), %ymm0
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm13[0,1,2,3],ymm0[4,5],ymm13[6,7]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm13 = xmm14[0],mem[0]
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm9, %ymm13, %ymm9
-; AVX2-ONLY-NEXT:    vbroadcastsd 32(%rcx), %ymm13
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3,4,5],ymm13[6,7]
-; AVX2-ONLY-NEXT:    vmovaps (%rdi), %xmm13
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm14 = xmm15[0],mem[0]
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm13, %ymm14, %ymm14
-; AVX2-ONLY-NEXT:    vbroadcastsd %xmm4, %ymm4
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm14[0,1,2,3],ymm4[4,5],ymm14[6,7]
-; AVX2-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rdi
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm2[1],ymm1[3],ymm2[3]
-; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,3,3]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],mem[6,7]
-; AVX2-ONLY-NEXT:    vbroadcastsd 24(%rcx), %ymm2
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovaps (%rdx), %xmm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm15 = ymm15[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0,1],ymm15[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps (%r9), %xmm15
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm8 = ymm8[0],ymm10[0],ymm8[2],ymm10[2]
+; AVX2-ONLY-NEXT:    vmovaps (%rax), %xmm10
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm15[1],xmm10[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm0[0,1,2,3],ymm8[4,5,6,7]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm7[0],ymm9[0],ymm7[2],ymm9[2]
+; AVX2-ONLY-NEXT:    vmovaps 32(%rcx), %ymm9
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm7 = xmm13[1],xmm14[1]
+; AVX2-ONLY-NEXT:    vmovaps 32(%r8), %ymm13
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm0[4,5,6,7]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm13[0],mem[0],ymm13[2],mem[2]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm6 = ymm6[0],ymm9[0],ymm6[2],ymm9[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm6 = ymm6[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vbroadcastsd 8(%rcx), %ymm0
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm14 = mem[0,0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm14[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovaps (%r8), %xmm14
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm14, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],mem[0],ymm2[2],mem[2]
+; AVX2-ONLY-NEXT:    vmovaps (%r8), %ymm2
+; AVX2-ONLY-NEXT:    vmovaps (%r9), %ymm0
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm5 = ymm2[0],ymm0[0],ymm2[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3],ymm5[2,3]
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm4 = xmm4[0],xmm12[0]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm11, %ymm4, %ymm4
+; AVX2-ONLY-NEXT:    vbroadcastsd 32(%rcx), %ymm5
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm5[6,7]
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm5 = xmm14[0],xmm15[0]
+; AVX2-ONLY-NEXT:    vmovaps (%rdi), %xmm11
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm11, %ymm5, %ymm5
+; AVX2-ONLY-NEXT:    vbroadcastsd %xmm10, %ymm10
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm10[4,5],ymm5[6,7]
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm2[1],ymm0[1],ymm2[3],ymm0[3]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,3,3]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm13 = xmm13[0],mem[0]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm2 = xmm2[0],mem[0]
-; AVX2-ONLY-NEXT:    vmovaps %xmm2, 16(%rdi)
-; AVX2-ONLY-NEXT:    vmovaps %xmm13, (%rdi)
-; AVX2-ONLY-NEXT:    vmovaps %ymm7, 128(%rdi)
-; AVX2-ONLY-NEXT:    vmovaps %ymm11, 64(%rdi)
-; AVX2-ONLY-NEXT:    vmovaps %ymm10, 320(%rdi)
-; AVX2-ONLY-NEXT:    vmovaps %ymm1, 192(%rdi)
-; AVX2-ONLY-NEXT:    vmovaps %ymm12, 384(%rdi)
-; AVX2-ONLY-NEXT:    vmovaps %ymm8, 256(%rdi)
-; AVX2-ONLY-NEXT:    vmovaps %ymm4, 32(%rdi)
-; AVX2-ONLY-NEXT:    vmovaps %ymm6, 96(%rdi)
-; AVX2-ONLY-NEXT:    vmovaps %ymm5, 352(%rdi)
-; AVX2-ONLY-NEXT:    vmovaps %ymm9, 224(%rdi)
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 416(%rdi)
-; AVX2-ONLY-NEXT:    vmovaps %ymm3, 288(%rdi)
+; AVX2-ONLY-NEXT:    vbroadcastsd 24(%rcx), %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rdi
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm2 = ymm9[1],ymm13[1],ymm9[3],ymm13[3]
+; AVX2-ONLY-NEXT:    vbroadcastsd 56(%r9), %ymm9
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm2[2,3],ymm9[2,3]
+; AVX2-ONLY-NEXT:    vmovaps (%rdx), %xmm9
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],mem[6,7]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm10 = xmm11[0],mem[0]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm9 = xmm9[0],mem[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm9, 16(%rdi)
+; AVX2-ONLY-NEXT:    vmovaps %xmm10, (%rdi)
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 128(%rdi)
+; AVX2-ONLY-NEXT:    vmovaps %ymm3, 64(%rdi)
+; AVX2-ONLY-NEXT:    vmovaps %ymm6, 352(%rdi)
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 320(%rdi)
+; AVX2-ONLY-NEXT:    vmovaps %ymm2, 416(%rdi)
+; AVX2-ONLY-NEXT:    vmovaps %ymm8, 96(%rdi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 160(%rdi)
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 192(%rdi)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 160(%rdi)
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 384(%rdi)
+; AVX2-ONLY-NEXT:    vmovaps %ymm5, 32(%rdi)
+; AVX2-ONLY-NEXT:    vmovaps %ymm4, 224(%rdi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 288(%rdi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 256(%rdi)
+; AVX2-ONLY-NEXT:    popq %rax
 ; AVX2-ONLY-NEXT:    vzeroupper
 ; AVX2-ONLY-NEXT:    retq
 ;
 ; AVX512F-ONLY-SLOW-LABEL: store_i64_stride7_vf8:
 ; AVX512F-ONLY-SLOW:       # %bb.0:
 ; AVX512F-ONLY-SLOW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
 ; AVX512F-ONLY-SLOW-NEXT:    movq {{[0-9]+}}(%rsp), %r10
 ; AVX512F-ONLY-SLOW-NEXT:    vmovdqa64 (%rdi), %zmm0

 ; SSE-NEXT:    movaps %xmm0, 16(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, (%rax)
 ; SSE-NEXT:    addq $536, %rsp # imm = 0x218
 ; SSE-NEXT:    retq
 ;
 ; AVX1-ONLY-LABEL: store_i64_stride7_vf16:
 ; AVX1-ONLY:       # %bb.0:
-; AVX1-ONLY-NEXT:    subq $552, %rsp # imm = 0x228
+; AVX1-ONLY-NEXT:    subq $408, %rsp # imm = 0x198
 ; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; AVX1-ONLY-NEXT:    vmovaps (%rsi), %ymm9
-; AVX1-ONLY-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps (%rdx), %ymm0
-; AVX1-ONLY-NEXT:    vmovaps (%r9), %ymm10
-; AVX1-ONLY-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps (%r8), %xmm3
-; AVX1-ONLY-NEXT:    vmovaps 16(%r8), %xmm1
-; AVX1-ONLY-NEXT:    vmovaps (%rax), %xmm2
-; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 16(%rax), %xmm4
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm3, %ymm5
-; AVX1-ONLY-NEXT:    vmovaps (%r9), %xmm2
-; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = xmm2[0,1,0,1]
-; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm8
+; AVX1-ONLY-NEXT:    vmovapd (%r9), %ymm7
+; AVX1-ONLY-NEXT:    vmovupd %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd (%rax), %ymm6
+; AVX1-ONLY-NEXT:    vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps (%r8), %xmm15
+; AVX1-ONLY-NEXT:    vmovapd 16(%r8), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps 32(%r8), %xmm0
+; AVX1-ONLY-NEXT:    vmovapd (%rax), %xmm14
+; AVX1-ONLY-NEXT:    vmovaps 32(%rax), %xmm8
 ; AVX1-ONLY-NEXT:    vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 16(%rdi), %xmm7
-; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm2
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm6, %ymm6
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm6[2,3],ymm5[4,5],ymm6[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vbroadcastsd 8(%rcx), %ymm5
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm0[0,1,2,3],ymm5[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vmovaps (%rsi), %xmm6
-; AVX1-ONLY-NEXT:    vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm6 = xmm6[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm6[0,1],ymm5[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm5[0,1,2,3,4,5],ymm3[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm3 = ymm9[1],ymm7[1],ymm9[3],ymm7[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1],ymm0[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 16(%rcx), %xmm14
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm14[2,3,2,3]
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm10[1],ymm1[1],ymm10[3],ymm1[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 32(%rsi), %xmm0
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm0[0]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 32(%rcx), %ymm1, %ymm3
-; AVX1-ONLY-NEXT:    vmovaps 32(%rdx), %xmm4
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm1, %ymm1
-; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm1[0],ymm3[1],ymm1[2],ymm3[2]
-; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 32(%r8), %xmm1
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX1-ONLY-NEXT:    vmovaps 32(%rax), %xmm5
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm14, %ymm15, %ymm2
+; AVX1-ONLY-NEXT:    vmovapd (%r9), %xmm13
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm13[0,1,0,1]
+; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm5
 ; AVX1-ONLY-NEXT:    vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 32(%r9), %xmm13
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm3 = xmm1[0],xmm13[0]
+; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm4
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm3, %ymm3
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2,3],ymm2[4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm7[0],ymm1[1],ymm7[3],ymm1[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm1[0,1,2],ymm6[3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = mem[2,3],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm1[1],ymm2[1],ymm1[2],ymm2[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 32(%rsi), %xmm1
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm4[0],xmm1[0]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 32(%rcx), %ymm2, %ymm3
+; AVX1-ONLY-NEXT:    vmovaps 32(%rdx), %xmm5
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm2, %ymm2
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm2 = ymm2[0],ymm3[1],ymm2[2],ymm3[2]
+; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 32(%r9), %xmm7
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm7[0,1,0,1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm2, %ymm2
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm0, %ymm3
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3],ymm3[4,5],ymm2[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm4[1]
+; AVX1-ONLY-NEXT:    vmovapd 32(%r9), %ymm6
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm5, %ymm0
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
 ; AVX1-ONLY-NEXT:    vbroadcastsd 40(%rcx), %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 32(%rsi), %ymm11
-; AVX1-ONLY-NEXT:    vmovaps 48(%rdi), %xmm0
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm11[1],ymm0[1],ymm11[3],ymm0[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 48(%rax), %xmm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 32(%r9), %ymm10
-; AVX1-ONLY-NEXT:    vmovaps 48(%r8), %xmm0
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm10[1],ymm0[1],ymm10[3],ymm0[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 48(%rcx), %xmm8
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm8[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vmovapd 32(%rax), %ymm11
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 64(%r9), %xmm9
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm9[0,1,0,1]
+; AVX1-ONLY-NEXT:    vmovapd 48(%r8), %xmm0
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm6[0],ymm0[1],ymm6[3],ymm0[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm0[0,1,2],ymm11[3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = mem[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[2],ymm1[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 64(%r9), %xmm10
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm10[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vmovaps 64(%r8), %xmm1
-; AVX1-ONLY-NEXT:    vmovaps 64(%rax), %xmm7
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm7, %ymm1, %ymm2
+; AVX1-ONLY-NEXT:    vmovapd 64(%rax), %xmm5
+; AVX1-ONLY-NEXT:    vmovaps 64(%r8), %xmm4
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm4, %ymm3
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1],ymm0[2,3],ymm3[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 64(%r9), %ymm3
+; AVX1-ONLY-NEXT:    vmovapd 80(%r8), %xmm0
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm3[0],ymm0[1],ymm3[3],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovapd 64(%rax), %ymm12
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm8 = ymm0[0,1,2],ymm12[3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = mem[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[1],ymm8[1],ymm0[2],ymm8[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 96(%rsi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %xmm8
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm9 = xmm8[0],xmm1[0]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 96(%rcx), %ymm9, %ymm2
+; AVX1-ONLY-NEXT:    vmovaps 96(%rdx), %xmm0
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm9, %ymm9
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm2 = ymm9[0],ymm2[1],ymm9[2],ymm2[2]
+; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX1-ONLY-NEXT:    vbroadcastsd 104(%rcx), %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 96(%r8), %xmm1
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 64(%rdx), %ymm0
-; AVX1-ONLY-NEXT:    vbroadcastsd 72(%rcx), %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3],ymm2[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 64(%rsi), %xmm12
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm12[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 64(%rsi), %ymm6
-; AVX1-ONLY-NEXT:    vmovaps 80(%rdi), %xmm1
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm6[1],ymm1[1],ymm6[3],ymm1[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 80(%rax), %xmm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 64(%r9), %ymm5
-; AVX1-ONLY-NEXT:    vmovaps 80(%r8), %xmm0
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm5[1],ymm0[1],ymm5[3],ymm0[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 80(%rcx), %xmm4
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm15 = xmm4[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1],ymm0[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 96(%rsi), %xmm2
-; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %xmm15
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm0 = xmm15[0],xmm2[0]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 96(%rcx), %ymm0, %ymm3
-; AVX1-ONLY-NEXT:    vmovaps 96(%rdx), %xmm1
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[0],ymm3[1],ymm0[2],ymm3[2]
+; AVX1-ONLY-NEXT:    vmovapd 112(%rcx), %xmm9
+; AVX1-ONLY-NEXT:    vmovapd 112(%rdx), %xmm0
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm9[0],ymm0[2],ymm9[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],mem[2],ymm0[3]
+; AVX1-ONLY-NEXT:    vbroadcastsd 112(%r9), %ymm2
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vmovaps 96(%r8), %xmm3
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm15 = xmm3[0],mem[0]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 96(%rax), %ymm15, %ymm15
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3,4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, (%rsp) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm2[1],xmm1[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 104(%rcx), %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
+; AVX1-ONLY-NEXT:    vbroadcastsd 8(%rcx), %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovapd (%rdx), %ymm2
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm2, %ymm15
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm15[2,3],ymm0[4,5],ymm15[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm13[1],xmm14[1]
+; AVX1-ONLY-NEXT:    vmovapd (%rdi), %ymm13
+; AVX1-ONLY-NEXT:    vmovapd (%rsi), %ymm15
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm13 = ymm13[0],ymm15[0],ymm13[2],ymm15[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm13[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps (%r8), %ymm0
+; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0],mem[0],ymm0[2],mem[2]
+; AVX1-ONLY-NEXT:    vmovaps 16(%rdx), %xmm13
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm13 = xmm13[0],mem[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm13[0,1,2,3],ymm0[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 112(%rcx), %xmm0
-; AVX1-ONLY-NEXT:    vmovapd 112(%rdx), %xmm1
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1],mem[2],ymm1[3]
-; AVX1-ONLY-NEXT:    vbroadcastsd 112(%r9), %ymm2
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm15 = ymm1[0,1,2],ymm2[3]
-; AVX1-ONLY-NEXT:    vmovaps (%rdi), %ymm1
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm1 = ymm1[0],mem[0],ymm1[2],mem[2]
-; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm2 = xmm2[1],mem[1]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps (%r8), %ymm2
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm2 = ymm2[0],mem[0],ymm2[2],mem[2]
-; AVX1-ONLY-NEXT:    vmovaps 16(%rdx), %xmm3
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm3 = xmm3[0],xmm14[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm3[0,1,2,3],ymm2[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %ymm2
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm2[0],ymm11[0],ymm2[2],ymm11[2]
-; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm3 = xmm13[1],mem[1]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps 32(%r8), %ymm3
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm3[0],ymm10[0],ymm3[2],ymm10[2]
-; AVX1-ONLY-NEXT:    vmovaps 48(%rdx), %xmm10
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm8 = xmm10[0],xmm8[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3],ymm3[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %ymm3
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm3[0],ymm6[0],ymm3[2],ymm6[2]
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm6 = xmm9[1],xmm7[1]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm6[0,1,2,3],ymm3[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps 64(%r8), %ymm6
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm5 = ymm6[0],ymm5[0],ymm6[2],ymm5[2]
-; AVX1-ONLY-NEXT:    vmovaps 80(%rdx), %xmm6
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm4 = xmm6[0],xmm4[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovapd 96(%rdi), %ymm5
-; AVX1-ONLY-NEXT:    vmovapd 96(%rsi), %ymm6
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm5 = ymm5[0],ymm6[0],ymm5[2],ymm6[2]
-; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm7 = mem[0,0]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm5 = ymm7[0],ymm5[1,2,3]
-; AVX1-ONLY-NEXT:    vmovapd 112(%rdi), %xmm7
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm7 = ymm7[0,1],mem[2,3]
-; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm6 = ymm6[0,0,3,2]
-; AVX1-ONLY-NEXT:    vmovapd 96(%rax), %ymm9
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm6 = ymm9[2,3],ymm6[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm6 = ymm6[0],ymm7[1],ymm6[2],ymm7[3]
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],mem[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 120(%r9), %ymm7
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm7[2],ymm0[3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm5 = ymm5[0],ymm9[1],ymm5[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm9[3]
-; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm7 = xmm7[0],xmm12[0]
-; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm9, %xmm9 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm9 = xmm9[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovapd 16(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm2 = ymm15[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 $19, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm2 = mem[2,3],ymm2[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm13 = ymm2[0],ymm0[1],ymm2[2],ymm0[3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm0 = xmm7[1],mem[1]
+; AVX1-ONLY-NEXT:    vmovapd 32(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovapd 32(%rsi), %ymm7
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm2[0],ymm7[0],ymm2[2],ymm7[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm15 = ymm0[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 32(%r8), %ymm0
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm6[0],ymm0[2],ymm6[2]
+; AVX1-ONLY-NEXT:    vmovapd 48(%rdx), %xmm2
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm2 = xmm2[0],mem[0]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm6 = ymm2[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm0 = ymm7[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm11[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 48(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm2[0,1],mem[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm7 = ymm0[0],ymm2[1],ymm0[2],ymm2[3]
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
+; AVX1-ONLY-NEXT:    vbroadcastsd 72(%rcx), %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovapd 64(%rdx), %ymm11
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm11, %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3],ymm0[4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm2 = xmm10[1],xmm5[1]
+; AVX1-ONLY-NEXT:    vmovapd 64(%rdi), %ymm4
+; AVX1-ONLY-NEXT:    vmovapd 64(%rsi), %ymm5
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm4[0],ymm5[0],ymm4[2],ymm5[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm2[0,1],ymm4[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 64(%r8), %ymm4
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm4[0],ymm3[0],ymm4[2],ymm3[2]
+; AVX1-ONLY-NEXT:    vmovapd 80(%rdx), %xmm4
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm4 = xmm4[0],mem[0]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm4[0,1],ymm3[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 80(%rdi), %xmm4
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm4 = ymm4[0,1],ymm11[2,3]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm5 = ymm5[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm5 = ymm12[2,3],ymm5[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm5 = ymm5[0],ymm4[1],ymm5[2],ymm4[3]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = mem[0,1,0,1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm4, %ymm4
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 96(%rax), %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm4[2,3],ymm1[4,5],ymm4[6,7]
+; AVX1-ONLY-NEXT:    vmovapd 96(%rdi), %ymm4
+; AVX1-ONLY-NEXT:    vmovapd 96(%rsi), %ymm8
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm4[0],ymm8[0],ymm4[2],ymm8[2]
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm10 = mem[0,0]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm4 = ymm10[0],ymm4[1,2,3]
+; AVX1-ONLY-NEXT:    vmovapd 112(%rdi), %xmm10
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm10 = ymm10[0,1],mem[2,3]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm8 = ymm8[0,0,3,2]
+; AVX1-ONLY-NEXT:    vmovapd 96(%rax), %ymm11
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm8 = ymm11[2,3],ymm8[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm8 = ymm8[0],ymm10[1],ymm8[2],ymm10[3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm9 = xmm9[1],mem[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 120(%r9), %ymm10
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm9 = ymm9[0,1],ymm10[2],ymm9[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm4 = ymm4[0],ymm11[1],ymm4[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm9 = ymm9[0,1,2],ymm11[3]
 ; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; AVX1-ONLY-NEXT:    vmovaps 64(%rdx), %xmm10
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm10 = xmm10[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps (%rdx), %xmm11
+; AVX1-ONLY-NEXT:    vmovaps 64(%rdx), %xmm11
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm11 = xmm11[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps %xmm11, 16(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm9, (%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm10, 464(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm7, 448(%rax)
-; AVX1-ONLY-NEXT:    vmovapd %ymm6, 832(%rax)
-; AVX1-ONLY-NEXT:    vmovapd %ymm5, 768(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %ymm4, 576(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %ymm3, 544(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %ymm8, 352(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %ymm2, 320(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %ymm14, 128(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %ymm1, 96(%rax)
-; AVX1-ONLY-NEXT:    vmovapd %ymm0, 864(%rax)
-; AVX1-ONLY-NEXT:    vmovapd %ymm15, 800(%rax)
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm12 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm12 = xmm12[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps (%rdx), %xmm14
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm14 = xmm14[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps %xmm14, 16(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm12, (%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm11, 464(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm10, 448(%rax)
+; AVX1-ONLY-NEXT:    vmovapd %ymm8, 832(%rax)
+; AVX1-ONLY-NEXT:    vmovapd %ymm4, 768(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm1, 704(%rax)
+; AVX1-ONLY-NEXT:    vmovapd %ymm5, 608(%rax)
+; AVX1-ONLY-NEXT:    vmovapd %ymm3, 576(%rax)
+; AVX1-ONLY-NEXT:    vmovapd %ymm2, 544(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 512(%rax)
+; AVX1-ONLY-NEXT:    vmovapd %ymm7, 384(%rax)
+; AVX1-ONLY-NEXT:    vmovapd %ymm6, 352(%rax)
+; AVX1-ONLY-NEXT:    vmovapd %ymm15, 320(%rax)
+; AVX1-ONLY-NEXT:    vmovapd %ymm13, 160(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 128(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 96(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 64(%rax)
+; AVX1-ONLY-NEXT:    vmovapd %ymm9, 864(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 800(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 736(%rax)
-; AVX1-ONLY-NEXT:    vmovups (%rsp), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 704(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 672(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 640(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 608(%rax)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 512(%rax)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 480(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 416(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 384(%rax)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 288(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 256(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 224(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 192(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 160(%rax)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 64(%rax)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 32(%rax)
-; AVX1-ONLY-NEXT:    addq $552, %rsp # imm = 0x228
+; AVX1-ONLY-NEXT:    addq $408, %rsp # imm = 0x198
 ; AVX1-ONLY-NEXT:    vzeroupper
 ; AVX1-ONLY-NEXT:    retq
 ;
 ; AVX2-ONLY-LABEL: store_i64_stride7_vf16:
 ; AVX2-ONLY:       # %bb.0:
-; AVX2-ONLY-NEXT:    subq $424, %rsp # imm = 0x1A8
+; AVX2-ONLY-NEXT:    subq $520, %rsp # imm = 0x208
 ; AVX2-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; AVX2-ONLY-NEXT:    vmovaps 64(%rdi), %ymm5
-; AVX2-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 32(%rdi), %ymm1
-; AVX2-ONLY-NEXT:    vmovaps (%rdi), %ymm3
-; AVX2-ONLY-NEXT:    vmovaps (%rsi), %ymm15
-; AVX2-ONLY-NEXT:    vmovaps 32(%rsi), %ymm8
-; AVX2-ONLY-NEXT:    vmovaps 64(%rsi), %ymm13
+; AVX2-ONLY-NEXT:    vmovaps 32(%rdi), %ymm6
+; AVX2-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps (%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovaps (%rsi), %ymm13
+; AVX2-ONLY-NEXT:    vmovaps 32(%rsi), %ymm7
+; AVX2-ONLY-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovaps 32(%rdx), %ymm11
-; AVX2-ONLY-NEXT:    vmovaps (%rdx), %ymm9
-; AVX2-ONLY-NEXT:    vmovaps 16(%rax), %xmm0
-; AVX2-ONLY-NEXT:    vmovaps 32(%rax), %xmm7
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm2 = ymm3[1],ymm15[1],ymm3[3],ymm15[3]
+; AVX2-ONLY-NEXT:    vmovaps (%rdx), %ymm4
+; AVX2-ONLY-NEXT:    vmovaps 16(%rax), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps 32(%rax), %xmm9
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm2 = ymm0[1],ymm13[1],ymm0[3],ymm13[3]
 ; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,2,3,3]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm9[6,7]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 32(%rdi), %xmm0
-; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm4[6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 32(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vmovaps 32(%r9), %xmm3
 ; AVX2-ONLY-NEXT:    vmovaps 32(%r8), %xmm2
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm4 = xmm2[0],mem[0]
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm7, %ymm4, %ymm4
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3,4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 32(%rdx), %xmm4
-; AVX2-ONLY-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm4[1]
-; AVX2-ONLY-NEXT:    vbroadcastsd 40(%rcx), %ymm4
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm4[4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT:    vmovaps 48(%rax), %xmm4
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm5 = xmm2[0],xmm3[0]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm9, %ymm5, %ymm5
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm5[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 32(%rdx), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, (%rsp) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 32(%rsi), %xmm5
+; AVX2-ONLY-NEXT:    vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm5[1],xmm1[1]
+; AVX2-ONLY-NEXT:    vbroadcastsd 40(%rcx), %ymm5
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm5[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 48(%rax), %xmm5
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm8[1],ymm1[3],ymm8[3]
-; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,3,3]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm11[6,7]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1],ymm0[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm5[1],ymm13[1],ymm5[3],ymm13[3]
-; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm2 = ymm0[0,2,3,3]
-; AVX2-ONLY-NEXT:    vmovaps 64(%rdx), %ymm0
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT:    vmovaps 80(%rax), %xmm4
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm4[0,1],ymm2[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 96(%r8), %xmm2
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm4 = xmm2[0],mem[0]
-; AVX2-ONLY-NEXT:    vinsertf128 $1, 96(%rax), %ymm4, %ymm4
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm6[1],ymm7[1],ymm6[3],ymm7[3]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,3,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm11[6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm5[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 64(%rdi), %ymm8
+; AVX2-ONLY-NEXT:    vmovaps 64(%rsi), %ymm12
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm8[1],ymm12[1],ymm8[3],ymm12[3]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,3,3]
+; AVX2-ONLY-NEXT:    vmovaps 64(%rdx), %ymm7
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm7[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 80(%rax), %xmm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 96(%r8), %xmm1
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm2 = xmm1[0],mem[0]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, 96(%rax), %ymm2, %ymm2
 ; AVX2-ONLY-NEXT:    vmovaps 96(%rdi), %xmm5
 ; AVX2-ONLY-NEXT:    vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; 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:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 96(%rdx), %xmm5
-; AVX2-ONLY-NEXT:    vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm4 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm4 = xmm4[1],xmm5[1]
-; AVX2-ONLY-NEXT:    vbroadcastsd 104(%rcx), %ymm5
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5],ymm4[6,7]
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm4[0,1,2,3,4,5],ymm2[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 96(%rdi), %ymm6
-; AVX2-ONLY-NEXT:    vmovaps 96(%rsi), %ymm10
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm2 = ymm6[1],ymm10[1],ymm6[3],ymm10[3]
-; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,2,3,3]
-; AVX2-ONLY-NEXT:    vmovaps 96(%rdx), %ymm14
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm14[6,7]
-; AVX2-ONLY-NEXT:    vmovaps 112(%rax), %xmm4
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm4[0,1],ymm2[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vbroadcastsd 8(%rcx), %ymm2
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm9[0,1,2,3],ymm2[4,5],ymm9[6,7]
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm4 = mem[0,0]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm4[0,1],ymm2[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovaps (%r8), %xmm12
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm12, %ymm0, %ymm5
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm5[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm3[0],ymm15[0],ymm3[2],ymm15[2]
-; AVX2-ONLY-NEXT:    vmovaps (%rax), %xmm5
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm3 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm3 = xmm3[1],xmm5[1]
-; 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:    vunpcklpd {{.*#+}} ymm9 = ymm9[0],mem[0],ymm9[2],mem[2]
-; AVX2-ONLY-NEXT:    vmovaps (%r8), %ymm4
-; AVX2-ONLY-NEXT:    vmovaps (%r9), %ymm3
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm15 = ymm4[0],ymm3[0],ymm4[2],ymm3[2]
-; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm9[2,3],ymm15[2,3]
+; AVX2-ONLY-NEXT:    vmovaps 96(%rdx), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 96(%rsi), %xmm5
+; AVX2-ONLY-NEXT:    vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm2 = xmm5[1],xmm2[1]
+; AVX2-ONLY-NEXT:    vbroadcastsd 104(%rcx), %ymm5
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm5[4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 96(%rdi), %ymm10
+; AVX2-ONLY-NEXT:    vmovaps 96(%rsi), %ymm6
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm10[1],ymm6[1],ymm10[3],ymm6[3]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,3,3]
+; AVX2-ONLY-NEXT:    vmovaps 96(%rdx), %ymm2
 ; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm8[0],ymm1[2],ymm8[2]
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm8 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm7 = xmm8[1],xmm7[1]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm7[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 112(%rax), %xmm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm11[0],mem[0],ymm11[2],mem[2]
-; AVX2-ONLY-NEXT:    vmovaps 32(%r8), %ymm8
-; AVX2-ONLY-NEXT:    vmovaps 32(%r9), %ymm7
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm9 = ymm8[0],ymm7[0],ymm8[2],ymm7[2]
-; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3],ymm9[2,3]
-; AVX2-ONLY-NEXT:    vmovups %ymm1, (%rsp) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vbroadcastsd 72(%rcx), %ymm1
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm9 = mem[0,0]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm9[0,1],ymm1[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovaps 64(%r8), %xmm11
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm11, %ymm0, %ymm9
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm9[6,7]
+; AVX2-ONLY-NEXT:    vbroadcastsd 8(%rcx), %ymm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm4[0,1,2,3],ymm1[4,5],ymm4[6,7]
+; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm2 = mem[0,0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovaps (%r8), %xmm14
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm14, %ymm0, %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm13[0],ymm1[2],ymm13[2]
-; AVX2-ONLY-NEXT:    vmovaps 64(%rax), %xmm9
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm13[0],ymm0[2],ymm13[2]
+; AVX2-ONLY-NEXT:    vmovaps (%rax), %xmm13
+; AVX2-ONLY-NEXT:    vmovaps (%r9), %xmm5
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm2 = xmm5[1],xmm13[1]
+; 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:    vunpcklpd {{.*#+}} ymm4 = ymm4[0],mem[0],ymm4[2],mem[2]
+; AVX2-ONLY-NEXT:    vmovaps (%r8), %ymm2
+; AVX2-ONLY-NEXT:    vmovaps (%r9), %ymm1
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm15 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm4[2,3],ymm15[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm4 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm4 = ymm0[0],mem[0],ymm0[2],mem[2]
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm3 = xmm3[1],xmm9[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm4[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm9 = ymm11[0],mem[0],ymm11[2],mem[2]
+; AVX2-ONLY-NEXT:    vmovaps 32(%r8), %ymm4
+; AVX2-ONLY-NEXT:    vmovaps 32(%r9), %ymm3
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm15 = ymm4[0],ymm3[0],ymm4[2],ymm3[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm9[2,3],ymm15[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastsd 72(%rcx), %ymm9
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm7[0,1,2,3],ymm9[4,5],ymm7[6,7]
 ; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm15 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm15 = xmm15[1],xmm9[1]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm15[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:    vunpcklpd {{.*#+}} ymm15 = ymm0[0],mem[0],ymm0[2],mem[2]
-; AVX2-ONLY-NEXT:    vmovaps 64(%r8), %ymm2
-; AVX2-ONLY-NEXT:    vmovaps 64(%r9), %ymm1
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm13 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
-; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm15 = ymm15[2,3],ymm13[2,3]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm6 = ymm6[0],ymm10[0],ymm6[2],ymm10[2]
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm10 = mem[0,0]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm10[0,1],ymm6[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vbroadcastsd 120(%rcx), %ymm10
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm10 = xmm10[0,1],mem[2,3]
-; AVX2-ONLY-NEXT:    vbroadcastsd 120(%r9), %ymm13
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm10[0,1,2,3],ymm13[4,5],ymm10[6,7]
-; AVX2-ONLY-NEXT:    vmovaps 96(%rax), %ymm0
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm6[0,1],ymm0[2,3],ymm6[4,5,6,7]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm12[0],mem[0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm15 = ymm15[0,1],ymm9[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovaps 64(%r8), %xmm9
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm9, %ymm0, %ymm11
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1,2,3,4,5],ymm11[6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm11 = ymm8[0],ymm12[0],ymm8[2],ymm12[2]
+; AVX2-ONLY-NEXT:    vmovaps 64(%r9), %xmm15
+; AVX2-ONLY-NEXT:    vmovaps 64(%rax), %xmm8
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm12 = xmm15[1],xmm8[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm12[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:    vunpcklpd {{.*#+}} ymm7 = ymm7[0],mem[0],ymm7[2],mem[2]
+; AVX2-ONLY-NEXT:    vmovaps 64(%r8), %ymm11
+; AVX2-ONLY-NEXT:    vmovaps 64(%r9), %ymm0
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm12 = ymm11[0],ymm0[0],ymm11[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm12 = ymm7[2,3],ymm12[2,3]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm6 = ymm10[0],ymm6[0],ymm10[2],ymm6[2]
+; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm7 = mem[0,0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm7[0,1],ymm6[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm5 = xmm14[0],xmm5[0]
 ; AVX2-ONLY-NEXT:    vmovaps (%rdi), %xmm6
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm0
-; AVX2-ONLY-NEXT:    vbroadcastsd %xmm5, %ymm5
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm12 = ymm0[0,1,2,3],ymm5[4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm4[1],ymm3[1],ymm4[3],ymm3[3]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm5, %ymm5
+; AVX2-ONLY-NEXT:    vbroadcastsd %xmm13, %ymm7
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm7[4,5],ymm5[6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm2[1],ymm1[1],ymm2[3],ymm1[3]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,3,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],mem[6,7]
+; AVX2-ONLY-NEXT:    vbroadcastsd 24(%rcx), %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm2[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm1 = xmm1[0],mem[0]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, (%rsp), %ymm1, %ymm1 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    vbroadcastsd 32(%rcx), %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm1[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm4[1],ymm3[1],ymm4[3],ymm3[3]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,3,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],mem[6,7]
+; AVX2-ONLY-NEXT:    vbroadcastsd 56(%rcx), %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm2[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm9[0],xmm15[0]
+; AVX2-ONLY-NEXT:    vmovaps 64(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
+; AVX2-ONLY-NEXT:    vbroadcastsd %xmm8, %ymm3
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm1[0,1,2,3],ymm3[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm11[1],ymm0[1],ymm11[3],ymm0[3]
 ; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,3,3]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
-; AVX2-ONLY-NEXT:    vbroadcastsd 24(%rcx), %ymm3
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vbroadcastsd 88(%rcx), %ymm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm15 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
+; AVX2-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm0 = xmm0[0],mem[0]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload
-; AVX2-ONLY-NEXT:    vbroadcastsd 32(%rcx), %ymm4
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm4[6,7]
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm4 = ymm8[1],ymm7[1],ymm8[3],ymm7[3]
-; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm4 = ymm4[0,2,3,3]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],mem[6,7]
-; AVX2-ONLY-NEXT:    vbroadcastsd 56(%rcx), %ymm5
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1],ymm4[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm5 = xmm11[0],mem[0]
-; AVX2-ONLY-NEXT:    vmovaps 64(%rdi), %xmm8
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm5, %ymm5
-; AVX2-ONLY-NEXT:    vbroadcastsd %xmm9, %ymm7
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm5[0,1,2,3],ymm7[4,5],ymm5[6,7]
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm2[1],ymm1[1],ymm2[3],ymm1[3]
-; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,3,3]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],mem[6,7]
-; AVX2-ONLY-NEXT:    vbroadcastsd 88(%rcx), %ymm2
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm2 = xmm2[0],mem[0]
-; AVX2-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 16-byte Folded Reload
-; AVX2-ONLY-NEXT:    vbroadcastsd 96(%rcx), %ymm5
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm5[6,7]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm5 = ymm14[0],mem[0],ymm14[2],mem[2]
-; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm5 = ymm5[2,3],mem[2,3]
-; AVX2-ONLY-NEXT:    vbroadcastsd 112(%r9), %ymm9
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],ymm9[6,7]
+; AVX2-ONLY-NEXT:    vbroadcastsd 96(%rcx), %ymm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 96(%rcx), %ymm1
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm3[0],ymm1[0],ymm3[2],ymm1[2]
+; AVX2-ONLY-NEXT:    vmovaps 96(%r8), %ymm5
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm3[2,3],ymm5[2,3]
+; AVX2-ONLY-NEXT:    vbroadcastsd 112(%r9), %ymm8
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm8[6,7]
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm5[1],ymm1[3],ymm5[3]
+; AVX2-ONLY-NEXT:    vbroadcastsd 120(%r9), %ymm5
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3],ymm5[2,3]
+; AVX2-ONLY-NEXT:    vmovaps 96(%rax), %ymm5
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm10[0,1],ymm5[2,3],ymm10[4,5,6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm5[6,7]
 ; AVX2-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm8 = xmm8[0],mem[0]
-; AVX2-ONLY-NEXT:    vmovaps 64(%rdx), %xmm9
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm9 = xmm9[0],mem[0]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm2 = xmm2[0],mem[0]
+; AVX2-ONLY-NEXT:    vmovaps 64(%rdx), %xmm5
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm5 = xmm5[0],mem[0]
 ; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm6 = xmm6[0],mem[0]
-; AVX2-ONLY-NEXT:    vmovaps (%rdx), %xmm11
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm11 = xmm11[0],mem[0]
-; AVX2-ONLY-NEXT:    vmovaps %xmm11, 16(%rax)
+; AVX2-ONLY-NEXT:    vmovaps (%rdx), %xmm9
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm9 = xmm9[0],mem[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm9, 16(%rax)
 ; AVX2-ONLY-NEXT:    vmovaps %xmm6, (%rax)
-; AVX2-ONLY-NEXT:    vmovaps %xmm9, 464(%rax)
-; AVX2-ONLY-NEXT:    vmovaps %xmm8, 448(%rax)
-; AVX2-ONLY-NEXT:    vmovaps %ymm5, 800(%rax)
-; AVX2-ONLY-NEXT:    vmovaps %ymm10, 768(%rax)
-; AVX2-ONLY-NEXT:    vmovaps %ymm15, 576(%rax)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm5, 544(%rax)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm5, 512(%rax)
-; AVX2-ONLY-NEXT:    vmovups (%rsp), %ymm5 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm5, 352(%rax)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm5, 320(%rax)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm5, 128(%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, 64(%rax)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm5, 832(%rax)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm5, 736(%rax)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm5, 704(%rax)
-; AVX2-ONLY-NEXT:    vmovaps %ymm2, 672(%rax)
-; AVX2-ONLY-NEXT:    vmovaps %ymm1, 640(%rax)
+; AVX2-ONLY-NEXT:    vmovaps %xmm5, 464(%rax)
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, 448(%rax)
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 864(%rax)
+; AVX2-ONLY-NEXT:    vmovaps %ymm3, 800(%rax)
+; AVX2-ONLY-NEXT:    vmovaps %ymm8, 768(%rax)
+; AVX2-ONLY-NEXT:    vmovaps %ymm12, 576(%rax)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm1, 608(%rax)
-; AVX2-ONLY-NEXT:    vmovaps %ymm7, 480(%rax)
-; AVX2-ONLY-NEXT:    vmovaps %ymm4, 416(%rax)
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 544(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 512(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 352(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 320(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 128(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 96(%rax)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm1, 384(%rax)
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 64(%rax)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm1, 288(%rax)
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 832(%rax)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm1, 256(%rax)
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 224(%rax)
-; AVX2-ONLY-NEXT:    vmovaps %ymm3, 192(%rax)
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 736(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm1, 704(%rax)
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 672(%rax)
+; AVX2-ONLY-NEXT:    vmovaps %ymm15, 640(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 608(%rax)
+; AVX2-ONLY-NEXT:    vmovaps %ymm13, 480(%rax)
+; AVX2-ONLY-NEXT:    vmovaps %ymm4, 416(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 384(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 288(%rax)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 256(%rax)
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 224(%rax)
+; AVX2-ONLY-NEXT:    vmovaps %ymm14, 192(%rax)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 160(%rax)
-; AVX2-ONLY-NEXT:    vmovaps %ymm12, 32(%rax)
-; AVX2-ONLY-NEXT:    vmovaps %ymm13, 864(%rax)
-; AVX2-ONLY-NEXT:    addq $424, %rsp # imm = 0x1A8
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 32(%rax)
+; AVX2-ONLY-NEXT:    addq $520, %rsp # imm = 0x208
 ; AVX2-ONLY-NEXT:    vzeroupper
 ; AVX2-ONLY-NEXT:    retq
 ;
 ; AVX512F-ONLY-SLOW-LABEL: store_i64_stride7_vf16:
 ; AVX512F-ONLY-SLOW:       # %bb.0:
 ; AVX512F-ONLY-SLOW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
 ; AVX512F-ONLY-SLOW-NEXT:    vmovdqa64 (%rdi), %zmm0
 ; AVX512F-ONLY-SLOW-NEXT:    vmovdqa64 64(%rdi), %zmm6

 ; SSE-NEXT:    movaps %xmm0, 16(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, (%rax)
 ; SSE-NEXT:    addq $1432, %rsp # imm = 0x598
 ; SSE-NEXT:    retq
 ;
 ; AVX1-ONLY-LABEL: store_i64_stride7_vf32:
 ; AVX1-ONLY:       # %bb.0:
-; AVX1-ONLY-NEXT:    subq $1320, %rsp # imm = 0x528
+; AVX1-ONLY-NEXT:    subq $1384, %rsp # imm = 0x568
 ; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; AVX1-ONLY-NEXT:    vmovaps (%rdi), %ymm3
-; AVX1-ONLY-NEXT:    vmovaps (%rsi), %ymm1
-; AVX1-ONLY-NEXT:    vmovaps (%rdx), %ymm2
-; AVX1-ONLY-NEXT:    vmovaps (%r8), %ymm7
-; AVX1-ONLY-NEXT:    vmovaps (%r9), %ymm4
-; AVX1-ONLY-NEXT:    vmovaps (%r8), %xmm9
-; AVX1-ONLY-NEXT:    vmovaps 16(%r8), %xmm6
-; AVX1-ONLY-NEXT:    vmovaps 32(%r8), %xmm0
-; AVX1-ONLY-NEXT:    vmovaps (%rax), %xmm10
-; AVX1-ONLY-NEXT:    vmovaps 16(%rax), %xmm8
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm9, %ymm5
-; AVX1-ONLY-NEXT:    vmovaps (%r9), %xmm11
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm12 = xmm11[0,1,0,1]
-; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm13
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm13, %ymm12, %ymm12
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm12[2,3],ymm5[4,5],ymm12[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vbroadcastsd 8(%rcx), %ymm5
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm2[0,1,2,3],ymm5[4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT:    vmovaps (%rsi), %xmm12
-; AVX1-ONLY-NEXT:    vmovaps %xmm12, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm12 = xmm12[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm12 = ymm12[0,1],ymm5[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovaps 32(%rax), %xmm5
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm9, %ymm0, %ymm9
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm12[0,1,2,3,4,5],ymm9[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 16(%rdi), %xmm9
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm3[0],ymm1[0],ymm3[2],ymm1[2]
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm10 = xmm11[1],xmm10[1]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm10[0,1,2,3],ymm3[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm7[0],ymm4[0],ymm7[2],ymm4[2]
-; AVX1-ONLY-NEXT:    vmovaps 16(%rcx), %xmm7
-; AVX1-ONLY-NEXT:    vmovaps 16(%rdx), %xmm10
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm10 = xmm10[0],xmm7[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm10[0,1,2,3],ymm3[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovapd (%r9), %ymm6
+; AVX1-ONLY-NEXT:    vmovupd %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd (%rax), %ymm7
+; AVX1-ONLY-NEXT:    vmovupd %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps (%r8), %xmm3
 ; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm3
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm9[1],ymm1[3],ymm9[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm8[0,1],ymm1[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm7[2,3,2,3]
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm2 = ymm4[1],ymm6[1],ymm4[3],ymm6[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 16(%r8), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps 32(%r8), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps (%rax), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 32(%rax), %xmm8
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm3, %ymm2
+; AVX1-ONLY-NEXT:    vmovaps (%r9), %xmm3
+; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[0,1,0,1]
+; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm5
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm4
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm3, %ymm3
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2,3],ymm2[4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm6[0],ymm1[1],ymm6[3],ymm1[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm1[0,1,2],ymm7[3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = mem[2,3],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm1[1],ymm2[1],ymm1[2],ymm2[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rsi), %xmm1
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm3[0],xmm1[0]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 32(%rcx), %ymm2, %ymm4
-; AVX1-ONLY-NEXT:    vmovaps 32(%rdx), %xmm6
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm2, %ymm2
-; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm2 = ymm2[0],ymm4[1],ymm2[2],ymm4[2]
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm4[0],xmm1[0]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 32(%rcx), %ymm2, %ymm3
+; AVX1-ONLY-NEXT:    vmovaps 32(%rdx), %xmm5
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm2, %ymm2
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm2 = ymm2[0],ymm3[1],ymm2[2],ymm3[2]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm2
-; AVX1-ONLY-NEXT:    vmovaps 32(%r9), %xmm3
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm4 = xmm0[0],xmm3[0]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm4, %ymm4
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm4[0,1,2,3,4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 32(%r9), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,0,1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm2, %ymm2
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm0, %ymm3
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3],ymm3[4,5],ymm2[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %ymm2
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm6[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 40(%rcx), %ymm4
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm4[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 32(%rsi), %ymm4
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 32(%r8), %ymm0
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm3[1],xmm5[1]
-; AVX1-ONLY-NEXT:    vmovaps 32(%r9), %ymm3
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm2[0],ymm4[0],ymm2[2],ymm4[2]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 48(%rcx), %xmm1
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[2],ymm3[2]
-; AVX1-ONLY-NEXT:    vmovaps 48(%rdx), %xmm2
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm2[0],xmm1[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 48(%rdi), %xmm0
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm4[1],ymm0[1],ymm4[3],ymm0[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 48(%rax), %xmm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 48(%r8), %xmm0
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm3[1],ymm0[1],ymm3[3],ymm0[3]
+; AVX1-ONLY-NEXT:    vmovapd 32(%r9), %ymm3
+; AVX1-ONLY-NEXT:    vmovupd %ymm3, (%rsp) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm5, %ymm0
 ; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vbroadcastsd 40(%rcx), %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vmovapd 32(%rax), %ymm2
+; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 48(%r8), %xmm0
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm3[0],ymm0[1],ymm3[3],ymm0[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm0[0,1,2],ymm2[3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = mem[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[2],ymm1[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 64(%r9), %xmm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[0,1,0,1]
-; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %xmm2
-; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,0,1]
+; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovaps 64(%rax), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 64(%r8), %xmm2
-; AVX1-ONLY-NEXT:    vmovaps 64(%rax), %xmm3
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm2, %ymm4
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm4[0,1],ymm1[2,3],ymm4[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 64(%rdx), %ymm1
-; AVX1-ONLY-NEXT:    vbroadcastsd 72(%rcx), %ymm4
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm1[0,1,2,3],ymm4[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 64(%rsi), %xmm11
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm5 = xmm11[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1],ymm4[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm4[0,1,2,3,4,5],ymm2[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm3[1]
-; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %ymm2
-; AVX1-ONLY-NEXT:    vmovaps 64(%rsi), %ymm3
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm2[0],ymm3[0],ymm2[2],ymm3[2]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 64(%r8), %ymm0
-; AVX1-ONLY-NEXT:    vmovaps 64(%r9), %ymm2
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm2[0],ymm0[2],ymm2[2]
-; AVX1-ONLY-NEXT:    vmovaps 80(%rcx), %xmm4
-; AVX1-ONLY-NEXT:    vmovaps 80(%rdx), %xmm5
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm5 = xmm5[0],xmm4[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 80(%rdi), %xmm0
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm3[1],ymm0[1],ymm3[3],ymm0[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 80(%rax), %xmm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 80(%r8), %xmm0
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm2[1],ymm0[1],ymm2[3],ymm0[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm4[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm2, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 64(%r9), %ymm1
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 80(%r8), %xmm0
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[3],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovapd 64(%rax), %ymm1
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm0[0,1,2],ymm1[3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = mem[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[2],ymm1[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rsi), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, 96(%rcx), %ymm2, %ymm3
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rdx), %xmm4
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm2, %ymm2
 ; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm2 = ymm2[0],ymm3[1],ymm2[2],ymm3[2]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vmovaps 96(%r9), %xmm2
-; AVX1-ONLY-NEXT:    vmovaps 96(%r8), %xmm3
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm5 = xmm3[0],xmm2[0]
-; AVX1-ONLY-NEXT:    vmovaps 96(%rax), %xmm6
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm5, %ymm5
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm5[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,0,1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm2, %ymm1
+; AVX1-ONLY-NEXT:    vmovaps 96(%r8), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps 96(%rax), %xmm3
+; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm2, %ymm3
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3],ymm3[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm4[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 104(%rcx), %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm2[1],xmm6[1]
-; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %ymm1
-; AVX1-ONLY-NEXT:    vmovaps 96(%rsi), %ymm2
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 96(%r8), %ymm0
-; AVX1-ONLY-NEXT:    vmovaps 96(%r9), %ymm1
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
-; AVX1-ONLY-NEXT:    vmovaps 112(%rcx), %xmm3
-; AVX1-ONLY-NEXT:    vmovaps 112(%rdx), %xmm4
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm4 = xmm4[0],xmm3[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 112(%rdi), %xmm0
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm2[1],ymm0[1],ymm2[3],ymm0[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 112(%rax), %xmm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 112(%r8), %xmm0
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm3[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm4, %ymm1
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX1-ONLY-NEXT:    vbroadcastsd 104(%rcx), %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 96(%r9), %ymm12
+; AVX1-ONLY-NEXT:    vmovapd 112(%r8), %xmm0
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm12[0],ymm0[1],ymm12[3],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovapd 96(%rax), %ymm1
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm0[0,1,2],ymm1[3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = mem[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[2],ymm1[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 128(%r9), %xmm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[0,1,0,1]
-; AVX1-ONLY-NEXT:    vmovaps 128(%rdi), %xmm12
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm12, %ymm1, %ymm1
-; AVX1-ONLY-NEXT:    vmovaps 128(%r8), %xmm2
-; AVX1-ONLY-NEXT:    vmovaps 128(%rax), %xmm3
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm2, %ymm4
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm4[0,1],ymm1[2,3],ymm4[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 128(%rdx), %ymm1
-; AVX1-ONLY-NEXT:    vbroadcastsd 136(%rcx), %ymm4
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm1[0,1,2,3],ymm4[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 128(%rsi), %xmm10
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm5 = xmm10[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1],ymm4[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm4[0,1,2,3,4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm3[1]
-; AVX1-ONLY-NEXT:    vmovaps 128(%rdi), %ymm2
-; AVX1-ONLY-NEXT:    vmovaps 128(%rsi), %ymm3
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm2[0],ymm3[0],ymm2[2],ymm3[2]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 128(%r8), %ymm0
-; AVX1-ONLY-NEXT:    vmovaps 128(%r9), %ymm2
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm2[0],ymm0[2],ymm2[2]
-; AVX1-ONLY-NEXT:    vmovaps 144(%rcx), %xmm4
-; AVX1-ONLY-NEXT:    vmovaps 144(%rdx), %xmm5
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm5 = xmm5[0],xmm4[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm5[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 144(%rdi), %xmm0
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm3[1],ymm0[1],ymm3[3],ymm0[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 144(%rax), %xmm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 144(%r8), %xmm0
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm2[1],ymm0[1],ymm2[3],ymm0[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm4[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,0,1]
+; AVX1-ONLY-NEXT:    vmovaps 128(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovaps 128(%r8), %xmm13
+; AVX1-ONLY-NEXT:    vmovapd 128(%rax), %xmm14
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm14, %ymm13, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 128(%r9), %ymm1
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 144(%r8), %xmm0
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[3],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovapd 128(%rax), %ymm1
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm0[0,1,2],ymm1[3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = mem[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[2],ymm1[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 160(%rsi), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps 160(%rdi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, 160(%rcx), %ymm2, %ymm3
-; AVX1-ONLY-NEXT:    vmovaps 160(%rdx), %xmm4
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm2, %ymm2
+; AVX1-ONLY-NEXT:    vmovaps 160(%rdx), %xmm5
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm2, %ymm2
 ; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm2 = ymm2[0],ymm3[1],ymm2[2],ymm3[2]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vmovaps 160(%r9), %xmm2
-; AVX1-ONLY-NEXT:    vmovaps 160(%r8), %xmm3
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm5 = xmm3[0],xmm2[0]
-; AVX1-ONLY-NEXT:    vmovaps 160(%rax), %xmm6
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm5, %ymm5
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm5[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vmovapd 160(%r9), %xmm11
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm11[0,1,0,1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm2, %ymm1
+; AVX1-ONLY-NEXT:    vmovapd 160(%rax), %xmm7
+; AVX1-ONLY-NEXT:    vmovaps 160(%r8), %xmm2
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm7, %ymm2, %ymm3
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3],ymm3[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm4[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 168(%rcx), %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm2[1],xmm6[1]
-; AVX1-ONLY-NEXT:    vmovaps 160(%rdi), %ymm1
-; AVX1-ONLY-NEXT:    vmovaps 160(%rsi), %ymm2
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 160(%r8), %ymm0
-; AVX1-ONLY-NEXT:    vmovaps 160(%r9), %ymm1
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
-; AVX1-ONLY-NEXT:    vmovaps 176(%rcx), %xmm3
-; AVX1-ONLY-NEXT:    vmovaps 176(%rdx), %xmm4
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm4 = xmm4[0],xmm3[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm5, %ymm1
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX1-ONLY-NEXT:    vbroadcastsd 168(%rcx), %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 176(%rdi), %xmm0
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm2[1],ymm0[1],ymm2[3],ymm0[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 176(%rax), %xmm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovapd 160(%r9), %ymm5
+; AVX1-ONLY-NEXT:    vmovapd 176(%r8), %xmm1
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm5[0],ymm1[1],ymm5[3],ymm1[2]
+; AVX1-ONLY-NEXT:    vmovapd 160(%rax), %ymm9
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm1[0,1,2],ymm9[3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = mem[2,3],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm1[1],ymm2[1],ymm1[2],ymm2[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 192(%r9), %xmm10
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm10[0,1,0,1]
+; AVX1-ONLY-NEXT:    vmovaps 192(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm2
+; AVX1-ONLY-NEXT:    vmovaps 192(%r8), %xmm4
+; AVX1-ONLY-NEXT:    vmovapd 192(%rax), %xmm6
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm4, %ymm15
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1],ymm2[2,3],ymm15[4,5],ymm2[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 176(%r8), %xmm0
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm3[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 224(%rsi), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps 224(%rdi), %xmm15
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm0 = xmm15[0],xmm2[0]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 224(%rcx), %ymm0, %ymm3
+; AVX1-ONLY-NEXT:    vmovaps 224(%rdx), %xmm1
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[0],ymm3[1],ymm0[2],ymm3[2]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm2[2,3,2,3]
+; AVX1-ONLY-NEXT:    vbroadcastsd 232(%rcx), %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 224(%r8), %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 192(%r9), %xmm0
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[0,1,0,1]
-; AVX1-ONLY-NEXT:    vmovaps 192(%rdi), %xmm9
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm9, %ymm1, %ymm1
-; AVX1-ONLY-NEXT:    vmovaps 192(%r8), %xmm4
-; AVX1-ONLY-NEXT:    vmovaps 192(%rax), %xmm5
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm4, %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3],ymm2[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 240(%rcx), %xmm1
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 192(%rdx), %ymm1
-; AVX1-ONLY-NEXT:    vbroadcastsd 200(%rcx), %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 192(%rsi), %xmm8
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm7 = xmm8[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm7[0,1],ymm6[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm0, %ymm4
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm6[0,1,2,3,4,5],ymm4[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm2, (%rsp) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm5[1]
-; AVX1-ONLY-NEXT:    vmovaps 192(%rdi), %ymm4
-; AVX1-ONLY-NEXT:    vmovaps 192(%rsi), %ymm5
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm4[0],ymm5[0],ymm4[2],ymm5[2]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm4[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 240(%rdx), %xmm0
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vmovaps 224(%r8), %ymm8
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm8[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vbroadcastsd 240(%r9), %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 192(%r8), %ymm0
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm0[0],mem[0],ymm0[2],mem[2]
-; AVX1-ONLY-NEXT:    vmovaps 208(%rcx), %xmm7
-; AVX1-ONLY-NEXT:    vmovaps 208(%rdx), %xmm6
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm6 = xmm6[0],xmm7[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm6[0,1,2,3],ymm4[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
+; AVX1-ONLY-NEXT:    vbroadcastsd 8(%rcx), %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovapd (%rdx), %ymm1
+; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm3 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm3[2,3],ymm0[4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm2 = xmm0[1],mem[1]
+; AVX1-ONLY-NEXT:    vmovapd (%rdi), %ymm3
+; AVX1-ONLY-NEXT:    vmovapd (%rsi), %ymm0
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm3[0],ymm0[0],ymm3[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm2[0,1],ymm3[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps (%r8), %ymm2
+; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm2, %ymm2 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm2 = ymm2[0],mem[0],ymm2[2],mem[2]
+; AVX1-ONLY-NEXT:    vmovaps 16(%rdx), %xmm3
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm3 = xmm3[0],mem[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3],ymm2[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 16(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm0 = ymm0[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 $19, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = mem[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm0 = xmm0[1],mem[1]
+; AVX1-ONLY-NEXT:    vmovapd 32(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovapd 32(%rsi), %ymm2
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 32(%r8), %ymm0
+; AVX1-ONLY-NEXT:    vunpcklpd (%rsp), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0],mem[0],ymm0[2],mem[2]
+; AVX1-ONLY-NEXT:    vmovaps 48(%rdx), %xmm1
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm1[0],mem[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 208(%rdi), %xmm4
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm4 = ymm5[1],ymm4[1],ymm5[3],ymm4[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm4[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 208(%rax), %xmm4
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1],ymm1[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm0 = ymm2[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 $19, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = mem[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 48(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1],mem[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
+; AVX1-ONLY-NEXT:    vbroadcastsd 72(%rcx), %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovapd 64(%rdx), %ymm1
+; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm2 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3],ymm0[4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm0 = xmm0[1],mem[1]
+; AVX1-ONLY-NEXT:    vmovapd 64(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovapd 64(%rsi), %ymm3
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm2[0],ymm3[0],ymm2[2],ymm3[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 64(%r8), %ymm0
+; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0],mem[0],ymm0[2],mem[2]
+; AVX1-ONLY-NEXT:    vmovaps 80(%rdx), %xmm2
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm2 = xmm2[0],mem[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 224(%rsi), %xmm1
-; AVX1-ONLY-NEXT:    vmovaps 224(%rdi), %xmm4
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm5 = xmm4[0],xmm1[0]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 224(%rcx), %ymm5, %ymm6
-; AVX1-ONLY-NEXT:    vmovaps 224(%rdx), %xmm0
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm5, %ymm5
-; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm6 = ymm5[0],ymm6[1],ymm5[2],ymm6[2]
-; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm5 = mem[0,0]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm5, %ymm4
-; AVX1-ONLY-NEXT:    vmovapd 224(%r8), %ymm2
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 224(%rax), %ymm2, %ymm5
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1],ymm4[2,3],ymm5[4,5],ymm4[6,7]
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
-; AVX1-ONLY-NEXT:    vbroadcastsd 232(%rcx), %ymm4
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm4[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 224(%r8), %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vmovapd 224(%rdi), %ymm0
-; AVX1-ONLY-NEXT:    vmovapd 224(%rsi), %ymm1
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
-; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm15 = mem[0,0]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm15 = ymm15[0],ymm0[1,2,3]
-; AVX1-ONLY-NEXT:    vmovapd 240(%rcx), %xmm3
-; AVX1-ONLY-NEXT:    vmovapd 240(%rdx), %xmm14
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm14 = ymm14[0],ymm3[0],ymm14[2],ymm3[2]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm14[0,1],ymm2[2],ymm14[3]
-; AVX1-ONLY-NEXT:    vbroadcastsd 240(%r9), %ymm14
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm2[0,1,2],ymm14[3]
-; AVX1-ONLY-NEXT:    vmovapd 240(%rdi), %xmm14
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm14 = ymm14[0,1],mem[2,3]
-; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm1[0,0,3,2]
-; AVX1-ONLY-NEXT:    vmovapd 224(%rax), %ymm0
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm0[2,3],ymm1[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0],ymm14[1],ymm1[2],ymm14[3]
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm7 = xmm7[1],mem[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 216(%r9), %ymm14
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3],ymm14[4,5],ymm7[6,7]
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm3 = xmm3[1],mem[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 248(%r9), %ymm14
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm3[0,1],ymm14[2],ymm3[3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm14 = ymm15[0],ymm0[1],ymm15[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm3[0,1,2],ymm0[3]
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm3 = xmm12[0],xmm10[0]
-; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm15 = xmm10[0],xmm11[0]
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm8 = xmm9[0],xmm8[0]
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm13, %xmm9 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm9 = xmm13[0],mem[0]
-; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rsi
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 128(%rdx), %xmm13
+; AVX1-ONLY-NEXT:    vmovapd 80(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm3[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 $19, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm1 = mem[2,3],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm0 = xmm0[1],mem[1]
+; AVX1-ONLY-NEXT:    vmovapd 96(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovapd 96(%rsi), %ymm2
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 96(%r8), %ymm0
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm12[0],ymm0[2],ymm12[2]
+; AVX1-ONLY-NEXT:    vmovapd 112(%rdx), %xmm1
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm1[0],mem[0]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm0 = ymm2[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 $19, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = mem[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 112(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1],mem[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
+; AVX1-ONLY-NEXT:    vbroadcastsd 136(%rcx), %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovapd 128(%rdx), %ymm1
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm13, %ymm1, %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3],ymm0[4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, (%rsp) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm14[1]
+; AVX1-ONLY-NEXT:    vmovapd 128(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovapd 128(%rsi), %ymm3
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm2[0],ymm3[0],ymm2[2],ymm3[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 128(%r8), %ymm0
+; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0],mem[0],ymm0[2],mem[2]
+; AVX1-ONLY-NEXT:    vmovaps 144(%rdx), %xmm2
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm2 = xmm2[0],mem[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 144(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm3[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 $19, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm1 = mem[2,3],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm11[1],xmm7[1]
+; AVX1-ONLY-NEXT:    vmovapd 160(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovapd 160(%rsi), %ymm2
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 160(%r8), %ymm0
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm5[0],ymm0[2],ymm5[2]
+; AVX1-ONLY-NEXT:    vmovapd 176(%rdx), %xmm1
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm1[0],mem[0]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm0 = ymm2[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm9[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 176(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1],mem[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
+; AVX1-ONLY-NEXT:    vbroadcastsd 200(%rcx), %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovapd 192(%rdx), %ymm2
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm2, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm10[1],xmm6[1]
+; AVX1-ONLY-NEXT:    vmovapd 192(%rdi), %ymm3
+; AVX1-ONLY-NEXT:    vmovapd 192(%rsi), %ymm4
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm3[0],ymm4[0],ymm3[2],ymm4[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm11 = ymm0[0,1],ymm3[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 192(%r8), %ymm0
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[2],mem[2]
+; AVX1-ONLY-NEXT:    vmovapd 208(%rcx), %xmm9
+; AVX1-ONLY-NEXT:    vmovapd 208(%rdx), %xmm6
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm6 = xmm6[0],xmm9[0]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm6 = ymm6[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 208(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm2 = ymm4[0,0,3,2]
+; AVX1-ONLY-NEXT:    vmovapd 192(%rax), %ymm4
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm4[2,3],ymm2[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm7 = ymm2[0],ymm0[1],ymm2[2],ymm0[3]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm2 = mem[0,0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 224(%rax), %ymm8, %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm2[0,1],ymm0[2,3],ymm2[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovapd 224(%rdi), %ymm2
+; AVX1-ONLY-NEXT:    vmovapd 224(%rsi), %ymm8
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm2[0],ymm8[0],ymm2[2],ymm8[2]
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm10 = mem[0,0]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm10 = ymm10[0],ymm2[1,2,3]
+; AVX1-ONLY-NEXT:    vmovapd 240(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm2[0,1],mem[2,3]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm8 = ymm8[0,0,3,2]
+; AVX1-ONLY-NEXT:    vmovapd 224(%rax), %ymm15
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm8 = ymm15[2,3],ymm8[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm8 = ymm8[0],ymm2[1],ymm8[2],ymm2[3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm2 = xmm9[1],mem[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 216(%r9), %ymm9
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm2[0,1],ymm9[2],ymm2[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm5 = ymm2[0,1,2],ymm4[3]
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm4 = xmm0[1],mem[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 248(%r9), %ymm9
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm4 = ymm4[0,1],ymm9[2],ymm4[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm9 = ymm10[0],ymm15[1],ymm10[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm15 = ymm4[0,1,2],ymm15[3]
+; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm4 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps 128(%rdx), %xmm10
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm10 = xmm10[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm2 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps 64(%rdx), %xmm14
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm14 = xmm14[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps 192(%rdx), %xmm13
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm13 = xmm13[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps 64(%rdx), %xmm12
+; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps (%rdx), %xmm12
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm12 = xmm12[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps 192(%rdx), %xmm11
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm11 = xmm11[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps (%rdx), %xmm10
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm10 = xmm10[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps %xmm10, 16(%rsi)
-; AVX1-ONLY-NEXT:    vmovaps %xmm9, (%rsi)
-; AVX1-ONLY-NEXT:    vmovaps %xmm11, 1360(%rsi)
-; AVX1-ONLY-NEXT:    vmovaps %xmm8, 1344(%rsi)
-; AVX1-ONLY-NEXT:    vmovaps %xmm12, 464(%rsi)
-; AVX1-ONLY-NEXT:    vmovaps %xmm15, 448(%rsi)
-; AVX1-ONLY-NEXT:    vmovaps %xmm13, 912(%rsi)
-; AVX1-ONLY-NEXT:    vmovaps %xmm3, 896(%rsi)
-; AVX1-ONLY-NEXT:    vmovapd %ymm0, 1760(%rsi)
-; AVX1-ONLY-NEXT:    vmovapd %ymm1, 1728(%rsi)
-; AVX1-ONLY-NEXT:    vmovapd %ymm2, 1696(%rsi)
-; AVX1-ONLY-NEXT:    vmovapd %ymm14, 1664(%rsi)
-; AVX1-ONLY-NEXT:    vmovaps %ymm4, 1632(%rsi)
-; AVX1-ONLY-NEXT:    vmovaps %ymm5, 1600(%rsi)
-; AVX1-ONLY-NEXT:    vmovapd %ymm6, 1568(%rsi)
-; AVX1-ONLY-NEXT:    vmovaps %ymm7, 1536(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps %xmm12, 16(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, (%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm13, 1360(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, 1344(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm14, 464(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, 448(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm10, 912(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm4, 896(%rax)
+; AVX1-ONLY-NEXT:    vmovapd %ymm8, 1728(%rax)
+; AVX1-ONLY-NEXT:    vmovapd %ymm9, 1664(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %ymm3, 1600(%rax)
+; AVX1-ONLY-NEXT:    vmovapd %ymm7, 1504(%rax)
+; AVX1-ONLY-NEXT:    vmovapd %ymm6, 1472(%rax)
+; AVX1-ONLY-NEXT:    vmovapd %ymm11, 1440(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1504(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1408(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1472(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1280(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1440(%rsi)
-; AVX1-ONLY-NEXT:    vmovups (%rsp), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1408(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1248(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1376(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1216(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1312(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1056(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1280(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1024(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1248(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 992(%rax)
+; AVX1-ONLY-NEXT:    vmovups (%rsp), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 960(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1216(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 832(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1184(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 800(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1152(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 768(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1120(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 608(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1088(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 576(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1056(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 544(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1024(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 512(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 992(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 384(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 960(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 352(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 928(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 320(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 864(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 160(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 832(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 128(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 800(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 96(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 768(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 64(%rax)
+; AVX1-ONLY-NEXT:    vmovapd %ymm15, 1760(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 736(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1696(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 704(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1632(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 672(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1568(%rax)
+; AVX1-ONLY-NEXT:    vmovapd %ymm5, 1536(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 640(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1376(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 608(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1312(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 576(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1184(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 544(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1152(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 512(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1120(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 480(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1088(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 416(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 928(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 384(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 864(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 352(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 736(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 320(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 704(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 288(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 672(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 256(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 640(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 224(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 480(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 192(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 416(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 160(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 288(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 128(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 256(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 96(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 224(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 64(%rsi)
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 192(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 32(%rsi)
-; AVX1-ONLY-NEXT:    addq $1320, %rsp # imm = 0x528
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 32(%rax)
+; AVX1-ONLY-NEXT:    addq $1384, %rsp # imm = 0x568
 ; AVX1-ONLY-NEXT:    vzeroupper
 ; AVX1-ONLY-NEXT:    retq
 ;
 ; AVX2-ONLY-LABEL: store_i64_stride7_vf32:
 ; AVX2-ONLY:       # %bb.0:
-; AVX2-ONLY-NEXT:    subq $1464, %rsp # imm = 0x5B8
+; AVX2-ONLY-NEXT:    subq $1768, %rsp # imm = 0x6E8
 ; AVX2-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; AVX2-ONLY-NEXT:    vmovaps 32(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovaps 32(%rdi), %ymm12
 ; AVX2-ONLY-NEXT:    vmovaps (%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovaps (%rsi), %ymm2
-; AVX2-ONLY-NEXT:    vmovaps (%rdx), %ymm3
-; AVX2-ONLY-NEXT:    vmovaps (%r8), %ymm13
-; AVX2-ONLY-NEXT:    vmovaps (%r9), %ymm9
-; AVX2-ONLY-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm4 = mem[0,0]
-; AVX2-ONLY-NEXT:    vbroadcastsd 8(%rcx), %ymm5
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm3[0,1,2,3],ymm5[4,5],ymm3[6,7]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm4[0,1],ymm5[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovaps (%r8), %xmm6
-; AVX2-ONLY-NEXT:    vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 32(%r8), %xmm4
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm6
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],ymm6[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm6 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
-; AVX2-ONLY-NEXT:    vmovaps (%rax), %xmm10
-; AVX2-ONLY-NEXT:    vmovaps %xmm10, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 16(%rax), %xmm7
-; AVX2-ONLY-NEXT:    vmovaps 32(%rax), %xmm5
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm8 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm8 = xmm8[1],xmm10[1]
-; 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:    vunpcklpd {{.*#+}} ymm6 = ymm13[0],ymm9[0],ymm13[2],ymm9[2]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm8 = ymm3[0],mem[0],ymm3[2],mem[2]
-; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm6 = ymm8[2,3],ymm6[2,3]
-; AVX2-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 32(%rsi), %ymm4
+; AVX2-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 32(%rdx), %ymm14
+; AVX2-ONLY-NEXT:    vmovaps (%rdx), %ymm10
+; AVX2-ONLY-NEXT:    vmovaps 16(%rax), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps 32(%rax), %xmm3
+; AVX2-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm2[1],ymm1[3],ymm2[3]
 ; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,3,3]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm3[6,7]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm7[0,1],ymm1[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 32(%rdi), %xmm1
-; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm2 = xmm4[0],mem[0]
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm2, %ymm2
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 32(%rdx), %xmm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm10[6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 32(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-ONLY-NEXT:    vmovaps 32(%r9), %xmm2
 ; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1]
+; AVX2-ONLY-NEXT:    vmovaps 32(%r8), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm2[0]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm2, %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 32(%rdx), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 32(%rsi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm2[1],xmm0[1]
 ; AVX2-ONLY-NEXT:    vbroadcastsd 40(%rcx), %ymm2
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
-; AVX2-ONLY-NEXT:    vmovaps 32(%rsi), %ymm2
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm0, %ymm3
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm3[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm5[1]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm0[0],ymm2[0],ymm0[2],ymm2[2]
-; 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:    vmovaps 32(%rdx), %ymm1
-; AVX2-ONLY-NEXT:    vmovaps 32(%r8), %ymm3
-; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 32(%r9), %ymm4
-; AVX2-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm3[0],ymm4[0],ymm3[2],ymm4[2]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm1[0],mem[0],ymm1[2],mem[2]
-; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm4[2,3],ymm3[2,3]
-; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm2[1],ymm0[3],ymm2[3]
-; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,3,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 48(%rax), %xmm2
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX2-ONLY-NEXT:    vmovaps 48(%rax), %xmm1
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 64(%rdx), %ymm0
-; AVX2-ONLY-NEXT:    vbroadcastsd 72(%rcx), %ymm1
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm2 = mem[0,0]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovaps 64(%r8), %xmm2
-; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 64(%rax), %xmm2
-; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1]
-; AVX2-ONLY-NEXT:    vmovaps 64(%rdi), %ymm2
-; AVX2-ONLY-NEXT:    vmovaps 64(%rsi), %ymm3
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm2[0],ymm3[0],ymm2[2],ymm3[2]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm4[4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 64(%r8), %ymm1
-; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 64(%r9), %ymm4
-; AVX2-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm4[0],ymm1[2],ymm4[2]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm0[0],mem[0],ymm0[2],mem[2]
-; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm4[2,3],ymm1[2,3]
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm12[1],ymm4[1],ymm12[3],ymm4[3]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,3,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm14[6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 64(%rdi), %ymm1
 ; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm2[1],ymm3[1],ymm2[3],ymm3[3]
-; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,3,3]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 64(%rsi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,3,3]
+; AVX2-ONLY-NEXT:    vmovaps 64(%rdx), %ymm3
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[6,7]
 ; AVX2-ONLY-NEXT:    vmovaps 80(%rax), %xmm1
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 96(%r9), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-ONLY-NEXT:    vmovaps 96(%r8), %xmm0
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm0[0],xmm1[0]
 ; AVX2-ONLY-NEXT:    vmovaps 96(%rax), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
-; AVX2-ONLY-NEXT:    vmovaps 96(%rdi), %xmm3
-; AVX2-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 96(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 96(%rdx), %xmm3
-; AVX2-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm3[1]
-; AVX2-ONLY-NEXT:    vbroadcastsd 104(%rcx), %ymm3
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 96(%rdx), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 96(%rsi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm2[1],xmm1[1]
+; AVX2-ONLY-NEXT:    vbroadcastsd 104(%rcx), %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm2[1]
-; AVX2-ONLY-NEXT:    vmovaps 96(%rdi), %ymm1
-; AVX2-ONLY-NEXT:    vmovaps 96(%rsi), %ymm2
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm1[0],ymm2[0],ymm1[2],ymm2[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 96(%rdx), %ymm0
-; AVX2-ONLY-NEXT:    vmovaps 96(%r8), %ymm3
-; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 96(%r9), %ymm4
-; AVX2-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm3[0],ymm4[0],ymm3[2],ymm4[2]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm0[0],mem[0],ymm0[2],mem[2]
-; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm4[2,3],ymm3[2,3]
-; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm2[1],ymm1[3],ymm2[3]
-; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,3,3]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 96(%rdi), %ymm6
+; AVX2-ONLY-NEXT:    vmovaps 96(%rsi), %ymm4
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm6[1],ymm4[1],ymm6[3],ymm4[3]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,3,3]
+; AVX2-ONLY-NEXT:    vmovaps 96(%rdx), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vmovaps 112(%rax), %xmm1
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 128(%rdx), %ymm0
-; AVX2-ONLY-NEXT:    vbroadcastsd 136(%rcx), %ymm1
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm2 = mem[0,0]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovaps 128(%r8), %xmm14
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm14, %ymm0, %ymm2
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 128(%rax), %xmm2
-; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1]
-; AVX2-ONLY-NEXT:    vmovaps 128(%rdi), %ymm2
-; AVX2-ONLY-NEXT:    vmovaps 128(%rsi), %ymm3
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm2[0],ymm3[0],ymm2[2],ymm3[2]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm4[4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 128(%r8), %ymm12
-; AVX2-ONLY-NEXT:    vmovaps 128(%r9), %ymm11
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm12[0],ymm11[0],ymm12[2],ymm11[2]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm0[0],mem[0],ymm0[2],mem[2]
-; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm4[2,3],ymm1[2,3]
+; AVX2-ONLY-NEXT:    vmovaps 128(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 128(%rsi), %ymm1
 ; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm2[1],ymm3[1],ymm2[3],ymm3[3]
-; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,3,3]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,3,3]
+; AVX2-ONLY-NEXT:    vmovaps 128(%rdx), %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
 ; AVX2-ONLY-NEXT:    vmovaps 144(%rax), %xmm1
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 160(%r9), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-ONLY-NEXT:    vmovaps 160(%r8), %xmm0
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
-; AVX2-ONLY-NEXT:    vmovaps 160(%rax), %xmm2
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
-; AVX2-ONLY-NEXT:    vmovaps 160(%rdi), %xmm10
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm0, %ymm3
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm0[0],xmm1[0]
+; AVX2-ONLY-NEXT:    vmovaps 160(%rax), %xmm5
+; AVX2-ONLY-NEXT:    vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm1, %ymm1
+; AVX2-ONLY-NEXT:    vmovaps 160(%rdi), %xmm5
+; AVX2-ONLY-NEXT:    vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm5
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm5[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 160(%rdx), %xmm9
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm9[1]
-; AVX2-ONLY-NEXT:    vbroadcastsd 168(%rcx), %ymm3
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 160(%rdx), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 160(%rsi), %xmm5
+; AVX2-ONLY-NEXT:    vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm5[1],xmm1[1]
+; AVX2-ONLY-NEXT:    vbroadcastsd 168(%rcx), %ymm5
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm5[4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm2[1]
-; AVX2-ONLY-NEXT:    vmovaps 160(%rdi), %ymm1
-; AVX2-ONLY-NEXT:    vmovaps 160(%rsi), %ymm2
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm1[0],ymm2[0],ymm1[2],ymm2[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 160(%rdx), %ymm0
-; AVX2-ONLY-NEXT:    vmovaps 160(%r8), %ymm8
-; AVX2-ONLY-NEXT:    vmovaps 160(%r9), %ymm7
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm8[0],ymm7[0],ymm8[2],ymm7[2]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm0[0],mem[0],ymm0[2],mem[2]
-; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm4[2,3],ymm3[2,3]
-; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm2[1],ymm1[3],ymm2[3]
-; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,3,3]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 160(%rdi), %ymm11
+; AVX2-ONLY-NEXT:    vmovaps 160(%rsi), %ymm7
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm11[1],ymm7[1],ymm11[3],ymm7[3]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,3,3]
+; AVX2-ONLY-NEXT:    vmovaps 160(%rdx), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vmovaps 176(%rax), %xmm1
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 192(%rdx), %ymm0
-; AVX2-ONLY-NEXT:    vbroadcastsd 200(%rcx), %ymm1
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm2 = mem[0,0]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovaps 192(%r8), %xmm4
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm0, %ymm2
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 192(%rax), %xmm5
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm5[1]
-; AVX2-ONLY-NEXT:    vmovaps 192(%rdi), %ymm2
-; AVX2-ONLY-NEXT:    vmovaps 192(%rsi), %ymm3
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm6 = ymm2[0],ymm3[0],ymm2[2],ymm3[2]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm6[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovaps 192(%rdi), %ymm13
+; AVX2-ONLY-NEXT:    vmovaps 192(%rsi), %ymm15
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm13[1],ymm15[1],ymm13[3],ymm15[3]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm0[0,2,3,3]
+; AVX2-ONLY-NEXT:    vmovaps 192(%rdx), %ymm9
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm9[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 208(%rax), %xmm5
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm5[0,1],ymm1[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 192(%r8), %ymm1
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],mem[0],ymm1[2],mem[2]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm6 = ymm0[0],mem[0],ymm0[2],mem[2]
-; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm6[2,3],ymm1[2,3]
+; AVX2-ONLY-NEXT:    vmovaps 224(%rdx), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 224(%rsi), %xmm5
+; AVX2-ONLY-NEXT:    vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm5[1],xmm1[1]
+; AVX2-ONLY-NEXT:    vbroadcastsd 232(%rcx), %ymm5
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm5[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, 224(%r8), %ymm0, %ymm5
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm5[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm2[1],ymm3[1],ymm2[3],ymm3[3]
-; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,3,3]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT:    vmovaps 208(%rax), %xmm1
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovaps 224(%rdi), %ymm1
+; AVX2-ONLY-NEXT:    vmovaps 224(%rsi), %ymm5
+; AVX2-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm5 = ymm1[1],ymm5[1],ymm1[3],ymm5[3]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,2,3,3]
+; AVX2-ONLY-NEXT:    vmovaps 224(%rdx), %ymm8
+; AVX2-ONLY-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],ymm8[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 240(%rax), %xmm8
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm8[0,1],ymm5[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastsd 8(%rcx), %ymm5
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm10[0,1,2,3],ymm5[4,5],ymm10[6,7]
+; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm8 = mem[0,0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm8[0,1],ymm5[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovaps (%r8), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm8
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],ymm8[6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm5 = ymm5[0],mem[0],ymm5[2],mem[2]
+; AVX2-ONLY-NEXT:    vmovaps (%rax), %xmm8
+; AVX2-ONLY-NEXT:    vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps (%r9), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm8 = xmm0[1],xmm8[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm8[0,1,2,3],ymm5[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm5 = ymm10[0],mem[0],ymm10[2],mem[2]
+; AVX2-ONLY-NEXT:    vmovaps (%r8), %ymm0
 ; 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(%r8), %ymm6
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
-; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm6[0,1],ymm0[0,1]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3],ymm2[4,5,6,7]
-; AVX2-ONLY-NEXT:    vbroadcastsd 224(%rax), %ymm2
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 224(%rdx), %xmm2
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1]
-; AVX2-ONLY-NEXT:    vbroadcastsd 232(%rcx), %ymm3
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5],ymm1[6,7]
-; AVX2-ONLY-NEXT:    vinsertf128 $1, 224(%r8), %ymm0, %ymm3
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm3[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 224(%rsi), %ymm1
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm3 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm15 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm15[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovaps (%r9), %ymm8
+; AVX2-ONLY-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm8 = ymm0[0],ymm8[0],ymm0[2],ymm8[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm5 = ymm5[2,3],ymm8[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm5 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm5 = ymm12[0],mem[0],ymm12[2],mem[2]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm8 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm8 = xmm8[1],mem[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm8[0,1,2,3],ymm5[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm5 = ymm14[0],mem[0],ymm14[2],mem[2]
+; AVX2-ONLY-NEXT:    vmovaps 32(%r8), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 32(%r9), %ymm8
+; AVX2-ONLY-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm8 = ymm0[0],ymm8[0],ymm0[2],ymm8[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm5 = ymm5[2,3],ymm8[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastsd 72(%rcx), %ymm5
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm3[0,1,2,3],ymm5[4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm8 = mem[0,0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm8[0,1],ymm5[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovaps 64(%r8), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm8
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],ymm8[6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm5 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm5, %ymm5 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm5 = ymm5[0],mem[0],ymm5[2],mem[2]
+; AVX2-ONLY-NEXT:    vmovaps 64(%r9), %xmm14
+; AVX2-ONLY-NEXT:    vmovaps 64(%rax), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm8 = xmm14[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm8[0,1,2,3],ymm5[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm3[0],mem[0],ymm3[2],mem[2]
+; AVX2-ONLY-NEXT:    vmovaps 64(%r8), %ymm12
+; AVX2-ONLY-NEXT:    vmovaps 64(%r9), %ymm10
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm5 = ymm12[0],ymm10[0],ymm12[2],ymm10[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm3[2,3],ymm5[2,3]
 ; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
-; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,3,3]
-; AVX2-ONLY-NEXT:    vmovaps 224(%rdx), %ymm1
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX2-ONLY-NEXT:    vmovaps 240(%rax), %xmm3
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm6[0],ymm4[0],ymm6[2],ymm4[2]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm4 = xmm4[1],mem[1]
+; 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:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm3[0],mem[0],ymm3[2],mem[2]
+; AVX2-ONLY-NEXT:    vmovaps 96(%r8), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 96(%r9), %ymm4
+; AVX2-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm0[0],ymm4[0],ymm0[2],ymm4[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm3[2,3],ymm4[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastsd 136(%rcx), %ymm3
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm2[0,1,2,3],ymm3[4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm4 = mem[0,0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1],ymm3[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovaps 128(%r8), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm4
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm4[6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm4 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm4 = ymm3[0],mem[0],ymm3[2],mem[2]
+; AVX2-ONLY-NEXT:    vmovaps 128(%rax), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 128(%r9), %xmm3
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm5 = xmm3[1],xmm0[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3],ymm4[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm2[0],mem[0],ymm2[2],mem[2]
+; AVX2-ONLY-NEXT:    vmovaps 128(%r8), %ymm5
+; AVX2-ONLY-NEXT:    vmovaps 128(%r9), %ymm4
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm6 = ymm5[0],ymm4[0],ymm5[2],ymm4[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm2[2,3],ymm6[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm11[0],ymm7[0],ymm11[2],ymm7[2]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm6 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm6 = xmm6[1],mem[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm6[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm0[0],mem[0],ymm0[2],mem[2]
+; AVX2-ONLY-NEXT:    vmovaps 160(%r8), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 160(%r9), %ymm11
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm6 = ymm0[0],ymm11[0],ymm0[2],ymm11[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm2[2,3],ymm6[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastsd 200(%rcx), %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm9[0,1,2,3],ymm2[4,5],ymm9[6,7]
+; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm6 = mem[0,0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm6[0,1],ymm2[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovaps 192(%r8), %xmm6
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm7
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm7[6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm7 = ymm13[0],ymm15[0],ymm13[2],ymm15[2]
+; AVX2-ONLY-NEXT:    vmovaps 192(%rax), %xmm8
+; AVX2-ONLY-NEXT:    vmovaps 192(%r9), %xmm2
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm13 = xmm2[1],xmm8[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm13[0,1,2,3],ymm7[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 192(%rcx), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm9[0],ymm0[0],ymm9[2],ymm0[2]
+; AVX2-ONLY-NEXT:    vmovaps 192(%r8), %ymm13
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm7 = ymm13[0],mem[0],ymm13[2],mem[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm7[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 224(%r8), %ymm7
+; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm15 = ymm7[0,1],ymm1[0,1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1],ymm0[2,3],ymm15[4,5,6,7]
+; AVX2-ONLY-NEXT:    vbroadcastsd 224(%rax), %ymm15
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm15[4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm1[0],mem[0],ymm1[2],mem[2]
+; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; 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:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
-; AVX2-ONLY-NEXT:    vmovaps (%rdi), %xmm3
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm0
+; AVX2-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm0 = xmm0[0],mem[0]
+; AVX2-ONLY-NEXT:    vmovaps (%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-ONLY-NEXT:    vbroadcastsd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 16-byte Folded Reload
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm15[4,5],ymm0[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm0 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm0 = ymm13[1],mem[1],ymm13[3],mem[3]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
 ; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,3,3]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
 ; AVX2-ONLY-NEXT:    vbroadcastsd 24(%rcx), %ymm15
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1],ymm0[2,3,4,5,6,7]
 ; 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:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
+; AVX2-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm0 = xmm0[0],mem[0]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload
 ; AVX2-ONLY-NEXT:    vbroadcastsd 32(%rcx), %ymm15
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm15[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
 ; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
 ; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,3,3]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
 ; AVX2-ONLY-NEXT:    vbroadcastsd 56(%rcx), %ymm15
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm15[0,1],ymm0[2,3,4,5,6,7]
 ; 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:    vunpcklpd {{.*#+}} xmm15 = xmm0[0],mem[0]
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm14 = xmm0[0],xmm14[0]
 ; AVX2-ONLY-NEXT:    vmovaps 64(%rdi), %xmm0
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm15, %ymm15
-; AVX2-ONLY-NEXT:    vbroadcastsd {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 16-byte Folded Reload
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm15[0,1,2,3],ymm13[4,5],ymm15[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm13 = ymm13[1],mem[1],ymm13[3],mem[3]
-; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm13 = ymm13[0,2,3,3]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5],mem[6,7]
-; AVX2-ONLY-NEXT:    vbroadcastsd 88(%rcx), %ymm15
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm15[0,1],ymm13[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm13 # 16-byte Reload
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm13 = xmm13[0],mem[0]
-; AVX2-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 16-byte Folded Reload
-; AVX2-ONLY-NEXT:    vbroadcastsd 96(%rcx), %ymm15
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5],ymm15[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm13 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm13, %ymm13 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm13 = ymm13[1],mem[1],ymm13[3],mem[3]
-; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm13 = ymm13[0,2,3,3]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5],mem[6,7]
-; AVX2-ONLY-NEXT:    vbroadcastsd 120(%rcx), %ymm15
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm15[0,1],ymm13[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm13 = xmm14[0],mem[0]
-; AVX2-ONLY-NEXT:    vmovaps 128(%rdi), %xmm14
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm14, %ymm13, %ymm13
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm14, %ymm14
 ; AVX2-ONLY-NEXT:    vbroadcastsd {{[-0-9]+}}(%r{{[sb]}}p), %ymm15 # 16-byte Folded Reload
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3],ymm15[4,5],ymm13[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm13, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm11 = ymm12[1],ymm11[1],ymm12[3],ymm11[3]
-; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm11 = ymm11[0,2,3,3]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm11[0,1,2,3,4,5],mem[6,7]
-; AVX2-ONLY-NEXT:    vbroadcastsd 152(%rcx), %ymm12
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm11 = ymm12[0,1],ymm11[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm10 = xmm10[0],mem[0]
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm9, %ymm10, %ymm9
-; AVX2-ONLY-NEXT:    vbroadcastsd 160(%rcx), %ymm10
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm9 = ymm9[0,1,2,3,4,5],ymm10[6,7]
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm7 = ymm8[1],ymm7[1],ymm8[3],ymm7[3]
-; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm7 = ymm7[0,2,3,3]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],mem[6,7]
-; AVX2-ONLY-NEXT:    vbroadcastsd 184(%rcx), %ymm8
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm8[0,1],ymm7[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm4 = xmm4[0],mem[0]
-; AVX2-ONLY-NEXT:    vmovaps 192(%rdi), %xmm8
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm4, %ymm4
-; AVX2-ONLY-NEXT:    vbroadcastsd %xmm5, %ymm5
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm5[4,5],ymm4[6,7]
-; AVX2-ONLY-NEXT:    vbroadcastsd 216(%rcx), %ymm5
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm5 = xmm5[0,1],mem[2,3]
-; AVX2-ONLY-NEXT:    vbroadcastsd 216(%r9), %ymm10
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3],ymm10[4,5],ymm5[6,7]
-; AVX2-ONLY-NEXT:    vmovaps 224(%rdi), %xmm10
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm10 = xmm10[0],mem[0]
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm10, %ymm2
-; AVX2-ONLY-NEXT:    vbroadcastsd 224(%rcx), %ymm10
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm10[6,7]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],mem[0],ymm1[2],mem[2]
-; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3],ymm6[2,3]
-; AVX2-ONLY-NEXT:    vbroadcastsd 240(%r9), %ymm6
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm6[6,7]
-; AVX2-ONLY-NEXT:    vbroadcastsd 248(%rcx), %ymm6
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm6 = xmm6[0,1],mem[2,3]
-; AVX2-ONLY-NEXT:    vbroadcastsd 248(%r9), %ymm10
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm10[4,5],ymm6[6,7]
-; AVX2-ONLY-NEXT:    vmovaps 224(%rax), %ymm10
-; AVX2-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm12 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm12 = mem[0,1],ymm10[2,3],mem[4,5,6,7]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm10[6,7]
-; AVX2-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rdi
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm14 = ymm14[0,1,2,3],ymm15[4,5],ymm14[6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm14, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm10 = ymm12[1],ymm10[1],ymm12[3],ymm10[3]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm10 = ymm10[0,2,3,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3,4,5],mem[6,7]
+; AVX2-ONLY-NEXT:    vbroadcastsd 88(%rcx), %ymm12
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm12[0,1],ymm10[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm10 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm10, %xmm10 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm10 = xmm10[0],mem[0]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    vbroadcastsd 96(%rcx), %ymm12
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3,4,5],ymm12[6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm9 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm9, %ymm10 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm10 = ymm9[1],mem[1],ymm9[3],mem[3]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm10 = ymm10[0,2,3,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm10[0,1,2,3,4,5],mem[6,7]
+; AVX2-ONLY-NEXT:    vbroadcastsd 120(%rcx), %ymm12
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm10 = ymm12[0,1],ymm10[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm10, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm9 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm3 = xmm9[0],xmm3[0]
+; AVX2-ONLY-NEXT:    vmovaps 128(%rdi), %xmm14
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm14, %ymm3, %ymm3
+; AVX2-ONLY-NEXT:    vbroadcastsd {{[-0-9]+}}(%r{{[sb]}}p), %ymm12 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm12[4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm3 = ymm5[1],ymm4[1],ymm5[3],ymm4[3]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,2,3,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],mem[6,7]
+; AVX2-ONLY-NEXT:    vbroadcastsd 152(%rcx), %ymm4
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1],ymm3[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm3 = xmm3[0],mem[0]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    vbroadcastsd 160(%rcx), %ymm5
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm5[6,7]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm5 = ymm4[1],ymm11[1],ymm4[3],ymm11[3]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm5 = ymm5[0,2,3,3]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm5[0,1,2,3,4,5],mem[6,7]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm10 = xmm14[0],mem[0]
+; AVX2-ONLY-NEXT:    vbroadcastsd 184(%rcx), %ymm9
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm9[0,1],ymm5[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm6[0],xmm2[0]
+; AVX2-ONLY-NEXT:    vmovaps 192(%rdi), %xmm9
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm9, %ymm2, %ymm2
+; AVX2-ONLY-NEXT:    vbroadcastsd %xmm8, %ymm6
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm6[4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 224(%rdi), %xmm6
+; AVX2-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm6 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm6 = xmm6[0],mem[0]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6, %ymm6 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    vbroadcastsd 224(%rcx), %ymm8
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3,4,5],ymm8[6,7]
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm8 = ymm4[1],ymm13[1],ymm4[3],ymm13[3]
+; AVX2-ONLY-NEXT:    vbroadcastsd 216(%r9), %ymm11
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm8 = ymm8[2,3],ymm11[2,3]
+; AVX2-ONLY-NEXT:    vmovaps 224(%rcx), %ymm11
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm13 = ymm4[0],ymm11[0],ymm4[2],ymm11[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm13 = ymm13[2,3],ymm7[2,3]
+; AVX2-ONLY-NEXT:    vbroadcastsd 240(%r9), %ymm15
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm13 = ymm13[0,1,2,3,4,5],ymm15[6,7]
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm7 = ymm11[1],ymm7[1],ymm11[3],ymm7[3]
+; AVX2-ONLY-NEXT:    vbroadcastsd 248(%r9), %ymm11
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm7 = ymm7[2,3],ymm11[2,3]
+; AVX2-ONLY-NEXT:    vmovaps 224(%rax), %ymm11
+; AVX2-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm11, %ymm15 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm15 = mem[0,1],ymm11[2,3],mem[4,5,6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm7 = ymm7[0,1,2,3,4,5],ymm11[6,7]
+; AVX2-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rdi
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm8 = ymm8[0,1,2,3,4,5],mem[6,7]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm11 = xmm14[0],mem[0]
 ; AVX2-ONLY-NEXT:    vmovaps 128(%rdx), %xmm14
 ; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm14 = xmm14[0],mem[0]
 ; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
-; AVX2-ONLY-NEXT:    vmovaps 64(%rdx), %xmm15
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm15 = xmm15[0],mem[0]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm8 = xmm8[0],mem[0]
-; AVX2-ONLY-NEXT:    vmovaps 192(%rdx), %xmm13
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm13 = xmm13[0],mem[0]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm3 = xmm3[0],mem[0]
-; AVX2-ONLY-NEXT:    vmovaps (%rdx), %xmm11
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm11 = xmm11[0],mem[0]
-; AVX2-ONLY-NEXT:    vmovaps %xmm11, 16(%rdi)
-; AVX2-ONLY-NEXT:    vmovaps %xmm3, (%rdi)
-; AVX2-ONLY-NEXT:    vmovaps %xmm13, 1360(%rdi)
-; AVX2-ONLY-NEXT:    vmovaps %xmm8, 1344(%rdi)
-; AVX2-ONLY-NEXT:    vmovaps %xmm15, 464(%rdi)
+; AVX2-ONLY-NEXT:    vmovaps 64(%rdx), %xmm10
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm10 = xmm10[0],mem[0]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm9 = xmm9[0],mem[0]
+; AVX2-ONLY-NEXT:    vmovaps 192(%rdx), %xmm12
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm12 = xmm12[0],mem[0]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm1[0],mem[0]
+; AVX2-ONLY-NEXT:    vmovaps (%rdx), %xmm4
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm4 = xmm4[0],mem[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 16(%rdi)
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, (%rdi)
+; AVX2-ONLY-NEXT:    vmovaps %xmm12, 1360(%rdi)
+; AVX2-ONLY-NEXT:    vmovaps %xmm9, 1344(%rdi)
+; AVX2-ONLY-NEXT:    vmovaps %xmm10, 464(%rdi)
 ; AVX2-ONLY-NEXT:    vmovaps %xmm0, 448(%rdi)
 ; AVX2-ONLY-NEXT:    vmovaps %xmm14, 912(%rdi)
-; AVX2-ONLY-NEXT:    vmovaps %xmm10, 896(%rdi)
-; AVX2-ONLY-NEXT:    vmovaps %ymm6, 1760(%rdi)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1728(%rdi)
-; AVX2-ONLY-NEXT:    vmovaps %ymm1, 1696(%rdi)
-; AVX2-ONLY-NEXT:    vmovaps %ymm12, 1664(%rdi)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1632(%rdi)
+; AVX2-ONLY-NEXT:    vmovaps %xmm11, 896(%rdi)
+; AVX2-ONLY-NEXT:    vmovaps %ymm7, 1760(%rdi)
+; AVX2-ONLY-NEXT:    vmovaps %ymm13, 1696(%rdi)
+; AVX2-ONLY-NEXT:    vmovaps %ymm15, 1664(%rdi)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1600(%rdi)
-; AVX2-ONLY-NEXT:    vmovaps %ymm2, 1568(%rdi)
-; AVX2-ONLY-NEXT:    vmovaps %ymm5, 1536(%rdi)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1504(%rdi)
+; AVX2-ONLY-NEXT:    vmovaps %ymm8, 1536(%rdi)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1472(%rdi)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1440(%rdi)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1408(%rdi)
-; AVX2-ONLY-NEXT:    vmovaps %ymm4, 1376(%rdi)
-; AVX2-ONLY-NEXT:    vmovaps %ymm7, 1312(%rdi)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1280(%rdi)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1248(%rdi)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1216(%rdi)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1024(%rdi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 992(%rdi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 960(%rdi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 800(%rdi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 768(%rdi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 576(%rdi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 544(%rdi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 512(%rdi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 352(%rdi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 320(%rdi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 128(%rdi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 96(%rdi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 64(%rdi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1728(%rdi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1632(%rdi)
+; AVX2-ONLY-NEXT:    vmovaps %ymm6, 1568(%rdi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1504(%rdi)
+; AVX2-ONLY-NEXT:    vmovaps %ymm2, 1376(%rdi)
+; AVX2-ONLY-NEXT:    vmovaps %ymm5, 1312(%rdi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1280(%rdi)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1184(%rdi)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1152(%rdi)
-; AVX2-ONLY-NEXT:    vmovaps %ymm9, 1120(%rdi)
+; AVX2-ONLY-NEXT:    vmovaps %ymm3, 1120(%rdi)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1088(%rdi)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1056(%rdi)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1024(%rdi)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 992(%rdi)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 960(%rdi)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 928(%rdi)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 864(%rdi)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 832(%rdi)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 800(%rdi)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 768(%rdi)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 736(%rdi)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 704(%rdi)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 672(%rdi)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 640(%rdi)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 608(%rdi)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 576(%rdi)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 544(%rdi)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 512(%rdi)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 480(%rdi)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 416(%rdi)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 384(%rdi)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 352(%rdi)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 320(%rdi)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 288(%rdi)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 256(%rdi)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 224(%rdi)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 192(%rdi)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 160(%rdi)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 128(%rdi)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 96(%rdi)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 64(%rdi)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 32(%rdi)
-; AVX2-ONLY-NEXT:    addq $1464, %rsp # imm = 0x5B8
+; AVX2-ONLY-NEXT:    addq $1768, %rsp # imm = 0x6E8
 ; AVX2-ONLY-NEXT:    vzeroupper
 ; AVX2-ONLY-NEXT:    retq
 ;
 ; AVX512F-ONLY-SLOW-LABEL: store_i64_stride7_vf32:
 ; AVX512F-ONLY-SLOW:       # %bb.0:
 ; AVX512F-ONLY-SLOW-NEXT:    subq $2120, %rsp # imm = 0x848
 ; AVX512F-ONLY-SLOW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
 ; AVX512F-ONLY-SLOW-NEXT:    vmovdqa64 (%rdi), %zmm8

 ; SSE-NEXT:    movaps %xmm0, 16(%rax)
 ; SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; SSE-NEXT:    movaps %xmm0, (%rax)
 ; SSE-NEXT:    addq $3224, %rsp # imm = 0xC98
 ; SSE-NEXT:    retq
 ;
 ; AVX1-ONLY-LABEL: store_i64_stride7_vf64:
 ; AVX1-ONLY:       # %bb.0:
-; AVX1-ONLY-NEXT:    subq $3960, %rsp # imm = 0xF78
+; AVX1-ONLY-NEXT:    subq $3320, %rsp # imm = 0xCF8
 ; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; AVX1-ONLY-NEXT:    vmovaps (%rsi), %ymm7
-; AVX1-ONLY-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps (%rdx), %ymm0
-; AVX1-ONLY-NEXT:    vmovaps (%r9), %ymm8
-; AVX1-ONLY-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps (%r8), %xmm2
-; AVX1-ONLY-NEXT:    vmovaps (%rax), %xmm1
-; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm2, %ymm3
+; AVX1-ONLY-NEXT:    vmovapd (%r9), %ymm4
+; AVX1-ONLY-NEXT:    vmovupd %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd (%rax), %ymm5
+; AVX1-ONLY-NEXT:    vmovupd %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps (%r8), %xmm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps (%rax), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-ONLY-NEXT:    vmovaps (%r9), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm1[0,1,0,1]
-; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm6
-; AVX1-ONLY-NEXT:    vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 16(%rdi), %xmm5
-; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm1
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm4, %ymm4
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm4[2,3],ymm3[4,5],ymm4[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vbroadcastsd 8(%rcx), %ymm3
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm0[0,1,2,3],ymm3[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vmovaps (%rsi), %xmm4
-; AVX1-ONLY-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm4[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1],ymm3[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1,2,3,4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 16(%rax), %xmm2
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm3 = ymm7[1],ymm5[1],ymm7[3],ymm5[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 16(%rcx), %xmm0
-; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
-; AVX1-ONLY-NEXT:    vmovaps 16(%r8), %xmm2
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm2 = ymm8[1],ymm2[1],ymm8[3],ymm2[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,1,0,1]
+; AVX1-ONLY-NEXT:    vmovaps (%rdi), %xmm3
+; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 16(%r8), %xmm0
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm4[0],ymm0[1],ymm4[3],ymm0[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm0[0,1,2],ymm5[3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = mem[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[2],ymm1[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rsi), %xmm0
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 32(%rcx), %ymm2, %ymm3
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm2[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 32(%rcx), %ymm1, %ymm3
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rdx), %xmm4
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm2, %ymm2
-; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm2 = ymm2[0],ymm3[1],ymm2[2],ymm3[2]
-; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm1[0],ymm3[1],ymm1[2],ymm3[2]
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 32(%r8), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps 32(%rax), %xmm5
 ; AVX1-ONLY-NEXT:    vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
 ; AVX1-ONLY-NEXT:    vmovaps 32(%r9), %xmm3
 ; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 32(%r8), %xmm2
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm3 = xmm2[0],xmm3[0]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm3, %ymm3
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm4[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 40(%rcx), %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 32(%rsi), %ymm1
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 48(%rdi), %xmm0
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 48(%rax), %xmm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 32(%r9), %ymm1
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 48(%r8), %xmm0
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 48(%rcx), %xmm1
-; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[0,1,0,1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm3, %ymm2
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm1, %ymm3
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3],ymm3[4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 32(%r9), %ymm3
+; AVX1-ONLY-NEXT:    vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm4, %ymm1
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX1-ONLY-NEXT:    vbroadcastsd 40(%rcx), %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovapd 32(%rax), %ymm2
+; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 48(%r8), %xmm0
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm3[0],ymm0[1],ymm3[3],ymm0[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm0[0,1,2],ymm2[3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = mem[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[2],ymm1[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 64(%r9), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vmovaps 64(%r8), %xmm1
-; AVX1-ONLY-NEXT:    vmovaps 64(%rax), %xmm2
-; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 64(%rdx), %ymm0
-; AVX1-ONLY-NEXT:    vbroadcastsd 72(%rcx), %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3],ymm2[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 64(%rsi), %xmm3
-; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 64(%rsi), %ymm2
+; AVX1-ONLY-NEXT:    vmovaps 64(%r8), %xmm2
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 80(%rdi), %xmm1
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm2[1],ymm1[1],ymm2[3],ymm1[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 80(%rax), %xmm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 64(%r9), %ymm1
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 80(%r8), %xmm0
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 80(%rcx), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps 64(%rax), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm2, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 64(%r9), %ymm1
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 80(%r8), %xmm0
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[3],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovapd 64(%rax), %ymm1
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm0[0,1,2],ymm1[3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = mem[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[2],ymm1[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rsi), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, 96(%rcx), %ymm2, %ymm3
 ; AVX1-ONLY-NEXT:    vmovaps 96(%rdx), %xmm4
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm2, %ymm2
 ; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm2 = ymm2[0],ymm3[1],ymm2[2],ymm3[2]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vmovaps 96(%r9), %xmm3
+; AVX1-ONLY-NEXT:    vmovaps 96(%r9), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,0,1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm2, %ymm1
+; AVX1-ONLY-NEXT:    vmovaps 96(%rax), %xmm3
 ; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 96(%r8), %xmm2
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm3 = xmm2[0],xmm3[0]
-; AVX1-ONLY-NEXT:    vmovaps 96(%rax), %xmm5
-; AVX1-ONLY-NEXT:    vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm3, %ymm3
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm4[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 104(%rcx), %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 96(%rsi), %ymm1
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 112(%rdi), %xmm0
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 112(%rax), %xmm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 96(%r9), %ymm1
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm2, %ymm3
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3],ymm3[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 112(%r8), %xmm0
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 112(%rcx), %xmm1
-; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm4, %ymm1
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX1-ONLY-NEXT:    vbroadcastsd 104(%rcx), %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 96(%r9), %ymm1
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 112(%r8), %xmm0
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[3],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovapd 96(%rax), %ymm1
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm0[0,1,2],ymm1[3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = mem[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[2],ymm1[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 128(%r9), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vmovaps 128(%rdi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vmovaps 128(%r8), %xmm1
-; AVX1-ONLY-NEXT:    vmovaps 128(%rax), %xmm2
-; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 128(%rdx), %ymm0
-; AVX1-ONLY-NEXT:    vbroadcastsd 136(%rcx), %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3],ymm2[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 128(%rsi), %xmm3
-; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 128(%rsi), %ymm2
+; AVX1-ONLY-NEXT:    vmovaps 128(%r8), %xmm2
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 144(%rdi), %xmm1
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm2[1],ymm1[1],ymm2[3],ymm1[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 144(%rax), %xmm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 128(%r9), %ymm1
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 144(%r8), %xmm0
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 144(%rcx), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps 128(%rax), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm2, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 128(%r9), %ymm1
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 144(%r8), %xmm0
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[3],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovapd 128(%rax), %ymm1
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm0[0,1,2],ymm1[3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = mem[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[2],ymm1[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 160(%rsi), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps 160(%rdi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, 160(%rcx), %ymm2, %ymm3
 ; AVX1-ONLY-NEXT:    vmovaps 160(%rdx), %xmm4
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm2, %ymm2
 ; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm2 = ymm2[0],ymm3[1],ymm2[2],ymm3[2]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vmovaps 160(%r9), %xmm3
-; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 160(%r9), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,0,1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm2, %ymm1
 ; AVX1-ONLY-NEXT:    vmovaps 160(%r8), %xmm2
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm3 = xmm2[0],xmm3[0]
-; AVX1-ONLY-NEXT:    vmovaps 160(%rax), %xmm5
-; AVX1-ONLY-NEXT:    vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm3, %ymm3
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm4[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 168(%rcx), %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 160(%rsi), %ymm1
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 176(%rdi), %xmm0
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 176(%rax), %xmm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 160(%r9), %ymm1
+; AVX1-ONLY-NEXT:    vmovaps 160(%rax), %xmm3
+; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm2, %ymm3
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3],ymm3[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 176(%r8), %xmm0
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 176(%rcx), %xmm1
-; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm4, %ymm1
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX1-ONLY-NEXT:    vbroadcastsd 168(%rcx), %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 160(%r9), %ymm1
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 176(%r8), %xmm0
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[3],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovapd 160(%rax), %ymm1
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm0[0,1,2],ymm1[3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = mem[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[2],ymm1[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 192(%r9), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vmovaps 192(%rdi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vmovaps 192(%r8), %xmm1
-; AVX1-ONLY-NEXT:    vmovaps 192(%rax), %xmm2
-; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 192(%rdx), %ymm0
-; AVX1-ONLY-NEXT:    vbroadcastsd 200(%rcx), %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3],ymm2[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 192(%rsi), %xmm3
-; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 192(%rsi), %ymm2
+; AVX1-ONLY-NEXT:    vmovaps 192(%r8), %xmm2
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 208(%rdi), %xmm1
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm2[1],ymm1[1],ymm2[3],ymm1[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 208(%rax), %xmm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 192(%r9), %ymm1
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 208(%r8), %xmm0
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 208(%rcx), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps 192(%rax), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm2, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 192(%r9), %ymm1
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 208(%r8), %xmm0
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[3],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovapd 192(%rax), %ymm1
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm0[0,1,2],ymm1[3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = mem[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[2],ymm1[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 224(%rsi), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps 224(%rdi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, 224(%rcx), %ymm2, %ymm3
 ; AVX1-ONLY-NEXT:    vmovaps 224(%rdx), %xmm4
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm2, %ymm2
 ; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm2 = ymm2[0],ymm3[1],ymm2[2],ymm3[2]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vmovaps 224(%r9), %xmm3
+; AVX1-ONLY-NEXT:    vmovaps 224(%r9), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,0,1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm2, %ymm1
+; AVX1-ONLY-NEXT:    vmovaps 224(%rax), %xmm3
 ; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 224(%r8), %xmm2
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm3 = xmm2[0],xmm3[0]
-; AVX1-ONLY-NEXT:    vmovaps 224(%rax), %xmm5
-; AVX1-ONLY-NEXT:    vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm3, %ymm3
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm4[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 232(%rcx), %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 224(%rsi), %ymm1
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 240(%rdi), %xmm0
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 240(%rax), %xmm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 224(%r9), %ymm1
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm2, %ymm3
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3],ymm3[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 240(%r8), %xmm0
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 240(%rcx), %xmm1
-; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm4, %ymm1
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX1-ONLY-NEXT:    vbroadcastsd 232(%rcx), %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 224(%r9), %ymm1
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 240(%r8), %xmm0
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[3],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovapd 224(%rax), %ymm1
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm0[0,1,2],ymm1[3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = mem[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[2],ymm1[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 256(%r9), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vmovaps 256(%rdi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vmovaps 256(%r8), %xmm1
-; AVX1-ONLY-NEXT:    vmovaps 256(%rax), %xmm2
-; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 256(%rdx), %ymm0
-; AVX1-ONLY-NEXT:    vbroadcastsd 264(%rcx), %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3],ymm2[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 256(%rsi), %xmm3
-; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 256(%rsi), %ymm2
+; AVX1-ONLY-NEXT:    vmovaps 256(%r8), %xmm2
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 272(%rdi), %xmm1
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm2[1],ymm1[1],ymm2[3],ymm1[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 272(%rax), %xmm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 256(%r9), %ymm1
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 272(%r8), %xmm0
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 272(%rcx), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps 256(%rax), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm2, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 256(%r9), %ymm1
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 272(%r8), %xmm0
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[3],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovapd 256(%rax), %ymm1
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm0[0,1,2],ymm1[3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = mem[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[2],ymm1[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 288(%rsi), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps 288(%rdi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, 288(%rcx), %ymm2, %ymm3
 ; AVX1-ONLY-NEXT:    vmovaps 288(%rdx), %xmm4
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm2, %ymm2
 ; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm2 = ymm2[0],ymm3[1],ymm2[2],ymm3[2]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vmovaps 288(%r9), %xmm3
-; AVX1-ONLY-NEXT:    vmovaps %xmm3, (%rsp) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 288(%r9), %xmm2
+; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm2[0,1,0,1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm2, %ymm1
+; AVX1-ONLY-NEXT:    vmovaps 288(%rax), %xmm3
+; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 288(%r8), %xmm2
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm3 = xmm2[0],xmm3[0]
-; AVX1-ONLY-NEXT:    vmovaps 288(%rax), %xmm5
-; AVX1-ONLY-NEXT:    vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm3, %ymm3
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm2, %ymm3
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3],ymm3[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm4[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 296(%rcx), %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 288(%rsi), %ymm1
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 304(%rdi), %xmm0
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 304(%rax), %xmm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 288(%r9), %ymm1
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 304(%r8), %xmm0
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 304(%rcx), %xmm1
-; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm4, %ymm1
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX1-ONLY-NEXT:    vbroadcastsd 296(%rcx), %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 288(%r9), %ymm1
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 304(%r8), %xmm0
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[3],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovapd 288(%rax), %ymm1
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm0[0,1,2],ymm1[3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = mem[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[2],ymm1[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 320(%r9), %xmm0
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,1,0,1]
 ; AVX1-ONLY-NEXT:    vmovaps 320(%rdi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vmovaps 320(%r8), %xmm1
-; AVX1-ONLY-NEXT:    vmovaps 320(%rax), %xmm2
-; AVX1-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 320(%rdx), %ymm0
-; AVX1-ONLY-NEXT:    vbroadcastsd 328(%rcx), %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm0[0,1,2,3],ymm2[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 320(%rsi), %xmm3
-; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm3[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 320(%rsi), %ymm2
+; AVX1-ONLY-NEXT:    vmovaps 320(%r8), %xmm2
 ; AVX1-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 336(%rdi), %xmm1
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm2[1],ymm1[1],ymm2[3],ymm1[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 336(%rax), %xmm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 320(%r9), %ymm1
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 336(%r8), %xmm0
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 336(%rcx), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps 320(%rax), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm2, %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 320(%r9), %ymm1
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 336(%r8), %xmm0
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[3],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovapd 320(%rax), %ymm1
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm0[0,1,2],ymm1[3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = mem[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[1],ymm2[1],ymm0[2],ymm2[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 352(%rsi), %xmm0
-; AVX1-ONLY-NEXT:    vmovaps 352(%rdi), %xmm1
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 352(%rcx), %ymm2, %ymm3
-; AVX1-ONLY-NEXT:    vmovaps 352(%rdx), %xmm4
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm2, %ymm2
-; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm2 = ymm2[0],ymm3[1],ymm2[2],ymm3[2]
-; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vmovaps 352(%r9), %xmm3
-; AVX1-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 352(%r8), %xmm2
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm3 = xmm2[0],xmm3[0]
-; AVX1-ONLY-NEXT:    vmovaps 352(%rax), %xmm5
-; AVX1-ONLY-NEXT:    vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 352(%rdi), %xmm2
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm3 = xmm2[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 352(%rcx), %ymm3, %ymm4
+; AVX1-ONLY-NEXT:    vmovaps 352(%rdx), %xmm5
 ; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm3, %ymm3
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm4[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 360(%rcx), %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 352(%rsi), %ymm1
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 368(%rdi), %xmm0
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 368(%rax), %xmm1
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 352(%r9), %ymm1
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm3[0],ymm4[1],ymm3[2],ymm4[2]
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 352(%r9), %xmm12
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm3 = xmm12[0,1,0,1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm3, %ymm2
+; AVX1-ONLY-NEXT:    vmovapd 352(%rax), %xmm9
+; AVX1-ONLY-NEXT:    vmovaps 352(%r8), %xmm3
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm9, %ymm3, %ymm4
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm4[0,1],ymm2[2,3],ymm4[4,5],ymm2[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 368(%r8), %xmm0
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 368(%rcx), %xmm15
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm15[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 384(%r9), %xmm14
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm14[0,1,0,1]
-; AVX1-ONLY-NEXT:    vmovaps 384(%rdi), %xmm0
-; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm1
-; AVX1-ONLY-NEXT:    vmovaps 384(%r8), %xmm2
-; AVX1-ONLY-NEXT:    vmovaps 384(%rax), %xmm13
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm13, %ymm2, %ymm3
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1],ymm1[2,3],ymm3[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 384(%rdx), %ymm1
-; AVX1-ONLY-NEXT:    vbroadcastsd 392(%rcx), %ymm3
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm1[0,1,2,3],ymm3[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 384(%rsi), %xmm0
-; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm4 = xmm0[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1],ymm3[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm2[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 384(%rsi), %ymm10
-; AVX1-ONLY-NEXT:    vmovaps 400(%rdi), %xmm2
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm2 = ymm10[1],ymm2[1],ymm10[3],ymm2[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 400(%rax), %xmm2
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm1[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 384(%r9), %ymm8
-; AVX1-ONLY-NEXT:    vmovaps 400(%r8), %xmm1
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm8[1],ymm1[1],ymm8[3],ymm1[3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],mem[6,7]
-; AVX1-ONLY-NEXT:    vmovaps 400(%rcx), %xmm7
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm2 = xmm7[2,3,2,3]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm1[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm5, %ymm2
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX1-ONLY-NEXT:    vbroadcastsd 360(%rcx), %ymm3
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3],ymm0[4,5],ymm2[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 352(%r9), %ymm4
+; AVX1-ONLY-NEXT:    vmovapd 368(%r8), %xmm0
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm4[0],ymm0[1],ymm4[3],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovapd 352(%rax), %ymm11
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm0[0,1,2],ymm11[3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = mem[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[1],ymm2[1],ymm0[2],ymm2[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 384(%r9), %xmm13
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm13[0,1,0,1]
+; AVX1-ONLY-NEXT:    vmovaps 384(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-ONLY-NEXT:    vmovapd 384(%rax), %xmm8
+; AVX1-ONLY-NEXT:    vmovaps 384(%r8), %xmm5
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm5, %ymm2
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 384(%r9), %ymm10
+; AVX1-ONLY-NEXT:    vmovapd 400(%r8), %xmm0
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm10[0],ymm0[1],ymm10[3],ymm0[2]
+; AVX1-ONLY-NEXT:    vmovapd 384(%rax), %ymm1
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, (%rsp) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm0[0,1,2],ymm1[3]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = mem[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[1],ymm2[1],ymm0[2],ymm2[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 416(%rsi), %xmm1
 ; AVX1-ONLY-NEXT:    vmovaps 416(%rdi), %xmm2
 ; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm3 = xmm2[0],xmm1[0]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 416(%rcx), %ymm3, %ymm4
-; AVX1-ONLY-NEXT:    vmovaps 416(%rdx), %xmm6
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm3, %ymm3
-; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm3[0],ymm4[1],ymm3[2],ymm4[2]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 416(%rcx), %ymm3, %ymm7
+; AVX1-ONLY-NEXT:    vmovaps 416(%rdx), %xmm15
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm3, %ymm3
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm0 = ymm3[0],ymm7[1],ymm3[2],ymm7[2]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm4
-; AVX1-ONLY-NEXT:    vmovapd 416(%r9), %xmm5
-; AVX1-ONLY-NEXT:    vmovapd 416(%r8), %xmm9
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm11 = xmm9[0],xmm5[0]
+; AVX1-ONLY-NEXT:    vmovapd 416(%r9), %xmm3
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm7 = xmm3[0,1,0,1]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm7, %ymm7
 ; AVX1-ONLY-NEXT:    vmovapd 416(%rax), %xmm2
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm11, %ymm11
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm11[0,1,2,3,4,5],ymm4[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm6[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 424(%rcx), %ymm4
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm4[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm9, %ymm0, %ymm4
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm4[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 448(%rsi), %xmm0
-; AVX1-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
-; AVX1-ONLY-NEXT:    vbroadcastsd 456(%rcx), %ymm4
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm4[2,3,4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovapd 448(%rdx), %ymm6
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 448(%r8), %ymm6, %ymm4
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm4[2,3],ymm1[4,5],ymm4[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 464(%rcx), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 416(%r8), %xmm0
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm14
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm6 = ymm14[0,1],ymm7[2,3],ymm14[4,5],ymm7[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm15, %ymm0
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; AVX1-ONLY-NEXT:    vbroadcastsd 424(%rcx), %ymm7
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm7[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 464(%rdx), %xmm1
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
-; AVX1-ONLY-NEXT:    vmovaps 448(%r8), %ymm3
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3],ymm3[4,5],ymm4[6,7]
-; AVX1-ONLY-NEXT:    vbroadcastsd 464(%r9), %ymm9
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3,4,5],ymm9[6,7]
+; AVX1-ONLY-NEXT:    vmovaps 464(%rcx), %xmm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 464(%rdx), %xmm0
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vmovaps 448(%r8), %ymm15
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm15[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vbroadcastsd 464(%r9), %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 480(%rsi), %xmm9
-; AVX1-ONLY-NEXT:    vmovaps 480(%rdi), %xmm12
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm11 = xmm12[0],xmm9[0]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 480(%rcx), %ymm11, %ymm1
-; AVX1-ONLY-NEXT:    vmovaps 480(%rdx), %xmm0
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm11, %ymm11
-; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm11[0],ymm1[1],ymm11[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vmovaps 480(%rsi), %xmm0
+; AVX1-ONLY-NEXT:    vmovaps 480(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm0[0]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 480(%rcx), %ymm1, %ymm7
+; AVX1-ONLY-NEXT:    vmovaps 480(%rdx), %xmm14
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm14, %ymm1, %ymm1
+; AVX1-ONLY-NEXT:    vshufpd {{.*#+}} ymm1 = ymm1[0],ymm7[1],ymm1[2],ymm7[2]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm1 = xmm9[2,3,2,3]
-; AVX1-ONLY-NEXT:    vbroadcastsd 488(%rcx), %ymm9
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm9[4,5],ymm1[6,7]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 480(%r8), %ymm0, %ymm0
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX1-ONLY-NEXT:    vbroadcastsd 488(%rcx), %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 480(%r8), %ymm14, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 496(%rcx), %xmm9
-; AVX1-ONLY-NEXT:    vmovapd 496(%rdx), %xmm0
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm9[0],ymm0[2],ymm9[2]
-; AVX1-ONLY-NEXT:    vmovapd 480(%r8), %ymm11
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm11[2],ymm0[3]
-; AVX1-ONLY-NEXT:    vbroadcastsd 496(%r9), %ymm1
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3]
+; AVX1-ONLY-NEXT:    vmovaps 496(%rcx), %xmm1
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps 496(%rdx), %xmm0
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
+; AVX1-ONLY-NEXT:    vmovaps 480(%r8), %ymm0
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vbroadcastsd 496(%r9), %ymm7
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm7[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
+; AVX1-ONLY-NEXT:    vbroadcastsd 8(%rcx), %ymm7
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm7[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovapd (%rdx), %ymm7
+; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm14 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm14[2,3],ymm1[4,5],ymm14[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm1 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm1 = xmm0[1],mem[1]
+; AVX1-ONLY-NEXT:    vmovapd (%rdi), %ymm14
+; AVX1-ONLY-NEXT:    vmovapd (%rsi), %ymm0
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm14 = ymm14[0],ymm0[0],ymm14[2],ymm0[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm14[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovaps (%r8), %ymm1
+; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm1 = ymm1[0],mem[0],ymm1[2],mem[2]
+; AVX1-ONLY-NEXT:    vmovaps 16(%rdx), %xmm14
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm14 = xmm14[0],mem[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm14[0,1,2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 16(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm7[2,3]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm0 = ymm0[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 $19, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = mem[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm0 = xmm0[1],mem[1]
+; AVX1-ONLY-NEXT:    vmovapd 32(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovapd 32(%rsi), %ymm7
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm7[0],ymm1[2],ymm7[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps (%rdi), %ymm0
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0],mem[0],ymm0[2],mem[2]
-; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm1 = xmm1[1],mem[1]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps (%r8), %ymm0
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0],mem[0],ymm0[2],mem[2]
-; AVX1-ONLY-NEXT:    vmovaps 16(%rdx), %xmm1
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm1 = xmm1[0],mem[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 32(%rdi), %ymm0
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0],mem[0],ymm0[2],mem[2]
-; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm1 = xmm1[1],mem[1]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 32(%r8), %ymm0
 ; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0],mem[0],ymm0[2],mem[2]
 ; AVX1-ONLY-NEXT:    vmovaps 48(%rdx), %xmm1
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm1 = xmm1[0],mem[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 64(%rdi), %ymm0
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0],mem[0],ymm0[2],mem[2]
-; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm1 = xmm1[1],mem[1]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm1[0],mem[0]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm0 = ymm7[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 $19, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = mem[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 48(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1],mem[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
+; AVX1-ONLY-NEXT:    vbroadcastsd 72(%rcx), %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovapd 64(%rdx), %ymm1
+; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm7 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm7[2,3],ymm0[4,5],ymm7[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm0 = xmm0[1],mem[1]
+; AVX1-ONLY-NEXT:    vmovapd 64(%rdi), %ymm7
+; AVX1-ONLY-NEXT:    vmovapd 64(%rsi), %ymm14
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm7 = ymm7[0],ymm14[0],ymm7[2],ymm14[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm7[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 64(%r8), %ymm0
 ; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0],mem[0],ymm0[2],mem[2]
-; AVX1-ONLY-NEXT:    vmovaps 80(%rdx), %xmm1
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm1 = xmm1[0],mem[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 96(%rdi), %ymm0
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0],mem[0],ymm0[2],mem[2]
-; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm1 = xmm1[1],mem[1]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 80(%rdx), %xmm7
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm7 = xmm7[0],mem[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm7[0,1,2,3],ymm0[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 80(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm14[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 $19, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm1 = mem[2,3],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm0 = xmm0[1],mem[1]
+; AVX1-ONLY-NEXT:    vmovapd 96(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovapd 96(%rsi), %ymm7
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm7[0],ymm1[2],ymm7[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 96(%r8), %ymm0
 ; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0],mem[0],ymm0[2],mem[2]
 ; AVX1-ONLY-NEXT:    vmovaps 112(%rdx), %xmm1
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm1 = xmm1[0],mem[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 128(%rdi), %ymm0
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0],mem[0],ymm0[2],mem[2]
-; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm1 = xmm1[1],mem[1]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm1[0],mem[0]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm0 = ymm7[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 $19, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = mem[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 112(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1],mem[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
+; AVX1-ONLY-NEXT:    vbroadcastsd 136(%rcx), %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovapd 128(%rdx), %ymm1
+; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm7 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm7[2,3],ymm0[4,5],ymm7[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm0 = xmm0[1],mem[1]
+; AVX1-ONLY-NEXT:    vmovapd 128(%rdi), %ymm7
+; AVX1-ONLY-NEXT:    vmovapd 128(%rsi), %ymm14
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm7 = ymm7[0],ymm14[0],ymm7[2],ymm14[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm7[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 128(%r8), %ymm0
 ; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0],mem[0],ymm0[2],mem[2]
-; AVX1-ONLY-NEXT:    vmovaps 144(%rdx), %xmm1
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm1 = xmm1[0],mem[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 160(%rdi), %ymm0
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0],mem[0],ymm0[2],mem[2]
-; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm1 = xmm1[1],mem[1]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 144(%rdx), %xmm7
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm7 = xmm7[0],mem[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm7[0,1,2,3],ymm0[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 144(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm14[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 $19, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm1 = mem[2,3],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm0 = xmm0[1],mem[1]
+; AVX1-ONLY-NEXT:    vmovapd 160(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovapd 160(%rsi), %ymm7
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm7[0],ymm1[2],ymm7[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 160(%r8), %ymm0
 ; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0],mem[0],ymm0[2],mem[2]
 ; AVX1-ONLY-NEXT:    vmovaps 176(%rdx), %xmm1
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm1 = xmm1[0],mem[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 192(%rdi), %ymm0
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0],mem[0],ymm0[2],mem[2]
-; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm1 = xmm1[1],mem[1]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm1[0],mem[0]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm0 = ymm7[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 $19, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = mem[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 176(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1],mem[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
+; AVX1-ONLY-NEXT:    vbroadcastsd 200(%rcx), %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovapd 192(%rdx), %ymm1
+; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm7 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm7[2,3],ymm0[4,5],ymm7[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm0 = xmm0[1],mem[1]
+; AVX1-ONLY-NEXT:    vmovapd 192(%rdi), %ymm7
+; AVX1-ONLY-NEXT:    vmovapd 192(%rsi), %ymm14
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm7 = ymm7[0],ymm14[0],ymm7[2],ymm14[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm7[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 192(%r8), %ymm0
 ; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0],mem[0],ymm0[2],mem[2]
-; AVX1-ONLY-NEXT:    vmovaps 208(%rdx), %xmm1
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm1 = xmm1[0],mem[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 224(%rdi), %ymm0
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0],mem[0],ymm0[2],mem[2]
-; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm1 = xmm1[1],mem[1]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 208(%rdx), %xmm7
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm7 = xmm7[0],mem[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm7[0,1,2,3],ymm0[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 208(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm14[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 $19, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm1 = mem[2,3],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm0 = xmm0[1],mem[1]
+; AVX1-ONLY-NEXT:    vmovapd 224(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovapd 224(%rsi), %ymm7
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm7[0],ymm1[2],ymm7[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 224(%r8), %ymm0
 ; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0],mem[0],ymm0[2],mem[2]
 ; AVX1-ONLY-NEXT:    vmovaps 240(%rdx), %xmm1
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm1 = xmm1[0],mem[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 256(%rdi), %ymm0
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0],mem[0],ymm0[2],mem[2]
-; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm1 = xmm1[1],mem[1]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm1[0],mem[0]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm0 = ymm7[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 $19, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = mem[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 240(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1],mem[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
+; AVX1-ONLY-NEXT:    vbroadcastsd 264(%rcx), %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovapd 256(%rdx), %ymm1
+; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm7 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm7[2,3],ymm0[4,5],ymm7[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm0 = xmm0[1],mem[1]
+; AVX1-ONLY-NEXT:    vmovapd 256(%rdi), %ymm7
+; AVX1-ONLY-NEXT:    vmovapd 256(%rsi), %ymm14
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm7 = ymm7[0],ymm14[0],ymm7[2],ymm14[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm7[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 256(%r8), %ymm0
 ; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0],mem[0],ymm0[2],mem[2]
-; AVX1-ONLY-NEXT:    vmovaps 272(%rdx), %xmm1
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm1 = xmm1[0],mem[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 288(%rdi), %ymm0
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0],mem[0],ymm0[2],mem[2]
-; AVX1-ONLY-NEXT:    vmovaps (%rsp), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm1 = xmm1[1],mem[1]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 272(%rdx), %xmm7
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm7 = xmm7[0],mem[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm7[0,1,2,3],ymm0[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 272(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm14[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 $19, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm1 = mem[2,3],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm0 = xmm0[1],mem[1]
+; AVX1-ONLY-NEXT:    vmovapd 288(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovapd 288(%rsi), %ymm7
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm7[0],ymm1[2],ymm7[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 288(%r8), %ymm0
 ; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0],mem[0],ymm0[2],mem[2]
 ; AVX1-ONLY-NEXT:    vmovaps 304(%rdx), %xmm1
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm1 = xmm1[0],mem[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 320(%rdi), %ymm0
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0],mem[0],ymm0[2],mem[2]
-; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm1 = xmm1[1],mem[1]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm1[0],mem[0]
 ; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm0 = ymm7[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 $19, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm0 = mem[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 304(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1],mem[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
+; AVX1-ONLY-NEXT:    vbroadcastsd 328(%rcx), %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovapd 320(%rdx), %ymm1
+; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm7 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm7[2,3],ymm0[4,5],ymm7[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX1-ONLY-NEXT:    # xmm0 = xmm0[1],mem[1]
+; AVX1-ONLY-NEXT:    vmovapd 320(%rdi), %ymm7
+; AVX1-ONLY-NEXT:    vmovapd 320(%rsi), %ymm14
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm7 = ymm7[0],ymm14[0],ymm7[2],ymm14[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm7[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovaps 320(%r8), %ymm0
 ; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
 ; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0],mem[0],ymm0[2],mem[2]
-; AVX1-ONLY-NEXT:    vmovaps 336(%rdx), %xmm1
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm1 = xmm1[0],mem[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 352(%rdi), %ymm0
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0],mem[0],ymm0[2],mem[2]
-; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm1 = xmm1[1],mem[1]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 352(%r8), %ymm0
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
-; AVX1-ONLY-NEXT:    # ymm0 = ymm0[0],mem[0],ymm0[2],mem[2]
-; AVX1-ONLY-NEXT:    vmovaps 368(%rdx), %xmm1
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm15[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovaps 336(%rdx), %xmm7
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm7 = xmm7[0],mem[0]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm7[0,1,2,3],ymm0[4,5,6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 384(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovapd 336(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm14[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 $19, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm1 = mem[2,3],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm12[1],xmm9[1]
+; AVX1-ONLY-NEXT:    vmovapd 352(%rdi), %ymm1
+; AVX1-ONLY-NEXT:    vmovapd 352(%rsi), %ymm7
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm7[0],ymm1[2],ymm7[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 352(%r8), %ymm0
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm4[0],ymm0[2],ymm4[2]
+; AVX1-ONLY-NEXT:    vmovapd 368(%rdx), %xmm1
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm1[0],mem[0]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm0 = ymm7[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm11[2,3],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 368(%rdi), %xmm1
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1],mem[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
+; AVX1-ONLY-NEXT:    vbroadcastsd 392(%rcx), %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovapd 384(%rdx), %ymm1
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm1, %ymm4
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm4[2,3],ymm0[4,5],ymm4[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm13[1],xmm8[1]
+; AVX1-ONLY-NEXT:    vmovapd 384(%rdi), %ymm4
+; AVX1-ONLY-NEXT:    vmovapd 384(%rsi), %ymm5
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm4[0],ymm5[0],ymm4[2],ymm5[2]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm4[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 384(%r8), %ymm0
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm10[0],ymm0[2],ymm10[2]
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm14[1],xmm13[1]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovaps 384(%r8), %ymm0
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm8[0],ymm0[2],ymm8[2]
-; AVX1-ONLY-NEXT:    vmovaps 400(%rdx), %xmm1
-; AVX1-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm7[0]
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm5[1],xmm2[1]
+; AVX1-ONLY-NEXT:    vmovapd 400(%rdx), %xmm4
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm4 = xmm4[0],mem[0]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm4[0,1],ymm0[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 400(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm5[0,0,3,2]
+; AVX1-ONLY-NEXT:    vperm2f128 $19, (%rsp), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX1-ONLY-NEXT:    # ymm1 = mem[2,3],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm3[1],xmm2[1]
 ; AVX1-ONLY-NEXT:    vmovapd 416(%rdi), %ymm1
-; AVX1-ONLY-NEXT:    vmovapd 416(%rsi), %ymm5
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm5[0],ymm1[2],ymm5[2]
+; AVX1-ONLY-NEXT:    vmovapd 416(%rsi), %ymm2
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovapd 416(%r8), %ymm0
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[2],mem[2]
-; AVX1-ONLY-NEXT:    vmovapd 432(%rcx), %xmm4
+; AVX1-ONLY-NEXT:    vmovapd 432(%rcx), %xmm8
 ; AVX1-ONLY-NEXT:    vmovapd 432(%rdx), %xmm1
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm1[0],xmm4[0]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm1[0],xmm8[0]
 ; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, (%rsp) # 32-byte Spill
 ; AVX1-ONLY-NEXT:    vmovapd 432(%rdi), %xmm0
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm0[0,1],mem[2,3]
-; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm5 = ymm5[0,0,3,2]
-; AVX1-ONLY-NEXT:    vmovapd 416(%rax), %ymm2
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm5 = ymm2[2,3],ymm5[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm5[0],ymm1[1],ymm5[2],ymm1[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],mem[2,3]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm1 = ymm2[0,0,3,2]
+; AVX1-ONLY-NEXT:    vmovapd 416(%rax), %ymm9
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm9[2,3],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 448(%rax), %ymm3, %ymm1
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
+; AVX1-ONLY-NEXT:    vmovaps 448(%rdi), %xmm6
+; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5,6,7]
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 448(%rax), %ymm15, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3],ymm1[4,5],ymm0[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
+; AVX1-ONLY-NEXT:    vbroadcastsd 456(%rcx), %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovapd 448(%rdx), %ymm3
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 448(%r8), %ymm3, %ymm1
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
+; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovapd 448(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovapd 448(%rsi), %ymm4
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm4[0],ymm0[2],ymm4[2]
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0],ymm0[1,2,3]
+; AVX1-ONLY-NEXT:    vmovapd 464(%rdi), %xmm0
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm3[2,3]
+; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm3 = ymm4[0,0,3,2]
+; AVX1-ONLY-NEXT:    vmovapd 448(%rax), %ymm4
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm4[2,3],ymm3[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm3[0],ymm0[1],ymm3[2],ymm0[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload
 ; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm3 = mem[0,0]
-; AVX1-ONLY-NEXT:    vmovaps 448(%rdi), %xmm14
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm14, %ymm3, %ymm3
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm3[2,3],ymm1[4,5],ymm3[6,7]
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm3[2,3],ymm0[4,5,6,7]
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vinsertf128 $1, 480(%rax), %ymm2, %ymm3
+; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1],ymm0[2,3],ymm3[4,5],ymm0[6,7]
 ; AVX1-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 448(%rdi), %ymm1
-; AVX1-ONLY-NEXT:    vmovapd 448(%rsi), %ymm3
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[2],ymm3[2]
-; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm10 = mem[0,0]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm15 = ymm10[0],ymm1[1,2,3]
-; AVX1-ONLY-NEXT:    vmovapd 464(%rdi), %xmm10
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm6 = ymm10[0,1],ymm6[2,3]
+; AVX1-ONLY-NEXT:    vmovapd 480(%rdi), %ymm0
+; AVX1-ONLY-NEXT:    vmovapd 480(%rsi), %ymm3
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[2],ymm3[2]
+; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm5 = mem[0,0]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm5[0],ymm0[1,2,3]
+; AVX1-ONLY-NEXT:    vmovapd 496(%rdi), %xmm5
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm5 = ymm5[0,1],mem[2,3]
 ; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm3 = ymm3[0,0,3,2]
-; AVX1-ONLY-NEXT:    vmovapd 448(%rax), %ymm0
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm0[2,3],ymm3[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm3[0],ymm6[1],ymm3[2],ymm6[3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm3 = mem[0,0]
-; AVX1-ONLY-NEXT:    vinsertf128 $1, %xmm12, %ymm3, %ymm3
-; AVX1-ONLY-NEXT:    vinsertf128 $1, 480(%rax), %ymm11, %ymm6
-; AVX1-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm6[0,1],ymm3[2,3],ymm6[4,5],ymm3[6,7]
-; AVX1-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovapd 480(%rdi), %ymm3
-; AVX1-ONLY-NEXT:    vmovapd 480(%rsi), %ymm6
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm3[0],ymm6[0],ymm3[2],ymm6[2]
-; AVX1-ONLY-NEXT:    vmovddup {{.*#+}} xmm12 = mem[0,0]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm12 = ymm12[0],ymm3[1,2,3]
-; AVX1-ONLY-NEXT:    vmovapd 496(%rdi), %xmm3
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm3[0,1],mem[2,3]
-; AVX1-ONLY-NEXT:    vpermilpd {{.*#+}} ymm6 = ymm6[0,0,3,2]
-; AVX1-ONLY-NEXT:    vmovapd 480(%rax), %ymm1
-; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm6 = ymm1[2,3],ymm6[2,3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm6[0],ymm3[1],ymm6[2],ymm3[3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm4 = xmm4[1],mem[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 440(%r9), %ymm6
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm4 = ymm4[0,1],ymm6[2],ymm4[3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm4[0,1,2],ymm2[3]
+; AVX1-ONLY-NEXT:    vmovapd 480(%rax), %ymm15
+; AVX1-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm15[2,3],ymm3[2,3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm3[0],ymm5[1],ymm3[2],ymm5[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm4 = xmm2[1],mem[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 472(%r9), %ymm6
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm4 = ymm4[0,1],ymm6[2],ymm4[3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm15[0],ymm0[1],ymm15[2,3]
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm3 = xmm8[1],mem[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 440(%r9), %ymm5
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm3[0,1],ymm5[2],ymm3[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm3[0,1,2],ymm9[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm4[0,1,2],ymm0[3]
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm3 = xmm2[1],mem[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 472(%r9), %ymm5
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm3 = ymm3[0,1],ymm5[2],ymm3[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0],ymm4[1],ymm1[2,3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm3[0,1,2],ymm4[3]
+; AVX1-ONLY-NEXT:    vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    vmovupd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],mem[1]
+; AVX1-ONLY-NEXT:    vbroadcastsd 504(%r9), %ymm3
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm1 = ymm1[0,1],ymm3[2],ymm1[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm15[1],ymm0[2,3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm9[1],mem[1]
-; AVX1-ONLY-NEXT:    vbroadcastsd 504(%r9), %ymm6
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm6[2],ymm0[3]
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm2 = ymm12[0],ymm1[1],ymm12[2,3]
-; AVX1-ONLY-NEXT:    vmovupd %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0,1,2],ymm1[3]
+; AVX1-ONLY-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1,2],ymm15[3]
 ; AVX1-ONLY-NEXT:    vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm5 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm5 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm10 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps 256(%rdx), %xmm0
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm7 = xmm0[0],mem[0]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm4 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm4 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm5 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps 128(%rdx), %xmm9
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm9 = xmm9[0],mem[0]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm12 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm12 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm4 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps 64(%rdx), %xmm8
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm8 = xmm8[0],mem[0]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm3 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm3 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm3 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps 192(%rdx), %xmm15
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm15 = xmm15[0],mem[0]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm2 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm2 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm14, %xmm14 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm14 = xmm14[0],mem[0]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm2 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps 320(%rdx), %xmm14
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm14 = xmm14[0],mem[0]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm6 = xmm6[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps 448(%rdx), %xmm13
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm13 = xmm13[0],mem[0]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm1 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm1 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps 384(%rdx), %xmm12
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm12 = xmm12[0],mem[0]
 ; AVX1-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX1-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
-; AVX1-ONLY-NEXT:    # xmm0 = xmm0[0],mem[0]
-; AVX1-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; AVX1-ONLY-NEXT:    vmovaps 256(%rdx), %xmm13
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm13 = xmm13[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps 128(%rdx), %xmm8
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm8 = xmm8[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps 64(%rdx), %xmm6
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm6 = xmm6[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps 192(%rdx), %xmm10
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm10 = xmm10[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps 320(%rdx), %xmm11
+; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
+; AVX1-ONLY-NEXT:    vmovaps (%rdx), %xmm11
 ; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm11 = xmm11[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps 448(%rdx), %xmm7
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm7 = xmm7[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps 384(%rdx), %xmm15
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm15 = xmm15[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps (%rdx), %xmm9
-; AVX1-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm9 = xmm9[0],mem[0]
-; AVX1-ONLY-NEXT:    vmovaps %xmm9, 16(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm11, 16(%rax)
 ; AVX1-ONLY-NEXT:    vmovaps %xmm0, (%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm15, 2704(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm12, 2704(%rax)
 ; AVX1-ONLY-NEXT:    vmovaps %xmm1, 2688(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm7, 3152(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm14, 3136(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm11, 2256(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm13, 3152(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm6, 3136(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm14, 2256(%rax)
 ; AVX1-ONLY-NEXT:    vmovaps %xmm2, 2240(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm10, 1360(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm15, 1360(%rax)
 ; AVX1-ONLY-NEXT:    vmovaps %xmm3, 1344(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm6, 464(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm12, 448(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm8, 912(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm4, 896(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm13, 1808(%rax)
-; AVX1-ONLY-NEXT:    vmovaps %xmm5, 1792(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm8, 464(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm4, 448(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm9, 912(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm5, 896(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm7, 1808(%rax)
+; AVX1-ONLY-NEXT:    vmovaps %xmm10, 1792(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 3520(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 3456(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 3392(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 3296(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 3232(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 3200(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 3168(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 3072(%rax)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovups (%rsp), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 3040(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 3008(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2848(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2816(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2784(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2752(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2624(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2592(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2560(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2400(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2368(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2336(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2304(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2176(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2144(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2112(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1952(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1920(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1888(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1856(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1728(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1696(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1664(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1504(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1472(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1440(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1408(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1280(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1248(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1216(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1056(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1024(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 992(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 960(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 832(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 800(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 768(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 608(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 576(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 544(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 512(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 384(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 352(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 320(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 160(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 128(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 96(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX1-ONLY-NEXT:    vmovaps %ymm0, 64(%rax)
+; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 3552(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 3488(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 3424(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 3360(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 3328(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 3264(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 3200(%rax)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 3104(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2976(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2944(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2912(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2880(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2848(%rax)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2752(%rax)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2720(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2656(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2624(%rax)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2528(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2496(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2464(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2432(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2400(%rax)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2304(%rax)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2272(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2208(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2176(%rax)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2080(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2048(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 2016(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1984(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1952(%rax)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1856(%rax)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1824(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1760(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1728(%rax)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1632(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1600(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1568(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1536(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1504(%rax)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1408(%rax)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1376(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1312(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1280(%rax)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1184(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1152(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1120(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1088(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 1056(%rax)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 960(%rax)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 928(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 864(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 832(%rax)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 736(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 704(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 672(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 640(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 608(%rax)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 512(%rax)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 480(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 416(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 384(%rax)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 288(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 256(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 224(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 192(%rax)
 ; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 160(%rax)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX1-ONLY-NEXT:    vmovaps %ymm0, 64(%rax)
-; AVX1-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX1-ONLY-NEXT:    vmovaps %ymm0, 32(%rax)
-; AVX1-ONLY-NEXT:    addq $3960, %rsp # imm = 0xF78
+; AVX1-ONLY-NEXT:    addq $3320, %rsp # imm = 0xCF8
 ; AVX1-ONLY-NEXT:    vzeroupper
 ; AVX1-ONLY-NEXT:    retq
 ;
 ; AVX2-ONLY-LABEL: store_i64_stride7_vf64:
 ; AVX2-ONLY:       # %bb.0:
-; AVX2-ONLY-NEXT:    subq $3624, %rsp # imm = 0xE28
+; AVX2-ONLY-NEXT:    subq $4264, %rsp # imm = 0x10A8
 ; AVX2-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %rax
-; AVX2-ONLY-NEXT:    vmovaps 32(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovaps 32(%rdi), %ymm3
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovaps (%rdi), %ymm1
-; AVX2-ONLY-NEXT:    vmovaps (%rsi), %ymm2
-; AVX2-ONLY-NEXT:    vmovaps (%rdx), %ymm3
-; AVX2-ONLY-NEXT:    vmovaps (%r8), %ymm6
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps (%rsi), %ymm15
+; AVX2-ONLY-NEXT:    vmovaps 32(%rsi), %ymm6
 ; AVX2-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps (%r9), %ymm7
-; AVX2-ONLY-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm4 = mem[0,0]
-; AVX2-ONLY-NEXT:    vbroadcastsd 8(%rcx), %ymm5
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm5 = ymm3[0,1,2,3],ymm5[4,5],ymm3[6,7]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm5[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovaps (%r8), %xmm9
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm9, %ymm0, %ymm5
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm5[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
-; AVX2-ONLY-NEXT:    vmovaps (%rax), %xmm11
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm5 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm5 = xmm5[1],xmm11[1]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3],ymm4[4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm6[0],ymm7[0],ymm6[2],ymm7[2]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm5 = ymm3[0],mem[0],ymm3[2],mem[2]
-; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm4 = ymm5[2,3],ymm4[2,3]
-; AVX2-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 16(%rax), %xmm4
-; AVX2-ONLY-NEXT:    vmovaps 32(%rax), %xmm5
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm2[1],ymm1[3],ymm2[3]
+; AVX2-ONLY-NEXT:    vmovaps 32(%rdx), %ymm5
+; AVX2-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps (%rdx), %ymm9
+; AVX2-ONLY-NEXT:    vmovaps 16(%rax), %xmm0
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm15[1],ymm1[3],ymm15[3]
 ; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,3,3]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm3[6,7]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm4[0,1],ymm1[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 32(%rdi), %xmm15
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm0, %ymm1
-; AVX2-ONLY-NEXT:    vmovaps 32(%r8), %xmm2
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm3 = xmm2[0],mem[0]
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm3, %ymm3
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3,4,5],ymm1[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 32(%rdx), %xmm3
-; AVX2-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm3[1]
-; AVX2-ONLY-NEXT:    vbroadcastsd 40(%rcx), %ymm3
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5],ymm1[6,7]
-; AVX2-ONLY-NEXT:    vmovaps 32(%rsi), %ymm3
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm5[1]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm0[0],ymm3[0],ymm0[2],ymm3[2]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 32(%rdx), %ymm1
-; AVX2-ONLY-NEXT:    vmovaps 32(%r8), %ymm2
-; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 32(%r9), %ymm4
-; AVX2-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm2[0],ymm4[0],ymm2[2],ymm4[2]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm1[0],mem[0],ymm1[2],mem[2]
-; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm4[2,3],ymm2[2,3]
-; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm3[1],ymm0[3],ymm3[3]
-; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,3,3]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
-; AVX2-ONLY-NEXT:    vmovaps 48(%rax), %xmm1
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm9[6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 64(%rdx), %ymm0
-; AVX2-ONLY-NEXT:    vbroadcastsd 72(%rcx), %ymm1
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm2 = mem[0,0]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovaps 64(%r8), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps 32(%rdi), %xmm0
+; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-ONLY-NEXT:    vmovaps 32(%r9), %xmm2
 ; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 64(%rax), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps 32(%r8), %xmm1
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm2 = xmm1[0],xmm2[0]
+; AVX2-ONLY-NEXT:    vmovaps 32(%rax), %xmm4
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm2, %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-ONLY-NEXT:    vmovaps 32(%rdx), %xmm2
 ; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
+; AVX2-ONLY-NEXT:    vmovaps 32(%rsi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1]
-; AVX2-ONLY-NEXT:    vmovaps 64(%rdi), %ymm2
-; AVX2-ONLY-NEXT:    vmovaps 64(%rsi), %ymm3
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm2[0],ymm3[0],ymm2[2],ymm3[2]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm4[4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 64(%r8), %ymm1
-; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 64(%r9), %ymm4
-; AVX2-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm4[0],ymm1[2],ymm4[2]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm0[0],mem[0],ymm0[2],mem[2]
-; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm4[2,3],ymm1[2,3]
-; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm2[1],ymm3[1],ymm2[3],ymm3[3]
-; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,3,3]
+; AVX2-ONLY-NEXT:    vbroadcastsd 40(%rcx), %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 48(%rax), %xmm0
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm3[1],ymm6[1],ymm3[3],ymm6[3]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,3,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm5[6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 64(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 64(%rsi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,3,3]
+; AVX2-ONLY-NEXT:    vmovaps 64(%rdx), %ymm7
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm7[6,7]
 ; AVX2-ONLY-NEXT:    vmovaps 80(%rax), %xmm1
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 96(%r9), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-ONLY-NEXT:    vmovaps 96(%r8), %xmm0
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm0[0],xmm1[0]
 ; AVX2-ONLY-NEXT:    vmovaps 96(%rax), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
-; AVX2-ONLY-NEXT:    vmovaps 96(%rdi), %xmm3
-; AVX2-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 96(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 96(%rdx), %xmm3
-; AVX2-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm3[1]
-; AVX2-ONLY-NEXT:    vbroadcastsd 104(%rcx), %ymm3
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 96(%rdx), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 96(%rsi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm2[1],xmm1[1]
+; AVX2-ONLY-NEXT:    vbroadcastsd 104(%rcx), %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm2[1]
-; AVX2-ONLY-NEXT:    vmovaps 96(%rdi), %ymm1
-; AVX2-ONLY-NEXT:    vmovaps 96(%rsi), %ymm2
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovaps 96(%rdi), %ymm0
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 96(%rdx), %ymm0
-; AVX2-ONLY-NEXT:    vmovaps 96(%r8), %ymm3
-; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 96(%r9), %ymm4
-; AVX2-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm3[0],ymm4[0],ymm3[2],ymm4[2]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm0[0],mem[0],ymm0[2],mem[2]
-; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm4[2,3],ymm3[2,3]
-; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm2[1],ymm1[3],ymm2[3]
-; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,3,3]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 96(%rsi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,3,3]
+; AVX2-ONLY-NEXT:    vmovaps 96(%rdx), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vmovaps 112(%rax), %xmm1
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 128(%rdx), %ymm0
-; AVX2-ONLY-NEXT:    vbroadcastsd 136(%rcx), %ymm1
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm2 = mem[0,0]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovaps 128(%r8), %xmm2
-; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 128(%rax), %xmm2
-; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1]
-; AVX2-ONLY-NEXT:    vmovaps 128(%rdi), %ymm2
-; AVX2-ONLY-NEXT:    vmovaps 128(%rsi), %ymm3
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm2[0],ymm3[0],ymm2[2],ymm3[2]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm4[4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 128(%r8), %ymm1
-; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 128(%r9), %ymm4
-; AVX2-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm4[0],ymm1[2],ymm4[2]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm0[0],mem[0],ymm0[2],mem[2]
-; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm4[2,3],ymm1[2,3]
+; AVX2-ONLY-NEXT:    vmovaps 128(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 128(%rsi), %ymm1
 ; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm2[1],ymm3[1],ymm2[3],ymm3[3]
-; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,3,3]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,3,3]
+; AVX2-ONLY-NEXT:    vmovaps 128(%rdx), %ymm6
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm6[6,7]
 ; AVX2-ONLY-NEXT:    vmovaps 144(%rax), %xmm1
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 160(%r9), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-ONLY-NEXT:    vmovaps 160(%r8), %xmm0
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm0[0],xmm1[0]
 ; AVX2-ONLY-NEXT:    vmovaps 160(%rax), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
-; AVX2-ONLY-NEXT:    vmovaps 160(%rdi), %xmm3
-; AVX2-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 160(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 160(%rdx), %xmm3
-; AVX2-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm3[1]
-; AVX2-ONLY-NEXT:    vbroadcastsd 168(%rcx), %ymm3
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 160(%rdx), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 160(%rsi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm2[1],xmm1[1]
+; AVX2-ONLY-NEXT:    vbroadcastsd 168(%rcx), %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm2[1]
-; AVX2-ONLY-NEXT:    vmovaps 160(%rdi), %ymm1
-; AVX2-ONLY-NEXT:    vmovaps 160(%rsi), %ymm2
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovaps 160(%rdi), %ymm0
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 160(%rdx), %ymm0
-; AVX2-ONLY-NEXT:    vmovaps 160(%r8), %ymm3
-; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 160(%r9), %ymm4
-; AVX2-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm3[0],ymm4[0],ymm3[2],ymm4[2]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm0[0],mem[0],ymm0[2],mem[2]
-; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm4[2,3],ymm3[2,3]
-; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm2[1],ymm1[3],ymm2[3]
-; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,3,3]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 160(%rsi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,3,3]
+; AVX2-ONLY-NEXT:    vmovaps 160(%rdx), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vmovaps 176(%rax), %xmm1
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 192(%rdx), %ymm0
-; AVX2-ONLY-NEXT:    vbroadcastsd 200(%rcx), %ymm1
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm2 = mem[0,0]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovaps 192(%r8), %xmm2
-; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 192(%rax), %xmm2
-; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1]
-; AVX2-ONLY-NEXT:    vmovaps 192(%rdi), %ymm2
-; AVX2-ONLY-NEXT:    vmovaps 192(%rsi), %ymm3
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm2[0],ymm3[0],ymm2[2],ymm3[2]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm4[4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 192(%r8), %ymm1
-; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 192(%r9), %ymm4
-; AVX2-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm4[0],ymm1[2],ymm4[2]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm0[0],mem[0],ymm0[2],mem[2]
-; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm4[2,3],ymm1[2,3]
+; AVX2-ONLY-NEXT:    vmovaps 192(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 192(%rsi), %ymm1
 ; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm2[1],ymm3[1],ymm2[3],ymm3[3]
-; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,3,3]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,3,3]
+; AVX2-ONLY-NEXT:    vmovaps 192(%rdx), %ymm5
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm5[6,7]
 ; AVX2-ONLY-NEXT:    vmovaps 208(%rax), %xmm1
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 224(%r9), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-ONLY-NEXT:    vmovaps 224(%r8), %xmm0
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm0[0],xmm1[0]
 ; AVX2-ONLY-NEXT:    vmovaps 224(%rax), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
-; AVX2-ONLY-NEXT:    vmovaps 224(%rdi), %xmm3
-; AVX2-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 224(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 224(%rdx), %xmm3
-; AVX2-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm3[1]
-; AVX2-ONLY-NEXT:    vbroadcastsd 232(%rcx), %ymm3
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 224(%rdx), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 224(%rsi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm2[1],xmm1[1]
+; AVX2-ONLY-NEXT:    vbroadcastsd 232(%rcx), %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm2[1]
-; AVX2-ONLY-NEXT:    vmovaps 224(%rdi), %ymm1
-; AVX2-ONLY-NEXT:    vmovaps 224(%rsi), %ymm2
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovaps 224(%rdi), %ymm0
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 224(%rdx), %ymm0
-; AVX2-ONLY-NEXT:    vmovaps 224(%r8), %ymm4
-; AVX2-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 224(%r9), %ymm3
-; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm4[0],ymm3[0],ymm4[2],ymm3[2]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm0[0],mem[0],ymm0[2],mem[2]
-; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm4[2,3],ymm3[2,3]
-; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm2[1],ymm1[3],ymm2[3]
-; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,3,3]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 224(%rsi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,3,3]
+; AVX2-ONLY-NEXT:    vmovaps 224(%rdx), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vmovaps 240(%rax), %xmm1
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 256(%rdx), %ymm0
-; AVX2-ONLY-NEXT:    vbroadcastsd 264(%rcx), %ymm1
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm2 = mem[0,0]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovaps 256(%r8), %xmm2
-; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 256(%rax), %xmm2
-; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1]
-; AVX2-ONLY-NEXT:    vmovaps 256(%rdi), %ymm2
-; AVX2-ONLY-NEXT:    vmovaps 256(%rsi), %ymm3
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm2[0],ymm3[0],ymm2[2],ymm3[2]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm4[4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 256(%r8), %ymm1
-; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 256(%r9), %ymm4
-; AVX2-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm4[0],ymm1[2],ymm4[2]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm0[0],mem[0],ymm0[2],mem[2]
-; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm4[2,3],ymm1[2,3]
+; AVX2-ONLY-NEXT:    vmovaps 256(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 256(%rsi), %ymm1
 ; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm2[1],ymm3[1],ymm2[3],ymm3[3]
-; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,3,3]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,3,3]
+; AVX2-ONLY-NEXT:    vmovaps 256(%rdx), %ymm4
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm4[6,7]
 ; AVX2-ONLY-NEXT:    vmovaps 272(%rax), %xmm1
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 288(%r9), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-ONLY-NEXT:    vmovaps 288(%r8), %xmm0
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm0[0],xmm1[0]
 ; AVX2-ONLY-NEXT:    vmovaps 288(%rax), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
-; AVX2-ONLY-NEXT:    vmovaps 288(%rdi), %xmm3
-; AVX2-ONLY-NEXT:    vmovaps %xmm3, (%rsp) # 16-byte Spill
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 288(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 288(%rdx), %xmm3
-; AVX2-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm3[1]
-; AVX2-ONLY-NEXT:    vbroadcastsd 296(%rcx), %ymm3
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 288(%rdx), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 288(%rsi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm2[1],xmm1[1]
+; AVX2-ONLY-NEXT:    vbroadcastsd 296(%rcx), %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm2[1]
-; AVX2-ONLY-NEXT:    vmovaps 288(%rdi), %ymm1
-; AVX2-ONLY-NEXT:    vmovaps 288(%rsi), %ymm2
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovaps 288(%rdi), %ymm0
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 288(%rdx), %ymm0
-; AVX2-ONLY-NEXT:    vmovaps 288(%r8), %ymm3
-; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 288(%r9), %ymm4
-; AVX2-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm3[0],ymm4[0],ymm3[2],ymm4[2]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm0[0],mem[0],ymm0[2],mem[2]
-; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm4[2,3],ymm3[2,3]
-; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm2[1],ymm1[3],ymm2[3]
-; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,3,3]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 288(%rsi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,3,3]
+; AVX2-ONLY-NEXT:    vmovaps 288(%rdx), %ymm14
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm14[6,7]
 ; AVX2-ONLY-NEXT:    vmovaps 304(%rax), %xmm1
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 320(%rdx), %ymm0
-; AVX2-ONLY-NEXT:    vbroadcastsd 328(%rcx), %ymm1
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm2 = mem[0,0]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovaps 320(%r8), %xmm7
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm7, %ymm0, %ymm2
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 320(%rax), %xmm2
-; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm2[1]
-; AVX2-ONLY-NEXT:    vmovaps 320(%rdi), %ymm2
-; AVX2-ONLY-NEXT:    vmovaps 320(%rsi), %ymm3
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm2[0],ymm3[0],ymm2[2],ymm3[2]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm4[4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 320(%r8), %ymm4
-; AVX2-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 320(%r9), %ymm1
-; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm4[0],ymm1[0],ymm4[2],ymm1[2]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm0[0],mem[0],ymm0[2],mem[2]
-; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm4[2,3],ymm1[2,3]
-; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm2[1],ymm3[1],ymm2[3],ymm3[3]
-; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,3,3]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 320(%rdi), %ymm10
+; AVX2-ONLY-NEXT:    vmovaps 320(%rsi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm10[1],ymm0[1],ymm10[3],ymm0[3]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,3,3]
+; AVX2-ONLY-NEXT:    vmovaps 320(%rdx), %ymm3
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm3[6,7]
 ; AVX2-ONLY-NEXT:    vmovaps 336(%rax), %xmm1
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 352(%r9), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-ONLY-NEXT:    vmovaps 352(%r8), %xmm0
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm0[0],xmm1[0]
 ; AVX2-ONLY-NEXT:    vmovaps 352(%rax), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
-; AVX2-ONLY-NEXT:    vmovaps 352(%rdi), %xmm3
-; AVX2-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 352(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 352(%rdx), %xmm3
-; AVX2-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm3[1]
-; AVX2-ONLY-NEXT:    vbroadcastsd 360(%rcx), %ymm3
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 352(%rdx), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 352(%rsi), %xmm2
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm2[1],xmm1[1]
+; AVX2-ONLY-NEXT:    vbroadcastsd 360(%rcx), %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm2[1]
-; AVX2-ONLY-NEXT:    vmovaps 352(%rdi), %ymm1
-; AVX2-ONLY-NEXT:    vmovaps 352(%rsi), %ymm2
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovaps 352(%rdi), %ymm0
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 352(%rdx), %ymm0
-; AVX2-ONLY-NEXT:    vmovaps 352(%r8), %ymm4
-; AVX2-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 352(%r9), %ymm3
-; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm4[0],ymm3[0],ymm4[2],ymm3[2]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm0[0],mem[0],ymm0[2],mem[2]
-; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm4[2,3],ymm3[2,3]
-; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm2[1],ymm1[3],ymm2[3]
-; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,3,3]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 352(%rsi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,3,3]
+; AVX2-ONLY-NEXT:    vmovaps 352(%rdx), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vmovaps 368(%rax), %xmm1
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 384(%rdx), %ymm0
-; AVX2-ONLY-NEXT:    vbroadcastsd 392(%rcx), %ymm1
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm2 = mem[0,0]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovaps 384(%r8), %xmm2
-; AVX2-ONLY-NEXT:    vmovaps %xmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm2
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 384(%rax), %xmm6
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm6[1]
-; AVX2-ONLY-NEXT:    vmovaps 384(%rdi), %ymm3
-; AVX2-ONLY-NEXT:    vmovaps 384(%rsi), %ymm4
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm3[0],ymm4[0],ymm3[2],ymm4[2]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovaps 384(%rdi), %ymm1
 ; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 384(%r8), %ymm1
-; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 384(%r9), %ymm2
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[2],ymm2[2]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm5 = ymm0[0],mem[0],ymm0[2],mem[2]
-; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm5[2,3],ymm1[2,3]
-; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm3[1],ymm4[1],ymm3[3],ymm4[3]
-; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,3,3]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 384(%rsi), %ymm0
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm0[1],ymm1[3],ymm0[3]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,3,3]
+; AVX2-ONLY-NEXT:    vmovaps 384(%rdx), %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
 ; AVX2-ONLY-NEXT:    vmovaps 400(%rax), %xmm1
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 416(%r9), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; AVX2-ONLY-NEXT:    vmovaps 416(%r8), %xmm0
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
-; AVX2-ONLY-NEXT:    vmovaps 416(%rax), %xmm3
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm1, %ymm1
-; AVX2-ONLY-NEXT:    vmovaps 416(%rdi), %xmm4
-; AVX2-ONLY-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm0, %ymm4
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm4[6,7]
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm1 = xmm0[0],xmm1[0]
+; AVX2-ONLY-NEXT:    vmovaps 416(%rax), %xmm8
+; AVX2-ONLY-NEXT:    vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm1, %ymm1
+; AVX2-ONLY-NEXT:    vmovaps 416(%rdi), %xmm8
+; AVX2-ONLY-NEXT:    vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm0, %ymm8
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm8[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 416(%rdx), %xmm12
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],xmm12[1]
-; AVX2-ONLY-NEXT:    vbroadcastsd 424(%rcx), %ymm4
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm4[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 416(%rdx), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 416(%rsi), %xmm8
+; AVX2-ONLY-NEXT:    vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm8[1],xmm1[1]
+; AVX2-ONLY-NEXT:    vbroadcastsd 424(%rcx), %ymm8
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm8[4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm0[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm3[1]
-; AVX2-ONLY-NEXT:    vmovaps 416(%rdi), %ymm1
-; AVX2-ONLY-NEXT:    vmovaps 416(%rsi), %ymm3
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm1[0],ymm3[0],ymm1[2],ymm3[2]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm4[4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 416(%r8), %ymm0
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[2],mem[2]
-; AVX2-ONLY-NEXT:    vmovaps 416(%rdx), %ymm4
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm5 = ymm4[0],mem[0],ymm4[2],mem[2]
-; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm5[2,3],ymm0[2,3]
+; AVX2-ONLY-NEXT:    vmovaps 416(%rdi), %ymm0
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm1[1],ymm3[1],ymm1[3],ymm3[3]
+; AVX2-ONLY-NEXT:    vmovaps 416(%rsi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
 ; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,3,3]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm4[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 416(%rdx), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm1[6,7]
 ; AVX2-ONLY-NEXT:    vmovaps 432(%rax), %xmm1
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 448(%rdi), %ymm1
-; AVX2-ONLY-NEXT:    vmovaps 448(%r8), %ymm13
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
-; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm13[0,1],ymm1[0,1]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1],ymm0[2,3],ymm3[4,5,6,7]
-; AVX2-ONLY-NEXT:    vbroadcastsd 448(%rax), %ymm3
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm3[4,5],ymm0[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 448(%rdx), %ymm8
-; AVX2-ONLY-NEXT:    vbroadcastsd 456(%rcx), %ymm3
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm8[0,1,2,3],ymm3[4,5],ymm8[6,7]
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm4 = mem[0,0]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1],ymm3[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vinsertf128 $1, 448(%r8), %ymm0, %ymm4
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm4[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 448(%rsi), %ymm3
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm4 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm5 = ymm1[0],ymm3[0],ymm1[2],ymm3[2]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1],ymm5[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm3[1],ymm1[3],ymm3[3]
+; AVX2-ONLY-NEXT:    vmovaps 448(%rdi), %ymm0
+; AVX2-ONLY-NEXT:    vmovaps 448(%rsi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
 ; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,3,3]
+; AVX2-ONLY-NEXT:    vmovaps 448(%rdx), %ymm11
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm11[6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm11, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 464(%rax), %xmm8
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm8[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 480(%rdx), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 480(%rsi), %xmm8
+; AVX2-ONLY-NEXT:    vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm8[1],xmm1[1]
+; AVX2-ONLY-NEXT:    vbroadcastsd 488(%rcx), %ymm8
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm8[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, 480(%r8), %ymm0, %ymm8
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm8[6,7]
-; AVX2-ONLY-NEXT:    vmovaps 464(%rax), %xmm3
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1],ymm1[2,3,4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 480(%rdi), %ymm1
-; AVX2-ONLY-NEXT:    vmovaps 480(%r8), %ymm14
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm3 = mem[0,0]
-; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm4 = ymm14[0,1],ymm1[0,1]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm4[0,1],ymm3[2,3],ymm4[4,5,6,7]
-; AVX2-ONLY-NEXT:    vbroadcastsd 480(%rax), %ymm4
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm4[4,5],ymm3[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 480(%rdx), %xmm10
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 480(%rdi), %ymm12
+; AVX2-ONLY-NEXT:    vmovaps 480(%rsi), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm12[1],ymm1[1],ymm12[3],ymm1[3]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,3,3]
+; AVX2-ONLY-NEXT:    vmovaps 480(%rdx), %ymm8
+; AVX2-ONLY-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm8[6,7]
+; AVX2-ONLY-NEXT:    vmovaps 496(%rax), %xmm8
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm8[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastsd 8(%rcx), %ymm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm9[0,1,2,3],ymm1[4,5],ymm9[6,7]
+; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm8 = mem[0,0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm8[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovaps (%r8), %xmm8
+; AVX2-ONLY-NEXT:    vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm0, %ymm8
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm8[6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],ymm15[0],ymm1[2],ymm15[2]
+; AVX2-ONLY-NEXT:    vmovaps (%rax), %xmm8
+; AVX2-ONLY-NEXT:    vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps (%r9), %xmm13
+; AVX2-ONLY-NEXT:    vmovaps %xmm13, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm8 = xmm13[1],xmm8[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm8[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:    vunpcklpd {{.*#+}} ymm1 = ymm9[0],mem[0],ymm9[2],mem[2]
+; AVX2-ONLY-NEXT:    vmovaps (%r8), %ymm8
+; AVX2-ONLY-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps (%r9), %ymm9
+; AVX2-ONLY-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm8 = ymm8[0],ymm9[0],ymm8[2],ymm9[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3],ymm8[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = ymm1[0],mem[0],ymm1[2],mem[2]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm8 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm8, %xmm8 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm8 = xmm8[1],mem[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm8[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],mem[0],ymm1[2],mem[2]
+; AVX2-ONLY-NEXT:    vmovaps 32(%r8), %ymm8
+; AVX2-ONLY-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 32(%r9), %ymm9
+; AVX2-ONLY-NEXT:    vmovups %ymm9, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm8 = ymm8[0],ymm9[0],ymm8[2],ymm9[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3],ymm8[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastsd 72(%rcx), %ymm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm7[0,1,2,3],ymm1[4,5],ymm7[6,7]
+; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm8 = mem[0,0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm8[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovaps 64(%r8), %xmm8
+; AVX2-ONLY-NEXT:    vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm8, %ymm0, %ymm8
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm8[6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = ymm1[0],mem[0],ymm1[2],mem[2]
+; AVX2-ONLY-NEXT:    vmovaps 64(%rax), %xmm8
+; AVX2-ONLY-NEXT:    vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 64(%r9), %xmm13
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm8 = xmm13[1],xmm8[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm8[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:    vunpcklpd {{.*#+}} ymm1 = ymm7[0],mem[0],ymm7[2],mem[2]
+; AVX2-ONLY-NEXT:    vmovaps 64(%r8), %ymm7
+; AVX2-ONLY-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 64(%r9), %ymm8
+; AVX2-ONLY-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm7 = ymm7[0],ymm8[0],ymm7[2],ymm8[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3],ymm7[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = ymm1[0],mem[0],ymm1[2],mem[2]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm7 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm7, %xmm7 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm7 = xmm7[1],mem[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm7[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],mem[0],ymm1[2],mem[2]
+; AVX2-ONLY-NEXT:    vmovaps 96(%r8), %ymm7
+; AVX2-ONLY-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 96(%r9), %ymm8
+; AVX2-ONLY-NEXT:    vmovups %ymm8, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm7 = ymm7[0],ymm8[0],ymm7[2],ymm8[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3],ymm7[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastsd 136(%rcx), %ymm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm6[0,1,2,3],ymm1[4,5],ymm6[6,7]
+; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm7 = mem[0,0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm7[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovaps 128(%r8), %xmm7
+; AVX2-ONLY-NEXT:    vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm7, %ymm0, %ymm7
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm7[6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = ymm1[0],mem[0],ymm1[2],mem[2]
+; AVX2-ONLY-NEXT:    vmovaps 128(%rax), %xmm7
+; AVX2-ONLY-NEXT:    vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 128(%r9), %xmm8
+; AVX2-ONLY-NEXT:    vmovaps %xmm8, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm7 = xmm8[1],xmm7[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm7[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm6[0],mem[0],ymm6[2],mem[2]
+; AVX2-ONLY-NEXT:    vmovaps 128(%r8), %ymm6
+; AVX2-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 128(%r9), %ymm7
+; AVX2-ONLY-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm6 = ymm6[0],ymm7[0],ymm6[2],ymm7[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3],ymm6[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = ymm1[0],mem[0],ymm1[2],mem[2]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm6 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm6, %xmm6 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm6 = xmm6[1],mem[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm6[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],mem[0],ymm1[2],mem[2]
+; AVX2-ONLY-NEXT:    vmovaps 160(%r8), %ymm6
+; AVX2-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 160(%r9), %ymm7
+; AVX2-ONLY-NEXT:    vmovups %ymm7, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm6 = ymm6[0],ymm7[0],ymm6[2],ymm7[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3],ymm6[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastsd 200(%rcx), %ymm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm5[0,1,2,3],ymm1[4,5],ymm5[6,7]
+; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm6 = mem[0,0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm6[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovaps 192(%r8), %xmm6
+; AVX2-ONLY-NEXT:    vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm6, %ymm0, %ymm6
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm6[6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = ymm1[0],mem[0],ymm1[2],mem[2]
+; AVX2-ONLY-NEXT:    vmovaps 192(%r9), %xmm7
+; AVX2-ONLY-NEXT:    vmovaps %xmm7, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 192(%rax), %xmm6
+; AVX2-ONLY-NEXT:    vmovaps %xmm6, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm6 = xmm7[1],xmm6[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm6[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm5[0],mem[0],ymm5[2],mem[2]
+; AVX2-ONLY-NEXT:    vmovaps 192(%r8), %ymm5
+; AVX2-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 192(%r9), %ymm6
+; AVX2-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm5 = ymm5[0],ymm6[0],ymm5[2],ymm6[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3],ymm5[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = ymm1[0],mem[0],ymm1[2],mem[2]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm5 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm5, %xmm5 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm5 = xmm5[1],mem[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm5[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],mem[0],ymm1[2],mem[2]
+; AVX2-ONLY-NEXT:    vmovaps 224(%r8), %ymm5
+; AVX2-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 224(%r9), %ymm6
+; AVX2-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm5 = ymm5[0],ymm6[0],ymm5[2],ymm6[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3],ymm5[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastsd 264(%rcx), %ymm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm4[0,1,2,3],ymm1[4,5],ymm4[6,7]
+; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm5 = mem[0,0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm5[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovaps 256(%r8), %xmm5
+; AVX2-ONLY-NEXT:    vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm5
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm5[6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = ymm1[0],mem[0],ymm1[2],mem[2]
+; AVX2-ONLY-NEXT:    vmovaps 256(%rax), %xmm5
+; AVX2-ONLY-NEXT:    vmovaps %xmm5, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 256(%r9), %xmm15
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm5 = xmm15[1],xmm5[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm5[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm4[0],mem[0],ymm4[2],mem[2]
+; AVX2-ONLY-NEXT:    vmovaps 256(%r8), %ymm4
+; AVX2-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 256(%r9), %ymm5
+; AVX2-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm4[0],ymm5[0],ymm4[2],ymm5[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3],ymm4[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = ymm1[0],mem[0],ymm1[2],mem[2]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm4 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm4, %xmm4 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm4 = xmm4[1],mem[1]
+; 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:    vunpcklpd {{.*#+}} ymm1 = ymm14[0],mem[0],ymm14[2],mem[2]
+; AVX2-ONLY-NEXT:    vmovaps 288(%r8), %ymm4
+; AVX2-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 288(%r9), %ymm5
+; AVX2-ONLY-NEXT:    vmovups %ymm5, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm4[0],ymm5[0],ymm4[2],ymm5[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3],ymm4[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastsd 328(%rcx), %ymm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm4 = mem[0,0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm4[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovaps 320(%r8), %xmm4
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm0, %ymm4
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm4[6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = ymm10[0],mem[0],ymm10[2],mem[2]
+; AVX2-ONLY-NEXT:    vmovaps 320(%r9), %xmm6
+; AVX2-ONLY-NEXT:    vmovaps 320(%rax), %xmm4
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm4 = xmm6[1],xmm4[1]
+; 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:    vunpcklpd {{.*#+}} ymm1 = ymm3[0],mem[0],ymm3[2],mem[2]
+; AVX2-ONLY-NEXT:    vmovaps 320(%r8), %ymm14
+; AVX2-ONLY-NEXT:    vmovaps 320(%r9), %ymm10
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm14[0],ymm10[0],ymm14[2],ymm10[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3],ymm3[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = ymm1[0],mem[0],ymm1[2],mem[2]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm3 = xmm3[1],mem[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],mem[0],ymm1[2],mem[2]
+; AVX2-ONLY-NEXT:    vmovaps 352(%r8), %ymm3
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 352(%r9), %ymm4
+; AVX2-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm3 = ymm3[0],ymm4[0],ymm3[2],ymm4[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3],ymm3[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastsd 392(%rcx), %ymm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5],ymm2[6,7]
 ; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm3 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm3 = xmm3[1],xmm10[1]
-; AVX2-ONLY-NEXT:    vbroadcastsd 488(%rcx), %ymm4
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm4[4,5],ymm3[6,7]
-; AVX2-ONLY-NEXT:    vinsertf128 $1, 480(%r8), %ymm0, %ymm4
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm4[6,7]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovaps 384(%r8), %xmm3
+; AVX2-ONLY-NEXT:    vmovaps %xmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm3
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = ymm1[0],mem[0],ymm1[2],mem[2]
+; AVX2-ONLY-NEXT:    vmovaps 384(%r9), %xmm4
+; AVX2-ONLY-NEXT:    vmovaps 384(%rax), %xmm3
+; AVX2-ONLY-NEXT:    vmovaps %xmm3, (%rsp) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} xmm3 = xmm4[1],xmm3[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm3[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:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],mem[0],ymm2[2],mem[2]
+; AVX2-ONLY-NEXT:    vmovaps 384(%r8), %ymm3
+; AVX2-ONLY-NEXT:    vmovaps 384(%r9), %ymm9
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm3[0],ymm9[0],ymm3[2],ymm9[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3],ymm2[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm1 = ymm1[0],mem[0],ymm1[2],mem[2]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm2 = xmm2[1],mem[1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3],ymm1[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 416(%rcx), %ymm1
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[2],ymm1[2]
+; AVX2-ONLY-NEXT:    vmovaps 416(%r8), %ymm2
+; AVX2-ONLY-NEXT:    vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm2[0],mem[0],ymm2[2],mem[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm1 = ymm1[2,3],ymm2[2,3]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 448(%r8), %ymm8
+; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm8[0,1],ymm0[0,1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT:    vbroadcastsd 448(%rax), %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vbroadcastsd 456(%rcx), %ymm1
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm11[0,1,2,3],ymm1[4,5],ymm11[6,7]
+; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm2 = mem[0,0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, 448(%r8), %ymm0, %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm2[6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[0],mem[0],ymm0[2],mem[2]
+; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 480(%rsi), %ymm3
-; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm4 = mem[0,0]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm5 = ymm1[0],ymm3[0],ymm1[2],ymm3[2]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1],ymm5[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovaps 480(%r8), %ymm7
+; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm7[0,1],ymm12[0,1]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3],ymm2[4,5,6,7]
+; AVX2-ONLY-NEXT:    vbroadcastsd 480(%rax), %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5],ymm0[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm1[1],ymm3[1],ymm1[3],ymm3[3]
-; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,3,3]
-; AVX2-ONLY-NEXT:    vmovaps 480(%rdx), %ymm5
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],ymm5[6,7]
-; AVX2-ONLY-NEXT:    vmovaps 496(%rax), %xmm3
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm12, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm12[0],mem[0],ymm12[2],mem[2]
+; AVX2-ONLY-NEXT:    vmovddup {{.*#+}} xmm2 = mem[0,0]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm9[0],mem[0]
-; AVX2-ONLY-NEXT:    vmovaps (%rdi), %xmm0
-; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm1
-; AVX2-ONLY-NEXT:    vbroadcastsd %xmm11, %ymm3
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm3[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm0 = xmm0[0],mem[0]
+; AVX2-ONLY-NEXT:    vmovaps (%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-ONLY-NEXT:    vbroadcastsd {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5],ymm0[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm1 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,3,3]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],mem[6,7]
-; AVX2-ONLY-NEXT:    vbroadcastsd 24(%rcx), %ymm3
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,3,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
+; AVX2-ONLY-NEXT:    vbroadcastsd 24(%rcx), %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm15[0],mem[0]
-; AVX2-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 16-byte Folded Reload
-; AVX2-ONLY-NEXT:    vbroadcastsd 32(%rcx), %ymm3
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm0 = xmm0[0],mem[0]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    vbroadcastsd 32(%rcx), %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm1 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,3,3]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],mem[6,7]
-; AVX2-ONLY-NEXT:    vbroadcastsd 56(%rcx), %ymm3
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,3,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
+; AVX2-ONLY-NEXT:    vbroadcastsd 56(%rcx), %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3,4,5,6,7]
 ; 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:    vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
-; AVX2-ONLY-NEXT:    vmovaps 64(%rdi), %xmm0
-; AVX2-ONLY-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm1
-; AVX2-ONLY-NEXT:    vbroadcastsd {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 16-byte Folded Reload
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm3[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm0 = xmm0[0],xmm13[0]
+; AVX2-ONLY-NEXT:    vmovaps 64(%rdi), %xmm13
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm13, %ymm0, %ymm0
+; AVX2-ONLY-NEXT:    vbroadcastsd {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5],ymm0[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm1 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,3,3]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],mem[6,7]
-; AVX2-ONLY-NEXT:    vbroadcastsd 88(%rcx), %ymm3
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,3,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
+; AVX2-ONLY-NEXT:    vbroadcastsd 88(%rcx), %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3,4,5,6,7]
 ; 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:    vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
-; AVX2-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 16-byte Folded Reload
-; AVX2-ONLY-NEXT:    vbroadcastsd 96(%rcx), %ymm3
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm0 = xmm0[0],mem[0]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    vbroadcastsd 96(%rcx), %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm1 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,3,3]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],mem[6,7]
-; AVX2-ONLY-NEXT:    vbroadcastsd 120(%rcx), %ymm3
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,3,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
+; AVX2-ONLY-NEXT:    vbroadcastsd 120(%rcx), %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3,4,5,6,7]
 ; 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:    vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
-; AVX2-ONLY-NEXT:    vmovaps 128(%rdi), %xmm11
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm11, %ymm1, %ymm1
-; AVX2-ONLY-NEXT:    vbroadcastsd {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 16-byte Folded Reload
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm3[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm0 = xmm0[0],mem[0]
+; AVX2-ONLY-NEXT:    vmovaps 128(%rdi), %xmm12
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm12, %ymm0, %ymm0
+; AVX2-ONLY-NEXT:    vbroadcastsd {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5],ymm0[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm1 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,3,3]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],mem[6,7]
-; AVX2-ONLY-NEXT:    vbroadcastsd 152(%rcx), %ymm3
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,3,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
+; AVX2-ONLY-NEXT:    vbroadcastsd 152(%rcx), %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3,4,5,6,7]
 ; 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:    vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
-; AVX2-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 16-byte Folded Reload
-; AVX2-ONLY-NEXT:    vbroadcastsd 160(%rcx), %ymm3
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm0 = xmm0[0],mem[0]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    vbroadcastsd 160(%rcx), %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm1 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,3,3]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],mem[6,7]
-; AVX2-ONLY-NEXT:    vbroadcastsd 184(%rcx), %ymm3
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,3,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
+; AVX2-ONLY-NEXT:    vbroadcastsd 184(%rcx), %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3,4,5,6,7]
 ; 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:    vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
-; AVX2-ONLY-NEXT:    vmovaps 192(%rdi), %xmm9
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm9, %ymm1, %ymm1
-; AVX2-ONLY-NEXT:    vbroadcastsd {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 16-byte Folded Reload
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm3[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm0 = xmm0[0],mem[0]
+; AVX2-ONLY-NEXT:    vmovaps 192(%rdi), %xmm11
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm11, %ymm0, %ymm0
+; AVX2-ONLY-NEXT:    vbroadcastsd {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5],ymm0[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm1 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,3,3]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],mem[6,7]
-; AVX2-ONLY-NEXT:    vbroadcastsd 216(%rcx), %ymm3
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,3,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
+; AVX2-ONLY-NEXT:    vbroadcastsd 216(%rcx), %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3,4,5,6,7]
 ; 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:    vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
-; AVX2-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 16-byte Folded Reload
-; AVX2-ONLY-NEXT:    vbroadcastsd 224(%rcx), %ymm3
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm0 = xmm0[0],mem[0]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    vbroadcastsd 224(%rcx), %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm1 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,3,3]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],mem[6,7]
-; AVX2-ONLY-NEXT:    vbroadcastsd 248(%rcx), %ymm3
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,3,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
+; AVX2-ONLY-NEXT:    vbroadcastsd 248(%rcx), %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3,4,5,6,7]
 ; 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:    vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
-; AVX2-ONLY-NEXT:    vmovaps 256(%rdi), %xmm15
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm15, %ymm1, %ymm1
-; AVX2-ONLY-NEXT:    vbroadcastsd {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 16-byte Folded Reload
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm3[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm0 = xmm0[0],xmm15[0]
+; AVX2-ONLY-NEXT:    vmovaps 256(%rdi), %xmm5
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm5, %ymm0, %ymm0
+; AVX2-ONLY-NEXT:    vbroadcastsd {{[-0-9]+}}(%r{{[sb]}}p), %ymm2 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm2[4,5],ymm0[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm1 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,3,3]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],mem[6,7]
-; AVX2-ONLY-NEXT:    vbroadcastsd 280(%rcx), %ymm3
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,3,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
+; AVX2-ONLY-NEXT:    vbroadcastsd 280(%rcx), %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps (%rsp), %xmm0 # 16-byte Reload
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
-; AVX2-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 16-byte Folded Reload
-; AVX2-ONLY-NEXT:    vbroadcastsd 288(%rcx), %ymm3
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm0 = xmm0[0],mem[0]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    vbroadcastsd 288(%rcx), %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm2[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm1 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,3,3]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],mem[6,7]
-; AVX2-ONLY-NEXT:    vbroadcastsd 312(%rcx), %ymm3
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,3,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
+; AVX2-ONLY-NEXT:    vbroadcastsd 312(%rcx), %ymm2
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1],ymm0[2,3,4,5,6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm7[0],mem[0]
-; AVX2-ONLY-NEXT:    vmovaps 320(%rdi), %xmm7
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm7, %ymm1, %ymm1
-; AVX2-ONLY-NEXT:    vbroadcastsd {{[-0-9]+}}(%r{{[sb]}}p), %ymm3 # 16-byte Folded Reload
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm3[4,5],ymm1[6,7]
+; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm0 = xmm0[0],xmm6[0]
+; AVX2-ONLY-NEXT:    vmovaps 320(%rdi), %xmm2
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX2-ONLY-NEXT:    vbroadcastsd {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm6[4,5],ymm0[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm1 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,3,3]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],mem[6,7]
-; AVX2-ONLY-NEXT:    vbroadcastsd 344(%rcx), %ymm3
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm14[1],ymm10[1],ymm14[3],ymm10[3]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,3,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
+; AVX2-ONLY-NEXT:    vbroadcastsd 344(%rcx), %ymm6
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm6[0,1],ymm0[2,3,4,5,6,7]
 ; 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:    vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
-; AVX2-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm1, %ymm1 # 16-byte Folded Reload
-; AVX2-ONLY-NEXT:    vbroadcastsd 352(%rcx), %ymm3
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm0 = xmm0[0],mem[0]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    vbroadcastsd 352(%rcx), %ymm6
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],ymm6[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm1 = ymm0[1],mem[1],ymm0[3],mem[3]
-; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm1 = ymm1[0,2,3,3]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3,4,5],mem[6,7]
-; AVX2-ONLY-NEXT:    vbroadcastsd 376(%rcx), %ymm3
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1],ymm1[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm0 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[0,2,3,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
+; AVX2-ONLY-NEXT:    vbroadcastsd 376(%rcx), %ymm6
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm6[0,1],ymm0[2,3,4,5,6,7]
 ; 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:    vunpcklpd {{.*#+}} xmm3 = xmm0[0],mem[0]
-; AVX2-ONLY-NEXT:    vmovaps 384(%rdi), %xmm4
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm4, %ymm3, %ymm3
-; AVX2-ONLY-NEXT:    vbroadcastsd %xmm6, %ymm6
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3],ymm6[4,5],ymm3[6,7]
+; AVX2-ONLY-NEXT:    vmovlhps {{.*#+}} xmm4 = xmm0[0],xmm4[0]
+; AVX2-ONLY-NEXT:    vmovaps 384(%rdi), %xmm1
+; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm1, %ymm4, %ymm4
+; AVX2-ONLY-NEXT:    vbroadcastsd (%rsp), %ymm6 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1,2,3],ymm6[4,5],ymm4[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm2 = ymm0[1],ymm2[1],ymm0[3],ymm2[3]
-; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm2 = ymm2[0,2,3,3]
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],mem[6,7]
-; AVX2-ONLY-NEXT:    vbroadcastsd 408(%rcx), %ymm3
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1],ymm2[2,3,4,5,6,7]
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm3 = ymm3[1],ymm9[1],ymm3[3],ymm9[3]
+; AVX2-ONLY-NEXT:    vpermpd {{.*#+}} ymm3 = ymm3[0,2,3,3]
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],mem[6,7]
+; AVX2-ONLY-NEXT:    vbroadcastsd 408(%rcx), %ymm4
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm4[0,1],ymm3[2,3,4,5,6,7]
 ; 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:    vunpcklpd {{.*#+}} xmm2 = xmm0[0],mem[0]
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm12, %ymm2, %ymm2
-; AVX2-ONLY-NEXT:    vbroadcastsd 416(%rcx), %ymm3
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3,4,5],ymm3[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vbroadcastsd 440(%rcx), %ymm2
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0,1],mem[2,3]
-; AVX2-ONLY-NEXT:    vbroadcastsd 440(%r9), %ymm3
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm3[4,5],ymm2[6,7]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm8[0],mem[0],ymm8[2],mem[2]
-; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm2[2,3],ymm13[2,3]
-; AVX2-ONLY-NEXT:    vbroadcastsd 464(%r9), %ymm3
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm3[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vbroadcastsd 472(%rcx), %ymm2
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0,1],mem[2,3]
-; AVX2-ONLY-NEXT:    vbroadcastsd 472(%r9), %ymm3
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5],ymm2[6,7]
-; AVX2-ONLY-NEXT:    vmovaps 448(%rax), %ymm3
-; AVX2-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm1 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm1 = mem[0,1],ymm3[2,3],mem[4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm3[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vmovaps 480(%rdi), %xmm2
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm2 = xmm2[0],mem[0]
-; AVX2-ONLY-NEXT:    vinsertf128 $1, %xmm10, %ymm2, %ymm2
-; AVX2-ONLY-NEXT:    vbroadcastsd 480(%rcx), %ymm3
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm3[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm2 = ymm5[0],mem[0],ymm5[2],mem[2]
-; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm2[2,3],ymm14[2,3]
-; AVX2-ONLY-NEXT:    vbroadcastsd 496(%r9), %ymm3
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm3[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vbroadcastsd 504(%rcx), %ymm2
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} xmm2 = xmm2[0,1],mem[2,3]
-; AVX2-ONLY-NEXT:    vbroadcastsd 504(%r9), %ymm3
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm3[4,5],ymm2[6,7]
-; AVX2-ONLY-NEXT:    vmovaps 480(%rax), %ymm8
-; AVX2-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm8, %ymm1 # 32-byte Folded Reload
-; AVX2-ONLY-NEXT:    # ymm1 = mem[0,1],ymm8[2,3],mem[4,5,6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm1, (%rsp) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm1 = ymm2[0,1,2,3,4,5],ymm8[6,7]
-; AVX2-ONLY-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm3 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm3 = xmm0[0],mem[0]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    vbroadcastsd 416(%rcx), %ymm4
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm4[6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 480(%rdi), %xmm3
+; AVX2-ONLY-NEXT:    vunpcklpd {{[-0-9]+}}(%r{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    # xmm3 = xmm3[0],mem[0]
+; AVX2-ONLY-NEXT:    vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 16-byte Folded Reload
+; AVX2-ONLY-NEXT:    vbroadcastsd 480(%rcx), %ymm4
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm3[0,1,2,3,4,5],ymm4[6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpckhpd {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm3 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm3 = ymm0[1],mem[1],ymm0[3],mem[3]
+; AVX2-ONLY-NEXT:    vbroadcastsd 440(%r9), %ymm4
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm0 = ymm3[2,3],ymm4[2,3]
+; AVX2-ONLY-NEXT:    vmovaps 448(%rcx), %ymm3
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm4[0],ymm3[0],ymm4[2],ymm3[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm4 = ymm4[2,3],ymm8[2,3]
+; AVX2-ONLY-NEXT:    vbroadcastsd 464(%r9), %ymm6
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm6[6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm3 = ymm3[1],ymm8[1],ymm3[3],ymm8[3]
+; AVX2-ONLY-NEXT:    vbroadcastsd 472(%r9), %ymm4
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm3[2,3],ymm4[2,3]
+; AVX2-ONLY-NEXT:    vmovaps 448(%rax), %ymm4
+; AVX2-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm4, %ymm6 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm6 = mem[0,1],ymm4[2,3],mem[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm6, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm4[6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vmovaps 480(%rcx), %ymm3
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm4 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} ymm4 = ymm4[0],ymm3[0],ymm4[2],ymm3[2]
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm4 = ymm4[2,3],ymm7[2,3]
+; AVX2-ONLY-NEXT:    vbroadcastsd 496(%r9), %ymm6
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm4 = ymm4[0,1,2,3,4,5],ymm6[6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vunpckhpd {{.*#+}} ymm3 = ymm3[1],ymm7[1],ymm3[3],ymm7[3]
+; AVX2-ONLY-NEXT:    vbroadcastsd 504(%r9), %ymm4
+; AVX2-ONLY-NEXT:    vperm2f128 {{.*#+}} ymm3 = ymm3[2,3],ymm4[2,3]
+; AVX2-ONLY-NEXT:    vmovaps 480(%rax), %ymm7
+; AVX2-ONLY-NEXT:    vblendps $243, {{[-0-9]+}}(%r{{[sb]}}p), %ymm7, %ymm4 # 32-byte Folded Reload
+; AVX2-ONLY-NEXT:    # ymm4 = mem[0,1],ymm7[2,3],mem[4,5,6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm7[6,7]
+; AVX2-ONLY-NEXT:    vmovups %ymm3, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; AVX2-ONLY-NEXT:    movq {{[0-9]+}}(%rsp), %r8
 ; AVX2-ONLY-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4,5],mem[6,7]
 ; AVX2-ONLY-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm3 = xmm15[0],mem[0]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm5 = xmm5[0],mem[0]
 ; AVX2-ONLY-NEXT:    vmovaps 256(%rdx), %xmm0
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm2 = xmm0[0],mem[0]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm11 = xmm11[0],mem[0]
-; AVX2-ONLY-NEXT:    vmovaps 128(%rdx), %xmm15
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm15 = xmm15[0],mem[0]
-; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0]
-; AVX2-ONLY-NEXT:    vmovaps 64(%rdx), %xmm12
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm3 = xmm0[0],mem[0]
 ; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm12 = xmm12[0],mem[0]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm9 = xmm9[0],mem[0]
-; AVX2-ONLY-NEXT:    vmovaps 192(%rdx), %xmm13
+; AVX2-ONLY-NEXT:    vmovaps 128(%rdx), %xmm14
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm14 = xmm14[0],mem[0]
 ; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm13 = xmm13[0],mem[0]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm7 = xmm7[0],mem[0]
-; AVX2-ONLY-NEXT:    vmovaps 320(%rdx), %xmm6
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm6 = xmm6[0],mem[0]
-; AVX2-ONLY-NEXT:    vmovaps 448(%rdi), %xmm10
+; AVX2-ONLY-NEXT:    vmovaps 64(%rdx), %xmm15
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm15 = xmm15[0],mem[0]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm11 = xmm11[0],mem[0]
+; AVX2-ONLY-NEXT:    vmovaps 192(%rdx), %xmm10
 ; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm10 = xmm10[0],mem[0]
-; AVX2-ONLY-NEXT:    vmovaps 448(%rdx), %xmm8
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm2 = xmm2[0],mem[0]
+; AVX2-ONLY-NEXT:    vmovaps 320(%rdx), %xmm9
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm9 = xmm9[0],mem[0]
+; AVX2-ONLY-NEXT:    vmovaps 448(%rdi), %xmm8
 ; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm8 = xmm8[0],mem[0]
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm4 = xmm4[0],mem[0]
-; AVX2-ONLY-NEXT:    vmovaps 384(%rdx), %xmm14
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm14 = xmm14[0],mem[0]
+; AVX2-ONLY-NEXT:    vmovaps 448(%rdx), %xmm7
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm7 = xmm7[0],mem[0]
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm1[0],mem[0]
+; AVX2-ONLY-NEXT:    vmovaps 384(%rdx), %xmm6
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm6 = xmm6[0],mem[0]
 ; AVX2-ONLY-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],mem[0]
-; AVX2-ONLY-NEXT:    vmovaps (%rdx), %xmm5
-; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm5 = xmm5[0],mem[0]
-; AVX2-ONLY-NEXT:    vmovaps %xmm5, 16(%r8)
+; AVX2-ONLY-NEXT:    vmovaps (%rdx), %xmm4
+; AVX2-ONLY-NEXT:    vunpcklpd {{.*#+}} xmm4 = xmm4[0],mem[0]
+; AVX2-ONLY-NEXT:    vmovaps %xmm4, 16(%r8)
 ; AVX2-ONLY-NEXT:    vmovaps %xmm0, (%r8)
-; AVX2-ONLY-NEXT:    vmovaps %xmm14, 2704(%r8)
-; AVX2-ONLY-NEXT:    vmovaps %xmm4, 2688(%r8)
-; AVX2-ONLY-NEXT:    vmovaps %xmm8, 3152(%r8)
-; AVX2-ONLY-NEXT:    vmovaps %xmm10, 3136(%r8)
-; AVX2-ONLY-NEXT:    vmovaps %xmm6, 2256(%r8)
-; AVX2-ONLY-NEXT:    vmovaps %xmm7, 2240(%r8)
-; AVX2-ONLY-NEXT:    vmovaps %xmm13, 1360(%r8)
-; AVX2-ONLY-NEXT:    vmovaps %xmm9, 1344(%r8)
-; AVX2-ONLY-NEXT:    vmovaps %xmm12, 464(%r8)
-; AVX2-ONLY-NEXT:    vmovaps %xmm1, 448(%r8)
-; AVX2-ONLY-NEXT:    vmovaps %xmm15, 912(%r8)
-; AVX2-ONLY-NEXT:    vmovaps %xmm11, 896(%r8)
-; AVX2-ONLY-NEXT:    vmovaps %xmm2, 1808(%r8)
-; AVX2-ONLY-NEXT:    vmovaps %xmm3, 1792(%r8)
+; AVX2-ONLY-NEXT:    vmovaps %xmm6, 2704(%r8)
+; AVX2-ONLY-NEXT:    vmovaps %xmm1, 2688(%r8)
+; AVX2-ONLY-NEXT:    vmovaps %xmm7, 3152(%r8)
+; AVX2-ONLY-NEXT:    vmovaps %xmm8, 3136(%r8)
+; AVX2-ONLY-NEXT:    vmovaps %xmm9, 2256(%r8)
+; AVX2-ONLY-NEXT:    vmovaps %xmm2, 2240(%r8)
+; AVX2-ONLY-NEXT:    vmovaps %xmm10, 1360(%r8)
+; AVX2-ONLY-NEXT:    vmovaps %xmm11, 1344(%r8)
+; AVX2-ONLY-NEXT:    vmovaps %xmm15, 464(%r8)
+; AVX2-ONLY-NEXT:    vmovaps %xmm13, 448(%r8)
+; AVX2-ONLY-NEXT:    vmovaps %xmm14, 912(%r8)
+; AVX2-ONLY-NEXT:    vmovaps %xmm12, 896(%r8)
+; AVX2-ONLY-NEXT:    vmovaps %xmm3, 1808(%r8)
+; AVX2-ONLY-NEXT:    vmovaps %xmm5, 1792(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 3552(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 3520(%r8)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 3488(%r8)
-; AVX2-ONLY-NEXT:    vmovups (%rsp), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 3456(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 3424(%r8)
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 3456(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 3392(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 3360(%r8)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 3328(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 3296(%r8)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 3264(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 3232(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 3200(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 3168(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 3104(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 3072(%r8)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 3040(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 3008(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 2816(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 2784(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 2752(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 2592(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 2560(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 2368(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 2336(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 2304(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 2144(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 2112(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1920(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1888(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1856(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1696(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1664(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1472(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1440(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1408(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1248(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1216(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1024(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 992(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 960(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 800(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 768(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 576(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 544(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 512(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 352(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 320(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 128(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 96(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 64(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 3520(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 3424(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 3360(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 3296(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; AVX2-ONLY-NEXT:    vmovaps %ymm0, 3072(%r8)
+; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 2976(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 2944(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 2912(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 2880(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 2848(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 2816(%r8)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 2784(%r8)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 2752(%r8)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 2720(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 2656(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 2624(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 2592(%r8)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 2560(%r8)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 2528(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 2496(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 2464(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 2432(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 2400(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 2368(%r8)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 2336(%r8)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 2304(%r8)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 2272(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 2208(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 2176(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 2144(%r8)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 2112(%r8)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 2080(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 2048(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 2016(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1984(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1952(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1920(%r8)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1888(%r8)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1856(%r8)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1824(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1760(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1728(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1696(%r8)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1664(%r8)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1632(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1600(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1568(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1536(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1504(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1472(%r8)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1440(%r8)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1408(%r8)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1376(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1312(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1280(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1248(%r8)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1216(%r8)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1184(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1152(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1120(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1088(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1056(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 1024(%r8)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 992(%r8)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 960(%r8)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 928(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 864(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 832(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 800(%r8)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 768(%r8)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 736(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 704(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 672(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 640(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 608(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 576(%r8)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 544(%r8)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 512(%r8)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 480(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 416(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 384(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 352(%r8)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 320(%r8)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 288(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 256(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 224(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 192(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 160(%r8)
 ; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 128(%r8)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 96(%r8)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
-; AVX2-ONLY-NEXT:    vmovaps %ymm0, 64(%r8)
-; AVX2-ONLY-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; AVX2-ONLY-NEXT:    vmovaps %ymm0, 32(%r8)
-; AVX2-ONLY-NEXT:    addq $3624, %rsp # imm = 0xE28
+; AVX2-ONLY-NEXT:    addq $4264, %rsp # imm = 0x10A8
 ; AVX2-ONLY-NEXT:    vzeroupper
 ; AVX2-ONLY-NEXT:    retq
 ;
 ; AVX512F-ONLY-SLOW-LABEL: store_i64_stride7_vf64:
 ; AVX512F-ONLY-SLOW:       # %bb.0:
 ; AVX512F-ONLY-SLOW-NEXT:    subq $6408, %rsp # imm = 0x1908
 ; AVX512F-ONLY-SLOW-NEXT:    movq {{[0-9]+}}(%rsp), %rax
 ; AVX512F-ONLY-SLOW-NEXT:    vmovdqa64 (%rdi), %zmm5

llvm/test/CodeGen/X86/vector-narrow-binop.ll

 ; by a width of at least 1 of the original vector elements.
 ; https://bugs.llvm.org/show_bug.cgi?id=39893
 
 define <2 x i8> @PR39893(<2 x i32> %x, <8 x i8> %y) {
 ; SSE-LABEL: PR39893:
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    pxor %xmm2, %xmm2
 ; SSE-NEXT:    psubd %xmm0, %xmm2
-; SSE-NEXT:    psrld $16, %xmm2
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[1,1,1,1]
-; SSE-NEXT:    punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3],xmm2[4],xmm0[4],xmm2[5],xmm0[5],xmm2[6],xmm0[6],xmm2[7],xmm0[7]
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
+; SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2
+; SSE-NEXT:    psrlq $16, %xmm2
+; SSE-NEXT:    packuswb %xmm2, %xmm2
 ; SSE-NEXT:    movdqa %xmm2, %xmm0
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: PR39893:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vpxor %xmm2, %xmm2, %xmm2
 ; AVX-NEXT:    vpsubd %xmm0, %xmm2, %xmm0
-; AVX-NEXT:    vpsrld $16, %xmm0, %xmm0
-; AVX-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[1,1,1,1]
 ; AVX-NEXT:    vpunpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
+; AVX-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[4,9,u,u,u,u,u,u,u,u,u,u,u,u,u,u]
 ; AVX-NEXT:    retq
   %sub = sub <2 x i32> <i32 0, i32 undef>, %x
   %bc = bitcast <2 x i32> %sub to <8 x i8>
   %shuffle = shufflevector <8 x i8> %y, <8 x i8> %bc, <2 x i32> <i32 10, i32 4>
   ret <2 x i8> %shuffle
 }
 
 define <2 x i8> @PR39893_2(<2 x float> %x) {

 ; SSE-NEXT:    movapd %xmm2, %xmm0
 ; SSE-NEXT:    retq
 ;
 ; AVX1-LABEL: fmul_v2f64:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vunpcklpd {{.*#+}} xmm2 = xmm1[0],xmm0[0]
 ; AVX1-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
 ; AVX1-NEXT:    vmulpd %xmm0, %xmm0, %xmm0
+; AVX1-NEXT:    vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
 ; AVX1-NEXT:    vmulpd %xmm2, %xmm2, %xmm1
+; AVX1-NEXT:    vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
 ; AVX1-NEXT:    vaddpd %xmm0, %xmm1, %xmm0
-; AVX1-NEXT:    vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: fmul_v2f64:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vunpcklpd {{.*#+}} xmm2 = xmm1[0],xmm0[0]
-; AVX2-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
-; AVX2-NEXT:    vmulpd %xmm0, %xmm0, %xmm0
-; AVX2-NEXT:    vmulpd %xmm2, %xmm2, %xmm1
-; AVX2-NEXT:    vaddpd %xmm0, %xmm1, %xmm0
-; AVX2-NEXT:    vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX2-NEXT:    vunpckhpd {{.*#+}} xmm2 = xmm0[1],xmm1[1]
+; AVX2-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm1[0],xmm0[0]
+; AVX2-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX2-NEXT:    vmulpd %ymm0, %ymm0, %ymm0
+; AVX2-NEXT:    vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX2-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
+; AVX2-NEXT:    vaddpd %ymm0, %ymm1, %ymm0
 ; AVX2-NEXT:    retq
 ;
 ; AVX512-LABEL: fmul_v2f64:
 ; AVX512:       # %bb.0:
 ; AVX512-NEXT:    vunpckhpd {{.*#+}} xmm2 = xmm0[1],xmm1[1]
 ; AVX512-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm1[0],xmm0[0]
-; AVX512-NEXT:    vmulpd %xmm0, %xmm0, %xmm0
-; AVX512-NEXT:    vfmadd231pd {{.*#+}} xmm0 = (xmm2 * xmm2) + xmm0
-; AVX512-NEXT:    vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX512-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX512-NEXT:    vmulpd %ymm0, %ymm0, %ymm0
+; AVX512-NEXT:    vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[3,3,3,3]
+; AVX512-NEXT:    vaddpd %ymm0, %ymm1, %ymm0
 ; AVX512-NEXT:    retq
   %s = shufflevector <2 x double> %x, <2 x double> %y, <4 x i32> <i32 2, i32 0, i32 1, i32 3>
   %bo = fmul fast <4 x double> %s, %s
   %ext = shufflevector <4 x double> %bo, <4 x double> undef, <4 x i32> <i32 2, i32 3, i32 undef, i32 undef>
   %add = fadd fast <4 x double> %bo, %ext
   %rdx = shufflevector <4 x double> %add, <4 x double> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef>
   ret <4 x double> %rdx
 }

llvm/test/CodeGen/X86/vector-shuffle-128-v4.ll

; AVX-NEXT:    retq
   %shuffle = shufflevector <4 x i32> %a, <4 x i32> %b, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
   ret <4 x i32> %shuffle
 }
 
 ; PR22391
 define <4 x i32> @shuffle_v4i32_2456(<4 x i32> %a, <4 x i32> %b) {
 ; SSE2-LABEL: shuffle_v4i32_2456:
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,3],xmm1[0,1]
+; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[0,0]
 ; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[1,2]
 ; SSE2-NEXT:    retq
 ;
 ; SSE3-LABEL: shuffle_v4i32_2456:
 ; SSE3:       # %bb.0:
-; SSE3-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,3],xmm1[0,1]
+; SSE3-NEXT:    shufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[0,0]
 ; SSE3-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[1,2]
 ; SSE3-NEXT:    retq
 ;
 ; SSSE3-LABEL: shuffle_v4i32_2456:
 ; SSSE3:       # %bb.0:
-; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
-; SSSE3-NEXT:    palignr {{.*#+}} xmm1 = xmm0[12,13,14,15],xmm1[0,1,2,3,4,5,6,7,8,9,10,11]
-; SSSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    palignr {{.*#+}} xmm2 = xmm0[8,9,10,11,12,13,14,15],xmm2[0,1,2,3,4,5,6,7]
+; SSSE3-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,2],xmm1[1,2]
+; SSSE3-NEXT:    movaps %xmm2, %xmm0
 ; SSSE3-NEXT:    retq
 ;
 ; SSE41-LABEL: shuffle_v4i32_2456:
 ; SSE41:       # %bb.0:
-; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
-; SSE41-NEXT:    palignr {{.*#+}} xmm1 = xmm0[12,13,14,15],xmm1[0,1,2,3,4,5,6,7,8,9,10,11]
-; SSE41-NEXT:    movdqa %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[2,3,2,3]
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[0,0,1,2]
+; SSE41-NEXT:    pblendw {{.*#+}} xmm0 = xmm2[0,1],xmm0[2,3,4,5,6,7]
 ; SSE41-NEXT:    retq
 ;
 ; AVX1OR2-LABEL: shuffle_v4i32_2456:
 ; AVX1OR2:       # %bb.0:
-; AVX1OR2-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[2,2,2,2]
-; AVX1OR2-NEXT:    vpalignr {{.*#+}} xmm0 = xmm0[12,13,14,15],xmm1[0,1,2,3,4,5,6,7,8,9,10,11]
+; AVX1OR2-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[2,3,2,3]
+; AVX1OR2-NEXT:    vpermilps {{.*#+}} xmm1 = xmm1[0,0,1,2]
+; AVX1OR2-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1,2,3]
 ; AVX1OR2-NEXT:    retq
 ;
 ; AVX512VL-LABEL: shuffle_v4i32_2456:
 ; AVX512VL:       # %bb.0:
 ; AVX512VL-NEXT:    vmovdqa {{.*#+}} xmm2 = [6,0,1,2]
 ; AVX512VL-NEXT:    vpermi2d %xmm0, %xmm1, %xmm2
 ; AVX512VL-NEXT:    vmovdqa %xmm2, %xmm0
 ; AVX512VL-NEXT:    retq

llvm/test/CodeGen/X86/vector-shuffle-combining-avx2.ll

; CHECK-NEXT:    ret{{[l|q]}}
   %2 = bitcast <8 x i32> %1 to <32 x i8>
   %3 = shufflevector <32 x i8> %2, <32 x i8> undef, <32 x i32> <i32 0, i32 1, i32 1, i32 2, i32 3, i32 4, i32 4, i32 5, i32 6, i32 7, i32 7, i32 8, i32 9, i32 10, i32 10, i32 11, i32 16, i32 17, i32 17, i32 18, i32 19, i32 20, i32 20, i32 21, i32 22, i32 23, i32 23, i32 24, i32 25, i32 26, i32 26, i32 27>
   ret <32 x i8> %3
 }
 
 define internal fastcc <8 x float> @PR34577(<8 x float> %inp0, <8 x float> %inp1, <8 x float> %inp2) {
 ; AVX2-LABEL: PR34577:
 ; AVX2:       # %bb.0: # %entry
-; AVX2-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[1,1,1,1]
-; AVX2-NEXT:    vxorps %xmm2, %xmm2, %xmm2
-; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7]
-; AVX2-NEXT:    vmovaps {{.*#+}} ymm2 = <u,u,7,2,u,u,3,2>
-; AVX2-NEXT:    vpermps %ymm1, %ymm2, %ymm1
-; AVX2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
+; AVX2-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = <u,u,7,2,u,u,3,2>
+; AVX2-NEXT:    vpermd %ymm1, %ymm2, %ymm1
+; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = zero,zero,zero,zero,zero,zero,zero,zero,ymm0[u,u,u,u,u,u,u,u,24,25,26,27,28,29,30,31,u,u,u,u,u,u,u,u]
+; AVX2-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]
 ; AVX2-NEXT:    ret{{[l|q]}}
 ;
 ; AVX512-LABEL: PR34577:
 ; AVX512:       # %bb.0: # %entry
 ; AVX512-NEXT:    # kill: def $ymm1 killed $ymm1 def $zmm1
-; AVX512-NEXT:    vpermpd {{.*#+}} ymm0 = ymm0[1,1,1,1]
+; AVX512-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX512-NEXT:    vxorps %xmm2, %xmm2, %xmm2
-; AVX512-NEXT:    vblendps {{.*#+}} ymm2 = ymm0[0,1],ymm2[2,3],ymm0[4,5],ymm2[6,7]
+; AVX512-NEXT:    vunpckhpd {{.*#+}} ymm2 = ymm0[1],ymm2[1],ymm0[3],ymm2[3]
 ; AVX512-NEXT:    vmovaps {{.*#+}} ymm0 = <23,18,7,2,20,u,3,2>
 ; AVX512-NEXT:    vpermi2ps %zmm2, %zmm1, %zmm0
 ; AVX512-NEXT:    # kill: def $ymm0 killed $ymm0 killed $zmm0
 ; AVX512-NEXT:    ret{{[l|q]}}
 entry:
   %shuf0 = shufflevector <8 x float> %inp0, <8 x float> %inp2, <8 x i32> <i32 1, i32 10, i32 11, i32 13, i32 2, i32 13, i32 5, i32 0>
   %sel = select <8 x i1> <i1 false, i1 true, i1 true, i1 false, i1 true, i1 false, i1 true, i1 false>, <8 x float> %shuf0, <8 x float> zeroinitializer
   %shuf1 = shufflevector <8 x float> zeroinitializer, <8 x float> %sel, <8 x i32> <i32 6, i32 11, i32 6, i32 15, i32 12, i32 11, i32 1, i32 3>