[DAGCombiner] Call SimplifyDemandedBits to simplify EXTRACT_VECTOR_ELT

[foad]

llvm/test/CodeGen/Mips/cconv/vector.ll

697–758

There are some regressions in this file but also some improvements. I haven't worked out what's going on yet.

llvm/test/CodeGen/Thumb2/mve-vecreduce-mla.ll

249–259

Regression here and in other cases that are now using muls instead of umull/umlal.

llvm/test/CodeGen/X86/vector-fshl-128.ll

178–179

Regression. Quite a few tests are now using pxor+punpckhdq instead of pshufd. I wonder if some kind of combine could spot this case and turn it back into pshufd.

[foad]

llvm/test/CodeGen/ARM/dagcombine-anyexttozeroext.ll

46–47

Regression.

llvm/test/CodeGen/ARM/vdup.ll

59–69

Regression in lots of cases in this file.

[foad]

Rebase on D88570.

[foad]

Rebase.

[RKSimon]

llvm/test/CodeGen/AMDGPU/sdiv64.ll

502

Is this a regression? It looks like we're lost track that we only need 1 element

llvm/test/CodeGen/ARM/func-argpassing-endian.ll

106 ↗

(On Diff #296629)

regression? we're no longer doing a single multiple load

llvm/test/CodeGen/ARM/vdup.ll

59–69

poor duplicate/splat detection?

[RKSimon]

@foad Any update on this?

[yubing]

llvm/test/CodeGen/X86/vector-fshr-128.ll

179–180

Hi, during combination before legalizeDAG. there are the following Nodes:

  t63: i32 = extract_vector_elt t62, Constant:i32<2>
  t64: i32 = extract_vector_elt t62, Constant:i32<3>
t66: v4i32 = BUILD_VECTOR t63, t64, undef:i32, undef:i32

SimplifyDemandedBits deduce that t64 is a zero, so t66 will transformed into a pxor and a punpckhqd instead of a single pshufd.

Maybe in such a case where a buildvector have elts which are extracted from the same vector, you shouldn't SimplifyDemandedBits for extract_vector_elt.

[foad]

Rebase.

[RKSimon]

llvm/test/CodeGen/X86/buildvec-insertvec.ll

783

this should simplify to "store i32 undef, store i32* undef" and be removed - can you check why it isn't please?

llvm/test/CodeGen/X86/vec_setcc.ll

223

why didn't this simplify?

[foad]

llvm/test/CodeGen/X86/buildvec-insertvec.ll

783

The value being stored is not undef, it's either -2147483648 or poison, depending on the value of %a0.

Anyway -simplifycfg would change the store into a trap + unreachable, but nothing in llc's codegen pipeline does that.

llvm/test/CodeGen/X86/vec_setcc.ll

223

What simplification are you expecting?

[foad]

llvm/test/CodeGen/X86/buildvec-insertvec.ll

783

I've looked into this more carefully now. Hopefully this answer makes more sense.

With my patch %4 = extractelement <4 x i32> zeroinitializer, i32 %2 is simplified into i32 0 based on the known bits of all elements. This happens before we simplify %2, because of the weird way that the DAG combiner runs top-down.

When we visit %2 we simplify it to -2147483648. After that, if we visited the original %4 again, we would simplify it to undef; but with my patch we have already simplified %4 to 0 so it's too late.

I'm not sure what to do about this -- other than change DAGCombine to run bottom-up ;-)

[RKSimon]

llvm/test/CodeGen/X86/buildvec-insertvec.ll

783

Do we need to tweak the out-of-range handling to ISD::EXTRACT_VECTOR_ELT indices? Either just for constant indices or we use computeKnownBits to work out if the minimum value always the exceeds the vector element count?

[RKSimon]

@foad Please can you rebase this?

[foad]

Rebase. There are two failing tests that I have not updated yet:

Failed Tests (2):

LLVM :: CodeGen/AMDGPU/cttz_zero_undef.ll
LLVM :: CodeGen/AMDGPU/scratch-simple.ll

[foad]

Maybe in such a case where a buildvector have elts which are extracted from the same vector, you shouldn't SimplifyDemandedBits for extract_vector_elt.

Yes, I am coming round to this idea: don't simplify EXTRACT_VECTOR_ELT to a constant if it is used by BUILD_VECTOR, because of the risk of breaking things like shuffle patterns.

[RKSimon]

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

2–8

Please can you add a common CHECK prefix:

; RUN: llc < %s -mtriple=x86_64-unknown-unknown | FileCheck %s --check-prefixes=CHECK,SSE
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse4a | FileCheck %s --check-prefixes=CHECK,SSE
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse4.1 | FileCheck %s --check-prefixes=CHECK,SSE
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx | FileCheck %s --check-prefixes=CHECK,AVX
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefixes=CHECK,AVX
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512dq,+avx512vl | FileCheck %s --check-prefixes=CHECK,AVX512
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512bw,+avx512vl | FileCheck %s --check-prefixes=CHECK,AVX512

llvm/test/CodeGen/X86/vec_setcc.ll

223

The 0'th index should be able to extract from the source of the _EXTEND_VECTOR_INREG using SimplifyMultipleUseDemandedBits

[foad]

Add a common CHECK prefix.

[RKSimon]

llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp

18650

How come this isn't picking up many of these test cases? Is it being run too late?

[foad]

llvm/test/CodeGen/X86/vec_setcc.ll

223

Well %eax is extracted directly from the result of the pcmpeqw, which is the source of the sign_extend_vector_inreg.

[RKSimon]

@foad Are you intending to take another look at this at all? I'm wondering if could help with some of the regressions in D127115

[foad]

Rebase.

[foad]

There are some outstanding lit test problems:

CodeGen/AMDGPU/bug-v4f64-subvector.ll: timeout
CodeGen/AMDGPU/scratch-simple.ll: fails
CodeGen/Hexagon/autohvx/hfnosplat_cp.ll: fails
CodeGen/Thumb2/mve-sext-masked-load.ll: fails machine verification for the -early-live-intervals RUN line

[foad]

In D88569#3594683, @RKSimon wrote:

@foad Are you intending to take another look at this at all? I'm wondering if could help with some of the regressions in D127115

TBH I had rather gone off this patch, because it seemed like it would interfere with tests where we want to recombine a bunch of extracts back into some kind of permute instruction. (If you see a bunch of extracts from the same source then you might try to do this, but if one of the extracts has been folded to a constant then it's much harder to spot.) But I have rebased it anyway.

[RKSimon]

@deadalnix This patch has a number of outstanding problems, and might not be worth it - but can you tell if it helps D127115 at all?

[deadalnix]

In D88569#3596712, @RKSimon wrote:

@deadalnix This patch has a number of outstanding problems, and might not be worth it - but can you tell if it helps D127115 at all?

Testing that now, I'll let you know.

[deadalnix]

In D88569#3597508, @deadalnix wrote:

Testing that now, I'll let you know.

So it definitively affects the codegen, but it's not clear if this is better or worse.

[RKSimon]

In D88569#3597648, @deadalnix wrote:

In D88569#3597508, @deadalnix wrote:

Testing that now, I'll let you know.

So it definitively affects the codegen, but it's not clear if this is better or worse.

Yeah that seems about right

@foad its up to you if you want to persevere, we already have SimplifyDemandedBits support for cases where all uses of the vector are extracts - it looks like we now know why it was never extended beyond that...

[foad]

This clearly needs work. The original motivation was to be able to remove SIFoldOperands::tryFoldCndMask which is a MIR optimisation that removes a v_cndmask (select) instruction if the values being selected are the same. We sometimes generate these when a 64-bit select is lowered to a pair of 32-bit selects, and the high or low halves of the 64-bit values were the same. I was hoping to fold this away during selection so we wouldn't have to do it later in SIFoldOperands.

llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp

if (VecVT.isScalableVector())
     return SDValue();
 
   // All the code from this point onwards assumes fixed width vectors, but it's
   // possible that some of the combinations could be made to work for scalable
   // vectors too.
   unsigned NumElts = VecVT.getVectorNumElements();
   unsigned VecEltBitWidth = VecVT.getScalarSizeInBits();
 
+  // Try to simplify the whole operation to a constant, or simplify its
+  // operands.
+  if (SimplifyDemandedBits(SDValue(N, 0)))
+    return SDValue(N, 0);
+
   // TODO: These transforms should not require the 'hasOneUse' restriction, but
   // there are regressions on multiple targets without it. We can end up with a
   // mess of scalar and vector code if we reduce only part of the DAG to scalar.
   if (IndexC && VecOp.getOpcode() == ISD::BITCAST && VecVT.isInteger() &&
       VecOp.hasOneUse()) {
     // The vector index of the LSBs of the source depend on the endian-ness.
     bool IsLE = DAG.getDataLayout().isLittleEndian();
     unsigned ExtractIndex = IndexC->getZExtValue();
if (llvm::all_of(VecOp->uses(), [&](SDNode *Use) {
     if (SimplifyDemandedVectorElts(VecOp, DemandedElts, true)) {
       // We simplified the vector operand of this extract element. If this
       // extract is not dead, visit it again so it is folded properly.
       if (N->getOpcode() != ISD::DELETED_NODE)
         AddToWorklist(N);
       return SDValue(N, 0);
     }
     APInt DemandedBits = APInt::getAllOnesValue(VecEltBitWidth);
     if (SimplifyDemandedBits(VecOp, DemandedBits, DemandedElts, true)) {
       // We simplified the vector operand of this extract element. If this
       // extract is not dead, visit it again so it is folded properly.
       if (N->getOpcode() != ISD::DELETED_NODE)
         AddToWorklist(N);
       return SDValue(N, 0);
     }
   }
 

llvm/test/CodeGen/AArch64/arm64-build-vector.ll

 ; The lowering of a widened f16 BUILD_VECTOR tries to optimize it by building
 ; an equivalent integer vector and BITCAST-ing that. This case checks that
 ; normalizing the vector generates a valid result. The choice of the
 ; constant prevents earlier passes from replacing the BUILD_VECTOR.
 define void @widen_f16_build_vector(half* %addr) {
 ; CHECK-LABEL: widen_f16_build_vector:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov w8, #13294
-; CHECK-NEXT:    dup.4h v0, w8
-; CHECK-NEXT:    str s0, [x0]
+; CHECK-NEXT:    movk w8, #13294, lsl #16
+; CHECK-NEXT:    str w8, [x0]
 ; CHECK-NEXT:    ret
   %1 = bitcast half* %addr to <2 x half>*
   store <2 x half> <half 0xH33EE, half 0xH33EE>, <2 x half>* %1, align 2
   ret void
 }
 
 ; Check that a single element vector is constructed with a mov
 define <1 x i64> @single_element_vector_i64(<1 x i64> %arg) {

llvm/test/CodeGen/AArch64/arm64-nvcast.ll

entry:
   %v2 = extractelement <3 x float> <float 0.000000e+00, float 2.000000e+00, float 0.000000e+00>, i32 %v1
   store float %v2, float* %p1, align 4
   ret void
 }
 
 define void @test2(float * %p1, i32 %v1) {
 ; CHECK-LABEL: test2:
 ; CHECK:       ; %bb.0: ; %entry
-; CHECK-NEXT:    sub sp, sp, #16 ; =16
-; CHECK-NEXT:    .cfi_def_cfa_offset 16
-; CHECK-NEXT:    ; kill: def $w1 killed $w1 def $x1
-; CHECK-NEXT:    movi.16b v0, #63
-; CHECK-NEXT:    and x8, x1, #0x3
-; CHECK-NEXT:    mov x9, sp
-; CHECK-NEXT:    str q0, [sp]
-; CHECK-NEXT:    bfi x9, x8, #2, #2
-; CHECK-NEXT:    ldr s0, [x9]
-; CHECK-NEXT:    str s0, [x0]
-; CHECK-NEXT:    add sp, sp, #16 ; =16
+; CHECK-NEXT:    mov w8, #1061109567
+; CHECK-NEXT:    str w8, [x0]
 ; CHECK-NEXT:    ret
 entry:
   %v2 = extractelement <3 x float> <float 0.7470588088035583, float 0.7470588088035583, float 0.7470588088035583>, i32 %v1
   store float %v2, float* %p1, align 4
   ret void
 }
 
 

llvm/test/CodeGen/AArch64/arm64-promote-const-complex-initializers.ll

 ; CHECK-NEXT:     ret
 
   ret [1 x <4 x float>] [<4 x float> bitcast (<1 x i128> <i128 84405977732342157929391748327801880576> to <4 x float>)]
 }
 
 define [1 x <4 x float>] @test2() {
 ; CHECK-LABEL:    .p2align    4               ; -- Begin function test2
 ; CHECK-NEXT: lCPI1_0:
-; CHECK-NEXT:     .long   0x00000000              ; float 0
-; CHECK-NEXT:     .long   0x00000000              ; float 0
-; CHECK-NEXT:     .long   0x00000000              ; float 0
-; CHECK-NEXT:     .long   0x3f800000              ; float 1
+; CHECK-NEXT:     .long   0x80000000              ; float -0
+; CHECK-NEXT:     .long   0x80000000              ; float -0
+; CHECK-NEXT:     .long   0x80000000              ; float -0
+; CHECK-NEXT:     .long   0xbf800000              ; float -1
 ; CHECK-NEXT:     .section    __TEXT,__text,regular,pure_instructions
 ; CHECK-NEXT:     .globl  _test2
 ; CHECK-NEXT:     .p2align    2
 ; CHECK-NEXT: _test2:                                 ; @test2
 ; CHECK-NEXT:     .cfi_startproc
 ; CHECK-NEXT: ; %bb.0:
 ; CHECK-NEXT: Lloh2:
 ; CHECK-NEXT:     adrp    x8, lCPI1_0@PAGE
 ; CHECK-NEXT: Lloh3:
-; CHECK-NEXT:     ldr q1, [x8, lCPI1_0@PAGEOFF]
-; CHECK-NEXT:     mov s2, v1[1]
-; CHECK-NEXT:     fneg    s0, s1
-; CHECK-NEXT:     mov s3, v1[2]
-; CHECK-NEXT:     fneg    s2, s2
-; CHECK-NEXT:     mov s1, v1[3]
-; CHECK-NEXT:     fneg    s3, s3
-; CHECK-NEXT:     mov.s   v0[1], v2[0]
-; CHECK-NEXT:     mov.s   v0[2], v3[0]
-; CHECK-NEXT:     fneg    s1, s1
-; CHECK-NEXT:     mov.s   v0[3], v1[0]
+; CHECK-NEXT:     ldr q0, [x8, lCPI1_0@PAGEOFF]
 ; CHECK-NEXT:     ret
 ;
   ret [1 x <4 x float>] [<4 x float>
     <float fneg (float extractelement (<4 x float> bitcast (<1 x i128> <i128 84405977732342157929391748327801880576> to <4 x float>), i32 0)),
      float fneg (float extractelement (<4 x float> bitcast (<1 x i128> <i128 84405977732342157929391748327801880576> to <4 x float>), i32 1)),
      float fneg (float extractelement (<4 x float> bitcast (<1 x i128> <i128 84405977732342157929391748327801880576> to <4 x float>), i32 2)),
      float fneg (float extractelement (<4 x float> bitcast (<1 x i128> <i128 84405977732342157929391748327801880576> to <4 x float>), i32 3))>]
 }

llvm/test/CodeGen/AArch64/cmp-select-sign.ll

; CHECK-NEXT:    ret
   %res = select <4 x i1> %c, <4 x i32> <i32 1, i32 1, i32 1, i32 1>, <4 x i32> <i32 -1, i32 -1, i32 -1, i32 1>
   ret <4 x i32> %res
 }
 
 ; i65 is not legal.
 define <4 x i65> @sign_4xi65(<4 x i65> %a) {
 ; CHECK-LABEL: sign_4xi65:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    sbfx x8, x1, #0, #1
-; CHECK-NEXT:    sbfx x9, x7, #0, #1
-; CHECK-NEXT:    orr x6, x9, #0x1
-; CHECK-NEXT:    lsr x7, x9, #63
-; CHECK-NEXT:    orr x9, x8, #0x1
-; CHECK-NEXT:    lsr x1, x8, #63
-; CHECK-NEXT:    fmov d0, x9
-; CHECK-NEXT:    sbfx x10, x5, #0, #1
-; CHECK-NEXT:    sbfx x11, x3, #0, #1
-; CHECK-NEXT:    mov v0.d[1], x1
-; CHECK-NEXT:    orr x2, x11, #0x1
-; CHECK-NEXT:    lsr x3, x11, #63
-; CHECK-NEXT:    orr x4, x10, #0x1
-; CHECK-NEXT:    lsr x5, x10, #63
-; CHECK-NEXT:    fmov x0, d0
+; CHECK-NEXT:    sbfx x8, x7, #0, #1
+; CHECK-NEXT:    sbfx x9, x5, #0, #1
+; CHECK-NEXT:    sbfx x10, x3, #0, #1
+; CHECK-NEXT:    sbfx x11, x1, #0, #1
+; CHECK-NEXT:    orr x0, x11, #0x1
+; CHECK-NEXT:    lsr x1, x11, #63
+; CHECK-NEXT:    orr x2, x10, #0x1
+; CHECK-NEXT:    lsr x3, x10, #63
+; CHECK-NEXT:    orr x4, x9, #0x1
+; CHECK-NEXT:    lsr x5, x9, #63
+; CHECK-NEXT:    orr x6, x8, #0x1
+; CHECK-NEXT:    lsr x7, x8, #63
 ; CHECK-NEXT:    ret
   %c = icmp sgt <4 x i65> %a, <i65 -1, i65 -1, i65 -1, i65 -1>
   %res = select <4 x i1> %c, <4 x i65> <i65 1, i65 1, i65 1, i65 1>, <4 x i65 > <i65 -1, i65 -1, i65 -1, i65 -1>
   ret <4 x i65> %res
 }
 
 declare void @use_4xi1(<4 x i1>)

llvm/test/CodeGen/AArch64/dag-numsignbits.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s -mtriple=aarch64-unknown | FileCheck %s
 
 ; PR32273
 
 define void @signbits_vXi1(<4 x i16> %a1) {
 ; CHECK-LABEL: signbits_vXi1:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    adrp x8, .LCPI0_0
 ; CHECK-NEXT:    ldr d1, [x8, :lo12:.LCPI0_0]
-; CHECK-NEXT:    adrp x8, .LCPI0_1
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $q0
-; CHECK-NEXT:    ldr d2, [x8, :lo12:.LCPI0_1]
+; CHECK-NEXT:    movi v2.4h, #1
 ; CHECK-NEXT:    dup v0.4h, v0.h[0]
+; CHECK-NEXT:    mov w1, wzr
 ; CHECK-NEXT:    add v0.4h, v0.4h, v1.4h
-; CHECK-NEXT:    movi v1.4h, #1
-; CHECK-NEXT:    cmgt v0.4h, v1.4h, v0.4h
-; CHECK-NEXT:    and v0.8b, v0.8b, v2.8b
-; CHECK-NEXT:    shl v0.4h, v0.4h, #15
-; CHECK-NEXT:    sshr v0.4h, v0.4h, #15
+; CHECK-NEXT:    cmgt v0.4h, v2.4h, v0.4h
 ; CHECK-NEXT:    umov w0, v0.h[0]
 ; CHECK-NEXT:    umov w3, v0.h[3]
-; CHECK-NEXT:    mov w1, wzr
 ; CHECK-NEXT:    mov w2, wzr
 ; CHECK-NEXT:    b foo
   %tmp3 = shufflevector <4 x i16> %a1, <4 x i16> undef, <4 x i32> zeroinitializer
   %tmp5 = add <4 x i16> %tmp3, <i16 18249, i16 6701, i16 -18744, i16 -25086>
   %tmp6 = icmp slt <4 x i16> %tmp5, <i16 1, i16 1, i16 1, i16 1>
   %tmp7 = and <4 x i1> %tmp6, <i1 true, i1 false, i1 false, i1 true>
   %tmp8 = sext <4 x i1> %tmp7 to <4 x i16>
   %tmp9 = extractelement <4 x i16> %tmp8, i32 0

llvm/test/CodeGen/AArch64/fptosi-sat-vector.ll

 }
 
 define <6 x i32> @test_signed_v6f32_v6i32(<6 x float> %f) {
 ; CHECK-LABEL: test_signed_v6f32_v6i32:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov w9, #-822083584
 ; CHECK-NEXT:    mov w11, #1325400063
 ; CHECK-NEXT:    fmov s6, w9
-; CHECK-NEXT:    fcvtzs w8, s5
+; CHECK-NEXT:    fcvtzs w8, s0
 ; CHECK-NEXT:    mov w10, #-2147483648
-; CHECK-NEXT:    fcmp s5, s6
 ; CHECK-NEXT:    fmov s7, w11
+; CHECK-NEXT:    fcmp s0, s6
 ; CHECK-NEXT:    mov w12, #2147483647
 ; CHECK-NEXT:    csel w8, w10, w8, lt
-; CHECK-NEXT:    fcmp s5, s7
-; CHECK-NEXT:    csel w8, w12, w8, gt
-; CHECK-NEXT:    fcmp s5, s5
-; CHECK-NEXT:    fcvtzs w13, s4
-; CHECK-NEXT:    csel w5, wzr, w8, vs
-; CHECK-NEXT:    fcmp s4, s6
-; CHECK-NEXT:    csel w8, w10, w13, lt
-; CHECK-NEXT:    fcmp s4, s7
-; CHECK-NEXT:    csel w8, w12, w8, gt
-; CHECK-NEXT:    fcmp s4, s4
-; CHECK-NEXT:    fcvtzs w14, s0
-; CHECK-NEXT:    csel w8, wzr, w8, vs
-; CHECK-NEXT:    fcmp s0, s6
-; CHECK-NEXT:    csel w13, w10, w14, lt
 ; CHECK-NEXT:    fcmp s0, s7
-; CHECK-NEXT:    csel w13, w12, w13, gt
+; CHECK-NEXT:    csel w8, w12, w8, gt
 ; CHECK-NEXT:    fcmp s0, s0
-; CHECK-NEXT:    fcvtzs w9, s1
-; CHECK-NEXT:    csel w0, wzr, w13, vs
+; CHECK-NEXT:    fcvtzs w13, s1
+; CHECK-NEXT:    csel w0, wzr, w8, vs
 ; CHECK-NEXT:    fcmp s1, s6
-; CHECK-NEXT:    csel w9, w10, w9, lt
+; CHECK-NEXT:    csel w8, w10, w13, lt
 ; CHECK-NEXT:    fcmp s1, s7
-; CHECK-NEXT:    csel w9, w12, w9, gt
+; CHECK-NEXT:    csel w8, w12, w8, gt
 ; CHECK-NEXT:    fcmp s1, s1
-; CHECK-NEXT:    fcvtzs w11, s2
-; CHECK-NEXT:    csel w1, wzr, w9, vs
+; CHECK-NEXT:    fcvtzs w14, s2
+; CHECK-NEXT:    csel w1, wzr, w8, vs
 ; CHECK-NEXT:    fcmp s2, s6
-; CHECK-NEXT:    csel w9, w10, w11, lt
+; CHECK-NEXT:    csel w8, w10, w14, lt
 ; CHECK-NEXT:    fcmp s2, s7
-; CHECK-NEXT:    csel w9, w12, w9, gt
+; CHECK-NEXT:    csel w8, w12, w8, gt
 ; CHECK-NEXT:    fcmp s2, s2
-; CHECK-NEXT:    fmov s4, w8
-; CHECK-NEXT:    fcvtzs w8, s3
-; CHECK-NEXT:    csel w2, wzr, w9, vs
+; CHECK-NEXT:    fcvtzs w15, s3
+; CHECK-NEXT:    csel w2, wzr, w8, vs
 ; CHECK-NEXT:    fcmp s3, s6
-; CHECK-NEXT:    csel w8, w10, w8, lt
+; CHECK-NEXT:    csel w8, w10, w15, lt
 ; CHECK-NEXT:    fcmp s3, s7
-; CHECK-NEXT:    mov v4.s[1], w5
 ; CHECK-NEXT:    csel w8, w12, w8, gt
 ; CHECK-NEXT:    fcmp s3, s3
+; CHECK-NEXT:    fcvtzs w9, s4
 ; CHECK-NEXT:    csel w3, wzr, w8, vs
-; CHECK-NEXT:    fmov w4, s4
+; CHECK-NEXT:    fcmp s4, s6
+; CHECK-NEXT:    csel w8, w10, w9, lt
+; CHECK-NEXT:    fcmp s4, s7
+; CHECK-NEXT:    csel w8, w12, w8, gt
+; CHECK-NEXT:    fcmp s4, s4
+; CHECK-NEXT:    fcvtzs w11, s5
+; CHECK-NEXT:    csel w4, wzr, w8, vs
+; CHECK-NEXT:    fcmp s5, s6
+; CHECK-NEXT:    csel w8, w10, w11, lt
+; CHECK-NEXT:    fcmp s5, s7
+; CHECK-NEXT:    csel w8, w12, w8, gt
+; CHECK-NEXT:    fcmp s5, s5
+; CHECK-NEXT:    csel w5, wzr, w8, vs
 ; CHECK-NEXT:    ret
     %x = call <6 x i32> @llvm.fptosi.sat.v6f32.v6i32(<6 x float> %f)
     ret <6 x i32> %x
 }
 
 define <7 x i32> @test_signed_v7f32_v7i32(<7 x float> %f) {
 ; CHECK-LABEL: test_signed_v7f32_v7i32:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov w9, #-822083584
 ; CHECK-NEXT:    mov w11, #1325400063
 ; CHECK-NEXT:    fmov s7, w9
-; CHECK-NEXT:    fcvtzs w8, s5
+; CHECK-NEXT:    fcvtzs w8, s0
 ; CHECK-NEXT:    mov w10, #-2147483648
-; CHECK-NEXT:    fcmp s5, s7
 ; CHECK-NEXT:    fmov s16, w11
+; CHECK-NEXT:    fcmp s0, s7
 ; CHECK-NEXT:    mov w12, #2147483647
 ; CHECK-NEXT:    csel w8, w10, w8, lt
-; CHECK-NEXT:    fcmp s5, s16
-; CHECK-NEXT:    csel w8, w12, w8, gt
-; CHECK-NEXT:    fcmp s5, s5
-; CHECK-NEXT:    fcvtzs w13, s4
-; CHECK-NEXT:    csel w8, wzr, w8, vs
-; CHECK-NEXT:    fcmp s4, s7
-; CHECK-NEXT:    csel w11, w10, w13, lt
-; CHECK-NEXT:    fcmp s4, s16
-; CHECK-NEXT:    csel w11, w12, w11, gt
-; CHECK-NEXT:    fcmp s4, s4
-; CHECK-NEXT:    fcvtzs w14, s6
-; CHECK-NEXT:    csel w11, wzr, w11, vs
-; CHECK-NEXT:    fcmp s6, s7
-; CHECK-NEXT:    csel w14, w10, w14, lt
-; CHECK-NEXT:    fcmp s6, s16
-; CHECK-NEXT:    csel w14, w12, w14, gt
-; CHECK-NEXT:    fcmp s6, s6
-; CHECK-NEXT:    fcvtzs w9, s0
-; CHECK-NEXT:    csel w6, wzr, w14, vs
-; CHECK-NEXT:    fcmp s0, s7
-; CHECK-NEXT:    csel w9, w10, w9, lt
 ; CHECK-NEXT:    fcmp s0, s16
-; CHECK-NEXT:    csel w9, w12, w9, gt
+; CHECK-NEXT:    csel w8, w12, w8, gt
 ; CHECK-NEXT:    fcmp s0, s0
 ; CHECK-NEXT:    fcvtzs w13, s1
-; CHECK-NEXT:    csel w0, wzr, w9, vs
+; CHECK-NEXT:    csel w0, wzr, w8, vs
 ; CHECK-NEXT:    fcmp s1, s7
-; CHECK-NEXT:    csel w9, w10, w13, lt
+; CHECK-NEXT:    csel w8, w10, w13, lt
 ; CHECK-NEXT:    fcmp s1, s16
-; CHECK-NEXT:    csel w9, w12, w9, gt
+; CHECK-NEXT:    csel w8, w12, w8, gt
 ; CHECK-NEXT:    fcmp s1, s1
-; CHECK-NEXT:    fmov s4, w11
-; CHECK-NEXT:    fcvtzs w11, s2
-; CHECK-NEXT:    csel w1, wzr, w9, vs
+; CHECK-NEXT:    fcvtzs w14, s2
+; CHECK-NEXT:    csel w1, wzr, w8, vs
 ; CHECK-NEXT:    fcmp s2, s7
-; CHECK-NEXT:    csel w9, w10, w11, lt
+; CHECK-NEXT:    csel w8, w10, w14, lt
 ; CHECK-NEXT:    fcmp s2, s16
-; CHECK-NEXT:    csel w9, w12, w9, gt
+; CHECK-NEXT:    csel w8, w12, w8, gt
 ; CHECK-NEXT:    fcmp s2, s2
-; CHECK-NEXT:    mov v4.s[1], w8
-; CHECK-NEXT:    fcvtzs w8, s3
-; CHECK-NEXT:    csel w2, wzr, w9, vs
+; CHECK-NEXT:    fcvtzs w15, s3
+; CHECK-NEXT:    csel w2, wzr, w8, vs
 ; CHECK-NEXT:    fcmp s3, s7
-; CHECK-NEXT:    csel w8, w10, w8, lt
+; CHECK-NEXT:    csel w8, w10, w15, lt
 ; CHECK-NEXT:    fcmp s3, s16
-; CHECK-NEXT:    mov v4.s[2], w6
 ; CHECK-NEXT:    csel w8, w12, w8, gt
 ; CHECK-NEXT:    fcmp s3, s3
+; CHECK-NEXT:    fcvtzs w16, s4
 ; CHECK-NEXT:    csel w3, wzr, w8, vs
-; CHECK-NEXT:    mov w5, v4.s[1]
-; CHECK-NEXT:    fmov w4, s4
+; CHECK-NEXT:    fcmp s4, s7
+; CHECK-NEXT:    csel w8, w10, w16, lt
+; CHECK-NEXT:    fcmp s4, s16
+; CHECK-NEXT:    csel w8, w12, w8, gt
+; CHECK-NEXT:    fcmp s4, s4
+; CHECK-NEXT:    fcvtzs w9, s5
+; CHECK-NEXT:    csel w4, wzr, w8, vs
+; CHECK-NEXT:    fcmp s5, s7
+; CHECK-NEXT:    csel w8, w10, w9, lt
+; CHECK-NEXT:    fcmp s5, s16
+; CHECK-NEXT:    csel w8, w12, w8, gt
+; CHECK-NEXT:    fcmp s5, s5
+; CHECK-NEXT:    fcvtzs w11, s6
+; CHECK-NEXT:    csel w5, wzr, w8, vs
+; CHECK-NEXT:    fcmp s6, s7
+; CHECK-NEXT:    csel w8, w10, w11, lt
+; CHECK-NEXT:    fcmp s6, s16
+; CHECK-NEXT:    csel w8, w12, w8, gt
+; CHECK-NEXT:    fcmp s6, s6
+; CHECK-NEXT:    csel w6, wzr, w8, vs
 ; CHECK-NEXT:    ret
     %x = call <7 x i32> @llvm.fptosi.sat.v7f32.v7i32(<7 x float> %f)
     ret <7 x i32> %x
 }
 
 define <8 x i32> @test_signed_v8f32_v8i32(<8 x float> %f) {
 ; CHECK-LABEL: test_signed_v8f32_v8i32:
 ; CHECK:       // %bb.0:

 ; CHECK-NEXT:    ret
     %x = call <5 x i32> @llvm.fptosi.sat.v5f16.v5i32(<5 x half> %f)
     ret <5 x i32> %x
 }
 
 define <6 x i32> @test_signed_v6f16_v6i32(<6 x half> %f) {
 ; CHECK-LABEL: test_signed_v6f16_v6i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ext v1.16b, v0.16b, v0.16b, #8
 ; CHECK-NEXT:    mov w8, #-822083584
-; CHECK-NEXT:    mov h2, v1.h[1]
+; CHECK-NEXT:    fcvt s1, h0
 ; CHECK-NEXT:    mov w10, #1325400063
-; CHECK-NEXT:    fmov s3, w8
-; CHECK-NEXT:    fcvt s2, h2
+; CHECK-NEXT:    fmov s2, w8
 ; CHECK-NEXT:    mov w9, #-2147483648
-; CHECK-NEXT:    fmov s4, w10
-; CHECK-NEXT:    fcvtzs w8, s2
-; CHECK-NEXT:    fcmp s2, s3
+; CHECK-NEXT:    fcvtzs w12, s1
+; CHECK-NEXT:    fcmp s1, s2
+; CHECK-NEXT:    fmov s3, w10
 ; CHECK-NEXT:    mov w11, #2147483647
-; CHECK-NEXT:    csel w8, w9, w8, lt
-; CHECK-NEXT:    fcmp s2, s4
+; CHECK-NEXT:    csel w8, w9, w12, lt
+; CHECK-NEXT:    fcmp s1, s3
 ; CHECK-NEXT:    csel w8, w11, w8, gt
-; CHECK-NEXT:    fcmp s2, s2
+; CHECK-NEXT:    fcmp s1, s1
+; CHECK-NEXT:    mov h1, v0.h[1]
 ; CHECK-NEXT:    fcvt s1, h1
 ; CHECK-NEXT:    fcvtzs w10, s1
-; CHECK-NEXT:    csel w5, wzr, w8, vs
+; CHECK-NEXT:    csel w0, wzr, w8, vs
+; CHECK-NEXT:    fcmp s1, s2
+; CHECK-NEXT:    csel w8, w9, w10, lt
 ; CHECK-NEXT:    fcmp s1, s3
+; CHECK-NEXT:    csel w8, w11, w8, gt
+; CHECK-NEXT:    fcmp s1, s1
+; CHECK-NEXT:    mov h1, v0.h[2]
+; CHECK-NEXT:    fcvt s1, h1
+; CHECK-NEXT:    fcvtzs w10, s1
+; CHECK-NEXT:    csel w1, wzr, w8, vs
+; CHECK-NEXT:    fcmp s1, s2
 ; CHECK-NEXT:    csel w8, w9, w10, lt
-; CHECK-NEXT:    fcmp s1, s4
-; CHECK-NEXT:    fcvt s2, h0
+; CHECK-NEXT:    fcmp s1, s3
 ; CHECK-NEXT:    csel w8, w11, w8, gt
 ; CHECK-NEXT:    fcmp s1, s1
-; CHECK-NEXT:    fcvtzs w10, s2
-; CHECK-NEXT:    csel w8, wzr, w8, vs
-; CHECK-NEXT:    fcmp s2, s3
-; CHECK-NEXT:    mov h1, v0.h[1]
-; CHECK-NEXT:    csel w10, w9, w10, lt
-; CHECK-NEXT:    fcmp s2, s4
-; CHECK-NEXT:    csel w10, w11, w10, gt
-; CHECK-NEXT:    fcmp s2, s2
+; CHECK-NEXT:    mov h1, v0.h[3]
 ; CHECK-NEXT:    fcvt s1, h1
-; CHECK-NEXT:    fcvtzs w12, s1
-; CHECK-NEXT:    csel w0, wzr, w10, vs
+; CHECK-NEXT:    fcvtzs w10, s1
+; CHECK-NEXT:    csel w2, wzr, w8, vs
+; CHECK-NEXT:    fcmp s1, s2
+; CHECK-NEXT:    ext v0.16b, v0.16b, v0.16b, #8
+; CHECK-NEXT:    csel w8, w9, w10, lt
 ; CHECK-NEXT:    fcmp s1, s3
-; CHECK-NEXT:    mov h2, v0.h[2]
-; CHECK-NEXT:    csel w10, w9, w12, lt
-; CHECK-NEXT:    fcmp s1, s4
-; CHECK-NEXT:    fcvt s2, h2
-; CHECK-NEXT:    csel w10, w11, w10, gt
+; CHECK-NEXT:    csel w8, w11, w8, gt
 ; CHECK-NEXT:    fcmp s1, s1
-; CHECK-NEXT:    fcvtzs w13, s2
-; CHECK-NEXT:    csel w1, wzr, w10, vs
-; CHECK-NEXT:    fcmp s2, s3
-; CHECK-NEXT:    mov h0, v0.h[3]
-; CHECK-NEXT:    csel w10, w9, w13, lt
-; CHECK-NEXT:    fcmp s2, s4
+; CHECK-NEXT:    fcvt s1, h0
+; CHECK-NEXT:    fcvtzs w10, s1
+; CHECK-NEXT:    csel w3, wzr, w8, vs
+; CHECK-NEXT:    fcmp s1, s2
+; CHECK-NEXT:    mov h0, v0.h[1]
+; CHECK-NEXT:    csel w8, w9, w10, lt
+; CHECK-NEXT:    fcmp s1, s3
 ; CHECK-NEXT:    fcvt s0, h0
-; CHECK-NEXT:    csel w10, w11, w10, gt
-; CHECK-NEXT:    fcmp s2, s2
-; CHECK-NEXT:    fmov s1, w8
-; CHECK-NEXT:    fcvtzs w8, s0
-; CHECK-NEXT:    csel w2, wzr, w10, vs
+; CHECK-NEXT:    csel w8, w11, w8, gt
+; CHECK-NEXT:    fcmp s1, s1
+; CHECK-NEXT:    fcvtzs w12, s0
+; CHECK-NEXT:    csel w4, wzr, w8, vs
+; CHECK-NEXT:    fcmp s0, s2
+; CHECK-NEXT:    csel w8, w9, w12, lt
 ; CHECK-NEXT:    fcmp s0, s3
-; CHECK-NEXT:    csel w8, w9, w8, lt
-; CHECK-NEXT:    fcmp s0, s4
-; CHECK-NEXT:    mov v1.s[1], w5
 ; CHECK-NEXT:    csel w8, w11, w8, gt
 ; CHECK-NEXT:    fcmp s0, s0
-; CHECK-NEXT:    csel w3, wzr, w8, vs
-; CHECK-NEXT:    fmov w4, s1
+; CHECK-NEXT:    csel w5, wzr, w8, vs
 ; CHECK-NEXT:    ret
     %x = call <6 x i32> @llvm.fptosi.sat.v6f16.v6i32(<6 x half> %f)
     ret <6 x i32> %x
 }
 
 define <7 x i32> @test_signed_v7f16_v7i32(<7 x half> %f) {
 ; CHECK-LABEL: test_signed_v7f16_v7i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ext v3.16b, v0.16b, v0.16b, #8
-; CHECK-NEXT:    mov w10, #-822083584
-; CHECK-NEXT:    mov h4, v3.h[1]
-; CHECK-NEXT:    mov w11, #1325400063
-; CHECK-NEXT:    fmov s2, w10
-; CHECK-NEXT:    fcvt s4, h4
-; CHECK-NEXT:    mov w8, #-2147483648
-; CHECK-NEXT:    fmov s1, w11
-; CHECK-NEXT:    fcvtzs w10, s4
-; CHECK-NEXT:    fcmp s4, s2
-; CHECK-NEXT:    mov w9, #2147483647
-; CHECK-NEXT:    csel w10, w8, w10, lt
-; CHECK-NEXT:    fcmp s4, s1
-; CHECK-NEXT:    csel w10, w9, w10, gt
-; CHECK-NEXT:    fcmp s4, s4
-; CHECK-NEXT:    fcvt s4, h3
-; CHECK-NEXT:    fcvtzs w11, s4
-; CHECK-NEXT:    csel w10, wzr, w10, vs
-; CHECK-NEXT:    fcmp s4, s2
-; CHECK-NEXT:    csel w11, w8, w11, lt
-; CHECK-NEXT:    fcmp s4, s1
-; CHECK-NEXT:    mov h3, v3.h[2]
-; CHECK-NEXT:    csel w11, w9, w11, gt
-; CHECK-NEXT:    fcmp s4, s4
-; CHECK-NEXT:    fcvt s3, h3
-; CHECK-NEXT:    fcvtzs w12, s3
-; CHECK-NEXT:    csel w11, wzr, w11, vs
-; CHECK-NEXT:    fcmp s3, s2
-; CHECK-NEXT:    csel w12, w8, w12, lt
-; CHECK-NEXT:    fcmp s3, s1
-; CHECK-NEXT:    fcvt s4, h0
-; CHECK-NEXT:    csel w12, w9, w12, gt
-; CHECK-NEXT:    fcmp s3, s3
-; CHECK-NEXT:    fcvtzs w13, s4
-; CHECK-NEXT:    csel w6, wzr, w12, vs
-; CHECK-NEXT:    fcmp s4, s2
-; CHECK-NEXT:    mov h3, v0.h[1]
-; CHECK-NEXT:    csel w12, w8, w13, lt
-; CHECK-NEXT:    fcmp s4, s1
-; CHECK-NEXT:    csel w12, w9, w12, gt
-; CHECK-NEXT:    fcmp s4, s4
-; CHECK-NEXT:    fcvt s3, h3
-; CHECK-NEXT:    fcvtzs w13, s3
-; CHECK-NEXT:    csel w0, wzr, w12, vs
-; CHECK-NEXT:    fcmp s3, s2
-; CHECK-NEXT:    mov h4, v0.h[2]
-; CHECK-NEXT:    csel w12, w8, w13, lt
-; CHECK-NEXT:    fcmp s3, s1
-; CHECK-NEXT:    fcvt s4, h4
-; CHECK-NEXT:    csel w12, w9, w12, gt
-; CHECK-NEXT:    fcmp s3, s3
-; CHECK-NEXT:    fmov s3, w11
-; CHECK-NEXT:    fcvtzs w11, s4
-; CHECK-NEXT:    csel w1, wzr, w12, vs
-; CHECK-NEXT:    fcmp s4, s2
-; CHECK-NEXT:    mov h0, v0.h[3]
-; CHECK-NEXT:    csel w11, w8, w11, lt
-; CHECK-NEXT:    fcmp s4, s1
+; CHECK-NEXT:    mov w8, #-822083584
+; CHECK-NEXT:    fcvt s1, h0
+; CHECK-NEXT:    mov w10, #1325400063
+; CHECK-NEXT:    fmov s2, w8
+; CHECK-NEXT:    mov w9, #-2147483648
+; CHECK-NEXT:    fcvtzs w12, s1
+; CHECK-NEXT:    fcmp s1, s2
+; CHECK-NEXT:    fmov s3, w10
+; CHECK-NEXT:    mov w11, #2147483647
+; CHECK-NEXT:    csel w8, w9, w12, lt
+; CHECK-NEXT:    fcmp s1, s3
+; CHECK-NEXT:    csel w8, w11, w8, gt
+; CHECK-NEXT:    fcmp s1, s1
+; CHECK-NEXT:    mov h1, v0.h[1]
+; CHECK-NEXT:    fcvt s1, h1
+; CHECK-NEXT:    fcvtzs w10, s1
+; CHECK-NEXT:    csel w0, wzr, w8, vs
+; CHECK-NEXT:    fcmp s1, s2
+; CHECK-NEXT:    csel w8, w9, w10, lt
+; CHECK-NEXT:    fcmp s1, s3
+; CHECK-NEXT:    csel w8, w11, w8, gt
+; CHECK-NEXT:    fcmp s1, s1
+; CHECK-NEXT:    mov h1, v0.h[2]
+; CHECK-NEXT:    fcvt s1, h1
+; CHECK-NEXT:    fcvtzs w10, s1
+; CHECK-NEXT:    csel w1, wzr, w8, vs
+; CHECK-NEXT:    fcmp s1, s2
+; CHECK-NEXT:    csel w8, w9, w10, lt
+; CHECK-NEXT:    fcmp s1, s3
+; CHECK-NEXT:    csel w8, w11, w8, gt
+; CHECK-NEXT:    fcmp s1, s1
+; CHECK-NEXT:    mov h1, v0.h[3]
+; CHECK-NEXT:    fcvt s1, h1
+; CHECK-NEXT:    fcvtzs w10, s1
+; CHECK-NEXT:    csel w2, wzr, w8, vs
+; CHECK-NEXT:    fcmp s1, s2
+; CHECK-NEXT:    ext v0.16b, v0.16b, v0.16b, #8
+; CHECK-NEXT:    csel w8, w9, w10, lt
+; CHECK-NEXT:    fcmp s1, s3
+; CHECK-NEXT:    csel w8, w11, w8, gt
+; CHECK-NEXT:    fcmp s1, s1
+; CHECK-NEXT:    fcvt s1, h0
+; CHECK-NEXT:    fcvtzs w10, s1
+; CHECK-NEXT:    csel w3, wzr, w8, vs
+; CHECK-NEXT:    fcmp s1, s2
+; CHECK-NEXT:    csel w8, w9, w10, lt
+; CHECK-NEXT:    fcmp s1, s3
+; CHECK-NEXT:    csel w8, w11, w8, gt
+; CHECK-NEXT:    fcmp s1, s1
+; CHECK-NEXT:    mov h1, v0.h[1]
+; CHECK-NEXT:    fcvt s1, h1
+; CHECK-NEXT:    fcvtzs w10, s1
+; CHECK-NEXT:    csel w4, wzr, w8, vs
+; CHECK-NEXT:    fcmp s1, s2
+; CHECK-NEXT:    mov h0, v0.h[2]
+; CHECK-NEXT:    csel w8, w9, w10, lt
+; CHECK-NEXT:    fcmp s1, s3
 ; CHECK-NEXT:    fcvt s0, h0
-; CHECK-NEXT:    csel w11, w9, w11, gt
-; CHECK-NEXT:    fcmp s4, s4
-; CHECK-NEXT:    mov v3.s[1], w10
-; CHECK-NEXT:    fcvtzs w10, s0
-; CHECK-NEXT:    csel w2, wzr, w11, vs
+; CHECK-NEXT:    csel w8, w11, w8, gt
+; CHECK-NEXT:    fcmp s1, s1
+; CHECK-NEXT:    fcvtzs w12, s0
+; CHECK-NEXT:    csel w5, wzr, w8, vs
 ; CHECK-NEXT:    fcmp s0, s2
-; CHECK-NEXT:    csel w8, w8, w10, lt
-; CHECK-NEXT:    fcmp s0, s1
-; CHECK-NEXT:    mov v3.s[2], w6
-; CHECK-NEXT:    csel w8, w9, w8, gt
+; CHECK-NEXT:    csel w8, w9, w12, lt
+; CHECK-NEXT:    fcmp s0, s3
+; CHECK-NEXT:    csel w8, w11, w8, gt
 ; CHECK-NEXT:    fcmp s0, s0
-; CHECK-NEXT:    csel w3, wzr, w8, vs
-; CHECK-NEXT:    mov w5, v3.s[1]
-; CHECK-NEXT:    fmov w4, s3
+; CHECK-NEXT:    csel w6, wzr, w8, vs
 ; CHECK-NEXT:    ret
     %x = call <7 x i32> @llvm.fptosi.sat.v7f16.v7i32(<7 x half> %f)
     ret <7 x i32> %x
 }
 
 define <8 x i32> @test_signed_v8f16_v8i32(<8 x half> %f) {
 ; CHECK-LABEL: test_signed_v8f16_v8i32:
 ; CHECK:       // %bb.0:

 ; CHECK-NEXT:    .cfi_offset w20, -16
 ; CHECK-NEXT:    .cfi_offset w21, -24
 ; CHECK-NEXT:    .cfi_offset w22, -32
 ; CHECK-NEXT:    .cfi_offset w30, -40
 ; CHECK-NEXT:    .cfi_offset b8, -48
 ; CHECK-NEXT:    .cfi_offset b9, -56
 ; CHECK-NEXT:    .cfi_offset b10, -64
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $q0
-; CHECK-NEXT:    mov s8, v0.s[1]
 ; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
-; CHECK-NEXT:    mov v0.16b, v8.16b
+; CHECK-NEXT:    // kill: def $s0 killed $s0 killed $q0
 ; CHECK-NEXT:    bl __fixsfti
+; CHECK-NEXT:    ldr q0, [sp] // 16-byte Folded Reload
 ; CHECK-NEXT:    mov w8, #-251658240
 ; CHECK-NEXT:    mov w9, #1895825407
 ; CHECK-NEXT:    fmov s9, w8
 ; CHECK-NEXT:    mov x21, #-34359738368
 ; CHECK-NEXT:    fmov s10, w9
-; CHECK-NEXT:    fcmp s8, s9
+; CHECK-NEXT:    fcmp s0, s9
 ; CHECK-NEXT:    mov x22, #34359738367
 ; CHECK-NEXT:    csel x8, xzr, x0, lt
 ; CHECK-NEXT:    csel x9, x21, x1, lt
-; CHECK-NEXT:    fcmp s8, s10
+; CHECK-NEXT:    fcmp s0, s10
+; CHECK-NEXT:    mov s8, v0.s[1]
 ; CHECK-NEXT:    csel x9, x22, x9, gt
 ; CHECK-NEXT:    csinv x8, x8, xzr, le
-; CHECK-NEXT:    fcmp s8, s8
-; CHECK-NEXT:    ldr q0, [sp] // 16-byte Folded Reload
+; CHECK-NEXT:    fcmp s0, s0
+; CHECK-NEXT:    mov v0.16b, v8.16b
 ; CHECK-NEXT:    csel x19, xzr, x8, vs
 ; CHECK-NEXT:    csel x20, xzr, x9, vs
-; CHECK-NEXT:    // kill: def $s0 killed $s0 killed $q0
 ; CHECK-NEXT:    bl __fixsfti
-; CHECK-NEXT:    ldr q0, [sp] // 16-byte Folded Reload
-; CHECK-NEXT:    mov x2, x19
-; CHECK-NEXT:    mov x3, x20
+; CHECK-NEXT:    fcmp s8, s9
+; CHECK-NEXT:    csel x8, xzr, x0, lt
+; CHECK-NEXT:    csel x9, x21, x1, lt
+; CHECK-NEXT:    fcmp s8, s10
+; CHECK-NEXT:    csel x9, x22, x9, gt
+; CHECK-NEXT:    csinv x8, x8, xzr, le
+; CHECK-NEXT:    fcmp s8, s8
+; CHECK-NEXT:    mov x0, x19
+; CHECK-NEXT:    mov x1, x20
 ; CHECK-NEXT:    ldp x20, x19, [sp, #64] // 16-byte Folded Reload
-; CHECK-NEXT:    fcmp s0, s9
-; CHECK-NEXT:    csel x8, x21, x1, lt
-; CHECK-NEXT:    csel x9, xzr, x0, lt
-; CHECK-NEXT:    fcmp s0, s10
-; CHECK-NEXT:    csinv x9, x9, xzr, le
-; CHECK-NEXT:    csel x8, x22, x8, gt
-; CHECK-NEXT:    fcmp s0, s0
-; CHECK-NEXT:    csel x9, xzr, x9, vs
 ; CHECK-NEXT:    ldp x22, x21, [sp, #48] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldr x30, [sp, #40] // 8-byte Folded Reload
 ; CHECK-NEXT:    ldp d9, d8, [sp, #24] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldr d10, [sp, #16] // 8-byte Folded Reload
-; CHECK-NEXT:    csel x1, xzr, x8, vs
-; CHECK-NEXT:    fmov d0, x9
-; CHECK-NEXT:    mov v0.d[1], x1
-; CHECK-NEXT:    fmov x0, d0
+; CHECK-NEXT:    csel x2, xzr, x8, vs
+; CHECK-NEXT:    csel x3, xzr, x9, vs
 ; CHECK-NEXT:    add sp, sp, #80 // =80
 ; CHECK-NEXT:    ret
     %x = call <2 x i100> @llvm.fptosi.sat.v2f32.v2i100(<2 x float> %f)
     ret <2 x i100> %x
 }
 
 define <2 x i128> @test_signed_v2f32_v2i128(<2 x float> %f) {
 ; CHECK-LABEL: test_signed_v2f32_v2i128:

 ; CHECK-NEXT:    .cfi_offset w20, -16
 ; CHECK-NEXT:    .cfi_offset w21, -24
 ; CHECK-NEXT:    .cfi_offset w22, -32
 ; CHECK-NEXT:    .cfi_offset w30, -40
 ; CHECK-NEXT:    .cfi_offset b8, -48
 ; CHECK-NEXT:    .cfi_offset b9, -56
 ; CHECK-NEXT:    .cfi_offset b10, -64
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $q0
-; CHECK-NEXT:    mov s8, v0.s[1]
 ; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
-; CHECK-NEXT:    mov v0.16b, v8.16b
+; CHECK-NEXT:    // kill: def $s0 killed $s0 killed $q0
 ; CHECK-NEXT:    bl __fixsfti
+; CHECK-NEXT:    ldr q0, [sp] // 16-byte Folded Reload
 ; CHECK-NEXT:    mov w8, #-16777216
 ; CHECK-NEXT:    mov w9, #2130706431
 ; CHECK-NEXT:    fmov s9, w8
 ; CHECK-NEXT:    mov x21, #-9223372036854775808
 ; CHECK-NEXT:    fmov s10, w9
-; CHECK-NEXT:    fcmp s8, s9
+; CHECK-NEXT:    fcmp s0, s9
 ; CHECK-NEXT:    mov x22, #9223372036854775807
 ; CHECK-NEXT:    csel x8, xzr, x0, lt
 ; CHECK-NEXT:    csel x9, x21, x1, lt
-; CHECK-NEXT:    fcmp s8, s10
+; CHECK-NEXT:    fcmp s0, s10
+; CHECK-NEXT:    mov s8, v0.s[1]
 ; CHECK-NEXT:    csel x9, x22, x9, gt
 ; CHECK-NEXT:    csinv x8, x8, xzr, le
-; CHECK-NEXT:    fcmp s8, s8
-; CHECK-NEXT:    ldr q0, [sp] // 16-byte Folded Reload
+; CHECK-NEXT:    fcmp s0, s0
+; CHECK-NEXT:    mov v0.16b, v8.16b
 ; CHECK-NEXT:    csel x19, xzr, x8, vs
 ; CHECK-NEXT:    csel x20, xzr, x9, vs
-; CHECK-NEXT:    // kill: def $s0 killed $s0 killed $q0
 ; CHECK-NEXT:    bl __fixsfti
-; CHECK-NEXT:    ldr q0, [sp] // 16-byte Folded Reload
-; CHECK-NEXT:    mov x2, x19
-; CHECK-NEXT:    mov x3, x20
+; CHECK-NEXT:    fcmp s8, s9
+; CHECK-NEXT:    csel x8, xzr, x0, lt
+; CHECK-NEXT:    csel x9, x21, x1, lt
+; CHECK-NEXT:    fcmp s8, s10
+; CHECK-NEXT:    csel x9, x22, x9, gt
+; CHECK-NEXT:    csinv x8, x8, xzr, le
+; CHECK-NEXT:    fcmp s8, s8
+; CHECK-NEXT:    mov x0, x19
+; CHECK-NEXT:    mov x1, x20
 ; CHECK-NEXT:    ldp x20, x19, [sp, #64] // 16-byte Folded Reload
-; CHECK-NEXT:    fcmp s0, s9
-; CHECK-NEXT:    csel x8, x21, x1, lt
-; CHECK-NEXT:    csel x9, xzr, x0, lt
-; CHECK-NEXT:    fcmp s0, s10
-; CHECK-NEXT:    csinv x9, x9, xzr, le
-; CHECK-NEXT:    csel x8, x22, x8, gt
-; CHECK-NEXT:    fcmp s0, s0
-; CHECK-NEXT:    csel x9, xzr, x9, vs
 ; CHECK-NEXT:    ldp x22, x21, [sp, #48] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldr x30, [sp, #40] // 8-byte Folded Reload
 ; CHECK-NEXT:    ldp d9, d8, [sp, #24] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldr d10, [sp, #16] // 8-byte Folded Reload
-; CHECK-NEXT:    csel x1, xzr, x8, vs
-; CHECK-NEXT:    fmov d0, x9
-; CHECK-NEXT:    mov v0.d[1], x1
-; CHECK-NEXT:    fmov x0, d0
+; CHECK-NEXT:    csel x2, xzr, x8, vs
+; CHECK-NEXT:    csel x3, xzr, x9, vs
 ; CHECK-NEXT:    add sp, sp, #80 // =80
 ; CHECK-NEXT:    ret
     %x = call <2 x i128> @llvm.fptosi.sat.v2f32.v2i128(<2 x float> %f)
     ret <2 x i128> %x
 }
 
 ;
 ; 2-Vector double to signed integer -- result size variation

 ; CHECK-NEXT:    .cfi_offset w19, -8
 ; CHECK-NEXT:    .cfi_offset w20, -16
 ; CHECK-NEXT:    .cfi_offset w21, -24
 ; CHECK-NEXT:    .cfi_offset w22, -32
 ; CHECK-NEXT:    .cfi_offset w30, -40
 ; CHECK-NEXT:    .cfi_offset b8, -48
 ; CHECK-NEXT:    .cfi_offset b9, -56
 ; CHECK-NEXT:    .cfi_offset b10, -64
-; CHECK-NEXT:    mov d8, v0.d[1]
 ; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
-; CHECK-NEXT:    mov v0.16b, v8.16b
+; CHECK-NEXT:    // kill: def $d0 killed $d0 killed $q0
 ; CHECK-NEXT:    bl __fixdfti
+; CHECK-NEXT:    ldr q0, [sp] // 16-byte Folded Reload
 ; CHECK-NEXT:    mov x8, #-4170333254945079296
 ; CHECK-NEXT:    mov x9, #5053038781909696511
 ; CHECK-NEXT:    fmov d9, x8
 ; CHECK-NEXT:    mov x21, #-34359738368
 ; CHECK-NEXT:    fmov d10, x9
-; CHECK-NEXT:    fcmp d8, d9
+; CHECK-NEXT:    fcmp d0, d9
 ; CHECK-NEXT:    mov x22, #34359738367
 ; CHECK-NEXT:    csel x8, xzr, x0, lt
 ; CHECK-NEXT:    csel x9, x21, x1, lt
-; CHECK-NEXT:    fcmp d8, d10
+; CHECK-NEXT:    fcmp d0, d10
+; CHECK-NEXT:    mov d8, v0.d[1]
 ; CHECK-NEXT:    csel x9, x22, x9, gt
 ; CHECK-NEXT:    csinv x8, x8, xzr, le
-; CHECK-NEXT:    fcmp d8, d8
-; CHECK-NEXT:    ldr q0, [sp] // 16-byte Folded Reload
+; CHECK-NEXT:    fcmp d0, d0
+; CHECK-NEXT:    mov v0.16b, v8.16b
 ; CHECK-NEXT:    csel x19, xzr, x8, vs
 ; CHECK-NEXT:    csel x20, xzr, x9, vs
-; CHECK-NEXT:    // kill: def $d0 killed $d0 killed $q0
 ; CHECK-NEXT:    bl __fixdfti
-; CHECK-NEXT:    ldr q0, [sp] // 16-byte Folded Reload
-; CHECK-NEXT:    mov x2, x19
-; CHECK-NEXT:    mov x3, x20
+; CHECK-NEXT:    fcmp d8, d9
+; CHECK-NEXT:    csel x8, xzr, x0, lt
+; CHECK-NEXT:    csel x9, x21, x1, lt
+; CHECK-NEXT:    fcmp d8, d10
+; CHECK-NEXT:    csel x9, x22, x9, gt
+; CHECK-NEXT:    csinv x8, x8, xzr, le
+; CHECK-NEXT:    fcmp d8, d8
+; CHECK-NEXT:    mov x0, x19
+; CHECK-NEXT:    mov x1, x20
 ; CHECK-NEXT:    ldp x20, x19, [sp, #64] // 16-byte Folded Reload
-; CHECK-NEXT:    fcmp d0, d9
-; CHECK-NEXT:    csel x8, x21, x1, lt
-; CHECK-NEXT:    csel x9, xzr, x0, lt
-; CHECK-NEXT:    fcmp d0, d10
-; CHECK-NEXT:    csinv x9, x9, xzr, le
-; CHECK-NEXT:    csel x8, x22, x8, gt
-; CHECK-NEXT:    fcmp d0, d0
-; CHECK-NEXT:    csel x9, xzr, x9, vs
 ; CHECK-NEXT:    ldp x22, x21, [sp, #48] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldr x30, [sp, #40] // 8-byte Folded Reload
 ; CHECK-NEXT:    ldp d9, d8, [sp, #24] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldr d10, [sp, #16] // 8-byte Folded Reload
-; CHECK-NEXT:    csel x1, xzr, x8, vs
-; CHECK-NEXT:    fmov d0, x9
-; CHECK-NEXT:    mov v0.d[1], x1
-; CHECK-NEXT:    fmov x0, d0
+; CHECK-NEXT:    csel x2, xzr, x8, vs
+; CHECK-NEXT:    csel x3, xzr, x9, vs
 ; CHECK-NEXT:    add sp, sp, #80 // =80
 ; CHECK-NEXT:    ret
     %x = call <2 x i100> @llvm.fptosi.sat.v2f64.v2i100(<2 x double> %f)
     ret <2 x i100> %x
 }
 
 define <2 x i128> @test_signed_v2f64_v2i128(<2 x double> %f) {
 ; CHECK-LABEL: test_signed_v2f64_v2i128:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    sub sp, sp, #80 // =80
 ; CHECK-NEXT:    str d10, [sp, #16] // 8-byte Folded Spill
 ; CHECK-NEXT:    stp d9, d8, [sp, #24] // 16-byte Folded Spill
 ; CHECK-NEXT:    str x30, [sp, #40] // 8-byte Folded Spill
 ; CHECK-NEXT:    stp x22, x21, [sp, #48] // 16-byte Folded Spill
 ; CHECK-NEXT:    stp x20, x19, [sp, #64] // 16-byte Folded Spill
 ; CHECK-NEXT:    .cfi_def_cfa_offset 80
 ; CHECK-NEXT:    .cfi_offset w19, -8
 ; CHECK-NEXT:    .cfi_offset w20, -16
 ; CHECK-NEXT:    .cfi_offset w21, -24
 ; CHECK-NEXT:    .cfi_offset w22, -32
 ; CHECK-NEXT:    .cfi_offset w30, -40
 ; CHECK-NEXT:    .cfi_offset b8, -48
 ; CHECK-NEXT:    .cfi_offset b9, -56
 ; CHECK-NEXT:    .cfi_offset b10, -64
-; CHECK-NEXT:    mov d8, v0.d[1]
 ; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
-; CHECK-NEXT:    mov v0.16b, v8.16b
+; CHECK-NEXT:    // kill: def $d0 killed $d0 killed $q0
 ; CHECK-NEXT:    bl __fixdfti
+; CHECK-NEXT:    ldr q0, [sp] // 16-byte Folded Reload
 ; CHECK-NEXT:    mov x8, #-4044232465378705408
 ; CHECK-NEXT:    mov x9, #5179139571476070399
 ; CHECK-NEXT:    fmov d9, x8
 ; CHECK-NEXT:    mov x21, #-9223372036854775808
 ; CHECK-NEXT:    fmov d10, x9
-; CHECK-NEXT:    fcmp d8, d9
+; CHECK-NEXT:    fcmp d0, d9
 ; CHECK-NEXT:    mov x22, #9223372036854775807
 ; CHECK-NEXT:    csel x8, xzr, x0, lt
 ; CHECK-NEXT:    csel x9, x21, x1, lt
-; CHECK-NEXT:    fcmp d8, d10
+; CHECK-NEXT:    fcmp d0, d10
+; CHECK-NEXT:    mov d8, v0.d[1]
 ; CHECK-NEXT:    csel x9, x22, x9, gt
 ; CHECK-NEXT:    csinv x8, x8, xzr, le
-; CHECK-NEXT:    fcmp d8, d8
-; CHECK-NEXT:    ldr q0, [sp] // 16-byte Folded Reload
+; CHECK-NEXT:    fcmp d0, d0
+; CHECK-NEXT:    mov v0.16b, v8.16b
 ; CHECK-NEXT:    csel x19, xzr, x8, vs
 ; CHECK-NEXT:    csel x20, xzr, x9, vs
-; CHECK-NEXT:    // kill: def $d0 killed $d0 killed $q0
 ; CHECK-NEXT:    bl __fixdfti
-; CHECK-NEXT:    ldr q0, [sp] // 16-byte Folded Reload
-; CHECK-NEXT:    mov x2, x19
-; CHECK-NEXT:    mov x3, x20
+; CHECK-NEXT:    fcmp d8, d9
+; CHECK-NEXT:    csel x8, xzr, x0, lt
+; CHECK-NEXT:    csel x9, x21, x1, lt
+; CHECK-NEXT:    fcmp d8, d10
+; CHECK-NEXT:    csel x9, x22, x9, gt
+; CHECK-NEXT:    csinv x8, x8, xzr, le
+; CHECK-NEXT:    fcmp d8, d8
+; CHECK-NEXT:    mov x0, x19
+; CHECK-NEXT:    mov x1, x20
 ; CHECK-NEXT:    ldp x20, x19, [sp, #64] // 16-byte Folded Reload
-; CHECK-NEXT:    fcmp d0, d9
-; CHECK-NEXT:    csel x8, x21, x1, lt
-; CHECK-NEXT:    csel x9, xzr, x0, lt
-; CHECK-NEXT:    fcmp d0, d10
-; CHECK-NEXT:    csinv x9, x9, xzr, le
-; CHECK-NEXT:    csel x8, x22, x8, gt
-; CHECK-NEXT:    fcmp d0, d0
-; CHECK-NEXT:    csel x9, xzr, x9, vs
 ; CHECK-NEXT:    ldp x22, x21, [sp, #48] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldr x30, [sp, #40] // 8-byte Folded Reload
 ; CHECK-NEXT:    ldp d9, d8, [sp, #24] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldr d10, [sp, #16] // 8-byte Folded Reload
-; CHECK-NEXT:    csel x1, xzr, x8, vs
-; CHECK-NEXT:    fmov d0, x9
-; CHECK-NEXT:    mov v0.d[1], x1
-; CHECK-NEXT:    fmov x0, d0
+; CHECK-NEXT:    csel x2, xzr, x8, vs
+; CHECK-NEXT:    csel x3, xzr, x9, vs
 ; CHECK-NEXT:    add sp, sp, #80 // =80
 ; CHECK-NEXT:    ret
     %x = call <2 x i128> @llvm.fptosi.sat.v2f64.v2i128(<2 x double> %f)
     ret <2 x i128> %x
 }
 
 ;
 ; 4-Vector half to signed integer -- result size variation

 ; CHECK-NEXT:    .cfi_offset w24, -48
 ; CHECK-NEXT:    .cfi_offset w25, -56
 ; CHECK-NEXT:    .cfi_offset w26, -64
 ; CHECK-NEXT:    .cfi_offset w30, -72
 ; CHECK-NEXT:    .cfi_offset b8, -80
 ; CHECK-NEXT:    .cfi_offset b9, -88
 ; CHECK-NEXT:    .cfi_offset b10, -96
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $q0
-; CHECK-NEXT:    mov h1, v0.h[1]
-; CHECK-NEXT:    fcvt s8, h1
+; CHECK-NEXT:    fcvt s8, h0
 ; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
 ; CHECK-NEXT:    mov v0.16b, v8.16b
 ; CHECK-NEXT:    bl __fixsfti
 ; CHECK-NEXT:    ldr q0, [sp] // 16-byte Folded Reload
 ; CHECK-NEXT:    mov w8, #-251658240
 ; CHECK-NEXT:    mov w9, #1895825407
 ; CHECK-NEXT:    fmov s9, w8
 ; CHECK-NEXT:    mov x25, #-34359738368
 ; CHECK-NEXT:    fmov s10, w9
 ; CHECK-NEXT:    fcmp s8, s9
 ; CHECK-NEXT:    mov x26, #34359738367
 ; CHECK-NEXT:    csel x8, xzr, x0, lt
 ; CHECK-NEXT:    csel x9, x25, x1, lt
 ; CHECK-NEXT:    fcmp s8, s10
-; CHECK-NEXT:    mov h0, v0.h[2]
+; CHECK-NEXT:    mov h0, v0.h[1]
 ; CHECK-NEXT:    csel x9, x26, x9, gt
 ; CHECK-NEXT:    csinv x8, x8, xzr, le
 ; CHECK-NEXT:    fcmp s8, s8
 ; CHECK-NEXT:    fcvt s8, h0
 ; CHECK-NEXT:    mov v0.16b, v8.16b
 ; CHECK-NEXT:    csel x19, xzr, x8, vs
 ; CHECK-NEXT:    csel x20, xzr, x9, vs
 ; CHECK-NEXT:    bl __fixsfti
 ; CHECK-NEXT:    ldr q0, [sp] // 16-byte Folded Reload
 ; CHECK-NEXT:    fcmp s8, s9
 ; CHECK-NEXT:    csel x8, xzr, x0, lt
 ; CHECK-NEXT:    csel x9, x25, x1, lt
-; CHECK-NEXT:    mov h0, v0.h[3]
+; CHECK-NEXT:    mov h0, v0.h[2]
 ; CHECK-NEXT:    fcmp s8, s10
 ; CHECK-NEXT:    csel x9, x26, x9, gt
 ; CHECK-NEXT:    csinv x8, x8, xzr, le
 ; CHECK-NEXT:    fcmp s8, s8
 ; CHECK-NEXT:    fcvt s8, h0
 ; CHECK-NEXT:    mov v0.16b, v8.16b
 ; CHECK-NEXT:    csel x21, xzr, x8, vs
 ; CHECK-NEXT:    csel x22, xzr, x9, vs
 ; CHECK-NEXT:    bl __fixsfti
 ; CHECK-NEXT:    ldr q0, [sp] // 16-byte Folded Reload
 ; CHECK-NEXT:    fcmp s8, s9
 ; CHECK-NEXT:    csel x8, xzr, x0, lt
 ; CHECK-NEXT:    csel x9, x25, x1, lt
+; CHECK-NEXT:    mov h0, v0.h[3]
 ; CHECK-NEXT:    fcmp s8, s10
 ; CHECK-NEXT:    csel x9, x26, x9, gt
 ; CHECK-NEXT:    csinv x8, x8, xzr, le
 ; CHECK-NEXT:    fcmp s8, s8
 ; CHECK-NEXT:    fcvt s8, h0
 ; CHECK-NEXT:    mov v0.16b, v8.16b
 ; CHECK-NEXT:    csel x23, xzr, x8, vs
 ; CHECK-NEXT:    csel x24, xzr, x9, vs
 ; CHECK-NEXT:    bl __fixsfti
 ; CHECK-NEXT:    fcmp s8, s9
-; CHECK-NEXT:    csel x8, x25, x1, lt
-; CHECK-NEXT:    csel x9, xzr, x0, lt
+; CHECK-NEXT:    csel x8, xzr, x0, lt
+; CHECK-NEXT:    csel x9, x25, x1, lt
 ; CHECK-NEXT:    fcmp s8, s10
-; CHECK-NEXT:    csinv x9, x9, xzr, le
-; CHECK-NEXT:    csel x8, x26, x8, gt
+; CHECK-NEXT:    csel x9, x26, x9, gt
+; CHECK-NEXT:    csinv x8, x8, xzr, le
 ; CHECK-NEXT:    fcmp s8, s8
-; CHECK-NEXT:    csel x9, xzr, x9, vs
-; CHECK-NEXT:    mov x2, x19
-; CHECK-NEXT:    mov x3, x20
-; CHECK-NEXT:    mov x4, x21
-; CHECK-NEXT:    mov x5, x22
-; CHECK-NEXT:    mov x6, x23
-; CHECK-NEXT:    mov x7, x24
+; CHECK-NEXT:    mov x0, x19
+; CHECK-NEXT:    mov x1, x20
+; CHECK-NEXT:    mov x2, x21
+; CHECK-NEXT:    mov x3, x22
+; CHECK-NEXT:    mov x4, x23
+; CHECK-NEXT:    mov x5, x24
 ; CHECK-NEXT:    ldp x20, x19, [sp, #96] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp x22, x21, [sp, #80] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp x24, x23, [sp, #64] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp x26, x25, [sp, #48] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldr x30, [sp, #40] // 8-byte Folded Reload
 ; CHECK-NEXT:    ldp d9, d8, [sp, #24] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldr d10, [sp, #16] // 8-byte Folded Reload
-; CHECK-NEXT:    csel x1, xzr, x8, vs
-; CHECK-NEXT:    fmov d0, x9
-; CHECK-NEXT:    mov v0.d[1], x1
-; CHECK-NEXT:    fmov x0, d0
+; CHECK-NEXT:    csel x6, xzr, x8, vs
+; CHECK-NEXT:    csel x7, xzr, x9, vs
 ; CHECK-NEXT:    add sp, sp, #112 // =112
 ; CHECK-NEXT:    ret
     %x = call <4 x i100> @llvm.fptosi.sat.v4f16.v4i100(<4 x half> %f)
     ret <4 x i100> %x
 }
 
 define <4 x i128> @test_signed_v4f16_v4i128(<4 x half> %f) {
 ; CHECK-LABEL: test_signed_v4f16_v4i128:

 ; CHECK-NEXT:    .cfi_offset w24, -48
 ; CHECK-NEXT:    .cfi_offset w25, -56
 ; CHECK-NEXT:    .cfi_offset w26, -64
 ; CHECK-NEXT:    .cfi_offset w30, -72
 ; CHECK-NEXT:    .cfi_offset b8, -80
 ; CHECK-NEXT:    .cfi_offset b9, -88
 ; CHECK-NEXT:    .cfi_offset b10, -96
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $q0
-; CHECK-NEXT:    mov h1, v0.h[1]
-; CHECK-NEXT:    fcvt s8, h1
+; CHECK-NEXT:    fcvt s8, h0
 ; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
 ; CHECK-NEXT:    mov v0.16b, v8.16b
 ; CHECK-NEXT:    bl __fixsfti
 ; CHECK-NEXT:    ldr q0, [sp] // 16-byte Folded Reload
 ; CHECK-NEXT:    mov w8, #-16777216
 ; CHECK-NEXT:    mov w9, #2130706431
 ; CHECK-NEXT:    fmov s9, w8
 ; CHECK-NEXT:    mov x25, #-9223372036854775808
 ; CHECK-NEXT:    fmov s10, w9
 ; CHECK-NEXT:    fcmp s8, s9
 ; CHECK-NEXT:    mov x26, #9223372036854775807
 ; CHECK-NEXT:    csel x8, xzr, x0, lt
 ; CHECK-NEXT:    csel x9, x25, x1, lt
 ; CHECK-NEXT:    fcmp s8, s10
-; CHECK-NEXT:    mov h0, v0.h[2]
+; CHECK-NEXT:    mov h0, v0.h[1]
 ; CHECK-NEXT:    csel x9, x26, x9, gt
 ; CHECK-NEXT:    csinv x8, x8, xzr, le
 ; CHECK-NEXT:    fcmp s8, s8
 ; CHECK-NEXT:    fcvt s8, h0
 ; CHECK-NEXT:    mov v0.16b, v8.16b
 ; CHECK-NEXT:    csel x19, xzr, x8, vs
 ; CHECK-NEXT:    csel x20, xzr, x9, vs
 ; CHECK-NEXT:    bl __fixsfti
 ; CHECK-NEXT:    ldr q0, [sp] // 16-byte Folded Reload
 ; CHECK-NEXT:    fcmp s8, s9
 ; CHECK-NEXT:    csel x8, xzr, x0, lt
 ; CHECK-NEXT:    csel x9, x25, x1, lt
-; CHECK-NEXT:    mov h0, v0.h[3]
+; CHECK-NEXT:    mov h0, v0.h[2]
 ; CHECK-NEXT:    fcmp s8, s10
 ; CHECK-NEXT:    csel x9, x26, x9, gt
 ; CHECK-NEXT:    csinv x8, x8, xzr, le
 ; CHECK-NEXT:    fcmp s8, s8
 ; CHECK-NEXT:    fcvt s8, h0
 ; CHECK-NEXT:    mov v0.16b, v8.16b
 ; CHECK-NEXT:    csel x21, xzr, x8, vs
 ; CHECK-NEXT:    csel x22, xzr, x9, vs
 ; CHECK-NEXT:    bl __fixsfti
 ; CHECK-NEXT:    ldr q0, [sp] // 16-byte Folded Reload
 ; CHECK-NEXT:    fcmp s8, s9
 ; CHECK-NEXT:    csel x8, xzr, x0, lt
 ; CHECK-NEXT:    csel x9, x25, x1, lt
+; CHECK-NEXT:    mov h0, v0.h[3]
 ; CHECK-NEXT:    fcmp s8, s10
 ; CHECK-NEXT:    csel x9, x26, x9, gt
 ; CHECK-NEXT:    csinv x8, x8, xzr, le
 ; CHECK-NEXT:    fcmp s8, s8
 ; CHECK-NEXT:    fcvt s8, h0
 ; CHECK-NEXT:    mov v0.16b, v8.16b
 ; CHECK-NEXT:    csel x23, xzr, x8, vs
 ; CHECK-NEXT:    csel x24, xzr, x9, vs
 ; CHECK-NEXT:    bl __fixsfti
 ; CHECK-NEXT:    fcmp s8, s9
-; CHECK-NEXT:    csel x8, x25, x1, lt
-; CHECK-NEXT:    csel x9, xzr, x0, lt
+; CHECK-NEXT:    csel x8, xzr, x0, lt
+; CHECK-NEXT:    csel x9, x25, x1, lt
 ; CHECK-NEXT:    fcmp s8, s10
-; CHECK-NEXT:    csinv x9, x9, xzr, le
-; CHECK-NEXT:    csel x8, x26, x8, gt
+; CHECK-NEXT:    csel x9, x26, x9, gt
+; CHECK-NEXT:    csinv x8, x8, xzr, le
 ; CHECK-NEXT:    fcmp s8, s8
-; CHECK-NEXT:    csel x9, xzr, x9, vs
-; CHECK-NEXT:    mov x2, x19
-; CHECK-NEXT:    mov x3, x20
-; CHECK-NEXT:    mov x4, x21
-; CHECK-NEXT:    mov x5, x22
-; CHECK-NEXT:    mov x6, x23
-; CHECK-NEXT:    mov x7, x24
+; CHECK-NEXT:    mov x0, x19
+; CHECK-NEXT:    mov x1, x20
+; CHECK-NEXT:    mov x2, x21
+; CHECK-NEXT:    mov x3, x22
+; CHECK-NEXT:    mov x4, x23
+; CHECK-NEXT:    mov x5, x24
 ; CHECK-NEXT:    ldp x20, x19, [sp, #96] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp x22, x21, [sp, #80] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp x24, x23, [sp, #64] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp x26, x25, [sp, #48] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldr x30, [sp, #40] // 8-byte Folded Reload
 ; CHECK-NEXT:    ldp d9, d8, [sp, #24] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldr d10, [sp, #16] // 8-byte Folded Reload
-; CHECK-NEXT:    csel x1, xzr, x8, vs
-; CHECK-NEXT:    fmov d0, x9
-; CHECK-NEXT:    mov v0.d[1], x1
-; CHECK-NEXT:    fmov x0, d0
+; CHECK-NEXT:    csel x6, xzr, x8, vs
+; CHECK-NEXT:    csel x7, xzr, x9, vs
 ; CHECK-NEXT:    add sp, sp, #112 // =112
 ; CHECK-NEXT:    ret
     %x = call <4 x i128> @llvm.fptosi.sat.v4f16.v4i128(<4 x half> %f)
     ret <4 x i128> %x
 }
 

llvm/test/CodeGen/AArch64/fptoui-sat-vector.ll

; CHECK-NEXT:    ret
     %x = call <5 x i32> @llvm.fptoui.sat.v5f32.v5i32(<5 x float> %f)
     ret <5 x i32> %x
 }
 
 define <6 x i32> @test_unsigned_v6f32_v6i32(<6 x float> %f) {
 ; CHECK-LABEL: test_unsigned_v6f32_v6i32:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov w9, #1333788671
-; CHECK-NEXT:    fcvtzu w8, s5
-; CHECK-NEXT:    fcmp s5, #0.0
+; CHECK-NEXT:    fcvtzu w8, s0
+; CHECK-NEXT:    fcmp s0, #0.0
 ; CHECK-NEXT:    fmov s6, w9
 ; CHECK-NEXT:    csel w8, wzr, w8, lt
-; CHECK-NEXT:    fcmp s5, s6
-; CHECK-NEXT:    fcvtzu w10, s4
-; CHECK-NEXT:    csinv w5, w8, wzr, le
-; CHECK-NEXT:    fcmp s4, #0.0
-; CHECK-NEXT:    csel w8, wzr, w10, lt
-; CHECK-NEXT:    fcmp s4, s6
-; CHECK-NEXT:    fcvtzu w11, s0
-; CHECK-NEXT:    csinv w8, w8, wzr, le
-; CHECK-NEXT:    fcmp s0, #0.0
-; CHECK-NEXT:    fmov s4, w8
-; CHECK-NEXT:    csel w8, wzr, w11, lt
 ; CHECK-NEXT:    fcmp s0, s6
-; CHECK-NEXT:    fcvtzu w12, s1
+; CHECK-NEXT:    fcvtzu w10, s1
 ; CHECK-NEXT:    csinv w0, w8, wzr, le
 ; CHECK-NEXT:    fcmp s1, #0.0
-; CHECK-NEXT:    csel w8, wzr, w12, lt
+; CHECK-NEXT:    csel w8, wzr, w10, lt
 ; CHECK-NEXT:    fcmp s1, s6
-; CHECK-NEXT:    fcvtzu w13, s2
+; CHECK-NEXT:    fcvtzu w11, s2
 ; CHECK-NEXT:    csinv w1, w8, wzr, le
 ; CHECK-NEXT:    fcmp s2, #0.0
-; CHECK-NEXT:    csel w8, wzr, w13, lt
+; CHECK-NEXT:    csel w8, wzr, w11, lt
 ; CHECK-NEXT:    fcmp s2, s6
-; CHECK-NEXT:    fcvtzu w9, s3
+; CHECK-NEXT:    fcvtzu w12, s3
 ; CHECK-NEXT:    csinv w2, w8, wzr, le
 ; CHECK-NEXT:    fcmp s3, #0.0
-; CHECK-NEXT:    mov v4.s[1], w5
-; CHECK-NEXT:    csel w8, wzr, w9, lt
+; CHECK-NEXT:    csel w8, wzr, w12, lt
 ; CHECK-NEXT:    fcmp s3, s6
+; CHECK-NEXT:    fcvtzu w13, s4
 ; CHECK-NEXT:    csinv w3, w8, wzr, le
-; CHECK-NEXT:    fmov w4, s4
+; CHECK-NEXT:    fcmp s4, #0.0
+; CHECK-NEXT:    csel w8, wzr, w13, lt
+; CHECK-NEXT:    fcmp s4, s6
+; CHECK-NEXT:    fcvtzu w9, s5
+; CHECK-NEXT:    csinv w4, w8, wzr, le
+; CHECK-NEXT:    fcmp s5, #0.0
+; CHECK-NEXT:    csel w8, wzr, w9, lt
+; CHECK-NEXT:    fcmp s5, s6
+; CHECK-NEXT:    csinv w5, w8, wzr, le
 ; CHECK-NEXT:    ret
     %x = call <6 x i32> @llvm.fptoui.sat.v6f32.v6i32(<6 x float> %f)
     ret <6 x i32> %x
 }
 
 define <7 x i32> @test_unsigned_v7f32_v7i32(<7 x float> %f) {
 ; CHECK-LABEL: test_unsigned_v7f32_v7i32:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov w9, #1333788671
-; CHECK-NEXT:    fcvtzu w8, s5
-; CHECK-NEXT:    fcmp s5, #0.0
+; CHECK-NEXT:    fcvtzu w8, s0
+; CHECK-NEXT:    fcmp s0, #0.0
 ; CHECK-NEXT:    fmov s7, w9
 ; CHECK-NEXT:    csel w8, wzr, w8, lt
-; CHECK-NEXT:    fcmp s5, s7
-; CHECK-NEXT:    fcvtzu w10, s4
-; CHECK-NEXT:    csinv w8, w8, wzr, le
-; CHECK-NEXT:    fcmp s4, #0.0
-; CHECK-NEXT:    csel w10, wzr, w10, lt
-; CHECK-NEXT:    fcmp s4, s7
-; CHECK-NEXT:    fcvtzu w11, s6
-; CHECK-NEXT:    csinv w10, w10, wzr, le
-; CHECK-NEXT:    fcmp s6, #0.0
-; CHECK-NEXT:    fmov s4, w10
-; CHECK-NEXT:    csel w10, wzr, w11, lt
-; CHECK-NEXT:    fcmp s6, s7
-; CHECK-NEXT:    fcvtzu w12, s0
-; CHECK-NEXT:    csinv w6, w10, wzr, le
-; CHECK-NEXT:    fcmp s0, #0.0
-; CHECK-NEXT:    mov v4.s[1], w8
-; CHECK-NEXT:    csel w8, wzr, w12, lt
 ; CHECK-NEXT:    fcmp s0, s7
-; CHECK-NEXT:    fcvtzu w13, s1
+; CHECK-NEXT:    fcvtzu w10, s1
 ; CHECK-NEXT:    csinv w0, w8, wzr, le
 ; CHECK-NEXT:    fcmp s1, #0.0
-; CHECK-NEXT:    csel w8, wzr, w13, lt
+; CHECK-NEXT:    csel w8, wzr, w10, lt
 ; CHECK-NEXT:    fcmp s1, s7
-; CHECK-NEXT:    fcvtzu w14, s2
+; CHECK-NEXT:    fcvtzu w11, s2
 ; CHECK-NEXT:    csinv w1, w8, wzr, le
 ; CHECK-NEXT:    fcmp s2, #0.0
-; CHECK-NEXT:    csel w8, wzr, w14, lt
+; CHECK-NEXT:    csel w8, wzr, w11, lt
 ; CHECK-NEXT:    fcmp s2, s7
-; CHECK-NEXT:    fcvtzu w9, s3
+; CHECK-NEXT:    fcvtzu w12, s3
 ; CHECK-NEXT:    csinv w2, w8, wzr, le
 ; CHECK-NEXT:    fcmp s3, #0.0
-; CHECK-NEXT:    mov v4.s[2], w6
-; CHECK-NEXT:    csel w8, wzr, w9, lt
+; CHECK-NEXT:    csel w8, wzr, w12, lt
 ; CHECK-NEXT:    fcmp s3, s7
+; CHECK-NEXT:    fcvtzu w13, s4
 ; CHECK-NEXT:    csinv w3, w8, wzr, le
-; CHECK-NEXT:    mov w5, v4.s[1]
-; CHECK-NEXT:    fmov w4, s4
+; CHECK-NEXT:    fcmp s4, #0.0
+; CHECK-NEXT:    csel w8, wzr, w13, lt
+; CHECK-NEXT:    fcmp s4, s7
+; CHECK-NEXT:    fcvtzu w14, s5
+; CHECK-NEXT:    csinv w4, w8, wzr, le
+; CHECK-NEXT:    fcmp s5, #0.0
+; CHECK-NEXT:    csel w8, wzr, w14, lt
+; CHECK-NEXT:    fcmp s5, s7
+; CHECK-NEXT:    fcvtzu w9, s6
+; CHECK-NEXT:    csinv w5, w8, wzr, le
+; CHECK-NEXT:    fcmp s6, #0.0
+; CHECK-NEXT:    csel w8, wzr, w9, lt
+; CHECK-NEXT:    fcmp s6, s7
+; CHECK-NEXT:    csinv w6, w8, wzr, le
 ; CHECK-NEXT:    ret
     %x = call <7 x i32> @llvm.fptoui.sat.v7f32.v7i32(<7 x float> %f)
     ret <7 x i32> %x
 }
 
 define <8 x i32> @test_unsigned_v8f32_v8i32(<8 x float> %f) {
 ; CHECK-LABEL: test_unsigned_v8f32_v8i32:
 ; CHECK:       // %bb.0:

 ; CHECK-NEXT:    ret
     %x = call <5 x i32> @llvm.fptoui.sat.v5f16.v5i32(<5 x half> %f)
     ret <5 x i32> %x
 }
 
 define <6 x i32> @test_unsigned_v6f16_v6i32(<6 x half> %f) {
 ; CHECK-LABEL: test_unsigned_v6f16_v6i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ext v1.16b, v0.16b, v0.16b, #8
-; CHECK-NEXT:    mov h2, v1.h[1]
+; CHECK-NEXT:    fcvt s1, h0
 ; CHECK-NEXT:    mov w8, #1333788671
-; CHECK-NEXT:    fcvt s2, h2
-; CHECK-NEXT:    fmov s3, w8
-; CHECK-NEXT:    fcvtzu w8, s2
-; CHECK-NEXT:    fcmp s2, #0.0
-; CHECK-NEXT:    csel w8, wzr, w8, lt
-; CHECK-NEXT:    fcmp s2, s3
-; CHECK-NEXT:    fcvt s1, h1
 ; CHECK-NEXT:    fcvtzu w9, s1
-; CHECK-NEXT:    csinv w5, w8, wzr, le
 ; CHECK-NEXT:    fcmp s1, #0.0
-; CHECK-NEXT:    fcvt s2, h0
+; CHECK-NEXT:    fmov s2, w8
 ; CHECK-NEXT:    csel w8, wzr, w9, lt
-; CHECK-NEXT:    fcmp s1, s3
+; CHECK-NEXT:    fcmp s1, s2
 ; CHECK-NEXT:    mov h1, v0.h[1]
-; CHECK-NEXT:    fcvtzu w9, s2
-; CHECK-NEXT:    csinv w8, w8, wzr, le
-; CHECK-NEXT:    fcmp s2, #0.0
-; CHECK-NEXT:    csel w9, wzr, w9, lt
-; CHECK-NEXT:    fcmp s2, s3
 ; CHECK-NEXT:    fcvt s1, h1
-; CHECK-NEXT:    mov h2, v0.h[2]
-; CHECK-NEXT:    fcvtzu w10, s1
-; CHECK-NEXT:    csinv w0, w9, wzr, le
+; CHECK-NEXT:    fcvtzu w9, s1
+; CHECK-NEXT:    csinv w0, w8, wzr, le
 ; CHECK-NEXT:    fcmp s1, #0.0
-; CHECK-NEXT:    fcvt s2, h2
-; CHECK-NEXT:    csel w9, wzr, w10, lt
-; CHECK-NEXT:    fcmp s1, s3
-; CHECK-NEXT:    mov h0, v0.h[3]
-; CHECK-NEXT:    fcvtzu w11, s2
-; CHECK-NEXT:    csinv w1, w9, wzr, le
-; CHECK-NEXT:    fcmp s2, #0.0
-; CHECK-NEXT:    fcvt s0, h0
-; CHECK-NEXT:    fmov s1, w8
-; CHECK-NEXT:    csel w8, wzr, w11, lt
-; CHECK-NEXT:    fcmp s2, s3
-; CHECK-NEXT:    fcvtzu w12, s0
+; CHECK-NEXT:    csel w8, wzr, w9, lt
+; CHECK-NEXT:    fcmp s1, s2
+; CHECK-NEXT:    mov h1, v0.h[2]
+; CHECK-NEXT:    fcvt s1, h1
+; CHECK-NEXT:    fcvtzu w9, s1
+; CHECK-NEXT:    csinv w1, w8, wzr, le
+; CHECK-NEXT:    fcmp s1, #0.0
+; CHECK-NEXT:    csel w8, wzr, w9, lt
+; CHECK-NEXT:    fcmp s1, s2
+; CHECK-NEXT:    mov h1, v0.h[3]
+; CHECK-NEXT:    fcvt s1, h1
+; CHECK-NEXT:    ext v0.16b, v0.16b, v0.16b, #8
+; CHECK-NEXT:    fcvtzu w9, s1
 ; CHECK-NEXT:    csinv w2, w8, wzr, le
-; CHECK-NEXT:    fcmp s0, #0.0
-; CHECK-NEXT:    mov v1.s[1], w5
-; CHECK-NEXT:    csel w8, wzr, w12, lt
-; CHECK-NEXT:    fcmp s0, s3
+; CHECK-NEXT:    fcmp s1, #0.0
+; CHECK-NEXT:    csel w8, wzr, w9, lt
+; CHECK-NEXT:    fcmp s1, s2
+; CHECK-NEXT:    fcvt s1, h0
+; CHECK-NEXT:    mov h0, v0.h[1]
+; CHECK-NEXT:    fcvtzu w9, s1
 ; CHECK-NEXT:    csinv w3, w8, wzr, le
-; CHECK-NEXT:    fmov w4, s1
+; CHECK-NEXT:    fcmp s1, #0.0
+; CHECK-NEXT:    fcvt s0, h0
+; CHECK-NEXT:    csel w8, wzr, w9, lt
+; CHECK-NEXT:    fcmp s1, s2
+; CHECK-NEXT:    fcvtzu w10, s0
+; CHECK-NEXT:    csinv w4, w8, wzr, le
+; CHECK-NEXT:    fcmp s0, #0.0
+; CHECK-NEXT:    csel w8, wzr, w10, lt
+; CHECK-NEXT:    fcmp s0, s2
+; CHECK-NEXT:    csinv w5, w8, wzr, le
 ; CHECK-NEXT:    ret
     %x = call <6 x i32> @llvm.fptoui.sat.v6f16.v6i32(<6 x half> %f)
     ret <6 x i32> %x
 }
 
 define <7 x i32> @test_unsigned_v7f16_v7i32(<7 x half> %f) {
 ; CHECK-LABEL: test_unsigned_v7f16_v7i32:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    ext v1.16b, v0.16b, v0.16b, #8
-; CHECK-NEXT:    mov h2, v1.h[1]
+; CHECK-NEXT:    fcvt s1, h0
 ; CHECK-NEXT:    mov w8, #1333788671
-; CHECK-NEXT:    fcvt s2, h2
-; CHECK-NEXT:    fmov s3, w8
-; CHECK-NEXT:    fcvtzu w8, s2
-; CHECK-NEXT:    fcmp s2, #0.0
-; CHECK-NEXT:    csel w8, wzr, w8, lt
-; CHECK-NEXT:    fcmp s2, s3
-; CHECK-NEXT:    fcvt s2, h1
-; CHECK-NEXT:    fcvtzu w9, s2
-; CHECK-NEXT:    csinv w8, w8, wzr, le
-; CHECK-NEXT:    fcmp s2, #0.0
-; CHECK-NEXT:    mov h1, v1.h[2]
-; CHECK-NEXT:    csel w9, wzr, w9, lt
-; CHECK-NEXT:    fcmp s2, s3
+; CHECK-NEXT:    fcvtzu w9, s1
+; CHECK-NEXT:    fcmp s1, #0.0
+; CHECK-NEXT:    fmov s2, w8
+; CHECK-NEXT:    csel w8, wzr, w9, lt
+; CHECK-NEXT:    fcmp s1, s2
+; CHECK-NEXT:    mov h1, v0.h[1]
 ; CHECK-NEXT:    fcvt s1, h1
-; CHECK-NEXT:    fcvtzu w10, s1
-; CHECK-NEXT:    csinv w9, w9, wzr, le
+; CHECK-NEXT:    fcvtzu w9, s1
+; CHECK-NEXT:    csinv w0, w8, wzr, le
 ; CHECK-NEXT:    fcmp s1, #0.0
-; CHECK-NEXT:    fcvt s2, h0
-; CHECK-NEXT:    csel w10, wzr, w10, lt
-; CHECK-NEXT:    fcmp s1, s3
+; CHECK-NEXT:    csel w8, wzr, w9, lt
+; CHECK-NEXT:    fcmp s1, s2
+; CHECK-NEXT:    mov h1, v0.h[2]
+; CHECK-NEXT:    fcvt s1, h1
+; CHECK-NEXT:    fcvtzu w9, s1
+; CHECK-NEXT:    csinv w1, w8, wzr, le
+; CHECK-NEXT:    fcmp s1, #0.0
+; CHECK-NEXT:    csel w8, wzr, w9, lt
+; CHECK-NEXT:    fcmp s1, s2
+; CHECK-NEXT:    mov h1, v0.h[3]
+; CHECK-NEXT:    fcvt s1, h1
+; CHECK-NEXT:    ext v0.16b, v0.16b, v0.16b, #8
+; CHECK-NEXT:    fcvtzu w9, s1
+; CHECK-NEXT:    csinv w2, w8, wzr, le
+; CHECK-NEXT:    fcmp s1, #0.0
+; CHECK-NEXT:    csel w8, wzr, w9, lt
+; CHECK-NEXT:    fcmp s1, s2
+; CHECK-NEXT:    fcvt s1, h0
+; CHECK-NEXT:    fcvtzu w9, s1
+; CHECK-NEXT:    csinv w3, w8, wzr, le
+; CHECK-NEXT:    fcmp s1, #0.0
+; CHECK-NEXT:    csel w8, wzr, w9, lt
+; CHECK-NEXT:    fcmp s1, s2
 ; CHECK-NEXT:    mov h1, v0.h[1]
-; CHECK-NEXT:    fcvtzu w11, s2
-; CHECK-NEXT:    csinv w6, w10, wzr, le
-; CHECK-NEXT:    fcmp s2, #0.0
-; CHECK-NEXT:    csel w10, wzr, w11, lt
-; CHECK-NEXT:    fcmp s2, s3
 ; CHECK-NEXT:    fcvt s1, h1
-; CHECK-NEXT:    mov h2, v0.h[2]
-; CHECK-NEXT:    fcvtzu w11, s1
-; CHECK-NEXT:    csinv w0, w10, wzr, le
+; CHECK-NEXT:    mov h0, v0.h[2]
+; CHECK-NEXT:    fcvtzu w9, s1
+; CHECK-NEXT:    csinv w4, w8, wzr, le
 ; CHECK-NEXT:    fcmp s1, #0.0
-; CHECK-NEXT:    fcvt s2, h2
-; CHECK-NEXT:    csel w10, wzr, w11, lt
-; CHECK-NEXT:    fcmp s1, s3
-; CHECK-NEXT:    mov h0, v0.h[3]
-; CHECK-NEXT:    fcvtzu w12, s2
-; CHECK-NEXT:    fmov s1, w9
-; CHECK-NEXT:    csinv w1, w10, wzr, le
-; CHECK-NEXT:    fcmp s2, #0.0
 ; CHECK-NEXT:    fcvt s0, h0
-; CHECK-NEXT:    mov v1.s[1], w8
-; CHECK-NEXT:    csel w8, wzr, w12, lt
-; CHECK-NEXT:    fcmp s2, s3
-; CHECK-NEXT:    fcvtzu w13, s0
-; CHECK-NEXT:    csinv w2, w8, wzr, le
+; CHECK-NEXT:    csel w8, wzr, w9, lt
+; CHECK-NEXT:    fcmp s1, s2
+; CHECK-NEXT:    fcvtzu w10, s0
+; CHECK-NEXT:    csinv w5, w8, wzr, le
 ; CHECK-NEXT:    fcmp s0, #0.0
-; CHECK-NEXT:    mov v1.s[2], w6
-; CHECK-NEXT:    csel w8, wzr, w13, lt
-; CHECK-NEXT:    fcmp s0, s3
-; CHECK-NEXT:    csinv w3, w8, wzr, le
-; CHECK-NEXT:    mov w5, v1.s[1]
-; CHECK-NEXT:    fmov w4, s1
+; CHECK-NEXT:    csel w8, wzr, w10, lt
+; CHECK-NEXT:    fcmp s0, s2
+; CHECK-NEXT:    csinv w6, w8, wzr, le
 ; CHECK-NEXT:    ret
     %x = call <7 x i32> @llvm.fptoui.sat.v7f16.v7i32(<7 x half> %f)
     ret <7 x i32> %x
 }
 
 define <8 x i32> @test_unsigned_v8f16_v8i32(<8 x half> %f) {
 ; CHECK-LABEL: test_unsigned_v8f16_v8i32:
 ; CHECK:       // %bb.0:

 ; CHECK-NEXT:    .cfi_def_cfa_offset 64
 ; CHECK-NEXT:    .cfi_offset w19, -8
 ; CHECK-NEXT:    .cfi_offset w20, -16
 ; CHECK-NEXT:    .cfi_offset w21, -24
 ; CHECK-NEXT:    .cfi_offset w30, -32
 ; CHECK-NEXT:    .cfi_offset b8, -40
 ; CHECK-NEXT:    .cfi_offset b9, -48
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $q0
-; CHECK-NEXT:    mov s8, v0.s[1]
 ; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
-; CHECK-NEXT:    mov v0.16b, v8.16b
+; CHECK-NEXT:    // kill: def $s0 killed $s0 killed $q0
 ; CHECK-NEXT:    bl __fixunssfti
+; CHECK-NEXT:    ldr q0, [sp] // 16-byte Folded Reload
 ; CHECK-NEXT:    mov w8, #1904214015
-; CHECK-NEXT:    fcmp s8, #0.0
 ; CHECK-NEXT:    fmov s9, w8
 ; CHECK-NEXT:    mov x21, #68719476735
+; CHECK-NEXT:    fcmp s0, #0.0
+; CHECK-NEXT:    mov s8, v0.s[1]
 ; CHECK-NEXT:    csel x9, xzr, x0, lt
 ; CHECK-NEXT:    csel x10, xzr, x1, lt
-; CHECK-NEXT:    fcmp s8, s9
-; CHECK-NEXT:    ldr q0, [sp] // 16-byte Folded Reload
+; CHECK-NEXT:    fcmp s0, s9
+; CHECK-NEXT:    mov v0.16b, v8.16b
 ; CHECK-NEXT:    csel x19, x21, x10, gt
 ; CHECK-NEXT:    csinv x20, x9, xzr, le
-; CHECK-NEXT:    // kill: def $s0 killed $s0 killed $q0
 ; CHECK-NEXT:    bl __fixunssfti
-; CHECK-NEXT:    ldr q0, [sp] // 16-byte Folded Reload
-; CHECK-NEXT:    mov x2, x20
-; CHECK-NEXT:    mov x3, x19
-; CHECK-NEXT:    ldp x20, x19, [sp, #48] // 16-byte Folded Reload
-; CHECK-NEXT:    fcmp s0, #0.0
+; CHECK-NEXT:    fcmp s8, #0.0
 ; CHECK-NEXT:    csel x8, xzr, x0, lt
 ; CHECK-NEXT:    csel x9, xzr, x1, lt
-; CHECK-NEXT:    fcmp s0, s9
-; CHECK-NEXT:    csinv x8, x8, xzr, le
-; CHECK-NEXT:    csel x1, x21, x9, gt
+; CHECK-NEXT:    fcmp s8, s9
+; CHECK-NEXT:    csel x3, x21, x9, gt
+; CHECK-NEXT:    mov x0, x20
+; CHECK-NEXT:    mov x1, x19
+; CHECK-NEXT:    ldp x20, x19, [sp, #48] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp x30, x21, [sp, #32] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d9, d8, [sp, #16] // 16-byte Folded Reload
-; CHECK-NEXT:    fmov d0, x8
-; CHECK-NEXT:    mov v0.d[1], x1
-; CHECK-NEXT:    fmov x0, d0
+; CHECK-NEXT:    csinv x2, x8, xzr, le
 ; CHECK-NEXT:    add sp, sp, #64 // =64
 ; CHECK-NEXT:    ret
     %x = call <2 x i100> @llvm.fptoui.sat.v2f32.v2i100(<2 x float> %f)
     ret <2 x i100> %x
 }
 
 define <2 x i128> @test_unsigned_v2f32_v2i128(<2 x float> %f) {
 ; CHECK-LABEL: test_unsigned_v2f32_v2i128:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    sub sp, sp, #64 // =64
 ; CHECK-NEXT:    stp d9, d8, [sp, #16] // 16-byte Folded Spill
 ; CHECK-NEXT:    str x30, [sp, #32] // 8-byte Folded Spill
 ; CHECK-NEXT:    stp x20, x19, [sp, #48] // 16-byte Folded Spill
 ; CHECK-NEXT:    .cfi_def_cfa_offset 64
 ; CHECK-NEXT:    .cfi_offset w19, -8
 ; CHECK-NEXT:    .cfi_offset w20, -16
 ; CHECK-NEXT:    .cfi_offset w30, -32
 ; CHECK-NEXT:    .cfi_offset b8, -40
 ; CHECK-NEXT:    .cfi_offset b9, -48
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $q0
-; CHECK-NEXT:    mov s8, v0.s[1]
 ; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
-; CHECK-NEXT:    mov v0.16b, v8.16b
+; CHECK-NEXT:    // kill: def $s0 killed $s0 killed $q0
 ; CHECK-NEXT:    bl __fixunssfti
+; CHECK-NEXT:    ldr q0, [sp] // 16-byte Folded Reload
 ; CHECK-NEXT:    mov w8, #2139095039
-; CHECK-NEXT:    fcmp s8, #0.0
 ; CHECK-NEXT:    fmov s9, w8
+; CHECK-NEXT:    fcmp s0, #0.0
+; CHECK-NEXT:    mov s8, v0.s[1]
 ; CHECK-NEXT:    csel x9, xzr, x1, lt
 ; CHECK-NEXT:    csel x10, xzr, x0, lt
-; CHECK-NEXT:    fcmp s8, s9
-; CHECK-NEXT:    ldr q0, [sp] // 16-byte Folded Reload
+; CHECK-NEXT:    fcmp s0, s9
+; CHECK-NEXT:    mov v0.16b, v8.16b
 ; CHECK-NEXT:    csinv x19, x10, xzr, le
 ; CHECK-NEXT:    csinv x20, x9, xzr, le
-; CHECK-NEXT:    // kill: def $s0 killed $s0 killed $q0
 ; CHECK-NEXT:    bl __fixunssfti
-; CHECK-NEXT:    ldr q0, [sp] // 16-byte Folded Reload
-; CHECK-NEXT:    mov x2, x19
-; CHECK-NEXT:    mov x3, x20
+; CHECK-NEXT:    fcmp s8, #0.0
+; CHECK-NEXT:    csel x8, xzr, x1, lt
+; CHECK-NEXT:    csel x9, xzr, x0, lt
+; CHECK-NEXT:    fcmp s8, s9
+; CHECK-NEXT:    mov x0, x19
+; CHECK-NEXT:    mov x1, x20
 ; CHECK-NEXT:    ldp x20, x19, [sp, #48] // 16-byte Folded Reload
-; CHECK-NEXT:    fcmp s0, #0.0
-; CHECK-NEXT:    csel x8, xzr, x0, lt
-; CHECK-NEXT:    csel x9, xzr, x1, lt
-; CHECK-NEXT:    fcmp s0, s9
-; CHECK-NEXT:    csinv x8, x8, xzr, le
 ; CHECK-NEXT:    ldr x30, [sp, #32] // 8-byte Folded Reload
 ; CHECK-NEXT:    ldp d9, d8, [sp, #16] // 16-byte Folded Reload
-; CHECK-NEXT:    csinv x1, x9, xzr, le
-; CHECK-NEXT:    fmov d0, x8
-; CHECK-NEXT:    mov v0.d[1], x1
-; CHECK-NEXT:    fmov x0, d0
+; CHECK-NEXT:    csinv x2, x9, xzr, le
+; CHECK-NEXT:    csinv x3, x8, xzr, le
 ; CHECK-NEXT:    add sp, sp, #64 // =64
 ; CHECK-NEXT:    ret
     %x = call <2 x i128> @llvm.fptoui.sat.v2f32.v2i128(<2 x float> %f)
     ret <2 x i128> %x
 }
 
 ;
 ; 2-Vector double to unsigned integer -- result size variation

 ; CHECK-NEXT:    stp x20, x19, [sp, #48] // 16-byte Folded Spill
 ; CHECK-NEXT:    .cfi_def_cfa_offset 64
 ; CHECK-NEXT:    .cfi_offset w19, -8
 ; CHECK-NEXT:    .cfi_offset w20, -16
 ; CHECK-NEXT:    .cfi_offset w21, -24
 ; CHECK-NEXT:    .cfi_offset w30, -32
 ; CHECK-NEXT:    .cfi_offset b8, -40
 ; CHECK-NEXT:    .cfi_offset b9, -48
-; CHECK-NEXT:    mov d8, v0.d[1]
 ; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
-; CHECK-NEXT:    mov v0.16b, v8.16b
+; CHECK-NEXT:    // kill: def $d0 killed $d0 killed $q0
 ; CHECK-NEXT:    bl __fixunsdfti
+; CHECK-NEXT:    ldr q0, [sp] // 16-byte Folded Reload
 ; CHECK-NEXT:    mov x8, #5057542381537067007
-; CHECK-NEXT:    fcmp d8, #0.0
 ; CHECK-NEXT:    fmov d9, x8
 ; CHECK-NEXT:    mov x21, #68719476735
+; CHECK-NEXT:    fcmp d0, #0.0
+; CHECK-NEXT:    mov d8, v0.d[1]
 ; CHECK-NEXT:    csel x9, xzr, x0, lt
 ; CHECK-NEXT:    csel x10, xzr, x1, lt
-; CHECK-NEXT:    fcmp d8, d9
-; CHECK-NEXT:    ldr q0, [sp] // 16-byte Folded Reload
+; CHECK-NEXT:    fcmp d0, d9
+; CHECK-NEXT:    mov v0.16b, v8.16b
 ; CHECK-NEXT:    csel x19, x21, x10, gt
 ; CHECK-NEXT:    csinv x20, x9, xzr, le
-; CHECK-NEXT:    // kill: def $d0 killed $d0 killed $q0
 ; CHECK-NEXT:    bl __fixunsdfti
-; CHECK-NEXT:    ldr q0, [sp] // 16-byte Folded Reload
-; CHECK-NEXT:    mov x2, x20
-; CHECK-NEXT:    mov x3, x19
-; CHECK-NEXT:    ldp x20, x19, [sp, #48] // 16-byte Folded Reload
-; CHECK-NEXT:    fcmp d0, #0.0
+; CHECK-NEXT:    fcmp d8, #0.0
 ; CHECK-NEXT:    csel x8, xzr, x0, lt
 ; CHECK-NEXT:    csel x9, xzr, x1, lt
-; CHECK-NEXT:    fcmp d0, d9
-; CHECK-NEXT:    csinv x8, x8, xzr, le
-; CHECK-NEXT:    csel x1, x21, x9, gt
+; CHECK-NEXT:    fcmp d8, d9
+; CHECK-NEXT:    csel x3, x21, x9, gt
+; CHECK-NEXT:    mov x0, x20
+; CHECK-NEXT:    mov x1, x19
+; CHECK-NEXT:    ldp x20, x19, [sp, #48] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp x30, x21, [sp, #32] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d9, d8, [sp, #16] // 16-byte Folded Reload
-; CHECK-NEXT:    fmov d0, x8
-; CHECK-NEXT:    mov v0.d[1], x1
-; CHECK-NEXT:    fmov x0, d0
+; CHECK-NEXT:    csinv x2, x8, xzr, le
 ; CHECK-NEXT:    add sp, sp, #64 // =64
 ; CHECK-NEXT:    ret
     %x = call <2 x i100> @llvm.fptoui.sat.v2f64.v2i100(<2 x double> %f)
     ret <2 x i100> %x
 }
 
 define <2 x i128> @test_unsigned_v2f64_v2i128(<2 x double> %f) {
 ; CHECK-LABEL: test_unsigned_v2f64_v2i128:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    sub sp, sp, #64 // =64
 ; CHECK-NEXT:    stp d9, d8, [sp, #16] // 16-byte Folded Spill
 ; CHECK-NEXT:    str x30, [sp, #32] // 8-byte Folded Spill
 ; CHECK-NEXT:    stp x20, x19, [sp, #48] // 16-byte Folded Spill
 ; CHECK-NEXT:    .cfi_def_cfa_offset 64
 ; CHECK-NEXT:    .cfi_offset w19, -8
 ; CHECK-NEXT:    .cfi_offset w20, -16
 ; CHECK-NEXT:    .cfi_offset w30, -32
 ; CHECK-NEXT:    .cfi_offset b8, -40
 ; CHECK-NEXT:    .cfi_offset b9, -48
-; CHECK-NEXT:    mov d8, v0.d[1]
 ; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
-; CHECK-NEXT:    mov v0.16b, v8.16b
+; CHECK-NEXT:    // kill: def $d0 killed $d0 killed $q0
 ; CHECK-NEXT:    bl __fixunsdfti
+; CHECK-NEXT:    ldr q0, [sp] // 16-byte Folded Reload
 ; CHECK-NEXT:    mov x8, #5183643171103440895
-; CHECK-NEXT:    fcmp d8, #0.0
 ; CHECK-NEXT:    fmov d9, x8
+; CHECK-NEXT:    fcmp d0, #0.0
+; CHECK-NEXT:    mov d8, v0.d[1]
 ; CHECK-NEXT:    csel x9, xzr, x1, lt
 ; CHECK-NEXT:    csel x10, xzr, x0, lt
-; CHECK-NEXT:    fcmp d8, d9
-; CHECK-NEXT:    ldr q0, [sp] // 16-byte Folded Reload
+; CHECK-NEXT:    fcmp d0, d9
+; CHECK-NEXT:    mov v0.16b, v8.16b
 ; CHECK-NEXT:    csinv x19, x10, xzr, le
 ; CHECK-NEXT:    csinv x20, x9, xzr, le
-; CHECK-NEXT:    // kill: def $d0 killed $d0 killed $q0
 ; CHECK-NEXT:    bl __fixunsdfti
-; CHECK-NEXT:    ldr q0, [sp] // 16-byte Folded Reload
-; CHECK-NEXT:    mov x2, x19
-; CHECK-NEXT:    mov x3, x20
+; CHECK-NEXT:    fcmp d8, #0.0
+; CHECK-NEXT:    csel x8, xzr, x1, lt
+; CHECK-NEXT:    csel x9, xzr, x0, lt
+; CHECK-NEXT:    fcmp d8, d9
+; CHECK-NEXT:    mov x0, x19
+; CHECK-NEXT:    mov x1, x20
 ; CHECK-NEXT:    ldp x20, x19, [sp, #48] // 16-byte Folded Reload
-; CHECK-NEXT:    fcmp d0, #0.0
-; CHECK-NEXT:    csel x8, xzr, x0, lt
-; CHECK-NEXT:    csel x9, xzr, x1, lt
-; CHECK-NEXT:    fcmp d0, d9
-; CHECK-NEXT:    csinv x8, x8, xzr, le
 ; CHECK-NEXT:    ldr x30, [sp, #32] // 8-byte Folded Reload
 ; CHECK-NEXT:    ldp d9, d8, [sp, #16] // 16-byte Folded Reload
-; CHECK-NEXT:    csinv x1, x9, xzr, le
-; CHECK-NEXT:    fmov d0, x8
-; CHECK-NEXT:    mov v0.d[1], x1
-; CHECK-NEXT:    fmov x0, d0
+; CHECK-NEXT:    csinv x2, x9, xzr, le
+; CHECK-NEXT:    csinv x3, x8, xzr, le
 ; CHECK-NEXT:    add sp, sp, #64 // =64
 ; CHECK-NEXT:    ret
     %x = call <2 x i128> @llvm.fptoui.sat.v2f64.v2i128(<2 x double> %f)
     ret <2 x i128> %x
 }
 
 ;
 ; 4-Vector half to unsigned integer -- result size variation

 ; CHECK-NEXT:    .cfi_offset w22, -32
 ; CHECK-NEXT:    .cfi_offset w23, -40
 ; CHECK-NEXT:    .cfi_offset w24, -48
 ; CHECK-NEXT:    .cfi_offset w25, -56
 ; CHECK-NEXT:    .cfi_offset w30, -64
 ; CHECK-NEXT:    .cfi_offset b8, -72
 ; CHECK-NEXT:    .cfi_offset b9, -80
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $q0
-; CHECK-NEXT:    mov h1, v0.h[2]
+; CHECK-NEXT:    mov h1, v0.h[1]
 ; CHECK-NEXT:    fcvt s8, h1
 ; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
 ; CHECK-NEXT:    mov v0.16b, v8.16b
 ; CHECK-NEXT:    bl __fixunssfti
 ; CHECK-NEXT:    ldr q0, [sp] // 16-byte Folded Reload
 ; CHECK-NEXT:    mov w8, #1904214015
 ; CHECK-NEXT:    fcmp s8, #0.0
 ; CHECK-NEXT:    fmov s9, w8
-; CHECK-NEXT:    mov h0, v0.h[1]
+; CHECK-NEXT:    mov h0, v0.h[2]
 ; CHECK-NEXT:    csel x9, xzr, x0, lt
 ; CHECK-NEXT:    csel x10, xzr, x1, lt
 ; CHECK-NEXT:    fcmp s8, s9
 ; CHECK-NEXT:    fcvt s8, h0
 ; CHECK-NEXT:    mov x25, #68719476735
 ; CHECK-NEXT:    mov v0.16b, v8.16b
 ; CHECK-NEXT:    csel x19, x25, x10, gt
 ; CHECK-NEXT:    csinv x20, x9, xzr, le
 ; CHECK-NEXT:    bl __fixunssfti
 ; CHECK-NEXT:    ldr q0, [sp] // 16-byte Folded Reload
 ; CHECK-NEXT:    fcmp s8, #0.0
 ; CHECK-NEXT:    csel x8, xzr, x0, lt
 ; CHECK-NEXT:    csel x9, xzr, x1, lt
-; CHECK-NEXT:    mov h0, v0.h[3]
 ; CHECK-NEXT:    fcmp s8, s9
 ; CHECK-NEXT:    fcvt s8, h0
 ; CHECK-NEXT:    mov v0.16b, v8.16b
 ; CHECK-NEXT:    csel x21, x25, x9, gt
 ; CHECK-NEXT:    csinv x22, x8, xzr, le
 ; CHECK-NEXT:    bl __fixunssfti
 ; CHECK-NEXT:    ldr q0, [sp] // 16-byte Folded Reload
 ; CHECK-NEXT:    fcmp s8, #0.0
 ; CHECK-NEXT:    csel x8, xzr, x0, lt
 ; CHECK-NEXT:    csel x9, xzr, x1, lt
+; CHECK-NEXT:    mov h0, v0.h[3]
 ; CHECK-NEXT:    fcmp s8, s9
 ; CHECK-NEXT:    fcvt s8, h0
 ; CHECK-NEXT:    mov v0.16b, v8.16b
 ; CHECK-NEXT:    csel x23, x25, x9, gt
 ; CHECK-NEXT:    csinv x24, x8, xzr, le
 ; CHECK-NEXT:    bl __fixunssfti
 ; CHECK-NEXT:    fcmp s8, #0.0
 ; CHECK-NEXT:    csel x8, xzr, x0, lt
 ; CHECK-NEXT:    csel x9, xzr, x1, lt
 ; CHECK-NEXT:    fcmp s8, s9
-; CHECK-NEXT:    csinv x8, x8, xzr, le
-; CHECK-NEXT:    csel x1, x25, x9, gt
-; CHECK-NEXT:    mov x2, x22
-; CHECK-NEXT:    mov x3, x21
-; CHECK-NEXT:    mov x4, x20
-; CHECK-NEXT:    mov x5, x19
-; CHECK-NEXT:    mov x6, x24
-; CHECK-NEXT:    mov x7, x23
+; CHECK-NEXT:    csel x7, x25, x9, gt
+; CHECK-NEXT:    mov x0, x24
+; CHECK-NEXT:    mov x1, x23
+; CHECK-NEXT:    mov x2, x20
+; CHECK-NEXT:    mov x3, x19
+; CHECK-NEXT:    mov x4, x22
+; CHECK-NEXT:    mov x5, x21
 ; CHECK-NEXT:    ldp x20, x19, [sp, #80] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp x22, x21, [sp, #64] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp x24, x23, [sp, #48] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp x30, x25, [sp, #32] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp d9, d8, [sp, #16] // 16-byte Folded Reload
-; CHECK-NEXT:    fmov d0, x8
-; CHECK-NEXT:    mov v0.d[1], x1
-; CHECK-NEXT:    fmov x0, d0
+; CHECK-NEXT:    csinv x6, x8, xzr, le
 ; CHECK-NEXT:    add sp, sp, #96 // =96
 ; CHECK-NEXT:    ret
     %x = call <4 x i100> @llvm.fptoui.sat.v4f16.v4i100(<4 x half> %f)
     ret <4 x i100> %x
 }
 
 define <4 x i128> @test_unsigned_v4f16_v4i128(<4 x half> %f) {
 ; CHECK-LABEL: test_unsigned_v4f16_v4i128:

 ; CHECK-NEXT:    .cfi_offset w21, -24
 ; CHECK-NEXT:    .cfi_offset w22, -32
 ; CHECK-NEXT:    .cfi_offset w23, -40
 ; CHECK-NEXT:    .cfi_offset w24, -48
 ; CHECK-NEXT:    .cfi_offset w30, -64
 ; CHECK-NEXT:    .cfi_offset b8, -72
 ; CHECK-NEXT:    .cfi_offset b9, -80
 ; CHECK-NEXT:    // kill: def $d0 killed $d0 def $q0
-; CHECK-NEXT:    mov h1, v0.h[1]
-; CHECK-NEXT:    fcvt s8, h1
+; CHECK-NEXT:    fcvt s8, h0
 ; CHECK-NEXT:    str q0, [sp] // 16-byte Folded Spill
 ; CHECK-NEXT:    mov v0.16b, v8.16b
 ; CHECK-NEXT:    bl __fixunssfti
 ; CHECK-NEXT:    ldr q0, [sp] // 16-byte Folded Reload
 ; CHECK-NEXT:    mov w8, #2139095039
 ; CHECK-NEXT:    fcmp s8, #0.0
 ; CHECK-NEXT:    fmov s9, w8
-; CHECK-NEXT:    mov h0, v0.h[2]
+; CHECK-NEXT:    mov h0, v0.h[1]
 ; CHECK-NEXT:    csel x9, xzr, x1, lt
 ; CHECK-NEXT:    csel x10, xzr, x0, lt
 ; CHECK-NEXT:    fcmp s8, s9
 ; CHECK-NEXT:    fcvt s8, h0
 ; CHECK-NEXT:    mov v0.16b, v8.16b
 ; CHECK-NEXT:    csinv x19, x10, xzr, le
 ; CHECK-NEXT:    csinv x20, x9, xzr, le
 ; CHECK-NEXT:    bl __fixunssfti
 ; CHECK-NEXT:    ldr q0, [sp] // 16-byte Folded Reload
 ; CHECK-NEXT:    fcmp s8, #0.0
 ; CHECK-NEXT:    csel x8, xzr, x1, lt
 ; CHECK-NEXT:    csel x9, xzr, x0, lt
-; CHECK-NEXT:    mov h0, v0.h[3]
+; CHECK-NEXT:    mov h0, v0.h[2]
 ; CHECK-NEXT:    fcmp s8, s9
 ; CHECK-NEXT:    fcvt s8, h0
 ; CHECK-NEXT:    mov v0.16b, v8.16b
 ; CHECK-NEXT:    csinv x21, x9, xzr, le
 ; CHECK-NEXT:    csinv x22, x8, xzr, le
 ; CHECK-NEXT:    bl __fixunssfti
 ; CHECK-NEXT:    ldr q0, [sp] // 16-byte Folded Reload
 ; CHECK-NEXT:    fcmp s8, #0.0
 ; CHECK-NEXT:    csel x8, xzr, x1, lt
 ; CHECK-NEXT:    csel x9, xzr, x0, lt
+; CHECK-NEXT:    mov h0, v0.h[3]
 ; CHECK-NEXT:    fcmp s8, s9
 ; CHECK-NEXT:    fcvt s8, h0
 ; CHECK-NEXT:    mov v0.16b, v8.16b
 ; CHECK-NEXT:    csinv x23, x9, xzr, le
 ; CHECK-NEXT:    csinv x24, x8, xzr, le
 ; CHECK-NEXT:    bl __fixunssfti
 ; CHECK-NEXT:    fcmp s8, #0.0
-; CHECK-NEXT:    csel x8, xzr, x0, lt
-; CHECK-NEXT:    csel x9, xzr, x1, lt
+; CHECK-NEXT:    csel x8, xzr, x1, lt
+; CHECK-NEXT:    csel x9, xzr, x0, lt
 ; CHECK-NEXT:    fcmp s8, s9
-; CHECK-NEXT:    csinv x8, x8, xzr, le
-; CHECK-NEXT:    mov x2, x19
-; CHECK-NEXT:    mov x3, x20
-; CHECK-NEXT:    mov x4, x21
-; CHECK-NEXT:    mov x5, x22
-; CHECK-NEXT:    mov x6, x23
-; CHECK-NEXT:    mov x7, x24
+; CHECK-NEXT:    mov x0, x19
+; CHECK-NEXT:    mov x1, x20
+; CHECK-NEXT:    mov x2, x21
+; CHECK-NEXT:    mov x3, x22
+; CHECK-NEXT:    mov x4, x23
+; CHECK-NEXT:    mov x5, x24
 ; CHECK-NEXT:    ldp x20, x19, [sp, #80] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp x22, x21, [sp, #64] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldp x24, x23, [sp, #48] // 16-byte Folded Reload
 ; CHECK-NEXT:    ldr x30, [sp, #32] // 8-byte Folded Reload
 ; CHECK-NEXT:    ldp d9, d8, [sp, #16] // 16-byte Folded Reload
-; CHECK-NEXT:    csinv x1, x9, xzr, le
-; CHECK-NEXT:    fmov d0, x8
-; CHECK-NEXT:    mov v0.d[1], x1
-; CHECK-NEXT:    fmov x0, d0
+; CHECK-NEXT:    csinv x6, x9, xzr, le
+; CHECK-NEXT:    csinv x7, x8, xzr, le
 ; CHECK-NEXT:    add sp, sp, #96 // =96
 ; CHECK-NEXT:    ret
     %x = call <4 x i128> @llvm.fptoui.sat.v4f16.v4i128(<4 x half> %f)
     ret <4 x i128> %x
 }
 

llvm/test/CodeGen/AArch64/popcount.ll

 ; CHECK-NEXT:    fmov d0, x0
 ; CHECK-NEXT:    mov v0.d[1], x1
 ; CHECK-NEXT:    cnt v0.16b, v0.16b
 ; CHECK-NEXT:    uaddlv h1, v0.16b
 ; CHECK-NEXT:    // implicit-def: $q0
 ; CHECK-NEXT:    mov v0.16b, v1.16b
 ; CHECK-NEXT:    fmov w0, s0
 ; CHECK-NEXT:    // kill: def $x0 killed $w0
-; CHECK-NEXT:    movi v0.2d, #0000000000000000
-; CHECK-NEXT:    mov x1, v0.d[1]
+; CHECK-NEXT:    mov x1, xzr
 ; CHECK-NEXT:    ret
 Entry:
   %1 = tail call <1 x i128> @llvm.ctpop.v1.i128(<1 x i128> %0)
   ret <1 x i128> %1
 }
 
 declare <1 x i128> @llvm.ctpop.v1.i128(<1 x i128>)

llvm/test/CodeGen/AArch64/sadd_sat_vec.ll

 ; CHECK-NEXT:    ret
   %z = call <8 x i64> @llvm.sadd.sat.v8i64(<8 x i64> %x, <8 x i64> %y)
   ret <8 x i64> %z
 }
 
 define <2 x i128> @v2i128(<2 x i128> %x, <2 x i128> %y) nounwind {
 ; CHECK-LABEL: v2i128:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    adds x8, x2, x6
-; CHECK-NEXT:    adcs x12, x3, x7
+; CHECK-NEXT:    adds x8, x0, x4
+; CHECK-NEXT:    adcs x12, x1, x5
 ; CHECK-NEXT:    mov x9, #9223372036854775807
-; CHECK-NEXT:    eor x10, x3, x7
+; CHECK-NEXT:    eor x10, x1, x5
 ; CHECK-NEXT:    cmp x12, #0 // =0
-; CHECK-NEXT:    eor x13, x3, x12
+; CHECK-NEXT:    eor x13, x1, x12
 ; CHECK-NEXT:    cinv x14, x9, ge
 ; CHECK-NEXT:    bics xzr, x13, x10
 ; CHECK-NEXT:    asr x10, x12, #63
-; CHECK-NEXT:    csel x2, x10, x8, lt
-; CHECK-NEXT:    csel x3, x14, x12, lt
-; CHECK-NEXT:    adds x8, x0, x4
-; CHECK-NEXT:    adcs x10, x1, x5
-; CHECK-NEXT:    eor x11, x1, x5
+; CHECK-NEXT:    csel x0, x10, x8, lt
+; CHECK-NEXT:    csel x1, x14, x12, lt
+; CHECK-NEXT:    adds x8, x2, x6
+; CHECK-NEXT:    adcs x10, x3, x7
+; CHECK-NEXT:    eor x11, x3, x7
 ; CHECK-NEXT:    cmp x10, #0 // =0
-; CHECK-NEXT:    eor x12, x1, x10
+; CHECK-NEXT:    eor x13, x3, x10
+; CHECK-NEXT:    asr x12, x10, #63
 ; CHECK-NEXT:    cinv x9, x9, ge
-; CHECK-NEXT:    bics xzr, x12, x11
-; CHECK-NEXT:    asr x11, x10, #63
-; CHECK-NEXT:    csel x8, x11, x8, lt
-; CHECK-NEXT:    csel x1, x9, x10, lt
-; CHECK-NEXT:    fmov d0, x8
-; CHECK-NEXT:    mov v0.d[1], x1
-; CHECK-NEXT:    fmov x0, d0
+; CHECK-NEXT:    bics xzr, x13, x11
+; CHECK-NEXT:    csel x2, x12, x8, lt
+; CHECK-NEXT:    csel x3, x9, x10, lt
 ; CHECK-NEXT:    ret
   %z = call <2 x i128> @llvm.sadd.sat.v2i128(<2 x i128> %x, <2 x i128> %y)
   ret <2 x i128> %z
 }

llvm/test/CodeGen/AArch64/ssub_sat_vec.ll

 ; CHECK-NEXT:    ret
   %z = call <8 x i64> @llvm.ssub.sat.v8i64(<8 x i64> %x, <8 x i64> %y)
   ret <8 x i64> %z
 }
 
 define <2 x i128> @v2i128(<2 x i128> %x, <2 x i128> %y) nounwind {
 ; CHECK-LABEL: v2i128:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    subs x8, x2, x6
-; CHECK-NEXT:    sbcs x12, x3, x7
+; CHECK-NEXT:    subs x8, x0, x4
+; CHECK-NEXT:    sbcs x12, x1, x5
 ; CHECK-NEXT:    mov x9, #9223372036854775807
-; CHECK-NEXT:    eor x10, x3, x7
+; CHECK-NEXT:    eor x10, x1, x5
 ; CHECK-NEXT:    cmp x12, #0 // =0
-; CHECK-NEXT:    eor x13, x3, x12
+; CHECK-NEXT:    eor x13, x1, x12
 ; CHECK-NEXT:    cinv x14, x9, ge
 ; CHECK-NEXT:    tst x10, x13
 ; CHECK-NEXT:    asr x10, x12, #63
-; CHECK-NEXT:    csel x2, x10, x8, lt
-; CHECK-NEXT:    csel x3, x14, x12, lt
-; CHECK-NEXT:    subs x8, x0, x4
-; CHECK-NEXT:    sbcs x10, x1, x5
-; CHECK-NEXT:    eor x11, x1, x5
+; CHECK-NEXT:    csel x0, x10, x8, lt
+; CHECK-NEXT:    csel x1, x14, x12, lt
+; CHECK-NEXT:    subs x8, x2, x6
+; CHECK-NEXT:    sbcs x10, x3, x7
+; CHECK-NEXT:    eor x11, x3, x7
 ; CHECK-NEXT:    cmp x10, #0 // =0
-; CHECK-NEXT:    eor x12, x1, x10
+; CHECK-NEXT:    eor x13, x3, x10
+; CHECK-NEXT:    asr x12, x10, #63
 ; CHECK-NEXT:    cinv x9, x9, ge
-; CHECK-NEXT:    tst x11, x12
-; CHECK-NEXT:    asr x11, x10, #63
-; CHECK-NEXT:    csel x8, x11, x8, lt
-; CHECK-NEXT:    csel x1, x9, x10, lt
-; CHECK-NEXT:    fmov d0, x8
-; CHECK-NEXT:    mov v0.d[1], x1
-; CHECK-NEXT:    fmov x0, d0
+; CHECK-NEXT:    tst x11, x13
+; CHECK-NEXT:    csel x2, x12, x8, lt
+; CHECK-NEXT:    csel x3, x9, x10, lt
 ; CHECK-NEXT:    ret
   %z = call <2 x i128> @llvm.ssub.sat.v2i128(<2 x i128> %x, <2 x i128> %y)
   ret <2 x i128> %z
 }

llvm/test/CodeGen/AArch64/uadd_sat_vec.ll

 ; CHECK-NEXT:    ret
   %z = call <8 x i64> @llvm.uadd.sat.v8i64(<8 x i64> %x, <8 x i64> %y)
   ret <8 x i64> %z
 }
 
 define <2 x i128> @v2i128(<2 x i128> %x, <2 x i128> %y) nounwind {
 ; CHECK-LABEL: v2i128:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    adds x8, x2, x6
-; CHECK-NEXT:    adcs x9, x3, x7
-; CHECK-NEXT:    cmp x8, x2
-; CHECK-NEXT:    cset w10, lo
-; CHECK-NEXT:    cmp x9, x3
-; CHECK-NEXT:    cset w11, lo
-; CHECK-NEXT:    csel w10, w10, w11, eq
-; CHECK-NEXT:    cmp w10, #0 // =0
-; CHECK-NEXT:    csinv x3, x9, xzr, eq
-; CHECK-NEXT:    csinv x2, x8, xzr, eq
 ; CHECK-NEXT:    adds x8, x0, x4
 ; CHECK-NEXT:    adcs x9, x1, x5
 ; CHECK-NEXT:    cmp x8, x0
 ; CHECK-NEXT:    cset w10, lo
 ; CHECK-NEXT:    cmp x9, x1
 ; CHECK-NEXT:    cset w11, lo
 ; CHECK-NEXT:    csel w10, w10, w11, eq
 ; CHECK-NEXT:    cmp w10, #0 // =0
-; CHECK-NEXT:    csinv x8, x8, xzr, eq
 ; CHECK-NEXT:    csinv x1, x9, xzr, eq
-; CHECK-NEXT:    fmov d0, x8
-; CHECK-NEXT:    mov v0.d[1], x1
-; CHECK-NEXT:    fmov x0, d0
+; CHECK-NEXT:    csinv x0, x8, xzr, eq
+; CHECK-NEXT:    adds x8, x2, x6
+; CHECK-NEXT:    adcs x9, x3, x7
+; CHECK-NEXT:    cmp x8, x2
+; CHECK-NEXT:    cset w10, lo
+; CHECK-NEXT:    cmp x9, x3
+; CHECK-NEXT:    cset w11, lo
+; CHECK-NEXT:    csel w10, w10, w11, eq
+; CHECK-NEXT:    cmp w10, #0 // =0
+; CHECK-NEXT:    csinv x2, x8, xzr, eq
+; CHECK-NEXT:    csinv x3, x9, xzr, eq
 ; CHECK-NEXT:    ret
   %z = call <2 x i128> @llvm.uadd.sat.v2i128(<2 x i128> %x, <2 x i128> %y)
   ret <2 x i128> %z
 }

llvm/test/CodeGen/AArch64/usub_sat_vec.ll

 ; CHECK-NEXT:    ret
   %z = call <8 x i64> @llvm.usub.sat.v8i64(<8 x i64> %x, <8 x i64> %y)
   ret <8 x i64> %z
 }
 
 define <2 x i128> @v2i128(<2 x i128> %x, <2 x i128> %y) nounwind {
 ; CHECK-LABEL: v2i128:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    subs x8, x2, x6
-; CHECK-NEXT:    sbcs x9, x3, x7
-; CHECK-NEXT:    cmp x8, x2
-; CHECK-NEXT:    cset w10, hi
-; CHECK-NEXT:    cmp x9, x3
-; CHECK-NEXT:    cset w11, hi
-; CHECK-NEXT:    csel w10, w10, w11, eq
-; CHECK-NEXT:    cmp w10, #0 // =0
-; CHECK-NEXT:    csel x3, xzr, x9, ne
-; CHECK-NEXT:    csel x2, xzr, x8, ne
 ; CHECK-NEXT:    subs x8, x0, x4
 ; CHECK-NEXT:    sbcs x9, x1, x5
 ; CHECK-NEXT:    cmp x8, x0
 ; CHECK-NEXT:    cset w10, hi
 ; CHECK-NEXT:    cmp x9, x1
 ; CHECK-NEXT:    cset w11, hi
 ; CHECK-NEXT:    csel w10, w10, w11, eq
 ; CHECK-NEXT:    cmp w10, #0 // =0
-; CHECK-NEXT:    csel x8, xzr, x8, ne
 ; CHECK-NEXT:    csel x1, xzr, x9, ne
-; CHECK-NEXT:    fmov d0, x8
-; CHECK-NEXT:    mov v0.d[1], x1
-; CHECK-NEXT:    fmov x0, d0
+; CHECK-NEXT:    csel x0, xzr, x8, ne
+; CHECK-NEXT:    subs x8, x2, x6
+; CHECK-NEXT:    sbcs x9, x3, x7
+; CHECK-NEXT:    cmp x8, x2
+; CHECK-NEXT:    cset w10, hi
+; CHECK-NEXT:    cmp x9, x3
+; CHECK-NEXT:    cset w11, hi
+; CHECK-NEXT:    csel w10, w10, w11, eq
+; CHECK-NEXT:    cmp w10, #0 // =0
+; CHECK-NEXT:    csel x2, xzr, x8, ne
+; CHECK-NEXT:    csel x3, xzr, x9, ne
 ; CHECK-NEXT:    ret
   %z = call <2 x i128> @llvm.usub.sat.v2i128(<2 x i128> %x, <2 x i128> %y)
   ret <2 x i128> %z
 }

llvm/test/CodeGen/AArch64/vecreduce-and-legalization.ll

 ; CHECK-NEXT:    ret
   %b = call i8 @llvm.vector.reduce.and.v3i8(<3 x i8> %a)
   ret i8 %b
 }
 
 define i8 @test_v9i8(<9 x i8> %a) nounwind {
 ; CHECK-LABEL: test_v9i8:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    mov w8, #-1
-; CHECK-NEXT:    mov v1.16b, v0.16b
-; CHECK-NEXT:    mov v1.b[9], w8
-; CHECK-NEXT:    mov v1.b[10], w8
-; CHECK-NEXT:    mov v1.b[11], w8
-; CHECK-NEXT:    mov v1.b[13], w8
-; CHECK-NEXT:    ext v1.16b, v1.16b, v1.16b, #8
+; CHECK-NEXT:    ext v1.16b, v0.16b, v0.16b, #8
 ; CHECK-NEXT:    and v1.8b, v0.8b, v1.8b
-; CHECK-NEXT:    umov w8, v1.b[1]
+; CHECK-NEXT:    umov w8, v0.b[1]
 ; CHECK-NEXT:    umov w9, v1.b[0]
 ; CHECK-NEXT:    and w8, w9, w8
-; CHECK-NEXT:    umov w9, v1.b[2]
+; CHECK-NEXT:    umov w9, v0.b[2]
 ; CHECK-NEXT:    and w8, w8, w9
-; CHECK-NEXT:    umov w9, v1.b[3]
+; CHECK-NEXT:    umov w9, v0.b[3]
 ; CHECK-NEXT:    and w8, w8, w9
 ; CHECK-NEXT:    umov w9, v0.b[4]
 ; CHECK-NEXT:    and w8, w8, w9
-; CHECK-NEXT:    umov w9, v1.b[5]
+; CHECK-NEXT:    umov w9, v0.b[5]
 ; CHECK-NEXT:    and w8, w8, w9
 ; CHECK-NEXT:    umov w9, v0.b[6]
 ; CHECK-NEXT:    and w8, w8, w9
 ; CHECK-NEXT:    umov w9, v0.b[7]
 ; CHECK-NEXT:    and w0, w8, w9
 ; CHECK-NEXT:    ret
   %b = call i8 @llvm.vector.reduce.and.v9i8(<9 x i8> %a)
   ret i8 %b

llvm/test/CodeGen/AMDGPU/copy-illegal-type.ll

 ; VI-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
 ; VI-NEXT:    s_mov_b32 s3, 0xf000
 ; VI-NEXT:    s_mov_b32 s2, -1
 ; VI-NEXT:    s_mov_b32 s10, s2
 ; VI-NEXT:    s_mov_b32 s11, s3
 ; VI-NEXT:    s_waitcnt lgkmcnt(0)
 ; VI-NEXT:    s_mov_b32 s8, s6
 ; VI-NEXT:    s_mov_b32 s9, s7
-; VI-NEXT:    buffer_load_ubyte v0, off, s[8:11], 0
-; VI-NEXT:    buffer_load_ubyte v1, off, s[8:11], 0 offset:1
-; VI-NEXT:    buffer_load_ubyte v2, off, s[8:11], 0 offset:2
+; VI-NEXT:    buffer_load_ushort v0, off, s[8:11], 0
+; VI-NEXT:    buffer_load_ubyte v1, off, s[8:11], 0 offset:2
 ; VI-NEXT:    s_mov_b32 s0, s4
 ; VI-NEXT:    s_mov_b32 s1, s5
-; VI-NEXT:    s_waitcnt vmcnt(2)
+; VI-NEXT:    s_waitcnt vmcnt(1)
 ; VI-NEXT:    buffer_store_byte v0, off, s[0:3], 0
-; VI-NEXT:    s_waitcnt vmcnt(2)
-; VI-NEXT:    buffer_store_byte v1, off, s[0:3], 0 offset:1
-; VI-NEXT:    s_waitcnt vmcnt(2)
-; VI-NEXT:    buffer_store_byte v2, off, s[0:3], 0 offset:2
+; VI-NEXT:    s_waitcnt vmcnt(1)
+; VI-NEXT:    buffer_store_byte v1, off, s[0:3], 0 offset:2
+; VI-NEXT:    v_lshrrev_b16_e32 v0, 8, v0
+; VI-NEXT:    buffer_store_byte v0, off, s[0:3], 0 offset:1
 ; VI-NEXT:    s_endpgm
   %val = load <3 x i8>, <3 x i8> addrspace(1)* %in, align 1
   store <3 x i8> %val, <3 x i8> addrspace(1)* %out, align 1
   ret void
 }
 
 define amdgpu_kernel void @test_copy_v4i8_volatile_load(<4 x i8> addrspace(1)* %out, <4 x i8> addrspace(1)* %in) nounwind {
 ; SI-LABEL: test_copy_v4i8_volatile_load:

llvm/test/CodeGen/AMDGPU/fptoui.f16.ll

entry:
   ret void
 }
 
 ; Need to make sure we promote f16 to f32 when converting f16 to i64. Existing
 ; test checks code generated for 'i64 = fp_to_uint f32'.
 
 ; GCN-LABEL: {{^}}fptoui_v2f16_to_v2i64
 ; GCN: buffer_load_dword v[[A_F16_0:[0-9]+]]
-; GCN: v_mov_b32_e32 v[[R_I64_1_High:[0-9]+]], 0
 ; SI: v_lshrrev_b32_e32 v[[A_F16_1:[0-9]+]], 16, v[[A_F16_0]]
 ; SI: v_cvt_f32_f16_e32 v[[A_F32_0:[0-9]+]], v[[A_F16_0]]
 ; SI: v_cvt_f32_f16_e32 v[[A_F32_1:[0-9]+]], v[[A_F16_1]]
 ; SI: v_cvt_u32_f32_e32 v[[R_I64_0_Low:[0-9]+]], v[[A_F32_0]]
 ; SI: v_cvt_u32_f32_e32 v[[R_I64_1_Low:[0-9]+]], v[[A_F32_1]]
 ; VI: v_cvt_f32_f16_sdwa v[[A_F32_1:[0-9]+]], v[[A_F16_0]] dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
 ; VI: v_cvt_f32_f16_e32 v[[A_F32_0:[0-9]+]], v[[A_F16_0]]
 ; VI: v_cvt_u32_f32_e32 v[[R_I64_1_Low:[0-9]+]], v[[A_F32_1]]
 ; VI: v_cvt_u32_f32_e32 v[[R_I64_0_Low:[0-9]+]], v[[A_F32_0]]
 ; GCN: v_mov_b32_e32 v[[R_I64_0_High:[0-9]+]], 0
+; GCN: v_mov_b32_e32 v[[R_I64_1_High:[0-9]+]], v[[R_I64_0_High]]
 ; GCN: buffer_store_dwordx4 v{{\[}}[[R_I64_0_Low]]{{\:}}[[R_I64_1_High]]{{\]}}
 ; GCN: s_endpgm
 define amdgpu_kernel void @fptoui_v2f16_to_v2i64(
     <2 x i64> addrspace(1)* %r,
     <2 x half> addrspace(1)* %a) {
 entry:
   %a.val = load <2 x half>, <2 x half> addrspace(1)* %a
   %r.val = fptoui <2 x half> %a.val to <2 x i64>

llvm/test/CodeGen/AMDGPU/load-constant-i16.ll

 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s3, 0xf000
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s2, -1
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, v1
 ; GCN-NOHSA-VI-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s0, s4
 ; GCN-NOHSA-VI-NEXT:    s_load_dword s4, s[6:7], 0x0
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s1, s5
 ; GCN-NOHSA-VI-NEXT:    s_waitcnt lgkmcnt(0)
-; GCN-NOHSA-VI-NEXT:    s_and_b32 s5, s4, 0xffff
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s4, s4, 16
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s5
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s4
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s5, s4, 16
+; GCN-NOHSA-VI-NEXT:    s_and_b32 s4, s4, 0xffff
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s4
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s5
 ; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0
 ; GCN-NOHSA-VI-NEXT:    s_endpgm
 ;
 ; EG-LABEL: constant_zextload_v2i16_to_v2i64:
 ; EG:       ; %bb.0:
 ; EG-NEXT:    ALU 0, @8, KC0[CB0:0-32], KC1[]
 ; EG-NEXT:    TEX 0 @6
 ; EG-NEXT:    ALU 6, @9, KC0[CB0:0-32], KC1[]

 ; GCN-NOHSA-SI-NEXT:    s_load_dwordx2 s[2:3], s[2:3], 0x0
 ; GCN-NOHSA-SI-NEXT:    s_mov_b32 s6, -1
 ; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v1, 0
 ; GCN-NOHSA-SI-NEXT:    s_mov_b32 s8, 0xffff
 ; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v3, v1
 ; GCN-NOHSA-SI-NEXT:    s_mov_b32 s4, s0
 ; GCN-NOHSA-SI-NEXT:    s_mov_b32 s5, s1
 ; GCN-NOHSA-SI-NEXT:    s_waitcnt lgkmcnt(0)
-; GCN-NOHSA-SI-NEXT:    s_lshr_b32 s0, s3, 16
-; GCN-NOHSA-SI-NEXT:    s_lshr_b32 s1, s2, 16
+; GCN-NOHSA-SI-NEXT:    s_lshr_b32 s0, s2, 16
+; GCN-NOHSA-SI-NEXT:    s_lshr_b32 s1, s3, 16
 ; GCN-NOHSA-SI-NEXT:    s_and_b32 s3, s3, s8
 ; GCN-NOHSA-SI-NEXT:    s_and_b32 s2, s2, s8
 ; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v0, s3
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v2, s0
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v2, s1
 ; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[4:7], 0 offset:16
 ; GCN-NOHSA-SI-NEXT:    s_waitcnt expcnt(0)
 ; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v0, s2
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v2, s1
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v2, s0
 ; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[4:7], 0
 ; GCN-NOHSA-SI-NEXT:    s_endpgm
 ;
 ; GCN-HSA-LABEL: constant_zextload_v4i16_to_v4i64:
 ; GCN-HSA:       ; %bb.0:
 ; GCN-HSA-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GCN-HSA-NEXT:    s_mov_b32 s6, 0xffff
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v3, v1
 ; GCN-HSA-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-HSA-NEXT:    s_load_dwordx2 s[2:3], s[2:3], 0x0
 ; GCN-HSA-NEXT:    s_waitcnt lgkmcnt(0)
-; GCN-HSA-NEXT:    s_lshr_b32 s4, s3, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s5, s2, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s5, s3, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s4, s2, 16
 ; GCN-HSA-NEXT:    s_and_b32 s7, s2, s6
 ; GCN-HSA-NEXT:    s_and_b32 s2, s3, s6
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 16
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s4
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s5
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s7
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s5
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s4
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-HSA-NEXT:    s_endpgm
 ;
 ; GCN-NOHSA-VI-LABEL: constant_zextload_v4i16_to_v4i64:
 ; GCN-NOHSA-VI:       ; %bb.0:
 ; GCN-NOHSA-VI-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s8, 0xffff
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, 0
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s3, 0xf000
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s2, -1
 ; GCN-NOHSA-VI-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s0, s4
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s1, s5
 ; GCN-NOHSA-VI-NEXT:    s_load_dwordx2 s[4:5], s[6:7], 0x0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, v1
 ; GCN-NOHSA-VI-NEXT:    s_waitcnt lgkmcnt(0)
-; GCN-NOHSA-VI-NEXT:    s_and_b32 s7, s5, s8
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s5, s5, 16
-; GCN-NOHSA-VI-NEXT:    s_and_b32 s6, s4, s8
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s4, s4, 16
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s7
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s5
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s7, s5, 16
+; GCN-NOHSA-VI-NEXT:    s_and_b32 s5, s5, s8
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s6, s4, 16
+; GCN-NOHSA-VI-NEXT:    s_and_b32 s4, s4, s8
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s5
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s7
 ; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:16
 ; GCN-NOHSA-VI-NEXT:    s_nop 0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s6
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s4
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s4
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s6
 ; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0
 ; GCN-NOHSA-VI-NEXT:    s_endpgm
 ;
 ; EG-LABEL: constant_zextload_v4i16_to_v4i64:
 ; EG:       ; %bb.0:
 ; EG-NEXT:    ALU 0, @8, KC0[CB0:0-32], KC1[]
 ; EG-NEXT:    TEX 0 @6
 ; EG-NEXT:    ALU 18, @9, KC0[CB0:0-32], KC1[]

 ; GCN-NOHSA-SI-NEXT:    s_load_dwordx4 s[8:11], s[2:3], 0x0
 ; GCN-NOHSA-SI-NEXT:    s_mov_b32 s6, -1
 ; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v1, 0
 ; GCN-NOHSA-SI-NEXT:    s_mov_b32 s2, 0xffff
 ; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v3, v1
 ; GCN-NOHSA-SI-NEXT:    s_mov_b32 s4, s0
 ; GCN-NOHSA-SI-NEXT:    s_mov_b32 s5, s1
 ; GCN-NOHSA-SI-NEXT:    s_waitcnt lgkmcnt(0)
-; GCN-NOHSA-SI-NEXT:    s_lshr_b32 s0, s9, 16
-; GCN-NOHSA-SI-NEXT:    s_lshr_b32 s1, s11, 16
+; GCN-NOHSA-SI-NEXT:    s_lshr_b32 s0, s8, 16
+; GCN-NOHSA-SI-NEXT:    s_lshr_b32 s1, s9, 16
 ; GCN-NOHSA-SI-NEXT:    s_lshr_b32 s3, s10, 16
-; GCN-NOHSA-SI-NEXT:    s_lshr_b32 s12, s8, 16
+; GCN-NOHSA-SI-NEXT:    s_lshr_b32 s12, s11, 16
 ; GCN-NOHSA-SI-NEXT:    s_and_b32 s8, s8, s2
-; GCN-NOHSA-SI-NEXT:    s_and_b32 s10, s10, s2
+; GCN-NOHSA-SI-NEXT:    s_and_b32 s9, s9, s2
 ; GCN-NOHSA-SI-NEXT:    s_and_b32 s11, s11, s2
-; GCN-NOHSA-SI-NEXT:    s_and_b32 s2, s9, s2
+; GCN-NOHSA-SI-NEXT:    s_and_b32 s2, s10, s2
 ; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v0, s11
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v2, s1
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v2, s12
 ; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[4:7], 0 offset:48
 ; GCN-NOHSA-SI-NEXT:    s_waitcnt expcnt(0)
 ; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v0, s2
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v2, s0
-; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[4:7], 0 offset:16
-; GCN-NOHSA-SI-NEXT:    s_waitcnt expcnt(0)
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v0, s10
 ; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v2, s3
 ; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[4:7], 0 offset:32
 ; GCN-NOHSA-SI-NEXT:    s_waitcnt expcnt(0)
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v0, s9
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v2, s1
+; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[4:7], 0 offset:16
+; GCN-NOHSA-SI-NEXT:    s_waitcnt expcnt(0)
 ; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v0, s8
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v2, s12
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v2, s0
 ; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[4:7], 0
 ; GCN-NOHSA-SI-NEXT:    s_endpgm
 ;
 ; GCN-HSA-LABEL: constant_zextload_v8i16_to_v8i64:
 ; GCN-HSA:       ; %bb.0:
 ; GCN-HSA-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GCN-HSA-NEXT:    s_mov_b32 s8, 0xffff
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v3, v1
 ; GCN-HSA-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-HSA-NEXT:    s_load_dwordx4 s[4:7], s[2:3], 0x0
 ; GCN-HSA-NEXT:    s_waitcnt lgkmcnt(0)
-; GCN-HSA-NEXT:    s_lshr_b32 s9, s5, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s9, s4, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s10, s5, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s11, s6, 16
 ; GCN-HSA-NEXT:    s_lshr_b32 s2, s7, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s10, s6, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s11, s4, 16
 ; GCN-HSA-NEXT:    s_and_b32 s3, s7, s8
 ; GCN-HSA-NEXT:    s_and_b32 s4, s4, s8
-; GCN-HSA-NEXT:    s_and_b32 s6, s6, s8
 ; GCN-HSA-NEXT:    s_and_b32 s5, s5, s8
+; GCN-HSA-NEXT:    s_and_b32 s6, s6, s8
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 48
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
-; GCN-HSA-NEXT:    s_add_u32 s2, s0, 16
+; GCN-HSA-NEXT:    s_add_u32 s2, s0, 32
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
-; GCN-HSA-NEXT:    s_add_u32 s2, s0, 32
-; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s5
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s9
+; GCN-HSA-NEXT:    s_add_u32 s2, s0, 16
+; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s6
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s11
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
-; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s6
+; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s5
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s10
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s4
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s11
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s9
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-HSA-NEXT:    s_endpgm
 ;
 ; GCN-NOHSA-VI-LABEL: constant_zextload_v8i16_to_v8i64:
 ; GCN-NOHSA-VI:       ; %bb.0:
 ; GCN-NOHSA-VI-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s8, 0xffff
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, 0
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s3, 0xf000
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s2, -1
 ; GCN-NOHSA-VI-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s0, s4
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s1, s5
 ; GCN-NOHSA-VI-NEXT:    s_load_dwordx4 s[4:7], s[6:7], 0x0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, v1
 ; GCN-NOHSA-VI-NEXT:    s_waitcnt lgkmcnt(0)
-; GCN-NOHSA-VI-NEXT:    s_and_b32 s9, s4, s8
-; GCN-NOHSA-VI-NEXT:    s_and_b32 s10, s5, s8
-; GCN-NOHSA-VI-NEXT:    s_and_b32 s11, s6, s8
-; GCN-NOHSA-VI-NEXT:    s_and_b32 s8, s7, s8
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s7, s7, 16
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s6, s6, 16
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s8
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s7
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s12, s7, 16
+; GCN-NOHSA-VI-NEXT:    s_and_b32 s7, s7, s8
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s11, s6, 16
+; GCN-NOHSA-VI-NEXT:    s_and_b32 s6, s6, s8
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s7
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s12
 ; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:48
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s5, s5, 16
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s11
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s6
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s10, s5, 16
+; GCN-NOHSA-VI-NEXT:    s_and_b32 s5, s5, s8
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s6
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s11
 ; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:32
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s4, s4, 16
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s10
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s5
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s9, s4, 16
+; GCN-NOHSA-VI-NEXT:    s_and_b32 s4, s4, s8
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s5
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s10
 ; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:16
 ; GCN-NOHSA-VI-NEXT:    s_nop 0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s9
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s4
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s4
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s9
 ; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0
 ; GCN-NOHSA-VI-NEXT:    s_endpgm
 ;
 ; EG-LABEL: constant_zextload_v8i16_to_v8i64:
 ; EG:       ; %bb.0:
 ; EG-NEXT:    ALU 0, @10, KC0[CB0:0-32], KC1[]
 ; EG-NEXT:    TEX 0 @8
 ; EG-NEXT:    ALU 30, @11, KC0[CB0:0-32], KC1[]

 ; GCN-NOHSA-SI-NEXT:    s_load_dwordx8 s[4:11], s[14:15], 0x0
 ; GCN-NOHSA-SI-NEXT:    s_mov_b32 s2, -1
 ; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v1, 0
 ; GCN-NOHSA-SI-NEXT:    s_mov_b32 s14, 0xffff
 ; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v3, v1
 ; GCN-NOHSA-SI-NEXT:    s_mov_b32 s0, s12
 ; GCN-NOHSA-SI-NEXT:    s_mov_b32 s1, s13
 ; GCN-NOHSA-SI-NEXT:    s_waitcnt lgkmcnt(0)
-; GCN-NOHSA-SI-NEXT:    s_lshr_b32 s12, s5, 16
-; GCN-NOHSA-SI-NEXT:    s_lshr_b32 s13, s7, 16
-; GCN-NOHSA-SI-NEXT:    s_lshr_b32 s15, s11, 16
-; GCN-NOHSA-SI-NEXT:    s_lshr_b32 s16, s9, 16
-; GCN-NOHSA-SI-NEXT:    s_lshr_b32 s17, s8, 16
-; GCN-NOHSA-SI-NEXT:    s_lshr_b32 s18, s10, 16
-; GCN-NOHSA-SI-NEXT:    s_lshr_b32 s19, s6, 16
-; GCN-NOHSA-SI-NEXT:    s_lshr_b32 s20, s4, 16
+; GCN-NOHSA-SI-NEXT:    s_lshr_b32 s12, s4, 16
+; GCN-NOHSA-SI-NEXT:    s_lshr_b32 s13, s5, 16
+; GCN-NOHSA-SI-NEXT:    s_lshr_b32 s15, s6, 16
+; GCN-NOHSA-SI-NEXT:    s_lshr_b32 s16, s7, 16
+; GCN-NOHSA-SI-NEXT:    s_lshr_b32 s17, s10, 16
+; GCN-NOHSA-SI-NEXT:    s_lshr_b32 s18, s11, 16
+; GCN-NOHSA-SI-NEXT:    s_lshr_b32 s19, s8, 16
+; GCN-NOHSA-SI-NEXT:    s_lshr_b32 s20, s9, 16
 ; GCN-NOHSA-SI-NEXT:    s_and_b32 s4, s4, s14
-; GCN-NOHSA-SI-NEXT:    s_and_b32 s6, s6, s14
-; GCN-NOHSA-SI-NEXT:    s_and_b32 s10, s10, s14
-; GCN-NOHSA-SI-NEXT:    s_and_b32 s8, s8, s14
 ; GCN-NOHSA-SI-NEXT:    s_and_b32 s5, s5, s14
+; GCN-NOHSA-SI-NEXT:    s_and_b32 s6, s6, s14
 ; GCN-NOHSA-SI-NEXT:    s_and_b32 s7, s7, s14
-; GCN-NOHSA-SI-NEXT:    s_and_b32 s9, s9, s14
+; GCN-NOHSA-SI-NEXT:    s_and_b32 s10, s10, s14
 ; GCN-NOHSA-SI-NEXT:    s_and_b32 s11, s11, s14
+; GCN-NOHSA-SI-NEXT:    s_and_b32 s9, s9, s14
+; GCN-NOHSA-SI-NEXT:    s_and_b32 s8, s8, s14
 ; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v0, s9
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v2, s16
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v2, s20
 ; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:80
 ; GCN-NOHSA-SI-NEXT:    s_waitcnt expcnt(0)
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v0, s11
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v2, s15
-; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:112
-; GCN-NOHSA-SI-NEXT:    s_waitcnt expcnt(0)
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v0, s7
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v2, s13
-; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:48
-; GCN-NOHSA-SI-NEXT:    s_waitcnt expcnt(0)
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v0, s5
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v2, s12
-; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:16
-; GCN-NOHSA-SI-NEXT:    s_waitcnt expcnt(0)
 ; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v0, s8
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v2, s17
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v2, s19
 ; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:64
 ; GCN-NOHSA-SI-NEXT:    s_waitcnt expcnt(0)
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v0, s10
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v0, s11
 ; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v2, s18
+; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:112
+; GCN-NOHSA-SI-NEXT:    s_waitcnt expcnt(0)
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v0, s10
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v2, s17
 ; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:96
 ; GCN-NOHSA-SI-NEXT:    s_waitcnt expcnt(0)
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v0, s7
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v2, s16
+; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:48
+; GCN-NOHSA-SI-NEXT:    s_waitcnt expcnt(0)
 ; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v0, s6
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v2, s19
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v2, s15
 ; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:32
 ; GCN-NOHSA-SI-NEXT:    s_waitcnt expcnt(0)
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v0, s5
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v2, s13
+; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:16
+; GCN-NOHSA-SI-NEXT:    s_waitcnt expcnt(0)
 ; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v0, s4
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v2, s20
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v2, s12
 ; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0
 ; GCN-NOHSA-SI-NEXT:    s_endpgm
 ;
 ; GCN-HSA-LABEL: constant_zextload_v16i16_to_v16i64:
 ; GCN-HSA:       ; %bb.0:
 ; GCN-HSA-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GCN-HSA-NEXT:    s_mov_b32 s12, 0xffff
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v3, v1
 ; GCN-HSA-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-HSA-NEXT:    s_load_dwordx8 s[4:11], s[2:3], 0x0
 ; GCN-HSA-NEXT:    s_waitcnt lgkmcnt(0)
-; GCN-HSA-NEXT:    s_lshr_b32 s13, s5, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s14, s7, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s15, s11, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s2, s9, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s16, s8, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s13, s4, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s14, s5, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s15, s6, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s16, s7, 16
 ; GCN-HSA-NEXT:    s_lshr_b32 s17, s10, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s18, s6, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s19, s4, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s18, s11, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s19, s8, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s2, s9, 16
 ; GCN-HSA-NEXT:    s_and_b32 s3, s9, s12
 ; GCN-HSA-NEXT:    s_and_b32 s4, s4, s12
-; GCN-HSA-NEXT:    s_and_b32 s6, s6, s12
-; GCN-HSA-NEXT:    s_and_b32 s10, s10, s12
-; GCN-HSA-NEXT:    s_and_b32 s8, s8, s12
 ; GCN-HSA-NEXT:    s_and_b32 s5, s5, s12
+; GCN-HSA-NEXT:    s_and_b32 s6, s6, s12
 ; GCN-HSA-NEXT:    s_and_b32 s7, s7, s12
+; GCN-HSA-NEXT:    s_and_b32 s10, s10, s12
 ; GCN-HSA-NEXT:    s_and_b32 s11, s11, s12
+; GCN-HSA-NEXT:    s_and_b32 s8, s8, s12
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x50
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
-; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x70
+; GCN-HSA-NEXT:    s_add_u32 s2, s0, 64
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
-; GCN-HSA-NEXT:    s_add_u32 s2, s0, 48
-; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s11
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s15
+; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x70
+; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s8
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s19
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
-; GCN-HSA-NEXT:    s_add_u32 s2, s0, 16
-; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s7
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s14
+; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x60
+; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s11
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s18
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
-; GCN-HSA-NEXT:    s_add_u32 s2, s0, 64
-; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s5
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s13
+; GCN-HSA-NEXT:    s_add_u32 s2, s0, 48
+; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s10
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s17
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
-; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x60
-; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s8
+; GCN-HSA-NEXT:    s_add_u32 s2, s0, 32
+; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s7
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s16
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
-; GCN-HSA-NEXT:    s_add_u32 s2, s0, 32
-; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s10
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s17
+; GCN-HSA-NEXT:    s_add_u32 s2, s0, 16
+; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s6
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s15
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
-; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s6
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s18
+; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s5
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s14
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s4
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s19
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s13
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-HSA-NEXT:    s_endpgm
 ;
 ; GCN-NOHSA-VI-LABEL: constant_zextload_v16i16_to_v16i64:
 ; GCN-NOHSA-VI:       ; %bb.0:
 ; GCN-NOHSA-VI-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s12, 0xffff
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, 0
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s3, 0xf000
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s2, -1
 ; GCN-NOHSA-VI-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s0, s4
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s1, s5
 ; GCN-NOHSA-VI-NEXT:    s_load_dwordx8 s[4:11], s[6:7], 0x0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, v1
 ; GCN-NOHSA-VI-NEXT:    s_waitcnt lgkmcnt(0)
-; GCN-NOHSA-VI-NEXT:    s_and_b32 s13, s4, s12
-; GCN-NOHSA-VI-NEXT:    s_and_b32 s14, s5, s12
-; GCN-NOHSA-VI-NEXT:    s_and_b32 s15, s6, s12
-; GCN-NOHSA-VI-NEXT:    s_and_b32 s16, s7, s12
-; GCN-NOHSA-VI-NEXT:    s_and_b32 s17, s8, s12
-; GCN-NOHSA-VI-NEXT:    s_and_b32 s18, s9, s12
-; GCN-NOHSA-VI-NEXT:    s_and_b32 s19, s10, s12
-; GCN-NOHSA-VI-NEXT:    s_and_b32 s12, s11, s12
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s11, s11, 16
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s10, s10, 16
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s12
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s11
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s20, s11, 16
+; GCN-NOHSA-VI-NEXT:    s_and_b32 s11, s11, s12
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s19, s10, 16
+; GCN-NOHSA-VI-NEXT:    s_and_b32 s10, s10, s12
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s11
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s20
 ; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:112
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s9, s9, 16
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s19
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s10
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s18, s9, 16
+; GCN-NOHSA-VI-NEXT:    s_and_b32 s9, s9, s12
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s10
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s19
 ; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:96
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s8, s8, 16
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s18
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s9
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s17, s8, 16
+; GCN-NOHSA-VI-NEXT:    s_and_b32 s8, s8, s12
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s9
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s18
 ; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:80
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s7, s7, 16
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s17
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s8
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s16, s7, 16
+; GCN-NOHSA-VI-NEXT:    s_and_b32 s7, s7, s12
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s8
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s17
 ; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:64
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s6, s6, 16
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s16
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s7
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s15, s6, 16
+; GCN-NOHSA-VI-NEXT:    s_and_b32 s6, s6, s12
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s7
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s16
 ; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:48
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s5, s5, 16
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s15
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s6
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s14, s5, 16
+; GCN-NOHSA-VI-NEXT:    s_and_b32 s5, s5, s12
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s6
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s15
 ; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:32
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s4, s4, 16
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s14
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s5
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s13, s4, 16
+; GCN-NOHSA-VI-NEXT:    s_and_b32 s4, s4, s12
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s5
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s14
 ; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:16
 ; GCN-NOHSA-VI-NEXT:    s_nop 0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s13
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s4
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s4
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s13
 ; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0
 ; GCN-NOHSA-VI-NEXT:    s_endpgm
 ;
 ; EG-LABEL: constant_zextload_v16i16_to_v16i64:
 ; EG:       ; %bb.0:
 ; EG-NEXT:    ALU 0, @16, KC0[CB0:0-32], KC1[]
 ; EG-NEXT:    TEX 1 @12
 ; EG-NEXT:    ALU 62, @17, KC0[CB0:0-32], KC1[]

 define amdgpu_kernel void @constant_zextload_v32i16_to_v32i64(<32 x i64> addrspace(1)* %out, <32 x i16> addrspace(4)* %in) #0 {
 ; GCN-NOHSA-SI-LABEL: constant_zextload_v32i16_to_v32i64:
 ; GCN-NOHSA-SI:       ; %bb.0:
 ; GCN-NOHSA-SI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
 ; GCN-NOHSA-SI-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-NOHSA-SI-NEXT:    s_load_dwordx16 s[4:19], s[2:3], 0x0
 ; GCN-NOHSA-SI-NEXT:    s_mov_b32 s2, 0xffff
 ; GCN-NOHSA-SI-NEXT:    s_waitcnt lgkmcnt(0)
-; GCN-NOHSA-SI-NEXT:    s_lshr_b32 s20, s5, 16
-; GCN-NOHSA-SI-NEXT:    s_lshr_b32 s21, s7, 16
-; GCN-NOHSA-SI-NEXT:    s_lshr_b32 s22, s9, 16
-; GCN-NOHSA-SI-NEXT:    s_lshr_b32 s23, s11, 16
-; GCN-NOHSA-SI-NEXT:    s_lshr_b32 s24, s13, 16
-; GCN-NOHSA-SI-NEXT:    s_lshr_b32 s25, s15, 16
-; GCN-NOHSA-SI-NEXT:    s_lshr_b32 s26, s17, 16
-; GCN-NOHSA-SI-NEXT:    s_lshr_b32 s27, s19, 16
-; GCN-NOHSA-SI-NEXT:    s_and_b32 s28, s4, s2
-; GCN-NOHSA-SI-NEXT:    s_and_b32 s29, s6, s2
-; GCN-NOHSA-SI-NEXT:    s_and_b32 s30, s8, s2
-; GCN-NOHSA-SI-NEXT:    s_and_b32 s31, s10, s2
-; GCN-NOHSA-SI-NEXT:    s_and_b32 s33, s12, s2
-; GCN-NOHSA-SI-NEXT:    s_and_b32 s34, s14, s2
-; GCN-NOHSA-SI-NEXT:    s_and_b32 s35, s16, s2
-; GCN-NOHSA-SI-NEXT:    s_and_b32 s36, s18, s2
+; GCN-NOHSA-SI-NEXT:    s_lshr_b32 s20, s4, 16
+; GCN-NOHSA-SI-NEXT:    s_lshr_b32 s21, s5, 16
+; GCN-NOHSA-SI-NEXT:    s_lshr_b32 s22, s6, 16
+; GCN-NOHSA-SI-NEXT:    s_lshr_b32 s23, s7, 16
+; GCN-NOHSA-SI-NEXT:    s_lshr_b32 s24, s8, 16
+; GCN-NOHSA-SI-NEXT:    s_lshr_b32 s25, s9, 16
+; GCN-NOHSA-SI-NEXT:    s_lshr_b32 s26, s10, 16
+; GCN-NOHSA-SI-NEXT:    s_lshr_b32 s27, s11, 16
+; GCN-NOHSA-SI-NEXT:    s_and_b32 s4, s4, s2
 ; GCN-NOHSA-SI-NEXT:    s_and_b32 s5, s5, s2
+; GCN-NOHSA-SI-NEXT:    s_and_b32 s6, s6, s2
 ; GCN-NOHSA-SI-NEXT:    s_and_b32 s7, s7, s2
+; GCN-NOHSA-SI-NEXT:    s_and_b32 s8, s8, s2
 ; GCN-NOHSA-SI-NEXT:    s_and_b32 s9, s9, s2
+; GCN-NOHSA-SI-NEXT:    s_and_b32 s10, s10, s2
 ; GCN-NOHSA-SI-NEXT:    s_and_b32 s11, s11, s2
-; GCN-NOHSA-SI-NEXT:    s_and_b32 s13, s13, s2
-; GCN-NOHSA-SI-NEXT:    s_and_b32 s15, s15, s2
-; GCN-NOHSA-SI-NEXT:    s_and_b32 s17, s17, s2
-; GCN-NOHSA-SI-NEXT:    s_and_b32 s19, s19, s2
-; GCN-NOHSA-SI-NEXT:    s_lshr_b32 s18, s18, 16
-; GCN-NOHSA-SI-NEXT:    s_lshr_b32 s16, s16, 16
-; GCN-NOHSA-SI-NEXT:    s_lshr_b32 s14, s14, 16
+; GCN-NOHSA-SI-NEXT:    s_and_b32 s28, s12, s2
+; GCN-NOHSA-SI-NEXT:    s_and_b32 s29, s13, s2
+; GCN-NOHSA-SI-NEXT:    s_and_b32 s30, s14, s2
+; GCN-NOHSA-SI-NEXT:    s_and_b32 s31, s15, s2
+; GCN-NOHSA-SI-NEXT:    s_and_b32 s33, s16, s2
+; GCN-NOHSA-SI-NEXT:    s_and_b32 s34, s17, s2
+; GCN-NOHSA-SI-NEXT:    s_and_b32 s35, s18, s2
+; GCN-NOHSA-SI-NEXT:    s_and_b32 s36, s19, s2
 ; GCN-NOHSA-SI-NEXT:    s_lshr_b32 s12, s12, 16
-; GCN-NOHSA-SI-NEXT:    s_lshr_b32 s10, s10, 16
-; GCN-NOHSA-SI-NEXT:    s_lshr_b32 s8, s8, 16
-; GCN-NOHSA-SI-NEXT:    s_lshr_b32 s6, s6, 16
-; GCN-NOHSA-SI-NEXT:    s_lshr_b32 s4, s4, 16
+; GCN-NOHSA-SI-NEXT:    s_lshr_b32 s13, s13, 16
+; GCN-NOHSA-SI-NEXT:    s_lshr_b32 s14, s14, 16
+; GCN-NOHSA-SI-NEXT:    s_lshr_b32 s15, s15, 16
+; GCN-NOHSA-SI-NEXT:    s_lshr_b32 s16, s16, 16
+; GCN-NOHSA-SI-NEXT:    s_lshr_b32 s17, s17, 16
+; GCN-NOHSA-SI-NEXT:    s_lshr_b32 s18, s18, 16
+; GCN-NOHSA-SI-NEXT:    s_lshr_b32 s19, s19, 16
 ; GCN-NOHSA-SI-NEXT:    s_mov_b32 s3, 0xf000
 ; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v1, 0
 ; GCN-NOHSA-SI-NEXT:    s_mov_b32 s2, -1
 ; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v3, v1
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v0, s19
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v2, s27
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v0, s36
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v2, s19
 ; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:240
 ; GCN-NOHSA-SI-NEXT:    s_waitcnt expcnt(0)
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v0, s17
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v2, s26
-; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:208
-; GCN-NOHSA-SI-NEXT:    s_waitcnt expcnt(0)
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v0, s15
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v2, s25
-; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:176
-; GCN-NOHSA-SI-NEXT:    s_waitcnt expcnt(0)
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v0, s13
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v2, s24
-; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:144
-; GCN-NOHSA-SI-NEXT:    s_waitcnt expcnt(0)
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v0, s11
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v2, s23
-; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:112
-; GCN-NOHSA-SI-NEXT:    s_waitcnt expcnt(0)
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v0, s9
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v2, s22
-; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:80
-; GCN-NOHSA-SI-NEXT:    s_waitcnt expcnt(0)
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v0, s7
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v2, s21
-; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:48
-; GCN-NOHSA-SI-NEXT:    s_waitcnt expcnt(0)
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v0, s5
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v2, s20
-; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:16
-; GCN-NOHSA-SI-NEXT:    s_waitcnt expcnt(0)
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v0, s36
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v0, s35
 ; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v2, s18
 ; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:224
 ; GCN-NOHSA-SI-NEXT:    s_waitcnt expcnt(0)
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v0, s35
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v0, s34
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v2, s17
+; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:208
+; GCN-NOHSA-SI-NEXT:    s_waitcnt expcnt(0)
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v0, s33
 ; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v2, s16
 ; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:192
 ; GCN-NOHSA-SI-NEXT:    s_waitcnt expcnt(0)
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v0, s34
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v0, s31
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v2, s15
+; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:176
+; GCN-NOHSA-SI-NEXT:    s_waitcnt expcnt(0)
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v0, s30
 ; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v2, s14
 ; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:160
 ; GCN-NOHSA-SI-NEXT:    s_waitcnt expcnt(0)
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v0, s33
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v0, s29
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v2, s13
+; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:144
+; GCN-NOHSA-SI-NEXT:    s_waitcnt expcnt(0)
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v0, s28
 ; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v2, s12
 ; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:128
 ; GCN-NOHSA-SI-NEXT:    s_waitcnt expcnt(0)
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v0, s31
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v2, s10
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v0, s11
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v2, s27
+; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:112
+; GCN-NOHSA-SI-NEXT:    s_waitcnt expcnt(0)
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v0, s10
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v2, s26
 ; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:96
 ; GCN-NOHSA-SI-NEXT:    s_waitcnt expcnt(0)
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v0, s30
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v2, s8
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v0, s9
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v2, s25
+; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:80
+; GCN-NOHSA-SI-NEXT:    s_waitcnt expcnt(0)
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v0, s8
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v2, s24
 ; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:64
 ; GCN-NOHSA-SI-NEXT:    s_waitcnt expcnt(0)
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v0, s29
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v2, s6
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v0, s7
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v2, s23
+; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:48
+; GCN-NOHSA-SI-NEXT:    s_waitcnt expcnt(0)
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v0, s6
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v2, s22
 ; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:32
 ; GCN-NOHSA-SI-NEXT:    s_waitcnt expcnt(0)
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v0, s28
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v2, s4
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v0, s5
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v2, s21
+; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:16
+; GCN-NOHSA-SI-NEXT:    s_waitcnt expcnt(0)
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v0, s4
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v2, s20
 ; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0
 ; GCN-NOHSA-SI-NEXT:    s_endpgm
 ;
 ; GCN-HSA-LABEL: constant_zextload_v32i16_to_v32i64:
 ; GCN-HSA:       ; %bb.0:
 ; GCN-HSA-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GCN-HSA-NEXT:    s_mov_b32 s20, 0xffff
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v3, v1
 ; GCN-HSA-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-HSA-NEXT:    s_load_dwordx16 s[4:19], s[2:3], 0x0
 ; GCN-HSA-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-HSA-NEXT:    s_and_b32 s21, s4, s20
-; GCN-HSA-NEXT:    s_and_b32 s22, s6, s20
-; GCN-HSA-NEXT:    s_and_b32 s23, s8, s20
-; GCN-HSA-NEXT:    s_and_b32 s24, s10, s20
-; GCN-HSA-NEXT:    s_and_b32 s25, s12, s20
-; GCN-HSA-NEXT:    s_and_b32 s26, s14, s20
-; GCN-HSA-NEXT:    s_and_b32 s27, s16, s20
-; GCN-HSA-NEXT:    s_and_b32 s28, s18, s20
-; GCN-HSA-NEXT:    s_and_b32 s29, s5, s20
-; GCN-HSA-NEXT:    s_and_b32 s30, s7, s20
-; GCN-HSA-NEXT:    s_and_b32 s31, s9, s20
-; GCN-HSA-NEXT:    s_and_b32 s33, s11, s20
-; GCN-HSA-NEXT:    s_and_b32 s34, s13, s20
-; GCN-HSA-NEXT:    s_and_b32 s35, s15, s20
-; GCN-HSA-NEXT:    s_and_b32 s36, s17, s20
+; GCN-HSA-NEXT:    s_and_b32 s22, s5, s20
+; GCN-HSA-NEXT:    s_and_b32 s23, s6, s20
+; GCN-HSA-NEXT:    s_and_b32 s24, s7, s20
+; GCN-HSA-NEXT:    s_and_b32 s25, s8, s20
+; GCN-HSA-NEXT:    s_and_b32 s26, s9, s20
+; GCN-HSA-NEXT:    s_and_b32 s27, s10, s20
+; GCN-HSA-NEXT:    s_and_b32 s28, s11, s20
+; GCN-HSA-NEXT:    s_and_b32 s29, s12, s20
+; GCN-HSA-NEXT:    s_and_b32 s30, s13, s20
+; GCN-HSA-NEXT:    s_and_b32 s31, s14, s20
+; GCN-HSA-NEXT:    s_and_b32 s33, s15, s20
+; GCN-HSA-NEXT:    s_and_b32 s34, s16, s20
+; GCN-HSA-NEXT:    s_and_b32 s35, s17, s20
+; GCN-HSA-NEXT:    s_and_b32 s36, s18, s20
 ; GCN-HSA-NEXT:    s_and_b32 s20, s19, s20
 ; GCN-HSA-NEXT:    s_lshr_b32 s19, s19, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s4, s4, 16
 ; GCN-HSA-NEXT:    s_lshr_b32 s5, s5, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s6, s6, 16
 ; GCN-HSA-NEXT:    s_lshr_b32 s7, s7, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s8, s8, 16
 ; GCN-HSA-NEXT:    s_lshr_b32 s9, s9, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s10, s10, 16
 ; GCN-HSA-NEXT:    s_lshr_b32 s11, s11, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s12, s12, 16
 ; GCN-HSA-NEXT:    s_lshr_b32 s13, s13, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s14, s14, 16
 ; GCN-HSA-NEXT:    s_lshr_b32 s15, s15, 16
+; GCN-HSA-NEXT:    s_lshr_b32 s16, s16, 16
 ; GCN-HSA-NEXT:    s_lshr_b32 s17, s17, 16
 ; GCN-HSA-NEXT:    s_lshr_b32 s18, s18, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s16, s16, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s14, s14, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s12, s12, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s10, s10, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s8, s8, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s6, s6, 16
-; GCN-HSA-NEXT:    s_lshr_b32 s4, s4, 16
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0xf0
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
-; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0xd0
+; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0xe0
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v7, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v6, s2
-; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0xb0
+; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0xd0
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v8, s2
-; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x90
+; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0xc0
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v10, s2
-; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x70
+; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0xb0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s20
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s19
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s36
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s17
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s18
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[6:7], v[0:3]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s35
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s15
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s17
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[8:9], v[0:3]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
-; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x50
+; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0xa0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s34
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s13
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s16
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[10:11], v[0:3]
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s33
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s11
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s15
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
-; GCN-HSA-NEXT:    s_add_u32 s2, s0, 48
+; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x90
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s31
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s9
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s14
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
-; GCN-HSA-NEXT:    s_add_u32 s2, s0, 16
+; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x80
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s30
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s7
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s13
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
-; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0xe0
+; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x70
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s29
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s5
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s12
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
-; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0xc0
+; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x60
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s28
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s18
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s11
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
-; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0xa0
+; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x50
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s27
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s16
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s10
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
-; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x80
+; GCN-HSA-NEXT:    s_add_u32 s2, s0, 64
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s26
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s14
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s9
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
-; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x60
+; GCN-HSA-NEXT:    s_add_u32 s2, s0, 48
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s25
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s12
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s8
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
-; GCN-HSA-NEXT:    s_add_u32 s2, s0, 64
+; GCN-HSA-NEXT:    s_add_u32 s2, s0, 32
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s24
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s10
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s7
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
-; GCN-HSA-NEXT:    s_add_u32 s2, s0, 32
+; GCN-HSA-NEXT:    s_add_u32 s2, s0, 16
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s23
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s8
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s6
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s22
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s6
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s5
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s21
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s4
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
 ; GCN-HSA-NEXT:    s_endpgm
 ;
 ; GCN-NOHSA-VI-LABEL: constant_zextload_v32i16_to_v32i64:
 ; GCN-NOHSA-VI:       ; %bb.0:
 ; GCN-NOHSA-VI-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s20, 0xffff
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, 0
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s3, 0xf000
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s2, -1
 ; GCN-NOHSA-VI-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s0, s4
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s1, s5
 ; GCN-NOHSA-VI-NEXT:    s_load_dwordx16 s[4:19], s[6:7], 0x0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, v1
 ; GCN-NOHSA-VI-NEXT:    s_waitcnt lgkmcnt(0)
-; GCN-NOHSA-VI-NEXT:    s_and_b32 s21, s4, s20
-; GCN-NOHSA-VI-NEXT:    s_and_b32 s22, s5, s20
-; GCN-NOHSA-VI-NEXT:    s_and_b32 s23, s6, s20
-; GCN-NOHSA-VI-NEXT:    s_and_b32 s24, s7, s20
-; GCN-NOHSA-VI-NEXT:    s_and_b32 s25, s8, s20
-; GCN-NOHSA-VI-NEXT:    s_and_b32 s26, s9, s20
-; GCN-NOHSA-VI-NEXT:    s_and_b32 s27, s10, s20
-; GCN-NOHSA-VI-NEXT:    s_and_b32 s28, s11, s20
-; GCN-NOHSA-VI-NEXT:    s_and_b32 s29, s12, s20
-; GCN-NOHSA-VI-NEXT:    s_and_b32 s30, s13, s20
-; GCN-NOHSA-VI-NEXT:    s_and_b32 s31, s14, s20
-; GCN-NOHSA-VI-NEXT:    s_and_b32 s33, s15, s20
-; GCN-NOHSA-VI-NEXT:    s_and_b32 s34, s16, s20
-; GCN-NOHSA-VI-NEXT:    s_and_b32 s35, s17, s20
-; GCN-NOHSA-VI-NEXT:    s_and_b32 s36, s18, s20
-; GCN-NOHSA-VI-NEXT:    s_and_b32 s20, s19, s20
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s19, s19, 16
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s18, s18, 16
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s20
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s19
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s37, s19, 16
+; GCN-NOHSA-VI-NEXT:    s_and_b32 s19, s19, s20
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s36, s18, 16
+; GCN-NOHSA-VI-NEXT:    s_and_b32 s18, s18, s20
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s19
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s37
 ; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:240
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s17, s17, 16
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s36
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s18
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s35, s17, 16
+; GCN-NOHSA-VI-NEXT:    s_and_b32 s17, s17, s20
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s18
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s36
 ; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:224
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s16, s16, 16
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s35
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s17
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s34, s16, 16
+; GCN-NOHSA-VI-NEXT:    s_and_b32 s16, s16, s20
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s17
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s35
 ; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:208
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s15, s15, 16
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s34
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s16
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s33, s15, 16
+; GCN-NOHSA-VI-NEXT:    s_and_b32 s15, s15, s20
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s16
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s34
 ; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:192
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s14, s14, 16
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s33
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s15
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s31, s14, 16
+; GCN-NOHSA-VI-NEXT:    s_and_b32 s14, s14, s20
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s15
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s33
 ; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:176
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s13, s13, 16
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s31
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s14
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s30, s13, 16
+; GCN-NOHSA-VI-NEXT:    s_and_b32 s13, s13, s20
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s14
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s31
 ; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:160
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s12, s12, 16
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s30
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s13
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s29, s12, 16
+; GCN-NOHSA-VI-NEXT:    s_and_b32 s12, s12, s20
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s13
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s30
 ; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:144
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s11, s11, 16
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s29
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s12
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s28, s11, 16
+; GCN-NOHSA-VI-NEXT:    s_and_b32 s11, s11, s20
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s12
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s29
 ; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:128
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s10, s10, 16
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s28
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s11
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s27, s10, 16
+; GCN-NOHSA-VI-NEXT:    s_and_b32 s10, s10, s20
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s11
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s28
 ; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:112
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s9, s9, 16
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s27
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s10
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s26, s9, 16
+; GCN-NOHSA-VI-NEXT:    s_and_b32 s9, s9, s20
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s10
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s27
 ; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:96
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s8, s8, 16
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s26
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s9
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s25, s8, 16
+; GCN-NOHSA-VI-NEXT:    s_and_b32 s8, s8, s20
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s9
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s26
 ; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:80
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s7, s7, 16
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s25
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s8
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s24, s7, 16
+; GCN-NOHSA-VI-NEXT:    s_and_b32 s7, s7, s20
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s8
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s25
 ; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:64
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s6, s6, 16
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s24
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s7
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s23, s6, 16
+; GCN-NOHSA-VI-NEXT:    s_and_b32 s6, s6, s20
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s7
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s24
 ; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:48
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s5, s5, 16
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s23
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s6
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s22, s5, 16
+; GCN-NOHSA-VI-NEXT:    s_and_b32 s5, s5, s20
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s6
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s23
 ; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:32
-; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s4, s4, 16
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s22
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s5
+; GCN-NOHSA-VI-NEXT:    s_lshr_b32 s21, s4, 16
+; GCN-NOHSA-VI-NEXT:    s_and_b32 s4, s4, s20
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s5
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s22
 ; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:16
 ; GCN-NOHSA-VI-NEXT:    s_nop 0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s21
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s4
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v0, s4
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, s21
 ; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0
 ; GCN-NOHSA-VI-NEXT:    s_endpgm
 ;
 ; EG-LABEL: constant_zextload_v32i16_to_v32i64:
 ; EG:       ; %bb.0:
 ; EG-NEXT:    ALU 0, @30, KC0[CB0:0-32], KC1[]
 ; EG-NEXT:    TEX 2 @22
 ; EG-NEXT:    ALU 33, @31, KC0[], KC1[]

llvm/test/CodeGen/AMDGPU/load-global-i16.ll

 ; GCN-NOHSA-VI-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s3, 0xf000
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s2, -1
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s10, s2
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s11, s3
 ; GCN-NOHSA-VI-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s8, s6
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s9, s7
-; GCN-NOHSA-VI-NEXT:    buffer_load_dword v2, off, s[8:11], 0
+; GCN-NOHSA-VI-NEXT:    buffer_load_dword v0, off, s[8:11], 0
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, 0
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s0, s4
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s1, s5
 ; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, v1
 ; GCN-NOHSA-VI-NEXT:    s_waitcnt vmcnt(0)
-; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v0, 0xffff, v2
-; GCN-NOHSA-VI-NEXT:    v_lshrrev_b32_e32 v2, 16, v2
+; GCN-NOHSA-VI-NEXT:    v_lshrrev_b32_e32 v2, 16, v0
+; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v0, 0xffff, v0
 ; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0
 ; GCN-NOHSA-VI-NEXT:    s_endpgm
 ;
 ; EG-LABEL: global_zextload_v2i16_to_v2i64:
 ; EG:       ; %bb.0:
 ; EG-NEXT:    ALU 0, @8, KC0[CB0:0-32], KC1[]
 ; EG-NEXT:    TEX 0 @6
 ; EG-NEXT:    ALU 6, @9, KC0[CB0:0-32], KC1[]

 ; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v1, 0
 ; GCN-NOHSA-SI-NEXT:    s_mov_b32 s2, 0xffff
 ; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v3, v1
 ; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v5, v1
 ; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v7, v1
 ; GCN-NOHSA-SI-NEXT:    s_mov_b32 s4, s0
 ; GCN-NOHSA-SI-NEXT:    s_mov_b32 s5, s1
 ; GCN-NOHSA-SI-NEXT:    s_waitcnt vmcnt(0)
-; GCN-NOHSA-SI-NEXT:    v_lshrrev_b32_e32 v2, 16, v9
 ; GCN-NOHSA-SI-NEXT:    v_lshrrev_b32_e32 v6, 16, v8
+; GCN-NOHSA-SI-NEXT:    v_lshrrev_b32_e32 v2, 16, v9
 ; GCN-NOHSA-SI-NEXT:    v_and_b32_e32 v4, s2, v8
 ; GCN-NOHSA-SI-NEXT:    v_and_b32_e32 v0, s2, v9
 ; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[4:7], 0 offset:16
 ; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[4:7], off, s[4:7], 0
 ; GCN-NOHSA-SI-NEXT:    s_endpgm
 ;
 ; GCN-HSA-LABEL: global_zextload_v4i16_to_v4i64:
 ; GCN-HSA:       ; %bb.0:

 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s3, 0xf000
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s2, -1
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s10, s2
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s11, s3
 ; GCN-NOHSA-VI-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s8, s6
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s9, s7
 ; GCN-NOHSA-VI-NEXT:    buffer_load_dwordx2 v[8:9], off, s[8:11], 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, 0
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s6, 0xffff
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, 0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v7, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, v1
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s0, s4
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s1, s5
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, 0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, v5
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, v1
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v7, v1
 ; GCN-NOHSA-VI-NEXT:    s_waitcnt vmcnt(0)
-; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v4, s6, v9
-; GCN-NOHSA-VI-NEXT:    v_lshrrev_b32_e32 v6, 16, v9
-; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v0, s6, v8
-; GCN-NOHSA-VI-NEXT:    v_lshrrev_b32_e32 v2, 16, v8
-; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[4:7], off, s[0:3], 0 offset:16
-; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0
+; GCN-NOHSA-VI-NEXT:    v_lshrrev_b32_e32 v2, 16, v9
+; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v0, s6, v9
+; GCN-NOHSA-VI-NEXT:    v_lshrrev_b32_e32 v6, 16, v8
+; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v4, s6, v8
+; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:16
+; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[4:7], off, s[0:3], 0
 ; GCN-NOHSA-VI-NEXT:    s_endpgm
 ;
 ; EG-LABEL: global_zextload_v4i16_to_v4i64:
 ; EG:       ; %bb.0:
 ; EG-NEXT:    ALU 0, @8, KC0[CB0:0-32], KC1[]
 ; EG-NEXT:    TEX 0 @6
 ; EG-NEXT:    ALU 18, @9, KC0[CB0:0-32], KC1[]
 ; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T5.XYZW, T8.X, 0

 }
 
 define amdgpu_kernel void @global_zextload_v8i16_to_v8i64(<8 x i64> addrspace(1)* %out, <8 x i16> addrspace(1)* %in) #0 {
 ; GCN-NOHSA-SI-LABEL: global_zextload_v8i16_to_v8i64:
 ; GCN-NOHSA-SI:       ; %bb.0:
 ; GCN-NOHSA-SI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
 ; GCN-NOHSA-SI-NEXT:    s_mov_b32 s7, 0xf000
 ; GCN-NOHSA-SI-NEXT:    s_mov_b32 s6, -1
-; GCN-NOHSA-SI-NEXT:    s_mov_b32 s12, 0xffff
 ; GCN-NOHSA-SI-NEXT:    s_mov_b32 s10, s6
 ; GCN-NOHSA-SI-NEXT:    s_mov_b32 s11, s7
 ; GCN-NOHSA-SI-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-NOHSA-SI-NEXT:    s_mov_b32 s8, s2
 ; GCN-NOHSA-SI-NEXT:    s_mov_b32 s9, s3
 ; GCN-NOHSA-SI-NEXT:    buffer_load_dwordx4 v[0:3], off, s[8:11], 0
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v7, 0
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v9, 0
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v15, 0
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v11, v9
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v17, v9
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v19, v9
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v13, v9
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v5, v9
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v4, 0
+; GCN-NOHSA-SI-NEXT:    s_mov_b32 s2, 0xffff
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v6, v4
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v8, v4
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v10, v4
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v12, v4
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v14, v4
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v16, v4
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v18, v4
 ; GCN-NOHSA-SI-NEXT:    s_mov_b32 s4, s0
 ; GCN-NOHSA-SI-NEXT:    s_mov_b32 s5, s1
 ; GCN-NOHSA-SI-NEXT:    s_waitcnt vmcnt(0)
-; GCN-NOHSA-SI-NEXT:    v_lshrrev_b32_e32 v18, 16, v1
-; GCN-NOHSA-SI-NEXT:    v_lshrrev_b32_e32 v10, 16, v3
-; GCN-NOHSA-SI-NEXT:    v_lshrrev_b32_e32 v14, 16, v2
-; GCN-NOHSA-SI-NEXT:    v_lshrrev_b32_e32 v6, 16, v0
-; GCN-NOHSA-SI-NEXT:    v_and_b32_e32 v4, s12, v0
-; GCN-NOHSA-SI-NEXT:    v_and_b32_e32 v12, s12, v2
-; GCN-NOHSA-SI-NEXT:    v_and_b32_e32 v16, s12, v1
-; GCN-NOHSA-SI-NEXT:    v_and_b32_e32 v8, s12, v3
-; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[8:11], off, s[4:7], 0 offset:48
-; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[16:19], off, s[4:7], 0 offset:16
-; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[12:15], off, s[4:7], 0 offset:32
-; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[4:7], off, s[4:7], 0
+; GCN-NOHSA-SI-NEXT:    v_lshrrev_b32_e32 v17, 16, v0
+; GCN-NOHSA-SI-NEXT:    v_lshrrev_b32_e32 v13, 16, v1
+; GCN-NOHSA-SI-NEXT:    v_lshrrev_b32_e32 v9, 16, v2
+; GCN-NOHSA-SI-NEXT:    v_lshrrev_b32_e32 v5, 16, v3
+; GCN-NOHSA-SI-NEXT:    v_and_b32_e32 v15, s2, v0
+; GCN-NOHSA-SI-NEXT:    v_and_b32_e32 v11, s2, v1
+; GCN-NOHSA-SI-NEXT:    v_and_b32_e32 v7, s2, v2
+; GCN-NOHSA-SI-NEXT:    v_and_b32_e32 v3, s2, v3
+; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[3:6], off, s[4:7], 0 offset:48
+; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[7:10], off, s[4:7], 0 offset:32
+; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[11:14], off, s[4:7], 0 offset:16
+; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[15:18], off, s[4:7], 0
 ; GCN-NOHSA-SI-NEXT:    s_endpgm
 ;
 ; GCN-HSA-LABEL: global_zextload_v8i16_to_v8i64:
 ; GCN-HSA:       ; %bb.0:
 ; GCN-HSA-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v12, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v4, 0
 ; GCN-HSA-NEXT:    s_mov_b32 s4, 0xffff
-; GCN-HSA-NEXT:    v_mov_b32_e32 v14, v12
-; GCN-HSA-NEXT:    v_mov_b32_e32 v15, v12
+; GCN-HSA-NEXT:    v_mov_b32_e32 v6, v4
+; GCN-HSA-NEXT:    v_mov_b32_e32 v7, v4
 ; GCN-HSA-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s2
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s3
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 48
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
-; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
-; GCN-HSA-NEXT:    s_add_u32 s2, s0, 16
+; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s3
+; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s2
+; GCN-HSA-NEXT:    s_add_u32 s2, s0, 32
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v17, s1
-; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s0
-; GCN-HSA-NEXT:    s_add_u32 s0, s0, 32
+; GCN-HSA-NEXT:    v_mov_b32_e32 v9, v4
+; GCN-HSA-NEXT:    v_mov_b32_e32 v11, v4
+; GCN-HSA-NEXT:    v_mov_b32_e32 v13, v4
+; GCN-HSA-NEXT:    v_mov_b32_e32 v17, v4
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v19, s3
-; GCN-HSA-NEXT:    v_mov_b32_e32 v8, v12
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v18, s2
-; GCN-HSA-NEXT:    s_addc_u32 s1, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v6, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v10, 0
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(0)
-; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v13, 16, v3
-; GCN-HSA-NEXT:    v_and_b32_e32 v11, s4, v3
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[11:14]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v4, v12
-; GCN-HSA-NEXT:    v_mov_b32_e32 v13, v12
-; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v14, 16, v1
-; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v5, 16, v0
-; GCN-HSA-NEXT:    v_and_b32_e32 v3, s4, v0
-; GCN-HSA-NEXT:    v_and_b32_e32 v12, s4, v1
+; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v5, 16, v3
+; GCN-HSA-NEXT:    v_and_b32_e32 v3, s4, v3
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[14:15], v[3:6]
+; GCN-HSA-NEXT:    v_mov_b32_e32 v15, v4
+; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s1
+; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s0
+; GCN-HSA-NEXT:    s_add_u32 s0, s0, 16
+; GCN-HSA-NEXT:    s_addc_u32 s1, s1, 0
+; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v16, 16, v0
+; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v12, 16, v1
+; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v8, 16, v2
+; GCN-HSA-NEXT:    v_and_b32_e32 v14, s4, v0
+; GCN-HSA-NEXT:    v_and_b32_e32 v10, s4, v1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s0
-; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v9, 16, v2
-; GCN-HSA-NEXT:    v_and_b32_e32 v7, s4, v2
+; GCN-HSA-NEXT:    v_and_b32_e32 v6, s4, v2
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s1
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[18:19], v[12:15]
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[0:1], v[7:10]
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[16:17], v[3:6]
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[18:19], v[6:9]
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[0:1], v[10:13]
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[3:4], v[14:17]
 ; GCN-HSA-NEXT:    s_endpgm
 ;
 ; GCN-NOHSA-VI-LABEL: global_zextload_v8i16_to_v8i64:
 ; GCN-NOHSA-VI:       ; %bb.0:
 ; GCN-NOHSA-VI-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s3, 0xf000
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s2, -1
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s10, s2
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s11, s3
 ; GCN-NOHSA-VI-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s8, s6
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s9, s7
 ; GCN-NOHSA-VI-NEXT:    buffer_load_dwordx4 v[0:3], off, s[8:11], 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, 0
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s6, 0xffff
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v17, 0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v19, 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v6, v4
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s0, s4
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s1, s5
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v7, 0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v11, 0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v15, 0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v13, v17
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v9, v17
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, v17
-; GCN-NOHSA-VI-NEXT:    s_waitcnt vmcnt(0)
-; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v16, s6, v3
-; GCN-NOHSA-VI-NEXT:    v_lshrrev_b32_e32 v18, 16, v3
-; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v4, s6, v0
-; GCN-NOHSA-VI-NEXT:    v_lshrrev_b32_e32 v6, 16, v0
-; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v8, s6, v1
-; GCN-NOHSA-VI-NEXT:    v_lshrrev_b32_e32 v10, 16, v1
-; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v12, s6, v2
-; GCN-NOHSA-VI-NEXT:    v_lshrrev_b32_e32 v14, 16, v2
-; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[16:19], off, s[0:3], 0 offset:48
-; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[12:15], off, s[0:3], 0 offset:32
-; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[8:11], off, s[0:3], 0 offset:16
-; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[4:7], off, s[0:3], 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v8, v4
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v10, v4
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v12, v4
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v14, v4
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v16, v4
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v18, v4
+; GCN-NOHSA-VI-NEXT:    s_waitcnt vmcnt(0)
+; GCN-NOHSA-VI-NEXT:    v_lshrrev_b32_e32 v5, 16, v3
+; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v3, s6, v3
+; GCN-NOHSA-VI-NEXT:    v_lshrrev_b32_e32 v17, 16, v0
+; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v15, s6, v0
+; GCN-NOHSA-VI-NEXT:    v_lshrrev_b32_e32 v13, 16, v1
+; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v11, s6, v1
+; GCN-NOHSA-VI-NEXT:    v_lshrrev_b32_e32 v9, 16, v2
+; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v7, s6, v2
+; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[3:6], off, s[0:3], 0 offset:48
+; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[7:10], off, s[0:3], 0 offset:32
+; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[11:14], off, s[0:3], 0 offset:16
+; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[15:18], off, s[0:3], 0
 ; GCN-NOHSA-VI-NEXT:    s_endpgm
 ;
 ; EG-LABEL: global_zextload_v8i16_to_v8i64:
 ; EG:       ; %bb.0:
 ; EG-NEXT:    ALU 0, @10, KC0[CB0:0-32], KC1[]
 ; EG-NEXT:    TEX 0 @8
 ; EG-NEXT:    ALU 30, @11, KC0[CB0:0-32], KC1[]
 ; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T8.XYZW, T14.X, 0

 ; GCN-NOHSA-SI-NEXT:    s_mov_b32 s2, -1
 ; GCN-NOHSA-SI-NEXT:    s_mov_b32 s10, s2
 ; GCN-NOHSA-SI-NEXT:    s_mov_b32 s11, s3
 ; GCN-NOHSA-SI-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-NOHSA-SI-NEXT:    s_mov_b32 s8, s6
 ; GCN-NOHSA-SI-NEXT:    s_mov_b32 s9, s7
 ; GCN-NOHSA-SI-NEXT:    buffer_load_dwordx4 v[0:3], off, s[8:11], 0
 ; GCN-NOHSA-SI-NEXT:    s_mov_b32 s0, 0xffff
-; GCN-NOHSA-SI-NEXT:    buffer_load_dwordx4 v[4:7], off, s[8:11], 0 offset:16
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v4, 0
+; GCN-NOHSA-SI-NEXT:    buffer_load_dwordx4 v[6:9], off, s[8:11], 0 offset:16
 ; GCN-NOHSA-SI-NEXT:    s_waitcnt vmcnt(1)
-; GCN-NOHSA-SI-NEXT:    v_lshrrev_b32_e32 v10, 16, v1
-; GCN-NOHSA-SI-NEXT:    v_lshrrev_b32_e32 v14, 16, v2
-; GCN-NOHSA-SI-NEXT:    v_lshrrev_b32_e32 v18, 16, v0
-; GCN-NOHSA-SI-NEXT:    v_and_b32_e32 v16, s0, v0
-; GCN-NOHSA-SI-NEXT:    v_and_b32_e32 v12, s0, v2
-; GCN-NOHSA-SI-NEXT:    v_and_b32_e32 v8, s0, v1
-; GCN-NOHSA-SI-NEXT:    v_lshrrev_b32_e32 v2, 16, v3
-; GCN-NOHSA-SI-NEXT:    v_and_b32_e32 v0, s0, v3
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v20, 0
+; GCN-NOHSA-SI-NEXT:    v_lshrrev_b32_e32 v12, 16, v0
+; GCN-NOHSA-SI-NEXT:    v_lshrrev_b32_e32 v16, 16, v1
+; GCN-NOHSA-SI-NEXT:    v_lshrrev_b32_e32 v20, 16, v2
+; GCN-NOHSA-SI-NEXT:    v_and_b32_e32 v10, s0, v0
+; GCN-NOHSA-SI-NEXT:    v_and_b32_e32 v14, s0, v1
+; GCN-NOHSA-SI-NEXT:    v_and_b32_e32 v18, s0, v2
+; GCN-NOHSA-SI-NEXT:    v_lshrrev_b32_e32 v24, 16, v3
+; GCN-NOHSA-SI-NEXT:    v_and_b32_e32 v22, s0, v3
 ; GCN-NOHSA-SI-NEXT:    s_waitcnt vmcnt(0)
-; GCN-NOHSA-SI-NEXT:    v_lshrrev_b32_e32 v21, 16, v5
-; GCN-NOHSA-SI-NEXT:    v_lshrrev_b32_e32 v25, 16, v6
-; GCN-NOHSA-SI-NEXT:    v_and_b32_e32 v23, s0, v6
-; GCN-NOHSA-SI-NEXT:    v_lshrrev_b32_e32 v6, 16, v4
-; GCN-NOHSA-SI-NEXT:    v_and_b32_e32 v4, s0, v4
-; GCN-NOHSA-SI-NEXT:    v_lshrrev_b32_e32 v28, 16, v7
-; GCN-NOHSA-SI-NEXT:    v_and_b32_e32 v26, s0, v7
-; GCN-NOHSA-SI-NEXT:    v_and_b32_e32 v19, s0, v5
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v22, v20
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v27, v20
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v29, v20
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v1, v20
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v3, v20
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v9, v20
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v11, v20
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v5, v20
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v24, v20
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v13, v20
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v17, v20
+; GCN-NOHSA-SI-NEXT:    v_lshrrev_b32_e32 v5, 16, v7
+; GCN-NOHSA-SI-NEXT:    v_lshrrev_b32_e32 v28, 16, v8
+; GCN-NOHSA-SI-NEXT:    v_and_b32_e32 v26, s0, v8
+; GCN-NOHSA-SI-NEXT:    v_lshrrev_b32_e32 v32, 16, v9
+; GCN-NOHSA-SI-NEXT:    v_and_b32_e32 v30, s0, v9
+; GCN-NOHSA-SI-NEXT:    v_lshrrev_b32_e32 v36, 16, v6
+; GCN-NOHSA-SI-NEXT:    v_and_b32_e32 v34, s0, v6
+; GCN-NOHSA-SI-NEXT:    v_and_b32_e32 v3, s0, v7
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v6, v4
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v35, v4
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v37, v4
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v31, v4
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v33, v4
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v27, v4
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v29, v4
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v23, v4
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v25, v4
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v19, v4
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v21, v4
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v15, v4
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v17, v4
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v11, v4
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v13, v4
 ; GCN-NOHSA-SI-NEXT:    s_mov_b32 s0, s4
 ; GCN-NOHSA-SI-NEXT:    s_mov_b32 s1, s5
-; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[19:22], off, s[0:3], 0 offset:80
-; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[26:29], off, s[0:3], 0 offset:112
-; GCN-NOHSA-SI-NEXT:    s_waitcnt expcnt(1)
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v19, 0
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v15, 0
-; GCN-NOHSA-SI-NEXT:    s_waitcnt expcnt(0)
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v26, 0
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v7, 0
-; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:48
-; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[8:11], off, s[0:3], 0 offset:16
-; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[4:7], off, s[0:3], 0 offset:64
-; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[23:26], off, s[0:3], 0 offset:96
-; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[12:15], off, s[0:3], 0 offset:32
-; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[16:19], off, s[0:3], 0
+; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[3:6], off, s[0:3], 0 offset:80
+; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[34:37], off, s[0:3], 0 offset:64
+; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[30:33], off, s[0:3], 0 offset:112
+; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[26:29], off, s[0:3], 0 offset:96
+; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[22:25], off, s[0:3], 0 offset:48
+; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[18:21], off, s[0:3], 0 offset:32
+; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[14:17], off, s[0:3], 0 offset:16
+; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[10:13], off, s[0:3], 0
 ; GCN-NOHSA-SI-NEXT:    s_endpgm
 ;
 ; GCN-HSA-LABEL: global_zextload_v16i16_to_v16i64:
 ; GCN-HSA:       ; %bb.0:
 ; GCN-HSA-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v8, 0
 ; GCN-HSA-NEXT:    s_mov_b32 s6, 0xffff
-; GCN-HSA-NEXT:    v_mov_b32_e32 v10, v8
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v12, v8
+; GCN-HSA-NEXT:    v_mov_b32_e32 v14, v8
 ; GCN-HSA-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-HSA-NEXT:    s_add_u32 s4, s2, 16
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s5, s3, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s4
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s2
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s5
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[4:7], v[4:5]
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 48
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-HSA-NEXT:    s_add_u32 s4, s0, 16
+; GCN-HSA-NEXT:    s_add_u32 s4, s0, 32
 ; GCN-HSA-NEXT:    s_addc_u32 s5, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s5
-; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s4
-; GCN-HSA-NEXT:    s_add_u32 s4, s0, 0x50
+; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s5
+; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s4
+; GCN-HSA-NEXT:    s_add_u32 s4, s0, 16
 ; GCN-HSA-NEXT:    s_addc_u32 s5, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s3
-; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s2
+; GCN-HSA-NEXT:    v_mov_b32_e32 v10, v8
+; GCN-HSA-NEXT:    v_mov_b32_e32 v17, v8
 ; GCN-HSA-NEXT:    s_waitcnt vmcnt(0)
-; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v11, 16, v5
-; GCN-HSA-NEXT:    v_and_b32_e32 v9, s6, v5
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[13:14], v[9:12]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s5
-; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s4
-; GCN-HSA-NEXT:    s_add_u32 s4, s0, 0x70
+; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v13, 16, v6
+; GCN-HSA-NEXT:    v_and_b32_e32 v11, s6, v6
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[15:16], v[11:14]
+; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s5
+; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s4
+; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v13, 16, v5
+; GCN-HSA-NEXT:    v_and_b32_e32 v11, s6, v5
+; GCN-HSA-NEXT:    v_mov_b32_e32 v6, s1
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[15:16], v[11:14]
+; GCN-HSA-NEXT:    s_add_u32 s4, s0, 0x50
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s0
+; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v13, 16, v4
+; GCN-HSA-NEXT:    v_and_b32_e32 v11, s6, v4
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[5:6], v[11:14]
 ; GCN-HSA-NEXT:    s_addc_u32 s5, s1, 0
-; GCN-HSA-NEXT:    s_add_u32 s2, s0, 32
-; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v11, 16, v1
-; GCN-HSA-NEXT:    v_and_b32_e32 v9, s6, v1
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[13:14], v[9:12]
-; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s5
-; GCN-HSA-NEXT:    v_mov_b32_e32 v18, s3
-; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s4
-; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v11, 16, v3
-; GCN-HSA-NEXT:    v_and_b32_e32 v9, s6, v3
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[13:14], v[9:12]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v17, s2
+; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s4
+; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s5
+; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v13, 16, v1
+; GCN-HSA-NEXT:    v_and_b32_e32 v11, s6, v1
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[11:14]
+; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s3
+; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s0, 64
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v9, 16, v7
 ; GCN-HSA-NEXT:    v_and_b32_e32 v7, s6, v7
-; GCN-HSA-NEXT:    v_mov_b32_e32 v14, 0
-; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v13, 16, v6
-; GCN-HSA-NEXT:    v_and_b32_e32 v11, s6, v6
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[15:16], v[7:10]
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[17:18], v[11:14]
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s1
-; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s0
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[14:15], v[7:10]
+; GCN-HSA-NEXT:    v_mov_b32_e32 v4, v8
+; GCN-HSA-NEXT:    v_mov_b32_e32 v6, v8
+; GCN-HSA-NEXT:    v_mov_b32_e32 v11, v8
+; GCN-HSA-NEXT:    v_mov_b32_e32 v13, v8
+; GCN-HSA-NEXT:    v_mov_b32_e32 v15, v8
+; GCN-HSA-NEXT:    v_mov_b32_e32 v8, s3
+; GCN-HSA-NEXT:    v_mov_b32_e32 v7, s2
+; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x70
+; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    s_add_u32 s0, s0, 0x60
+; GCN-HSA-NEXT:    v_mov_b32_e32 v19, s3
 ; GCN-HSA-NEXT:    s_addc_u32 s1, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v18, s3
-; GCN-HSA-NEXT:    v_mov_b32_e32 v1, v8
-; GCN-HSA-NEXT:    v_mov_b32_e32 v7, v8
-; GCN-HSA-NEXT:    v_mov_b32_e32 v3, v14
-; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v8, 16, v0
-; GCN-HSA-NEXT:    v_and_b32_e32 v6, s6, v0
-; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v13, 16, v2
-; GCN-HSA-NEXT:    v_and_b32_e32 v11, s6, v2
-; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v2, 16, v4
-; GCN-HSA-NEXT:    v_and_b32_e32 v0, s6, v4
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, s1
-; GCN-HSA-NEXT:    v_mov_b32_e32 v9, v14
-; GCN-HSA-NEXT:    v_mov_b32_e32 v17, s2
-; GCN-HSA-NEXT:    v_mov_b32_e32 v4, s0
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[15:16], v[0:3]
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[17:18], v[6:9]
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[4:5], v[11:14]
+; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v12, 16, v3
+; GCN-HSA-NEXT:    v_and_b32_e32 v10, s6, v3
+; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v5, 16, v0
+; GCN-HSA-NEXT:    v_and_b32_e32 v3, s6, v0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v18, s2
+; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v16, 16, v2
+; GCN-HSA-NEXT:    v_and_b32_e32 v14, s6, v2
+; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s1
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[7:8], v[3:6]
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[18:19], v[10:13]
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[0:1], v[14:17]
 ; GCN-HSA-NEXT:    s_endpgm
 ;
 ; GCN-NOHSA-VI-LABEL: global_zextload_v16i16_to_v16i64:
 ; GCN-NOHSA-VI:       ; %bb.0:
 ; GCN-NOHSA-VI-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s3, 0xf000
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s2, -1
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s10, s2
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s11, s3
 ; GCN-NOHSA-VI-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s8, s6
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s9, s7
 ; GCN-NOHSA-VI-NEXT:    buffer_load_dwordx4 v[0:3], off, s[8:11], 0
 ; GCN-NOHSA-VI-NEXT:    buffer_load_dwordx4 v[4:7], off, s[8:11], 0 offset:16
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s0, 0xffff
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v28, 0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v30, 0
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s1, s5
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v25, v28
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v21, v28
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v17, v28
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v13, v28
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v9, v28
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v11, 0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v15, 0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v19, 0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v23, 0
 ; GCN-NOHSA-VI-NEXT:    s_waitcnt vmcnt(1)
-; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v8, s0, v0
 ; GCN-NOHSA-VI-NEXT:    v_lshrrev_b32_e32 v10, 16, v0
-; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v0, s0, v2
-; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v12, s0, v1
-; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v16, s0, v3
-; GCN-NOHSA-VI-NEXT:    v_lshrrev_b32_e32 v18, 16, v3
 ; GCN-NOHSA-VI-NEXT:    s_waitcnt vmcnt(0)
-; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v20, s0, v4
+; GCN-NOHSA-VI-NEXT:    v_lshrrev_b32_e32 v26, 16, v5
 ; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v24, s0, v5
-; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v3, s0, v6
-; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v27, s0, v7
-; GCN-NOHSA-VI-NEXT:    v_lshrrev_b32_e32 v29, 16, v7
-; GCN-NOHSA-VI-NEXT:    s_mov_b32 s0, s4
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, 0
+; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v8, s0, v0
+; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v12, s0, v1
+; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v16, s0, v2
+; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v0, s0, v3
 ; GCN-NOHSA-VI-NEXT:    v_lshrrev_b32_e32 v22, 16, v4
-; GCN-NOHSA-VI-NEXT:    v_lshrrev_b32_e32 v26, 16, v5
-; GCN-NOHSA-VI-NEXT:    v_lshrrev_b32_e32 v5, 16, v6
-; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[27:30], off, s[0:3], 0 offset:112
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, v28
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v6, 0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v27, 0
+; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v20, s0, v4
+; GCN-NOHSA-VI-NEXT:    v_lshrrev_b32_e32 v30, 16, v6
+; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v28, s0, v6
+; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v4, s0, v7
+; GCN-NOHSA-VI-NEXT:    v_lshrrev_b32_e32 v6, 16, v7
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v7, v5
+; GCN-NOHSA-VI-NEXT:    s_mov_b32 s0, s4
 ; GCN-NOHSA-VI-NEXT:    v_lshrrev_b32_e32 v14, 16, v1
-; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[3:6], off, s[0:3], 0 offset:96
-; GCN-NOHSA-VI-NEXT:    v_lshrrev_b32_e32 v2, 16, v2
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, v28
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, 0
+; GCN-NOHSA-VI-NEXT:    v_lshrrev_b32_e32 v18, 16, v2
+; GCN-NOHSA-VI-NEXT:    v_lshrrev_b32_e32 v2, 16, v3
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v29, v5
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v31, v5
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v25, v5
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v27, v5
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v21, v5
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v23, v5
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v1, v5
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v3, v5
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v17, v5
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v19, v5
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v13, v5
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v15, v5
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v9, v5
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v11, v5
+; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[4:7], off, s[0:3], 0 offset:112
+; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[28:31], off, s[0:3], 0 offset:96
 ; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[24:27], off, s[0:3], 0 offset:80
 ; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[20:23], off, s[0:3], 0 offset:64
-; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[16:19], off, s[0:3], 0 offset:48
-; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:32
+; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:48
+; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[16:19], off, s[0:3], 0 offset:32
 ; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[12:15], off, s[0:3], 0 offset:16
 ; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[8:11], off, s[0:3], 0
 ; GCN-NOHSA-VI-NEXT:    s_endpgm
 ;
 ; EG-LABEL: global_zextload_v16i16_to_v16i64:
 ; EG:       ; %bb.0:
 ; EG-NEXT:    ALU 0, @16, KC0[CB0:0-32], KC1[]
 ; EG-NEXT:    TEX 1 @12

 ; GCN-NOHSA-SI-NEXT:    s_mov_b32 s14, -1
 ; GCN-NOHSA-SI-NEXT:    s_mov_b32 s15, 0xe8f000
 ; GCN-NOHSA-SI-NEXT:    s_add_u32 s12, s12, s3
 ; GCN-NOHSA-SI-NEXT:    s_addc_u32 s13, s13, 0
 ; GCN-NOHSA-SI-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x9
 ; GCN-NOHSA-SI-NEXT:    s_mov_b32 s3, 0xf000
 ; GCN-NOHSA-SI-NEXT:    s_mov_b32 s2, -1
 ; GCN-NOHSA-SI-NEXT:    s_mov_b32 s0, 0xffff
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v1, 0
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v27, 0
 ; GCN-NOHSA-SI-NEXT:    s_mov_b32 s10, s2
 ; GCN-NOHSA-SI-NEXT:    s_mov_b32 s11, s3
 ; GCN-NOHSA-SI-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-NOHSA-SI-NEXT:    s_mov_b32 s8, s6
 ; GCN-NOHSA-SI-NEXT:    s_mov_b32 s9, s7
 ; GCN-NOHSA-SI-NEXT:    buffer_load_dwordx4 v[2:5], off, s[8:11], 0
 ; GCN-NOHSA-SI-NEXT:    buffer_load_dwordx4 v[6:9], off, s[8:11], 0 offset:16
 ; GCN-NOHSA-SI-NEXT:    buffer_load_dwordx4 v[10:13], off, s[8:11], 0 offset:32
 ; GCN-NOHSA-SI-NEXT:    buffer_load_dwordx4 v[14:17], off, s[8:11], 0 offset:48
 ; GCN-NOHSA-SI-NEXT:    s_waitcnt vmcnt(3)
-; GCN-NOHSA-SI-NEXT:    v_lshrrev_b32_e32 v23, 16, v3
-; GCN-NOHSA-SI-NEXT:    v_lshrrev_b32_e32 v21, 16, v4
-; GCN-NOHSA-SI-NEXT:    v_lshrrev_b32_e32 v20, 16, v2
-; GCN-NOHSA-SI-NEXT:    v_and_b32_e32 v18, s0, v2
+; GCN-NOHSA-SI-NEXT:    v_lshrrev_b32_e32 v20, 16, v3
+; GCN-NOHSA-SI-NEXT:    v_lshrrev_b32_e32 v21, 16, v2
+; GCN-NOHSA-SI-NEXT:    v_lshrrev_b32_e32 v23, 16, v5
+; GCN-NOHSA-SI-NEXT:    v_and_b32_e32 v18, s0, v3
 ; GCN-NOHSA-SI-NEXT:    buffer_store_dword v18, off, s[12:15], 0 offset:4 ; 4-byte Folded Spill
 ; GCN-NOHSA-SI-NEXT:    s_waitcnt vmcnt(0)
 ; GCN-NOHSA-SI-NEXT:    buffer_store_dword v19, off, s[12:15], 0 offset:8 ; 4-byte Folded Spill
 ; GCN-NOHSA-SI-NEXT:    buffer_store_dword v20, off, s[12:15], 0 offset:12 ; 4-byte Folded Spill
 ; GCN-NOHSA-SI-NEXT:    buffer_store_dword v21, off, s[12:15], 0 offset:16 ; 4-byte Folded Spill
-; GCN-NOHSA-SI-NEXT:    s_waitcnt expcnt(2)
-; GCN-NOHSA-SI-NEXT:    v_and_b32_e32 v19, s0, v4
+; GCN-NOHSA-SI-NEXT:    v_and_b32_e32 v0, s0, v2
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v2, v21
+; GCN-NOHSA-SI-NEXT:    v_and_b32_e32 v22, s0, v5
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v24, v23
+; GCN-NOHSA-SI-NEXT:    v_lshrrev_b32_e32 v32, 16, v4
+; GCN-NOHSA-SI-NEXT:    v_and_b32_e32 v30, s0, v4
+; GCN-NOHSA-SI-NEXT:    v_lshrrev_b32_e32 v36, 16, v7
+; GCN-NOHSA-SI-NEXT:    v_and_b32_e32 v34, s0, v7
+; GCN-NOHSA-SI-NEXT:    s_waitcnt expcnt(1)
+; GCN-NOHSA-SI-NEXT:    v_lshrrev_b32_e32 v20, 16, v6
+; GCN-NOHSA-SI-NEXT:    v_and_b32_e32 v18, s0, v6
+; GCN-NOHSA-SI-NEXT:    v_lshrrev_b32_e32 v40, 16, v9
+; GCN-NOHSA-SI-NEXT:    v_and_b32_e32 v38, s0, v9
+; GCN-NOHSA-SI-NEXT:    v_lshrrev_b32_e32 v44, 16, v8
+; GCN-NOHSA-SI-NEXT:    v_and_b32_e32 v42, s0, v8
+; GCN-NOHSA-SI-NEXT:    v_lshrrev_b32_e32 v48, 16, v11
+; GCN-NOHSA-SI-NEXT:    v_and_b32_e32 v46, s0, v11
+; GCN-NOHSA-SI-NEXT:    v_lshrrev_b32_e32 v11, 16, v10
+; GCN-NOHSA-SI-NEXT:    v_and_b32_e32 v9, s0, v10
+; GCN-NOHSA-SI-NEXT:    v_lshrrev_b32_e32 v52, 16, v13
+; GCN-NOHSA-SI-NEXT:    v_and_b32_e32 v50, s0, v13
+; GCN-NOHSA-SI-NEXT:    v_lshrrev_b32_e32 v56, 16, v12
+; GCN-NOHSA-SI-NEXT:    v_and_b32_e32 v54, s0, v12
+; GCN-NOHSA-SI-NEXT:    v_lshrrev_b32_e32 v28, 16, v16
+; GCN-NOHSA-SI-NEXT:    v_lshrrev_b32_e32 v60, 16, v15
+; GCN-NOHSA-SI-NEXT:    v_and_b32_e32 v58, s0, v15
+; GCN-NOHSA-SI-NEXT:    v_lshrrev_b32_e32 v15, 16, v14
+; GCN-NOHSA-SI-NEXT:    v_and_b32_e32 v13, s0, v14
+; GCN-NOHSA-SI-NEXT:    v_lshrrev_b32_e32 v6, 16, v17
+; GCN-NOHSA-SI-NEXT:    v_and_b32_e32 v4, s0, v17
+; GCN-NOHSA-SI-NEXT:    v_and_b32_e32 v26, s0, v16
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v29, v27
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v5, v27
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v7, v27
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v14, v27
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v16, v27
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v59, v27
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v61, v27
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v55, v27
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v57, v27
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v51, v27
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v53, v27
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v10, v27
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v12, v27
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v47, v27
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v49, v27
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v43, v27
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v45, v27
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v39, v27
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v41, v27
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v19, v27
 ; GCN-NOHSA-SI-NEXT:    s_waitcnt expcnt(0)
-; GCN-NOHSA-SI-NEXT:    buffer_store_dword v19, off, s[12:15], 0 offset:20 ; 4-byte Folded Spill
-; GCN-NOHSA-SI-NEXT:    s_waitcnt vmcnt(0)
-; GCN-NOHSA-SI-NEXT:    buffer_store_dword v20, off, s[12:15], 0 offset:24 ; 4-byte Folded Spill
-; GCN-NOHSA-SI-NEXT:    buffer_store_dword v21, off, s[12:15], 0 offset:28 ; 4-byte Folded Spill
-; GCN-NOHSA-SI-NEXT:    buffer_store_dword v22, off, s[12:15], 0 offset:32 ; 4-byte Folded Spill
-; GCN-NOHSA-SI-NEXT:    v_and_b32_e32 v4, s0, v3
-; GCN-NOHSA-SI-NEXT:    v_lshrrev_b32_e32 v32, 16, v5
-; GCN-NOHSA-SI-NEXT:    v_and_b32_e32 v30, s0, v5
-; GCN-NOHSA-SI-NEXT:    v_lshrrev_b32_e32 v36, 16, v6
-; GCN-NOHSA-SI-NEXT:    v_and_b32_e32 v34, s0, v6
-; GCN-NOHSA-SI-NEXT:    v_lshrrev_b32_e32 v28, 16, v8
-; GCN-NOHSA-SI-NEXT:    v_and_b32_e32 v26, s0, v8
-; GCN-NOHSA-SI-NEXT:    v_lshrrev_b32_e32 v40, 16, v7
-; GCN-NOHSA-SI-NEXT:    v_and_b32_e32 v38, s0, v7
-; GCN-NOHSA-SI-NEXT:    v_lshrrev_b32_e32 v44, 16, v9
-; GCN-NOHSA-SI-NEXT:    v_and_b32_e32 v42, s0, v9
-; GCN-NOHSA-SI-NEXT:    v_lshrrev_b32_e32 v48, 16, v10
-; GCN-NOHSA-SI-NEXT:    v_and_b32_e32 v46, s0, v10
-; GCN-NOHSA-SI-NEXT:    v_lshrrev_b32_e32 v24, 16, v12
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v21, v27
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v35, v27
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v37, v27
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v31, v27
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v33, v27
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v23, v27
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v25, v27
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v1, v27
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v3, v27
+; GCN-NOHSA-SI-NEXT:    buffer_store_dword v0, off, s[12:15], 0 offset:20 ; 4-byte Folded Spill
+; GCN-NOHSA-SI-NEXT:    s_waitcnt vmcnt(0)
+; GCN-NOHSA-SI-NEXT:    buffer_store_dword v1, off, s[12:15], 0 offset:24 ; 4-byte Folded Spill
+; GCN-NOHSA-SI-NEXT:    buffer_store_dword v2, off, s[12:15], 0 offset:28 ; 4-byte Folded Spill
+; GCN-NOHSA-SI-NEXT:    buffer_store_dword v3, off, s[12:15], 0 offset:32 ; 4-byte Folded Spill
 ; GCN-NOHSA-SI-NEXT:    s_waitcnt expcnt(0)
-; GCN-NOHSA-SI-NEXT:    v_and_b32_e32 v22, s0, v12
-; GCN-NOHSA-SI-NEXT:    v_lshrrev_b32_e32 v52, 16, v11
-; GCN-NOHSA-SI-NEXT:    v_and_b32_e32 v50, s0, v11
-; GCN-NOHSA-SI-NEXT:    v_lshrrev_b32_e32 v56, 16, v13
-; GCN-NOHSA-SI-NEXT:    v_and_b32_e32 v54, s0, v13
-; GCN-NOHSA-SI-NEXT:    v_lshrrev_b32_e32 v2, 16, v17
-; GCN-NOHSA-SI-NEXT:    v_lshrrev_b32_e32 v60, 16, v14
-; GCN-NOHSA-SI-NEXT:    v_and_b32_e32 v58, s0, v14
-; GCN-NOHSA-SI-NEXT:    v_lshrrev_b32_e32 v20, 16, v16
-; GCN-NOHSA-SI-NEXT:    v_and_b32_e32 v18, s0, v16
-; GCN-NOHSA-SI-NEXT:    v_lshrrev_b32_e32 v10, 16, v15
-; GCN-NOHSA-SI-NEXT:    v_and_b32_e32 v8, s0, v15
-; GCN-NOHSA-SI-NEXT:    v_and_b32_e32 v0, s0, v17
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v3, v1
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v9, v1
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v11, v1
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v55, v1
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v57, v1
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v51, v1
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v53, v1
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v43, v1
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v45, v1
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v39, v1
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v41, v1
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v31, v1
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v33, v1
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v6, v23
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v5, v1
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v7, v1
-; GCN-NOHSA-SI-NEXT:    buffer_store_dword v4, off, s[12:15], 0 offset:36 ; 4-byte Folded Spill
+; GCN-NOHSA-SI-NEXT:    buffer_load_dword v0, off, s[12:15], 0 offset:4 ; 4-byte Folded Reload
+; GCN-NOHSA-SI-NEXT:    buffer_load_dword v1, off, s[12:15], 0 offset:8 ; 4-byte Folded Reload
+; GCN-NOHSA-SI-NEXT:    buffer_load_dword v2, off, s[12:15], 0 offset:12 ; 4-byte Folded Reload
+; GCN-NOHSA-SI-NEXT:    buffer_load_dword v3, off, s[12:15], 0 offset:16 ; 4-byte Folded Reload
 ; GCN-NOHSA-SI-NEXT:    s_waitcnt vmcnt(0)
-; GCN-NOHSA-SI-NEXT:    buffer_store_dword v5, off, s[12:15], 0 offset:40 ; 4-byte Folded Spill
-; GCN-NOHSA-SI-NEXT:    buffer_store_dword v6, off, s[12:15], 0 offset:44 ; 4-byte Folded Spill
-; GCN-NOHSA-SI-NEXT:    buffer_store_dword v7, off, s[12:15], 0 offset:48 ; 4-byte Folded Spill
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v19, v1
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v59, v1
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v23, v1
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v47, v1
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v27, v1
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v35, v1
-; GCN-NOHSA-SI-NEXT:    s_waitcnt expcnt(0)
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v1, v27
+; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v3, v27
+; GCN-NOHSA-SI-NEXT:    s_mov_b32 s0, s4
+; GCN-NOHSA-SI-NEXT:    s_mov_b32 s1, s5
+; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[26:29], off, s[0:3], 0 offset:224
+; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[4:7], off, s[0:3], 0 offset:240
+; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[13:16], off, s[0:3], 0 offset:192
+; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[58:61], off, s[0:3], 0 offset:208
+; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[54:57], off, s[0:3], 0 offset:160
+; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[50:53], off, s[0:3], 0 offset:176
+; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[9:12], off, s[0:3], 0 offset:128
+; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[46:49], off, s[0:3], 0 offset:144
+; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[42:45], off, s[0:3], 0 offset:96
+; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[38:41], off, s[0:3], 0 offset:112
+; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[18:21], off, s[0:3], 0 offset:64
+; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[34:37], off, s[0:3], 0 offset:80
+; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[30:33], off, s[0:3], 0 offset:32
+; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[22:25], off, s[0:3], 0 offset:48
 ; GCN-NOHSA-SI-NEXT:    buffer_load_dword v4, off, s[12:15], 0 offset:20 ; 4-byte Folded Reload
 ; GCN-NOHSA-SI-NEXT:    buffer_load_dword v5, off, s[12:15], 0 offset:24 ; 4-byte Folded Reload
 ; GCN-NOHSA-SI-NEXT:    buffer_load_dword v6, off, s[12:15], 0 offset:28 ; 4-byte Folded Reload
 ; GCN-NOHSA-SI-NEXT:    buffer_load_dword v7, off, s[12:15], 0 offset:32 ; 4-byte Folded Reload
 ; GCN-NOHSA-SI-NEXT:    s_waitcnt vmcnt(0)
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v5, v1
-; GCN-NOHSA-SI-NEXT:    buffer_load_dword v12, off, s[12:15], 0 offset:4 ; 4-byte Folded Reload
-; GCN-NOHSA-SI-NEXT:    buffer_load_dword v13, off, s[12:15], 0 offset:8 ; 4-byte Folded Reload
-; GCN-NOHSA-SI-NEXT:    buffer_load_dword v14, off, s[12:15], 0 offset:12 ; 4-byte Folded Reload
-; GCN-NOHSA-SI-NEXT:    buffer_load_dword v15, off, s[12:15], 0 offset:16 ; 4-byte Folded Reload
-; GCN-NOHSA-SI-NEXT:    s_waitcnt vmcnt(0)
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v13, v1
-; GCN-NOHSA-SI-NEXT:    s_mov_b32 s0, s4
-; GCN-NOHSA-SI-NEXT:    s_mov_b32 s1, s5
-; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:240
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v21, 0
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v61, 0
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v25, 0
-; GCN-NOHSA-SI-NEXT:    s_waitcnt expcnt(0)
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v0, v12
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v1, v13
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v2, v14
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v3, 0
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v7, 0
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v37, 0
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v29, 0
-; GCN-NOHSA-SI-NEXT:    v_mov_b32_e32 v49, 0
-; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[8:11], off, s[0:3], 0 offset:208
-; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[54:57], off, s[0:3], 0 offset:176
-; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[50:53], off, s[0:3], 0 offset:144
-; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[42:45], off, s[0:3], 0 offset:112
-; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[38:41], off, s[0:3], 0 offset:80
-; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[30:33], off, s[0:3], 0 offset:48
-; GCN-NOHSA-SI-NEXT:    s_waitcnt expcnt(5)
-; GCN-NOHSA-SI-NEXT:    buffer_load_dword v8, off, s[12:15], 0 offset:36 ; 4-byte Folded Reload
-; GCN-NOHSA-SI-NEXT:    buffer_load_dword v9, off, s[12:15], 0 offset:40 ; 4-byte Folded Reload
-; GCN-NOHSA-SI-NEXT:    buffer_load_dword v10, off, s[12:15], 0 offset:44 ; 4-byte Folded Reload
-; GCN-NOHSA-SI-NEXT:    buffer_load_dword v11, off, s[12:15], 0 offset:48 ; 4-byte Folded Reload
-; GCN-NOHSA-SI-NEXT:    s_waitcnt vmcnt(0)
-; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[8:11], off, s[0:3], 0 offset:16
-; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[18:21], off, s[0:3], 0 offset:224
-; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[58:61], off, s[0:3], 0 offset:192
-; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[22:25], off, s[0:3], 0 offset:160
-; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[46:49], off, s[0:3], 0 offset:128
-; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[26:29], off, s[0:3], 0 offset:96
-; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[34:37], off, s[0:3], 0 offset:64
-; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[4:7], off, s[0:3], 0 offset:32
-; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0
+; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[4:7], off, s[0:3], 0
+; GCN-NOHSA-SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:16
 ; GCN-NOHSA-SI-NEXT:    s_endpgm
 ;
 ; GCN-HSA-LABEL: global_zextload_v32i16_to_v32i64:
 ; GCN-HSA:       ; %bb.0:
 ; GCN-HSA-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, 0
-; GCN-HSA-NEXT:    s_mov_b32 s16, 0xffff
+; GCN-HSA-NEXT:    s_mov_b32 s18, 0xffff
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v3, v1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v5, v1
 ; GCN-HSA-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-HSA-NEXT:    s_add_u32 s4, s2, 16
 ; GCN-HSA-NEXT:    s_addc_u32 s5, s3, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v7, s5
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v6, s4
 ; GCN-HSA-NEXT:    s_add_u32 s4, s2, 32
 ; GCN-HSA-NEXT:    s_addc_u32 s5, s3, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s5
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v10, s4
+; GCN-HSA-NEXT:    v_mov_b32_e32 v19, s3
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[6:9], v[6:7]
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[10:13], v[10:11]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v19, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v18, s2
 ; GCN-HSA-NEXT:    s_add_u32 s2, s2, 48
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s3, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s3
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s2
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[14:17], v[14:15]
 ; GCN-HSA-NEXT:    flat_load_dwordx4 v[18:21], v[18:19]
-; GCN-HSA-NEXT:    s_add_u32 s2, s0, 48
+; GCN-HSA-NEXT:    s_add_u32 s2, s0, 32
 ; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-HSA-NEXT:    s_add_u32 s4, s0, 16
+; GCN-HSA-NEXT:    s_add_u32 s4, s0, 48
 ; GCN-HSA-NEXT:    s_addc_u32 s5, s1, 0
-; GCN-HSA-NEXT:    s_add_u32 s6, s0, 0xf0
+; GCN-HSA-NEXT:    s_add_u32 s6, s0, 16
 ; GCN-HSA-NEXT:    s_addc_u32 s7, s1, 0
-; GCN-HSA-NEXT:    s_add_u32 s8, s0, 0xd0
+; GCN-HSA-NEXT:    s_add_u32 s8, s0, 0xe0
 ; GCN-HSA-NEXT:    s_addc_u32 s9, s1, 0
-; GCN-HSA-NEXT:    s_add_u32 s10, s0, 0xb0
+; GCN-HSA-NEXT:    s_add_u32 s10, s0, 0xf0
 ; GCN-HSA-NEXT:    s_addc_u32 s11, s1, 0
-; GCN-HSA-NEXT:    s_add_u32 s12, s0, 0x90
+; GCN-HSA-NEXT:    s_add_u32 s12, s0, 0xc0
 ; GCN-HSA-NEXT:    s_addc_u32 s13, s1, 0
-; GCN-HSA-NEXT:    s_add_u32 s14, s0, 0x70
-; GCN-HSA-NEXT:    s_addc_u32 s15, s1, 0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v23, s15
-; GCN-HSA-NEXT:    v_mov_b32_e32 v22, s14
-; GCN-HSA-NEXT:    s_add_u32 s14, s0, 0x50
+; GCN-HSA-NEXT:    s_add_u32 s14, s0, 0xd0
 ; GCN-HSA-NEXT:    s_addc_u32 s15, s1, 0
-; GCN-HSA-NEXT:    s_waitcnt vmcnt(3)
-; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v4, 16, v9
-; GCN-HSA-NEXT:    v_and_b32_e32 v2, s16, v9
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[22:23], v[2:5]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v23, s15
-; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v4, 16, v7
-; GCN-HSA-NEXT:    v_and_b32_e32 v2, s16, v7
-; GCN-HSA-NEXT:    v_mov_b32_e32 v22, s14
+; GCN-HSA-NEXT:    s_add_u32 s16, s0, 0xa0
+; GCN-HSA-NEXT:    s_addc_u32 s17, s1, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v23, s17
+; GCN-HSA-NEXT:    v_mov_b32_e32 v22, s16
+; GCN-HSA-NEXT:    s_add_u32 s16, s0, 0xb0
+; GCN-HSA-NEXT:    s_addc_u32 s17, s1, 0
+; GCN-HSA-NEXT:    s_waitcnt vmcnt(2)
+; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v4, 16, v12
+; GCN-HSA-NEXT:    v_and_b32_e32 v2, s18, v12
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[22:23], v[2:5]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v23, s11
-; GCN-HSA-NEXT:    s_waitcnt vmcnt(4)
+; GCN-HSA-NEXT:    s_nop 0
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v4, 16, v13
-; GCN-HSA-NEXT:    v_and_b32_e32 v2, s16, v13
-; GCN-HSA-NEXT:    v_mov_b32_e32 v22, s10
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[22:23], v[2:5]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v23, s13
-; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v4, 16, v11
-; GCN-HSA-NEXT:    v_and_b32_e32 v2, s16, v11
-; GCN-HSA-NEXT:    v_mov_b32_e32 v22, s12
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[22:23], v[2:5]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v23, s5
-; GCN-HSA-NEXT:    v_mov_b32_e32 v22, s4
-; GCN-HSA-NEXT:    s_waitcnt vmcnt(4)
-; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v4, 16, v19
-; GCN-HSA-NEXT:    v_and_b32_e32 v2, s16, v19
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[22:23], v[2:5]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v23, s7
-; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v4, 16, v17
-; GCN-HSA-NEXT:    v_and_b32_e32 v2, s16, v17
-; GCN-HSA-NEXT:    v_mov_b32_e32 v22, s6
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[22:23], v[2:5]
-; GCN-HSA-NEXT:    s_add_u32 s4, s0, 32
-; GCN-HSA-NEXT:    v_mov_b32_e32 v23, s9
-; GCN-HSA-NEXT:    s_addc_u32 s5, s1, 0
-; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v4, 16, v15
-; GCN-HSA-NEXT:    v_and_b32_e32 v2, s16, v15
-; GCN-HSA-NEXT:    v_mov_b32_e32 v22, s8
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[22:23], v[2:5]
-; GCN-HSA-NEXT:    s_add_u32 s6, s0, 0xe0
-; GCN-HSA-NEXT:    v_mov_b32_e32 v7, 0
+; GCN-HSA-NEXT:    v_and_b32_e32 v2, s18, v13
+; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s16
+; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s17
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[12:13], v[2:5]
+; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s5
+; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s4
+; GCN-HSA-NEXT:    s_waitcnt vmcnt(2)
+; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v4, 16, v21
+; GCN-HSA-NEXT:    v_and_b32_e32 v2, s18, v21
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[12:13], v[2:5]
+; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s1
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v4, 16, v18
-; GCN-HSA-NEXT:    v_and_b32_e32 v2, s16, v18
-; GCN-HSA-NEXT:    v_mov_b32_e32 v18, s1
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, v7
-; GCN-HSA-NEXT:    v_mov_b32_e32 v17, s0
-; GCN-HSA-NEXT:    s_addc_u32 s7, s1, 0
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[17:18], v[2:5]
-; GCN-HSA-NEXT:    v_and_b32_e32 v0, s16, v21
+; GCN-HSA-NEXT:    v_and_b32_e32 v2, s18, v18
+; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s0
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[12:13], v[2:5]
+; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s7
+; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v4, 16, v19
+; GCN-HSA-NEXT:    v_and_b32_e32 v2, s18, v19
+; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s6
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[12:13], v[2:5]
+; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s9
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v4, 16, v16
-; GCN-HSA-NEXT:    v_and_b32_e32 v2, s16, v16
-; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s7
-; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s6
-; GCN-HSA-NEXT:    s_add_u32 s6, s0, 0xc0
-; GCN-HSA-NEXT:    s_addc_u32 s7, s1, 0
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[15:16], v[2:5]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v16, s3
+; GCN-HSA-NEXT:    v_and_b32_e32 v2, s18, v16
+; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s8
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[12:13], v[2:5]
+; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s11
+; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v4, 16, v17
+; GCN-HSA-NEXT:    v_and_b32_e32 v2, s18, v17
+; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s10
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[12:13], v[2:5]
+; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s12
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v4, 16, v14
-; GCN-HSA-NEXT:    v_and_b32_e32 v2, s16, v14
-; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s7
-; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s6
-; GCN-HSA-NEXT:    s_add_u32 s6, s0, 0xa0
-; GCN-HSA-NEXT:    s_addc_u32 s7, s1, 0
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[13:14], v[2:5]
-; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s2
-; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v4, 16, v12
-; GCN-HSA-NEXT:    v_and_b32_e32 v2, s16, v12
-; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s7
-; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s6
-; GCN-HSA-NEXT:    s_add_u32 s6, s0, 0x80
-; GCN-HSA-NEXT:    s_addc_u32 s7, s1, 0
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[11:12], v[2:5]
-; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x60
+; GCN-HSA-NEXT:    v_and_b32_e32 v2, s18, v14
+; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s13
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[12:13], v[2:5]
+; GCN-HSA-NEXT:    s_add_u32 s4, s0, 0x80
+; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s14
+; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v4, 16, v15
+; GCN-HSA-NEXT:    v_and_b32_e32 v2, s18, v15
+; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s15
+; GCN-HSA-NEXT:    s_addc_u32 s5, s1, 0
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[12:13], v[2:5]
+; GCN-HSA-NEXT:    v_mov_b32_e32 v13, s5
+; GCN-HSA-NEXT:    v_mov_b32_e32 v12, s4
+; GCN-HSA-NEXT:    s_add_u32 s4, s0, 0x90
 ; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v4, 16, v10
-; GCN-HSA-NEXT:    v_and_b32_e32 v2, s16, v10
-; GCN-HSA-NEXT:    v_mov_b32_e32 v10, s7
-; GCN-HSA-NEXT:    v_mov_b32_e32 v9, s6
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[9:10], v[2:5]
-; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
-; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v2, 16, v21
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[15:16], v[0:3]
-; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v11, 16, v6
-; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s4
-; GCN-HSA-NEXT:    v_and_b32_e32 v9, s16, v6
-; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v6, 16, v20
-; GCN-HSA-NEXT:    v_and_b32_e32 v4, s16, v20
-; GCN-HSA-NEXT:    v_mov_b32_e32 v5, v1
-; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s5
-; GCN-HSA-NEXT:    s_add_u32 s0, s0, 64
+; GCN-HSA-NEXT:    v_and_b32_e32 v2, s18, v10
+; GCN-HSA-NEXT:    s_addc_u32 s5, s1, 0
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[12:13], v[2:5]
+; GCN-HSA-NEXT:    v_mov_b32_e32 v15, s3
+; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v4, 16, v11
+; GCN-HSA-NEXT:    v_and_b32_e32 v2, s18, v11
+; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s5
+; GCN-HSA-NEXT:    v_mov_b32_e32 v10, s4
+; GCN-HSA-NEXT:    s_add_u32 s4, s0, 0x60
+; GCN-HSA-NEXT:    s_addc_u32 s5, s1, 0
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[10:11], v[2:5]
+; GCN-HSA-NEXT:    v_mov_b32_e32 v11, s5
+; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v4, 16, v8
+; GCN-HSA-NEXT:    v_and_b32_e32 v2, s18, v8
+; GCN-HSA-NEXT:    v_mov_b32_e32 v10, s4
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[10:11], v[2:5]
+; GCN-HSA-NEXT:    v_mov_b32_e32 v14, s2
+; GCN-HSA-NEXT:    s_add_u32 s2, s0, 0x70
+; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v2, 16, v20
+; GCN-HSA-NEXT:    v_and_b32_e32 v0, s18, v20
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[14:15], v[0:3]
+; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
+; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s2
+; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s3
+; GCN-HSA-NEXT:    s_add_u32 s2, s0, 64
+; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v10, 16, v7
+; GCN-HSA-NEXT:    v_and_b32_e32 v8, s18, v7
+; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v13, 16, v6
+; GCN-HSA-NEXT:    v_and_b32_e32 v11, s18, v6
+; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v6, 16, v9
+; GCN-HSA-NEXT:    v_and_b32_e32 v4, s18, v9
+; GCN-HSA-NEXT:    v_mov_b32_e32 v7, v1
+; GCN-HSA-NEXT:    s_addc_u32 s3, s1, 0
 ; GCN-HSA-NEXT:    flat_store_dwordx4 v[2:3], v[4:7]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v2, s2
-; GCN-HSA-NEXT:    v_lshrrev_b32_e32 v14, 16, v8
-; GCN-HSA-NEXT:    v_and_b32_e32 v12, s16, v8
-; GCN-HSA-NEXT:    v_mov_b32_e32 v13, v1
-; GCN-HSA-NEXT:    v_mov_b32_e32 v15, v7
+; GCN-HSA-NEXT:    s_add_u32 s0, s0, 0x50
+; GCN-HSA-NEXT:    v_mov_b32_e32 v12, v1
+; GCN-HSA-NEXT:    v_mov_b32_e32 v14, v1
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v3, s3
-; GCN-HSA-NEXT:    v_mov_b32_e32 v10, v1
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[2:3], v[11:14]
+; GCN-HSA-NEXT:    v_mov_b32_e32 v9, v1
+; GCN-HSA-NEXT:    v_mov_b32_e32 v11, v1
 ; GCN-HSA-NEXT:    s_addc_u32 s1, s1, 0
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v0, s0
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[2:3], v[12:15]
 ; GCN-HSA-NEXT:    v_mov_b32_e32 v1, s1
-; GCN-HSA-NEXT:    v_mov_b32_e32 v12, v7
-; GCN-HSA-NEXT:    flat_store_dwordx4 v[0:1], v[9:12]
+; GCN-HSA-NEXT:    flat_store_dwordx4 v[0:1], v[8:11]
 ; GCN-HSA-NEXT:    s_endpgm
 ;
 ; GCN-NOHSA-VI-LABEL: global_zextload_v32i16_to_v32i64:
 ; GCN-NOHSA-VI:       ; %bb.0:
 ; GCN-NOHSA-VI-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s3, 0xf000
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s2, -1
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s10, s2
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s11, s3
 ; GCN-NOHSA-VI-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s8, s6
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s9, s7
 ; GCN-NOHSA-VI-NEXT:    buffer_load_dwordx4 v[0:3], off, s[8:11], 0
 ; GCN-NOHSA-VI-NEXT:    buffer_load_dwordx4 v[4:7], off, s[8:11], 0 offset:16
-; GCN-NOHSA-VI-NEXT:    buffer_load_dwordx4 v[31:34], off, s[8:11], 0 offset:32
-; GCN-NOHSA-VI-NEXT:    buffer_load_dwordx4 v[35:38], off, s[8:11], 0 offset:48
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s0, 0xffff
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v57, 0
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s1, s5
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v11, 0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v15, 0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v27, 0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v23, 0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v19, 0
-; GCN-NOHSA-VI-NEXT:    s_waitcnt vmcnt(3)
-; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v8, s0, v1
+; GCN-NOHSA-VI-NEXT:    s_waitcnt vmcnt(1)
 ; GCN-NOHSA-VI-NEXT:    v_lshrrev_b32_e32 v10, 16, v1
 ; GCN-NOHSA-VI-NEXT:    s_waitcnt vmcnt(0)
-; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v48, s0, v36
-; GCN-NOHSA-VI-NEXT:    v_lshrrev_b32_e32 v50, 16, v36
-; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v54, s0, v38
-; GCN-NOHSA-VI-NEXT:    v_lshrrev_b32_e32 v56, 16, v38
-; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v36, s0, v37
-; GCN-NOHSA-VI-NEXT:    v_lshrrev_b32_e32 v38, 16, v37
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v37, 0
-; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v1, s0, v3
-; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v12, s0, v0
-; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v16, s0, v2
-; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v20, s0, v5
-; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v24, s0, v4
-; GCN-NOHSA-VI-NEXT:    v_lshrrev_b32_e32 v26, 16, v4
-; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v4, s0, v6
-; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v28, s0, v7
-; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v39, s0, v32
-; GCN-NOHSA-VI-NEXT:    v_lshrrev_b32_e32 v41, 16, v32
-; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v32, s0, v34
-; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v42, s0, v31
-; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v45, s0, v33
+; GCN-NOHSA-VI-NEXT:    v_lshrrev_b32_e32 v20, 16, v5
+; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v18, s0, v5
+; GCN-NOHSA-VI-NEXT:    v_lshrrev_b32_e32 v24, 16, v4
+; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v22, s0, v4
+; GCN-NOHSA-VI-NEXT:    v_lshrrev_b32_e32 v28, 16, v7
+; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v26, s0, v7
+; GCN-NOHSA-VI-NEXT:    v_lshrrev_b32_e32 v31, 16, v6
+; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v29, s0, v6
+; GCN-NOHSA-VI-NEXT:    buffer_load_dwordx4 v[4:7], off, s[8:11], 0 offset:32
+; GCN-NOHSA-VI-NEXT:    buffer_load_dwordx4 v[32:35], off, s[8:11], 0 offset:48
+; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v8, s0, v1
+; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v11, s0, v0
+; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v14, s0, v3
+; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v1, s0, v2
+; GCN-NOHSA-VI-NEXT:    v_lshrrev_b32_e32 v16, 16, v3
+; GCN-NOHSA-VI-NEXT:    v_lshrrev_b32_e32 v13, 16, v0
+; GCN-NOHSA-VI-NEXT:    v_lshrrev_b32_e32 v3, 16, v2
+; GCN-NOHSA-VI-NEXT:    s_waitcnt vmcnt(1)
+; GCN-NOHSA-VI-NEXT:    v_lshrrev_b32_e32 v41, 16, v4
+; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v39, s0, v4
+; GCN-NOHSA-VI-NEXT:    v_lshrrev_b32_e32 v47, 16, v6
+; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v45, s0, v6
+; GCN-NOHSA-VI-NEXT:    s_waitcnt vmcnt(0)
+; GCN-NOHSA-VI-NEXT:    v_lshrrev_b32_e32 v6, 16, v33
+; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v4, s0, v33
+; GCN-NOHSA-VI-NEXT:    v_lshrrev_b32_e32 v53, 16, v35
 ; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v51, s0, v35
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v55, v37
+; GCN-NOHSA-VI-NEXT:    v_lshrrev_b32_e32 v35, 16, v34
+; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v33, s0, v34
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v34, 0
+; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v36, s0, v5
+; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v42, s0, v7
+; GCN-NOHSA-VI-NEXT:    v_and_b32_e32 v48, s0, v32
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v52, v34
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v54, v34
 ; GCN-NOHSA-VI-NEXT:    s_mov_b32 s0, s4
-; GCN-NOHSA-VI-NEXT:    v_lshrrev_b32_e32 v53, 16, v35
-; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[54:57], off, s[0:3], 0 offset:240
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v52, v37
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v54, 0
-; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[51:54], off, s[0:3], 0 offset:192
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v49, v37
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v51, 0
-; GCN-NOHSA-VI-NEXT:    v_lshrrev_b32_e32 v47, 16, v33
-; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[48:51], off, s[0:3], 0 offset:208
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v46, v37
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v48, 0
-; GCN-NOHSA-VI-NEXT:    v_lshrrev_b32_e32 v22, 16, v5
-; GCN-NOHSA-VI-NEXT:    v_lshrrev_b32_e32 v30, 16, v7
-; GCN-NOHSA-VI-NEXT:    v_lshrrev_b32_e32 v44, 16, v31
+; GCN-NOHSA-VI-NEXT:    v_lshrrev_b32_e32 v50, 16, v32
+; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[51:54], off, s[0:3], 0 offset:240
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v49, v34
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v51, v34
+; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[48:51], off, s[0:3], 0 offset:192
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v46, v34
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v48, v34
+; GCN-NOHSA-VI-NEXT:    v_lshrrev_b32_e32 v44, 16, v7
 ; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[45:48], off, s[0:3], 0 offset:160
-; GCN-NOHSA-VI-NEXT:    v_lshrrev_b32_e32 v6, 16, v6
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v45, 0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v43, v37
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v7, 0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, v37
-; GCN-NOHSA-VI-NEXT:    v_lshrrev_b32_e32 v18, 16, v2
-; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[42:45], off, s[0:3], 0 offset:128
-; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[4:7], off, s[0:3], 0 offset:96
-; GCN-NOHSA-VI-NEXT:    v_lshrrev_b32_e32 v14, 16, v0
-; GCN-NOHSA-VI-NEXT:    v_lshrrev_b32_e32 v3, 16, v3
-; GCN-NOHSA-VI-NEXT:    v_lshrrev_b32_e32 v34, 16, v34
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v35, 0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v33, v37
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v42, 0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v40, v37
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v31, 0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, 0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v29, v37
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, v37
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v13, v37
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v9, v37
-; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[39:42], off, s[0:3], 0 offset:144
-; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[32:35], off, s[0:3], 0 offset:176
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v39, 0
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v25, v37
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v21, v37
-; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v17, v37
-; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[28:31], off, s[0:3], 0 offset:112
-; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[24:27], off, s[0:3], 0 offset:64
-; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[20:23], off, s[0:3], 0 offset:80
-; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[36:39], off, s[0:3], 0 offset:224
-; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[16:19], off, s[0:3], 0 offset:32
-; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[1:4], off, s[0:3], 0 offset:48
-; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[12:15], off, s[0:3], 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v43, v34
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v45, v34
+; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[42:45], off, s[0:3], 0 offset:176
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v40, v34
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v42, v34
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v30, v34
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v32, v34
+; GCN-NOHSA-VI-NEXT:    v_lshrrev_b32_e32 v38, 16, v5
+; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[39:42], off, s[0:3], 0 offset:128
+; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[29:32], off, s[0:3], 0 offset:96
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v5, v34
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v7, v34
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v37, v34
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v39, v34
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v27, v34
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v29, v34
+; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[4:7], off, s[0:3], 0 offset:208
+; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[36:39], off, s[0:3], 0 offset:144
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v23, v34
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v25, v34
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v19, v34
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v21, v34
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v36, v34
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v2, v34
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v4, v34
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v15, v34
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v17, v34
+; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[26:29], off, s[0:3], 0 offset:112
+; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[22:25], off, s[0:3], 0 offset:64
+; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[18:21], off, s[0:3], 0 offset:80
+; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[33:36], off, s[0:3], 0 offset:224
+; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[1:4], off, s[0:3], 0 offset:32
+; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[14:17], off, s[0:3], 0 offset:48
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v12, v34
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v14, v34
+; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[11:14], off, s[0:3], 0
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v9, v34
+; GCN-NOHSA-VI-NEXT:    v_mov_b32_e32 v11, v34
 ; GCN-NOHSA-VI-NEXT:    buffer_store_dwordx4 v[8:11], off, s[0:3], 0 offset:16
 ; GCN-NOHSA-VI-NEXT:    s_endpgm
 ;
 ; EG-LABEL: global_zextload_v32i16_to_v32i64:
 ; EG:       ; %bb.0:
 ; EG-NEXT:    ALU 0, @30, KC0[CB0:0-32], KC1[]
 ; EG-NEXT:    TEX 2 @22
 ; EG-NEXT:    ALU 33, @31, KC0[], KC1[]

llvm/test/CodeGen/AMDGPU/saddsat.ll

; GFX9-NEXT:    s_setpc_b64 s[30:31]
   %result = call <3 x i16> @llvm.sadd.sat.v3i16(<3 x i16> %lhs, <3 x i16> %rhs)
   ret <3 x i16> %result
 }
 
 define <2 x float> @v_saddsat_v4i16(<4 x i16> %lhs, <4 x i16> %rhs) {
 ; GFX6-LABEL: v_saddsat_v4i16:
 ; GFX6:       ; %bb.0:
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    v_bfe_i32 v8, v1, 0, 16
+; GFX6-NEXT:    v_bfe_i32 v1, v6, 0, 16
+; GFX6-NEXT:    v_bfe_i32 v2, v2, 0, 16
 ; GFX6-NEXT:    v_bfe_i32 v4, v4, 0, 16
 ; GFX6-NEXT:    v_bfe_i32 v0, v0, 0, 16
 ; GFX6-NEXT:    v_bfe_i32 v5, v5, 0, 16
-; GFX6-NEXT:    v_bfe_i32 v1, v1, 0, 16
-; GFX6-NEXT:    v_add_i32_e32 v1, vcc, v1, v5
+; GFX6-NEXT:    v_bfe_i32 v6, v7, 0, 16
+; GFX6-NEXT:    v_bfe_i32 v3, v3, 0, 16
+; GFX6-NEXT:    v_add_i32_e32 v1, vcc, v2, v1
 ; GFX6-NEXT:    s_movk_i32 s4, 0x7fff
+; GFX6-NEXT:    v_add_i32_e32 v3, vcc, v3, v6
+; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v8, v5
 ; GFX6-NEXT:    v_add_i32_e32 v0, vcc, v0, v4
-; GFX6-NEXT:    v_min_i32_e32 v1, s4, v1
+; GFX6-NEXT:    v_min_i32_e32 v3, s4, v3
 ; GFX6-NEXT:    s_movk_i32 s5, 0x8000
+; GFX6-NEXT:    v_min_i32_e32 v1, s4, v1
+; GFX6-NEXT:    v_min_i32_e32 v2, s4, v2
 ; GFX6-NEXT:    v_min_i32_e32 v0, s4, v0
+; GFX6-NEXT:    v_max_i32_e32 v3, s5, v3
+; GFX6-NEXT:    v_max_i32_e32 v2, s5, v2
 ; GFX6-NEXT:    v_max_i32_e32 v1, s5, v1
-; GFX6-NEXT:    v_max_i32_e32 v0, s5, v0
 ; GFX6-NEXT:    s_mov_b32 s6, 0xffff
-; GFX6-NEXT:    v_bfe_i32 v6, v6, 0, 16
-; GFX6-NEXT:    v_bfe_i32 v2, v2, 0, 16
-; GFX6-NEXT:    v_bfe_i32 v7, v7, 0, 16
-; GFX6-NEXT:    v_bfe_i32 v3, v3, 0, 16
-; GFX6-NEXT:    v_lshlrev_b32_e32 v1, 16, v1
+; GFX6-NEXT:    v_max_i32_e32 v0, s5, v0
+; GFX6-NEXT:    v_lshlrev_b32_e32 v3, 16, v3
+; GFX6-NEXT:    v_and_b32_e32 v1, s6, v1
+; GFX6-NEXT:    v_lshlrev_b32_e32 v2, 16, v2
 ; GFX6-NEXT:    v_and_b32_e32 v0, s6, v0
-; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v2, v6
-; GFX6-NEXT:    v_or_b32_e32 v0, v0, v1
-; GFX6-NEXT:    v_add_i32_e32 v1, vcc, v3, v7
-; GFX6-NEXT:    v_min_i32_e32 v1, s4, v1
-; GFX6-NEXT:    v_min_i32_e32 v2, s4, v2
-; GFX6-NEXT:    v_max_i32_e32 v1, s5, v1
-; GFX6-NEXT:    v_max_i32_e32 v2, s5, v2
-; GFX6-NEXT:    v_lshlrev_b32_e32 v1, 16, v1
-; GFX6-NEXT:    v_and_b32_e32 v2, s6, v2
-; GFX6-NEXT:    v_or_b32_e32 v1, v2, v1
+; GFX6-NEXT:    v_or_b32_e32 v1, v1, v3
+; GFX6-NEXT:    v_or_b32_e32 v0, v0, v2
 ; GFX6-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX8-LABEL: v_saddsat_v4i16:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX8-NEXT:    v_lshrrev_b32_e32 v4, 16, v2
 ; GFX8-NEXT:    v_lshrrev_b32_e32 v5, 16, v0
 ; GFX8-NEXT:    v_add_u16_e32 v6, v5, v4

llvm/test/CodeGen/AMDGPU/sdiv64.ll

; GCN-IR-NEXT:    s_setpc_b64 s[30:31]
   %result = sdiv i64 %x, %y
   ret i64 %result
 }
 
 define amdgpu_kernel void @s_test_sdiv24_64(i64 addrspace(1)* %out, i64 %x, i64 %y) {
 ; GCN-LABEL: s_test_sdiv24_64:
 ; GCN:       ; %bb.0:
 ; GCN-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x9
-; GCN-NEXT:    s_load_dword s1, s[0:1], 0xe
+; GCN-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xd
 ; GCN-NEXT:    s_mov_b32 s3, 0xf000
 ; GCN-NEXT:    s_mov_b32 s2, -1
 ; GCN-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-NEXT:    s_mov_b32 s0, s4
 ; GCN-NEXT:    s_ashr_i64 s[8:9], s[0:1], 40
 ; GCN-NEXT:    v_cvt_f32_i32_e32 v0, s8
 ; GCN-NEXT:    s_mov_b32 s1, s5
 ; GCN-NEXT:    s_ashr_i64 s[4:5], s[6:7], 40

 ; GCN-NEXT:    v_bfe_i32 v0, v0, 0, 24
 ; GCN-NEXT:    v_ashrrev_i32_e32 v1, 31, v0
 ; GCN-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
 ; GCN-NEXT:    s_endpgm
 ;
 ; GCN-IR-LABEL: s_test_sdiv24_64:
 ; GCN-IR:       ; %bb.0:
 ; GCN-IR-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x9
-; GCN-IR-NEXT:    s_load_dword s1, s[0:1], 0xe
+; GCN-IR-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xd
 ; GCN-IR-NEXT:    s_mov_b32 s3, 0xf000
 ; GCN-IR-NEXT:    s_mov_b32 s2, -1
 ; GCN-IR-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-IR-NEXT:    s_mov_b32 s0, s4
 ; GCN-IR-NEXT:    s_ashr_i64 s[8:9], s[0:1], 40
 ; GCN-IR-NEXT:    v_cvt_f32_i32_e32 v0, s8
 ; GCN-IR-NEXT:    s_mov_b32 s1, s5
 ; GCN-IR-NEXT:    s_ashr_i64 s[4:5], s[6:7], 40
; GCN-IR-NEXT:    s_endpgm
   store i64 %result, i64 addrspace(1)* %out
   ret void
 }
 
 define amdgpu_kernel void @s_test_sdiv31_64(i64 addrspace(1)* %out, i64 %x, i64 %y) {
 ; GCN-LABEL: s_test_sdiv31_64:
 ; GCN:       ; %bb.0:
 ; GCN-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x9
-; GCN-NEXT:    s_load_dword s1, s[0:1], 0xe
+; GCN-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xd
 ; GCN-NEXT:    s_mov_b32 s3, 0xf000
 ; GCN-NEXT:    s_mov_b32 s2, -1
 ; GCN-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-NEXT:    s_mov_b32 s0, s4
 ; GCN-NEXT:    s_ashr_i64 s[8:9], s[0:1], 33
 ; GCN-NEXT:    v_cvt_f32_i32_e32 v0, s8
 ; GCN-NEXT:    s_mov_b32 s1, s5
 ; GCN-NEXT:    s_ashr_i64 s[4:5], s[6:7], 33

 ; GCN-NEXT:    v_bfe_i32 v0, v0, 0, 31
 ; GCN-NEXT:    v_ashrrev_i32_e32 v1, 31, v0
 ; GCN-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
 ; GCN-NEXT:    s_endpgm
 ;
 ; GCN-IR-LABEL: s_test_sdiv31_64:
 ; GCN-IR:       ; %bb.0:
 ; GCN-IR-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x9
-; GCN-IR-NEXT:    s_load_dword s1, s[0:1], 0xe
+; GCN-IR-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xd
 ; GCN-IR-NEXT:    s_mov_b32 s3, 0xf000
 ; GCN-IR-NEXT:    s_mov_b32 s2, -1
 ; GCN-IR-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-IR-NEXT:    s_mov_b32 s0, s4
 ; GCN-IR-NEXT:    s_ashr_i64 s[8:9], s[0:1], 33
 ; GCN-IR-NEXT:    v_cvt_f32_i32_e32 v0, s8
 ; GCN-IR-NEXT:    s_mov_b32 s1, s5
 ; GCN-IR-NEXT:    s_ashr_i64 s[4:5], s[6:7], 33
; GCN-IR-NEXT:    s_endpgm
   store i64 %result, i64 addrspace(1)* %out
   ret void
 }
 
 define amdgpu_kernel void @s_test_sdiv23_64(i64 addrspace(1)* %out, i64 %x, i64 %y) {
 ; GCN-LABEL: s_test_sdiv23_64:
 ; GCN:       ; %bb.0:
 ; GCN-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x9
-; GCN-NEXT:    s_load_dword s1, s[0:1], 0xe
+; GCN-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xd
 ; GCN-NEXT:    s_mov_b32 s3, 0xf000
 ; GCN-NEXT:    s_mov_b32 s2, -1
 ; GCN-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-NEXT:    s_mov_b32 s0, s4
 ; GCN-NEXT:    s_ashr_i64 s[8:9], s[0:1], 41
 ; GCN-NEXT:    v_cvt_f32_i32_e32 v0, s8
 ; GCN-NEXT:    s_mov_b32 s1, s5
 ; GCN-NEXT:    s_ashr_i64 s[4:5], s[6:7], 41

 ; GCN-NEXT:    v_bfe_i32 v0, v0, 0, 23
 ; GCN-NEXT:    v_ashrrev_i32_e32 v1, 31, v0
 ; GCN-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
 ; GCN-NEXT:    s_endpgm
 ;
 ; GCN-IR-LABEL: s_test_sdiv23_64:
 ; GCN-IR:       ; %bb.0:
 ; GCN-IR-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x9
-; GCN-IR-NEXT:    s_load_dword s1, s[0:1], 0xe
+; GCN-IR-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xd
 ; GCN-IR-NEXT:    s_mov_b32 s3, 0xf000
 ; GCN-IR-NEXT:    s_mov_b32 s2, -1
 ; GCN-IR-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-IR-NEXT:    s_mov_b32 s0, s4
 ; GCN-IR-NEXT:    s_ashr_i64 s[8:9], s[0:1], 41
 ; GCN-IR-NEXT:    v_cvt_f32_i32_e32 v0, s8
 ; GCN-IR-NEXT:    s_mov_b32 s1, s5
 ; GCN-IR-NEXT:    s_ashr_i64 s[4:5], s[6:7], 41
; GCN-IR-NEXT:    s_endpgm
   store i64 %result, i64 addrspace(1)* %out
   ret void
 }
 
 define amdgpu_kernel void @s_test_sdiv25_64(i64 addrspace(1)* %out, i64 %x, i64 %y) {
 ; GCN-LABEL: s_test_sdiv25_64:
 ; GCN:       ; %bb.0:
 ; GCN-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x9
-; GCN-NEXT:    s_load_dword s1, s[0:1], 0xe
+; GCN-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xd
 ; GCN-NEXT:    s_mov_b32 s3, 0xf000
 ; GCN-NEXT:    s_mov_b32 s2, -1
 ; GCN-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-NEXT:    s_mov_b32 s0, s4
 ; GCN-NEXT:    s_ashr_i64 s[8:9], s[0:1], 39
 ; GCN-NEXT:    v_cvt_f32_i32_e32 v0, s8
 ; GCN-NEXT:    s_mov_b32 s1, s5
 ; GCN-NEXT:    s_ashr_i64 s[4:5], s[6:7], 39

 ; GCN-NEXT:    v_bfe_i32 v0, v0, 0, 25
 ; GCN-NEXT:    v_ashrrev_i32_e32 v1, 31, v0
 ; GCN-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
 ; GCN-NEXT:    s_endpgm
 ;
 ; GCN-IR-LABEL: s_test_sdiv25_64:
 ; GCN-IR:       ; %bb.0:
 ; GCN-IR-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x9
-; GCN-IR-NEXT:    s_load_dword s1, s[0:1], 0xe
+; GCN-IR-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xd
 ; GCN-IR-NEXT:    s_mov_b32 s3, 0xf000
 ; GCN-IR-NEXT:    s_mov_b32 s2, -1
 ; GCN-IR-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-IR-NEXT:    s_mov_b32 s0, s4
 ; GCN-IR-NEXT:    s_ashr_i64 s[8:9], s[0:1], 39
 ; GCN-IR-NEXT:    v_cvt_f32_i32_e32 v0, s8
 ; GCN-IR-NEXT:    s_mov_b32 s1, s5
 ; GCN-IR-NEXT:    s_ashr_i64 s[4:5], s[6:7], 39

llvm/test/CodeGen/AMDGPU/shift-i128.ll

; GCN-NEXT:    s_setpc_b64 s[30:31]
   ret i128 %shl
 }
 
 define i128 @v_shl_i128_kv(i128 %rhs) {
 ; GCN-LABEL: v_shl_i128_kv:
 ; GCN:       ; %bb.0:
 ; GCN-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GCN-NEXT:    v_sub_i32_e32 v1, vcc, 64, v0
-; GCN-NEXT:    v_lshr_b64 v[2:3], 17, v1
-; GCN-NEXT:    v_subrev_i32_e32 v1, vcc, 64, v0
-; GCN-NEXT:    v_lshl_b64 v[4:5], 17, v1
+; GCN-NEXT:    v_lshr_b64 v[1:2], 17, v1
+; GCN-NEXT:    v_subrev_i32_e32 v2, vcc, 64, v0
+; GCN-NEXT:    v_lshl_b64 v[2:3], 17, v2
 ; GCN-NEXT:    v_cmp_gt_u32_e32 vcc, 64, v0
-; GCN-NEXT:    v_cndmask_b32_e32 v1, v4, v2, vcc
+; GCN-NEXT:    v_lshl_b64 v[4:5], 17, v0
 ; GCN-NEXT:    v_cmp_ne_u32_e64 s[4:5], 0, v0
+; GCN-NEXT:    v_cndmask_b32_e32 v1, v2, v1, vcc
+; GCN-NEXT:    v_cndmask_b32_e64 v3, v3, 0, vcc
 ; GCN-NEXT:    v_cndmask_b32_e64 v2, 0, v1, s[4:5]
-; GCN-NEXT:    v_lshl_b64 v[0:1], 17, v0
-; GCN-NEXT:    v_cndmask_b32_e32 v3, v5, v3, vcc
+; GCN-NEXT:    v_cndmask_b32_e32 v0, 0, v4, vcc
+; GCN-NEXT:    v_cndmask_b32_e32 v1, 0, v5, vcc
 ; GCN-NEXT:    v_cndmask_b32_e64 v3, 0, v3, s[4:5]
-; GCN-NEXT:    v_cndmask_b32_e32 v0, 0, v0, vcc
-; GCN-NEXT:    v_cndmask_b32_e32 v1, 0, v1, vcc
 ; GCN-NEXT:    s_setpc_b64 s[30:31]
   %shl = shl i128 17, %rhs
   ret i128 %shl
 }
 
 define i128 @v_lshr_i128_kv(i128 %rhs) {
 ; GCN-LABEL: v_lshr_i128_kv:
 ; GCN:       ; %bb.0:
 ; GCN-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GCN-NEXT:    s_movk_i32 s4, 0x41
 ; GCN-NEXT:    s_mov_b32 s5, 0
+; GCN-NEXT:    s_movk_i32 s4, 0x41
 ; GCN-NEXT:    v_lshr_b64 v[1:2], s[4:5], v0
 ; GCN-NEXT:    v_cmp_gt_u32_e32 vcc, 64, v0
-; GCN-NEXT:    v_mov_b32_e32 v3, s4
-; GCN-NEXT:    v_cmp_ne_u32_e64 s[4:5], 0, v0
 ; GCN-NEXT:    v_cndmask_b32_e32 v1, 0, v1, vcc
-; GCN-NEXT:    s_and_b64 vcc, s[4:5], vcc
-; GCN-NEXT:    v_cndmask_b32_e64 v0, v3, v1, s[4:5]
-; GCN-NEXT:    v_cndmask_b32_e32 v1, 0, v2, vcc
+; GCN-NEXT:    v_mov_b32_e32 v2, s4
+; GCN-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v0
+; GCN-NEXT:    v_cndmask_b32_e32 v0, v2, v1, vcc
+; GCN-NEXT:    v_mov_b32_e32 v1, 0
 ; GCN-NEXT:    v_mov_b32_e32 v2, 0
 ; GCN-NEXT:    v_mov_b32_e32 v3, 0
 ; GCN-NEXT:    s_setpc_b64 s[30:31]
   %shl = lshr i128 65, %rhs
   ret i128 %shl
 }
 
 define i128 @v_ashr_i128_kv(i128 %rhs) {
 ; GCN-LABEL: v_ashr_i128_kv:
 ; GCN:       ; %bb.0:
 ; GCN-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GCN-NEXT:    v_lshr_b64 v[1:2], 33, v0
 ; GCN-NEXT:    v_cmp_gt_u32_e32 vcc, 64, v0
-; GCN-NEXT:    v_cmp_ne_u32_e64 s[4:5], 0, v0
 ; GCN-NEXT:    v_cndmask_b32_e32 v1, 0, v1, vcc
-; GCN-NEXT:    s_and_b64 vcc, s[4:5], vcc
-; GCN-NEXT:    v_cndmask_b32_e64 v0, 33, v1, s[4:5]
-; GCN-NEXT:    v_cndmask_b32_e32 v1, 0, v2, vcc
+; GCN-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v0
+; GCN-NEXT:    v_cndmask_b32_e32 v0, 33, v1, vcc
+; GCN-NEXT:    v_mov_b32_e32 v1, 0
 ; GCN-NEXT:    v_mov_b32_e32 v2, 0
 ; GCN-NEXT:    v_mov_b32_e32 v3, 0
 ; GCN-NEXT:    s_setpc_b64 s[30:31]
   %shl = ashr i128 33, %rhs
   ret i128 %shl
 }
 
 define amdgpu_kernel void @s_shl_i128_ss(i128 %lhs, i128 %rhs) {

llvm/test/CodeGen/AMDGPU/srem64.ll

; GCN-IR-NEXT:    s_setpc_b64 s[30:31]
   %result = srem i64 %x, %y
   ret i64 %result
 }
 
 define amdgpu_kernel void @s_test_srem23_64(i64 addrspace(1)* %out, i64 %x, i64 %y) {
 ; GCN-LABEL: s_test_srem23_64:
 ; GCN:       ; %bb.0:
 ; GCN-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x9
-; GCN-NEXT:    s_load_dword s1, s[0:1], 0xe
+; GCN-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xd
 ; GCN-NEXT:    s_mov_b32 s3, 0xf000
 ; GCN-NEXT:    s_mov_b32 s2, -1
 ; GCN-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-NEXT:    s_ashr_i64 s[6:7], s[6:7], 41
 ; GCN-NEXT:    s_ashr_i64 s[0:1], s[0:1], 41
 ; GCN-NEXT:    v_cvt_f32_i32_e32 v0, s0
 ; GCN-NEXT:    v_cvt_f32_i32_e32 v1, s6
 ; GCN-NEXT:    s_xor_b32 s1, s6, s0

 ; GCN-NEXT:    v_bfe_i32 v0, v0, 0, 23
 ; GCN-NEXT:    v_ashrrev_i32_e32 v1, 31, v0
 ; GCN-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
 ; GCN-NEXT:    s_endpgm
 ;
 ; GCN-IR-LABEL: s_test_srem23_64:
 ; GCN-IR:       ; %bb.0:
 ; GCN-IR-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x9
-; GCN-IR-NEXT:    s_load_dword s1, s[0:1], 0xe
+; GCN-IR-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xd
 ; GCN-IR-NEXT:    s_mov_b32 s3, 0xf000
 ; GCN-IR-NEXT:    s_mov_b32 s2, -1
 ; GCN-IR-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-IR-NEXT:    s_ashr_i64 s[6:7], s[6:7], 41
 ; GCN-IR-NEXT:    s_ashr_i64 s[0:1], s[0:1], 41
 ; GCN-IR-NEXT:    v_cvt_f32_i32_e32 v0, s0
 ; GCN-IR-NEXT:    v_cvt_f32_i32_e32 v1, s6
 ; GCN-IR-NEXT:    s_xor_b32 s1, s6, s0
; GCN-IR-NEXT:    s_endpgm
   store i64 %result, i64 addrspace(1)* %out
   ret void
 }
 
 define amdgpu_kernel void @s_test_srem24_64(i64 addrspace(1)* %out, i64 %x, i64 %y) {
 ; GCN-LABEL: s_test_srem24_64:
 ; GCN:       ; %bb.0:
 ; GCN-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x9
-; GCN-NEXT:    s_load_dword s1, s[0:1], 0xe
+; GCN-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xd
 ; GCN-NEXT:    s_mov_b32 s3, 0xf000
 ; GCN-NEXT:    s_mov_b32 s2, -1
 ; GCN-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-NEXT:    s_ashr_i64 s[6:7], s[6:7], 40
 ; GCN-NEXT:    s_ashr_i64 s[0:1], s[0:1], 40
 ; GCN-NEXT:    v_cvt_f32_i32_e32 v0, s0
 ; GCN-NEXT:    v_cvt_f32_i32_e32 v1, s6
 ; GCN-NEXT:    s_xor_b32 s1, s6, s0

 ; GCN-NEXT:    v_bfe_i32 v0, v0, 0, 24
 ; GCN-NEXT:    v_ashrrev_i32_e32 v1, 31, v0
 ; GCN-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
 ; GCN-NEXT:    s_endpgm
 ;
 ; GCN-IR-LABEL: s_test_srem24_64:
 ; GCN-IR:       ; %bb.0:
 ; GCN-IR-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x9
-; GCN-IR-NEXT:    s_load_dword s1, s[0:1], 0xe
+; GCN-IR-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xd
 ; GCN-IR-NEXT:    s_mov_b32 s3, 0xf000
 ; GCN-IR-NEXT:    s_mov_b32 s2, -1
 ; GCN-IR-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-IR-NEXT:    s_ashr_i64 s[6:7], s[6:7], 40
 ; GCN-IR-NEXT:    s_ashr_i64 s[0:1], s[0:1], 40
 ; GCN-IR-NEXT:    v_cvt_f32_i32_e32 v0, s0
 ; GCN-IR-NEXT:    v_cvt_f32_i32_e32 v1, s6
 ; GCN-IR-NEXT:    s_xor_b32 s1, s6, s0
; GCN-IR-NEXT:    s_setpc_b64 s[30:31]
   %result = srem i64 %1, %2
   ret i64 %result
 }
 
 define amdgpu_kernel void @s_test_srem25_64(i64 addrspace(1)* %out, i64 %x, i64 %y) {
 ; GCN-LABEL: s_test_srem25_64:
 ; GCN:       ; %bb.0:
 ; GCN-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x9
-; GCN-NEXT:    s_load_dword s1, s[0:1], 0xe
+; GCN-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xd
 ; GCN-NEXT:    s_mov_b32 s3, 0xf000
 ; GCN-NEXT:    s_mov_b32 s2, -1
 ; GCN-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-NEXT:    s_ashr_i64 s[6:7], s[6:7], 39
 ; GCN-NEXT:    s_ashr_i64 s[0:1], s[0:1], 39
 ; GCN-NEXT:    v_cvt_f32_i32_e32 v0, s0
 ; GCN-NEXT:    v_cvt_f32_i32_e32 v1, s6
 ; GCN-NEXT:    s_xor_b32 s1, s6, s0

 ; GCN-NEXT:    v_bfe_i32 v0, v0, 0, 25
 ; GCN-NEXT:    v_ashrrev_i32_e32 v1, 31, v0
 ; GCN-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
 ; GCN-NEXT:    s_endpgm
 ;
 ; GCN-IR-LABEL: s_test_srem25_64:
 ; GCN-IR:       ; %bb.0:
 ; GCN-IR-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x9
-; GCN-IR-NEXT:    s_load_dword s1, s[0:1], 0xe
+; GCN-IR-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xd
 ; GCN-IR-NEXT:    s_mov_b32 s3, 0xf000
 ; GCN-IR-NEXT:    s_mov_b32 s2, -1
 ; GCN-IR-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-IR-NEXT:    s_ashr_i64 s[6:7], s[6:7], 39
 ; GCN-IR-NEXT:    s_ashr_i64 s[0:1], s[0:1], 39
 ; GCN-IR-NEXT:    v_cvt_f32_i32_e32 v0, s0
 ; GCN-IR-NEXT:    v_cvt_f32_i32_e32 v1, s6
 ; GCN-IR-NEXT:    s_xor_b32 s1, s6, s0
; GCN-IR-NEXT:    s_endpgm
   store i64 %result, i64 addrspace(1)* %out
   ret void
 }
 
 define amdgpu_kernel void @s_test_srem31_64(i64 addrspace(1)* %out, i64 %x, i64 %y) {
 ; GCN-LABEL: s_test_srem31_64:
 ; GCN:       ; %bb.0:
 ; GCN-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x9
-; GCN-NEXT:    s_load_dword s1, s[0:1], 0xe
+; GCN-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xd
 ; GCN-NEXT:    s_mov_b32 s3, 0xf000
 ; GCN-NEXT:    s_mov_b32 s2, -1
 ; GCN-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-NEXT:    s_ashr_i64 s[6:7], s[6:7], 33
 ; GCN-NEXT:    s_ashr_i64 s[0:1], s[0:1], 33
 ; GCN-NEXT:    v_cvt_f32_i32_e32 v0, s0
 ; GCN-NEXT:    v_cvt_f32_i32_e32 v1, s6
 ; GCN-NEXT:    s_xor_b32 s1, s6, s0

 ; GCN-NEXT:    v_bfe_i32 v0, v0, 0, 31
 ; GCN-NEXT:    v_ashrrev_i32_e32 v1, 31, v0
 ; GCN-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
 ; GCN-NEXT:    s_endpgm
 ;
 ; GCN-IR-LABEL: s_test_srem31_64:
 ; GCN-IR:       ; %bb.0:
 ; GCN-IR-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x9
-; GCN-IR-NEXT:    s_load_dword s1, s[0:1], 0xe
+; GCN-IR-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xd
 ; GCN-IR-NEXT:    s_mov_b32 s3, 0xf000
 ; GCN-IR-NEXT:    s_mov_b32 s2, -1
 ; GCN-IR-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-IR-NEXT:    s_ashr_i64 s[6:7], s[6:7], 33
 ; GCN-IR-NEXT:    s_ashr_i64 s[0:1], s[0:1], 33
 ; GCN-IR-NEXT:    v_cvt_f32_i32_e32 v0, s0
 ; GCN-IR-NEXT:    v_cvt_f32_i32_e32 v1, s6
 ; GCN-IR-NEXT:    s_xor_b32 s1, s6, s0

llvm/test/CodeGen/AMDGPU/ssubsat.ll

; GFX9-NEXT:    s_setpc_b64 s[30:31]
   %result = call <3 x i16> @llvm.ssub.sat.v3i16(<3 x i16> %lhs, <3 x i16> %rhs)
   ret <3 x i16> %result
 }
 
 define <2 x float> @v_ssubsat_v4i16(<4 x i16> %lhs, <4 x i16> %rhs) {
 ; GFX6-LABEL: v_ssubsat_v4i16:
 ; GFX6:       ; %bb.0:
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX6-NEXT:    v_bfe_i32 v8, v1, 0, 16
+; GFX6-NEXT:    v_bfe_i32 v1, v6, 0, 16
+; GFX6-NEXT:    v_bfe_i32 v2, v2, 0, 16
 ; GFX6-NEXT:    v_bfe_i32 v4, v4, 0, 16
 ; GFX6-NEXT:    v_bfe_i32 v0, v0, 0, 16
 ; GFX6-NEXT:    v_bfe_i32 v5, v5, 0, 16
-; GFX6-NEXT:    v_bfe_i32 v1, v1, 0, 16
-; GFX6-NEXT:    v_sub_i32_e32 v1, vcc, v1, v5
+; GFX6-NEXT:    v_bfe_i32 v6, v7, 0, 16
+; GFX6-NEXT:    v_bfe_i32 v3, v3, 0, 16
+; GFX6-NEXT:    v_sub_i32_e32 v1, vcc, v2, v1
 ; GFX6-NEXT:    s_movk_i32 s4, 0x7fff
+; GFX6-NEXT:    v_sub_i32_e32 v3, vcc, v3, v6
+; GFX6-NEXT:    v_sub_i32_e32 v2, vcc, v8, v5
 ; GFX6-NEXT:    v_sub_i32_e32 v0, vcc, v0, v4
-; GFX6-NEXT:    v_min_i32_e32 v1, s4, v1
+; GFX6-NEXT:    v_min_i32_e32 v3, s4, v3
 ; GFX6-NEXT:    s_movk_i32 s5, 0x8000
+; GFX6-NEXT:    v_min_i32_e32 v1, s4, v1
+; GFX6-NEXT:    v_min_i32_e32 v2, s4, v2
 ; GFX6-NEXT:    v_min_i32_e32 v0, s4, v0
+; GFX6-NEXT:    v_max_i32_e32 v3, s5, v3
+; GFX6-NEXT:    v_max_i32_e32 v2, s5, v2
 ; GFX6-NEXT:    v_max_i32_e32 v1, s5, v1
-; GFX6-NEXT:    v_max_i32_e32 v0, s5, v0
 ; GFX6-NEXT:    s_mov_b32 s6, 0xffff
-; GFX6-NEXT:    v_bfe_i32 v6, v6, 0, 16
-; GFX6-NEXT:    v_bfe_i32 v2, v2, 0, 16
-; GFX6-NEXT:    v_bfe_i32 v7, v7, 0, 16
-; GFX6-NEXT:    v_bfe_i32 v3, v3, 0, 16
-; GFX6-NEXT:    v_lshlrev_b32_e32 v1, 16, v1
+; GFX6-NEXT:    v_max_i32_e32 v0, s5, v0
+; GFX6-NEXT:    v_lshlrev_b32_e32 v3, 16, v3
+; GFX6-NEXT:    v_and_b32_e32 v1, s6, v1
+; GFX6-NEXT:    v_lshlrev_b32_e32 v2, 16, v2
 ; GFX6-NEXT:    v_and_b32_e32 v0, s6, v0
-; GFX6-NEXT:    v_sub_i32_e32 v2, vcc, v2, v6
-; GFX6-NEXT:    v_or_b32_e32 v0, v0, v1
-; GFX6-NEXT:    v_sub_i32_e32 v1, vcc, v3, v7
-; GFX6-NEXT:    v_min_i32_e32 v1, s4, v1
-; GFX6-NEXT:    v_min_i32_e32 v2, s4, v2
-; GFX6-NEXT:    v_max_i32_e32 v1, s5, v1
-; GFX6-NEXT:    v_max_i32_e32 v2, s5, v2
-; GFX6-NEXT:    v_lshlrev_b32_e32 v1, 16, v1
-; GFX6-NEXT:    v_and_b32_e32 v2, s6, v2
-; GFX6-NEXT:    v_or_b32_e32 v1, v2, v1
+; GFX6-NEXT:    v_or_b32_e32 v1, v1, v3
+; GFX6-NEXT:    v_or_b32_e32 v0, v0, v2
 ; GFX6-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX8-LABEL: v_ssubsat_v4i16:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX8-NEXT:    v_lshrrev_b32_e32 v4, 16, v2
 ; GFX8-NEXT:    v_lshrrev_b32_e32 v5, 16, v0
 ; GFX8-NEXT:    v_sub_u16_e32 v6, v5, v4

llvm/test/CodeGen/AMDGPU/sub.v2i16.ll

 ; GFX9-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    global_load_dword v2, v0, s[6:7] glc
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    global_load_dword v3, v0, s[2:3] glc
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    s_mov_b32 s7, 0xf000
 ; GFX9-NEXT:    s_mov_b32 s6, -1
-; GFX9-NEXT:    v_pk_sub_i16 v2, v2, v3
-; GFX9-NEXT:    v_and_b32_e32 v0, 0xffff, v2
-; GFX9-NEXT:    v_lshrrev_b32_e32 v2, 16, v2
+; GFX9-NEXT:    v_pk_sub_i16 v0, v2, v3
+; GFX9-NEXT:    v_lshrrev_b32_e32 v2, 16, v0
+; GFX9-NEXT:    v_and_b32_e32 v0, 0xffff, v0
 ; GFX9-NEXT:    v_mov_b32_e32 v3, v1
 ; GFX9-NEXT:    buffer_store_dwordx4 v[0:3], off, s[4:7], 0
 ; GFX9-NEXT:    s_endpgm
 ;
 ; VI-LABEL: v_test_sub_v2i16_zext_to_v2i64:
 ; VI:       ; %bb.0:
 ; VI-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
 ; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x34

llvm/test/CodeGen/AMDGPU/uaddsat.ll

; GFX9-NEXT:    s_setpc_b64 s[30:31]
   ret <3 x i16> %result
 }
 
 define <2 x float> @v_uaddsat_v4i16(<4 x i16> %lhs, <4 x i16> %rhs) {
 ; GFX6-LABEL: v_uaddsat_v4i16:
 ; GFX6:       ; %bb.0:
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    s_mov_b32 s4, 0xffff
-; GFX6-NEXT:    v_and_b32_e32 v5, s4, v5
-; GFX6-NEXT:    v_and_b32_e32 v1, s4, v1
-; GFX6-NEXT:    v_add_i32_e32 v1, vcc, v1, v5
-; GFX6-NEXT:    v_and_b32_e32 v4, s4, v4
-; GFX6-NEXT:    v_and_b32_e32 v0, s4, v0
-; GFX6-NEXT:    v_add_i32_e32 v0, vcc, v0, v4
-; GFX6-NEXT:    v_min_u32_e32 v1, s4, v1
-; GFX6-NEXT:    v_and_b32_e32 v7, s4, v7
+; GFX6-NEXT:    v_and_b32_e32 v8, s4, v1
+; GFX6-NEXT:    v_and_b32_e32 v1, s4, v7
 ; GFX6-NEXT:    v_and_b32_e32 v3, s4, v3
+; GFX6-NEXT:    v_add_i32_e32 v1, vcc, v3, v1
 ; GFX6-NEXT:    v_and_b32_e32 v6, s4, v6
 ; GFX6-NEXT:    v_and_b32_e32 v2, s4, v2
-; GFX6-NEXT:    v_min_u32_e32 v0, s4, v0
+; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v2, v6
+; GFX6-NEXT:    v_min_u32_e32 v1, s4, v1
+; GFX6-NEXT:    v_and_b32_e32 v5, s4, v5
+; GFX6-NEXT:    v_min_u32_e32 v2, s4, v2
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v1, 16, v1
-; GFX6-NEXT:    v_or_b32_e32 v0, v0, v1
-; GFX6-NEXT:    v_add_i32_e32 v1, vcc, v2, v6
-; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v3, v7
+; GFX6-NEXT:    v_or_b32_e32 v1, v2, v1
+; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v8, v5
+; GFX6-NEXT:    v_and_b32_e32 v4, s4, v4
+; GFX6-NEXT:    v_and_b32_e32 v0, s4, v0
+; GFX6-NEXT:    v_add_i32_e32 v0, vcc, v0, v4
 ; GFX6-NEXT:    v_min_u32_e32 v2, s4, v2
-; GFX6-NEXT:    v_min_u32_e32 v1, s4, v1
+; GFX6-NEXT:    v_min_u32_e32 v0, s4, v0
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v2, 16, v2
-; GFX6-NEXT:    v_or_b32_e32 v1, v1, v2
+; GFX6-NEXT:    v_or_b32_e32 v0, v0, v2
 ; GFX6-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX8-LABEL: v_uaddsat_v4i16:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX8-NEXT:    v_add_u16_sdwa v4, v0, v2 clamp dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
 ; GFX8-NEXT:    v_add_u16_e64 v0, v0, v2 clamp
 ; GFX8-NEXT:    v_add_u16_sdwa v2, v1, v3 clamp dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1

llvm/test/CodeGen/AMDGPU/usubsat.ll

; GFX9-NEXT:    s_setpc_b64 s[30:31]
   ret <3 x i16> %result
 }
 
 define <2 x float> @v_usubsat_v4i16(<4 x i16> %lhs, <4 x i16> %rhs) {
 ; GFX6-LABEL: v_usubsat_v4i16:
 ; GFX6:       ; %bb.0:
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    s_mov_b32 s4, 0xffff
-; GFX6-NEXT:    v_and_b32_e32 v9, s4, v5
-; GFX6-NEXT:    v_and_b32_e32 v1, s4, v1
-; GFX6-NEXT:    v_max_u32_e32 v1, v1, v9
-; GFX6-NEXT:    v_and_b32_e32 v4, s4, v4
-; GFX6-NEXT:    v_and_b32_e32 v0, s4, v0
-; GFX6-NEXT:    v_max_u32_e32 v0, v0, v4
-; GFX6-NEXT:    v_sub_i32_e32 v1, vcc, v1, v5
-; GFX6-NEXT:    v_and_b32_e32 v8, s4, v7
+; GFX6-NEXT:    v_and_b32_e32 v9, s4, v1
+; GFX6-NEXT:    v_and_b32_e32 v1, s4, v7
 ; GFX6-NEXT:    v_and_b32_e32 v3, s4, v3
+; GFX6-NEXT:    v_max_u32_e32 v1, v3, v1
 ; GFX6-NEXT:    v_and_b32_e32 v6, s4, v6
 ; GFX6-NEXT:    v_and_b32_e32 v2, s4, v2
-; GFX6-NEXT:    v_sub_i32_e32 v0, vcc, v0, v4
+; GFX6-NEXT:    v_max_u32_e32 v2, v2, v6
+; GFX6-NEXT:    v_sub_i32_e32 v1, vcc, v1, v7
+; GFX6-NEXT:    v_and_b32_e32 v8, s4, v5
+; GFX6-NEXT:    v_sub_i32_e32 v2, vcc, v2, v6
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v1, 16, v1
-; GFX6-NEXT:    v_or_b32_e32 v0, v0, v1
-; GFX6-NEXT:    v_max_u32_e32 v1, v2, v6
-; GFX6-NEXT:    v_max_u32_e32 v2, v3, v8
-; GFX6-NEXT:    v_sub_i32_e32 v2, vcc, v2, v7
-; GFX6-NEXT:    v_sub_i32_e32 v1, vcc, v1, v6
+; GFX6-NEXT:    v_or_b32_e32 v1, v2, v1
+; GFX6-NEXT:    v_max_u32_e32 v2, v9, v8
+; GFX6-NEXT:    v_and_b32_e32 v4, s4, v4
+; GFX6-NEXT:    v_and_b32_e32 v0, s4, v0
+; GFX6-NEXT:    v_max_u32_e32 v0, v0, v4
+; GFX6-NEXT:    v_sub_i32_e32 v2, vcc, v2, v5
+; GFX6-NEXT:    v_sub_i32_e32 v0, vcc, v0, v4
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v2, 16, v2
-; GFX6-NEXT:    v_or_b32_e32 v1, v1, v2
+; GFX6-NEXT:    v_or_b32_e32 v0, v0, v2
 ; GFX6-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX8-LABEL: v_usubsat_v4i16:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX8-NEXT:    v_sub_u16_sdwa v4, v0, v2 clamp dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
 ; GFX8-NEXT:    v_sub_u16_e64 v0, v0, v2 clamp
 ; GFX8-NEXT:    v_sub_u16_sdwa v2, v1, v3 clamp dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1

llvm/test/CodeGen/ARM/dagcombine-anyexttozeroext.ll

; CHECK-NEXT:    bx lr
   ret float %3
 }
 
 ; Make sure we generate zext from <4 x i8> to <4 x 32>.
 define <4 x i32> @h(<4 x i8> *%in) {
 ; CHECK-LABEL: h:
 ; CHECK:       @ %bb.0:
 ; CHECK-NEXT:    vld1.32 {d16[0]}, [r0:32]
 ; CHECK-NEXT:    vmovl.u8 q8, d16
-; CHECK-NEXT:    vmovl.u16 q8, d16
-; CHECK-NEXT:    vmov r0, r1, d16
-; CHECK-NEXT:    vmov r2, r3, d17
+; CHECK-NEXT:    vmov.u16 r0, d16[2]
+; CHECK-NEXT:    vmov.u16 r1, d16[3]
+; CHECK-NEXT:    vmov.u16 r12, d16[1]
+; CHECK-NEXT:    uxtb r0, r0
+; CHECK-NEXT:    vmov.32 d19[0], r0
+; CHECK-NEXT:    uxtb r1, r1
+; CHECK-NEXT:    vmov.u16 r0, d16[0]
+; CHECK-NEXT:    vmov.32 d19[1], r1
+; CHECK-NEXT:    uxtb r1, r12
+; CHECK-NEXT:    vmov r2, r3, d19
+; CHECK-NEXT:    uxtb r0, r0
 ; CHECK-NEXT:    bx lr
   %1 = load <4 x i8>, <4 x i8>* %in, align 4
   %2 = extractelement <4 x i8> %1, i32 0
   %3 = zext i8 %2 to i32
   %4 = insertelement <4 x i32> undef, i32 %3, i32 0
   %5 = extractelement <4 x i8> %1, i32 1
   %6 = zext i8 %5 to i32
   %7 = insertelement <4 x i32> %4, i32 %6, i32 1

llvm/test/CodeGen/ARM/fp16-insert-extract.ll

 define <8 x half> @test_vset_laneq_f16_1(<8 x half> %a, float %fb) nounwind {
 ; CHECKHARD-LABEL: test_vset_laneq_f16_1:
 ; CHECKHARD:       @ %bb.0: @ %entry
 ; CHECKHARD-NEXT:    vcvtt.f16.f32 s0, s4
 ; CHECKHARD-NEXT:    bx lr
 ;
 ; CHECKSOFT-LABEL: test_vset_laneq_f16_1:
 ; CHECKSOFT:       @ %bb.0: @ %entry
-; CHECKSOFT-NEXT:    vmov d1, r2, r3
-; CHECKSOFT-NEXT:    vldr s4, [sp]
-; CHECKSOFT-NEXT:    vmov d0, r0, r1
-; CHECKSOFT-NEXT:    vcvtt.f16.f32 s0, s4
-; CHECKSOFT-NEXT:    vmov r2, r3, d1
-; CHECKSOFT-NEXT:    vmov r0, r1, d0
+; CHECKSOFT-NEXT:    vldr s0, [sp]
+; CHECKSOFT-NEXT:    vmov d2, r0, r1
+; CHECKSOFT-NEXT:    vcvtt.f16.f32 s4, s0
+; CHECKSOFT-NEXT:    vmov r0, r1, d2
 ; CHECKSOFT-NEXT:    bx lr
 entry:
   %b = fptrunc float %fb to half
   %x = insertelement <8 x half> %a, half %b, i32 1
   ret <8 x half> %x
 }
 
 define <8 x half> @test_vset_laneq_f16_7(<8 x half> %a, float %fb) nounwind {
 ; CHECKHARD-LABEL: test_vset_laneq_f16_7:
 ; CHECKHARD:       @ %bb.0: @ %entry
 ; CHECKHARD-NEXT:    vcvtt.f16.f32 s3, s4
 ; CHECKHARD-NEXT:    bx lr
 ;
 ; CHECKSOFT-LABEL: test_vset_laneq_f16_7:
 ; CHECKSOFT:       @ %bb.0: @ %entry
 ; CHECKSOFT-NEXT:    vmov d1, r2, r3
 ; CHECKSOFT-NEXT:    vldr s4, [sp]
 ; CHECKSOFT-NEXT:    vmov d0, r0, r1
 ; CHECKSOFT-NEXT:    vcvtt.f16.f32 s3, s4
-; CHECKSOFT-NEXT:    vmov r0, r1, d0
 ; CHECKSOFT-NEXT:    vmov r2, r3, d1
 ; CHECKSOFT-NEXT:    bx lr
 entry:
   %b = fptrunc float %fb to half
   %x = insertelement <8 x half> %a, half %b, i32 7
   ret <8 x half> %x
 }
 

llvm/test/CodeGen/ARM/vdup.ll

; CHECK-NEXT:    mov pc, lr
   %tmp1 = insertelement <2 x float> zeroinitializer, float %A, i32 0
   %tmp2 = insertelement <2 x float> %tmp1, float %A, i32 1
   ret <2 x float> %tmp2
 }
 
 define <16 x i8> @v_dupQ8(i8 %A) nounwind {
 ; CHECK-LABEL: v_dupQ8:
 ; CHECK:       @ %bb.0:
-; CHECK-NEXT:    vdup.8 q8, r0
+; CHECK-NEXT:    vmov.i32 d17, #0x0
+; CHECK-NEXT:    vdup.8 d16, r0
+; CHECK-NEXT:    vmov.8 d17[0], r0
+; CHECK-NEXT:    vmov.8 d17[1], r0
+; CHECK-NEXT:    vmov.8 d17[2], r0
+; CHECK-NEXT:    vmov.8 d17[3], r0
+; CHECK-NEXT:    vmov.8 d17[4], r0
+; CHECK-NEXT:    vmov.8 d17[5], r0
+; CHECK-NEXT:    vmov.8 d17[6], r0
+; CHECK-NEXT:    vmov.8 d17[7], r0
 ; CHECK-NEXT:    vmov r0, r1, d16
 ; CHECK-NEXT:    vmov r2, r3, d17
 ; CHECK-NEXT:    mov pc, lr
   %tmp1 = insertelement <16 x i8> zeroinitializer, i8 %A, i32 0
   %tmp2 = insertelement <16 x i8> %tmp1, i8 %A, i32 1
   %tmp3 = insertelement <16 x i8> %tmp2, i8 %A, i32 2
   %tmp4 = insertelement <16 x i8> %tmp3, i8 %A, i32 3
   %tmp5 = insertelement <16 x i8> %tmp4, i8 %A, i32 4
   %tmp6 = insertelement <16 x i8> %tmp5, i8 %A, i32 5
   %tmp7 = insertelement <16 x i8> %tmp6, i8 %A, i32 6
   %tmp8 = insertelement <16 x i8> %tmp7, i8 %A, i32 7
   %tmp9 = insertelement <16 x i8> %tmp8, i8 %A, i32 8
   %tmp10 = insertelement <16 x i8> %tmp9, i8 %A, i32 9
   %tmp11 = insertelement <16 x i8> %tmp10, i8 %A, i32 10
   %tmp12 = insertelement <16 x i8> %tmp11, i8 %A, i32 11
   %tmp13 = insertelement <16 x i8> %tmp12, i8 %A, i32 12
   %tmp14 = insertelement <16 x i8> %tmp13, i8 %A, i32 13
   %tmp15 = insertelement <16 x i8> %tmp14, i8 %A, i32 14
   %tmp16 = insertelement <16 x i8> %tmp15, i8 %A, i32 15
   ret <16 x i8> %tmp16
 }
 
 define <8 x i16> @v_dupQ16(i16 %A) nounwind {
 ; CHECK-LABEL: v_dupQ16:
 ; CHECK:       @ %bb.0:
-; CHECK-NEXT:    vdup.16 q8, r0
+; CHECK-NEXT:    vmov.i32 d17, #0x0
+; CHECK-NEXT:    vdup.16 d16, r0
+; CHECK-NEXT:    vmov.16 d17[0], r0
+; CHECK-NEXT:    vmov.16 d17[1], r0
+; CHECK-NEXT:    vmov.16 d17[2], r0
+; CHECK-NEXT:    vmov.16 d17[3], r0
 ; CHECK-NEXT:    vmov r0, r1, d16
 ; CHECK-NEXT:    vmov r2, r3, d17
 ; CHECK-NEXT:    mov pc, lr
   %tmp1 = insertelement <8 x i16> zeroinitializer, i16 %A, i32 0
   %tmp2 = insertelement <8 x i16> %tmp1, i16 %A, i32 1
   %tmp3 = insertelement <8 x i16> %tmp2, i16 %A, i32 2
   %tmp4 = insertelement <8 x i16> %tmp3, i16 %A, i32 3
   %tmp5 = insertelement <8 x i16> %tmp4, i16 %A, i32 4
   %tmp6 = insertelement <8 x i16> %tmp5, i16 %A, i32 5
   %tmp7 = insertelement <8 x i16> %tmp6, i16 %A, i32 6
   %tmp8 = insertelement <8 x i16> %tmp7, i16 %A, i32 7
   ret <8 x i16> %tmp8
 }
 
 define <4 x i32> @v_dupQ32(i32 %A) nounwind {
 ; CHECK-LABEL: v_dupQ32:
 ; CHECK:       @ %bb.0:
-; CHECK-NEXT:    vdup.32 q8, r0
+; CHECK-NEXT:    vdup.32 d16, r0
+; CHECK-NEXT:    vmov.32 d17[0], r0
+; CHECK-NEXT:    vmov.32 d17[1], r0
 ; CHECK-NEXT:    vmov r0, r1, d16
 ; CHECK-NEXT:    vmov r2, r3, d17
 ; CHECK-NEXT:    mov pc, lr
   %tmp1 = insertelement <4 x i32> zeroinitializer, i32 %A, i32 0
   %tmp2 = insertelement <4 x i32> %tmp1, i32 %A, i32 1
   %tmp3 = insertelement <4 x i32> %tmp2, i32 %A, i32 2
   %tmp4 = insertelement <4 x i32> %tmp3, i32 %A, i32 3
   ret <4 x i32> %tmp4
 }
 
 define <4 x float> @v_dupQfloat(float %A) nounwind {
 ; CHECK-LABEL: v_dupQfloat:
 ; CHECK:       @ %bb.0:
-; CHECK-NEXT:    vdup.32 q8, r0
-; CHECK-NEXT:    vmov r0, r1, d16
-; CHECK-NEXT:    vmov r2, r3, d17
+; CHECK-NEXT:    vmov s0, r0
+; CHECK-NEXT:    vmov.f32 s1, s0
+; CHECK-NEXT:    vmov r0, r1, d0
+; CHECK-NEXT:    vmov.f32 s2, s0
+; CHECK-NEXT:    vmov.f32 s3, s0
+; CHECK-NEXT:    vmov r2, r3, d1
 ; CHECK-NEXT:    mov pc, lr
   %tmp1 = insertelement <4 x float> zeroinitializer, float %A, i32 0
   %tmp2 = insertelement <4 x float> %tmp1, float %A, i32 1
   %tmp3 = insertelement <4 x float> %tmp2, float %A, i32 2
   %tmp4 = insertelement <4 x float> %tmp3, float %A, i32 3
   ret <4 x float> %tmp4
 }
 
; CHECK-NEXT:    mov pc, lr
   %2 = shufflevector <8 x i8> %1, <8 x i8> undef, <8 x i32> zeroinitializer
   store <8 x i8> %2, <8 x i8>* %ptr, align 8
   ret void
 }
 
 define <4 x i32> @tdupi(i32 %x, i32 %y) {
 ; CHECK-LABEL: tdupi:
 ; CHECK:       @ %bb.0:
-; CHECK-NEXT:    vdup.32 q8, r0
+; CHECK-NEXT:    vdup.32 d16, r0
+; CHECK-NEXT:    vmov.32 d17[0], r0
 ; CHECK-NEXT:    vmov.32 d17[1], r1
 ; CHECK-NEXT:    vmov r0, r1, d16
 ; CHECK-NEXT:    vmov r2, r3, d17
 ; CHECK-NEXT:    mov pc, lr
   %1 = insertelement <4 x i32> undef, i32 %x, i32 0
   %2 = insertelement <4 x i32> %1, i32 %x, i32 1
   %3 = insertelement <4 x i32> %2, i32 %x, i32 2
   %4 = insertelement <4 x i32> %3, i32 %y, i32 3
   ret <4 x i32> %4
 }
 
 define <4 x float> @tdupf(float %x, float %y) {
 ; CHECK-LABEL: tdupf:
 ; CHECK:       @ %bb.0:
-; CHECK-NEXT:    vdup.32 q0, r0
+; CHECK-NEXT:    vmov s0, r0
 ; CHECK-NEXT:    vmov s3, r1
+; CHECK-NEXT:    vmov.f32 s1, s0
 ; CHECK-NEXT:    vmov r0, r1, d0
+; CHECK-NEXT:    vmov.f32 s2, s0
 ; CHECK-NEXT:    vmov r2, r3, d1
 ; CHECK-NEXT:    mov pc, lr
   %1 = insertelement <4 x float> undef, float %x, i32 0
   %2 = insertelement <4 x float> %1, float %x, i32 1
   %3 = insertelement <4 x float> %2, float %x, i32 2
   %4 = insertelement <4 x float> %3, float %y, i32 3
   ret <4 x float> %4
 }
 
 ; This test checks that when splatting an element from a vector into another,
 ; the value isn't moved out to GPRs first.
 define <4 x i32> @tduplane(<4 x i32> %invec) {
 ; CHECK-LABEL: tduplane:
 ; CHECK:       @ %bb.0:
 ; CHECK-NEXT:    vmov d16, r0, r1
-; CHECK-NEXT:    mov r0, #255
+; CHECK-NEXT:    vmov.32 r0, d16[1]
+; CHECK-NEXT:    vmov d17, r2, r3
 ; CHECK-NEXT:    vdup.32 q8, d16[1]
+; CHECK-NEXT:    vmov.32 d17[0], r0
+; CHECK-NEXT:    mov r0, #255
 ; CHECK-NEXT:    vmov.32 d17[1], r0
 ; CHECK-NEXT:    vmov r0, r1, d16
 ; CHECK-NEXT:    vmov r2, r3, d17
 ; CHECK-NEXT:    mov pc, lr
   %in = extractelement <4 x i32> %invec, i32 1
   %1 = insertelement <4 x i32> undef, i32 %in, i32 0
   %2 = insertelement <4 x i32> %1, i32 %in, i32 1
   %3 = insertelement <4 x i32> %2, i32 %in, i32 2

llvm/test/CodeGen/ARM/vldlane.ll

 ; CHECK-NEXT:    mov pc, lr
   %tmp1 = load <2 x float>, <2 x float>* %B
   %tmp2 = load float, float* %A, align 4
   %tmp3 = insertelement <2 x float> %tmp1, float %tmp2, i32 1
   ret <2 x float> %tmp3
 }
 
 define <16 x i8> @vld1laneQi8(i8* %A, <16 x i8>* %B) nounwind {
-; CHECK-LABEL: vld1laneQi8:
-; CHECK:       @ %bb.0:
-; CHECK-NEXT:    vld1.64 {d16, d17}, [r1]
-; CHECK-NEXT:    vld1.8 {d17[1]}, [r0]
-; CHECK-NEXT:    vmov r0, r1, d16
-; CHECK-NEXT:    vmov r2, r3, d17
-; CHECK-NEXT:    mov pc, lr
+; DEFAULT-LABEL: vld1laneQi8:
+; DEFAULT:       @ %bb.0:
+; DEFAULT-NEXT:    vld1.64 {d16, d17}, [r1]
+; DEFAULT-NEXT:    vorr q9, q8, q8
+; DEFAULT-NEXT:    vld1.8 {d19[1]}, [r0]
+; DEFAULT-NEXT:    vmov r0, r1, d16
+; DEFAULT-NEXT:    vmov r2, r3, d19
+; DEFAULT-NEXT:    mov pc, lr
+;
+; BASIC-LABEL: vld1laneQi8:
+; BASIC:       @ %bb.0:
+; BASIC-NEXT:    vld1.64 {d18, d19}, [r1]
+; BASIC-NEXT:    vorr q8, q9, q9
+; BASIC-NEXT:    vld1.8 {d17[1]}, [r0]
+; BASIC-NEXT:    vmov r0, r1, d18
+; BASIC-NEXT:    vmov r2, r3, d17
+; BASIC-NEXT:    mov pc, lr
   %tmp1 = load <16 x i8>, <16 x i8>* %B
   %tmp2 = load i8, i8* %A, align 8
   %tmp3 = insertelement <16 x i8> %tmp1, i8 %tmp2, i32 9
   ret <16 x i8> %tmp3
 }
 
 define <8 x i16> @vld1laneQi16(i16* %A, <8 x i16>* %B) nounwind {
-; CHECK-LABEL: vld1laneQi16:
-; CHECK:       @ %bb.0:
-; CHECK-NEXT:    vld1.64 {d16, d17}, [r1]
-; CHECK-NEXT:    vld1.16 {d17[1]}, [r0:16]
-; CHECK-NEXT:    vmov r0, r1, d16
-; CHECK-NEXT:    vmov r2, r3, d17
-; CHECK-NEXT:    mov pc, lr
+; DEFAULT-LABEL: vld1laneQi16:
+; DEFAULT:       @ %bb.0:
+; DEFAULT-NEXT:    vld1.64 {d16, d17}, [r1]
+; DEFAULT-NEXT:    vorr q9, q8, q8
+; DEFAULT-NEXT:    vld1.16 {d19[1]}, [r0:16]
+; DEFAULT-NEXT:    vmov r0, r1, d16
+; DEFAULT-NEXT:    vmov r2, r3, d19
+; DEFAULT-NEXT:    mov pc, lr
+;
+; BASIC-LABEL: vld1laneQi16:
+; BASIC:       @ %bb.0:
+; BASIC-NEXT:    vld1.64 {d18, d19}, [r1]
+; BASIC-NEXT:    vorr q8, q9, q9
+; BASIC-NEXT:    vld1.16 {d17[1]}, [r0:16]
+; BASIC-NEXT:    vmov r0, r1, d18
+; BASIC-NEXT:    vmov r2, r3, d17
+; BASIC-NEXT:    mov pc, lr
   %tmp1 = load <8 x i16>, <8 x i16>* %B
   %tmp2 = load i16, i16* %A, align 8
   %tmp3 = insertelement <8 x i16> %tmp1, i16 %tmp2, i32 5
   ret <8 x i16> %tmp3
 }
 
 define <4 x i32> @vld1laneQi32(i32* %A, <4 x i32>* %B) nounwind {
-; CHECK-LABEL: vld1laneQi32:
-; CHECK:       @ %bb.0:
-; CHECK-NEXT:    vld1.64 {d16, d17}, [r1]
-; CHECK-NEXT:    vld1.32 {d17[1]}, [r0:32]
-; CHECK-NEXT:    vmov r0, r1, d16
-; CHECK-NEXT:    vmov r2, r3, d17
-; CHECK-NEXT:    mov pc, lr
+; DEFAULT-LABEL: vld1laneQi32:
+; DEFAULT:       @ %bb.0:
+; DEFAULT-NEXT:    vld1.64 {d16, d17}, [r1]
+; DEFAULT-NEXT:    vorr q9, q8, q8
+; DEFAULT-NEXT:    vld1.32 {d19[1]}, [r0:32]
+; DEFAULT-NEXT:    vmov r0, r1, d16
+; DEFAULT-NEXT:    vmov r2, r3, d19
+; DEFAULT-NEXT:    mov pc, lr
+;
+; BASIC-LABEL: vld1laneQi32:
+; BASIC:       @ %bb.0:
+; BASIC-NEXT:    vld1.64 {d18, d19}, [r1]
+; BASIC-NEXT:    vorr q8, q9, q9
+; BASIC-NEXT:    vld1.32 {d17[1]}, [r0:32]
+; BASIC-NEXT:    vmov r0, r1, d18
+; BASIC-NEXT:    vmov r2, r3, d17
+; BASIC-NEXT:    mov pc, lr
   %tmp1 = load <4 x i32>, <4 x i32>* %B
   %tmp2 = load i32, i32* %A, align 8
   %tmp3 = insertelement <4 x i32> %tmp1, i32 %tmp2, i32 3
   ret <4 x i32> %tmp3
 }
 
 define <4 x float> @vld1laneQf(float* %A, <4 x float>* %B) nounwind {
-; CHECK-LABEL: vld1laneQf:
-; CHECK:       @ %bb.0:
-; CHECK-NEXT:    vld1.64 {d16, d17}, [r1]
-; CHECK-NEXT:    vld1.32 {d16[0]}, [r0:32]
-; CHECK-NEXT:    vmov r2, r3, d17
-; CHECK-NEXT:    vmov r0, r1, d16
-; CHECK-NEXT:    mov pc, lr
+; DEFAULT-LABEL: vld1laneQf:
+; DEFAULT:       @ %bb.0:
+; DEFAULT-NEXT:    vld1.64 {d16, d17}, [r1]
+; DEFAULT-NEXT:    vorr q9, q8, q8
+; DEFAULT-NEXT:    vmov r2, r3, d17
+; DEFAULT-NEXT:    vld1.32 {d18[0]}, [r0:32]
+; DEFAULT-NEXT:    vmov r0, r1, d18
+; DEFAULT-NEXT:    mov pc, lr
+;
+; BASIC-LABEL: vld1laneQf:
+; BASIC:       @ %bb.0:
+; BASIC-NEXT:    vld1.64 {d18, d19}, [r1]
+; BASIC-NEXT:    vorr q8, q9, q9
+; BASIC-NEXT:    vmov r2, r3, d19
+; BASIC-NEXT:    vld1.32 {d16[0]}, [r0:32]
+; BASIC-NEXT:    vmov r0, r1, d16
+; BASIC-NEXT:    mov pc, lr
   %tmp1 = load <4 x float>, <4 x float>* %B
   %tmp2 = load float, float* %A
   %tmp3 = insertelement <4 x float> %tmp1, float %tmp2, i32 0
   ret <4 x float> %tmp3
 }
 
 %struct.__neon_int8x8x2_t = type { <8 x i8>,  <8 x i8> }
 %struct.__neon_int16x4x2_t = type { <4 x i16>, <4 x i16> }

llvm/test/CodeGen/ARM/vzip.ll

 ; but which the current handling of two-result vzip can't do - thus ending up
 ; as a vtrn.
 define <8 x i16> @vzip_lower_shufflemask_undef_rev(<4 x i16>* %A, <4 x i16>* %B) {
 ; CHECK-LABEL: vzip_lower_shufflemask_undef_rev:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    vldr d16, [r1]
 ; CHECK-NEXT:    vldr d19, [r0]
 ; CHECK-NEXT:    vtrn.16 d19, d16
-; CHECK-NEXT:    vmov r0, r1, d18
+; CHECK-NEXT:    vmov r0, r1, d16
 ; CHECK-NEXT:    vmov r2, r3, d19
 ; CHECK-NEXT:    mov pc, lr
 entry:
   %tmp1 = load <4 x i16>, <4 x i16>* %A
   %tmp2 = load <4 x i16>, <4 x i16>* %B
   %0 = shufflevector <4 x i16> %tmp1, <4 x i16> %tmp2, <8 x i32> <i32 undef, i32 undef, i32 undef, i32 undef, i32 0, i32 4, i32 undef, i32 undef>
   ret <8 x i16> %0
 }

llvm/test/CodeGen/Mips/cconv/vector.ll

 ; MIPS64-NEXT:    andi $2, $2, 65535
 ; MIPS64-NEXT:    or $2, $2, $3
 ; MIPS64-NEXT:    dsll $2, $2, 32
 ; MIPS64-NEXT:    dsrl $2, $2, 32
 ; MIPS64-NEXT:    or $2, $2, $1
 ; MIPS64-NEXT:    jr $ra
 ; MIPS64-NEXT:    nop
 ;
-; MIPS32R5EB-LABEL: i8_8:
-; MIPS32R5EB:       # %bb.0:
-; MIPS32R5EB-NEXT:    addiu $sp, $sp, -48
-; MIPS32R5EB-NEXT:    .cfi_def_cfa_offset 48
-; MIPS32R5EB-NEXT:    sw $ra, 44($sp) # 4-byte Folded Spill
-; MIPS32R5EB-NEXT:    sw $fp, 40($sp) # 4-byte Folded Spill
-; MIPS32R5EB-NEXT:    .cfi_offset 31, -4
-; MIPS32R5EB-NEXT:    .cfi_offset 30, -8
-; MIPS32R5EB-NEXT:    move $fp, $sp
-; MIPS32R5EB-NEXT:    .cfi_def_cfa_register 30
-; MIPS32R5EB-NEXT:    addiu $1, $zero, -16
-; MIPS32R5EB-NEXT:    and $sp, $sp, $1
-; MIPS32R5EB-NEXT:    sw $6, 24($sp)
-; MIPS32R5EB-NEXT:    lbu $1, 25($sp)
-; MIPS32R5EB-NEXT:    lbu $2, 24($sp)
-; MIPS32R5EB-NEXT:    sw $7, 28($sp)
-; MIPS32R5EB-NEXT:    insert.h $w0[0], $2
-; MIPS32R5EB-NEXT:    insert.h $w0[1], $1
-; MIPS32R5EB-NEXT:    lbu $1, 26($sp)
-; MIPS32R5EB-NEXT:    sw $4, 32($sp)
-; MIPS32R5EB-NEXT:    insert.h $w0[2], $1
-; MIPS32R5EB-NEXT:    lbu $1, 27($sp)
-; MIPS32R5EB-NEXT:    insert.h $w0[3], $1
-; MIPS32R5EB-NEXT:    lbu $1, 28($sp)
-; MIPS32R5EB-NEXT:    sw $5, 36($sp)
-; MIPS32R5EB-NEXT:    insert.h $w0[4], $1
-; MIPS32R5EB-NEXT:    lbu $1, 33($sp)
-; MIPS32R5EB-NEXT:    lbu $2, 32($sp)
-; MIPS32R5EB-NEXT:    insert.h $w1[0], $2
-; MIPS32R5EB-NEXT:    insert.h $w1[1], $1
-; MIPS32R5EB-NEXT:    lbu $1, 29($sp)
-; MIPS32R5EB-NEXT:    lbu $2, 34($sp)
-; MIPS32R5EB-NEXT:    insert.h $w1[2], $2
-; MIPS32R5EB-NEXT:    insert.h $w0[5], $1
-; MIPS32R5EB-NEXT:    lbu $1, 35($sp)
-; MIPS32R5EB-NEXT:    lbu $2, 31($sp)
-; MIPS32R5EB-NEXT:    lbu $3, 30($sp)
-; MIPS32R5EB-NEXT:    lbu $4, 39($sp)
-; MIPS32R5EB-NEXT:    insert.h $w0[6], $3
-; MIPS32R5EB-NEXT:    insert.h $w0[7], $2
-; MIPS32R5EB-NEXT:    insert.h $w1[3], $1
-; MIPS32R5EB-NEXT:    lbu $1, 36($sp)
-; MIPS32R5EB-NEXT:    insert.h $w1[4], $1
-; MIPS32R5EB-NEXT:    lbu $1, 37($sp)
-; MIPS32R5EB-NEXT:    insert.h $w1[5], $1
-; MIPS32R5EB-NEXT:    lbu $1, 38($sp)
-; MIPS32R5EB-NEXT:    insert.h $w1[6], $1
-; MIPS32R5EB-NEXT:    insert.h $w1[7], $4
-; MIPS32R5EB-NEXT:    addv.h $w0, $w1, $w0
-; MIPS32R5EB-NEXT:    copy_s.h $1, $w0[0]
-; MIPS32R5EB-NEXT:    copy_s.h $2, $w0[1]
-; MIPS32R5EB-NEXT:    copy_s.h $3, $w0[2]
-; MIPS32R5EB-NEXT:    copy_s.h $4, $w0[3]
-; MIPS32R5EB-NEXT:    copy_s.h $5, $w0[4]
-; MIPS32R5EB-NEXT:    copy_s.h $6, $w0[5]
-; MIPS32R5EB-NEXT:    copy_s.h $7, $w0[6]
-; MIPS32R5EB-NEXT:    copy_s.h $8, $w0[7]
-; MIPS32R5EB-NEXT:    sb $8, 23($sp)
-; MIPS32R5EB-NEXT:    sb $7, 22($sp)
-; MIPS32R5EB-NEXT:    sb $6, 21($sp)
-; MIPS32R5EB-NEXT:    sb $5, 20($sp)
-; MIPS32R5EB-NEXT:    sb $4, 19($sp)
-; MIPS32R5EB-NEXT:    sb $3, 18($sp)
-; MIPS32R5EB-NEXT:    sb $2, 17($sp)
-; MIPS32R5EB-NEXT:    sb $1, 16($sp)
-; MIPS32R5EB-NEXT:    lw $1, 20($sp)
-; MIPS32R5EB-NEXT:    sw $1, 12($sp)
-; MIPS32R5EB-NEXT:    lw $1, 16($sp)
-; MIPS32R5EB-NEXT:    sw $1, 4($sp)
-; MIPS32R5EB-NEXT:    ld.w $w0, 0($sp)
-; MIPS32R5EB-NEXT:    copy_s.w $2, $w0[1]
-; MIPS32R5EB-NEXT:    copy_s.w $3, $w0[3]
-; MIPS32R5EB-NEXT:    move $sp, $fp
-; MIPS32R5EB-NEXT:    lw $fp, 40($sp) # 4-byte Folded Reload
-; MIPS32R5EB-NEXT:    lw $ra, 44($sp) # 4-byte Folded Reload
-; MIPS32R5EB-NEXT:    addiu $sp, $sp, 48
-; MIPS32R5EB-NEXT:    jr $ra
-; MIPS32R5EB-NEXT:    nop
+; MIPS32R5-LABEL: i8_8:
+; MIPS32R5:       # %bb.0:
+; MIPS32R5-NEXT:    addiu $sp, $sp, -24
+; MIPS32R5-NEXT:    .cfi_def_cfa_offset 24
+; MIPS32R5-NEXT:    sw $6, 8($sp)
+; MIPS32R5-NEXT:    lbu $1, 9($sp)
+; MIPS32R5-NEXT:    lbu $2, 8($sp)
+; MIPS32R5-NEXT:    sw $7, 12($sp)
+; MIPS32R5-NEXT:    insert.h $w0[0], $2
+; MIPS32R5-NEXT:    insert.h $w0[1], $1
+; MIPS32R5-NEXT:    lbu $1, 10($sp)
+; MIPS32R5-NEXT:    sw $4, 16($sp)
+; MIPS32R5-NEXT:    insert.h $w0[2], $1
+; MIPS32R5-NEXT:    lbu $1, 11($sp)
+; MIPS32R5-NEXT:    insert.h $w0[3], $1
+; MIPS32R5-NEXT:    lbu $1, 12($sp)
+; MIPS32R5-NEXT:    sw $5, 20($sp)
+; MIPS32R5-NEXT:    insert.h $w0[4], $1
+; MIPS32R5-NEXT:    lbu $1, 17($sp)
+; MIPS32R5-NEXT:    lbu $2, 16($sp)
+; MIPS32R5-NEXT:    insert.h $w1[0], $2
+; MIPS32R5-NEXT:    insert.h $w1[1], $1
+; MIPS32R5-NEXT:    lbu $1, 13($sp)
+; MIPS32R5-NEXT:    lbu $2, 18($sp)
+; MIPS32R5-NEXT:    insert.h $w1[2], $2
+; MIPS32R5-NEXT:    insert.h $w0[5], $1
+; MIPS32R5-NEXT:    lbu $1, 19($sp)
+; MIPS32R5-NEXT:    lbu $2, 15($sp)
+; MIPS32R5-NEXT:    lbu $3, 14($sp)
+; MIPS32R5-NEXT:    lbu $4, 23($sp)
+; MIPS32R5-NEXT:    insert.h $w0[6], $3
+; MIPS32R5-NEXT:    insert.h $w0[7], $2
+; MIPS32R5-NEXT:    insert.h $w1[3], $1
+; MIPS32R5-NEXT:    lbu $1, 20($sp)
+; MIPS32R5-NEXT:    insert.h $w1[4], $1
+; MIPS32R5-NEXT:    lbu $1, 21($sp)
+; MIPS32R5-NEXT:    insert.h $w1[5], $1
+; MIPS32R5-NEXT:    lbu $1, 22($sp)
+; MIPS32R5-NEXT:    insert.h $w1[6], $1
+; MIPS32R5-NEXT:    insert.h $w1[7], $4
+; MIPS32R5-NEXT:    addv.h $w0, $w1, $w0
+; MIPS32R5-NEXT:    copy_s.h $1, $w0[4]
+; MIPS32R5-NEXT:    copy_s.h $2, $w0[5]
+; MIPS32R5-NEXT:    copy_s.h $3, $w0[6]
+; MIPS32R5-NEXT:    copy_s.h $4, $w0[7]
+; MIPS32R5-NEXT:    copy_s.h $5, $w0[0]
+; MIPS32R5-NEXT:    copy_s.h $6, $w0[1]
+; MIPS32R5-NEXT:    copy_s.h $7, $w0[2]
+; MIPS32R5-NEXT:    copy_s.h $8, $w0[3]
+; MIPS32R5-NEXT:    sb $8, 3($sp)
+; MIPS32R5-NEXT:    sb $7, 2($sp)
+; MIPS32R5-NEXT:    sb $6, 1($sp)
+; MIPS32R5-NEXT:    sb $5, 0($sp)
+; MIPS32R5-NEXT:    sb $4, 7($sp)
+; MIPS32R5-NEXT:    sb $3, 6($sp)
+; MIPS32R5-NEXT:    sb $2, 5($sp)
+; MIPS32R5-NEXT:    sb $1, 4($sp)
+; MIPS32R5-NEXT:    lw $2, 0($sp)
+; MIPS32R5-NEXT:    lw $3, 4($sp)
+; MIPS32R5-NEXT:    addiu $sp, $sp, 24
+; MIPS32R5-NEXT:    jr $ra
+; MIPS32R5-NEXT:    nop
 ;
 ; MIPS64R5-LABEL: i8_8:
 ; MIPS64R5:       # %bb.0:
 ; MIPS64R5-NEXT:    daddiu $sp, $sp, -32
 ; MIPS64R5-NEXT:    .cfi_def_cfa_offset 32
 ; MIPS64R5-NEXT:    sd $5, 16($sp)
 ; MIPS64R5-NEXT:    lbu $1, 17($sp)
 ; MIPS64R5-NEXT:    lbu $2, 16($sp)

 ; MIPS64R5-NEXT:    sb $4, 11($sp)
 ; MIPS64R5-NEXT:    sb $3, 10($sp)
 ; MIPS64R5-NEXT:    sb $2, 9($sp)
 ; MIPS64R5-NEXT:    sb $1, 8($sp)
 ; MIPS64R5-NEXT:    ld $2, 8($sp)
 ; MIPS64R5-NEXT:    daddiu $sp, $sp, 32
 ; MIPS64R5-NEXT:    jr $ra
 ; MIPS64R5-NEXT:    nop
-;
-; MIPS32R5EL-LABEL: i8_8:
-; MIPS32R5EL:       # %bb.0:
-; MIPS32R5EL-NEXT:    addiu $sp, $sp, -48
-; MIPS32R5EL-NEXT:    .cfi_def_cfa_offset 48
-; MIPS32R5EL-NEXT:    sw $ra, 44($sp) # 4-byte Folded Spill
-; MIPS32R5EL-NEXT:    sw $fp, 40($sp) # 4-byte Folded Spill
-; MIPS32R5EL-NEXT:    .cfi_offset 31, -4
-; MIPS32R5EL-NEXT:    .cfi_offset 30, -8
-; MIPS32R5EL-NEXT:    move $fp, $sp
-; MIPS32R5EL-NEXT:    .cfi_def_cfa_register 30
-; MIPS32R5EL-NEXT:    addiu $1, $zero, -16
-; MIPS32R5EL-NEXT:    and $sp, $sp, $1
-; MIPS32R5EL-NEXT:    sw $6, 24($sp)
-; MIPS32R5EL-NEXT:    lbu $1, 25($sp)
-; MIPS32R5EL-NEXT:    lbu $2, 24($sp)
-; MIPS32R5EL-NEXT:    sw $7, 28($sp)
-; MIPS32R5EL-NEXT:    insert.h $w0[0], $2
-; MIPS32R5EL-NEXT:    insert.h $w0[1], $1
-; MIPS32R5EL-NEXT:    lbu $1, 26($sp)
-; MIPS32R5EL-NEXT:    sw $4, 32($sp)
-; MIPS32R5EL-NEXT:    insert.h $w0[2], $1
-; MIPS32R5EL-NEXT:    lbu $1, 27($sp)
-; MIPS32R5EL-NEXT:    insert.h $w0[3], $1
-; MIPS32R5EL-NEXT:    lbu $1, 28($sp)
-; MIPS32R5EL-NEXT:    sw $5, 36($sp)
-; MIPS32R5EL-NEXT:    insert.h $w0[4], $1
-; MIPS32R5EL-NEXT:    lbu $1, 33($sp)
-; MIPS32R5EL-NEXT:    lbu $2, 32($sp)
-; MIPS32R5EL-NEXT:    insert.h $w1[0], $2
-; MIPS32R5EL-NEXT:    insert.h $w1[1], $1
-; MIPS32R5EL-NEXT:    lbu $1, 29($sp)
-; MIPS32R5EL-NEXT:    lbu $2, 34($sp)
-; MIPS32R5EL-NEXT:    insert.h $w1[2], $2
-; MIPS32R5EL-NEXT:    insert.h $w0[5], $1
-; MIPS32R5EL-NEXT:    lbu $1, 35($sp)
-; MIPS32R5EL-NEXT:    lbu $2, 31($sp)
-; MIPS32R5EL-NEXT:    lbu $3, 30($sp)
-; MIPS32R5EL-NEXT:    lbu $4, 39($sp)
-; MIPS32R5EL-NEXT:    insert.h $w0[6], $3
-; MIPS32R5EL-NEXT:    insert.h $w0[7], $2
-; MIPS32R5EL-NEXT:    insert.h $w1[3], $1
-; MIPS32R5EL-NEXT:    lbu $1, 36($sp)
-; MIPS32R5EL-NEXT:    insert.h $w1[4], $1
-; MIPS32R5EL-NEXT:    lbu $1, 37($sp)
-; MIPS32R5EL-NEXT:    insert.h $w1[5], $1
-; MIPS32R5EL-NEXT:    lbu $1, 38($sp)
-; MIPS32R5EL-NEXT:    insert.h $w1[6], $1
-; MIPS32R5EL-NEXT:    insert.h $w1[7], $4
-; MIPS32R5EL-NEXT:    addv.h $w0, $w1, $w0
-; MIPS32R5EL-NEXT:    copy_s.h $1, $w0[0]
-; MIPS32R5EL-NEXT:    copy_s.h $2, $w0[1]
-; MIPS32R5EL-NEXT:    copy_s.h $3, $w0[2]
-; MIPS32R5EL-NEXT:    copy_s.h $4, $w0[3]
-; MIPS32R5EL-NEXT:    copy_s.h $5, $w0[4]
-; MIPS32R5EL-NEXT:    copy_s.h $6, $w0[5]
-; MIPS32R5EL-NEXT:    copy_s.h $7, $w0[6]
-; MIPS32R5EL-NEXT:    copy_s.h $8, $w0[7]
-; MIPS32R5EL-NEXT:    sb $8, 23($sp)
-; MIPS32R5EL-NEXT:    sb $7, 22($sp)
-; MIPS32R5EL-NEXT:    sb $6, 21($sp)
-; MIPS32R5EL-NEXT:    sb $5, 20($sp)
-; MIPS32R5EL-NEXT:    sb $4, 19($sp)
-; MIPS32R5EL-NEXT:    sb $3, 18($sp)
-; MIPS32R5EL-NEXT:    sb $2, 17($sp)
-; MIPS32R5EL-NEXT:    sb $1, 16($sp)
-; MIPS32R5EL-NEXT:    lw $1, 20($sp)
-; MIPS32R5EL-NEXT:    sw $1, 8($sp)
-; MIPS32R5EL-NEXT:    lw $1, 16($sp)
-; MIPS32R5EL-NEXT:    sw $1, 0($sp)
-; MIPS32R5EL-NEXT:    ld.w $w0, 0($sp)
-; MIPS32R5EL-NEXT:    copy_s.w $2, $w0[0]
-; MIPS32R5EL-NEXT:    copy_s.w $3, $w0[2]
-; MIPS32R5EL-NEXT:    move $sp, $fp
-; MIPS32R5EL-NEXT:    lw $fp, 40($sp) # 4-byte Folded Reload
-; MIPS32R5EL-NEXT:    lw $ra, 44($sp) # 4-byte Folded Reload
-; MIPS32R5EL-NEXT:    addiu $sp, $sp, 48
-; MIPS32R5EL-NEXT:    jr $ra
-; MIPS32R5EL-NEXT:    nop
   %1 = add <8 x i8> %a, %b
   ret <8 x i8> %1
 }
 
 define <16 x i8> @i8_16(<16 x i8> %a, <16 x i8> %b) {
 ; MIPS32-LABEL: i8_16:
 ; MIPS32:       # %bb.0:
 ; MIPS32-NEXT:    lw $1, 24($sp)

 ; MIPS64-NEXT:    sll $2, $2, 16
 ; MIPS64-NEXT:    or $2, $4, $2
 ; MIPS64-NEXT:    dsll $2, $2, 32
 ; MIPS64-NEXT:    dsrl $2, $2, 32
 ; MIPS64-NEXT:    or $2, $2, $1
 ; MIPS64-NEXT:    jr $ra
 ; MIPS64-NEXT:    nop
 ;
-; MIPS32R5EB-LABEL: i16_4:
-; MIPS32R5EB:       # %bb.0:
-; MIPS32R5EB-NEXT:    addiu $sp, $sp, -48
-; MIPS32R5EB-NEXT:    .cfi_def_cfa_offset 48
-; MIPS32R5EB-NEXT:    sw $ra, 44($sp) # 4-byte Folded Spill
-; MIPS32R5EB-NEXT:    sw $fp, 40($sp) # 4-byte Folded Spill
-; MIPS32R5EB-NEXT:    .cfi_offset 31, -4
-; MIPS32R5EB-NEXT:    .cfi_offset 30, -8
-; MIPS32R5EB-NEXT:    move $fp, $sp
-; MIPS32R5EB-NEXT:    .cfi_def_cfa_register 30
-; MIPS32R5EB-NEXT:    addiu $1, $zero, -16
-; MIPS32R5EB-NEXT:    and $sp, $sp, $1
-; MIPS32R5EB-NEXT:    sw $6, 24($sp)
-; MIPS32R5EB-NEXT:    sw $7, 28($sp)
-; MIPS32R5EB-NEXT:    lhu $1, 26($sp)
-; MIPS32R5EB-NEXT:    lhu $2, 24($sp)
-; MIPS32R5EB-NEXT:    sw $4, 32($sp)
-; MIPS32R5EB-NEXT:    insert.w $w0[0], $2
-; MIPS32R5EB-NEXT:    insert.w $w0[1], $1
-; MIPS32R5EB-NEXT:    lhu $1, 28($sp)
-; MIPS32R5EB-NEXT:    sw $5, 36($sp)
-; MIPS32R5EB-NEXT:    insert.w $w0[2], $1
-; MIPS32R5EB-NEXT:    lhu $1, 30($sp)
-; MIPS32R5EB-NEXT:    insert.w $w0[3], $1
-; MIPS32R5EB-NEXT:    lhu $1, 34($sp)
-; MIPS32R5EB-NEXT:    lhu $2, 32($sp)
-; MIPS32R5EB-NEXT:    insert.w $w1[0], $2
-; MIPS32R5EB-NEXT:    insert.w $w1[1], $1
-; MIPS32R5EB-NEXT:    lhu $1, 36($sp)
-; MIPS32R5EB-NEXT:    insert.w $w1[2], $1
-; MIPS32R5EB-NEXT:    lhu $1, 38($sp)
-; MIPS32R5EB-NEXT:    insert.w $w1[3], $1
-; MIPS32R5EB-NEXT:    addv.w $w0, $w1, $w0
-; MIPS32R5EB-NEXT:    copy_s.w $1, $w0[0]
-; MIPS32R5EB-NEXT:    copy_s.w $2, $w0[1]
-; MIPS32R5EB-NEXT:    copy_s.w $3, $w0[2]
-; MIPS32R5EB-NEXT:    copy_s.w $4, $w0[3]
-; MIPS32R5EB-NEXT:    sh $4, 22($sp)
-; MIPS32R5EB-NEXT:    sh $3, 20($sp)
-; MIPS32R5EB-NEXT:    sh $2, 18($sp)
-; MIPS32R5EB-NEXT:    sh $1, 16($sp)
-; MIPS32R5EB-NEXT:    lw $1, 20($sp)
-; MIPS32R5EB-NEXT:    sw $1, 12($sp)
-; MIPS32R5EB-NEXT:    lw $1, 16($sp)
-; MIPS32R5EB-NEXT:    sw $1, 4($sp)
-; MIPS32R5EB-NEXT:    ld.w $w0, 0($sp)
-; MIPS32R5EB-NEXT:    copy_s.w $2, $w0[1]
-; MIPS32R5EB-NEXT:    copy_s.w $3, $w0[3]
-; MIPS32R5EB-NEXT:    move $sp, $fp
-; MIPS32R5EB-NEXT:    lw $fp, 40($sp) # 4-byte Folded Reload
-; MIPS32R5EB-NEXT:    lw $ra, 44($sp) # 4-byte Folded Reload
-; MIPS32R5EB-NEXT:    addiu $sp, $sp, 48
-; MIPS32R5EB-NEXT:    jr $ra
-; MIPS32R5EB-NEXT:    nop
+; MIPS32R5-LABEL: i16_4:
+; MIPS32R5:       # %bb.0:
+; MIPS32R5-NEXT:    addiu $sp, $sp, -24
+; MIPS32R5-NEXT:    .cfi_def_cfa_offset 24
+; MIPS32R5-NEXT:    sw $6, 8($sp)
+; MIPS32R5-NEXT:    sw $7, 12($sp)
+; MIPS32R5-NEXT:    lhu $1, 10($sp)
+; MIPS32R5-NEXT:    lhu $2, 8($sp)
+; MIPS32R5-NEXT:    sw $4, 16($sp)
+; MIPS32R5-NEXT:    insert.w $w0[0], $2
+; MIPS32R5-NEXT:    insert.w $w0[1], $1
+; MIPS32R5-NEXT:    lhu $1, 12($sp)
+; MIPS32R5-NEXT:    sw $5, 20($sp)
+; MIPS32R5-NEXT:    insert.w $w0[2], $1
+; MIPS32R5-NEXT:    lhu $1, 14($sp)
+; MIPS32R5-NEXT:    insert.w $w0[3], $1
+; MIPS32R5-NEXT:    lhu $1, 18($sp)
+; MIPS32R5-NEXT:    lhu $2, 16($sp)
+; MIPS32R5-NEXT:    insert.w $w1[0], $2
+; MIPS32R5-NEXT:    insert.w $w1[1], $1
+; MIPS32R5-NEXT:    lhu $1, 20($sp)
+; MIPS32R5-NEXT:    insert.w $w1[2], $1
+; MIPS32R5-NEXT:    lhu $1, 22($sp)
+; MIPS32R5-NEXT:    insert.w $w1[3], $1
+; MIPS32R5-NEXT:    addv.w $w0, $w1, $w0
+; MIPS32R5-NEXT:    copy_s.w $1, $w0[2]
+; MIPS32R5-NEXT:    copy_s.w $2, $w0[3]
+; MIPS32R5-NEXT:    copy_s.w $3, $w0[0]
+; MIPS32R5-NEXT:    copy_s.w $4, $w0[1]
+; MIPS32R5-NEXT:    sh $4, 2($sp)
+; MIPS32R5-NEXT:    sh $3, 0($sp)
+; MIPS32R5-NEXT:    sh $2, 6($sp)
+; MIPS32R5-NEXT:    sh $1, 4($sp)
+; MIPS32R5-NEXT:    lw $2, 0($sp)
+; MIPS32R5-NEXT:    lw $3, 4($sp)
+; MIPS32R5-NEXT:    addiu $sp, $sp, 24
+; MIPS32R5-NEXT:    jr $ra
+; MIPS32R5-NEXT:    nop
 ;
 ; MIPS64R5-LABEL: i16_4:
 ; MIPS64R5:       # %bb.0:
 ; MIPS64R5-NEXT:    daddiu $sp, $sp, -32
 ; MIPS64R5-NEXT:    .cfi_def_cfa_offset 32
 ; MIPS64R5-NEXT:    sd $5, 16($sp)
 ; MIPS64R5-NEXT:    sd $4, 24($sp)
 ; MIPS64R5-NEXT:    lhu $1, 18($sp)

 ; MIPS64R5-NEXT:    sh $4, 14($sp)
 ; MIPS64R5-NEXT:    sh $3, 12($sp)
 ; MIPS64R5-NEXT:    sh $2, 10($sp)
 ; MIPS64R5-NEXT:    sh $1, 8($sp)
 ; MIPS64R5-NEXT:    ld $2, 8($sp)
 ; MIPS64R5-NEXT:    daddiu $sp, $sp, 32
 ; MIPS64R5-NEXT:    jr $ra
 ; MIPS64R5-NEXT:    nop
-;
-; MIPS32R5EL-LABEL: i16_4:
-; MIPS32R5EL:       # %bb.0:
-; MIPS32R5EL-NEXT:    addiu $sp, $sp, -48
-; MIPS32R5EL-NEXT:    .cfi_def_cfa_offset 48
-; MIPS32R5EL-NEXT:    sw $ra, 44($sp) # 4-byte Folded Spill
-; MIPS32R5EL-NEXT:    sw $fp, 40($sp) # 4-byte Folded Spill
-; MIPS32R5EL-NEXT:    .cfi_offset 31, -4
-; MIPS32R5EL-NEXT:    .cfi_offset 30, -8
-; MIPS32R5EL-NEXT:    move $fp, $sp
-; MIPS32R5EL-NEXT:    .cfi_def_cfa_register 30
-; MIPS32R5EL-NEXT:    addiu $1, $zero, -16
-; MIPS32R5EL-NEXT:    and $sp, $sp, $1
-; MIPS32R5EL-NEXT:    sw $6, 24($sp)
-; MIPS32R5EL-NEXT:    sw $7, 28($sp)
-; MIPS32R5EL-NEXT:    lhu $1, 26($sp)
-; MIPS32R5EL-NEXT:    lhu $2, 24($sp)
-; MIPS32R5EL-NEXT:    sw $4, 32($sp)
-; MIPS32R5EL-NEXT:    insert.w $w0[0], $2
-; MIPS32R5EL-NEXT:    insert.w $w0[1], $1
-; MIPS32R5EL-NEXT:    lhu $1, 28($sp)
-; MIPS32R5EL-NEXT:    sw $5, 36($sp)
-; MIPS32R5EL-NEXT:    insert.w $w0[2], $1
-; MIPS32R5EL-NEXT:    lhu $1, 30($sp)
-; MIPS32R5EL-NEXT:    insert.w $w0[3], $1
-; MIPS32R5EL-NEXT:    lhu $1, 34($sp)
-; MIPS32R5EL-NEXT:    lhu $2, 32($sp)
-; MIPS32R5EL-NEXT:    insert.w $w1[0], $2
-; MIPS32R5EL-NEXT:    insert.w $w1[1], $1
-; MIPS32R5EL-NEXT:    lhu $1, 36($sp)
-; MIPS32R5EL-NEXT:    insert.w $w1[2], $1
-; MIPS32R5EL-NEXT:    lhu $1, 38($sp)
-; MIPS32R5EL-NEXT:    insert.w $w1[3], $1
-; MIPS32R5EL-NEXT:    addv.w $w0, $w1, $w0
-; MIPS32R5EL-NEXT:    copy_s.w $1, $w0[0]
-; MIPS32R5EL-NEXT:    copy_s.w $2, $w0[1]
-; MIPS32R5EL-NEXT:    copy_s.w $3, $w0[2]
-; MIPS32R5EL-NEXT:    copy_s.w $4, $w0[3]
-; MIPS32R5EL-NEXT:    sh $4, 22($sp)
-; MIPS32R5EL-NEXT:    sh $3, 20($sp)
-; MIPS32R5EL-NEXT:    sh $2, 18($sp)
-; MIPS32R5EL-NEXT:    sh $1, 16($sp)
-; MIPS32R5EL-NEXT:    lw $1, 20($sp)
-; MIPS32R5EL-NEXT:    sw $1, 8($sp)
-; MIPS32R5EL-NEXT:    lw $1, 16($sp)
-; MIPS32R5EL-NEXT:    sw $1, 0($sp)
-; MIPS32R5EL-NEXT:    ld.w $w0, 0($sp)
-; MIPS32R5EL-NEXT:    copy_s.w $2, $w0[0]
-; MIPS32R5EL-NEXT:    copy_s.w $3, $w0[2]
-; MIPS32R5EL-NEXT:    move $sp, $fp
-; MIPS32R5EL-NEXT:    lw $fp, 40($sp) # 4-byte Folded Reload
-; MIPS32R5EL-NEXT:    lw $ra, 44($sp) # 4-byte Folded Reload
-; MIPS32R5EL-NEXT:    addiu $sp, $sp, 48
-; MIPS32R5EL-NEXT:    jr $ra
-; MIPS32R5EL-NEXT:    nop
   %1 = add <4 x i16> %a, %b
   ret <4 x i16> %1
 }
 
 define <8 x i16> @i16_8(<8 x i16> %a, <8 x i16> %b) {
 ; MIPS32-LABEL: i16_8:
 ; MIPS32:       # %bb.0:
 ; MIPS32-NEXT:    lw $1, 24($sp)

 ; MIPS64-NEXT:    lui $1, %hi(%neg(%gp_rel(ret_8_i8)))
 ; MIPS64-NEXT:    daddu $1, $1, $25
 ; MIPS64-NEXT:    daddiu $1, $1, %lo(%neg(%gp_rel(ret_8_i8)))
 ; MIPS64-NEXT:    ld $1, %got_disp(gv8i8)($1)
 ; MIPS64-NEXT:    ld $2, 0($1)
 ; MIPS64-NEXT:    jr $ra
 ; MIPS64-NEXT:    nop
 ;
-; MIPS32R5EB-LABEL: ret_8_i8:
-; MIPS32R5EB:       # %bb.0:
-; MIPS32R5EB-NEXT:    addiu $sp, $sp, -32
-; MIPS32R5EB-NEXT:    .cfi_def_cfa_offset 32
-; MIPS32R5EB-NEXT:    sw $ra, 28($sp) # 4-byte Folded Spill
-; MIPS32R5EB-NEXT:    sw $fp, 24($sp) # 4-byte Folded Spill
-; MIPS32R5EB-NEXT:    .cfi_offset 31, -4
-; MIPS32R5EB-NEXT:    .cfi_offset 30, -8
-; MIPS32R5EB-NEXT:    move $fp, $sp
-; MIPS32R5EB-NEXT:    .cfi_def_cfa_register 30
-; MIPS32R5EB-NEXT:    addiu $1, $zero, -16
-; MIPS32R5EB-NEXT:    and $sp, $sp, $1
-; MIPS32R5EB-NEXT:    lui $1, %hi(gv8i8)
-; MIPS32R5EB-NEXT:    lw $2, %lo(gv8i8)($1)
-; MIPS32R5EB-NEXT:    sw $2, 4($sp)
-; MIPS32R5EB-NEXT:    addiu $1, $1, %lo(gv8i8)
-; MIPS32R5EB-NEXT:    lw $1, 4($1)
-; MIPS32R5EB-NEXT:    sw $1, 12($sp)
-; MIPS32R5EB-NEXT:    ld.w $w0, 0($sp)
-; MIPS32R5EB-NEXT:    copy_s.w $2, $w0[1]
-; MIPS32R5EB-NEXT:    copy_s.w $3, $w0[3]
-; MIPS32R5EB-NEXT:    move $sp, $fp
-; MIPS32R5EB-NEXT:    lw $fp, 24($sp) # 4-byte Folded Reload
-; MIPS32R5EB-NEXT:    lw $ra, 28($sp) # 4-byte Folded Reload
-; MIPS32R5EB-NEXT:    addiu $sp, $sp, 32
-; MIPS32R5EB-NEXT:    jr $ra
-; MIPS32R5EB-NEXT:    nop
+; MIPS32R5-LABEL: ret_8_i8:
+; MIPS32R5:       # %bb.0:
+; MIPS32R5-NEXT:    lui $1, %hi(gv8i8)
+; MIPS32R5-NEXT:    lw $2, %lo(gv8i8)($1)
+; MIPS32R5-NEXT:    addiu $1, $1, %lo(gv8i8)
+; MIPS32R5-NEXT:    lw $3, 4($1)
+; MIPS32R5-NEXT:    jr $ra
+; MIPS32R5-NEXT:    nop
 ;
 ; MIPS64R5-LABEL: ret_8_i8:
 ; MIPS64R5:       # %bb.0:
 ; MIPS64R5-NEXT:    lui $1, %hi(%neg(%gp_rel(ret_8_i8)))
 ; MIPS64R5-NEXT:    daddu $1, $1, $25
 ; MIPS64R5-NEXT:    daddiu $1, $1, %lo(%neg(%gp_rel(ret_8_i8)))
 ; MIPS64R5-NEXT:    ld $1, %got_disp(gv8i8)($1)
 ; MIPS64R5-NEXT:    ld $2, 0($1)
 ; MIPS64R5-NEXT:    jr $ra
 ; MIPS64R5-NEXT:    nop
-;
-; MIPS32R5EL-LABEL: ret_8_i8:
-; MIPS32R5EL:       # %bb.0:
-; MIPS32R5EL-NEXT:    addiu $sp, $sp, -32
-; MIPS32R5EL-NEXT:    .cfi_def_cfa_offset 32
-; MIPS32R5EL-NEXT:    sw $ra, 28($sp) # 4-byte Folded Spill
-; MIPS32R5EL-NEXT:    sw $fp, 24($sp) # 4-byte Folded Spill
-; MIPS32R5EL-NEXT:    .cfi_offset 31, -4
-; MIPS32R5EL-NEXT:    .cfi_offset 30, -8
-; MIPS32R5EL-NEXT:    move $fp, $sp
-; MIPS32R5EL-NEXT:    .cfi_def_cfa_register 30
-; MIPS32R5EL-NEXT:    addiu $1, $zero, -16
-; MIPS32R5EL-NEXT:    and $sp, $sp, $1
-; MIPS32R5EL-NEXT:    lui $1, %hi(gv8i8)
-; MIPS32R5EL-NEXT:    lw $2, %lo(gv8i8)($1)
-; MIPS32R5EL-NEXT:    sw $2, 0($sp)
-; MIPS32R5EL-NEXT:    addiu $1, $1, %lo(gv8i8)
-; MIPS32R5EL-NEXT:    lw $1, 4($1)
-; MIPS32R5EL-NEXT:    sw $1, 8($sp)
-; MIPS32R5EL-NEXT:    ld.w $w0, 0($sp)
-; MIPS32R5EL-NEXT:    copy_s.w $2, $w0[0]
-; MIPS32R5EL-NEXT:    copy_s.w $3, $w0[2]
-; MIPS32R5EL-NEXT:    move $sp, $fp
-; MIPS32R5EL-NEXT:    lw $fp, 24($sp) # 4-byte Folded Reload
-; MIPS32R5EL-NEXT:    lw $ra, 28($sp) # 4-byte Folded Reload
-; MIPS32R5EL-NEXT:    addiu $sp, $sp, 32
-; MIPS32R5EL-NEXT:    jr $ra
-; MIPS32R5EL-NEXT:    nop
   %1 = load <8 x i8>, <8 x i8> * @gv8i8
   ret <8 x i8> %1
 }
 
 define <16 x i8> @ret_16_i8() {
 ; MIPS32-LABEL: ret_16_i8:
 ; MIPS32:       # %bb.0:
 ; MIPS32-NEXT:    lui $1, %hi(gv16i8)

 ; MIPS64-NEXT:    lui $1, %hi(%neg(%gp_rel(ret_4_i16)))
 ; MIPS64-NEXT:    daddu $1, $1, $25
 ; MIPS64-NEXT:    daddiu $1, $1, %lo(%neg(%gp_rel(ret_4_i16)))
 ; MIPS64-NEXT:    ld $1, %got_disp(gv4i16)($1)
 ; MIPS64-NEXT:    ld $2, 0($1)
 ; MIPS64-NEXT:    jr $ra
 ; MIPS64-NEXT:    nop
 ;
-; MIPS32R5EB-LABEL: ret_4_i16:
-; MIPS32R5EB:       # %bb.0:
-; MIPS32R5EB-NEXT:    addiu $sp, $sp, -32
-; MIPS32R5EB-NEXT:    .cfi_def_cfa_offset 32
-; MIPS32R5EB-NEXT:    sw $ra, 28($sp) # 4-byte Folded Spill
-; MIPS32R5EB-NEXT:    sw $fp, 24($sp) # 4-byte Folded Spill
-; MIPS32R5EB-NEXT:    .cfi_offset 31, -4
-; MIPS32R5EB-NEXT:    .cfi_offset 30, -8
-; MIPS32R5EB-NEXT:    move $fp, $sp
-; MIPS32R5EB-NEXT:    .cfi_def_cfa_register 30
-; MIPS32R5EB-NEXT:    addiu $1, $zero, -16
-; MIPS32R5EB-NEXT:    and $sp, $sp, $1
-; MIPS32R5EB-NEXT:    lui $1, %hi(gv4i16)
-; MIPS32R5EB-NEXT:    lw $2, %lo(gv4i16)($1)
-; MIPS32R5EB-NEXT:    sw $2, 4($sp)
-; MIPS32R5EB-NEXT:    addiu $1, $1, %lo(gv4i16)
-; MIPS32R5EB-NEXT:    lw $1, 4($1)
-; MIPS32R5EB-NEXT:    sw $1, 12($sp)
-; MIPS32R5EB-NEXT:    ld.w $w0, 0($sp)
-; MIPS32R5EB-NEXT:    copy_s.w $2, $w0[1]
-; MIPS32R5EB-NEXT:    copy_s.w $3, $w0[3]
-; MIPS32R5EB-NEXT:    move $sp, $fp
-; MIPS32R5EB-NEXT:    lw $fp, 24($sp) # 4-byte Folded Reload
-; MIPS32R5EB-NEXT:    lw $ra, 28($sp) # 4-byte Folded Reload
-; MIPS32R5EB-NEXT:    addiu $sp, $sp, 32
-; MIPS32R5EB-NEXT:    jr $ra
-; MIPS32R5EB-NEXT:    nop
+; MIPS32R5-LABEL: ret_4_i16:
+; MIPS32R5:       # %bb.0:
+; MIPS32R5-NEXT:    lui $1, %hi(gv4i16)
+; MIPS32R5-NEXT:    lw $2, %lo(gv4i16)($1)
+; MIPS32R5-NEXT:    addiu $1, $1, %lo(gv4i16)
+; MIPS32R5-NEXT:    lw $3, 4($1)
+; MIPS32R5-NEXT:    jr $ra
+; MIPS32R5-NEXT:    nop
 ;
 ; MIPS64R5-LABEL: ret_4_i16:
 ; MIPS64R5:       # %bb.0:
 ; MIPS64R5-NEXT:    lui $1, %hi(%neg(%gp_rel(ret_4_i16)))
 ; MIPS64R5-NEXT:    daddu $1, $1, $25
 ; MIPS64R5-NEXT:    daddiu $1, $1, %lo(%neg(%gp_rel(ret_4_i16)))
 ; MIPS64R5-NEXT:    ld $1, %got_disp(gv4i16)($1)
 ; MIPS64R5-NEXT:    ld $2, 0($1)
 ; MIPS64R5-NEXT:    jr $ra
 ; MIPS64R5-NEXT:    nop
-;
-; MIPS32R5EL-LABEL: ret_4_i16:
-; MIPS32R5EL:       # %bb.0:
-; MIPS32R5EL-NEXT:    addiu $sp, $sp, -32
-; MIPS32R5EL-NEXT:    .cfi_def_cfa_offset 32
-; MIPS32R5EL-NEXT:    sw $ra, 28($sp) # 4-byte Folded Spill
-; MIPS32R5EL-NEXT:    sw $fp, 24($sp) # 4-byte Folded Spill
-; MIPS32R5EL-NEXT:    .cfi_offset 31, -4
-; MIPS32R5EL-NEXT:    .cfi_offset 30, -8
-; MIPS32R5EL-NEXT:    move $fp, $sp
-; MIPS32R5EL-NEXT:    .cfi_def_cfa_register 30
-; MIPS32R5EL-NEXT:    addiu $1, $zero, -16
-; MIPS32R5EL-NEXT:    and $sp, $sp, $1
-; MIPS32R5EL-NEXT:    lui $1, %hi(gv4i16)
-; MIPS32R5EL-NEXT:    lw $2, %lo(gv4i16)($1)
-; MIPS32R5EL-NEXT:    sw $2, 0($sp)
-; MIPS32R5EL-NEXT:    addiu $1, $1, %lo(gv4i16)
-; MIPS32R5EL-NEXT:    lw $1, 4($1)
-; MIPS32R5EL-NEXT:    sw $1, 8($sp)
-; MIPS32R5EL-NEXT:    ld.w $w0, 0($sp)
-; MIPS32R5EL-NEXT:    copy_s.w $2, $w0[0]
-; MIPS32R5EL-NEXT:    copy_s.w $3, $w0[2]
-; MIPS32R5EL-NEXT:    move $sp, $fp
-; MIPS32R5EL-NEXT:    lw $fp, 24($sp) # 4-byte Folded Reload
-; MIPS32R5EL-NEXT:    lw $ra, 28($sp) # 4-byte Folded Reload
-; MIPS32R5EL-NEXT:    addiu $sp, $sp, 32
-; MIPS32R5EL-NEXT:    jr $ra
-; MIPS32R5EL-NEXT:    nop
   %1 = load <4 x i16>, <4 x i16> * @gv4i16
   ret <4 x i16> %1
 }
 
 define <8 x i16> @ret_8_i16() {
 ; MIPS32-LABEL: ret_8_i16:
 ; MIPS32:       # %bb.0:
 ; MIPS32-NEXT:    lui $1, %hi(gv8i16)

 ; MIPS64-NEXT:    lui $1, %hi(%neg(%gp_rel(ret_2_i32)))
 ; MIPS64-NEXT:    daddu $1, $1, $25
 ; MIPS64-NEXT:    daddiu $1, $1, %lo(%neg(%gp_rel(ret_2_i32)))
 ; MIPS64-NEXT:    ld $1, %got_disp(gv2i32)($1)
 ; MIPS64-NEXT:    ld $2, 0($1)
 ; MIPS64-NEXT:    jr $ra
 ; MIPS64-NEXT:    nop
 ;
-; MIPS32R5EB-LABEL: ret_2_i32:
-; MIPS32R5EB:       # %bb.0:
-; MIPS32R5EB-NEXT:    addiu $sp, $sp, -32
-; MIPS32R5EB-NEXT:    .cfi_def_cfa_offset 32
-; MIPS32R5EB-NEXT:    sw $ra, 28($sp) # 4-byte Folded Spill
-; MIPS32R5EB-NEXT:    sw $fp, 24($sp) # 4-byte Folded Spill
-; MIPS32R5EB-NEXT:    .cfi_offset 31, -4
-; MIPS32R5EB-NEXT:    .cfi_offset 30, -8
-; MIPS32R5EB-NEXT:    move $fp, $sp
-; MIPS32R5EB-NEXT:    .cfi_def_cfa_register 30
-; MIPS32R5EB-NEXT:    addiu $1, $zero, -16
-; MIPS32R5EB-NEXT:    and $sp, $sp, $1
-; MIPS32R5EB-NEXT:    lui $1, %hi(gv2i32)
-; MIPS32R5EB-NEXT:    lw $2, %lo(gv2i32)($1)
-; MIPS32R5EB-NEXT:    sw $2, 4($sp)
-; MIPS32R5EB-NEXT:    addiu $1, $1, %lo(gv2i32)
-; MIPS32R5EB-NEXT:    lw $1, 4($1)
-; MIPS32R5EB-NEXT:    sw $1, 12($sp)
-; MIPS32R5EB-NEXT:    ld.w $w0, 0($sp)
-; MIPS32R5EB-NEXT:    copy_s.w $2, $w0[1]
-; MIPS32R5EB-NEXT:    copy_s.w $3, $w0[3]
-; MIPS32R5EB-NEXT:    move $sp, $fp
-; MIPS32R5EB-NEXT:    lw $fp, 24($sp) # 4-byte Folded Reload
-; MIPS32R5EB-NEXT:    lw $ra, 28($sp) # 4-byte Folded Reload
-; MIPS32R5EB-NEXT:    addiu $sp, $sp, 32
-; MIPS32R5EB-NEXT:    jr $ra
-; MIPS32R5EB-NEXT:    nop
+; MIPS32R5-LABEL: ret_2_i32:
+; MIPS32R5:       # %bb.0:
+; MIPS32R5-NEXT:    lui $1, %hi(gv2i32)
+; MIPS32R5-NEXT:    lw $2, %lo(gv2i32)($1)
+; MIPS32R5-NEXT:    addiu $1, $1, %lo(gv2i32)
+; MIPS32R5-NEXT:    lw $3, 4($1)
+; MIPS32R5-NEXT:    jr $ra
+; MIPS32R5-NEXT:    nop
 ;
 ; MIPS64R5-LABEL: ret_2_i32:
 ; MIPS64R5:       # %bb.0:
 ; MIPS64R5-NEXT:    lui $1, %hi(%neg(%gp_rel(ret_2_i32)))
 ; MIPS64R5-NEXT:    daddu $1, $1, $25
 ; MIPS64R5-NEXT:    daddiu $1, $1, %lo(%neg(%gp_rel(ret_2_i32)))
 ; MIPS64R5-NEXT:    ld $1, %got_disp(gv2i32)($1)
 ; MIPS64R5-NEXT:    ld $2, 0($1)
 ; MIPS64R5-NEXT:    jr $ra
 ; MIPS64R5-NEXT:    nop
-;
-; MIPS32R5EL-LABEL: ret_2_i32:
-; MIPS32R5EL:       # %bb.0:
-; MIPS32R5EL-NEXT:    addiu $sp, $sp, -32
-; MIPS32R5EL-NEXT:    .cfi_def_cfa_offset 32
-; MIPS32R5EL-NEXT:    sw $ra, 28($sp) # 4-byte Folded Spill
-; MIPS32R5EL-NEXT:    sw $fp, 24($sp) # 4-byte Folded Spill
-; MIPS32R5EL-NEXT:    .cfi_offset 31, -4
-; MIPS32R5EL-NEXT:    .cfi_offset 30, -8
-; MIPS32R5EL-NEXT:    move $fp, $sp
-; MIPS32R5EL-NEXT:    .cfi_def_cfa_register 30
-; MIPS32R5EL-NEXT:    addiu $1, $zero, -16
-; MIPS32R5EL-NEXT:    and $sp, $sp, $1
-; MIPS32R5EL-NEXT:    lui $1, %hi(gv2i32)
-; MIPS32R5EL-NEXT:    lw $2, %lo(gv2i32)($1)
-; MIPS32R5EL-NEXT:    sw $2, 0($sp)
-; MIPS32R5EL-NEXT:    addiu $1, $1, %lo(gv2i32)
-; MIPS32R5EL-NEXT:    lw $1, 4($1)
-; MIPS32R5EL-NEXT:    sw $1, 8($sp)
-; MIPS32R5EL-NEXT:    ld.w $w0, 0($sp)
-; MIPS32R5EL-NEXT:    copy_s.w $2, $w0[0]
-; MIPS32R5EL-NEXT:    copy_s.w $3, $w0[2]
-; MIPS32R5EL-NEXT:    move $sp, $fp
-; MIPS32R5EL-NEXT:    lw $fp, 24($sp) # 4-byte Folded Reload
-; MIPS32R5EL-NEXT:    lw $ra, 28($sp) # 4-byte Folded Reload
-; MIPS32R5EL-NEXT:    addiu $sp, $sp, 32
-; MIPS32R5EL-NEXT:    jr $ra
-; MIPS32R5EL-NEXT:    nop
   %1 = load <2 x i32>, <2 x i32> * @gv2i32
   ret <2 x i32> %1
 }
 
 define <4 x i32> @ret_4_i32() {
 ; MIPS32-LABEL: ret_4_i32:
 ; MIPS32:       # %bb.0:
 ; MIPS32-NEXT:    lui $1, %hi(gv4i32)

 ; MIPS64EB-NEXT:    sd $3, 8($1)
 ; MIPS64EB-NEXT:    sd $2, 0($1)
 ; MIPS64EB-NEXT:    ld $gp, 0($sp) # 8-byte Folded Reload
 ; MIPS64EB-NEXT:    ld $ra, 8($sp) # 8-byte Folded Reload
 ; MIPS64EB-NEXT:    daddiu $sp, $sp, 16
 ; MIPS64EB-NEXT:    jr $ra
 ; MIPS64EB-NEXT:    nop
 ;
-; MIPS32R5-LABEL: calli8_16:
-; MIPS32R5:       # %bb.0: # %entry
-; MIPS32R5-NEXT:    addiu $sp, $sp, -40
-; MIPS32R5-NEXT:    .cfi_def_cfa_offset 40
-; MIPS32R5-NEXT:    sw $ra, 36($sp) # 4-byte Folded Spill
-; MIPS32R5-NEXT:    .cfi_offset 31, -4
-; MIPS32R5-NEXT:    lui $1, %hi($CPI30_0)
-; MIPS32R5-NEXT:    addiu $1, $1, %lo($CPI30_0)
-; MIPS32R5-NEXT:    ld.w $w0, 0($1)
-; MIPS32R5-NEXT:    copy_s.w $4, $w0[0]
-; MIPS32R5-NEXT:    copy_s.w $5, $w0[1]
-; MIPS32R5-NEXT:    copy_s.w $6, $w0[2]
-; MIPS32R5-NEXT:    copy_s.w $7, $w0[3]
-; MIPS32R5-NEXT:    lui $1, %hi($CPI30_1)
-; MIPS32R5-NEXT:    addiu $1, $1, %lo($CPI30_1)
-; MIPS32R5-NEXT:    ld.w $w0, 0($1)
-; MIPS32R5-NEXT:    copy_s.w $1, $w0[0]
-; MIPS32R5-NEXT:    copy_s.w $2, $w0[1]
-; MIPS32R5-NEXT:    copy_s.w $3, $w0[2]
-; MIPS32R5-NEXT:    copy_s.w $8, $w0[3]
-; MIPS32R5-NEXT:    sw $8, 28($sp)
-; MIPS32R5-NEXT:    sw $3, 24($sp)
-; MIPS32R5-NEXT:    sw $2, 20($sp)
-; MIPS32R5-NEXT:    sw $1, 16($sp)
-; MIPS32R5-NEXT:    jal i8_16
-; MIPS32R5-NEXT:    nop
-; MIPS32R5-NEXT:    lui $1, %hi(gv16i8)
-; MIPS32R5-NEXT:    insert.w $w0[0], $2
-; MIPS32R5-NEXT:    insert.w $w0[1], $3
-; MIPS32R5-NEXT:    addiu $1, $1, %lo(gv16i8)
-; MIPS32R5-NEXT:    insert.w $w0[2], $4
-; MIPS32R5-NEXT:    insert.w $w0[3], $5
-; MIPS32R5-NEXT:    st.w $w0, 0($1)
-; MIPS32R5-NEXT:    lw $ra, 36($sp) # 4-byte Folded Reload
-; MIPS32R5-NEXT:    addiu $sp, $sp, 40
-; MIPS32R5-NEXT:    jr $ra
-; MIPS32R5-NEXT:    nop
+; MIPS32R5EB-LABEL: calli8_16:
+; MIPS32R5EB:       # %bb.0: # %entry
+; MIPS32R5EB-NEXT:    addiu $sp, $sp, -40
+; MIPS32R5EB-NEXT:    .cfi_def_cfa_offset 40
+; MIPS32R5EB-NEXT:    sw $ra, 36($sp) # 4-byte Folded Spill
+; MIPS32R5EB-NEXT:    .cfi_offset 31, -4
+; MIPS32R5EB-NEXT:    lui $1, 3080
+; MIPS32R5EB-NEXT:    ori $1, $1, 2314
+; MIPS32R5EB-NEXT:    lui $2, 1801
+; MIPS32R5EB-NEXT:    sw $1, 28($sp)
+; MIPS32R5EB-NEXT:    ori $1, $2, 1801
+; MIPS32R5EB-NEXT:    sw $1, 24($sp)
+; MIPS32R5EB-NEXT:    sw $1, 20($sp)
+; MIPS32R5EB-NEXT:    sw $1, 16($sp)
+; MIPS32R5EB-NEXT:    lui $1, 1543
+; MIPS32R5EB-NEXT:    ori $4, $1, 1543
+; MIPS32R5EB-NEXT:    ori $7, $1, 2314
+; MIPS32R5EB-NEXT:    move $5, $4
+; MIPS32R5EB-NEXT:    move $6, $4
+; MIPS32R5EB-NEXT:    jal i8_16
+; MIPS32R5EB-NEXT:    nop
+; MIPS32R5EB-NEXT:    insert.w $w0[0], $2
+; MIPS32R5EB-NEXT:    insert.w $w0[1], $3
+; MIPS32R5EB-NEXT:    insert.w $w0[2], $4
+; MIPS32R5EB-NEXT:    lui $1, %hi(gv16i8)
+; MIPS32R5EB-NEXT:    insert.w $w0[3], $5
+; MIPS32R5EB-NEXT:    addiu $1, $1, %lo(gv16i8)
+; MIPS32R5EB-NEXT:    st.w $w0, 0($1)
+; MIPS32R5EB-NEXT:    lw $ra, 36($sp) # 4-byte Folded Reload
+; MIPS32R5EB-NEXT:    addiu $sp, $sp, 40
+; MIPS32R5EB-NEXT:    jr $ra
+; MIPS32R5EB-NEXT:    nop
 ;
-; MIPS64R5-LABEL: calli8_16:
-; MIPS64R5:       # %bb.0: # %entry
-; MIPS64R5-NEXT:    daddiu $sp, $sp, -16
-; MIPS64R5-NEXT:    .cfi_def_cfa_offset 16
-; MIPS64R5-NEXT:    sd $ra, 8($sp) # 8-byte Folded Spill
-; MIPS64R5-NEXT:    sd $gp, 0($sp) # 8-byte Folded Spill
-; MIPS64R5-NEXT:    .cfi_offset 31, -8
-; MIPS64R5-NEXT:    .cfi_offset 28, -16
-; MIPS64R5-NEXT:    lui $1, %hi(%neg(%gp_rel(calli8_16)))
-; MIPS64R5-NEXT:    daddu $1, $1, $25
-; MIPS64R5-NEXT:    daddiu $gp, $1, %lo(%neg(%gp_rel(calli8_16)))
-; MIPS64R5-NEXT:    ld $1, %got_page(.LCPI30_0)($gp)
-; MIPS64R5-NEXT:    daddiu $1, $1, %got_ofst(.LCPI30_0)
-; MIPS64R5-NEXT:    ld.d $w0, 0($1)
-; MIPS64R5-NEXT:    copy_s.d $4, $w0[0]
-; MIPS64R5-NEXT:    copy_s.d $5, $w0[1]
-; MIPS64R5-NEXT:    ld $1, %got_page(.LCPI30_1)($gp)
-; MIPS64R5-NEXT:    daddiu $1, $1, %got_ofst(.LCPI30_1)
-; MIPS64R5-NEXT:    ld.d $w0, 0($1)
-; MIPS64R5-NEXT:    copy_s.d $6, $w0[0]
-; MIPS64R5-NEXT:    copy_s.d $7, $w0[1]
-; MIPS64R5-NEXT:    ld $25, %call16(i8_16)($gp)
-; MIPS64R5-NEXT:    jalr $25
-; MIPS64R5-NEXT:    nop
-; MIPS64R5-NEXT:    insert.d $w0[0], $2
-; MIPS64R5-NEXT:    insert.d $w0[1], $3
-; MIPS64R5-NEXT:    ld $1, %got_disp(gv16i8)($gp)
-; MIPS64R5-NEXT:    st.d $w0, 0($1)
-; MIPS64R5-NEXT:    ld $gp, 0($sp) # 8-byte Folded Reload
-; MIPS64R5-NEXT:    ld $ra, 8($sp) # 8-byte Folded Reload
-; MIPS64R5-NEXT:    daddiu $sp, $sp, 16
-; MIPS64R5-NEXT:    jr $ra
-; MIPS64R5-NEXT:    nop
+; MIPS64R5EB-LABEL: calli8_16:
+; MIPS64R5EB:       # %bb.0: # %entry
+; MIPS64R5EB-NEXT:    daddiu $sp, $sp, -16
+; MIPS64R5EB-NEXT:    .cfi_def_cfa_offset 16
+; MIPS64R5EB-NEXT:    sd $ra, 8($sp) # 8-byte Folded Spill
+; MIPS64R5EB-NEXT:    sd $gp, 0($sp) # 8-byte Folded Spill
+; MIPS64R5EB-NEXT:    .cfi_offset 31, -8
+; MIPS64R5EB-NEXT:    .cfi_offset 28, -16
+; MIPS64R5EB-NEXT:    lui $1, %hi(%neg(%gp_rel(calli8_16)))
+; MIPS64R5EB-NEXT:    daddu $1, $1, $25
+; MIPS64R5EB-NEXT:    daddiu $gp, $1, %lo(%neg(%gp_rel(calli8_16)))
+; MIPS64R5EB-NEXT:    lui $1, 1801
+; MIPS64R5EB-NEXT:    daddiu $1, $1, 1801
+; MIPS64R5EB-NEXT:    dsll $1, $1, 16
+; MIPS64R5EB-NEXT:    daddiu $1, $1, 1801
+; MIPS64R5EB-NEXT:    lui $2, 1543
+; MIPS64R5EB-NEXT:    dsll $1, $1, 16
+; MIPS64R5EB-NEXT:    daddiu $2, $2, 1543
+; MIPS64R5EB-NEXT:    dsll $2, $2, 16
+; MIPS64R5EB-NEXT:    daddiu $2, $2, 1543
+; MIPS64R5EB-NEXT:    dsll $2, $2, 16
+; MIPS64R5EB-NEXT:    daddiu $4, $2, 1543
+; MIPS64R5EB-NEXT:    daddiu $5, $2, 2314
+; MIPS64R5EB-NEXT:    daddiu $6, $1, 1801
+; MIPS64R5EB-NEXT:    lui $1, 225
+; MIPS64R5EB-NEXT:    daddiu $1, $1, 8417
+; MIPS64R5EB-NEXT:    dsll $1, $1, 16
+; MIPS64R5EB-NEXT:    daddiu $1, $1, 8577
+; MIPS64R5EB-NEXT:    dsll $1, $1, 19
+; MIPS64R5EB-NEXT:    daddiu $7, $1, 2314
+; MIPS64R5EB-NEXT:    ld $25, %call16(i8_16)($gp)
+; MIPS64R5EB-NEXT:    jalr $25
+; MIPS64R5EB-NEXT:    nop
+; MIPS64R5EB-NEXT:    ld $1, %got_disp(gv16i8)($gp)
+; MIPS64R5EB-NEXT:    insert.d $w0[0], $2
+; MIPS64R5EB-NEXT:    insert.d $w0[1], $3
+; MIPS64R5EB-NEXT:    st.d $w0, 0($1)
+; MIPS64R5EB-NEXT:    ld $gp, 0($sp) # 8-byte Folded Reload
+; MIPS64R5EB-NEXT:    ld $ra, 8($sp) # 8-byte Folded Reload
+; MIPS64R5EB-NEXT:    daddiu $sp, $sp, 16
+; MIPS64R5EB-NEXT:    jr $ra
+; MIPS64R5EB-NEXT:    nop
 ;
 ; MIPS32EL-LABEL: calli8_16:
 ; MIPS32EL:       # %bb.0: # %entry
 ; MIPS32EL-NEXT:    addiu $sp, $sp, -40
 ; MIPS32EL-NEXT:    .cfi_def_cfa_offset 40
 ; MIPS32EL-NEXT:    sw $ra, 36($sp) # 4-byte Folded Spill
 ; MIPS32EL-NEXT:    .cfi_offset 31, -4
 ; MIPS32EL-NEXT:    lui $1, 2569

 ; MIPS64EL-NEXT:    ld $1, %got_disp(gv16i8)($gp)
 ; MIPS64EL-NEXT:    sd $3, 8($1)
 ; MIPS64EL-NEXT:    sd $2, 0($1)
 ; MIPS64EL-NEXT:    ld $gp, 0($sp) # 8-byte Folded Reload
 ; MIPS64EL-NEXT:    ld $ra, 8($sp) # 8-byte Folded Reload
 ; MIPS64EL-NEXT:    daddiu $sp, $sp, 16
 ; MIPS64EL-NEXT:    jr $ra
 ; MIPS64EL-NEXT:    nop
+;
+; MIPS32R5EL-LABEL: calli8_16:
+; MIPS32R5EL:       # %bb.0: # %entry
+; MIPS32R5EL-NEXT:    addiu $sp, $sp, -40
+; MIPS32R5EL-NEXT:    .cfi_def_cfa_offset 40
+; MIPS32R5EL-NEXT:    sw $ra, 36($sp) # 4-byte Folded Spill
+; MIPS32R5EL-NEXT:    .cfi_offset 31, -4
+; MIPS32R5EL-NEXT:    lui $1, 2569
+; MIPS32R5EL-NEXT:    ori $2, $1, 2060
+; MIPS32R5EL-NEXT:    lui $3, 2311
+; MIPS32R5EL-NEXT:    sw $2, 28($sp)
+; MIPS32R5EL-NEXT:    ori $2, $3, 2311
+; MIPS32R5EL-NEXT:    sw $2, 24($sp)
+; MIPS32R5EL-NEXT:    sw $2, 20($sp)
+; MIPS32R5EL-NEXT:    sw $2, 16($sp)
+; MIPS32R5EL-NEXT:    lui $2, 1798
+; MIPS32R5EL-NEXT:    ori $4, $2, 1798
+; MIPS32R5EL-NEXT:    ori $7, $1, 1798
+; MIPS32R5EL-NEXT:    move $5, $4
+; MIPS32R5EL-NEXT:    move $6, $4
+; MIPS32R5EL-NEXT:    jal i8_16
+; MIPS32R5EL-NEXT:    nop
+; MIPS32R5EL-NEXT:    insert.w $w0[0], $2
+; MIPS32R5EL-NEXT:    insert.w $w0[1], $3
+; MIPS32R5EL-NEXT:    insert.w $w0[2], $4
+; MIPS32R5EL-NEXT:    lui $1, %hi(gv16i8)
+; MIPS32R5EL-NEXT:    insert.w $w0[3], $5
+; MIPS32R5EL-NEXT:    addiu $1, $1, %lo(gv16i8)
+; MIPS32R5EL-NEXT:    st.w $w0, 0($1)
+; MIPS32R5EL-NEXT:    lw $ra, 36($sp) # 4-byte Folded Reload
+; MIPS32R5EL-NEXT:    addiu $sp, $sp, 40
+; MIPS32R5EL-NEXT:    jr $ra
+; MIPS32R5EL-NEXT:    nop
+;
+; MIPS64R5EL-LABEL: calli8_16:
+; MIPS64R5EL:       # %bb.0: # %entry
+; MIPS64R5EL-NEXT:    daddiu $sp, $sp, -16
+; MIPS64R5EL-NEXT:    .cfi_def_cfa_offset 16
+; MIPS64R5EL-NEXT:    sd $ra, 8($sp) # 8-byte Folded Spill
+; MIPS64R5EL-NEXT:    sd $gp, 0($sp) # 8-byte Folded Spill
+; MIPS64R5EL-NEXT:    .cfi_offset 31, -8
+; MIPS64R5EL-NEXT:    .cfi_offset 28, -16
+; MIPS64R5EL-NEXT:    lui $1, %hi(%neg(%gp_rel(calli8_16)))
+; MIPS64R5EL-NEXT:    daddu $1, $1, $25
+; MIPS64R5EL-NEXT:    daddiu $gp, $1, %lo(%neg(%gp_rel(calli8_16)))
+; MIPS64R5EL-NEXT:    lui $1, 1285
+; MIPS64R5EL-NEXT:    daddiu $1, $1, -31869
+; MIPS64R5EL-NEXT:    dsll $1, $1, 16
+; MIPS64R5EL-NEXT:    daddiu $1, $1, 899
+; MIPS64R5EL-NEXT:    lui $2, 2311
+; MIPS64R5EL-NEXT:    daddiu $2, $2, 2311
+; MIPS64R5EL-NEXT:    dsll $2, $2, 16
+; MIPS64R5EL-NEXT:    daddiu $2, $2, 2311
+; MIPS64R5EL-NEXT:    dsll $2, $2, 16
+; MIPS64R5EL-NEXT:    dsll $1, $1, 17
+; MIPS64R5EL-NEXT:    lui $3, 899
+; MIPS64R5EL-NEXT:    daddiu $3, $3, 899
+; MIPS64R5EL-NEXT:    dsll $3, $3, 16
+; MIPS64R5EL-NEXT:    daddiu $3, $3, 899
+; MIPS64R5EL-NEXT:    dsll $3, $3, 17
+; MIPS64R5EL-NEXT:    daddiu $4, $3, 1798
+; MIPS64R5EL-NEXT:    daddiu $5, $1, 1798
+; MIPS64R5EL-NEXT:    daddiu $6, $2, 2311
+; MIPS64R5EL-NEXT:    lui $1, 642
+; MIPS64R5EL-NEXT:    daddiu $1, $1, 16899
+; MIPS64R5EL-NEXT:    dsll $1, $1, 18
+; MIPS64R5EL-NEXT:    daddiu $1, $1, 2311
+; MIPS64R5EL-NEXT:    dsll $1, $1, 16
+; MIPS64R5EL-NEXT:    daddiu $7, $1, 2311
+; MIPS64R5EL-NEXT:    ld $25, %call16(i8_16)($gp)
+; MIPS64R5EL-NEXT:    jalr $25
+; MIPS64R5EL-NEXT:    nop
+; MIPS64R5EL-NEXT:    ld $1, %got_disp(gv16i8)($gp)
+; MIPS64R5EL-NEXT:    insert.d $w0[0], $2
+; MIPS64R5EL-NEXT:    insert.d $w0[1], $3
+; MIPS64R5EL-NEXT:    st.d $w0, 0($1)
+; MIPS64R5EL-NEXT:    ld $gp, 0($sp) # 8-byte Folded Reload
+; MIPS64R5EL-NEXT:    ld $ra, 8($sp) # 8-byte Folded Reload
+; MIPS64R5EL-NEXT:    daddiu $sp, $sp, 16
+; MIPS64R5EL-NEXT:    jr $ra
+; MIPS64R5EL-NEXT:    nop
 entry:
   %0 = call <16 x i8> @i8_16(<16 x i8> <i8 6, i8 7,i8 6, i8 7,i8 6, i8 7,i8 6, i8 7,i8 6, i8 7,i8 6, i8 7, i8 6, i8 7, i8 9, i8 10>, <16 x i8> <i8 7, i8 9,i8 7, i8 9,i8 7, i8 9,i8 7, i8 9,i8 7, i8 9,i8 7, i8 9,i8 12, i8 8, i8 9, i8 10>)
   store <16 x i8> %0, <16 x i8> * @gv16i8
   ret void
 }
 
 define void @calli16_2() {
 ; MIPS32EB-LABEL: calli16_2:

 ; MIPS64EB-NEXT:    nop
 ;
 ; MIPS32R5EB-LABEL: calli16_8:
 ; MIPS32R5EB:       # %bb.0: # %entry
 ; MIPS32R5EB-NEXT:    addiu $sp, $sp, -40
 ; MIPS32R5EB-NEXT:    .cfi_def_cfa_offset 40
 ; MIPS32R5EB-NEXT:    sw $ra, 36($sp) # 4-byte Folded Spill
 ; MIPS32R5EB-NEXT:    .cfi_offset 31, -4
+; MIPS32R5EB-NEXT:    lui $1, 9
+; MIPS32R5EB-NEXT:    ori $5, $1, 10
+; MIPS32R5EB-NEXT:    sw $5, 28($sp)
+; MIPS32R5EB-NEXT:    lui $1, 12
+; MIPS32R5EB-NEXT:    ori $1, $1, 8
+; MIPS32R5EB-NEXT:    sw $1, 24($sp)
+; MIPS32R5EB-NEXT:    sw $5, 20($sp)
 ; MIPS32R5EB-NEXT:    lui $1, 6
-; MIPS32R5EB-NEXT:    ori $1, $1, 7
-; MIPS32R5EB-NEXT:    lui $2, 9
-; MIPS32R5EB-NEXT:    ori $2, $2, 10
-; MIPS32R5EB-NEXT:    fill.w $w0, $2
-; MIPS32R5EB-NEXT:    insert.w $w0[1], $1
-; MIPS32R5EB-NEXT:    splati.d $w0, $w0[0]
-; MIPS32R5EB-NEXT:    copy_s.w $4, $w0[0]
-; MIPS32R5EB-NEXT:    copy_s.w $5, $w0[1]
-; MIPS32R5EB-NEXT:    copy_s.w $6, $w0[2]
-; MIPS32R5EB-NEXT:    copy_s.w $7, $w0[3]
-; MIPS32R5EB-NEXT:    lui $1, %hi($CPI33_0)
-; MIPS32R5EB-NEXT:    addiu $1, $1, %lo($CPI33_0)
-; MIPS32R5EB-NEXT:    ld.w $w0, 0($1)
-; MIPS32R5EB-NEXT:    copy_s.w $1, $w0[0]
-; MIPS32R5EB-NEXT:    copy_s.w $2, $w0[1]
-; MIPS32R5EB-NEXT:    copy_s.w $3, $w0[2]
-; MIPS32R5EB-NEXT:    copy_s.w $8, $w0[3]
-; MIPS32R5EB-NEXT:    sw $8, 28($sp)
-; MIPS32R5EB-NEXT:    sw $3, 24($sp)
-; MIPS32R5EB-NEXT:    sw $2, 20($sp)
-; MIPS32R5EB-NEXT:    sw $1, 16($sp)
+; MIPS32R5EB-NEXT:    ori $4, $1, 7
+; MIPS32R5EB-NEXT:    sw $4, 16($sp)
+; MIPS32R5EB-NEXT:    move $6, $4
+; MIPS32R5EB-NEXT:    move $7, $5
 ; MIPS32R5EB-NEXT:    jal i16_8
 ; MIPS32R5EB-NEXT:    nop
-; MIPS32R5EB-NEXT:    lui $1, %hi(gv8i16)
-; MIPS32R5EB-NEXT:    addiu $1, $1, %lo(gv8i16)
 ; MIPS32R5EB-NEXT:    insert.w $w0[0], $2
 ; MIPS32R5EB-NEXT:    insert.w $w0[1], $3
 ; MIPS32R5EB-NEXT:    insert.w $w0[2], $4
+; MIPS32R5EB-NEXT:    lui $1, %hi(gv8i16)
 ; MIPS32R5EB-NEXT:    insert.w $w0[3], $5
+; MIPS32R5EB-NEXT:    addiu $1, $1, %lo(gv8i16)
 ; MIPS32R5EB-NEXT:    st.w $w0, 0($1)
 ; MIPS32R5EB-NEXT:    lw $ra, 36($sp) # 4-byte Folded Reload
 ; MIPS32R5EB-NEXT:    addiu $sp, $sp, 40
 ; MIPS32R5EB-NEXT:    jr $ra
 ; MIPS32R5EB-NEXT:    nop
 ;
 ; MIPS64R5EB-LABEL: calli16_8:
 ; MIPS64R5EB:       # %bb.0: # %entry
 ; MIPS64R5EB-NEXT:    daddiu $sp, $sp, -16
 ; MIPS64R5EB-NEXT:    .cfi_def_cfa_offset 16
 ; MIPS64R5EB-NEXT:    sd $ra, 8($sp) # 8-byte Folded Spill
 ; MIPS64R5EB-NEXT:    sd $gp, 0($sp) # 8-byte Folded Spill
 ; MIPS64R5EB-NEXT:    .cfi_offset 31, -8
 ; MIPS64R5EB-NEXT:    .cfi_offset 28, -16
 ; MIPS64R5EB-NEXT:    lui $1, %hi(%neg(%gp_rel(calli16_8)))
 ; MIPS64R5EB-NEXT:    daddu $1, $1, $25
 ; MIPS64R5EB-NEXT:    daddiu $gp, $1, %lo(%neg(%gp_rel(calli16_8)))
-; MIPS64R5EB-NEXT:    lui $1, 9
-; MIPS64R5EB-NEXT:    ori $1, $1, 10
-; MIPS64R5EB-NEXT:    lui $2, 6
-; MIPS64R5EB-NEXT:    ori $2, $2, 7
-; MIPS64R5EB-NEXT:    dinsu $1, $2, 32, 32
-; MIPS64R5EB-NEXT:    fill.d $w0, $1
-; MIPS64R5EB-NEXT:    copy_s.d $4, $w0[0]
-; MIPS64R5EB-NEXT:    copy_s.d $5, $w0[1]
-; MIPS64R5EB-NEXT:    ld $1, %got_page(.LCPI33_0)($gp)
-; MIPS64R5EB-NEXT:    daddiu $1, $1, %got_ofst(.LCPI33_0)
-; MIPS64R5EB-NEXT:    ld.d $w0, 0($1)
-; MIPS64R5EB-NEXT:    copy_s.d $6, $w0[0]
-; MIPS64R5EB-NEXT:    copy_s.d $7, $w0[1]
+; MIPS64R5EB-NEXT:    lui $1, 6
+; MIPS64R5EB-NEXT:    daddiu $1, $1, 7
+; MIPS64R5EB-NEXT:    dsll $1, $1, 16
+; MIPS64R5EB-NEXT:    daddiu $1, $1, 9
+; MIPS64R5EB-NEXT:    dsll $1, $1, 16
+; MIPS64R5EB-NEXT:    daddiu $4, $1, 10
+; MIPS64R5EB-NEXT:    lui $1, 2
+; MIPS64R5EB-NEXT:    daddiu $1, $1, -32767
+; MIPS64R5EB-NEXT:    dsll $1, $1, 19
+; MIPS64R5EB-NEXT:    daddiu $1, $1, 9
+; MIPS64R5EB-NEXT:    dsll $1, $1, 16
+; MIPS64R5EB-NEXT:    daddiu $7, $1, 10
 ; MIPS64R5EB-NEXT:    ld $25, %call16(i16_8)($gp)
+; MIPS64R5EB-NEXT:    move $5, $4
+; MIPS64R5EB-NEXT:    move $6, $4
 ; MIPS64R5EB-NEXT:    jalr $25
 ; MIPS64R5EB-NEXT:    nop
 ; MIPS64R5EB-NEXT:    ld $1, %got_disp(gv8i16)($gp)
 ; MIPS64R5EB-NEXT:    insert.d $w0[0], $2
 ; MIPS64R5EB-NEXT:    insert.d $w0[1], $3
 ; MIPS64R5EB-NEXT:    st.d $w0, 0($1)
 ; MIPS64R5EB-NEXT:    ld $gp, 0($sp) # 8-byte Folded Reload
 ; MIPS64R5EB-NEXT:    ld $ra, 8($sp) # 8-byte Folded Reload

 ;
 ; MIPS32R5EL-LABEL: calli16_8:
 ; MIPS32R5EL:       # %bb.0: # %entry
 ; MIPS32R5EL-NEXT:    addiu $sp, $sp, -40
 ; MIPS32R5EL-NEXT:    .cfi_def_cfa_offset 40
 ; MIPS32R5EL-NEXT:    sw $ra, 36($sp) # 4-byte Folded Spill
 ; MIPS32R5EL-NEXT:    .cfi_offset 31, -4
 ; MIPS32R5EL-NEXT:    lui $1, 10
-; MIPS32R5EL-NEXT:    ori $1, $1, 9
-; MIPS32R5EL-NEXT:    lui $2, 7
-; MIPS32R5EL-NEXT:    ori $2, $2, 6
-; MIPS32R5EL-NEXT:    fill.w $w0, $2
-; MIPS32R5EL-NEXT:    insert.w $w0[1], $1
-; MIPS32R5EL-NEXT:    splati.d $w0, $w0[0]
-; MIPS32R5EL-NEXT:    copy_s.w $4, $w0[0]
-; MIPS32R5EL-NEXT:    copy_s.w $5, $w0[1]
-; MIPS32R5EL-NEXT:    copy_s.w $6, $w0[2]
-; MIPS32R5EL-NEXT:    copy_s.w $7, $w0[3]
-; MIPS32R5EL-NEXT:    lui $1, %hi($CPI33_0)
-; MIPS32R5EL-NEXT:    addiu $1, $1, %lo($CPI33_0)
-; MIPS32R5EL-NEXT:    ld.w $w0, 0($1)
-; MIPS32R5EL-NEXT:    copy_s.w $1, $w0[0]
-; MIPS32R5EL-NEXT:    copy_s.w $2, $w0[1]
-; MIPS32R5EL-NEXT:    copy_s.w $3, $w0[2]
-; MIPS32R5EL-NEXT:    copy_s.w $8, $w0[3]
-; MIPS32R5EL-NEXT:    sw $8, 28($sp)
-; MIPS32R5EL-NEXT:    sw $3, 24($sp)
-; MIPS32R5EL-NEXT:    sw $2, 20($sp)
-; MIPS32R5EL-NEXT:    sw $1, 16($sp)
+; MIPS32R5EL-NEXT:    ori $5, $1, 9
+; MIPS32R5EL-NEXT:    sw $5, 28($sp)
+; MIPS32R5EL-NEXT:    lui $1, 8
+; MIPS32R5EL-NEXT:    ori $1, $1, 12
+; MIPS32R5EL-NEXT:    sw $1, 24($sp)
+; MIPS32R5EL-NEXT:    sw $5, 20($sp)
+; MIPS32R5EL-NEXT:    lui $1, 7
+; MIPS32R5EL-NEXT:    ori $4, $1, 6
+; MIPS32R5EL-NEXT:    sw $4, 16($sp)
+; MIPS32R5EL-NEXT:    move $6, $4
+; MIPS32R5EL-NEXT:    move $7, $5
 ; MIPS32R5EL-NEXT:    jal i16_8
 ; MIPS32R5EL-NEXT:    nop
-; MIPS32R5EL-NEXT:    lui $1, %hi(gv8i16)
-; MIPS32R5EL-NEXT:    addiu $1, $1, %lo(gv8i16)
 ; MIPS32R5EL-NEXT:    insert.w $w0[0], $2
 ; MIPS32R5EL-NEXT:    insert.w $w0[1], $3
 ; MIPS32R5EL-NEXT:    insert.w $w0[2], $4
+; MIPS32R5EL-NEXT:    lui $1, %hi(gv8i16)
 ; MIPS32R5EL-NEXT:    insert.w $w0[3], $5
+; MIPS32R5EL-NEXT:    addiu $1, $1, %lo(gv8i16)
 ; MIPS32R5EL-NEXT:    st.w $w0, 0($1)
 ; MIPS32R5EL-NEXT:    lw $ra, 36($sp) # 4-byte Folded Reload
 ; MIPS32R5EL-NEXT:    addiu $sp, $sp, 40
 ; MIPS32R5EL-NEXT:    jr $ra
 ; MIPS32R5EL-NEXT:    nop
 ;
 ; MIPS64R5EL-LABEL: calli16_8:
 ; MIPS64R5EL:       # %bb.0: # %entry
 ; MIPS64R5EL-NEXT:    daddiu $sp, $sp, -16
 ; MIPS64R5EL-NEXT:    .cfi_def_cfa_offset 16
 ; MIPS64R5EL-NEXT:    sd $ra, 8($sp) # 8-byte Folded Spill
 ; MIPS64R5EL-NEXT:    sd $gp, 0($sp) # 8-byte Folded Spill
 ; MIPS64R5EL-NEXT:    .cfi_offset 31, -8
 ; MIPS64R5EL-NEXT:    .cfi_offset 28, -16
 ; MIPS64R5EL-NEXT:    lui $1, %hi(%neg(%gp_rel(calli16_8)))
 ; MIPS64R5EL-NEXT:    daddu $1, $1, $25
 ; MIPS64R5EL-NEXT:    daddiu $gp, $1, %lo(%neg(%gp_rel(calli16_8)))
-; MIPS64R5EL-NEXT:    lui $1, 7
-; MIPS64R5EL-NEXT:    ori $1, $1, 6
-; MIPS64R5EL-NEXT:    lui $2, 10
-; MIPS64R5EL-NEXT:    ori $2, $2, 9
-; MIPS64R5EL-NEXT:    dinsu $1, $2, 32, 32
-; MIPS64R5EL-NEXT:    fill.d $w0, $1
-; MIPS64R5EL-NEXT:    copy_s.d $4, $w0[0]
-; MIPS64R5EL-NEXT:    copy_s.d $5, $w0[1]
-; MIPS64R5EL-NEXT:    ld $1, %got_page(.LCPI33_0)($gp)
-; MIPS64R5EL-NEXT:    daddiu $1, $1, %got_ofst(.LCPI33_0)
-; MIPS64R5EL-NEXT:    ld.d $w0, 0($1)
-; MIPS64R5EL-NEXT:    copy_s.d $6, $w0[0]
-; MIPS64R5EL-NEXT:    copy_s.d $7, $w0[1]
+; MIPS64R5EL-NEXT:    lui $1, 10
+; MIPS64R5EL-NEXT:    daddiu $1, $1, 9
+; MIPS64R5EL-NEXT:    dsll $1, $1, 16
+; MIPS64R5EL-NEXT:    daddiu $1, $1, 7
+; MIPS64R5EL-NEXT:    dsll $1, $1, 16
+; MIPS64R5EL-NEXT:    daddiu $4, $1, 6
+; MIPS64R5EL-NEXT:    lui $1, 1
+; MIPS64R5EL-NEXT:    daddiu $1, $1, 16385
+; MIPS64R5EL-NEXT:    dsll $1, $1, 16
+; MIPS64R5EL-NEXT:    daddiu $1, $1, 8193
+; MIPS64R5EL-NEXT:    dsll $1, $1, 19
+; MIPS64R5EL-NEXT:    daddiu $7, $1, 12
 ; MIPS64R5EL-NEXT:    ld $25, %call16(i16_8)($gp)
+; MIPS64R5EL-NEXT:    move $5, $4
+; MIPS64R5EL-NEXT:    move $6, $4
 ; MIPS64R5EL-NEXT:    jalr $25
 ; MIPS64R5EL-NEXT:    nop
 ; MIPS64R5EL-NEXT:    ld $1, %got_disp(gv8i16)($gp)
 ; MIPS64R5EL-NEXT:    insert.d $w0[0], $2
 ; MIPS64R5EL-NEXT:    insert.d $w0[1], $3
 ; MIPS64R5EL-NEXT:    st.d $w0, 0($1)
 ; MIPS64R5EL-NEXT:    ld $gp, 0($sp) # 8-byte Folded Reload
 ; MIPS64R5EL-NEXT:    ld $ra, 8($sp) # 8-byte Folded Reload

 ; MIPS32R5-NEXT:    insert.w $w0[3], $5
 ; MIPS32R5-NEXT:    addiu $1, $1, %lo(gv4i32)
 ; MIPS32R5-NEXT:    st.w $w0, 0($1)
 ; MIPS32R5-NEXT:    lw $ra, 36($sp) # 4-byte Folded Reload
 ; MIPS32R5-NEXT:    addiu $sp, $sp, 40
 ; MIPS32R5-NEXT:    jr $ra
 ; MIPS32R5-NEXT:    nop
 ;
-; MIPS64R5-LABEL: calli32_4:
-; MIPS64R5:       # %bb.0: # %entry
-; MIPS64R5-NEXT:    daddiu $sp, $sp, -16
-; MIPS64R5-NEXT:    .cfi_def_cfa_offset 16
-; MIPS64R5-NEXT:    sd $ra, 8($sp) # 8-byte Folded Spill
-; MIPS64R5-NEXT:    sd $gp, 0($sp) # 8-byte Folded Spill
-; MIPS64R5-NEXT:    .cfi_offset 31, -8
-; MIPS64R5-NEXT:    .cfi_offset 28, -16
-; MIPS64R5-NEXT:    lui $1, %hi(%neg(%gp_rel(calli32_4)))
-; MIPS64R5-NEXT:    daddu $1, $1, $25
-; MIPS64R5-NEXT:    daddiu $gp, $1, %lo(%neg(%gp_rel(calli32_4)))
-; MIPS64R5-NEXT:    ld $1, %got_page(.LCPI35_0)($gp)
-; MIPS64R5-NEXT:    daddiu $1, $1, %got_ofst(.LCPI35_0)
-; MIPS64R5-NEXT:    ld.d $w0, 0($1)
-; MIPS64R5-NEXT:    copy_s.d $4, $w0[0]
-; MIPS64R5-NEXT:    copy_s.d $5, $w0[1]
-; MIPS64R5-NEXT:    ld $1, %got_page(.LCPI35_1)($gp)
-; MIPS64R5-NEXT:    daddiu $1, $1, %got_ofst(.LCPI35_1)
-; MIPS64R5-NEXT:    ld.d $w0, 0($1)
-; MIPS64R5-NEXT:    copy_s.d $6, $w0[0]
-; MIPS64R5-NEXT:    copy_s.d $7, $w0[1]
-; MIPS64R5-NEXT:    ld $25, %call16(i32_4)($gp)
-; MIPS64R5-NEXT:    jalr $25
-; MIPS64R5-NEXT:    nop
-; MIPS64R5-NEXT:    insert.d $w0[0], $2
-; MIPS64R5-NEXT:    insert.d $w0[1], $3
-; MIPS64R5-NEXT:    ld $1, %got_disp(gv4i32)($gp)
-; MIPS64R5-NEXT:    st.d $w0, 0($1)
-; MIPS64R5-NEXT:    ld $gp, 0($sp) # 8-byte Folded Reload
-; MIPS64R5-NEXT:    ld $ra, 8($sp) # 8-byte Folded Reload
-; MIPS64R5-NEXT:    daddiu $sp, $sp, 16
-; MIPS64R5-NEXT:    jr $ra
-; MIPS64R5-NEXT:    nop
+; MIPS64R5EB-LABEL: calli32_4:
+; MIPS64R5EB:       # %bb.0: # %entry
+; MIPS64R5EB-NEXT:    daddiu $sp, $sp, -16
+; MIPS64R5EB-NEXT:    .cfi_def_cfa_offset 16
+; MIPS64R5EB-NEXT:    sd $ra, 8($sp) # 8-byte Folded Spill
+; MIPS64R5EB-NEXT:    sd $gp, 0($sp) # 8-byte Folded Spill
+; MIPS64R5EB-NEXT:    .cfi_offset 31, -8
+; MIPS64R5EB-NEXT:    .cfi_offset 28, -16
+; MIPS64R5EB-NEXT:    lui $1, %hi(%neg(%gp_rel(calli32_4)))
+; MIPS64R5EB-NEXT:    daddu $1, $1, $25
+; MIPS64R5EB-NEXT:    daddiu $gp, $1, %lo(%neg(%gp_rel(calli32_4)))
+; MIPS64R5EB-NEXT:    daddiu $1, $zero, 3
+; MIPS64R5EB-NEXT:    dsll $2, $1, 33
+; MIPS64R5EB-NEXT:    daddiu $4, $2, 7
+; MIPS64R5EB-NEXT:    dsll $1, $1, 34
+; MIPS64R5EB-NEXT:    daddiu $6, $1, 8
+; MIPS64R5EB-NEXT:    daddiu $1, $zero, 9
+; MIPS64R5EB-NEXT:    dsll $1, $1, 32
+; MIPS64R5EB-NEXT:    daddiu $5, $1, 10
+; MIPS64R5EB-NEXT:    ld $25, %call16(i32_4)($gp)
+; MIPS64R5EB-NEXT:    move $7, $5
+; MIPS64R5EB-NEXT:    jalr $25
+; MIPS64R5EB-NEXT:    nop
+; MIPS64R5EB-NEXT:    insert.d $w0[0], $2
+; MIPS64R5EB-NEXT:    insert.d $w0[1], $3
+; MIPS64R5EB-NEXT:    ld $1, %got_disp(gv4i32)($gp)
+; MIPS64R5EB-NEXT:    st.d $w0, 0($1)
+; MIPS64R5EB-NEXT:    ld $gp, 0($sp) # 8-byte Folded Reload
+; MIPS64R5EB-NEXT:    ld $ra, 8($sp) # 8-byte Folded Reload
+; MIPS64R5EB-NEXT:    daddiu $sp, $sp, 16
+; MIPS64R5EB-NEXT:    jr $ra
+; MIPS64R5EB-NEXT:    nop
 ;
 ; MIPS64EL-LABEL: calli32_4:
 ; MIPS64EL:       # %bb.0: # %entry
 ; MIPS64EL-NEXT:    daddiu $sp, $sp, -16
 ; MIPS64EL-NEXT:    .cfi_def_cfa_offset 16
 ; MIPS64EL-NEXT:    sd $ra, 8($sp) # 8-byte Folded Spill
 ; MIPS64EL-NEXT:    sd $gp, 0($sp) # 8-byte Folded Spill
 ; MIPS64EL-NEXT:    .cfi_offset 31, -8

 ; MIPS64EL-NEXT:    ld $1, %got_disp(gv4i32)($gp)
 ; MIPS64EL-NEXT:    sd $3, 8($1)
 ; MIPS64EL-NEXT:    sd $2, 0($1)
 ; MIPS64EL-NEXT:    ld $gp, 0($sp) # 8-byte Folded Reload
 ; MIPS64EL-NEXT:    ld $ra, 8($sp) # 8-byte Folded Reload
 ; MIPS64EL-NEXT:    daddiu $sp, $sp, 16
 ; MIPS64EL-NEXT:    jr $ra
 ; MIPS64EL-NEXT:    nop
+;
+; MIPS64R5EL-LABEL: calli32_4:
+; MIPS64R5EL:       # %bb.0: # %entry
+; MIPS64R5EL-NEXT:    daddiu $sp, $sp, -16
+; MIPS64R5EL-NEXT:    .cfi_def_cfa_offset 16
+; MIPS64R5EL-NEXT:    sd $ra, 8($sp) # 8-byte Folded Spill
+; MIPS64R5EL-NEXT:    sd $gp, 0($sp) # 8-byte Folded Spill
+; MIPS64R5EL-NEXT:    .cfi_offset 31, -8
+; MIPS64R5EL-NEXT:    .cfi_offset 28, -16
+; MIPS64R5EL-NEXT:    lui $1, %hi(%neg(%gp_rel(calli32_4)))
+; MIPS64R5EL-NEXT:    daddu $1, $1, $25
+; MIPS64R5EL-NEXT:    daddiu $gp, $1, %lo(%neg(%gp_rel(calli32_4)))
+; MIPS64R5EL-NEXT:    daddiu $1, $zero, 7
+; MIPS64R5EL-NEXT:    dsll $1, $1, 32
+; MIPS64R5EL-NEXT:    daddiu $4, $1, 6
+; MIPS64R5EL-NEXT:    daddiu $1, $zero, 1
+; MIPS64R5EL-NEXT:    dsll $1, $1, 35
+; MIPS64R5EL-NEXT:    daddiu $6, $1, 12
+; MIPS64R5EL-NEXT:    daddiu $1, $zero, 5
+; MIPS64R5EL-NEXT:    dsll $1, $1, 33
+; MIPS64R5EL-NEXT:    daddiu $5, $1, 9
+; MIPS64R5EL-NEXT:    ld $25, %call16(i32_4)($gp)
+; MIPS64R5EL-NEXT:    move $7, $5
+; MIPS64R5EL-NEXT:    jalr $25
+; MIPS64R5EL-NEXT:    nop
+; MIPS64R5EL-NEXT:    insert.d $w0[0], $2
+; MIPS64R5EL-NEXT:    insert.d $w0[1], $3
+; MIPS64R5EL-NEXT:    ld $1, %got_disp(gv4i32)($gp)
+; MIPS64R5EL-NEXT:    st.d $w0, 0($1)
+; MIPS64R5EL-NEXT:    ld $gp, 0($sp) # 8-byte Folded Reload
+; MIPS64R5EL-NEXT:    ld $ra, 8($sp) # 8-byte Folded Reload
+; MIPS64R5EL-NEXT:    daddiu $sp, $sp, 16
+; MIPS64R5EL-NEXT:    jr $ra
+; MIPS64R5EL-NEXT:    nop
 entry:
   %0 = call <4 x i32> @i32_4(<4 x i32> <i32 6, i32 7, i32 9, i32 10>, <4 x i32> <i32 12, i32 8, i32 9, i32 10>)
   store <4 x i32> %0, <4 x i32> * @gv4i32
   ret void
 }
 
 define void @calli64_2() {
 ; MIPS32EB-LABEL: calli64_2:

 ; MIPS64-NEXT:    sd $3, 8($1)
 ; MIPS64-NEXT:    sd $2, 0($1)
 ; MIPS64-NEXT:    ld $gp, 0($sp) # 8-byte Folded Reload
 ; MIPS64-NEXT:    ld $ra, 8($sp) # 8-byte Folded Reload
 ; MIPS64-NEXT:    daddiu $sp, $sp, 16
 ; MIPS64-NEXT:    jr $ra
 ; MIPS64-NEXT:    nop
 ;
-; MIPS32R5-LABEL: calli64_2:
-; MIPS32R5:       # %bb.0: # %entry
-; MIPS32R5-NEXT:    addiu $sp, $sp, -40
-; MIPS32R5-NEXT:    .cfi_def_cfa_offset 40
-; MIPS32R5-NEXT:    sw $ra, 36($sp) # 4-byte Folded Spill
-; MIPS32R5-NEXT:    .cfi_offset 31, -4
-; MIPS32R5-NEXT:    lui $1, %hi($CPI36_0)
-; MIPS32R5-NEXT:    addiu $1, $1, %lo($CPI36_0)
-; MIPS32R5-NEXT:    ld.w $w0, 0($1)
-; MIPS32R5-NEXT:    copy_s.w $4, $w0[0]
-; MIPS32R5-NEXT:    copy_s.w $5, $w0[1]
-; MIPS32R5-NEXT:    copy_s.w $6, $w0[2]
-; MIPS32R5-NEXT:    copy_s.w $7, $w0[3]
-; MIPS32R5-NEXT:    lui $1, %hi($CPI36_1)
-; MIPS32R5-NEXT:    addiu $1, $1, %lo($CPI36_1)
-; MIPS32R5-NEXT:    ld.w $w0, 0($1)
-; MIPS32R5-NEXT:    copy_s.w $1, $w0[0]
-; MIPS32R5-NEXT:    copy_s.w $2, $w0[1]
-; MIPS32R5-NEXT:    copy_s.w $3, $w0[2]
-; MIPS32R5-NEXT:    copy_s.w $8, $w0[3]
-; MIPS32R5-NEXT:    sw $8, 28($sp)
-; MIPS32R5-NEXT:    sw $3, 24($sp)
-; MIPS32R5-NEXT:    sw $2, 20($sp)
-; MIPS32R5-NEXT:    sw $1, 16($sp)
-; MIPS32R5-NEXT:    jal i64_2
-; MIPS32R5-NEXT:    nop
-; MIPS32R5-NEXT:    lui $1, %hi(gv2i64)
-; MIPS32R5-NEXT:    insert.w $w0[0], $2
-; MIPS32R5-NEXT:    insert.w $w0[1], $3
-; MIPS32R5-NEXT:    addiu $1, $1, %lo(gv2i64)
-; MIPS32R5-NEXT:    insert.w $w0[2], $4
-; MIPS32R5-NEXT:    insert.w $w0[3], $5
-; MIPS32R5-NEXT:    st.w $w0, 0($1)
-; MIPS32R5-NEXT:    lw $ra, 36($sp) # 4-byte Folded Reload
-; MIPS32R5-NEXT:    addiu $sp, $sp, 40
-; MIPS32R5-NEXT:    jr $ra
-; MIPS32R5-NEXT:    nop
+; MIPS32R5EB-LABEL: calli64_2:
+; MIPS32R5EB:       # %bb.0: # %entry
+; MIPS32R5EB-NEXT:    addiu $sp, $sp, -40
+; MIPS32R5EB-NEXT:    .cfi_def_cfa_offset 40
+; MIPS32R5EB-NEXT:    sw $ra, 36($sp) # 4-byte Folded Spill
+; MIPS32R5EB-NEXT:    .cfi_offset 31, -4
+; MIPS32R5EB-NEXT:    addiu $1, $zero, 8
+; MIPS32R5EB-NEXT:    sw $1, 28($sp)
+; MIPS32R5EB-NEXT:    addiu $1, $zero, 12
+; MIPS32R5EB-NEXT:    sw $1, 20($sp)
+; MIPS32R5EB-NEXT:    sw $zero, 24($sp)
+; MIPS32R5EB-NEXT:    sw $zero, 16($sp)
+; MIPS32R5EB-NEXT:    addiu $4, $zero, 0
+; MIPS32R5EB-NEXT:    addiu $5, $zero, 6
+; MIPS32R5EB-NEXT:    addiu $6, $zero, 0
+; MIPS32R5EB-NEXT:    addiu $7, $zero, 7
+; MIPS32R5EB-NEXT:    jal i64_2
+; MIPS32R5EB-NEXT:    nop
+; MIPS32R5EB-NEXT:    insert.w $w0[0], $2
+; MIPS32R5EB-NEXT:    insert.w $w0[1], $3
+; MIPS32R5EB-NEXT:    insert.w $w0[2], $4
+; MIPS32R5EB-NEXT:    lui $1, %hi(gv2i64)
+; MIPS32R5EB-NEXT:    insert.w $w0[3], $5
+; MIPS32R5EB-NEXT:    addiu $1, $1, %lo(gv2i64)
+; MIPS32R5EB-NEXT:    st.w $w0, 0($1)
+; MIPS32R5EB-NEXT:    lw $ra, 36($sp) # 4-byte Folded Reload
+; MIPS32R5EB-NEXT:    addiu $sp, $sp, 40
+; MIPS32R5EB-NEXT:    jr $ra
+; MIPS32R5EB-NEXT:    nop
 ;
 ; MIPS64R5-LABEL: calli64_2:
 ; MIPS64R5:       # %bb.0: # %entry
 ; MIPS64R5-NEXT:    daddiu $sp, $sp, -16
 ; MIPS64R5-NEXT:    .cfi_def_cfa_offset 16
 ; MIPS64R5-NEXT:    sd $ra, 8($sp) # 8-byte Folded Spill
 ; MIPS64R5-NEXT:    sd $gp, 0($sp) # 8-byte Folded Spill
 ; MIPS64R5-NEXT:    .cfi_offset 31, -8

 ; MIPS32EL-NEXT:    sw $5, 12($6)
 ; MIPS32EL-NEXT:    sw $4, 8($6)
 ; MIPS32EL-NEXT:    sw $3, 4($6)
 ; MIPS32EL-NEXT:    sw $2, %lo(gv2i64)($1)
 ; MIPS32EL-NEXT:    lw $ra, 36($sp) # 4-byte Folded Reload
 ; MIPS32EL-NEXT:    addiu $sp, $sp, 40
 ; MIPS32EL-NEXT:    jr $ra
 ; MIPS32EL-NEXT:    nop
+;
+; MIPS32R5EL-LABEL: calli64_2:
+; MIPS32R5EL:       # %bb.0: # %entry
+; MIPS32R5EL-NEXT:    addiu $sp, $sp, -40
+; MIPS32R5EL-NEXT:    .cfi_def_cfa_offset 40
+; MIPS32R5EL-NEXT:    sw $ra, 36($sp) # 4-byte Folded Spill
+; MIPS32R5EL-NEXT:    .cfi_offset 31, -4
+; MIPS32R5EL-NEXT:    addiu $1, $zero, 8
+; MIPS32R5EL-NEXT:    sw $1, 24($sp)
+; MIPS32R5EL-NEXT:    addiu $1, $zero, 12
+; MIPS32R5EL-NEXT:    sw $1, 16($sp)
+; MIPS32R5EL-NEXT:    sw $zero, 28($sp)
+; MIPS32R5EL-NEXT:    sw $zero, 20($sp)
+; MIPS32R5EL-NEXT:    addiu $4, $zero, 6
+; MIPS32R5EL-NEXT:    addiu $5, $zero, 0
+; MIPS32R5EL-NEXT:    addiu $6, $zero, 7
+; MIPS32R5EL-NEXT:    addiu $7, $zero, 0
+; MIPS32R5EL-NEXT:    jal i64_2
+; MIPS32R5EL-NEXT:    nop
+; MIPS32R5EL-NEXT:    insert.w $w0[0], $2
+; MIPS32R5EL-NEXT:    insert.w $w0[1], $3
+; MIPS32R5EL-NEXT:    insert.w $w0[2], $4
+; MIPS32R5EL-NEXT:    lui $1, %hi(gv2i64)
+; MIPS32R5EL-NEXT:    insert.w $w0[3], $5
+; MIPS32R5EL-NEXT:    addiu $1, $1, %lo(gv2i64)
+; MIPS32R5EL-NEXT:    st.w $w0, 0($1)
+; MIPS32R5EL-NEXT:    lw $ra, 36($sp) # 4-byte Folded Reload
+; MIPS32R5EL-NEXT:    addiu $sp, $sp, 40
+; MIPS32R5EL-NEXT:    jr $ra
+; MIPS32R5EL-NEXT:    nop
 entry:
   %0 = call <2 x i64> @i64_2(<2 x i64> <i64 6, i64 7>, <2 x i64> <i64 12, i64 8>)
   store <2 x i64> %0, <2 x i64> * @gv2i64
   ret void
 }
 
 declare <2 x float> @float2_extern(<2 x float>, <2 x float>)
 declare <4 x float> @float4_extern(<4 x float>, <4 x float>)

 ; MIPS32R5-NEXT:    swc1 $f0, 4($2)
 ; MIPS32R5-NEXT:    lwc1 $f0, 24($sp)
 ; MIPS32R5-NEXT:    swc1 $f0, %lo(gv2f32)($1)
 ; MIPS32R5-NEXT:    lw $ra, 36($sp) # 4-byte Folded Reload
 ; MIPS32R5-NEXT:    addiu $sp, $sp, 40
 ; MIPS32R5-NEXT:    jr $ra
 ; MIPS32R5-NEXT:    nop
 ;
-; MIPS64R5-LABEL: callfloat_2:
-; MIPS64R5:       # %bb.0: # %entry
-; MIPS64R5-NEXT:    daddiu $sp, $sp, -16
-; MIPS64R5-NEXT:    .cfi_def_cfa_offset 16
-; MIPS64R5-NEXT:    sd $ra, 8($sp) # 8-byte Folded Spill
-; MIPS64R5-NEXT:    sd $gp, 0($sp) # 8-byte Folded Spill
-; MIPS64R5-NEXT:    .cfi_offset 31, -8
-; MIPS64R5-NEXT:    .cfi_offset 28, -16
-; MIPS64R5-NEXT:    lui $1, %hi(%neg(%gp_rel(callfloat_2)))
-; MIPS64R5-NEXT:    daddu $1, $1, $25
-; MIPS64R5-NEXT:    daddiu $gp, $1, %lo(%neg(%gp_rel(callfloat_2)))
-; MIPS64R5-NEXT:    ld $1, %got_page(.LCPI37_0)($gp)
-; MIPS64R5-NEXT:    daddiu $1, $1, %got_ofst(.LCPI37_0)
-; MIPS64R5-NEXT:    ld.d $w0, 0($1)
-; MIPS64R5-NEXT:    copy_s.d $4, $w0[0]
-; MIPS64R5-NEXT:    ld $1, %got_page(.LCPI37_1)($gp)
-; MIPS64R5-NEXT:    daddiu $1, $1, %got_ofst(.LCPI37_1)
-; MIPS64R5-NEXT:    ld.d $w0, 0($1)
-; MIPS64R5-NEXT:    copy_s.d $5, $w0[0]
-; MIPS64R5-NEXT:    ld $25, %call16(float2_extern)($gp)
-; MIPS64R5-NEXT:    jalr $25
-; MIPS64R5-NEXT:    nop
-; MIPS64R5-NEXT:    ld $1, %got_disp(gv2f32)($gp)
-; MIPS64R5-NEXT:    sd $2, 0($1)
-; MIPS64R5-NEXT:    ld $gp, 0($sp) # 8-byte Folded Reload
-; MIPS64R5-NEXT:    ld $ra, 8($sp) # 8-byte Folded Reload
-; MIPS64R5-NEXT:    daddiu $sp, $sp, 16
-; MIPS64R5-NEXT:    jr $ra
-; MIPS64R5-NEXT:    nop
+; MIPS64R5EB-LABEL: callfloat_2:
+; MIPS64R5EB:       # %bb.0: # %entry
+; MIPS64R5EB-NEXT:    daddiu $sp, $sp, -16
+; MIPS64R5EB-NEXT:    .cfi_def_cfa_offset 16
+; MIPS64R5EB-NEXT:    sd $ra, 8($sp) # 8-byte Folded Spill
+; MIPS64R5EB-NEXT:    sd $gp, 0($sp) # 8-byte Folded Spill
+; MIPS64R5EB-NEXT:    .cfi_offset 31, -8
+; MIPS64R5EB-NEXT:    .cfi_offset 28, -16
+; MIPS64R5EB-NEXT:    lui $1, %hi(%neg(%gp_rel(callfloat_2)))
+; MIPS64R5EB-NEXT:    daddu $1, $1, $25
+; MIPS64R5EB-NEXT:    daddiu $gp, $1, %lo(%neg(%gp_rel(callfloat_2)))
+; MIPS64R5EB-NEXT:    daddiu $1, $zero, 383
+; MIPS64R5EB-NEXT:    dsll $4, $1, 23
+; MIPS64R5EB-NEXT:    daddiu $1, $zero, 261
+; MIPS64R5EB-NEXT:    dsll $1, $1, 33
+; MIPS64R5EB-NEXT:    daddiu $1, $1, 523
+; MIPS64R5EB-NEXT:    dsll $5, $1, 21
+; MIPS64R5EB-NEXT:    ld $25, %call16(float2_extern)($gp)
+; MIPS64R5EB-NEXT:    jalr $25
+; MIPS64R5EB-NEXT:    nop
+; MIPS64R5EB-NEXT:    ld $1, %got_disp(gv2f32)($gp)
+; MIPS64R5EB-NEXT:    sd $2, 0($1)
+; MIPS64R5EB-NEXT:    ld $gp, 0($sp) # 8-byte Folded Reload
+; MIPS64R5EB-NEXT:    ld $ra, 8($sp) # 8-byte Folded Reload
+; MIPS64R5EB-NEXT:    daddiu $sp, $sp, 16
+; MIPS64R5EB-NEXT:    jr $ra
+; MIPS64R5EB-NEXT:    nop
 ;
 ; MIPS64EL-LABEL: callfloat_2:
 ; MIPS64EL:       # %bb.0: # %entry
 ; MIPS64EL-NEXT:    daddiu $sp, $sp, -16
 ; MIPS64EL-NEXT:    .cfi_def_cfa_offset 16
 ; MIPS64EL-NEXT:    sd $ra, 8($sp) # 8-byte Folded Spill
 ; MIPS64EL-NEXT:    sd $gp, 0($sp) # 8-byte Folded Spill
 ; MIPS64EL-NEXT:    .cfi_offset 31, -8

 ; MIPS64EL-NEXT:    nop
 ; MIPS64EL-NEXT:    ld $1, %got_disp(gv2f32)($gp)
 ; MIPS64EL-NEXT:    sd $2, 0($1)
 ; MIPS64EL-NEXT:    ld $gp, 0($sp) # 8-byte Folded Reload
 ; MIPS64EL-NEXT:    ld $ra, 8($sp) # 8-byte Folded Reload
 ; MIPS64EL-NEXT:    daddiu $sp, $sp, 16
 ; MIPS64EL-NEXT:    jr $ra
 ; MIPS64EL-NEXT:    nop
+;
+; MIPS64R5EL-LABEL: callfloat_2:
+; MIPS64R5EL:       # %bb.0: # %entry
+; MIPS64R5EL-NEXT:    daddiu $sp, $sp, -16
+; MIPS64R5EL-NEXT:    .cfi_def_cfa_offset 16
+; MIPS64R5EL-NEXT:    sd $ra, 8($sp) # 8-byte Folded Spill
+; MIPS64R5EL-NEXT:    sd $gp, 0($sp) # 8-byte Folded Spill
+; MIPS64R5EL-NEXT:    .cfi_offset 31, -8
+; MIPS64R5EL-NEXT:    .cfi_offset 28, -16
+; MIPS64R5EL-NEXT:    lui $1, %hi(%neg(%gp_rel(callfloat_2)))
+; MIPS64R5EL-NEXT:    daddu $1, $1, $25
+; MIPS64R5EL-NEXT:    daddiu $gp, $1, %lo(%neg(%gp_rel(callfloat_2)))
+; MIPS64R5EL-NEXT:    daddiu $1, $zero, 383
+; MIPS64R5EL-NEXT:    dsll $4, $1, 55
+; MIPS64R5EL-NEXT:    daddiu $1, $zero, 523
+; MIPS64R5EL-NEXT:    dsll $1, $1, 31
+; MIPS64R5EL-NEXT:    daddiu $1, $1, 261
+; MIPS64R5EL-NEXT:    dsll $5, $1, 22
+; MIPS64R5EL-NEXT:    ld $25, %call16(float2_extern)($gp)
+; MIPS64R5EL-NEXT:    jalr $25
+; MIPS64R5EL-NEXT:    nop
+; MIPS64R5EL-NEXT:    ld $1, %got_disp(gv2f32)($gp)
+; MIPS64R5EL-NEXT:    sd $2, 0($1)
+; MIPS64R5EL-NEXT:    ld $gp, 0($sp) # 8-byte Folded Reload
+; MIPS64R5EL-NEXT:    ld $ra, 8($sp) # 8-byte Folded Reload
+; MIPS64R5EL-NEXT:    daddiu $sp, $sp, 16
+; MIPS64R5EL-NEXT:    jr $ra
+; MIPS64R5EL-NEXT:    nop
 entry:
   %0 = call <2 x float> @float2_extern(<2 x float> <float 0.0, float -1.0>, <2 x float> <float 12.0, float 14.0>)
   store <2 x float> %0, <2 x float> * @gv2f32
   ret void
 }
 
 define void @callfloat_4() {
 ; MIPS32-LABEL: callfloat_4:

 ; MIPS32R5-NEXT:    sw $ra, 76($sp) # 4-byte Folded Spill
 ; MIPS32R5-NEXT:    sw $fp, 72($sp) # 4-byte Folded Spill
 ; MIPS32R5-NEXT:    .cfi_offset 31, -4
 ; MIPS32R5-NEXT:    .cfi_offset 30, -8
 ; MIPS32R5-NEXT:    move $fp, $sp
 ; MIPS32R5-NEXT:    .cfi_def_cfa_register 30
 ; MIPS32R5-NEXT:    addiu $1, $zero, -16
 ; MIPS32R5-NEXT:    and $sp, $sp, $1
-; MIPS32R5-NEXT:    lui $1, %hi($CPI38_0)
-; MIPS32R5-NEXT:    addiu $1, $1, %lo($CPI38_0)
-; MIPS32R5-NEXT:    ld.w $w0, 0($1)
-; MIPS32R5-NEXT:    copy_s.w $6, $w0[0]
-; MIPS32R5-NEXT:    copy_s.w $7, $w0[1]
-; MIPS32R5-NEXT:    copy_s.w $1, $w0[2]
-; MIPS32R5-NEXT:    copy_s.w $2, $w0[3]
-; MIPS32R5-NEXT:    lui $3, %hi($CPI38_1)
-; MIPS32R5-NEXT:    addiu $3, $3, %lo($CPI38_1)
-; MIPS32R5-NEXT:    ld.w $w0, 0($3)
-; MIPS32R5-NEXT:    copy_s.w $3, $w0[0]
-; MIPS32R5-NEXT:    copy_s.w $4, $w0[1]
-; MIPS32R5-NEXT:    copy_s.w $5, $w0[2]
-; MIPS32R5-NEXT:    copy_s.w $8, $w0[3]
-; MIPS32R5-NEXT:    sw $8, 36($sp)
-; MIPS32R5-NEXT:    sw $5, 32($sp)
-; MIPS32R5-NEXT:    sw $4, 28($sp)
-; MIPS32R5-NEXT:    sw $3, 24($sp)
-; MIPS32R5-NEXT:    sw $2, 20($sp)
+; MIPS32R5-NEXT:    lui $1, 16704
+; MIPS32R5-NEXT:    lui $2, 16736
+; MIPS32R5-NEXT:    lui $3, 16752
+; MIPS32R5-NEXT:    lui $4, 16768
+; MIPS32R5-NEXT:    sw $4, 36($sp)
+; MIPS32R5-NEXT:    sw $3, 32($sp)
+; MIPS32R5-NEXT:    sw $2, 28($sp)
+; MIPS32R5-NEXT:    sw $1, 24($sp)
+; MIPS32R5-NEXT:    lui $1, 16512
+; MIPS32R5-NEXT:    sw $1, 20($sp)
+; MIPS32R5-NEXT:    lui $1, 16384
 ; MIPS32R5-NEXT:    sw $1, 16($sp)
 ; MIPS32R5-NEXT:    addiu $4, $sp, 48
+; MIPS32R5-NEXT:    addiu $6, $zero, 0
+; MIPS32R5-NEXT:    lui $7, 49024
 ; MIPS32R5-NEXT:    jal float4_extern
 ; MIPS32R5-NEXT:    nop
 ; MIPS32R5-NEXT:    lui $1, %hi(gv4f32)
 ; MIPS32R5-NEXT:    addiu $1, $1, %lo(gv4f32)
 ; MIPS32R5-NEXT:    ld.w $w0, 48($sp)
 ; MIPS32R5-NEXT:    st.w $w0, 0($1)
 ; MIPS32R5-NEXT:    move $sp, $fp
 ; MIPS32R5-NEXT:    lw $fp, 72($sp) # 4-byte Folded Reload
 ; MIPS32R5-NEXT:    lw $ra, 76($sp) # 4-byte Folded Reload
 ; MIPS32R5-NEXT:    addiu $sp, $sp, 80
 ; MIPS32R5-NEXT:    jr $ra
 ; MIPS32R5-NEXT:    nop
 ;
-; MIPS64R5-LABEL: callfloat_4:
-; MIPS64R5:       # %bb.0: # %entry
-; MIPS64R5-NEXT:    daddiu $sp, $sp, -16
-; MIPS64R5-NEXT:    .cfi_def_cfa_offset 16
-; MIPS64R5-NEXT:    sd $ra, 8($sp) # 8-byte Folded Spill
-; MIPS64R5-NEXT:    sd $gp, 0($sp) # 8-byte Folded Spill
-; MIPS64R5-NEXT:    .cfi_offset 31, -8
-; MIPS64R5-NEXT:    .cfi_offset 28, -16
-; MIPS64R5-NEXT:    lui $1, %hi(%neg(%gp_rel(callfloat_4)))
-; MIPS64R5-NEXT:    daddu $1, $1, $25
-; MIPS64R5-NEXT:    daddiu $gp, $1, %lo(%neg(%gp_rel(callfloat_4)))
-; MIPS64R5-NEXT:    ld $1, %got_page(.LCPI38_0)($gp)
-; MIPS64R5-NEXT:    daddiu $1, $1, %got_ofst(.LCPI38_0)
-; MIPS64R5-NEXT:    ld.d $w0, 0($1)
-; MIPS64R5-NEXT:    copy_s.d $4, $w0[0]
-; MIPS64R5-NEXT:    copy_s.d $5, $w0[1]
-; MIPS64R5-NEXT:    ld $1, %got_page(.LCPI38_1)($gp)
-; MIPS64R5-NEXT:    daddiu $1, $1, %got_ofst(.LCPI38_1)
-; MIPS64R5-NEXT:    ld.d $w0, 0($1)
-; MIPS64R5-NEXT:    copy_s.d $6, $w0[0]
-; MIPS64R5-NEXT:    copy_s.d $7, $w0[1]
-; MIPS64R5-NEXT:    ld $25, %call16(float4_extern)($gp)
-; MIPS64R5-NEXT:    jalr $25
-; MIPS64R5-NEXT:    nop
-; MIPS64R5-NEXT:    insert.d $w0[0], $2
-; MIPS64R5-NEXT:    insert.d $w0[1], $3
-; MIPS64R5-NEXT:    ld $1, %got_disp(gv4f32)($gp)
-; MIPS64R5-NEXT:    st.d $w0, 0($1)
-; MIPS64R5-NEXT:    ld $gp, 0($sp) # 8-byte Folded Reload
-; MIPS64R5-NEXT:    ld $ra, 8($sp) # 8-byte Folded Reload
-; MIPS64R5-NEXT:    daddiu $sp, $sp, 16
-; MIPS64R5-NEXT:    jr $ra
-; MIPS64R5-NEXT:    nop
+; MIPS64R5EB-LABEL: callfloat_4:
+; MIPS64R5EB:       # %bb.0: # %entry
+; MIPS64R5EB-NEXT:    daddiu $sp, $sp, -16
+; MIPS64R5EB-NEXT:    .cfi_def_cfa_offset 16
+; MIPS64R5EB-NEXT:    sd $ra, 8($sp) # 8-byte Folded Spill
+; MIPS64R5EB-NEXT:    sd $gp, 0($sp) # 8-byte Folded Spill
+; MIPS64R5EB-NEXT:    .cfi_offset 31, -8
+; MIPS64R5EB-NEXT:    .cfi_offset 28, -16
+; MIPS64R5EB-NEXT:    lui $1, %hi(%neg(%gp_rel(callfloat_4)))
+; MIPS64R5EB-NEXT:    daddu $1, $1, $25
+; MIPS64R5EB-NEXT:    daddiu $gp, $1, %lo(%neg(%gp_rel(callfloat_4)))
+; MIPS64R5EB-NEXT:    daddiu $1, $zero, 1
+; MIPS64R5EB-NEXT:    dsll $1, $1, 39
+; MIPS64R5EB-NEXT:    daddiu $1, $1, 129
+; MIPS64R5EB-NEXT:    daddiu $2, $zero, 261
+; MIPS64R5EB-NEXT:    dsll $2, $2, 33
+; MIPS64R5EB-NEXT:    daddiu $3, $zero, 383
+; MIPS64R5EB-NEXT:    dsll $4, $3, 23
+; MIPS64R5EB-NEXT:    dsll $5, $1, 23
+; MIPS64R5EB-NEXT:    daddiu $1, $2, 523
+; MIPS64R5EB-NEXT:    dsll $6, $1, 21
+; MIPS64R5EB-NEXT:    daddiu $1, $zero, 1047
+; MIPS64R5EB-NEXT:    dsll $1, $1, 29
+; MIPS64R5EB-NEXT:    daddiu $1, $1, 131
+; MIPS64R5EB-NEXT:    dsll $7, $1, 23
+; MIPS64R5EB-NEXT:    ld $25, %call16(float4_extern)($gp)
+; MIPS64R5EB-NEXT:    jalr $25
+; MIPS64R5EB-NEXT:    nop
+; MIPS64R5EB-NEXT:    insert.d $w0[0], $2
+; MIPS64R5EB-NEXT:    insert.d $w0[1], $3
+; MIPS64R5EB-NEXT:    ld $1, %got_disp(gv4f32)($gp)
+; MIPS64R5EB-NEXT:    st.d $w0, 0($1)
+; MIPS64R5EB-NEXT:    ld $gp, 0($sp) # 8-byte Folded Reload
+; MIPS64R5EB-NEXT:    ld $ra, 8($sp) # 8-byte Folded Reload
+; MIPS64R5EB-NEXT:    daddiu $sp, $sp, 16
+; MIPS64R5EB-NEXT:    jr $ra
+; MIPS64R5EB-NEXT:    nop
 ;
 ; MIPS64EL-LABEL: callfloat_4:
 ; MIPS64EL:       # %bb.0: # %entry
 ; MIPS64EL-NEXT:    daddiu $sp, $sp, -16
 ; MIPS64EL-NEXT:    .cfi_def_cfa_offset 16
 ; MIPS64EL-NEXT:    sd $ra, 8($sp) # 8-byte Folded Spill
 ; MIPS64EL-NEXT:    sd $gp, 0($sp) # 8-byte Folded Spill
 ; MIPS64EL-NEXT:    .cfi_offset 31, -8

 ; MIPS64EL-NEXT:    ld $1, %got_disp(gv4f32)($gp)
 ; MIPS64EL-NEXT:    sd $3, 8($1)
 ; MIPS64EL-NEXT:    sd $2, 0($1)
 ; MIPS64EL-NEXT:    ld $gp, 0($sp) # 8-byte Folded Reload
 ; MIPS64EL-NEXT:    ld $ra, 8($sp) # 8-byte Folded Reload
 ; MIPS64EL-NEXT:    daddiu $sp, $sp, 16
 ; MIPS64EL-NEXT:    jr $ra
 ; MIPS64EL-NEXT:    nop
+;
+; MIPS64R5EL-LABEL: callfloat_4:
+; MIPS64R5EL:       # %bb.0: # %entry
+; MIPS64R5EL-NEXT:    daddiu $sp, $sp, -16
+; MIPS64R5EL-NEXT:    .cfi_def_cfa_offset 16
+; MIPS64R5EL-NEXT:    sd $ra, 8($sp) # 8-byte Folded Spill
+; MIPS64R5EL-NEXT:    sd $gp, 0($sp) # 8-byte Folded Spill
+; MIPS64R5EL-NEXT:    .cfi_offset 31, -8
+; MIPS64R5EL-NEXT:    .cfi_offset 28, -16
+; MIPS64R5EL-NEXT:    lui $1, %hi(%neg(%gp_rel(callfloat_4)))
+; MIPS64R5EL-NEXT:    daddu $1, $1, $25
+; MIPS64R5EL-NEXT:    daddiu $gp, $1, %lo(%neg(%gp_rel(callfloat_4)))
+; MIPS64R5EL-NEXT:    daddiu $1, $zero, 129
+; MIPS64R5EL-NEXT:    dsll $1, $1, 25
+; MIPS64R5EL-NEXT:    daddiu $1, $1, 1
+; MIPS64R5EL-NEXT:    daddiu $2, $zero, 523
+; MIPS64R5EL-NEXT:    dsll $2, $2, 31
+; MIPS64R5EL-NEXT:    daddiu $3, $zero, 383
+; MIPS64R5EL-NEXT:    dsll $4, $3, 55
+; MIPS64R5EL-NEXT:    dsll $5, $1, 30
+; MIPS64R5EL-NEXT:    daddiu $1, $2, 261
+; MIPS64R5EL-NEXT:    dsll $6, $1, 22
+; MIPS64R5EL-NEXT:    daddiu $1, $zero, 131
+; MIPS64R5EL-NEXT:    dsll $1, $1, 35
+; MIPS64R5EL-NEXT:    daddiu $1, $1, 1047
+; MIPS64R5EL-NEXT:    dsll $7, $1, 20
+; MIPS64R5EL-NEXT:    ld $25, %call16(float4_extern)($gp)
+; MIPS64R5EL-NEXT:    jalr $25
+; MIPS64R5EL-NEXT:    nop
+; MIPS64R5EL-NEXT:    insert.d $w0[0], $2
+; MIPS64R5EL-NEXT:    insert.d $w0[1], $3
+; MIPS64R5EL-NEXT:    ld $1, %got_disp(gv4f32)($gp)
+; MIPS64R5EL-NEXT:    st.d $w0, 0($1)
+; MIPS64R5EL-NEXT:    ld $gp, 0($sp) # 8-byte Folded Reload
+; MIPS64R5EL-NEXT:    ld $ra, 8($sp) # 8-byte Folded Reload
+; MIPS64R5EL-NEXT:    daddiu $sp, $sp, 16
+; MIPS64R5EL-NEXT:    jr $ra
+; MIPS64R5EL-NEXT:    nop
 entry:
   %0 = call <4 x float> @float4_extern(<4 x float> <float 0.0, float -1.0, float 2.0, float 4.0>, <4 x float> <float 12.0, float 14.0, float 15.0, float 16.0>)
   store <4 x float> %0, <4 x float> * @gv4f32
   ret void
 }
 
 define void @calldouble_2() {
 ; MIPS32EB-LABEL: calldouble_2:

 ; MIPS64-NEXT:    sd $3, 8($1)
 ; MIPS64-NEXT:    sd $2, 0($1)
 ; MIPS64-NEXT:    ld $gp, 0($sp) # 8-byte Folded Reload
 ; MIPS64-NEXT:    ld $ra, 8($sp) # 8-byte Folded Reload
 ; MIPS64-NEXT:    daddiu $sp, $sp, 16
 ; MIPS64-NEXT:    jr $ra
 ; MIPS64-NEXT:    nop
 ;
-; MIPS32R5-LABEL: calldouble_2:
-; MIPS32R5:       # %bb.0: # %entry
-; MIPS32R5-NEXT:    addiu $sp, $sp, -80
-; MIPS32R5-NEXT:    .cfi_def_cfa_offset 80
-; MIPS32R5-NEXT:    sw $ra, 76($sp) # 4-byte Folded Spill
-; MIPS32R5-NEXT:    sw $fp, 72($sp) # 4-byte Folded Spill
-; MIPS32R5-NEXT:    .cfi_offset 31, -4
-; MIPS32R5-NEXT:    .cfi_offset 30, -8
-; MIPS32R5-NEXT:    move $fp, $sp
-; MIPS32R5-NEXT:    .cfi_def_cfa_register 30
-; MIPS32R5-NEXT:    addiu $1, $zero, -16
-; MIPS32R5-NEXT:    and $sp, $sp, $1
-; MIPS32R5-NEXT:    lui $1, %hi($CPI39_0)
-; MIPS32R5-NEXT:    addiu $1, $1, %lo($CPI39_0)
-; MIPS32R5-NEXT:    ld.w $w0, 0($1)
-; MIPS32R5-NEXT:    copy_s.w $6, $w0[0]
-; MIPS32R5-NEXT:    copy_s.w $7, $w0[1]
-; MIPS32R5-NEXT:    copy_s.w $1, $w0[2]
-; MIPS32R5-NEXT:    copy_s.w $2, $w0[3]
-; MIPS32R5-NEXT:    lui $3, %hi($CPI39_1)
-; MIPS32R5-NEXT:    addiu $3, $3, %lo($CPI39_1)
-; MIPS32R5-NEXT:    ld.w $w0, 0($3)
-; MIPS32R5-NEXT:    copy_s.w $3, $w0[0]
-; MIPS32R5-NEXT:    copy_s.w $4, $w0[1]
-; MIPS32R5-NEXT:    copy_s.w $5, $w0[2]
-; MIPS32R5-NEXT:    copy_s.w $8, $w0[3]
-; MIPS32R5-NEXT:    sw $8, 36($sp)
-; MIPS32R5-NEXT:    sw $5, 32($sp)
-; MIPS32R5-NEXT:    sw $4, 28($sp)
-; MIPS32R5-NEXT:    sw $3, 24($sp)
-; MIPS32R5-NEXT:    sw $2, 20($sp)
-; MIPS32R5-NEXT:    sw $1, 16($sp)
-; MIPS32R5-NEXT:    addiu $4, $sp, 48
-; MIPS32R5-NEXT:    jal double2_extern
-; MIPS32R5-NEXT:    nop
-; MIPS32R5-NEXT:    lui $1, %hi(gv2f64)
-; MIPS32R5-NEXT:    addiu $1, $1, %lo(gv2f64)
-; MIPS32R5-NEXT:    ld.d $w0, 48($sp)
-; MIPS32R5-NEXT:    st.d $w0, 0($1)
-; MIPS32R5-NEXT:    move $sp, $fp
-; MIPS32R5-NEXT:    lw $fp, 72($sp) # 4-byte Folded Reload
-; MIPS32R5-NEXT:    lw $ra, 76($sp) # 4-byte Folded Reload
-; MIPS32R5-NEXT:    addiu $sp, $sp, 80
-; MIPS32R5-NEXT:    jr $ra
-; MIPS32R5-NEXT:    nop
+; MIPS32R5EB-LABEL: calldouble_2:
+; MIPS32R5EB:       # %bb.0: # %entry
+; MIPS32R5EB-NEXT:    addiu $sp, $sp, -80
+; MIPS32R5EB-NEXT:    .cfi_def_cfa_offset 80
+; MIPS32R5EB-NEXT:    sw $ra, 76($sp) # 4-byte Folded Spill
+; MIPS32R5EB-NEXT:    sw $fp, 72($sp) # 4-byte Folded Spill
+; MIPS32R5EB-NEXT:    .cfi_offset 31, -4
+; MIPS32R5EB-NEXT:    .cfi_offset 30, -8
+; MIPS32R5EB-NEXT:    move $fp, $sp
+; MIPS32R5EB-NEXT:    .cfi_def_cfa_register 30
+; MIPS32R5EB-NEXT:    addiu $1, $zero, -16
+; MIPS32R5EB-NEXT:    and $sp, $sp, $1
+; MIPS32R5EB-NEXT:    lui $1, 16424
+; MIPS32R5EB-NEXT:    lui $2, 16428
+; MIPS32R5EB-NEXT:    sw $2, 32($sp)
+; MIPS32R5EB-NEXT:    sw $1, 24($sp)
+; MIPS32R5EB-NEXT:    lui $1, 49136
+; MIPS32R5EB-NEXT:    sw $1, 16($sp)
+; MIPS32R5EB-NEXT:    sw $zero, 36($sp)
+; MIPS32R5EB-NEXT:    sw $zero, 28($sp)
+; MIPS32R5EB-NEXT:    sw $zero, 20($sp)
+; MIPS32R5EB-NEXT:    addiu $4, $sp, 48
+; MIPS32R5EB-NEXT:    addiu $6, $zero, 0
+; MIPS32R5EB-NEXT:    addiu $7, $zero, 0
+; MIPS32R5EB-NEXT:    jal double2_extern
+; MIPS32R5EB-NEXT:    nop
+; MIPS32R5EB-NEXT:    lui $1, %hi(gv2f64)
+; MIPS32R5EB-NEXT:    addiu $1, $1, %lo(gv2f64)
+; MIPS32R5EB-NEXT:    ld.d $w0, 48($sp)
+; MIPS32R5EB-NEXT:    st.d $w0, 0($1)
+; MIPS32R5EB-NEXT:    move $sp, $fp
+; MIPS32R5EB-NEXT:    lw $fp, 72($sp) # 4-byte Folded Reload
+; MIPS32R5EB-NEXT:    lw $ra, 76($sp) # 4-byte Folded Reload
+; MIPS32R5EB-NEXT:    addiu $sp, $sp, 80
+; MIPS32R5EB-NEXT:    jr $ra
+; MIPS32R5EB-NEXT:    nop
 ;
 ; MIPS64R5-LABEL: calldouble_2:
 ; MIPS64R5:       # %bb.0: # %entry
 ; MIPS64R5-NEXT:    daddiu $sp, $sp, -16
 ; MIPS64R5-NEXT:    .cfi_def_cfa_offset 16
 ; MIPS64R5-NEXT:    sd $ra, 8($sp) # 8-byte Folded Spill
 ; MIPS64R5-NEXT:    sd $gp, 0($sp) # 8-byte Folded Spill
 ; MIPS64R5-NEXT:    .cfi_offset 31, -8
 ; MIPS64R5-NEXT:    .cfi_offset 28, -16
 ; MIPS64R5-NEXT:    lui $1, %hi(%neg(%gp_rel(calldouble_2)))
 ; MIPS64R5-NEXT:    daddu $1, $1, $25
 ; MIPS64R5-NEXT:    daddiu $gp, $1, %lo(%neg(%gp_rel(calldouble_2)))
-; MIPS64R5-NEXT:    ld $1, %got_page(.LCPI39_0)($gp)
-; MIPS64R5-NEXT:    daddiu $1, $1, %got_ofst(.LCPI39_0)
-; MIPS64R5-NEXT:    ld.d $w0, 0($1)
-; MIPS64R5-NEXT:    copy_s.d $4, $w0[0]
-; MIPS64R5-NEXT:    copy_s.d $5, $w0[1]
-; MIPS64R5-NEXT:    ld $1, %got_page(.LCPI39_1)($gp)
-; MIPS64R5-NEXT:    daddiu $1, $1, %got_ofst(.LCPI39_1)
-; MIPS64R5-NEXT:    ld.d $w0, 0($1)
-; MIPS64R5-NEXT:    copy_s.d $6, $w0[0]
-; MIPS64R5-NEXT:    copy_s.d $7, $w0[1]
+; MIPS64R5-NEXT:    daddiu $1, $zero, 3071
+; MIPS64R5-NEXT:    dsll $5, $1, 52
+; MIPS64R5-NEXT:    daddiu $1, $zero, 2053
+; MIPS64R5-NEXT:    dsll $6, $1, 51
+; MIPS64R5-NEXT:    daddiu $1, $zero, 4107
+; MIPS64R5-NEXT:    dsll $7, $1, 50
 ; MIPS64R5-NEXT:    ld $25, %call16(double2_extern)($gp)
+; MIPS64R5-NEXT:    daddiu $4, $zero, 0
 ; MIPS64R5-NEXT:    jalr $25
 ; MIPS64R5-NEXT:    nop
 ; MIPS64R5-NEXT:    insert.d $w0[0], $2
 ; MIPS64R5-NEXT:    insert.d $w0[1], $3
 ; MIPS64R5-NEXT:    ld $1, %got_disp(gv2f64)($gp)
 ; MIPS64R5-NEXT:    st.d $w0, 0($1)
 ; MIPS64R5-NEXT:    ld $gp, 0($sp) # 8-byte Folded Reload
 ; MIPS64R5-NEXT:    ld $ra, 8($sp) # 8-byte Folded Reload

 ; MIPS32EL-NEXT:    ldc1 $f0, 48($sp)
 ; MIPS32EL-NEXT:    sdc1 $f0, %lo(gv2f64)($1)
 ; MIPS32EL-NEXT:    move $sp, $fp
 ; MIPS32EL-NEXT:    lw $fp, 72($sp) # 4-byte Folded Reload
 ; MIPS32EL-NEXT:    lw $ra, 76($sp) # 4-byte Folded Reload
 ; MIPS32EL-NEXT:    addiu $sp, $sp, 80
 ; MIPS32EL-NEXT:    jr $ra
 ; MIPS32EL-NEXT:    nop
+;
+; MIPS32R5EL-LABEL: calldouble_2:
+; MIPS32R5EL:       # %bb.0: # %entry
+; MIPS32R5EL-NEXT:    addiu $sp, $sp, -80
+; MIPS32R5EL-NEXT:    .cfi_def_cfa_offset 80
+; MIPS32R5EL-NEXT:    sw $ra, 76($sp) # 4-byte Folded Spill
+; MIPS32R5EL-NEXT:    sw $fp, 72($sp) # 4-byte Folded Spill
+; MIPS32R5EL-NEXT:    .cfi_offset 31, -4
+; MIPS32R5EL-NEXT:    .cfi_offset 30, -8
+; MIPS32R5EL-NEXT:    move $fp, $sp
+; MIPS32R5EL-NEXT:    .cfi_def_cfa_register 30
+; MIPS32R5EL-NEXT:    addiu $1, $zero, -16
+; MIPS32R5EL-NEXT:    and $sp, $sp, $1
+; MIPS32R5EL-NEXT:    lui $1, 16424
+; MIPS32R5EL-NEXT:    lui $2, 16428
+; MIPS32R5EL-NEXT:    sw $2, 36($sp)
+; MIPS32R5EL-NEXT:    sw $1, 28($sp)
+; MIPS32R5EL-NEXT:    lui $1, 49136
+; MIPS32R5EL-NEXT:    sw $1, 20($sp)
+; MIPS32R5EL-NEXT:    sw $zero, 32($sp)
+; MIPS32R5EL-NEXT:    sw $zero, 24($sp)
+; MIPS32R5EL-NEXT:    sw $zero, 16($sp)
+; MIPS32R5EL-NEXT:    addiu $4, $sp, 48
+; MIPS32R5EL-NEXT:    addiu $6, $zero, 0
+; MIPS32R5EL-NEXT:    addiu $7, $zero, 0
+; MIPS32R5EL-NEXT:    jal double2_extern
+; MIPS32R5EL-NEXT:    nop
+; MIPS32R5EL-NEXT:    lui $1, %hi(gv2f64)
+; MIPS32R5EL-NEXT:    addiu $1, $1, %lo(gv2f64)
+; MIPS32R5EL-NEXT:    ld.d $w0, 48($sp)
+; MIPS32R5EL-NEXT:    st.d $w0, 0($1)
+; MIPS32R5EL-NEXT:    move $sp, $fp
+; MIPS32R5EL-NEXT:    lw $fp, 72($sp) # 4-byte Folded Reload
+; MIPS32R5EL-NEXT:    lw $ra, 76($sp) # 4-byte Folded Reload
+; MIPS32R5EL-NEXT:    addiu $sp, $sp, 80
+; MIPS32R5EL-NEXT:    jr $ra
+; MIPS32R5EL-NEXT:    nop
 entry:
   %0 = call <2 x double> @double2_extern(<2 x double> <double 0.0, double -1.0>, <2 x double> <double 12.0, double 14.0>)
   store <2 x double> %0, <2 x double> * @gv2f64
   ret void
 }
 
 ; The mixed tests show that due to alignment requirements, $5 is not used
 ; in argument passing.

llvm/test/CodeGen/PowerPC/pr45709.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -verify-machineinstrs -mtriple=powerpc64le-unknown-unknown \
 ; RUN:     -mcpu=pwr6 -ppc-asm-full-reg-names -mattr=-vsx \
 ; RUN:     -ppc-vsr-nums-as-vr < %s | FileCheck %s
 
 ; There is code in the SDAG to expand FMAX/FMIN with fast flags to SELECT_CC.
 ; On PPC, we had SELECT_CC legalized using Promote for all vector types
 ; (including the type that they are all promoted to - which caused an infinite
 ; loop in legalization). This test just ensures that we terminate on such input.
 define dso_local void @_ZN1a1bEv(<4 x float> %in) local_unnamed_addr #0 align 2 {
 ; CHECK-LABEL: _ZN1a1bEv:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    bc 12, 4*cr5+lt, .LBB0_6
+; CHECK-NEXT:    bc 12, 4*cr5+lt, .LBB0_4
 ; CHECK-NEXT:    b .LBB0_1
 ; CHECK-NEXT:  .LBB0_1: # %.preheader
 ; CHECK-NEXT:    b .LBB0_2
 ; CHECK-NEXT:  .LBB0_2:
 ; CHECK-NEXT:    b .LBB0_3
 ; CHECK-NEXT:  .LBB0_3:
 ; CHECK-NEXT:    addis r3, r2, .LCPI0_0@toc@ha
 ; CHECK-NEXT:    addi r3, r3, .LCPI0_0@toc@l
 ; CHECK-NEXT:    lvx v3, 0, r3
 ; CHECK-NEXT:    vperm v2, v2, v2, v3
-; CHECK-NEXT:    vxor v3, v3, v3
-; CHECK-NEXT:    addi r3, r1, -48
-; CHECK-NEXT:    stvx v3, 0, r3
 ; CHECK-NEXT:    addi r3, r1, -32
 ; CHECK-NEXT:    stvx v2, 0, r3
-; CHECK-NEXT:    lwz r3, -48(r1)
-; CHECK-NEXT:    lwz r4, -32(r1)
-; CHECK-NEXT:    cmpw r4, r3
-; CHECK-NEXT:    bc 12, gt, .LBB0_4
-; CHECK-NEXT:    b .LBB0_5
-; CHECK-NEXT:  .LBB0_4:
-; CHECK-NEXT:    addi r3, r4, 0
-; CHECK-NEXT:  .LBB0_5:
+; CHECK-NEXT:    lwz r3, -32(r1)
+; CHECK-NEXT:    srawi r4, r3, 31
+; CHECK-NEXT:    andc r3, r3, r4
 ; CHECK-NEXT:    cmpw r3, r3
-; CHECK-NEXT:    stw r3, -64(r1)
-; CHECK-NEXT:    addi r3, r1, -64
+; CHECK-NEXT:    stw r3, -48(r1)
+; CHECK-NEXT:    addi r3, r1, -48
 ; CHECK-NEXT:    lvx v2, 0, r3
 ; CHECK-NEXT:    addi r3, r1, -16
 ; CHECK-NEXT:    stvx v2, 0, r3
-; CHECK-NEXT:  .LBB0_6:
+; CHECK-NEXT:  .LBB0_4:
 ; CHECK-NEXT:    blr
   br i1 undef, label %7, label %1
 
 1:                                                ; preds = %1, %0
   br i1 undef, label %2, label %1
 
 2:                                                ; preds = %1
   %3 = shufflevector <4 x float> %in, <4 x float> undef, <4 x i32> <i32 2, i32 3, i32 1, i32 0>

llvm/test/CodeGen/Thumb2/active_lane_mask.ll

 ; CHECK-NEXT:    ldr r1, [sp, #56]
 ; CHECK-NEXT:    vmov q0[2], q0[0], r2, r1
 ; CHECK-NEXT:    ldr r1, [sp, #20]
 ; CHECK-NEXT:    ldr r2, [sp, #16]
 ; CHECK-NEXT:    vmov.32 q1[1], r1
 ; CHECK-NEXT:    ldr r1, [sp, #24]
 ; CHECK-NEXT:    vmov q1[2], q1[0], r2, r1
 ; CHECK-NEXT:    vpsel q0, q1, q0
-; CHECK-NEXT:    vmov r1, s2
-; CHECK-NEXT:    vmov.f32 s2, s1
-; CHECK-NEXT:    vmov r3, s0
+; CHECK-NEXT:    vmov r1, s0
 ; CHECK-NEXT:    vmov r2, s2
-; CHECK-NEXT:    strd r3, r2, [r0, #16]
-; CHECK-NEXT:    str r1, [r0, #24]
+; CHECK-NEXT:    vmov.f32 s2, s1
+; CHECK-NEXT:    vmov r3, s2
+; CHECK-NEXT:    strd r1, r3, [r0, #16]
+; CHECK-NEXT:    str r2, [r0, #24]
 ; CHECK-NEXT:    bx lr
 ; CHECK-NEXT:    .p2align 4
 ; CHECK-NEXT:  @ %bb.1:
 ; CHECK-NEXT:  .LCPI1_0:
 ; CHECK-NEXT:    .long 0 @ 0x0
 ; CHECK-NEXT:    .long 1 @ 0x1
 ; CHECK-NEXT:    .long 2 @ 0x2
 ; CHECK-NEXT:    .long 3 @ 0x3
; CHECK-NEXT:    .long 15 @ 0xf
   %active.lane.mask = call <16 x i1> @llvm.get.active.lane.mask.v16i1.i32(i32 %index, i32 %TC)
   %select = select <16 x i1> %active.lane.mask, <16 x i8> %V1, <16 x i8> %V2
   ret <16 x i8> %select
 }
 
 define void @test_width2(i32* nocapture readnone %x, i32* nocapture %y, i8 zeroext %m) {
 ; CHECK-LABEL: test_width2:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    push.w {r4, r5, r6, r7, r8, r9, lr}
+; CHECK-NEXT:    push {r4, r5, r7, lr}
 ; CHECK-NEXT:    sub sp, #4
-; CHECK-NEXT:    vpush {d8, d9}
-; CHECK-NEXT:    sub sp, #8
 ; CHECK-NEXT:    cmp r2, #0
-; CHECK-NEXT:    beq.w .LBB4_3
+; CHECK-NEXT:    beq .LBB4_3
 ; CHECK-NEXT:  @ %bb.1: @ %for.body.preheader
 ; CHECK-NEXT:    adds r0, r2, #1
-; CHECK-NEXT:    vmov q1[2], q1[0], r2, r2
+; CHECK-NEXT:    movs r3, #1
 ; CHECK-NEXT:    bic r0, r0, #1
-; CHECK-NEXT:    adr r2, .LCPI4_0
+; CHECK-NEXT:    vldr s0, .LCPI4_0
 ; CHECK-NEXT:    subs r0, #2
-; CHECK-NEXT:    movs r3, #1
-; CHECK-NEXT:    vmov.i64 q0, #0xffffffff
-; CHECK-NEXT:    vldrw.u32 q2, [r2]
+; CHECK-NEXT:    mov.w r12, #0
 ; CHECK-NEXT:    add.w lr, r3, r0, lsr #1
-; CHECK-NEXT:    mov.w r8, #0
-; CHECK-NEXT:    vand q1, q1, q0
+; CHECK-NEXT:    adr r0, .LCPI4_1
+; CHECK-NEXT:    vldrw.u32 q1, [r0]
+; CHECK-NEXT:    movs r3, #0
 ; CHECK-NEXT:  .LBB4_2: @ %vector.body
 ; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
-; CHECK-NEXT:    vmov q3[2], q3[0], r8, r8
-; CHECK-NEXT:    vmov r7, s6
-; CHECK-NEXT:    vand q3, q3, q0
-; CHECK-NEXT:    vmov r6, s7
-; CHECK-NEXT:    vmov r3, s14
-; CHECK-NEXT:    add.w r8, r8, #2
-; CHECK-NEXT:    vmov r9, s12
-; CHECK-NEXT:    vmov r2, s15
-; CHECK-NEXT:    vmov r0, s5
-; CHECK-NEXT:    adds r3, #1
-; CHECK-NEXT:    vmov q3[2], q3[0], r9, r3
-; CHECK-NEXT:    vand q3, q3, q0
-; CHECK-NEXT:    adc r12, r2, #0
-; CHECK-NEXT:    vmov r5, s14
-; CHECK-NEXT:    vmov r4, s15
-; CHECK-NEXT:    vmov r2, s4
-; CHECK-NEXT:    subs r7, r5, r7
-; CHECK-NEXT:    vmov r7, s12
-; CHECK-NEXT:    sbcs r4, r6
-; CHECK-NEXT:    vmov r6, s13
-; CHECK-NEXT:    mov.w r4, #0
+; CHECK-NEXT:    adds r0, r3, #1
+; CHECK-NEXT:    adc r4, r12, #0
+; CHECK-NEXT:    subs r5, r3, r2
+; CHECK-NEXT:    sbcs r5, r12, #0
+; CHECK-NEXT:    add.w r3, r3, #2
+; CHECK-NEXT:    mov.w r5, #0
 ; CHECK-NEXT:    it lo
-; CHECK-NEXT:    movlo r4, #1
-; CHECK-NEXT:    cmp r4, #0
-; CHECK-NEXT:    csetm r4, ne
-; CHECK-NEXT:    subs r2, r7, r2
-; CHECK-NEXT:    sbcs.w r0, r6, r0
+; CHECK-NEXT:    movlo r5, #1
+; CHECK-NEXT:    cmp r5, #0
+; CHECK-NEXT:    csetm r5, ne
+; CHECK-NEXT:    subs r0, r0, r2
+; CHECK-NEXT:    sbcs r0, r12, #0
 ; CHECK-NEXT:    mov.w r0, #0
 ; CHECK-NEXT:    it lo
 ; CHECK-NEXT:    movlo r0, #1
 ; CHECK-NEXT:    cmp r0, #0
 ; CHECK-NEXT:    csetm r0, ne
-; CHECK-NEXT:    vmov q3[2], q3[0], r0, r4
-; CHECK-NEXT:    vmov q3[3], q3[1], r0, r4
-; CHECK-NEXT:    eor.w r0, r5, r3
-; CHECK-NEXT:    orrs.w r0, r0, r12
+; CHECK-NEXT:    cmp r4, #0
+; CHECK-NEXT:    vmov q2[2], q2[0], r5, r0
+; CHECK-NEXT:    vmov q2[3], q2[1], r5, r0
 ; CHECK-NEXT:    cset r0, ne
 ; CHECK-NEXT:    cmp r0, #0
 ; CHECK-NEXT:    csetm r0, ne
-; CHECK-NEXT:    teq.w r7, r9
-; CHECK-NEXT:    cset r2, ne
-; CHECK-NEXT:    cmp r2, #0
-; CHECK-NEXT:    csetm r2, ne
-; CHECK-NEXT:    vmov q4[2], q4[0], r2, r0
-; CHECK-NEXT:    vmov q4[3], q4[1], r2, r0
-; CHECK-NEXT:    veor q4, q4, q2
-; CHECK-NEXT:    vand q4, q4, q3
-; CHECK-NEXT:    @ implicit-def: $q3
-; CHECK-NEXT:    vmov r2, s16
-; CHECK-NEXT:    vmov r0, s18
-; CHECK-NEXT:    and r2, r2, #1
-; CHECK-NEXT:    orr.w r3, r2, r0, lsl #1
-; CHECK-NEXT:    sub.w r2, r1, #8
-; CHECK-NEXT:    lsls r0, r3, #31
+; CHECK-NEXT:    vmov s2, r0
+; CHECK-NEXT:    vmov.f32 s1, s0
+; CHECK-NEXT:    vmov.f32 s3, s2
+; CHECK-NEXT:    veor q3, q0, q1
+; CHECK-NEXT:    vand q3, q3, q2
+; CHECK-NEXT:    @ implicit-def: $q2
+; CHECK-NEXT:    vmov r4, s12
+; CHECK-NEXT:    vmov r0, s14
+; CHECK-NEXT:    and r4, r4, #1
+; CHECK-NEXT:    orr.w r4, r4, r0, lsl #1
+; CHECK-NEXT:    lsls r0, r4, #31
+; CHECK-NEXT:    and r4, r4, #3
+; CHECK-NEXT:    sub.w r0, r1, #8
 ; CHECK-NEXT:    itt ne
-; CHECK-NEXT:    ldrne r0, [r2]
-; CHECK-NEXT:    vmovne.32 q3[0], r0
-; CHECK-NEXT:    and r0, r3, #3
-; CHECK-NEXT:    lsls r0, r0, #30
+; CHECK-NEXT:    ldrne r5, [r0]
+; CHECK-NEXT:    vmovne.32 q2[0], r5
+; CHECK-NEXT:    lsls r4, r4, #30
 ; CHECK-NEXT:    itt mi
-; CHECK-NEXT:    ldrmi r0, [r2, #4]
-; CHECK-NEXT:    vmovmi.32 q3[2], r0
-; CHECK-NEXT:    vmov r2, s16
-; CHECK-NEXT:    vmov r0, s18
-; CHECK-NEXT:    and r2, r2, #1
-; CHECK-NEXT:    orr.w r2, r2, r0, lsl #1
-; CHECK-NEXT:    lsls r0, r2, #31
+; CHECK-NEXT:    ldrmi r0, [r0, #4]
+; CHECK-NEXT:    vmovmi.32 q2[2], r0
+; CHECK-NEXT:    vmov r4, s12
+; CHECK-NEXT:    vmov r0, s14
+; CHECK-NEXT:    and r4, r4, #1
+; CHECK-NEXT:    orr.w r0, r4, r0, lsl #1
+; CHECK-NEXT:    lsls r4, r0, #31
+; CHECK-NEXT:    and r0, r0, #3
 ; CHECK-NEXT:    itt ne
-; CHECK-NEXT:    vmovne r0, s12
-; CHECK-NEXT:    strne r0, [r1]
-; CHECK-NEXT:    and r0, r2, #3
+; CHECK-NEXT:    vmovne r4, s8
+; CHECK-NEXT:    strne r4, [r1]
 ; CHECK-NEXT:    lsls r0, r0, #30
 ; CHECK-NEXT:    itt mi
-; CHECK-NEXT:    vmovmi r0, s14
+; CHECK-NEXT:    vmovmi r0, s10
 ; CHECK-NEXT:    strmi r0, [r1, #4]
 ; CHECK-NEXT:    adds r1, #8
 ; CHECK-NEXT:    le lr, .LBB4_2
 ; CHECK-NEXT:  .LBB4_3: @ %for.cond.cleanup
-; CHECK-NEXT:    add sp, #8
-; CHECK-NEXT:    vpop {d8, d9}
 ; CHECK-NEXT:    add sp, #4
-; CHECK-NEXT:    pop.w {r4, r5, r6, r7, r8, r9, pc}
+; CHECK-NEXT:    pop {r4, r5, r7, pc}
 ; CHECK-NEXT:    .p2align 4
 ; CHECK-NEXT:  @ %bb.4:
-; CHECK-NEXT:  .LCPI4_0:
+; CHECK-NEXT:  .LCPI4_1:
 ; CHECK-NEXT:    .long 1 @ 0x1
 ; CHECK-NEXT:    .long 0 @ 0x0
 ; CHECK-NEXT:    .long 1 @ 0x1
 ; CHECK-NEXT:    .long 0 @ 0x0
+; CHECK-NEXT:  .LCPI4_0:
+; CHECK-NEXT:    .long 0x00000000 @ float 0
 entry:
   %cmp9.not = icmp eq i8 %m, 0
   br i1 %cmp9.not, label %for.cond.cleanup, label %for.body.preheader
 
 for.body.preheader:                               ; preds = %entry
   %wide.trip.count = zext i8 %m to i32
   %n.rnd.up = add nuw nsw i32 %wide.trip.count, 1
   %n.vec = and i32 %n.rnd.up, 510

llvm/test/CodeGen/Thumb2/mve-laneinterleaving-cost.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=thumbv8.1m.main-none-none-eabi -mattr=+mve.fp -verify-machineinstrs %s -o - | FileCheck %s
 
 define arm_aapcs_vfpcc <4 x i32> @loads_i32(<4 x i32> *%A, <4 x i32> *%B, <4 x i32> *%C) {
 ; CHECK-LABEL: loads_i32:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .save {r4, lr}
 ; CHECK-NEXT:    push {r4, lr}
 ; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13}
 ; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13}
-; CHECK-NEXT:    vldrw.u32 q5, [r2]
-; CHECK-NEXT:    vldrw.u32 q6, [r1]
-; CHECK-NEXT:    vldrw.u32 q1, [r0]
-; CHECK-NEXT:    vmov.i64 q3, #0xffffffff
-; CHECK-NEXT:    vmov.f32 s8, s20
-; CHECK-NEXT:    vmov.f32 s16, s22
-; CHECK-NEXT:    vmov.f32 s10, s21
-; CHECK-NEXT:    vmov.f32 s18, s23
-; CHECK-NEXT:    vmov.f32 s20, s26
-; CHECK-NEXT:    vmov.f32 s22, s27
-; CHECK-NEXT:    vmov.f32 s0, s6
-; CHECK-NEXT:    vand q5, q5, q3
-; CHECK-NEXT:    vmov.f32 s2, s7
-; CHECK-NEXT:    vmov r0, s0
-; CHECK-NEXT:    vmov r2, s20
-; CHECK-NEXT:    vmov r1, s21
-; CHECK-NEXT:    vmov.f32 s26, s25
-; CHECK-NEXT:    vand q3, q6, q3
-; CHECK-NEXT:    vmov.f32 s6, s5
-; CHECK-NEXT:    asrs r3, r0, #31
-; CHECK-NEXT:    adds r0, r0, r2
-; CHECK-NEXT:    adcs r1, r3
-; CHECK-NEXT:    vmov r2, s16
-; CHECK-NEXT:    asrl r0, r1, r2
-; CHECK-NEXT:    vmov r2, s12
-; CHECK-NEXT:    vmov r1, s4
-; CHECK-NEXT:    vmov r3, s13
-; CHECK-NEXT:    adds r4, r1, r2
-; CHECK-NEXT:    vmov r2, s8
-; CHECK-NEXT:    asr.w r12, r1, #31
-; CHECK-NEXT:    adc.w r1, r12, r3
-; CHECK-NEXT:    asrl r4, r1, r2
+; CHECK-NEXT:    vldrw.u32 q0, [r0]
+; CHECK-NEXT:    vldrw.u32 q1, [r1]
+; CHECK-NEXT:    vldrw.u32 q2, [r2]
+; CHECK-NEXT:    vmov.f32 s16, s2
+; CHECK-NEXT:    vmov.f32 s20, s6
+; CHECK-NEXT:    vmov.f32 s22, s7
+; CHECK-NEXT:    vmov.f32 s18, s3
+; CHECK-NEXT:    vmov r0, s16
+; CHECK-NEXT:    vmov r1, s20
+; CHECK-NEXT:    vmov.f32 s24, s10
+; CHECK-NEXT:    vmov.f32 s26, s11
+; CHECK-NEXT:    vmov.f32 s2, s1
 ; CHECK-NEXT:    vmov r2, s22
-; CHECK-NEXT:    vmov r1, s2
-; CHECK-NEXT:    vmov q0[2], q0[0], r4, r0
-; CHECK-NEXT:    vmov r3, s23
-; CHECK-NEXT:    vmov r0, s6
-; CHECK-NEXT:    vmov r4, s14
-; CHECK-NEXT:    adds r2, r2, r1
-; CHECK-NEXT:    asr.w r12, r1, #31
-; CHECK-NEXT:    adc.w r1, r12, r3
-; CHECK-NEXT:    vmov r3, s18
-; CHECK-NEXT:    asrl r2, r1, r3
-; CHECK-NEXT:    vmov r3, s15
+; CHECK-NEXT:    vmov.f32 s6, s5
+; CHECK-NEXT:    vmov.f32 s12, s8
+; CHECK-NEXT:    vmov.f32 s14, s9
+; CHECK-NEXT:    adds.w r12, r0, r1
+; CHECK-NEXT:    asr.w r3, r0, #31
+; CHECK-NEXT:    adc r1, r3, #0
+; CHECK-NEXT:    vmov r0, s24
+; CHECK-NEXT:    asrl r12, r1, r0
+; CHECK-NEXT:    vmov r0, s18
+; CHECK-NEXT:    vmov r3, s6
+; CHECK-NEXT:    adds r2, r2, r0
+; CHECK-NEXT:    asr.w r1, r0, #31
+; CHECK-NEXT:    vmov r0, s26
+; CHECK-NEXT:    adc r1, r1, #0
+; CHECK-NEXT:    asrl r2, r1, r0
+; CHECK-NEXT:    vmov r0, s2
 ; CHECK-NEXT:    asrs r1, r0, #31
-; CHECK-NEXT:    adds r0, r0, r4
-; CHECK-NEXT:    adcs r1, r3
-; CHECK-NEXT:    vmov r3, s10
+; CHECK-NEXT:    adds r0, r0, r3
+; CHECK-NEXT:    adc r1, r1, #0
+; CHECK-NEXT:    vmov r3, s14
 ; CHECK-NEXT:    asrl r0, r1, r3
+; CHECK-NEXT:    vmov r3, s4
+; CHECK-NEXT:    vmov r1, s0
+; CHECK-NEXT:    adds r4, r1, r3
+; CHECK-NEXT:    vmov r3, s8
+; CHECK-NEXT:    asr.w lr, r1, #31
+; CHECK-NEXT:    adc r1, lr, #0
+; CHECK-NEXT:    asrl r4, r1, r3
+; CHECK-NEXT:    vmov q0[2], q0[0], r4, r12
 ; CHECK-NEXT:    vmov q0[3], q0[1], r0, r2
 ; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13}
 ; CHECK-NEXT:    pop {r4, pc}
 entry:
   %a = load <4 x i32>, <4 x i32> *%A, align 4
   %b = load <4 x i32>, <4 x i32> *%B, align 4
   %c = load <4 x i32>, <4 x i32> *%C, align 4
   %sa = sext <4 x i32> %a to <4 x i64>
entry:
   %sh = ashr <16 x i16> %add, %sc
   %t = trunc <16 x i16> %sh to <16 x i8>
   ret <16 x i8> %t
 }
 
 define arm_aapcs_vfpcc void @load_store_i32(<4 x i32> *%A, <4 x i32> *%B, <4 x i32> *%C, <4 x i32> *%D) {
 ; CHECK-LABEL: load_store_i32:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .save {r4, lr}
-; CHECK-NEXT:    push {r4, lr}
+; CHECK-NEXT:    .save {r4, r6, r8, lr}
+; CHECK-NEXT:    push.w {r4, r6, r8, lr}
 ; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13}
 ; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13}
-; CHECK-NEXT:    vldrw.u32 q5, [r2]
-; CHECK-NEXT:    vldrw.u32 q6, [r1]
 ; CHECK-NEXT:    vldrw.u32 q0, [r0]
-; CHECK-NEXT:    vmov.i64 q3, #0xffffffff
-; CHECK-NEXT:    vmov.f32 s4, s20
-; CHECK-NEXT:    vmov.f32 s16, s22
-; CHECK-NEXT:    vmov.f32 s6, s21
-; CHECK-NEXT:    vmov.f32 s18, s23
-; CHECK-NEXT:    vmov.f32 s20, s26
-; CHECK-NEXT:    vmov.f32 s22, s27
-; CHECK-NEXT:    vmov.f32 s8, s2
-; CHECK-NEXT:    vand q5, q5, q3
-; CHECK-NEXT:    vmov.f32 s10, s3
-; CHECK-NEXT:    vmov r0, s8
-; CHECK-NEXT:    vmov r2, s20
-; CHECK-NEXT:    vmov r1, s21
-; CHECK-NEXT:    vmov.f32 s26, s25
-; CHECK-NEXT:    vand q3, q6, q3
+; CHECK-NEXT:    vldrw.u32 q1, [r1]
+; CHECK-NEXT:    vldrw.u32 q2, [r2]
+; CHECK-NEXT:    vmov.f32 s12, s2
+; CHECK-NEXT:    vmov.f32 s16, s6
+; CHECK-NEXT:    vmov.f32 s18, s7
+; CHECK-NEXT:    vmov.f32 s14, s3
+; CHECK-NEXT:    vmov.f32 s20, s10
+; CHECK-NEXT:    vmov.f32 s22, s11
 ; CHECK-NEXT:    vmov.f32 s2, s1
-; CHECK-NEXT:    vmov lr, s13
-; CHECK-NEXT:    asr.w r12, r0, #31
-; CHECK-NEXT:    adds r0, r0, r2
-; CHECK-NEXT:    adc.w r1, r1, r12
-; CHECK-NEXT:    vmov r2, s16
-; CHECK-NEXT:    asrl r0, r1, r2
-; CHECK-NEXT:    vmov r2, s12
-; CHECK-NEXT:    vmov r1, s0
-; CHECK-NEXT:    adds r4, r1, r2
-; CHECK-NEXT:    vmov r2, s4
-; CHECK-NEXT:    asr.w r12, r1, #31
-; CHECK-NEXT:    adc.w r1, r12, lr
-; CHECK-NEXT:    asrl r4, r1, r2
+; CHECK-NEXT:    vmov.f32 s6, s5
+; CHECK-NEXT:    vmov.f32 s24, s8
+; CHECK-NEXT:    vmov r1, s18
+; CHECK-NEXT:    vmov r0, s14
+; CHECK-NEXT:    vmov.f32 s26, s9
 ; CHECK-NEXT:    vmov r2, s22
-; CHECK-NEXT:    vmov r1, s10
-; CHECK-NEXT:    vmov q2[2], q2[0], r4, r0
-; CHECK-NEXT:    vmov lr, s23
-; CHECK-NEXT:    vmov r0, s2
-; CHECK-NEXT:    vmov r4, s15
-; CHECK-NEXT:    adds r2, r2, r1
-; CHECK-NEXT:    asr.w r12, r1, #31
-; CHECK-NEXT:    adc.w r1, r12, lr
-; CHECK-NEXT:    vmov r12, s18
-; CHECK-NEXT:    asrl r2, r1, r12
+; CHECK-NEXT:    vmov r4, s16
+; CHECK-NEXT:    vmov r6, s4
+; CHECK-NEXT:    adds.w r8, r0, r1
 ; CHECK-NEXT:    asr.w r12, r0, #31
-; CHECK-NEXT:    vmov r1, s14
-; CHECK-NEXT:    adds r0, r0, r1
-; CHECK-NEXT:    adc.w r1, r12, r4
-; CHECK-NEXT:    vmov r4, s6
+; CHECK-NEXT:    adc r1, r12, #0
+; CHECK-NEXT:    asrl r8, r1, r2
+; CHECK-NEXT:    vmov r2, s6
+; CHECK-NEXT:    vmov r1, s2
+; CHECK-NEXT:    adds r2, r2, r1
+; CHECK-NEXT:    asr.w r0, r1, #31
+; CHECK-NEXT:    adc r1, r0, #0
+; CHECK-NEXT:    vmov r0, s26
+; CHECK-NEXT:    asrl r2, r1, r0
+; CHECK-NEXT:    vmov r0, s12
+; CHECK-NEXT:    asrs r1, r0, #31
+; CHECK-NEXT:    adds r0, r0, r4
+; CHECK-NEXT:    adc r1, r1, #0
+; CHECK-NEXT:    vmov r4, s20
 ; CHECK-NEXT:    asrl r0, r1, r4
-; CHECK-NEXT:    vmov q2[3], q2[1], r0, r2
-; CHECK-NEXT:    vstrw.32 q2, [r3]
+; CHECK-NEXT:    vmov r1, s0
+; CHECK-NEXT:    adds r6, r6, r1
+; CHECK-NEXT:    asr.w r4, r1, #31
+; CHECK-NEXT:    adc r1, r4, #0
+; CHECK-NEXT:    vmov r4, s8
+; CHECK-NEXT:    asrl r6, r1, r4
+; CHECK-NEXT:    vmov q0[2], q0[0], r6, r0
+; CHECK-NEXT:    vmov q0[3], q0[1], r2, r8
+; CHECK-NEXT:    vstrw.32 q0, [r3]
 ; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13}
-; CHECK-NEXT:    pop {r4, pc}
+; CHECK-NEXT:    pop.w {r4, r6, r8, pc}
 entry:
   %a = load <4 x i32>, <4 x i32> *%A, align 4
   %b = load <4 x i32>, <4 x i32> *%B, align 4
   %c = load <4 x i32>, <4 x i32> *%C, align 4
   %sa = sext <4 x i32> %a to <4 x i64>
   %sb = zext <4 x i32> %b to <4 x i64>
   %sc = zext <4 x i32> %c to <4 x i64>
   %add = add <4 x i64> %sa, %sb
entry:
   store <16 x i8> %t, <16 x i8> *%D, align 4
   ret void
 }
 
 
 define arm_aapcs_vfpcc void @load_one_store_i32(<4 x i32> *%A, <4 x i32> *%D) {
 ; CHECK-LABEL: load_one_store_i32:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .save {r4, r5, r7, lr}
-; CHECK-NEXT:    push {r4, r5, r7, lr}
+; CHECK-NEXT:    .save {r4, r5, r6, lr}
+; CHECK-NEXT:    push {r4, r5, r6, lr}
 ; CHECK-NEXT:    vldrw.u32 q0, [r0]
 ; CHECK-NEXT:    vmov.f32 s4, s2
 ; CHECK-NEXT:    vmov.f32 s6, s3
 ; CHECK-NEXT:    vmov.f32 s2, s1
 ; CHECK-NEXT:    vmov r2, s6
 ; CHECK-NEXT:    adds.w r12, r2, r2
 ; CHECK-NEXT:    asr.w r3, r2, #31
-; CHECK-NEXT:    adc.w r7, r3, r2, asr #31
-; CHECK-NEXT:    vmov r3, s4
-; CHECK-NEXT:    asrl r12, r7, r2
-; CHECK-NEXT:    adds r0, r3, r3
-; CHECK-NEXT:    asr.w r5, r3, #31
-; CHECK-NEXT:    adc.w r5, r5, r3, asr #31
-; CHECK-NEXT:    asrl r0, r5, r3
-; CHECK-NEXT:    vmov r3, s0
-; CHECK-NEXT:    adds r4, r3, r3
-; CHECK-NEXT:    asr.w r5, r3, #31
-; CHECK-NEXT:    adc.w r5, r5, r3, asr #31
-; CHECK-NEXT:    asrl r4, r5, r3
-; CHECK-NEXT:    vmov q1[2], q1[0], r4, r0
-; CHECK-NEXT:    vmov r0, s2
+; CHECK-NEXT:    adc.w r3, r3, r2, asr #31
+; CHECK-NEXT:    asrl r12, r3, r2
+; CHECK-NEXT:    vmov r3, s2
+; CHECK-NEXT:    adds r2, r3, r3
+; CHECK-NEXT:    asr.w r0, r3, #31
+; CHECK-NEXT:    adc.w r5, r0, r3, asr #31
+; CHECK-NEXT:    vmov r0, s4
+; CHECK-NEXT:    asrl r2, r5, r3
 ; CHECK-NEXT:    adds r4, r0, r0
-; CHECK-NEXT:    asr.w r2, r0, #31
-; CHECK-NEXT:    adc.w r3, r2, r0, asr #31
+; CHECK-NEXT:    asr.w r3, r0, #31
+; CHECK-NEXT:    adc.w r3, r3, r0, asr #31
 ; CHECK-NEXT:    asrl r4, r3, r0
-; CHECK-NEXT:    vmov q1[3], q1[1], r4, r12
-; CHECK-NEXT:    vstrw.32 q1, [r1]
-; CHECK-NEXT:    pop {r4, r5, r7, pc}
+; CHECK-NEXT:    vmov r0, s0
+; CHECK-NEXT:    adds r6, r0, r0
+; CHECK-NEXT:    asr.w r3, r0, #31
+; CHECK-NEXT:    adc.w r3, r3, r0, asr #31
+; CHECK-NEXT:    asrl r6, r3, r0
+; CHECK-NEXT:    vmov q0[2], q0[0], r6, r4
+; CHECK-NEXT:    vmov q0[3], q0[1], r2, r12
+; CHECK-NEXT:    vstrw.32 q0, [r1]
+; CHECK-NEXT:    pop {r4, r5, r6, pc}
 entry:
   %a = load <4 x i32>, <4 x i32> *%A, align 4
   %sa = sext <4 x i32> %a to <4 x i64>
   %add = add <4 x i64> %sa, %sa
   %sh = ashr <4 x i64> %add, %sa
   %t = trunc <4 x i64> %sh to <4 x i32>
   store <4 x i32> %t, <4 x i32> *%D, align 4
   ret void

 }
 
 
 define arm_aapcs_vfpcc void @mul_i32(<4 x i32> *%A, <4 x i32> *%B, i64 %C, <4 x i32> *%D) {
 ; CHECK-LABEL: mul_i32:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .save {r4, r5, r6, r7, lr}
 ; CHECK-NEXT:    push {r4, r5, r6, r7, lr}
-; CHECK-NEXT:    vldrw.u32 q0, [r1]
-; CHECK-NEXT:    vldrw.u32 q1, [r0]
+; CHECK-NEXT:    vldrw.u32 q1, [r1]
+; CHECK-NEXT:    vldrw.u32 q0, [r0]
+; CHECK-NEXT:    vmov r0, s4
+; CHECK-NEXT:    vmov r1, s0
+; CHECK-NEXT:    vmov.f32 s8, s4
+; CHECK-NEXT:    vmov.f32 s10, s5
+; CHECK-NEXT:    smull r12, r3, r1, r0
+; CHECK-NEXT:    vmov r0, s10
 ; CHECK-NEXT:    vmov.f32 s8, s0
-; CHECK-NEXT:    vmov.f32 s12, s4
 ; CHECK-NEXT:    vmov.f32 s10, s1
-; CHECK-NEXT:    vmov.f32 s14, s5
-; CHECK-NEXT:    vmov r0, s8
-; CHECK-NEXT:    vmov r1, s12
-; CHECK-NEXT:    vmov r5, s10
-; CHECK-NEXT:    vmov r4, s14
-; CHECK-NEXT:    vmov.f32 s8, s2
-; CHECK-NEXT:    vmov.f32 s10, s3
-; CHECK-NEXT:    vmov.f32 s0, s6
-; CHECK-NEXT:    vmov.f32 s2, s7
-; CHECK-NEXT:    vmullb.s32 q1, q0, q2
-; CHECK-NEXT:    vmov r7, s7
-; CHECK-NEXT:    vmov r6, s4
-; CHECK-NEXT:    smull r0, r3, r1, r0
+; CHECK-NEXT:    asrl r12, r3, r2
 ; CHECK-NEXT:    ldr r1, [sp, #20]
-; CHECK-NEXT:    asrl r0, r3, r2
-; CHECK-NEXT:    smull r12, r5, r4, r5
-; CHECK-NEXT:    vmov r4, s6
+; CHECK-NEXT:    vmov r5, s10
+; CHECK-NEXT:    vmov.f32 s8, s6
+; CHECK-NEXT:    vmov.f32 s10, s7
+; CHECK-NEXT:    vmov.f32 s4, s2
+; CHECK-NEXT:    vmov.f32 s6, s3
+; CHECK-NEXT:    vmullb.s32 q0, q1, q2
+; CHECK-NEXT:    vmov r7, s3
+; CHECK-NEXT:    vmov r4, s2
 ; CHECK-NEXT:    asrl r4, r7, r2
-; CHECK-NEXT:    vmov r7, s5
+; CHECK-NEXT:    vmov r6, s0
+; CHECK-NEXT:    vmov r7, s1
 ; CHECK-NEXT:    asrl r6, r7, r2
-; CHECK-NEXT:    asrl r12, r5, r2
-; CHECK-NEXT:    vmov q0[2], q0[0], r0, r6
-; CHECK-NEXT:    vmov q0[3], q0[1], r12, r4
+; CHECK-NEXT:    vmov q0[2], q0[0], r12, r6
+; CHECK-NEXT:    smull r0, r5, r5, r0
+; CHECK-NEXT:    asrl r0, r5, r2
+; CHECK-NEXT:    vmov q0[3], q0[1], r0, r4
 ; CHECK-NEXT:    vstrw.32 q0, [r1]
 ; CHECK-NEXT:    pop {r4, r5, r6, r7, pc}
 entry:
   %a = load <4 x i32>, <4 x i32> *%A, align 4
   %b = load <4 x i32>, <4 x i32> *%B, align 4
   %i = insertelement <4 x i64> undef, i64 %C, i32 0
   %c = shufflevector <4 x i64> %i, <4 x i64> undef, <4 x i32> zeroinitializer
   %sa = sext <4 x i32> %a to <4 x i64>

llvm/test/CodeGen/Thumb2/mve-laneinterleaving.ll

entry:
   ret <32 x i8> %t
 }
 
 define arm_aapcs_vfpcc <4 x i32> @ext_add_ashr_trunc_i32(<4 x i32> %a, <4 x i32> %b) {
 ; CHECK-LABEL: ext_add_ashr_trunc_i32:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .save {r4, lr}
 ; CHECK-NEXT:    push {r4, lr}
-; CHECK-NEXT:    .vsave {d8, d9}
-; CHECK-NEXT:    vpush {d8, d9}
-; CHECK-NEXT:    vmov.f32 s16, s6
-; CHECK-NEXT:    vmov.i64 q3, #0xffffffff
-; CHECK-NEXT:    vmov.f32 s18, s7
 ; CHECK-NEXT:    vmov.f32 s8, s2
-; CHECK-NEXT:    vand q4, q4, q3
 ; CHECK-NEXT:    vmov.f32 s10, s3
+; CHECK-NEXT:    vmov.f32 s12, s6
 ; CHECK-NEXT:    vmov r0, s8
-; CHECK-NEXT:    vmov r3, s16
-; CHECK-NEXT:    vmov r2, s17
-; CHECK-NEXT:    vmov.f32 s6, s5
-; CHECK-NEXT:    vand q3, q1, q3
+; CHECK-NEXT:    vmov.f32 s14, s7
+; CHECK-NEXT:    vmov r2, s12
 ; CHECK-NEXT:    vmov.f32 s2, s1
-; CHECK-NEXT:    vmov lr, s19
+; CHECK-NEXT:    vmov r3, s14
 ; CHECK-NEXT:    asrs r1, r0, #31
-; CHECK-NEXT:    adds r0, r0, r3
-; CHECK-NEXT:    adcs r1, r2
-; CHECK-NEXT:    vmov r2, s12
-; CHECK-NEXT:    lsrl r0, r1, #1
-; CHECK-NEXT:    vmov r3, s13
-; CHECK-NEXT:    vmov r1, s0
-; CHECK-NEXT:    adds r2, r2, r1
-; CHECK-NEXT:    asr.w r12, r1, #31
-; CHECK-NEXT:    adc.w r1, r12, r3
+; CHECK-NEXT:    adds r2, r2, r0
+; CHECK-NEXT:    vmov r0, s10
+; CHECK-NEXT:    adc r1, r1, #0
 ; CHECK-NEXT:    lsrl r2, r1, #1
-; CHECK-NEXT:    vmov r3, s18
-; CHECK-NEXT:    vmov r1, s10
-; CHECK-NEXT:    vmov q1[2], q1[0], r2, r0
-; CHECK-NEXT:    vmov r0, s2
-; CHECK-NEXT:    vmov r2, s15
-; CHECK-NEXT:    adds r4, r1, r3
-; CHECK-NEXT:    vmov r3, s14
-; CHECK-NEXT:    asr.w r12, r1, #31
-; CHECK-NEXT:    adc.w r1, r12, lr
+; CHECK-NEXT:    adds r4, r0, r3
+; CHECK-NEXT:    vmov r3, s0
+; CHECK-NEXT:    asr.w r1, r0, #31
+; CHECK-NEXT:    vmov r0, s4
+; CHECK-NEXT:    vmov.f32 s6, s5
+; CHECK-NEXT:    adc r1, r1, #0
 ; CHECK-NEXT:    lsrl r4, r1, #1
-; CHECK-NEXT:    asrs r1, r0, #31
 ; CHECK-NEXT:    adds r0, r0, r3
-; CHECK-NEXT:    adcs r1, r2
+; CHECK-NEXT:    asr.w r12, r3, #31
+; CHECK-NEXT:    adc r3, r12, #0
+; CHECK-NEXT:    lsrl r0, r3, #1
+; CHECK-NEXT:    vmov q2[2], q2[0], r0, r2
+; CHECK-NEXT:    vmov r0, s2
+; CHECK-NEXT:    vmov r2, s6
+; CHECK-NEXT:    asrs r1, r0, #31
+; CHECK-NEXT:    adds r0, r0, r2
+; CHECK-NEXT:    adc r1, r1, #0
 ; CHECK-NEXT:    lsrl r0, r1, #1
-; CHECK-NEXT:    vmov q1[3], q1[1], r0, r4
-; CHECK-NEXT:    vmov q0, q1
-; CHECK-NEXT:    vpop {d8, d9}
+; CHECK-NEXT:    vmov q2[3], q2[1], r0, r4
+; CHECK-NEXT:    vmov q0, q2
 ; CHECK-NEXT:    pop {r4, pc}
 entry:
   %sa = sext <4 x i32> %a to <4 x i64>
   %sb = zext <4 x i32> %b to <4 x i64>
   %add = add <4 x i64> %sa, %sb
   %sh = ashr <4 x i64> %add, <i64 1, i64 1, i64 1, i64 1>
   %t = trunc <4 x i64> %sh to <4 x i32>
   ret <4 x i32> %t

 
 define arm_aapcs_vfpcc <4 x i32> @ext_ops_trunc_i32(<4 x i32> %a, <4 x i32> %b) {
 ; CHECK-LABEL: ext_ops_trunc_i32:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .save {r4, r5, r6, r7, r8, r9, r10, r11, lr}
 ; CHECK-NEXT:    push.w {r4, r5, r6, r7, r8, r9, r10, r11, lr}
 ; CHECK-NEXT:    .pad #4
 ; CHECK-NEXT:    sub sp, #4
-; CHECK-NEXT:    .vsave {d8, d9, d10, d11}
-; CHECK-NEXT:    vpush {d8, d9, d10, d11}
-; CHECK-NEXT:    vmov.f32 s8, s6
-; CHECK-NEXT:    vmov.i64 q4, #0xffffffff
-; CHECK-NEXT:    vmov.f32 s10, s7
+; CHECK-NEXT:    .vsave {d8, d9}
+; CHECK-NEXT:    vpush {d8, d9}
 ; CHECK-NEXT:    vmov.f32 s12, s2
-; CHECK-NEXT:    vand q2, q2, q4
+; CHECK-NEXT:    vmov.i64 q4, #0xffffffff
+; CHECK-NEXT:    vmov.f32 s8, s6
 ; CHECK-NEXT:    vmov.f32 s14, s3
-; CHECK-NEXT:    vmov r10, s12
-; CHECK-NEXT:    vmov r4, s8
-; CHECK-NEXT:    vmov r1, s9
-; CHECK-NEXT:    vmov.f32 s6, s5
-; CHECK-NEXT:    vand q1, q1, q4
+; CHECK-NEXT:    vmov.f32 s10, s7
+; CHECK-NEXT:    vmov r11, s12
+; CHECK-NEXT:    vmov r5, s8
 ; CHECK-NEXT:    vmov.f32 s2, s1
-; CHECK-NEXT:    vmov r9, s0
-; CHECK-NEXT:    vmov r11, s4
-; CHECK-NEXT:    vmov r7, s5
-; CHECK-NEXT:    adds.w r2, r10, r4
-; CHECK-NEXT:    asr.w r0, r10, #31
-; CHECK-NEXT:    adc.w r5, r0, r1
-; CHECK-NEXT:    asrl r2, r5, r4
-; CHECK-NEXT:    subs r6, r2, r4
-; CHECK-NEXT:    sbc.w r12, r5, r1
-; CHECK-NEXT:    adds.w r0, r9, r11
-; CHECK-NEXT:    asr.w r2, r9, #31
-; CHECK-NEXT:    adc.w r3, r2, r7
-; CHECK-NEXT:    umull r8, r5, r6, r4
-; CHECK-NEXT:    asrl r0, r3, r11
-; CHECK-NEXT:    subs.w r0, r0, r11
-; CHECK-NEXT:    mla r5, r12, r4, r5
-; CHECK-NEXT:    sbc.w r12, r3, r7
-; CHECK-NEXT:    umull r2, r1, r0, r11
-; CHECK-NEXT:    muls r0, r7, r0
-; CHECK-NEXT:    vmov r7, s14
-; CHECK-NEXT:    orr.w lr, r1, r0
-; CHECK-NEXT:    rsb.w r0, r10, #0
-; CHECK-NEXT:    lsll r8, r5, r0
-; CHECK-NEXT:    vmov r0, s10
-; CHECK-NEXT:    vmov r1, s11
-; CHECK-NEXT:    lsll r8, r5, r4
-; CHECK-NEXT:    vmov r5, s6
-; CHECK-NEXT:    eor.w r4, r4, r10
-; CHECK-NEXT:    orr.w r4, r4, r10, asr #31
-; CHECK-NEXT:    asrs r3, r7, #31
-; CHECK-NEXT:    adds r6, r7, r0
-; CHECK-NEXT:    adcs r3, r1
-; CHECK-NEXT:    asrl r6, r3, r0
-; CHECK-NEXT:    subs r6, r6, r0
-; CHECK-NEXT:    sbc.w r1, r3, r1
-; CHECK-NEXT:    umull r6, r3, r6, r0
-; CHECK-NEXT:    mla r1, r1, r0, r3
-; CHECK-NEXT:    rsbs r3, r7, #0
-; CHECK-NEXT:    lsll r6, r1, r3
-; CHECK-NEXT:    lsll r6, r1, r0
-; CHECK-NEXT:    eors r0, r7
-; CHECK-NEXT:    vmov q3[2], q3[0], r8, r6
-; CHECK-NEXT:    vmov r6, s2
-; CHECK-NEXT:    vmov r1, s7
-; CHECK-NEXT:    orr.w r0, r0, r7, asr #31
-; CHECK-NEXT:    adds.w r8, r6, r5
-; CHECK-NEXT:    eor.w r7, r6, r5
-; CHECK-NEXT:    asr.w r3, r6, #31
-; CHECK-NEXT:    orr.w r7, r7, r6, asr #31
-; CHECK-NEXT:    adcs r3, r1
-; CHECK-NEXT:    asrl r8, r3, r5
-; CHECK-NEXT:    subs.w r8, r8, r5
-; CHECK-NEXT:    sbcs r3, r1
-; CHECK-NEXT:    cmp r0, #0
-; CHECK-NEXT:    cset r0, eq
-; CHECK-NEXT:    mul r1, r8, r1
-; CHECK-NEXT:    cmp r0, #0
-; CHECK-NEXT:    csetm r0, ne
+; CHECK-NEXT:    vmov.f32 s6, s5
+; CHECK-NEXT:    vmov r10, s2
+; CHECK-NEXT:    adds.w r2, r11, r5
+; CHECK-NEXT:    asr.w r0, r11, #31
+; CHECK-NEXT:    adc r1, r0, #0
+; CHECK-NEXT:    asrl r2, r1, r5
+; CHECK-NEXT:    subs r0, r2, r5
+; CHECK-NEXT:    vmov r2, s14
+; CHECK-NEXT:    sbc lr, r1, #0
+; CHECK-NEXT:    vmov r1, s6
+; CHECK-NEXT:    umull r8, r0, r0, r5
+; CHECK-NEXT:    asr.w r6, r10, #31
+; CHECK-NEXT:    adds.w r4, r10, r1
+; CHECK-NEXT:    asr.w r3, r2, #31
+; CHECK-NEXT:    adc r7, r6, #0
+; CHECK-NEXT:    asrl r4, r7, r1
+; CHECK-NEXT:    subs r6, r4, r1
+; CHECK-NEXT:    sbc r7, r7, #0
+; CHECK-NEXT:    umull r12, r4, r6, r1
+; CHECK-NEXT:    vmov r6, s10
+; CHECK-NEXT:    vand q2, q2, q4
+; CHECK-NEXT:    mla r9, r7, r1, r4
+; CHECK-NEXT:    eor.w r7, r11, r5
+; CHECK-NEXT:    orr.w r7, r7, r11, asr #31
+; CHECK-NEXT:    eor.w r4, r2, r6
+; CHECK-NEXT:    orr.w r4, r4, r2, asr #31
 ; CHECK-NEXT:    cmp r4, #0
 ; CHECK-NEXT:    cset r4, eq
 ; CHECK-NEXT:    cmp r4, #0
 ; CHECK-NEXT:    csetm r4, ne
-; CHECK-NEXT:    vmov.32 q4[1], r4
-; CHECK-NEXT:    vmov q4[2], q4[0], r4, r0
-; CHECK-NEXT:    umull r4, r0, r8, r5
-; CHECK-NEXT:    orrs r0, r1
-; CHECK-NEXT:    eor.w r1, r9, r11
-; CHECK-NEXT:    orr.w r1, r1, r9, asr #31
-; CHECK-NEXT:    cmp r1, #0
-; CHECK-NEXT:    cset r1, eq
-; CHECK-NEXT:    cmp r1, #0
-; CHECK-NEXT:    csetm r1, ne
 ; CHECK-NEXT:    cmp r7, #0
 ; CHECK-NEXT:    cset r7, eq
-; CHECK-NEXT:    vmov.32 q0[1], r1
 ; CHECK-NEXT:    cmp r7, #0
 ; CHECK-NEXT:    csetm r7, ne
-; CHECK-NEXT:    vmov q0[2], q0[0], r1, r7
-; CHECK-NEXT:    mla r7, r3, r5, r0
-; CHECK-NEXT:    rsbs r1, r6, #0
-; CHECK-NEXT:    vbic q5, q1, q0
-; CHECK-NEXT:    mla r3, r12, r11, lr
-; CHECK-NEXT:    rsb.w r0, r9, #0
-; CHECK-NEXT:    lsll r4, r7, r1
-; CHECK-NEXT:    vbic q1, q2, q4
-; CHECK-NEXT:    lsll r2, r3, r0
-; CHECK-NEXT:    vand q2, q3, q4
-; CHECK-NEXT:    lsll r4, r7, r5
-; CHECK-NEXT:    lsll r2, r3, r11
-; CHECK-NEXT:    vorr q1, q2, q1
-; CHECK-NEXT:    vmov q2[2], q2[0], r2, r4
-; CHECK-NEXT:    vand q0, q2, q0
-; CHECK-NEXT:    vorr q0, q0, q5
+; CHECK-NEXT:    vmov.32 q3[1], r7
+; CHECK-NEXT:    vmov q3[2], q3[0], r7, r4
+; CHECK-NEXT:    adds r4, r2, r6
+; CHECK-NEXT:    adc r3, r3, #0
+; CHECK-NEXT:    mla r7, lr, r5, r0
+; CHECK-NEXT:    asrl r4, r3, r6
+; CHECK-NEXT:    subs r4, r4, r6
+; CHECK-NEXT:    sbc r3, r3, #0
+; CHECK-NEXT:    umull r4, r0, r4, r6
+; CHECK-NEXT:    mla r3, r3, r6, r0
+; CHECK-NEXT:    rsbs r0, r2, #0
+; CHECK-NEXT:    lsll r4, r3, r0
+; CHECK-NEXT:    rsb.w r0, r11, #0
+; CHECK-NEXT:    lsll r8, r7, r0
+; CHECK-NEXT:    lsll r4, r3, r6
+; CHECK-NEXT:    lsll r8, r7, r5
+; CHECK-NEXT:    vmov r5, s0
+; CHECK-NEXT:    vmov r3, s4
+; CHECK-NEXT:    vand q1, q1, q4
+; CHECK-NEXT:    vbic q4, q2, q3
+; CHECK-NEXT:    adds r2, r5, r3
+; CHECK-NEXT:    asr.w r0, r5, #31
+; CHECK-NEXT:    adc r7, r0, #0
+; CHECK-NEXT:    asrl r2, r7, r3
+; CHECK-NEXT:    subs r0, r2, r3
+; CHECK-NEXT:    sbc r2, r7, #0
+; CHECK-NEXT:    umull r6, r0, r0, r3
+; CHECK-NEXT:    mla r7, r2, r3, r0
+; CHECK-NEXT:    eor.w r0, r10, r1
+; CHECK-NEXT:    orr.w r0, r0, r10, asr #31
+; CHECK-NEXT:    cmp r0, #0
+; CHECK-NEXT:    eor.w r2, r5, r3
+; CHECK-NEXT:    cset r0, eq
+; CHECK-NEXT:    orr.w r2, r2, r5, asr #31
+; CHECK-NEXT:    cmp r0, #0
+; CHECK-NEXT:    csetm r0, ne
+; CHECK-NEXT:    cmp r2, #0
+; CHECK-NEXT:    cset r2, eq
+; CHECK-NEXT:    cmp r2, #0
+; CHECK-NEXT:    csetm r2, ne
+; CHECK-NEXT:    vmov.32 q0[1], r2
+; CHECK-NEXT:    vmov q0[2], q0[0], r2, r0
+; CHECK-NEXT:    rsb.w r0, r10, #0
+; CHECK-NEXT:    lsll r12, r9, r0
+; CHECK-NEXT:    rsbs r0, r5, #0
+; CHECK-NEXT:    lsll r6, r7, r0
+; CHECK-NEXT:    vbic q2, q1, q0
+; CHECK-NEXT:    vmov q1[2], q1[0], r8, r4
+; CHECK-NEXT:    lsll r12, r9, r1
+; CHECK-NEXT:    lsll r6, r7, r3
+; CHECK-NEXT:    vand q1, q1, q3
+; CHECK-NEXT:    vmov q3[2], q3[0], r6, r12
+; CHECK-NEXT:    vorr q1, q1, q4
+; CHECK-NEXT:    vand q0, q3, q0
+; CHECK-NEXT:    vorr q0, q0, q2
 ; CHECK-NEXT:    vmov.f32 s1, s2
 ; CHECK-NEXT:    vmov.f32 s2, s4
 ; CHECK-NEXT:    vmov.f32 s3, s6
-; CHECK-NEXT:    vpop {d8, d9, d10, d11}
+; CHECK-NEXT:    vpop {d8, d9}
 ; CHECK-NEXT:    add sp, #4
 ; CHECK-NEXT:    pop.w {r4, r5, r6, r7, r8, r9, r10, r11, pc}
 entry:
   %sa = sext <4 x i32> %a to <4 x i64>
   %sb = zext <4 x i32> %b to <4 x i64>
   %add = add <4 x i64> %sa, %sb
   %ashr = ashr <4 x i64> %add, %sb
   %sub = sub <4 x i64> %ashr, %sb

llvm/test/CodeGen/Thumb2/mve-sext-masked-load.ll

entry:
   ret <4 x float> %0
 }
 
 define arm_aapcs_vfpcc <4 x double> @foo_v4i32(<4 x i32>* nocapture readonly %pSrc, i32 %blockSize, <4 x i32> %a) {
 ; CHECK-LABEL: foo_v4i32:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .save {r7, lr}
 ; CHECK-NEXT:    push {r7, lr}
-; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13}
-; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13}
+; CHECK-NEXT:    .vsave {d8, d9, d10, d11}
+; CHECK-NEXT:    vpush {d8, d9, d10, d11}
 ; CHECK-NEXT:    vpt.s32 lt, q0, zr
 ; CHECK-NEXT:    vldrwt.u32 q5, [r0]
-; CHECK-NEXT:    vmov.f64 d8, d10
-; CHECK-NEXT:    vmov.f32 s18, s21
+; CHECK-NEXT:    vmov.f64 d8, d11
+; CHECK-NEXT:    vmov.f32 s18, s23
 ; CHECK-NEXT:    vmov r0, s18
 ; CHECK-NEXT:    asrs r1, r0, #31
 ; CHECK-NEXT:    bl __aeabi_l2d
 ; CHECK-NEXT:    vmov r2, s16
 ; CHECK-NEXT:    vmov d9, r0, r1
 ; CHECK-NEXT:    asrs r3, r2, #31
 ; CHECK-NEXT:    mov r0, r2
 ; CHECK-NEXT:    mov r1, r3
 ; CHECK-NEXT:    bl __aeabi_l2d
-; CHECK-NEXT:    vmov.f64 d12, d11
-; CHECK-NEXT:    vmov.f32 s26, s23
+; CHECK-NEXT:    vmov.f32 s22, s21
 ; CHECK-NEXT:    vmov d8, r0, r1
-; CHECK-NEXT:    vmov r2, s26
+; CHECK-NEXT:    vmov r2, s22
 ; CHECK-NEXT:    asrs r3, r2, #31
 ; CHECK-NEXT:    mov r0, r2
 ; CHECK-NEXT:    mov r1, r3
 ; CHECK-NEXT:    bl __aeabi_l2d
-; CHECK-NEXT:    vmov r2, s24
+; CHECK-NEXT:    vmov r2, s20
 ; CHECK-NEXT:    vmov d11, r0, r1
 ; CHECK-NEXT:    asrs r3, r2, #31
 ; CHECK-NEXT:    mov r0, r2
 ; CHECK-NEXT:    mov r1, r3
 ; CHECK-NEXT:    bl __aeabi_l2d
 ; CHECK-NEXT:    vmov d10, r0, r1
-; CHECK-NEXT:    vmov q0, q4
-; CHECK-NEXT:    vmov q1, q5
-; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13}
+; CHECK-NEXT:    vmov q1, q4
+; CHECK-NEXT:    vmov q0, q5
+; CHECK-NEXT:    vpop {d8, d9, d10, d11}
 ; CHECK-NEXT:    pop {r7, pc}
 entry:
   %active.lane.mask = icmp slt <4 x i32> %a, zeroinitializer
   %wide.masked.load = call <4 x i32> @llvm.masked.load.v4i32.p0v4i32(<4 x i32>* %pSrc, i32 4, <4 x i1> %active.lane.mask, <4 x i32> undef)
   %0 = sitofp <4 x i32> %wide.masked.load to <4 x double>
   ret <4 x double> %0
 }
 
 declare <4 x i16> @llvm.masked.load.v4i16.p0v4i16(<4 x i16>*, i32 immarg, <4 x i1>, <4 x i16>)
 
 declare <8 x i8> @llvm.masked.load.v8i8.p0v8i8(<8 x i8>*, i32 immarg, <8 x i1>, <8 x i8>)
 
 declare <4 x i8> @llvm.masked.load.v4i8.p0v4i8(<4 x i8>*, i32 immarg, <4 x i1>, <4 x i8>)
 
 declare <4 x i32> @llvm.masked.load.v4i32.p0v4i32(<4 x i32>*, i32 immarg, <4 x i1>, <4 x i32>)

llvm/test/CodeGen/Thumb2/mve-shuffle.ll

entry:
   %s2 = shufflevector <16 x i16> %src, <16 x i16> undef, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
   %r = add <8 x i16> %s1, %s2
   ret <8 x i16> %r
 }
 
 define arm_aapcs_vfpcc <8 x i16> @shuffle3step_i16(<32 x i16> %src) {
 ; CHECK-LABEL: shuffle3step_i16:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .vsave {d8, d9, d10, d11}
-; CHECK-NEXT:    vpush {d8, d9, d10, d11}
+; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13}
+; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13}
 ; CHECK-NEXT:    vmovx.f16 s16, s2
 ; CHECK-NEXT:    vmov.f32 s12, s1
 ; CHECK-NEXT:    vins.f16 s12, s16
 ; CHECK-NEXT:    vmovx.f16 s16, s5
 ; CHECK-NEXT:    vmov.f32 s13, s4
-; CHECK-NEXT:    vmovx.f16 s20, s11
+; CHECK-NEXT:    vmov.u16 r0, q1[5]
 ; CHECK-NEXT:    vins.f16 s13, s16
-; CHECK-NEXT:    vmov.f32 s19, s10
-; CHECK-NEXT:    vins.f16 s19, s20
+; CHECK-NEXT:    vmovx.f16 s16, s11
+; CHECK-NEXT:    vmov.f32 s23, s10
+; CHECK-NEXT:    vins.f16 s23, s16
 ; CHECK-NEXT:    vmov.f32 s14, s7
-; CHECK-NEXT:    vmov.f32 s18, s8
-; CHECK-NEXT:    vmov.u16 r0, q1[5]
+; CHECK-NEXT:    vmov.f32 s22, s8
+; CHECK-NEXT:    vmov q4, q5
+; CHECK-NEXT:    vmovnb.i32 q4, q3
+; CHECK-NEXT:    vmov.f32 s14, s18
+; CHECK-NEXT:    vmovx.f16 s16, s0
+; CHECK-NEXT:    vins.f16 s16, s2
+; CHECK-NEXT:    vmov.f32 s15, s23
+; CHECK-NEXT:    vmovx.f16 s17, s3
+; CHECK-NEXT:    vins.f16 s17, s5
+; CHECK-NEXT:    vins.f16 s18, s8
 ; CHECK-NEXT:    vmov q5, q4
-; CHECK-NEXT:    vmovnb.i32 q5, q3
-; CHECK-NEXT:    vmov.f32 s14, s22
+; CHECK-NEXT:    vmovx.f16 s19, s9
+; CHECK-NEXT:    vins.f16 s19, s11
+; CHECK-NEXT:    vmov.16 q5[4], r0
+; CHECK-NEXT:    vmov q6, q4
+; CHECK-NEXT:    vmovnb.i32 q6, q5
 ; CHECK-NEXT:    vmovx.f16 s20, s1
-; CHECK-NEXT:    vmov.f32 s15, s19
-; CHECK-NEXT:    vmov.f32 s16, s0
-; CHECK-NEXT:    vins.f16 s16, s20
+; CHECK-NEXT:    vins.f16 s0, s20
 ; CHECK-NEXT:    vmovx.f16 s20, s4
-; CHECK-NEXT:    vmov.f32 s17, s3
-; CHECK-NEXT:    vins.f16 s17, s20
+; CHECK-NEXT:    vins.f16 s3, s20
 ; CHECK-NEXT:    vmovx.f16 s20, s7
-; CHECK-NEXT:    vmov.f32 s18, s6
-; CHECK-NEXT:    vins.f16 s18, s20
-; CHECK-NEXT:    vmovx.f16 s20, s10
-; CHECK-NEXT:    vmov.f32 s19, s9
-; CHECK-NEXT:    vins.f16 s19, s20
-; CHECK-NEXT:    vmovx.f16 s20, s0
-; CHECK-NEXT:    vins.f16 s20, s2
-; CHECK-NEXT:    vmovx.f16 s21, s3
-; CHECK-NEXT:    vins.f16 s2, s8
-; CHECK-NEXT:    vmovx.f16 s3, s9
-; CHECK-NEXT:    vins.f16 s21, s5
-; CHECK-NEXT:    vins.f16 s3, s11
-; CHECK-NEXT:    vmov.16 q5[4], r0
-; CHECK-NEXT:    vmov q1, q0
-; CHECK-NEXT:    vmovnb.i32 q1, q5
-; CHECK-NEXT:    vmov.f32 s22, s6
-; CHECK-NEXT:    vmov.f32 s23, s3
-; CHECK-NEXT:    vadd.i16 q0, q4, q5
+; CHECK-NEXT:    vmov.f32 s1, s3
+; CHECK-NEXT:    vins.f16 s6, s20
+; CHECK-NEXT:    vmov.f32 s2, s6
+; CHECK-NEXT:    vmovx.f16 s4, s10
+; CHECK-NEXT:    vins.f16 s9, s4
+; CHECK-NEXT:    vmov.f32 s18, s26
+; CHECK-NEXT:    vmov.f32 s3, s9
+; CHECK-NEXT:    vadd.i16 q0, q0, q4
 ; CHECK-NEXT:    vadd.i16 q0, q0, q3
-; CHECK-NEXT:    vpop {d8, d9, d10, d11}
+; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13}
 ; CHECK-NEXT:    bx lr
 entry:
   %s1 = shufflevector <32 x i16> %src, <32 x i16> undef, <8 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21>
   %s2 = shufflevector <32 x i16> %src, <32 x i16> undef, <8 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22>
   %s3 = shufflevector <32 x i16> %src, <32 x i16> undef, <8 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23>
   %a = add <8 x i16> %s1, %s2
   %r = add <8 x i16> %a, %s3
   ret <8 x i16> %r
entry:
   %s2 = shufflevector <32 x i8> %src, <32 x i8> undef, <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31>
   %r = add <16 x i8> %s1, %s2
   ret <16 x i8> %r
 }
 
 define arm_aapcs_vfpcc <16 x i8> @shuffle3step_i8(<64 x i8> %src) {
 ; CHECK-LABEL: shuffle3step_i8:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13}
-; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13}
+; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13, d14, d15}
+; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13, d14, d15}
 ; CHECK-NEXT:    vmov.u8 r0, q0[1]
 ; CHECK-NEXT:    vmov.8 q3[0], r0
 ; CHECK-NEXT:    vmov.u8 r0, q0[4]
 ; CHECK-NEXT:    vmov.8 q3[1], r0
 ; CHECK-NEXT:    vmov.u8 r0, q0[7]
 ; CHECK-NEXT:    vmov.8 q3[2], r0
 ; CHECK-NEXT:    vmov.u8 r0, q0[10]
 ; CHECK-NEXT:    vmov.8 q3[3], r0
 ; CHECK-NEXT:    vmov.u8 r0, q0[13]
-; CHECK-NEXT:    vmov.8 q3[4], r0
+; CHECK-NEXT:    vmov q4, q3
+; CHECK-NEXT:    vmov.8 q4[4], r0
 ; CHECK-NEXT:    vmov.u8 r0, q1[0]
-; CHECK-NEXT:    vmov.8 q3[5], r0
+; CHECK-NEXT:    vmov.8 q4[5], r0
 ; CHECK-NEXT:    vmov.u8 r0, q1[3]
-; CHECK-NEXT:    vmov.8 q3[6], r0
+; CHECK-NEXT:    vmov.8 q4[6], r0
 ; CHECK-NEXT:    vmov.u8 r0, q1[6]
-; CHECK-NEXT:    vmov.8 q3[7], r0
+; CHECK-NEXT:    vmov.8 q4[7], r0
 ; CHECK-NEXT:    vmov.u8 r0, q1[9]
-; CHECK-NEXT:    vmov.8 q3[8], r0
+; CHECK-NEXT:    vmov q5, q4
+; CHECK-NEXT:    vmov.f32 s13, s17
+; CHECK-NEXT:    vmov.8 q5[8], r0
 ; CHECK-NEXT:    vmov.u8 r0, q1[12]
-; CHECK-NEXT:    vmov.8 q3[9], r0
+; CHECK-NEXT:    vmov.8 q5[9], r0
 ; CHECK-NEXT:    vmov.u8 r0, q1[15]
-; CHECK-NEXT:    vmov.8 q3[10], r0
+; CHECK-NEXT:    vmov.8 q5[10], r0
 ; CHECK-NEXT:    vmov.u8 r0, q2[5]
-; CHECK-NEXT:    vmov.8 q4[12], r0
+; CHECK-NEXT:    vmov.8 q6[12], r0
 ; CHECK-NEXT:    vmov.u8 r0, q2[8]
-; CHECK-NEXT:    vmov.8 q4[13], r0
+; CHECK-NEXT:    vmov.8 q6[13], r0
 ; CHECK-NEXT:    vmov.u8 r0, q2[11]
-; CHECK-NEXT:    vmov.8 q4[14], r0
+; CHECK-NEXT:    vmov.8 q6[14], r0
 ; CHECK-NEXT:    vmov.u8 r0, q2[14]
-; CHECK-NEXT:    vmov.8 q4[15], r0
+; CHECK-NEXT:    vmov.8 q6[15], r0
 ; CHECK-NEXT:    vmov.u8 r0, q2[2]
-; CHECK-NEXT:    vmov q5, q3
 ; CHECK-NEXT:    vmov.8 q5[11], r0
 ; CHECK-NEXT:    vmov.u8 r0, q0[0]
-; CHECK-NEXT:    vmov.f32 s14, s22
-; CHECK-NEXT:    vmov.f32 s15, s19
 ; CHECK-NEXT:    vmov.8 q4[0], r0
 ; CHECK-NEXT:    vmov.u8 r0, q0[3]
 ; CHECK-NEXT:    vmov.8 q4[1], r0
 ; CHECK-NEXT:    vmov.u8 r0, q0[6]
 ; CHECK-NEXT:    vmov.8 q4[2], r0
 ; CHECK-NEXT:    vmov.u8 r0, q0[9]
 ; CHECK-NEXT:    vmov.8 q4[3], r0
+; CHECK-NEXT:    vmov.f32 s14, s22
 ; CHECK-NEXT:    vmov.u8 r0, q0[12]
-; CHECK-NEXT:    vmov.8 q4[4], r0
+; CHECK-NEXT:    vmov q5, q4
+; CHECK-NEXT:    vmov.8 q5[4], r0
 ; CHECK-NEXT:    vmov.u8 r0, q0[15]
-; CHECK-NEXT:    vmov.8 q4[5], r0
+; CHECK-NEXT:    vmov.8 q5[5], r0
 ; CHECK-NEXT:    vmov.u8 r0, q1[2]
-; CHECK-NEXT:    vmov.8 q4[6], r0
+; CHECK-NEXT:    vmov.8 q5[6], r0
 ; CHECK-NEXT:    vmov.u8 r0, q1[5]
-; CHECK-NEXT:    vmov.8 q4[7], r0
+; CHECK-NEXT:    vmov.8 q5[7], r0
+; CHECK-NEXT:    vmov.f32 s15, s27
 ; CHECK-NEXT:    vmov.u8 r0, q1[8]
-; CHECK-NEXT:    vmov.8 q4[8], r0
+; CHECK-NEXT:    vmov q6, q5
+; CHECK-NEXT:    vmov.8 q6[8], r0
 ; CHECK-NEXT:    vmov.u8 r0, q1[11]
-; CHECK-NEXT:    vmov.8 q4[9], r0
+; CHECK-NEXT:    vmov.8 q6[9], r0
 ; CHECK-NEXT:    vmov.u8 r0, q1[14]
-; CHECK-NEXT:    vmov.8 q4[10], r0
+; CHECK-NEXT:    vmov.8 q6[10], r0
 ; CHECK-NEXT:    vmov.u8 r0, q2[4]
-; CHECK-NEXT:    vmov.8 q5[12], r0
+; CHECK-NEXT:    vmov.8 q7[12], r0
 ; CHECK-NEXT:    vmov.u8 r0, q2[7]
-; CHECK-NEXT:    vmov.8 q5[13], r0
+; CHECK-NEXT:    vmov.8 q7[13], r0
 ; CHECK-NEXT:    vmov.u8 r0, q2[10]
-; CHECK-NEXT:    vmov.8 q5[14], r0
+; CHECK-NEXT:    vmov.8 q7[14], r0
 ; CHECK-NEXT:    vmov.u8 r0, q2[13]
-; CHECK-NEXT:    vmov.8 q5[15], r0
+; CHECK-NEXT:    vmov.8 q7[15], r0
 ; CHECK-NEXT:    vmov.u8 r0, q2[1]
-; CHECK-NEXT:    vmov q6, q4
 ; CHECK-NEXT:    vmov.8 q6[11], r0
-; CHECK-NEXT:    vmov.u8 r0, q0[2]
+; CHECK-NEXT:    vmov.f32 s17, s21
 ; CHECK-NEXT:    vmov.f32 s18, s26
-; CHECK-NEXT:    vmov.f32 s19, s23
+; CHECK-NEXT:    vmov.u8 r0, q0[2]
+; CHECK-NEXT:    vmov.f32 s19, s31
 ; CHECK-NEXT:    vadd.i8 q3, q4, q3
 ; CHECK-NEXT:    vmov.8 q4[0], r0
 ; CHECK-NEXT:    vmov.u8 r0, q0[5]
 ; CHECK-NEXT:    vmov.8 q4[1], r0
 ; CHECK-NEXT:    vmov.u8 r0, q0[8]
 ; CHECK-NEXT:    vmov.8 q4[2], r0
 ; CHECK-NEXT:    vmov.u8 r0, q0[11]
 ; CHECK-NEXT:    vmov.8 q4[3], r0
 ; CHECK-NEXT:    vmov.u8 r0, q0[14]
-; CHECK-NEXT:    vmov.8 q4[4], r0
+; CHECK-NEXT:    vmov q0, q4
+; CHECK-NEXT:    vmov.8 q0[4], r0
 ; CHECK-NEXT:    vmov.u8 r0, q1[1]
-; CHECK-NEXT:    vmov.8 q4[5], r0
+; CHECK-NEXT:    vmov.8 q0[5], r0
 ; CHECK-NEXT:    vmov.u8 r0, q1[4]
-; CHECK-NEXT:    vmov.8 q4[6], r0
+; CHECK-NEXT:    vmov.8 q0[6], r0
 ; CHECK-NEXT:    vmov.u8 r0, q2[6]
-; CHECK-NEXT:    vmov.8 q0[12], r0
+; CHECK-NEXT:    vmov.8 q5[12], r0
 ; CHECK-NEXT:    vmov.u8 r0, q2[9]
-; CHECK-NEXT:    vmov.8 q0[13], r0
+; CHECK-NEXT:    vmov.8 q5[13], r0
 ; CHECK-NEXT:    vmov.u8 r0, q2[12]
-; CHECK-NEXT:    vmov.8 q0[14], r0
+; CHECK-NEXT:    vmov.8 q5[14], r0
 ; CHECK-NEXT:    vmov.u8 r0, q2[15]
-; CHECK-NEXT:    vmov.8 q0[15], r0
+; CHECK-NEXT:    vmov.8 q5[15], r0
 ; CHECK-NEXT:    vmov.u8 r0, q1[10]
-; CHECK-NEXT:    vmov.8 q5[8], r0
+; CHECK-NEXT:    vmov.8 q6[8], r0
 ; CHECK-NEXT:    vmov.u8 r0, q1[13]
-; CHECK-NEXT:    vmov.8 q5[9], r0
+; CHECK-NEXT:    vmov.8 q6[9], r0
 ; CHECK-NEXT:    vmov.u8 r0, q2[0]
-; CHECK-NEXT:    vmov.8 q5[10], r0
+; CHECK-NEXT:    vmov.8 q6[10], r0
 ; CHECK-NEXT:    vmov.u8 r0, q2[3]
-; CHECK-NEXT:    vmov.8 q5[11], r0
+; CHECK-NEXT:    vmov.8 q6[11], r0
 ; CHECK-NEXT:    vmov.u8 r0, q1[7]
-; CHECK-NEXT:    vmov.8 q4[7], r0
-; CHECK-NEXT:    vmov.f32 s18, s22
-; CHECK-NEXT:    vmov.f32 s19, s3
+; CHECK-NEXT:    vmov.8 q0[7], r0
+; CHECK-NEXT:    vmov.f32 s17, s1
+; CHECK-NEXT:    vmov.f32 s18, s26
+; CHECK-NEXT:    vmov.f32 s19, s23
 ; CHECK-NEXT:    vadd.i8 q0, q3, q4
-; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13}
+; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13, d14, d15}
 ; CHECK-NEXT:    bx lr
 entry:
   %s1 = shufflevector <64 x i8> %src, <64 x i8> undef, <16 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21, i32 24, i32 27, i32 30, i32 33, i32 36, i32 39, i32 42, i32 45>
   %s2 = shufflevector <64 x i8> %src, <64 x i8> undef, <16 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22, i32 25, i32 28, i32 31, i32 34, i32 37, i32 40, i32 43, i32 46>
   %s3 = shufflevector <64 x i8> %src, <64 x i8> undef, <16 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23, i32 26, i32 29, i32 32, i32 35, i32 38, i32 41, i32 44, i32 47>
   %a = add <16 x i8> %s1, %s2
   %r = add <16 x i8> %a, %s3
   ret <16 x i8> %r
 }
 
 define arm_aapcs_vfpcc <16 x i8> @shuffle4step_i8(<64 x i8> %src) {
 ; CHECK-LABEL: shuffle4step_i8:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13}
-; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13}
+; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13, d14, d15}
+; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13, d14, d15}
+; CHECK-NEXT:    .pad #48
+; CHECK-NEXT:    sub sp, #48
 ; CHECK-NEXT:    vmov.u8 r0, q0[3]
+; CHECK-NEXT:    vmov q7, q3
 ; CHECK-NEXT:    vmov.8 q4[0], r0
 ; CHECK-NEXT:    vmov.u8 r0, q0[7]
 ; CHECK-NEXT:    vmov.8 q4[1], r0
 ; CHECK-NEXT:    vmov.u8 r0, q0[11]
 ; CHECK-NEXT:    vmov.8 q4[2], r0
 ; CHECK-NEXT:    vmov.u8 r0, q0[15]
 ; CHECK-NEXT:    vmov.8 q4[3], r0
 ; CHECK-NEXT:    vmov.u8 r0, q1[3]
-; CHECK-NEXT:    vmov.8 q4[4], r0
+; CHECK-NEXT:    vmov q5, q4
+; CHECK-NEXT:    vmov q3, q2
+; CHECK-NEXT:    vmov.8 q5[4], r0
 ; CHECK-NEXT:    vmov.u8 r0, q1[7]
-; CHECK-NEXT:    vmov.8 q4[5], r0
+; CHECK-NEXT:    vmov.8 q5[5], r0
 ; CHECK-NEXT:    vmov.u8 r0, q1[11]
-; CHECK-NEXT:    vmov.8 q4[6], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[3]
-; CHECK-NEXT:    vmov.8 q5[8], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[7]
-; CHECK-NEXT:    vmov.8 q5[9], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[11]
-; CHECK-NEXT:    vmov.8 q5[10], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[15]
-; CHECK-NEXT:    vmov.8 q5[11], r0
+; CHECK-NEXT:    vmov.8 q5[6], r0
 ; CHECK-NEXT:    vmov.u8 r0, q3[3]
-; CHECK-NEXT:    vmov.8 q5[12], r0
+; CHECK-NEXT:    vmov.8 q6[8], r0
 ; CHECK-NEXT:    vmov.u8 r0, q3[7]
-; CHECK-NEXT:    vmov.8 q5[13], r0
+; CHECK-NEXT:    vmov.8 q6[9], r0
 ; CHECK-NEXT:    vmov.u8 r0, q3[11]
-; CHECK-NEXT:    vmov.8 q5[14], r0
+; CHECK-NEXT:    vmov.8 q6[10], r0
 ; CHECK-NEXT:    vmov.u8 r0, q3[15]
-; CHECK-NEXT:    vmov.8 q5[15], r0
+; CHECK-NEXT:    vmov.8 q6[11], r0
+; CHECK-NEXT:    vmov q2, q0
+; CHECK-NEXT:    vmov q0, q1
+; CHECK-NEXT:    vmov q1, q7
+; CHECK-NEXT:    vmov.u8 r0, q7[3]
+; CHECK-NEXT:    vmov q7, q6
+; CHECK-NEXT:    vmov.8 q7[12], r0
+; CHECK-NEXT:    vmov.u8 r0, q1[7]
+; CHECK-NEXT:    vmov.8 q7[13], r0
+; CHECK-NEXT:    vmov.u8 r0, q1[11]
+; CHECK-NEXT:    vmov.8 q7[14], r0
 ; CHECK-NEXT:    vmov.u8 r0, q1[15]
-; CHECK-NEXT:    vmov.8 q4[7], r0
-; CHECK-NEXT:    vmov.u8 r0, q0[2]
-; CHECK-NEXT:    vmov.f32 s18, s22
-; CHECK-NEXT:    vmov.f32 s19, s23
-; CHECK-NEXT:    vmov.8 q5[0], r0
-; CHECK-NEXT:    vmov.u8 r0, q0[6]
-; CHECK-NEXT:    vmov.8 q5[1], r0
-; CHECK-NEXT:    vmov.u8 r0, q0[10]
-; CHECK-NEXT:    vmov.8 q5[2], r0
-; CHECK-NEXT:    vmov.u8 r0, q0[14]
-; CHECK-NEXT:    vmov.8 q5[3], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[2]
-; CHECK-NEXT:    vmov.8 q5[4], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[6]
-; CHECK-NEXT:    vmov.8 q5[5], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[10]
-; CHECK-NEXT:    vmov.8 q5[6], r0
+; CHECK-NEXT:    vmov.8 q7[15], r0
+; CHECK-NEXT:    vmov.u8 r0, q0[15]
+; CHECK-NEXT:    vmov.8 q5[7], r0
 ; CHECK-NEXT:    vmov.u8 r0, q2[2]
-; CHECK-NEXT:    vmov.8 q6[8], r0
+; CHECK-NEXT:    vmov.f32 s17, s21
+; CHECK-NEXT:    vmov.8 q5[0], r0
 ; CHECK-NEXT:    vmov.u8 r0, q2[6]
-; CHECK-NEXT:    vmov.8 q6[9], r0
+; CHECK-NEXT:    vmov.f32 s18, s26
+; CHECK-NEXT:    vmov.8 q5[1], r0
 ; CHECK-NEXT:    vmov.u8 r0, q2[10]
-; CHECK-NEXT:    vmov.8 q6[10], r0
+; CHECK-NEXT:    vmov.8 q5[2], r0
 ; CHECK-NEXT:    vmov.u8 r0, q2[14]
-; CHECK-NEXT:    vmov.8 q6[11], r0
+; CHECK-NEXT:    vmov.8 q5[3], r0
+; CHECK-NEXT:    vmov.u8 r0, q0[2]
+; CHECK-NEXT:    vmov q6, q5
+; CHECK-NEXT:    vstrw.32 q0, [sp, #32] @ 16-byte Spill
+; CHECK-NEXT:    vmov.8 q6[4], r0
+; CHECK-NEXT:    vmov.u8 r0, q0[6]
+; CHECK-NEXT:    vmov.8 q6[5], r0
+; CHECK-NEXT:    vmov.u8 r0, q0[10]
+; CHECK-NEXT:    vmov.8 q6[6], r0
 ; CHECK-NEXT:    vmov.u8 r0, q3[2]
-; CHECK-NEXT:    vmov.8 q6[12], r0
+; CHECK-NEXT:    vmov.8 q0[8], r0
 ; CHECK-NEXT:    vmov.u8 r0, q3[6]
-; CHECK-NEXT:    vmov.8 q6[13], r0
+; CHECK-NEXT:    vmov.8 q0[9], r0
 ; CHECK-NEXT:    vmov.u8 r0, q3[10]
-; CHECK-NEXT:    vmov.8 q6[14], r0
+; CHECK-NEXT:    vmov.8 q0[10], r0
 ; CHECK-NEXT:    vmov.u8 r0, q3[14]
-; CHECK-NEXT:    vmov.8 q6[15], r0
+; CHECK-NEXT:    vmov.8 q0[11], r0
+; CHECK-NEXT:    vmov.f32 s19, s31
+; CHECK-NEXT:    vmov.u8 r0, q1[2]
+; CHECK-NEXT:    vmov q7, q0
+; CHECK-NEXT:    vmov.8 q7[12], r0
+; CHECK-NEXT:    vmov.u8 r0, q1[6]
+; CHECK-NEXT:    vmov.8 q7[13], r0
+; CHECK-NEXT:    vmov.u8 r0, q1[10]
+; CHECK-NEXT:    vmov.8 q7[14], r0
 ; CHECK-NEXT:    vmov.u8 r0, q1[14]
-; CHECK-NEXT:    vmov.8 q5[7], r0
-; CHECK-NEXT:    vmov.u8 r0, q0[1]
-; CHECK-NEXT:    vmov.f32 s22, s26
-; CHECK-NEXT:    vmov.f32 s23, s27
-; CHECK-NEXT:    vadd.i8 q4, q5, q4
-; CHECK-NEXT:    vmov.8 q5[0], r0
-; CHECK-NEXT:    vmov.u8 r0, q0[5]
-; CHECK-NEXT:    vmov.8 q5[1], r0
-; CHECK-NEXT:    vmov.u8 r0, q0[9]
-; CHECK-NEXT:    vmov.8 q5[2], r0
-; CHECK-NEXT:    vmov.u8 r0, q0[13]
-; CHECK-NEXT:    vmov.8 q5[3], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[1]
-; CHECK-NEXT:    vmov.8 q5[4], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[5]
-; CHECK-NEXT:    vmov.8 q5[5], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[9]
-; CHECK-NEXT:    vmov.8 q5[6], r0
+; CHECK-NEXT:    vstrw.32 q1, [sp] @ 16-byte Spill
+; CHECK-NEXT:    vldrw.u32 q1, [sp, #32] @ 16-byte Reload
+; CHECK-NEXT:    vmov.8 q7[15], r0
+; CHECK-NEXT:    vmov.u8 r0, q1[14]
+; CHECK-NEXT:    vldrw.u32 q1, [sp] @ 16-byte Reload
+; CHECK-NEXT:    vmov.8 q6[7], r0
 ; CHECK-NEXT:    vmov.u8 r0, q2[1]
-; CHECK-NEXT:    vmov.8 q6[8], r0
+; CHECK-NEXT:    vmov.f32 s21, s25
+; CHECK-NEXT:    vmov.f32 s22, s2
+; CHECK-NEXT:    vmov.f32 s23, s31
+; CHECK-NEXT:    vadd.i8 q0, q5, q4
+; CHECK-NEXT:    vmov.8 q5[0], r0
 ; CHECK-NEXT:    vmov.u8 r0, q2[5]
-; CHECK-NEXT:    vmov.8 q6[9], r0
+; CHECK-NEXT:    vldrw.u32 q4, [sp, #32] @ 16-byte Reload
+; CHECK-NEXT:    vmov.8 q5[1], r0
 ; CHECK-NEXT:    vmov.u8 r0, q2[9]
-; CHECK-NEXT:    vmov.8 q6[10], r0
+; CHECK-NEXT:    vmov.8 q5[2], r0
 ; CHECK-NEXT:    vmov.u8 r0, q2[13]
-; CHECK-NEXT:    vmov.8 q6[11], r0
+; CHECK-NEXT:    vmov.8 q5[3], r0
+; CHECK-NEXT:    vmov.u8 r0, q4[1]
+; CHECK-NEXT:    vmov q6, q5
+; CHECK-NEXT:    vstrw.32 q0, [sp, #16] @ 16-byte Spill
+; CHECK-NEXT:    vmov.8 q6[4], r0
+; CHECK-NEXT:    vmov.u8 r0, q4[5]
+; CHECK-NEXT:    vmov.8 q6[5], r0
+; CHECK-NEXT:    vmov.u8 r0, q4[9]
+; CHECK-NEXT:    vmov.8 q6[6], r0
 ; CHECK-NEXT:    vmov.u8 r0, q3[1]
-; CHECK-NEXT:    vmov.8 q6[12], r0
+; CHECK-NEXT:    vmov.8 q7[8], r0
 ; CHECK-NEXT:    vmov.u8 r0, q3[5]
-; CHECK-NEXT:    vmov.8 q6[13], r0
+; CHECK-NEXT:    vmov.8 q7[9], r0
 ; CHECK-NEXT:    vmov.u8 r0, q3[9]
-; CHECK-NEXT:    vmov.8 q6[14], r0
+; CHECK-NEXT:    vmov.8 q7[10], r0
 ; CHECK-NEXT:    vmov.u8 r0, q3[13]
-; CHECK-NEXT:    vmov.8 q6[15], r0
+; CHECK-NEXT:    vmov.8 q7[11], r0
+; CHECK-NEXT:    vmov.u8 r0, q1[1]
+; CHECK-NEXT:    vmov q0, q7
+; CHECK-NEXT:    vmov.8 q0[12], r0
+; CHECK-NEXT:    vmov.u8 r0, q1[5]
+; CHECK-NEXT:    vmov.8 q0[13], r0
+; CHECK-NEXT:    vmov.u8 r0, q1[9]
+; CHECK-NEXT:    vmov.8 q0[14], r0
 ; CHECK-NEXT:    vmov.u8 r0, q1[13]
-; CHECK-NEXT:    vmov.8 q5[7], r0
-; CHECK-NEXT:    vmov.u8 r0, q0[0]
-; CHECK-NEXT:    vmov.f32 s22, s26
-; CHECK-NEXT:    vmov.f32 s23, s27
-; CHECK-NEXT:    vmov.8 q6[0], r0
-; CHECK-NEXT:    vmov.u8 r0, q0[4]
-; CHECK-NEXT:    vmov.8 q6[1], r0
-; CHECK-NEXT:    vmov.u8 r0, q0[8]
-; CHECK-NEXT:    vmov.8 q6[2], r0
-; CHECK-NEXT:    vmov.u8 r0, q0[12]
-; CHECK-NEXT:    vmov.8 q6[3], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[0]
-; CHECK-NEXT:    vmov.8 q6[4], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[4]
-; CHECK-NEXT:    vmov.8 q6[5], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[8]
-; CHECK-NEXT:    vmov.8 q6[6], r0
+; CHECK-NEXT:    vmov.8 q0[15], r0
+; CHECK-NEXT:    vmov.u8 r0, q4[13]
+; CHECK-NEXT:    vmov.8 q6[7], r0
 ; CHECK-NEXT:    vmov.u8 r0, q2[0]
-; CHECK-NEXT:    vmov.8 q0[8], r0
+; CHECK-NEXT:    vmov.f32 s21, s25
+; CHECK-NEXT:    vmov.8 q6[0], r0
 ; CHECK-NEXT:    vmov.u8 r0, q2[4]
-; CHECK-NEXT:    vmov.8 q0[9], r0
+; CHECK-NEXT:    vmov.f32 s22, s30
+; CHECK-NEXT:    vmov.8 q6[1], r0
 ; CHECK-NEXT:    vmov.u8 r0, q2[8]
-; CHECK-NEXT:    vmov.8 q0[10], r0
+; CHECK-NEXT:    vmov.8 q6[2], r0
 ; CHECK-NEXT:    vmov.u8 r0, q2[12]
-; CHECK-NEXT:    vmov.8 q0[11], r0
+; CHECK-NEXT:    vmov.8 q6[3], r0
+; CHECK-NEXT:    vmov.f32 s23, s3
+; CHECK-NEXT:    vmov.u8 r0, q4[0]
+; CHECK-NEXT:    vmov q0, q6
+; CHECK-NEXT:    vmov.8 q0[4], r0
+; CHECK-NEXT:    vmov.u8 r0, q4[4]
+; CHECK-NEXT:    vmov.8 q0[5], r0
+; CHECK-NEXT:    vmov.u8 r0, q4[8]
+; CHECK-NEXT:    vmov.8 q0[6], r0
 ; CHECK-NEXT:    vmov.u8 r0, q3[0]
-; CHECK-NEXT:    vmov.8 q0[12], r0
+; CHECK-NEXT:    vmov.8 q7[8], r0
 ; CHECK-NEXT:    vmov.u8 r0, q3[4]
-; CHECK-NEXT:    vmov.8 q0[13], r0
+; CHECK-NEXT:    vmov.8 q7[9], r0
 ; CHECK-NEXT:    vmov.u8 r0, q3[8]
-; CHECK-NEXT:    vmov.8 q0[14], r0
+; CHECK-NEXT:    vmov.8 q7[10], r0
 ; CHECK-NEXT:    vmov.u8 r0, q3[12]
-; CHECK-NEXT:    vmov.8 q0[15], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[12]
-; CHECK-NEXT:    vmov.8 q6[7], r0
-; CHECK-NEXT:    vmov.f32 s26, s2
-; CHECK-NEXT:    vmov.f32 s27, s3
+; CHECK-NEXT:    vmov.8 q7[11], r0
+; CHECK-NEXT:    vmov q2, q1
+; CHECK-NEXT:    vmov.u8 r0, q1[0]
+; CHECK-NEXT:    vmov q1, q7
+; CHECK-NEXT:    vmov.8 q1[12], r0
+; CHECK-NEXT:    vmov.u8 r0, q2[4]
+; CHECK-NEXT:    vmov.8 q1[13], r0
+; CHECK-NEXT:    vmov.u8 r0, q2[8]
+; CHECK-NEXT:    vmov.8 q1[14], r0
+; CHECK-NEXT:    vmov.u8 r0, q2[12]
+; CHECK-NEXT:    vmov.8 q1[15], r0
+; CHECK-NEXT:    vmov.u8 r0, q4[12]
+; CHECK-NEXT:    vmov.8 q0[7], r0
+; CHECK-NEXT:    vmov.f32 s25, s1
+; CHECK-NEXT:    vmov.f32 s26, s30
+; CHECK-NEXT:    vmov.f32 s27, s7
+; CHECK-NEXT:    vldrw.u32 q1, [sp, #16] @ 16-byte Reload
 ; CHECK-NEXT:    vadd.i8 q0, q6, q5
-; CHECK-NEXT:    vadd.i8 q0, q0, q4
-; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13}
+; CHECK-NEXT:    vadd.i8 q0, q0, q1
+; CHECK-NEXT:    add sp, #48
+; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13, d14, d15}
 ; CHECK-NEXT:    bx lr
 entry:
   %s1 = shufflevector <64 x i8> %src, <64 x i8> undef, <16 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28, i32 32, i32 36, i32 40, i32 44, i32 48, i32 52, i32 56, i32 60>
   %s2 = shufflevector <64 x i8> %src, <64 x i8> undef, <16 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29, i32 33, i32 37, i32 41, i32 45, i32 49, i32 53, i32 57, i32 61>
   %s3 = shufflevector <64 x i8> %src, <64 x i8> undef, <16 x i32> <i32 2, i32 6, i32 10, i32 14, i32 18, i32 22, i32 26, i32 30, i32 34, i32 38, i32 42, i32 46, i32 50, i32 54, i32 58, i32 62>
   %s4 = shufflevector <64 x i8> %src, <64 x i8> undef, <16 x i32> <i32 3, i32 7, i32 11, i32 15, i32 19, i32 23, i32 27, i32 31, i32 35, i32 39, i32 43, i32 47, i32 51, i32 55, i32 59, i32 63>
   %a1 = add <16 x i8> %s1, %s2
   %a2 = add <16 x i8> %s3, %s4

llvm/test/CodeGen/Thumb2/mve-soft-float-abi.ll

 entry:
   %sum = add <2 x i64> %lhs, %rhs
   ret <2 x i64> %sum
 }
 
 define <8 x half> @vector_add_f16(<8 x half> %lhs, <8 x half> %rhs) {
 ; CHECK-MVE-LABEL: vector_add_f16:
 ; CHECK-MVE:       @ %bb.0: @ %entry
-; CHECK-MVE-NEXT:    .save {r4, r5, r7, lr}
-; CHECK-MVE-NEXT:    push {r4, r5, r7, lr}
+; CHECK-MVE-NEXT:    .save {r4, r5, r6, r7, lr}
+; CHECK-MVE-NEXT:    push {r4, r5, r6, r7, lr}
+; CHECK-MVE-NEXT:    .pad #4
+; CHECK-MVE-NEXT:    sub sp, #4
 ; CHECK-MVE-NEXT:    .vsave {d8, d9, d10, d11, d12, d13}
 ; CHECK-MVE-NEXT:    vpush {d8, d9, d10, d11, d12, d13}
 ; CHECK-MVE-NEXT:    vmov d9, r2, r3
 ; CHECK-MVE-NEXT:    vmov d8, r0, r1
-; CHECK-MVE-NEXT:    add r0, sp, #64
+; CHECK-MVE-NEXT:    add r0, sp, #72
 ; CHECK-MVE-NEXT:    vldrw.u32 q6, [r0]
 ; CHECK-MVE-NEXT:    vmov.u16 r4, q4[0]
 ; CHECK-MVE-NEXT:    vmov.u16 r0, q6[0]
 ; CHECK-MVE-NEXT:    bl __aeabi_h2f
 ; CHECK-MVE-NEXT:    mov r5, r0
 ; CHECK-MVE-NEXT:    mov r0, r4
 ; CHECK-MVE-NEXT:    bl __aeabi_h2f
 ; CHECK-MVE-NEXT:    mov r1, r5

 ; CHECK-MVE-NEXT:    mov r5, r0
 ; CHECK-MVE-NEXT:    mov r0, r4
 ; CHECK-MVE-NEXT:    bl __aeabi_h2f
 ; CHECK-MVE-NEXT:    mov r1, r5
 ; CHECK-MVE-NEXT:    bl __aeabi_fadd
 ; CHECK-MVE-NEXT:    bl __aeabi_f2h
 ; CHECK-MVE-NEXT:    vmov.16 q5[3], r0
 ; CHECK-MVE-NEXT:    vmov.u16 r0, q6[4]
-; CHECK-MVE-NEXT:    vmov.u16 r4, q4[4]
+; CHECK-MVE-NEXT:    vmov r5, r4, d10
+; CHECK-MVE-NEXT:    vmov.u16 r6, q4[4]
 ; CHECK-MVE-NEXT:    bl __aeabi_h2f
-; CHECK-MVE-NEXT:    mov r5, r0
-; CHECK-MVE-NEXT:    mov r0, r4
+; CHECK-MVE-NEXT:    mov r7, r0
+; CHECK-MVE-NEXT:    mov r0, r6
 ; CHECK-MVE-NEXT:    bl __aeabi_h2f
-; CHECK-MVE-NEXT:    mov r1, r5
+; CHECK-MVE-NEXT:    mov r1, r7
 ; CHECK-MVE-NEXT:    bl __aeabi_fadd
 ; CHECK-MVE-NEXT:    bl __aeabi_f2h
 ; CHECK-MVE-NEXT:    vmov.16 q5[4], r0
 ; CHECK-MVE-NEXT:    vmov.u16 r0, q6[5]
-; CHECK-MVE-NEXT:    vmov.u16 r4, q4[5]
+; CHECK-MVE-NEXT:    vmov.u16 r6, q4[5]
 ; CHECK-MVE-NEXT:    bl __aeabi_h2f
-; CHECK-MVE-NEXT:    mov r5, r0
-; CHECK-MVE-NEXT:    mov r0, r4
+; CHECK-MVE-NEXT:    mov r7, r0
+; CHECK-MVE-NEXT:    mov r0, r6
 ; CHECK-MVE-NEXT:    bl __aeabi_h2f
-; CHECK-MVE-NEXT:    mov r1, r5
+; CHECK-MVE-NEXT:    mov r1, r7
 ; CHECK-MVE-NEXT:    bl __aeabi_fadd
 ; CHECK-MVE-NEXT:    bl __aeabi_f2h
 ; CHECK-MVE-NEXT:    vmov.16 q5[5], r0
 ; CHECK-MVE-NEXT:    vmov.u16 r0, q6[6]
-; CHECK-MVE-NEXT:    vmov.u16 r4, q4[6]
+; CHECK-MVE-NEXT:    vmov.u16 r6, q4[6]
 ; CHECK-MVE-NEXT:    bl __aeabi_h2f
-; CHECK-MVE-NEXT:    mov r5, r0
-; CHECK-MVE-NEXT:    mov r0, r4
+; CHECK-MVE-NEXT:    mov r7, r0
+; CHECK-MVE-NEXT:    mov r0, r6
 ; CHECK-MVE-NEXT:    bl __aeabi_h2f
-; CHECK-MVE-NEXT:    mov r1, r5
+; CHECK-MVE-NEXT:    mov r1, r7
 ; CHECK-MVE-NEXT:    bl __aeabi_fadd
 ; CHECK-MVE-NEXT:    bl __aeabi_f2h
 ; CHECK-MVE-NEXT:    vmov.16 q5[6], r0
 ; CHECK-MVE-NEXT:    vmov.u16 r0, q6[7]
-; CHECK-MVE-NEXT:    vmov.u16 r4, q4[7]
+; CHECK-MVE-NEXT:    vmov.u16 r6, q4[7]
 ; CHECK-MVE-NEXT:    bl __aeabi_h2f
-; CHECK-MVE-NEXT:    mov r5, r0
-; CHECK-MVE-NEXT:    mov r0, r4
+; CHECK-MVE-NEXT:    mov r7, r0
+; CHECK-MVE-NEXT:    mov r0, r6
 ; CHECK-MVE-NEXT:    bl __aeabi_h2f
-; CHECK-MVE-NEXT:    mov r1, r5
+; CHECK-MVE-NEXT:    mov r1, r7
 ; CHECK-MVE-NEXT:    bl __aeabi_fadd
 ; CHECK-MVE-NEXT:    bl __aeabi_f2h
 ; CHECK-MVE-NEXT:    vmov.16 q5[7], r0
-; CHECK-MVE-NEXT:    vmov r0, r1, d10
+; CHECK-MVE-NEXT:    mov r0, r5
 ; CHECK-MVE-NEXT:    vmov r2, r3, d11
+; CHECK-MVE-NEXT:    mov r1, r4
 ; CHECK-MVE-NEXT:    vpop {d8, d9, d10, d11, d12, d13}
-; CHECK-MVE-NEXT:    pop {r4, r5, r7, pc}
+; CHECK-MVE-NEXT:    add sp, #4
+; CHECK-MVE-NEXT:    pop {r4, r5, r6, r7, pc}
 ;
 ; CHECK-BE-LABEL: vector_add_f16:
 ; CHECK-BE:       @ %bb.0: @ %entry
 ; CHECK-BE-NEXT:    .save {r4, r5, r7, lr}
 ; CHECK-BE-NEXT:    push {r4, r5, r7, lr}
 ; CHECK-BE-NEXT:    .vsave {d8, d9, d10, d11, d12, d13}
 ; CHECK-BE-NEXT:    vpush {d8, d9, d10, d11, d12, d13}
 ; CHECK-BE-NEXT:    vmov d1, r3, r2

llvm/test/CodeGen/Thumb2/mve-vabdus.ll

 }
 
 define arm_aapcs_vfpcc <4 x i32> @vabd_s32(<4 x i32> %src1, <4 x i32> %src2) {
 ; CHECK-LABEL: vabd_s32:
 ; CHECK:       @ %bb.0:
 ; CHECK-NEXT:    .vsave {d8, d9}
 ; CHECK-NEXT:    vpush {d8, d9}
 ; CHECK-NEXT:    vmov.f32 s12, s2
-; CHECK-NEXT:    vmov.f32 s14, s3
 ; CHECK-NEXT:    vmov.f32 s16, s6
+; CHECK-NEXT:    vmov.f32 s14, s3
 ; CHECK-NEXT:    vmov r0, s12
 ; CHECK-NEXT:    vmov.f32 s18, s7
 ; CHECK-NEXT:    vmov r2, s16
-; CHECK-NEXT:    vmov.f32 s2, s1
-; CHECK-NEXT:    vmov.f32 s6, s5
 ; CHECK-NEXT:    vmov r3, s4
+; CHECK-NEXT:    vmov.f32 s6, s5
 ; CHECK-NEXT:    asrs r1, r0, #31
 ; CHECK-NEXT:    subs r0, r0, r2
 ; CHECK-NEXT:    sbc.w r1, r1, r2, asr #31
 ; CHECK-NEXT:    add.w r0, r0, r1, asr #31
 ; CHECK-NEXT:    eor.w r0, r0, r1, asr #31
 ; CHECK-NEXT:    vmov r1, s0
+; CHECK-NEXT:    vmov.f32 s2, s1
 ; CHECK-NEXT:    asrs r2, r1, #31
 ; CHECK-NEXT:    subs r1, r1, r3
 ; CHECK-NEXT:    sbc.w r2, r2, r3, asr #31
 ; CHECK-NEXT:    vmov r3, s6
 ; CHECK-NEXT:    add.w r1, r1, r2, asr #31
 ; CHECK-NEXT:    eor.w r1, r1, r2, asr #31
 ; CHECK-NEXT:    vmov r2, s18
 ; CHECK-NEXT:    vmov q2[2], q2[0], r1, r0
; CHECK-NEXT:    bx lr
   ret <8 x i16> %result
 }
 
 define arm_aapcs_vfpcc <4 x i32> @vabd_u32(<4 x i32> %src1, <4 x i32> %src2) {
 ; CHECK-LABEL: vabd_u32:
 ; CHECK:       @ %bb.0:
 ; CHECK-NEXT:    .vsave {d8, d9}
 ; CHECK-NEXT:    vpush {d8, d9}
-; CHECK-NEXT:    vmov.f32 s8, s6
-; CHECK-NEXT:    vmov.i64 q4, #0xffffffff
-; CHECK-NEXT:    vmov.f32 s12, s2
-; CHECK-NEXT:    vmov.f32 s10, s7
-; CHECK-NEXT:    vmov.f32 s14, s3
-; CHECK-NEXT:    vand q2, q2, q4
-; CHECK-NEXT:    vand q3, q3, q4
-; CHECK-NEXT:    vmov r2, s8
-; CHECK-NEXT:    vmov r3, s12
-; CHECK-NEXT:    vmov r0, s9
-; CHECK-NEXT:    vmov r1, s13
-; CHECK-NEXT:    vmov.f32 s6, s5
+; CHECK-NEXT:    vmov.f32 s12, s6
+; CHECK-NEXT:    movs r0, #0
+; CHECK-NEXT:    vmov.f32 s16, s2
+; CHECK-NEXT:    vmov.f32 s14, s7
+; CHECK-NEXT:    vmov.f32 s18, s3
+; CHECK-NEXT:    vmov r1, s12
+; CHECK-NEXT:    vmov r2, s16
+; CHECK-NEXT:    vmov r3, s0
 ; CHECK-NEXT:    vmov.f32 s2, s1
-; CHECK-NEXT:    vand q1, q1, q4
-; CHECK-NEXT:    vand q4, q0, q4
+; CHECK-NEXT:    subs r1, r2, r1
+; CHECK-NEXT:    sbc r2, r0, #0
+; CHECK-NEXT:    add.w r1, r1, r2, asr #31
+; CHECK-NEXT:    eor.w r1, r1, r2, asr #31
+; CHECK-NEXT:    vmov r2, s4
+; CHECK-NEXT:    vmov.f32 s6, s5
 ; CHECK-NEXT:    subs r2, r3, r2
-; CHECK-NEXT:    vmov r3, s4
-; CHECK-NEXT:    sbc.w r0, r1, r0
-; CHECK-NEXT:    add.w r1, r2, r0, asr #31
-; CHECK-NEXT:    vmov r2, s17
-; CHECK-NEXT:    eor.w r12, r1, r0, asr #31
-; CHECK-NEXT:    vmov r0, s16
-; CHECK-NEXT:    vmov r1, s5
-; CHECK-NEXT:    subs r0, r0, r3
-; CHECK-NEXT:    vmov r3, s14
-; CHECK-NEXT:    sbc.w r1, r2, r1
-; CHECK-NEXT:    vmov r2, s10
-; CHECK-NEXT:    add.w r0, r0, r1, asr #31
-; CHECK-NEXT:    eor.w r0, r0, r1, asr #31
-; CHECK-NEXT:    vmov r1, s15
-; CHECK-NEXT:    vmov q0[2], q0[0], r0, r12
-; CHECK-NEXT:    vmov r0, s11
+; CHECK-NEXT:    sbc r3, r0, #0
+; CHECK-NEXT:    add.w r2, r2, r3, asr #31
+; CHECK-NEXT:    eor.w r2, r2, r3, asr #31
+; CHECK-NEXT:    vmov r3, s2
+; CHECK-NEXT:    vmov q2[2], q2[0], r2, r1
+; CHECK-NEXT:    vmov r1, s14
+; CHECK-NEXT:    vmov r2, s18
+; CHECK-NEXT:    subs r1, r2, r1
+; CHECK-NEXT:    sbc r2, r0, #0
+; CHECK-NEXT:    add.w r1, r1, r2, asr #31
+; CHECK-NEXT:    eor.w r1, r1, r2, asr #31
+; CHECK-NEXT:    vmov r2, s6
 ; CHECK-NEXT:    subs r2, r3, r2
-; CHECK-NEXT:    vmov r3, s6
-; CHECK-NEXT:    sbc.w r0, r1, r0
-; CHECK-NEXT:    add.w r1, r2, r0, asr #31
-; CHECK-NEXT:    vmov r2, s19
-; CHECK-NEXT:    eor.w r12, r1, r0, asr #31
-; CHECK-NEXT:    vmov r0, s18
-; CHECK-NEXT:    vmov r1, s7
-; CHECK-NEXT:    subs r0, r0, r3
-; CHECK-NEXT:    sbc.w r1, r2, r1
-; CHECK-NEXT:    add.w r0, r0, r1, asr #31
-; CHECK-NEXT:    eor.w r0, r0, r1, asr #31
-; CHECK-NEXT:    vmov q0[3], q0[1], r0, r12
+; CHECK-NEXT:    sbc r0, r0, #0
+; CHECK-NEXT:    add.w r2, r2, r0, asr #31
+; CHECK-NEXT:    eor.w r0, r2, r0, asr #31
+; CHECK-NEXT:    vmov q2[3], q2[1], r0, r1
+; CHECK-NEXT:    vmov q0, q2
 ; CHECK-NEXT:    vpop {d8, d9}
 ; CHECK-NEXT:    bx lr
   %zextsrc1 = zext <4 x i32> %src1 to <4 x i64>
   %zextsrc2 = zext <4 x i32> %src2 to <4 x i64>
   %add1 = sub <4 x i64> %zextsrc1, %zextsrc2
   %add2 = sub <4 x i64> zeroinitializer, %add1
   %c = icmp sge <4 x i64> %add1, zeroinitializer
   %s = select <4 x i1> %c, <4 x i64> %add1, <4 x i64> %add2

 
 for.cond.cleanup:                                 ; preds = %vector.body
   ret void
 }
 
 define void @vabd_loop_s32(i32* nocapture readonly %x, i32* nocapture readonly %y, i32* noalias nocapture %z, i32 %n) {
 ; CHECK-LABEL: vabd_loop_s32:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .save {r4, r5, r6, r7, r8, r9, r10, r11, lr}
-; CHECK-NEXT:    push.w {r4, r5, r6, r7, r8, r9, r10, r11, lr}
+; CHECK-NEXT:    .save {r4, r5, r6, r7, r8, r9, lr}
+; CHECK-NEXT:    push.w {r4, r5, r6, r7, r8, r9, lr}
 ; CHECK-NEXT:    .pad #4
 ; CHECK-NEXT:    sub sp, #4
 ; CHECK-NEXT:    .vsave {d8, d9}
 ; CHECK-NEXT:    vpush {d8, d9}
 ; CHECK-NEXT:    mov.w lr, #256
-; CHECK-NEXT:    mov.w r12, #1
 ; CHECK-NEXT:    vmov.i32 q0, #0x0
 ; CHECK-NEXT:  .LBB8_1: @ %vector.body
 ; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
-; CHECK-NEXT:    vldrw.u32 q1, [r0], #16
-; CHECK-NEXT:    vldrw.u32 q2, [r1], #16
-; CHECK-NEXT:    vmov.f32 s12, s4
-; CHECK-NEXT:    vmov.f32 s16, s8
-; CHECK-NEXT:    vmov.f32 s14, s5
-; CHECK-NEXT:    vmov r3, s12
-; CHECK-NEXT:    vmov.f32 s18, s9
-; CHECK-NEXT:    vmov r5, s16
-; CHECK-NEXT:    vmov r7, s18
-; CHECK-NEXT:    asrs r4, r3, #31
-; CHECK-NEXT:    subs.w r8, r3, r5
-; CHECK-NEXT:    sbc.w r4, r4, r5, asr #31
-; CHECK-NEXT:    asrs r5, r4, #31
-; CHECK-NEXT:    movs r4, #0
-; CHECK-NEXT:    bfi r4, r5, #0, #4
-; CHECK-NEXT:    vmov r5, s14
-; CHECK-NEXT:    vmov.f32 s12, s6
-; CHECK-NEXT:    vmov.f32 s14, s7
+; CHECK-NEXT:    vldrw.u32 q2, [r0], #16
+; CHECK-NEXT:    vldrw.u32 q4, [r1], #16
 ; CHECK-NEXT:    vmov.f32 s4, s10
 ; CHECK-NEXT:    vmov.f32 s6, s11
+; CHECK-NEXT:    vmov.f32 s12, s18
+; CHECK-NEXT:    vmov r3, s4
+; CHECK-NEXT:    vmov.f32 s14, s19
+; CHECK-NEXT:    vmov r4, s12
+; CHECK-NEXT:    vmov.f32 s10, s9
+; CHECK-NEXT:    vmov.f32 s18, s17
+; CHECK-NEXT:    vmov r7, s8
+; CHECK-NEXT:    vmov r6, s18
+; CHECK-NEXT:    asr.w r12, r3, #31
+; CHECK-NEXT:    subs.w r8, r3, r4
+; CHECK-NEXT:    sbc.w r12, r12, r4, asr #31
+; CHECK-NEXT:    vmov r4, s10
+; CHECK-NEXT:    asrs r3, r4, #31
+; CHECK-NEXT:    subs r4, r4, r6
+; CHECK-NEXT:    sbc.w r9, r3, r6, asr #31
+; CHECK-NEXT:    vmov r6, s16
 ; CHECK-NEXT:    vmov r3, s6
-; CHECK-NEXT:    subs.w r9, r5, r7
-; CHECK-NEXT:    asr.w r6, r5, #31
-; CHECK-NEXT:    vmov r5, s12
-; CHECK-NEXT:    sbc.w r6, r6, r7, asr #31
-; CHECK-NEXT:    and.w r6, r12, r6, asr #31
-; CHECK-NEXT:    rsbs r6, r6, #0
-; CHECK-NEXT:    bfi r4, r6, #4, #4
-; CHECK-NEXT:    vmov r6, s14
-; CHECK-NEXT:    subs.w r10, r6, r3
-; CHECK-NEXT:    asr.w r7, r6, #31
-; CHECK-NEXT:    sbc.w r3, r7, r3, asr #31
-; CHECK-NEXT:    vmov r7, s4
-; CHECK-NEXT:    asrs r6, r5, #31
-; CHECK-NEXT:    asr.w r11, r3, #31
-; CHECK-NEXT:    and.w r3, r12, r3, asr #31
-; CHECK-NEXT:    rsbs r3, r3, #0
-; CHECK-NEXT:    subs r5, r5, r7
-; CHECK-NEXT:    sbc.w r6, r6, r7, asr #31
+; CHECK-NEXT:    subs r5, r7, r6
+; CHECK-NEXT:    asr.w r7, r7, #31
+; CHECK-NEXT:    vmov q2[2], q2[0], r5, r8
+; CHECK-NEXT:    vmov r5, s14
+; CHECK-NEXT:    sbc.w r6, r7, r6, asr #31
 ; CHECK-NEXT:    asrs r6, r6, #31
-; CHECK-NEXT:    vmov q1[2], q1[0], r6, r11
-; CHECK-NEXT:    vmov r6, s4
-; CHECK-NEXT:    vmov q1[2], q1[0], r8, r5
-; CHECK-NEXT:    vmov q1[3], q1[1], r9, r10
-; CHECK-NEXT:    and r6, r6, #1
-; CHECK-NEXT:    rsbs r6, r6, #0
-; CHECK-NEXT:    bfi r4, r6, #8, #4
-; CHECK-NEXT:    bfi r4, r3, #12, #4
-; CHECK-NEXT:    vmsr p0, r4
+; CHECK-NEXT:    subs r7, r3, r5
+; CHECK-NEXT:    asr.w r3, r3, #31
+; CHECK-NEXT:    vmov q2[3], q2[1], r4, r7
+; CHECK-NEXT:    mov.w r7, #0
+; CHECK-NEXT:    sbc.w r3, r3, r5, asr #31
+; CHECK-NEXT:    bfi r7, r6, #0, #4
+; CHECK-NEXT:    asr.w r4, r9, #31
+; CHECK-NEXT:    asr.w r6, r12, #31
+; CHECK-NEXT:    bfi r7, r4, #4, #4
+; CHECK-NEXT:    asrs r3, r3, #31
+; CHECK-NEXT:    bfi r7, r6, #8, #4
+; CHECK-NEXT:    bfi r7, r3, #12, #4
+; CHECK-NEXT:    vmsr p0, r7
 ; CHECK-NEXT:    vpst
-; CHECK-NEXT:    vsubt.i32 q1, q0, q1
-; CHECK-NEXT:    vstrb.8 q1, [r2], #16
+; CHECK-NEXT:    vsubt.i32 q2, q0, q2
+; CHECK-NEXT:    vstrb.8 q2, [r2], #16
 ; CHECK-NEXT:    le lr, .LBB8_1
 ; CHECK-NEXT:  @ %bb.2: @ %for.cond.cleanup
 ; CHECK-NEXT:    vpop {d8, d9}
 ; CHECK-NEXT:    add sp, #4
-; CHECK-NEXT:    pop.w {r4, r5, r6, r7, r8, r9, r10, r11, pc}
+; CHECK-NEXT:    pop.w {r4, r5, r6, r7, r8, r9, pc}
 entry:
   br label %vector.body
 
 vector.body:                                      ; preds = %vector.body, %entry
   %index = phi i32 [ 0, %entry ], [ %index.next, %vector.body ]
   %0 = getelementptr inbounds i32, i32* %x, i32 %index
   %1 = bitcast i32* %0 to <4 x i32>*
   %wide.load = load <4 x i32>, <4 x i32>* %1, align 4

 
 for.cond.cleanup:                                 ; preds = %vector.body
   ret void
 }
 
 define void @vabd_loop_u32(i32* nocapture readonly %x, i32* nocapture readonly %y, i32* noalias nocapture %z, i32 %n) {
 ; CHECK-LABEL: vabd_loop_u32:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .save {r4, r5, r6, r7, r8, r9, r10, r11, lr}
-; CHECK-NEXT:    push.w {r4, r5, r6, r7, r8, r9, r10, r11, lr}
-; CHECK-NEXT:    .pad #4
-; CHECK-NEXT:    sub sp, #4
-; CHECK-NEXT:    .vsave {d8, d9, d10, d11}
-; CHECK-NEXT:    vpush {d8, d9, d10, d11}
+; CHECK-NEXT:    .save {r4, r5, r6, r7, r8, lr}
+; CHECK-NEXT:    push.w {r4, r5, r6, r7, r8, lr}
+; CHECK-NEXT:    .vsave {d8, d9}
+; CHECK-NEXT:    vpush {d8, d9}
 ; CHECK-NEXT:    mov.w lr, #256
-; CHECK-NEXT:    vmov.i64 q0, #0xffffffff
-; CHECK-NEXT:    vmov.i32 q1, #0x0
+; CHECK-NEXT:    mov.w r12, #0
+; CHECK-NEXT:    vmov.i32 q0, #0x0
 ; CHECK-NEXT:  .LBB11_1: @ %vector.body
 ; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
 ; CHECK-NEXT:    vldrw.u32 q2, [r1], #16
-; CHECK-NEXT:    vmov.f32 s12, s8
-; CHECK-NEXT:    vmov.f32 s14, s9
-; CHECK-NEXT:    vand q4, q3, q0
-; CHECK-NEXT:    vldrw.u32 q3, [r0], #16
-; CHECK-NEXT:    vmov r3, s16
-; CHECK-NEXT:    vmov.f32 s20, s12
-; CHECK-NEXT:    vmov.f32 s22, s13
-; CHECK-NEXT:    vand q5, q5, q0
-; CHECK-NEXT:    vmov r4, s17
-; CHECK-NEXT:    vmov r6, s20
-; CHECK-NEXT:    vmov r5, s21
-; CHECK-NEXT:    vmov r7, s23
-; CHECK-NEXT:    subs.w r8, r6, r3
-; CHECK-NEXT:    vmov r3, s22
-; CHECK-NEXT:    sbc.w r4, r5, r4
-; CHECK-NEXT:    vmov r6, s19
-; CHECK-NEXT:    asrs r5, r4, #31
-; CHECK-NEXT:    movs r4, #0
-; CHECK-NEXT:    bfi r4, r5, #0, #4
-; CHECK-NEXT:    vmov r5, s18
-; CHECK-NEXT:    vmov.f32 s16, s10
-; CHECK-NEXT:    vmov.f32 s18, s11
-; CHECK-NEXT:    vand q2, q4, q0
-; CHECK-NEXT:    vmov.f32 s16, s14
-; CHECK-NEXT:    vmov.f32 s18, s15
-; CHECK-NEXT:    vand q3, q4, q0
-; CHECK-NEXT:    vmov r12, s12
-; CHECK-NEXT:    subs.w r9, r3, r5
-; CHECK-NEXT:    vmov r5, s14
-; CHECK-NEXT:    sbc.w r3, r7, r6
-; CHECK-NEXT:    movs r7, #1
-; CHECK-NEXT:    vmov r6, s15
-; CHECK-NEXT:    and.w r3, r7, r3, asr #31
-; CHECK-NEXT:    vmov r7, s10
-; CHECK-NEXT:    rsbs r3, r3, #0
-; CHECK-NEXT:    bfi r4, r3, #4, #4
-; CHECK-NEXT:    vmov r3, s11
-; CHECK-NEXT:    subs.w r10, r5, r7
-; CHECK-NEXT:    vmov r7, s9
-; CHECK-NEXT:    vmov r5, s13
-; CHECK-NEXT:    sbc.w r3, r6, r3
-; CHECK-NEXT:    vmov r6, s8
-; CHECK-NEXT:    asr.w r11, r3, #31
-; CHECK-NEXT:    subs.w r6, r12, r6
-; CHECK-NEXT:    sbc.w r7, r5, r7
-; CHECK-NEXT:    asrs r7, r7, #31
-; CHECK-NEXT:    vmov q2[2], q2[0], r7, r11
+; CHECK-NEXT:    vldrw.u32 q4, [r0], #16
+; CHECK-NEXT:    vmov.f32 s4, s10
+; CHECK-NEXT:    vmov.f32 s12, s18
+; CHECK-NEXT:    vmov.f32 s6, s11
+; CHECK-NEXT:    vmov.f32 s14, s19
+; CHECK-NEXT:    vmov r3, s4
+; CHECK-NEXT:    vmov r4, s12
+; CHECK-NEXT:    vmov.f32 s10, s9
+; CHECK-NEXT:    vmov.f32 s18, s17
 ; CHECK-NEXT:    vmov r7, s8
-; CHECK-NEXT:    vmov q2[2], q2[0], r8, r6
-; CHECK-NEXT:    vmov q2[3], q2[1], r9, r10
-; CHECK-NEXT:    and r7, r7, #1
-; CHECK-NEXT:    rsbs r7, r7, #0
-; CHECK-NEXT:    bfi r4, r7, #8, #4
-; CHECK-NEXT:    movs r7, #1
-; CHECK-NEXT:    and.w r3, r7, r3, asr #31
-; CHECK-NEXT:    rsbs r3, r3, #0
-; CHECK-NEXT:    bfi r4, r3, #12, #4
-; CHECK-NEXT:    vmsr p0, r4
+; CHECK-NEXT:    vmov r5, s10
+; CHECK-NEXT:    vmov r6, s18
+; CHECK-NEXT:    subs r4, r4, r3
+; CHECK-NEXT:    vmov r3, s16
+; CHECK-NEXT:    sbc r8, r12, #0
+; CHECK-NEXT:    subs r5, r6, r5
+; CHECK-NEXT:    sbc r6, r12, #0
+; CHECK-NEXT:    subs r3, r3, r7
+; CHECK-NEXT:    vmov q2[2], q2[0], r3, r4
+; CHECK-NEXT:    vmov r3, s6
+; CHECK-NEXT:    vmov r4, s14
+; CHECK-NEXT:    sbc r7, r12, #0
+; CHECK-NEXT:    subs r3, r4, r3
+; CHECK-NEXT:    asr.w r4, r7, #31
+; CHECK-NEXT:    vmov q2[3], q2[1], r5, r3
+; CHECK-NEXT:    mov.w r5, #0
+; CHECK-NEXT:    asr.w r3, r6, #31
+; CHECK-NEXT:    bfi r5, r4, #0, #4
+; CHECK-NEXT:    bfi r5, r3, #4, #4
+; CHECK-NEXT:    asr.w r3, r8, #31
+; CHECK-NEXT:    bfi r5, r3, #8, #4
+; CHECK-NEXT:    sbc r3, r12, #0
+; CHECK-NEXT:    asrs r3, r3, #31
+; CHECK-NEXT:    bfi r5, r3, #12, #4
+; CHECK-NEXT:    vmsr p0, r5
 ; CHECK-NEXT:    vpst
-; CHECK-NEXT:    vsubt.i32 q2, q1, q2
+; CHECK-NEXT:    vsubt.i32 q2, q0, q2
 ; CHECK-NEXT:    vstrb.8 q2, [r2], #16
 ; CHECK-NEXT:    le lr, .LBB11_1
 ; CHECK-NEXT:  @ %bb.2: @ %for.cond.cleanup
-; CHECK-NEXT:    vpop {d8, d9, d10, d11}
-; CHECK-NEXT:    add sp, #4
-; CHECK-NEXT:    pop.w {r4, r5, r6, r7, r8, r9, r10, r11, pc}
+; CHECK-NEXT:    vpop {d8, d9}
+; CHECK-NEXT:    pop.w {r4, r5, r6, r7, r8, pc}
 entry:
   br label %vector.body
 
 vector.body:                                      ; preds = %vector.body, %entry
   %index = phi i32 [ 0, %entry ], [ %index.next, %vector.body ]
   %0 = getelementptr inbounds i32, i32* %x, i32 %index
   %1 = bitcast i32* %0 to <4 x i32>*
   %wide.load = load <4 x i32>, <4 x i32>* %1, align 4

llvm/test/CodeGen/Thumb2/mve-vcreate.ll

 entry:
   %v1 = insertelement <4 x i32> undef, i32 %a, i32 0
   ret <4 x i32> %v1
 }
 
 define hidden <8 x i16> @create_i16(i16 zeroext %a, i16 zeroext %b, i16 zeroext %c, i16 zeroext %d, i16 zeroext %a2, i16 zeroext %b2, i16 zeroext %c2, i16 zeroext %d2) local_unnamed_addr #0 {
 ; CHECK-LABEL: create_i16:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .save {r4, r5, r7, lr}
-; CHECK-NEXT:    push {r4, r5, r7, lr}
+; CHECK-NEXT:    .save {r5, r6, r7, lr}
+; CHECK-NEXT:    push {r5, r6, r7, lr}
 ; CHECK-NEXT:    movs r7, #0
 ; CHECK-NEXT:    orr.w r0, r1, r0, lsl #16
 ; CHECK-NEXT:    lsll r2, r7, #16
-; CHECK-NEXT:    ldr r4, [sp, #24]
-; CHECK-NEXT:    orr.w r1, r0, r7
-; CHECK-NEXT:    ldrd r0, r7, [sp, #16]
 ; CHECK-NEXT:    movs r5, #0
-; CHECK-NEXT:    lsll r4, r5, #16
-; CHECK-NEXT:    orr.w r0, r7, r0, lsl #16
-; CHECK-NEXT:    orr.w r12, r0, r5
+; CHECK-NEXT:    orr.w r1, r0, r7
+; CHECK-NEXT:    ldrd r12, r7, [sp, #16]
+; CHECK-NEXT:    ldr r0, [sp, #24]
+; CHECK-NEXT:    lsll r0, r5, #16
+; CHECK-NEXT:    orr.w r7, r7, r12, lsl #16
+; CHECK-NEXT:    orrs r7, r5
+; CHECK-NEXT:    ldr r5, [sp, #28]
+; CHECK-NEXT:    orrs r0, r5
+; CHECK-NEXT:    vmov.32 q0[2], r0
 ; CHECK-NEXT:    orr.w r0, r2, r3
-; CHECK-NEXT:    ldr r2, [sp, #28]
-; CHECK-NEXT:    mov r3, r12
-; CHECK-NEXT:    orrs r2, r4
-; CHECK-NEXT:    pop {r4, r5, r7, pc}
+; CHECK-NEXT:    vmov.32 q0[3], r7
+; CHECK-NEXT:    vmov r12, lr, d1
+; CHECK-NEXT:    mov r2, r12
+; CHECK-NEXT:    mov r3, lr
+; CHECK-NEXT:    pop {r5, r6, r7, pc}
 entry:
   %conv = zext i16 %a to i64
   %shl = shl nuw i64 %conv, 48
   %conv1 = zext i16 %b to i64
   %shl2 = shl nuw nsw i64 %conv1, 32
   %or = or i64 %shl2, %shl
   %conv3 = zext i16 %c to i64
   %shl4 = shl nuw nsw i64 %conv3, 16
entry:
   %v7 = insertelement <8 x i16> %v6, i16 %a6, i32 6
   %v8 = insertelement <8 x i16> %v7, i16 %a7, i32 7
   ret <8 x i16> %v8
 }
 
 define hidden <16 x i8> @create_i8(i8 zeroext %a1, i8 zeroext %b1, i8 zeroext %c1, i8 zeroext %d1, i8 zeroext %a2, i8 zeroext %b2, i8 zeroext %c2, i8 zeroext %d2, i8 zeroext %a3, i8 zeroext %b3, i8 zeroext %c3, i8 zeroext %d3, i8 zeroext %a4, i8 zeroext %b4, i8 zeroext %c4, i8 zeroext %d4) local_unnamed_addr #0 {
 ; CHECK-LABEL: create_i8:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .save {r4, r5, r6, r7, r8, r9, r10, r11, lr}
-; CHECK-NEXT:    push.w {r4, r5, r6, r7, r8, r9, r10, r11, lr}
-; CHECK-NEXT:    ldr r4, [sp, #40]
+; CHECK-NEXT:    .save {r4, r5, r6, r7, r9, r11, lr}
+; CHECK-NEXT:    push.w {r4, r5, r6, r7, r9, r11, lr}
+; CHECK-NEXT:    ldr.w r12, [sp, #64]
 ; CHECK-NEXT:    mov.w r11, #0
-; CHECK-NEXT:    ldr r6, [sp, #36]
-; CHECK-NEXT:    movs r7, #0
-; CHECK-NEXT:    lsll r4, r11, #16
-; CHECK-NEXT:    mov r10, r1
-; CHECK-NEXT:    lsll r6, r7, #24
-; CHECK-NEXT:    mov r8, r3
-; CHECK-NEXT:    orr.w r1, r6, r4
-; CHECK-NEXT:    ldr r6, [sp, #44]
-; CHECK-NEXT:    movs r3, #0
-; CHECK-NEXT:    ldr r4, [sp, #72]
-; CHECK-NEXT:    lsll r6, r3, #8
+; CHECK-NEXT:    ldr r4, [sp, #60]
 ; CHECK-NEXT:    movs r5, #0
-; CHECK-NEXT:    orrs r1, r6
-; CHECK-NEXT:    ldr r6, [sp, #48]
-; CHECK-NEXT:    lsll r4, r5, #16
+; CHECK-NEXT:    lsll r12, r11, #16
+; CHECK-NEXT:    movs r7, #0
+; CHECK-NEXT:    lsll r4, r5, #24
+; CHECK-NEXT:    ldr r6, [sp, #28]
+; CHECK-NEXT:    orr.w r12, r12, r4
+; CHECK-NEXT:    ldr r4, [sp, #68]
+; CHECK-NEXT:    lsls r1, r1, #16
+; CHECK-NEXT:    orr.w r0, r1, r0, lsl #22
+; CHECK-NEXT:    lsll r4, r7, #8
 ; CHECK-NEXT:    mov.w r9, #0
-; CHECK-NEXT:    orr.w r12, r1, r6
-; CHECK-NEXT:    ldr r6, [sp, #68]
-; CHECK-NEXT:    movs r1, #0
-; CHECK-NEXT:    lsll r6, r1, #24
-; CHECK-NEXT:    orrs r6, r4
-; CHECK-NEXT:    ldr r4, [sp, #76]
-; CHECK-NEXT:    lsll r4, r9, #8
-; CHECK-NEXT:    orrs r6, r4
-; CHECK-NEXT:    ldr r4, [sp, #80]
-; CHECK-NEXT:    orr.w lr, r6, r4
-; CHECK-NEXT:    lsl.w r4, r10, #16
-; CHECK-NEXT:    orr.w r0, r4, r0, lsl #22
+; CHECK-NEXT:    orr.w r12, r12, r4
+; CHECK-NEXT:    ldr r4, [sp, #72]
 ; CHECK-NEXT:    orr.w r0, r0, r2, lsl #8
-; CHECK-NEXT:    add r0, r8
-; CHECK-NEXT:    orrs r0, r7
-; CHECK-NEXT:    orr.w r0, r0, r11
-; CHECK-NEXT:    orr.w r2, r0, r3
-; CHECK-NEXT:    ldr r0, [sp, #56]
-; CHECK-NEXT:    ldr r3, [sp, #52]
-; CHECK-NEXT:    lsls r0, r0, #16
-; CHECK-NEXT:    orr.w r0, r0, r3, lsl #22
-; CHECK-NEXT:    ldr r3, [sp, #60]
-; CHECK-NEXT:    orr.w r0, r0, r3, lsl #8
-; CHECK-NEXT:    ldr r3, [sp, #64]
+; CHECK-NEXT:    orr.w r4, r4, r12
 ; CHECK-NEXT:    add r0, r3
-; CHECK-NEXT:    orrs r0, r1
-; CHECK-NEXT:    mov r1, r2
+; CHECK-NEXT:    vmov.32 q0[2], r4
+; CHECK-NEXT:    ldr r4, [sp, #48]
+; CHECK-NEXT:    lsl.w r12, r4, #16
+; CHECK-NEXT:    ldr r4, [sp, #44]
+; CHECK-NEXT:    orr.w r12, r12, r4, lsl #22
+; CHECK-NEXT:    ldr r4, [sp, #52]
+; CHECK-NEXT:    orr.w r12, r12, r4, lsl #8
+; CHECK-NEXT:    ldr r4, [sp, #56]
+; CHECK-NEXT:    add r4, r12
+; CHECK-NEXT:    orrs r5, r4
+; CHECK-NEXT:    ldr r4, [sp, #32]
+; CHECK-NEXT:    orr.w r5, r5, r11
+; CHECK-NEXT:    orrs r7, r5
+; CHECK-NEXT:    movs r5, #0
+; CHECK-NEXT:    vmov.32 q0[3], r7
+; CHECK-NEXT:    movs r7, #0
+; CHECK-NEXT:    vmov r12, lr, d1
+; CHECK-NEXT:    lsll r6, r5, #24
+; CHECK-NEXT:    lsll r4, r7, #16
 ; CHECK-NEXT:    orrs r0, r5
-; CHECK-NEXT:    mov r2, lr
-; CHECK-NEXT:    orr.w r3, r0, r9
-; CHECK-NEXT:    mov r0, r12
-; CHECK-NEXT:    pop.w {r4, r5, r6, r7, r8, r9, r10, r11, pc}
+; CHECK-NEXT:    orrs r4, r6
+; CHECK-NEXT:    ldr r6, [sp, #36]
+; CHECK-NEXT:    orrs r0, r7
+; CHECK-NEXT:    lsll r6, r9, #8
+; CHECK-NEXT:    orrs r4, r6
+; CHECK-NEXT:    ldr r6, [sp, #40]
+; CHECK-NEXT:    orr.w r1, r0, r9
+; CHECK-NEXT:    orrs r4, r6
+; CHECK-NEXT:    mov r0, r4
+; CHECK-NEXT:    mov r2, r12
+; CHECK-NEXT:    mov r3, lr
+; CHECK-NEXT:    pop.w {r4, r5, r6, r7, r9, r11, pc}
 entry:
   %conv = zext i8 %a1 to i64
   %shl = shl nuw nsw i64 %conv, 54
   %conv1 = zext i8 %b1 to i64
   %shl2 = shl nuw nsw i64 %conv1, 48
   %or = or i64 %shl2, %shl
   %conv3 = zext i8 %c1 to i64
   %shl4 = shl nuw nsw i64 %conv3, 40

llvm/test/CodeGen/Thumb2/mve-vecreduce-add.ll

entry:
   %xx = sext <4 x i32> %x to <4 x i64>
   %z = call i64 @llvm.vector.reduce.add.v4i64(<4 x i64> %xx)
   ret i64 %z
 }
 
 define arm_aapcs_vfpcc i64 @add_v2i32_v2i64_zext(<2 x i32> %x) {
 ; CHECK-LABEL: add_v2i32_v2i64_zext:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.i64 q1, #0xffffffff
-; CHECK-NEXT:    vand q0, q0, q1
 ; CHECK-NEXT:    vmov r0, s2
-; CHECK-NEXT:    vmov r3, s0
-; CHECK-NEXT:    vmov r1, s3
-; CHECK-NEXT:    vmov r2, s1
-; CHECK-NEXT:    adds r0, r0, r3
-; CHECK-NEXT:    adcs r1, r2
+; CHECK-NEXT:    movs r2, #0
+; CHECK-NEXT:    vmov r1, s0
+; CHECK-NEXT:    adds r0, r0, r1
+; CHECK-NEXT:    adc r1, r2, #0
 ; CHECK-NEXT:    bx lr
 entry:
   %xx = zext <2 x i32> %x to <2 x i64>
   %z = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> %xx)
   ret i64 %z
 }
 
 define arm_aapcs_vfpcc i64 @add_v2i32_v2i64_sext(<2 x i32> %x) {
 ; CHECK-LABEL: add_v2i32_v2i64_sext:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    vmov r0, s0
-; CHECK-NEXT:    vmov q1, q0
 ; CHECK-NEXT:    vmov r2, s2
 ; CHECK-NEXT:    asrs r1, r0, #31
-; CHECK-NEXT:    vmov.32 q1[1], r1
-; CHECK-NEXT:    vmov r0, s4
 ; CHECK-NEXT:    adds r0, r0, r2
 ; CHECK-NEXT:    adc.w r1, r1, r2, asr #31
 ; CHECK-NEXT:    bx lr
 entry:
   %xx = sext <2 x i32> %x to <2 x i64>
   %z = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> %xx)
   ret i64 %z
 }

 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    vmov.u16 r0, q0[1]
 ; CHECK-NEXT:    vmov.u16 r1, q0[0]
 ; CHECK-NEXT:    vmov q2[2], q2[0], r1, r0
 ; CHECK-NEXT:    vmov.i64 q1, #0xffff
 ; CHECK-NEXT:    vand q2, q2, q1
 ; CHECK-NEXT:    vmov.u16 r2, q0[2]
 ; CHECK-NEXT:    vmov r0, s10
+; CHECK-NEXT:    vmov.u16 r3, q0[6]
 ; CHECK-NEXT:    vmov r1, s8
 ; CHECK-NEXT:    add r0, r1
 ; CHECK-NEXT:    vmov.u16 r1, q0[3]
-; CHECK-NEXT:    vmov q3[2], q3[0], r2, r1
+; CHECK-NEXT:    vmov q2[2], q2[0], r2, r1
 ; CHECK-NEXT:    vmov.u16 r2, q0[4]
-; CHECK-NEXT:    vand q3, q3, q1
-; CHECK-NEXT:    vmov r1, s12
+; CHECK-NEXT:    vand q2, q2, q1
+; CHECK-NEXT:    vmov r1, s8
 ; CHECK-NEXT:    add r0, r1
-; CHECK-NEXT:    vmov r1, s14
+; CHECK-NEXT:    vmov r1, s10
 ; CHECK-NEXT:    add r0, r1
 ; CHECK-NEXT:    vmov.u16 r1, q0[5]
-; CHECK-NEXT:    vmov q3[2], q3[0], r2, r1
-; CHECK-NEXT:    vmov r2, s11
-; CHECK-NEXT:    vand q3, q3, q1
-; CHECK-NEXT:    vmov r1, s12
-; CHECK-NEXT:    vmov r3, s14
+; CHECK-NEXT:    vmov q2[2], q2[0], r2, r1
+; CHECK-NEXT:    movs r2, #0
+; CHECK-NEXT:    vand q2, q2, q1
+; CHECK-NEXT:    vmov r1, s8
 ; CHECK-NEXT:    add r0, r1
-; CHECK-NEXT:    vmov r1, s15
-; CHECK-NEXT:    adds r0, r0, r3
-; CHECK-NEXT:    vmov.u16 r3, q0[6]
-; CHECK-NEXT:    adcs r1, r2
+; CHECK-NEXT:    vmov r1, s10
+; CHECK-NEXT:    adds r0, r0, r1
+; CHECK-NEXT:    adc r1, r2, #0
 ; CHECK-NEXT:    vmov.u16 r2, q0[7]
 ; CHECK-NEXT:    vmov q0[2], q0[0], r3, r2
 ; CHECK-NEXT:    vand q0, q0, q1
-; CHECK-NEXT:    vmov r3, s0
-; CHECK-NEXT:    vmov r2, s1
-; CHECK-NEXT:    adds r0, r0, r3
-; CHECK-NEXT:    vmov r3, s2
-; CHECK-NEXT:    adcs r1, r2
-; CHECK-NEXT:    vmov r2, s3
-; CHECK-NEXT:    adds r0, r0, r3
-; CHECK-NEXT:    adcs r1, r2
+; CHECK-NEXT:    vmov r2, s0
+; CHECK-NEXT:    adds r0, r0, r2
+; CHECK-NEXT:    vmov r2, s2
+; CHECK-NEXT:    adc r1, r1, #0
+; CHECK-NEXT:    adds r0, r0, r2
+; CHECK-NEXT:    adc r1, r1, #0
 ; CHECK-NEXT:    bx lr
 entry:
   %xx = zext <8 x i16> %x to <8 x i64>
   %z = call i64 @llvm.vector.reduce.add.v8i64(<8 x i64> %xx)
   ret i64 %z
 }
 
 define arm_aapcs_vfpcc i64 @add_v8i16_v8i64_sext(<8 x i16> %x) {
 ; CHECK-LABEL: add_v8i16_v8i64_sext:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.s16 r0, q0[1]
-; CHECK-NEXT:    vmov.s16 r1, q0[0]
-; CHECK-NEXT:    vmov q1[2], q1[0], r1, r0
-; CHECK-NEXT:    asrs r2, r0, #31
-; CHECK-NEXT:    asrs r1, r1, #31
-; CHECK-NEXT:    vmov q1[3], q1[1], r1, r2
-; CHECK-NEXT:    vmov r2, s6
-; CHECK-NEXT:    vmov r3, s4
-; CHECK-NEXT:    vmov r1, s5
-; CHECK-NEXT:    adds r2, r2, r3
-; CHECK-NEXT:    vmov.s16 r3, q0[2]
-; CHECK-NEXT:    adc.w r12, r1, r0, asr #31
-; CHECK-NEXT:    vmov.s16 r1, q0[3]
-; CHECK-NEXT:    vmov q1[2], q1[0], r3, r1
-; CHECK-NEXT:    asrs r0, r1, #31
-; CHECK-NEXT:    asrs r3, r3, #31
-; CHECK-NEXT:    vmov q1[3], q1[1], r3, r0
-; CHECK-NEXT:    vmov r3, s4
-; CHECK-NEXT:    vmov r0, s5
-; CHECK-NEXT:    adds r2, r2, r3
-; CHECK-NEXT:    vmov r3, s6
-; CHECK-NEXT:    adc.w r0, r0, r12
-; CHECK-NEXT:    adds r2, r2, r3
-; CHECK-NEXT:    vmov.s16 r3, q0[4]
-; CHECK-NEXT:    adc.w r12, r0, r1, asr #31
-; CHECK-NEXT:    vmov.s16 r1, q0[5]
-; CHECK-NEXT:    vmov q1[2], q1[0], r3, r1
-; CHECK-NEXT:    asrs r0, r1, #31
-; CHECK-NEXT:    asrs r3, r3, #31
-; CHECK-NEXT:    vmov q1[3], q1[1], r3, r0
-; CHECK-NEXT:    vmov r3, s4
-; CHECK-NEXT:    vmov r0, s5
-; CHECK-NEXT:    adds r2, r2, r3
-; CHECK-NEXT:    vmov r3, s6
-; CHECK-NEXT:    adc.w r0, r0, r12
-; CHECK-NEXT:    adds r2, r2, r3
-; CHECK-NEXT:    vmov.s16 r3, q0[7]
-; CHECK-NEXT:    adc.w r0, r0, r1, asr #31
-; CHECK-NEXT:    vmov.s16 r1, q0[6]
-; CHECK-NEXT:    adds r2, r2, r1
-; CHECK-NEXT:    adc.w r1, r0, r1, asr #31
-; CHECK-NEXT:    adds r0, r2, r3
-; CHECK-NEXT:    adc.w r1, r1, r3, asr #31
+; CHECK-NEXT:    vmov.s16 r0, q0[0]
+; CHECK-NEXT:    vmov.s16 r2, q0[1]
+; CHECK-NEXT:    asrs r1, r0, #31
+; CHECK-NEXT:    adds r0, r0, r2
+; CHECK-NEXT:    adc.w r1, r1, r2, asr #31
+; CHECK-NEXT:    vmov.s16 r2, q0[2]
+; CHECK-NEXT:    adds r0, r0, r2
+; CHECK-NEXT:    adc.w r1, r1, r2, asr #31
+; CHECK-NEXT:    vmov.s16 r2, q0[3]
+; CHECK-NEXT:    adds r0, r0, r2
+; CHECK-NEXT:    adc.w r1, r1, r2, asr #31
+; CHECK-NEXT:    vmov.s16 r2, q0[4]
+; CHECK-NEXT:    adds r0, r0, r2
+; CHECK-NEXT:    adc.w r1, r1, r2, asr #31
+; CHECK-NEXT:    vmov.s16 r2, q0[5]
+; CHECK-NEXT:    adds r0, r0, r2
+; CHECK-NEXT:    adc.w r1, r1, r2, asr #31
+; CHECK-NEXT:    vmov.s16 r2, q0[6]
+; CHECK-NEXT:    adds r0, r0, r2
+; CHECK-NEXT:    adc.w r1, r1, r2, asr #31
+; CHECK-NEXT:    vmov.s16 r2, q0[7]
+; CHECK-NEXT:    adds r0, r0, r2
+; CHECK-NEXT:    adc.w r1, r1, r2, asr #31
 ; CHECK-NEXT:    bx lr
 entry:
   %xx = sext <8 x i16> %x to <8 x i64>
   %z = call i64 @llvm.vector.reduce.add.v8i64(<8 x i64> %xx)
   ret i64 %z
 }
 
 define arm_aapcs_vfpcc i64 @add_v4i16_v4i64_zext(<4 x i16> %x) {
entry:
   %xx = sext <4 x i16> %x to <4 x i64>
   %z = call i64 @llvm.vector.reduce.add.v4i64(<4 x i64> %xx)
   ret i64 %z
 }
 
 define arm_aapcs_vfpcc i64 @add_v2i16_v2i64_zext(<2 x i16> %x) {
 ; CHECK-LABEL: add_v2i16_v2i64_zext:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.i64 q1, #0xffff
+; CHECK-NEXT:    vmov.i32 q1, #0xffff
 ; CHECK-NEXT:    vand q0, q0, q1
 ; CHECK-NEXT:    vmov r0, s2
 ; CHECK-NEXT:    vmov r1, s0
 ; CHECK-NEXT:    add r0, r1
-; CHECK-NEXT:    vmov r1, s3
+; CHECK-NEXT:    movs r1, #0
 ; CHECK-NEXT:    bx lr
 entry:
   %xx = zext <2 x i16> %x to <2 x i64>
   %z = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> %xx)
   ret i64 %z
 }
 
 define arm_aapcs_vfpcc i64 @add_v2i16_v2i64_sext(<2 x i16> %x) {

 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    vmov.u8 r0, q0[1]
 ; CHECK-NEXT:    vmov.u8 r1, q0[0]
 ; CHECK-NEXT:    vmov q2[2], q2[0], r1, r0
 ; CHECK-NEXT:    vmov.i64 q1, #0xff
 ; CHECK-NEXT:    vand q2, q2, q1
 ; CHECK-NEXT:    vmov.u8 r2, q0[2]
 ; CHECK-NEXT:    vmov r0, s10
+; CHECK-NEXT:    vmov.u8 r3, q0[6]
 ; CHECK-NEXT:    vmov r1, s8
 ; CHECK-NEXT:    add r0, r1
 ; CHECK-NEXT:    vmov.u8 r1, q0[3]
-; CHECK-NEXT:    vmov q3[2], q3[0], r2, r1
+; CHECK-NEXT:    vmov q2[2], q2[0], r2, r1
 ; CHECK-NEXT:    vmov.u8 r2, q0[4]
-; CHECK-NEXT:    vand q3, q3, q1
-; CHECK-NEXT:    vmov r1, s12
+; CHECK-NEXT:    vand q2, q2, q1
+; CHECK-NEXT:    vmov r1, s8
 ; CHECK-NEXT:    add r0, r1
-; CHECK-NEXT:    vmov r1, s14
+; CHECK-NEXT:    vmov r1, s10
 ; CHECK-NEXT:    add r0, r1
 ; CHECK-NEXT:    vmov.u8 r1, q0[5]
-; CHECK-NEXT:    vmov q3[2], q3[0], r2, r1
-; CHECK-NEXT:    vmov r2, s11
-; CHECK-NEXT:    vand q3, q3, q1
-; CHECK-NEXT:    vmov r1, s12
-; CHECK-NEXT:    vmov r3, s14
+; CHECK-NEXT:    vmov q2[2], q2[0], r2, r1
+; CHECK-NEXT:    movs r2, #0
+; CHECK-NEXT:    vand q2, q2, q1
+; CHECK-NEXT:    vmov r1, s8
 ; CHECK-NEXT:    add r0, r1
-; CHECK-NEXT:    vmov r1, s15
-; CHECK-NEXT:    adds r0, r0, r3
-; CHECK-NEXT:    vmov.u8 r3, q0[6]
-; CHECK-NEXT:    adcs r1, r2
+; CHECK-NEXT:    vmov r1, s10
+; CHECK-NEXT:    adds r0, r0, r1
+; CHECK-NEXT:    adc r1, r2, #0
 ; CHECK-NEXT:    vmov.u8 r2, q0[7]
 ; CHECK-NEXT:    vmov q2[2], q2[0], r3, r2
-; CHECK-NEXT:    vand q2, q2, q1
-; CHECK-NEXT:    vmov r3, s8
-; CHECK-NEXT:    vmov r2, s9
-; CHECK-NEXT:    adds r0, r0, r3
-; CHECK-NEXT:    vmov r3, s10
-; CHECK-NEXT:    adcs r1, r2
-; CHECK-NEXT:    vmov r2, s11
-; CHECK-NEXT:    adds r0, r0, r3
 ; CHECK-NEXT:    vmov.u8 r3, q0[8]
-; CHECK-NEXT:    adcs r1, r2
+; CHECK-NEXT:    vand q2, q2, q1
+; CHECK-NEXT:    vmov r2, s8
+; CHECK-NEXT:    adds r0, r0, r2
+; CHECK-NEXT:    vmov r2, s10
+; CHECK-NEXT:    adc r1, r1, #0
+; CHECK-NEXT:    adds r0, r0, r2
 ; CHECK-NEXT:    vmov.u8 r2, q0[9]
 ; CHECK-NEXT:    vmov q2[2], q2[0], r3, r2
+; CHECK-NEXT:    adc r1, r1, #0
 ; CHECK-NEXT:    vand q2, q2, q1
-; CHECK-NEXT:    vmov r3, s8
-; CHECK-NEXT:    vmov r2, s9
-; CHECK-NEXT:    adds r0, r0, r3
-; CHECK-NEXT:    vmov r3, s10
-; CHECK-NEXT:    adcs r1, r2
-; CHECK-NEXT:    vmov r2, s11
-; CHECK-NEXT:    adds r0, r0, r3
 ; CHECK-NEXT:    vmov.u8 r3, q0[10]
-; CHECK-NEXT:    adcs r1, r2
+; CHECK-NEXT:    vmov r2, s8
+; CHECK-NEXT:    adds r0, r0, r2
+; CHECK-NEXT:    vmov r2, s10
+; CHECK-NEXT:    adc r1, r1, #0
+; CHECK-NEXT:    adds r0, r0, r2
 ; CHECK-NEXT:    vmov.u8 r2, q0[11]
 ; CHECK-NEXT:    vmov q2[2], q2[0], r3, r2
+; CHECK-NEXT:    adc r1, r1, #0
 ; CHECK-NEXT:    vand q2, q2, q1
-; CHECK-NEXT:    vmov r3, s8
-; CHECK-NEXT:    vmov r2, s9
-; CHECK-NEXT:    adds r0, r0, r3
-; CHECK-NEXT:    vmov r3, s10
-; CHECK-NEXT:    adcs r1, r2
-; CHECK-NEXT:    vmov r2, s11
-; CHECK-NEXT:    adds r0, r0, r3
 ; CHECK-NEXT:    vmov.u8 r3, q0[12]
-; CHECK-NEXT:    adcs r1, r2
+; CHECK-NEXT:    vmov r2, s8
+; CHECK-NEXT:    adds r0, r0, r2
+; CHECK-NEXT:    vmov r2, s10
+; CHECK-NEXT:    adc r1, r1, #0
+; CHECK-NEXT:    adds r0, r0, r2
 ; CHECK-NEXT:    vmov.u8 r2, q0[13]
 ; CHECK-NEXT:    vmov q2[2], q2[0], r3, r2
+; CHECK-NEXT:    adc r1, r1, #0
 ; CHECK-NEXT:    vand q2, q2, q1
-; CHECK-NEXT:    vmov r3, s8
-; CHECK-NEXT:    vmov r2, s9
-; CHECK-NEXT:    adds r0, r0, r3
-; CHECK-NEXT:    vmov r3, s10
-; CHECK-NEXT:    adcs r1, r2
-; CHECK-NEXT:    vmov r2, s11
-; CHECK-NEXT:    adds r0, r0, r3
 ; CHECK-NEXT:    vmov.u8 r3, q0[14]
-; CHECK-NEXT:    adcs r1, r2
+; CHECK-NEXT:    vmov r2, s8
+; CHECK-NEXT:    adds r0, r0, r2
+; CHECK-NEXT:    vmov r2, s10
+; CHECK-NEXT:    adc r1, r1, #0
+; CHECK-NEXT:    adds r0, r0, r2
 ; CHECK-NEXT:    vmov.u8 r2, q0[15]
 ; CHECK-NEXT:    vmov q0[2], q0[0], r3, r2
+; CHECK-NEXT:    adc r1, r1, #0
 ; CHECK-NEXT:    vand q0, q0, q1
-; CHECK-NEXT:    vmov r3, s0
-; CHECK-NEXT:    vmov r2, s1
-; CHECK-NEXT:    adds r0, r0, r3
-; CHECK-NEXT:    vmov r3, s2
-; CHECK-NEXT:    adcs r1, r2
-; CHECK-NEXT:    vmov r2, s3
-; CHECK-NEXT:    adds r0, r0, r3
-; CHECK-NEXT:    adcs r1, r2
+; CHECK-NEXT:    vmov r2, s0
+; CHECK-NEXT:    adds r0, r0, r2
+; CHECK-NEXT:    vmov r2, s2
+; CHECK-NEXT:    adc r1, r1, #0
+; CHECK-NEXT:    adds r0, r0, r2
+; CHECK-NEXT:    adc r1, r1, #0
 ; CHECK-NEXT:    bx lr
 entry:
   %xx = zext <16 x i8> %x to <16 x i64>
   %z = call i64 @llvm.vector.reduce.add.v16i64(<16 x i64> %xx)
   ret i64 %z
 }
 
 define arm_aapcs_vfpcc i64 @add_v16i8_v16i64_sext(<16 x i8> %x) {
 ; CHECK-LABEL: add_v16i8_v16i64_sext:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.s8 r0, q0[1]
-; CHECK-NEXT:    vmov.s8 r1, q0[0]
-; CHECK-NEXT:    vmov q1[2], q1[0], r1, r0
-; CHECK-NEXT:    asrs r2, r0, #31
-; CHECK-NEXT:    asrs r1, r1, #31
-; CHECK-NEXT:    vmov q1[3], q1[1], r1, r2
-; CHECK-NEXT:    vmov r2, s6
-; CHECK-NEXT:    vmov r3, s4
-; CHECK-NEXT:    vmov r1, s5
-; CHECK-NEXT:    adds r2, r2, r3
-; CHECK-NEXT:    vmov.s8 r3, q0[2]
-; CHECK-NEXT:    adc.w r12, r1, r0, asr #31
-; CHECK-NEXT:    vmov.s8 r1, q0[3]
-; CHECK-NEXT:    vmov q1[2], q1[0], r3, r1
-; CHECK-NEXT:    asrs r0, r1, #31
-; CHECK-NEXT:    asrs r3, r3, #31
-; CHECK-NEXT:    vmov q1[3], q1[1], r3, r0
-; CHECK-NEXT:    vmov r3, s4
-; CHECK-NEXT:    vmov r0, s5
-; CHECK-NEXT:    adds r2, r2, r3
-; CHECK-NEXT:    vmov r3, s6
-; CHECK-NEXT:    adc.w r0, r0, r12
-; CHECK-NEXT:    adds r2, r2, r3
-; CHECK-NEXT:    vmov.s8 r3, q0[4]
-; CHECK-NEXT:    adc.w r12, r0, r1, asr #31
-; CHECK-NEXT:    vmov.s8 r1, q0[5]
-; CHECK-NEXT:    vmov q1[2], q1[0], r3, r1
-; CHECK-NEXT:    asrs r0, r1, #31
-; CHECK-NEXT:    asrs r3, r3, #31
-; CHECK-NEXT:    vmov q1[3], q1[1], r3, r0
-; CHECK-NEXT:    vmov r3, s4
-; CHECK-NEXT:    vmov r0, s5
-; CHECK-NEXT:    adds r2, r2, r3
-; CHECK-NEXT:    vmov r3, s6
-; CHECK-NEXT:    adc.w r0, r0, r12
-; CHECK-NEXT:    adds r2, r2, r3
-; CHECK-NEXT:    vmov.s8 r3, q0[6]
-; CHECK-NEXT:    adc.w r12, r0, r1, asr #31
-; CHECK-NEXT:    vmov.s8 r1, q0[7]
-; CHECK-NEXT:    vmov q1[2], q1[0], r3, r1
-; CHECK-NEXT:    asrs r0, r1, #31
-; CHECK-NEXT:    asrs r3, r3, #31
-; CHECK-NEXT:    vmov q1[3], q1[1], r3, r0
-; CHECK-NEXT:    vmov r3, s4
-; CHECK-NEXT:    vmov r0, s5
-; CHECK-NEXT:    adds r2, r2, r3
-; CHECK-NEXT:    vmov r3, s6
-; CHECK-NEXT:    adc.w r0, r0, r12
-; CHECK-NEXT:    adds r2, r2, r3
-; CHECK-NEXT:    vmov.s8 r3, q0[8]
-; CHECK-NEXT:    adc.w r12, r0, r1, asr #31
-; CHECK-NEXT:    vmov.s8 r1, q0[9]
-; CHECK-NEXT:    vmov q1[2], q1[0], r3, r1
-; CHECK-NEXT:    asrs r0, r1, #31
-; CHECK-NEXT:    asrs r3, r3, #31
-; CHECK-NEXT:    vmov q1[3], q1[1], r3, r0
-; CHECK-NEXT:    vmov r3, s4
-; CHECK-NEXT:    vmov r0, s5
-; CHECK-NEXT:    adds r2, r2, r3
-; CHECK-NEXT:    vmov r3, s6
-; CHECK-NEXT:    adc.w r0, r0, r12
-; CHECK-NEXT:    adds r2, r2, r3
-; CHECK-NEXT:    vmov.s8 r3, q0[10]
-; CHECK-NEXT:    adc.w r12, r0, r1, asr #31
-; CHECK-NEXT:    vmov.s8 r1, q0[11]
-; CHECK-NEXT:    vmov q1[2], q1[0], r3, r1
-; CHECK-NEXT:    asrs r0, r1, #31
-; CHECK-NEXT:    asrs r3, r3, #31
-; CHECK-NEXT:    vmov q1[3], q1[1], r3, r0
-; CHECK-NEXT:    vmov r3, s4
-; CHECK-NEXT:    vmov r0, s5
-; CHECK-NEXT:    adds r2, r2, r3
-; CHECK-NEXT:    vmov r3, s6
-; CHECK-NEXT:    adc.w r0, r0, r12
-; CHECK-NEXT:    adds r2, r2, r3
-; CHECK-NEXT:    vmov.s8 r3, q0[12]
-; CHECK-NEXT:    adc.w r12, r0, r1, asr #31
-; CHECK-NEXT:    vmov.s8 r1, q0[13]
-; CHECK-NEXT:    vmov q1[2], q1[0], r3, r1
-; CHECK-NEXT:    asrs r0, r1, #31
-; CHECK-NEXT:    asrs r3, r3, #31
-; CHECK-NEXT:    vmov q1[3], q1[1], r3, r0
-; CHECK-NEXT:    vmov r3, s4
-; CHECK-NEXT:    vmov r0, s5
-; CHECK-NEXT:    adds r2, r2, r3
-; CHECK-NEXT:    vmov r3, s6
-; CHECK-NEXT:    adc.w r0, r0, r12
-; CHECK-NEXT:    adds r2, r2, r3
-; CHECK-NEXT:    vmov.s8 r3, q0[15]
-; CHECK-NEXT:    adc.w r0, r0, r1, asr #31
-; CHECK-NEXT:    vmov.s8 r1, q0[14]
-; CHECK-NEXT:    adds r2, r2, r1
-; CHECK-NEXT:    adc.w r1, r0, r1, asr #31
-; CHECK-NEXT:    adds r0, r2, r3
-; CHECK-NEXT:    adc.w r1, r1, r3, asr #31
+; CHECK-NEXT:    vmov.s8 r0, q0[0]
+; CHECK-NEXT:    vmov.s8 r2, q0[1]
+; CHECK-NEXT:    asrs r1, r0, #31
+; CHECK-NEXT:    adds r0, r0, r2
+; CHECK-NEXT:    adc.w r1, r1, r2, asr #31
+; CHECK-NEXT:    vmov.s8 r2, q0[2]
+; CHECK-NEXT:    adds r0, r0, r2
+; CHECK-NEXT:    adc.w r1, r1, r2, asr #31
+; CHECK-NEXT:    vmov.s8 r2, q0[3]
+; CHECK-NEXT:    adds r0, r0, r2
+; CHECK-NEXT:    adc.w r1, r1, r2, asr #31
+; CHECK-NEXT:    vmov.s8 r2, q0[4]
+; CHECK-NEXT:    adds r0, r0, r2
+; CHECK-NEXT:    adc.w r1, r1, r2, asr #31
+; CHECK-NEXT:    vmov.s8 r2, q0[5]
+; CHECK-NEXT:    adds r0, r0, r2
+; CHECK-NEXT:    adc.w r1, r1, r2, asr #31
+; CHECK-NEXT:    vmov.s8 r2, q0[6]
+; CHECK-NEXT:    adds r0, r0, r2
+; CHECK-NEXT:    adc.w r1, r1, r2, asr #31
+; CHECK-NEXT:    vmov.s8 r2, q0[7]
+; CHECK-NEXT:    adds r0, r0, r2
+; CHECK-NEXT:    adc.w r1, r1, r2, asr #31
+; CHECK-NEXT:    vmov.s8 r2, q0[8]
+; CHECK-NEXT:    adds r0, r0, r2
+; CHECK-NEXT:    adc.w r1, r1, r2, asr #31
+; CHECK-NEXT:    vmov.s8 r2, q0[9]
+; CHECK-NEXT:    adds r0, r0, r2
+; CHECK-NEXT:    adc.w r1, r1, r2, asr #31
+; CHECK-NEXT:    vmov.s8 r2, q0[10]
+; CHECK-NEXT:    adds r0, r0, r2
+; CHECK-NEXT:    adc.w r1, r1, r2, asr #31
+; CHECK-NEXT:    vmov.s8 r2, q0[11]
+; CHECK-NEXT:    adds r0, r0, r2
+; CHECK-NEXT:    adc.w r1, r1, r2, asr #31
+; CHECK-NEXT:    vmov.s8 r2, q0[12]
+; CHECK-NEXT:    adds r0, r0, r2
+; CHECK-NEXT:    adc.w r1, r1, r2, asr #31
+; CHECK-NEXT:    vmov.s8 r2, q0[13]
+; CHECK-NEXT:    adds r0, r0, r2
+; CHECK-NEXT:    adc.w r1, r1, r2, asr #31
+; CHECK-NEXT:    vmov.s8 r2, q0[14]
+; CHECK-NEXT:    adds r0, r0, r2
+; CHECK-NEXT:    adc.w r1, r1, r2, asr #31
+; CHECK-NEXT:    vmov.s8 r2, q0[15]
+; CHECK-NEXT:    adds r0, r0, r2
+; CHECK-NEXT:    adc.w r1, r1, r2, asr #31
 ; CHECK-NEXT:    bx lr
 entry:
   %xx = sext <16 x i8> %x to <16 x i64>
   %z = call i64 @llvm.vector.reduce.add.v16i64(<16 x i64> %xx)
   ret i64 %z
 }
 
 define arm_aapcs_vfpcc i64 @add_v8i8_v8i64_zext(<8 x i8> %x) {
 ; CHECK-LABEL: add_v8i8_v8i64_zext:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    vmovlb.u8 q0, q0
 ; CHECK-NEXT:    vmov.i64 q1, #0xffff
 ; CHECK-NEXT:    vmov.u16 r0, q0[1]
 ; CHECK-NEXT:    vmov.u16 r1, q0[0]
 ; CHECK-NEXT:    vmov q2[2], q2[0], r1, r0
 ; CHECK-NEXT:    vmov.u16 r2, q0[2]
 ; CHECK-NEXT:    vand q2, q2, q1
+; CHECK-NEXT:    vmov.u16 r3, q0[6]
 ; CHECK-NEXT:    vmov r0, s10
 ; CHECK-NEXT:    vmov r1, s8
 ; CHECK-NEXT:    add r0, r1
 ; CHECK-NEXT:    vmov.u16 r1, q0[3]
-; CHECK-NEXT:    vmov q3[2], q3[0], r2, r1
+; CHECK-NEXT:    vmov q2[2], q2[0], r2, r1
 ; CHECK-NEXT:    vmov.u16 r2, q0[4]
-; CHECK-NEXT:    vand q3, q3, q1
-; CHECK-NEXT:    vmov r1, s12
+; CHECK-NEXT:    vand q2, q2, q1
+; CHECK-NEXT:    vmov r1, s8
 ; CHECK-NEXT:    add r0, r1
-; CHECK-NEXT:    vmov r1, s14
+; CHECK-NEXT:    vmov r1, s10
 ; CHECK-NEXT:    add r0, r1
 ; CHECK-NEXT:    vmov.u16 r1, q0[5]
-; CHECK-NEXT:    vmov q3[2], q3[0], r2, r1
-; CHECK-NEXT:    vmov r2, s11
-; CHECK-NEXT:    vand q3, q3, q1
-; CHECK-NEXT:    vmov r1, s12
-; CHECK-NEXT:    vmov r3, s14
+; CHECK-NEXT:    vmov q2[2], q2[0], r2, r1
+; CHECK-NEXT:    movs r2, #0
+; CHECK-NEXT:    vand q2, q2, q1
+; CHECK-NEXT:    vmov r1, s8
 ; CHECK-NEXT:    add r0, r1
-; CHECK-NEXT:    vmov r1, s15
-; CHECK-NEXT:    adds r0, r0, r3
-; CHECK-NEXT:    vmov.u16 r3, q0[6]
-; CHECK-NEXT:    adcs r1, r2
+; CHECK-NEXT:    vmov r1, s10
+; CHECK-NEXT:    adds r0, r0, r1
+; CHECK-NEXT:    adc r1, r2, #0
 ; CHECK-NEXT:    vmov.u16 r2, q0[7]
 ; CHECK-NEXT:    vmov q0[2], q0[0], r3, r2
 ; CHECK-NEXT:    vand q0, q0, q1
-; CHECK-NEXT:    vmov r3, s0
-; CHECK-NEXT:    vmov r2, s1
-; CHECK-NEXT:    adds r0, r0, r3
-; CHECK-NEXT:    vmov r3, s2
-; CHECK-NEXT:    adcs r1, r2
-; CHECK-NEXT:    vmov r2, s3
-; CHECK-NEXT:    adds r0, r0, r3
-; CHECK-NEXT:    adcs r1, r2
+; CHECK-NEXT:    vmov r2, s0
+; CHECK-NEXT:    adds r0, r0, r2
+; CHECK-NEXT:    vmov r2, s2
+; CHECK-NEXT:    adc r1, r1, #0
+; CHECK-NEXT:    adds r0, r0, r2
+; CHECK-NEXT:    adc r1, r1, #0
 ; CHECK-NEXT:    bx lr
 entry:
   %xx = zext <8 x i8> %x to <8 x i64>
   %z = call i64 @llvm.vector.reduce.add.v8i64(<8 x i64> %xx)
   ret i64 %z
 }
 
 define arm_aapcs_vfpcc i64 @add_v8i8_v8i64_sext(<8 x i8> %x) {
 ; CHECK-LABEL: add_v8i8_v8i64_sext:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.u16 r0, q0[1]
-; CHECK-NEXT:    vmov.u16 r1, q0[0]
+; CHECK-NEXT:    vmov.u16 r0, q0[0]
+; CHECK-NEXT:    vmov.u16 r2, q0[1]
 ; CHECK-NEXT:    sxtb r0, r0
-; CHECK-NEXT:    sxtb r1, r1
-; CHECK-NEXT:    vmov q1[2], q1[0], r1, r0
-; CHECK-NEXT:    asrs r2, r0, #31
-; CHECK-NEXT:    asrs r1, r1, #31
-; CHECK-NEXT:    vmov q1[3], q1[1], r1, r2
-; CHECK-NEXT:    vmov r2, s6
-; CHECK-NEXT:    vmov r3, s4
-; CHECK-NEXT:    vmov r1, s5
-; CHECK-NEXT:    adds r2, r2, r3
-; CHECK-NEXT:    vmov.u16 r3, q0[2]
-; CHECK-NEXT:    adc.w r12, r1, r0, asr #31
-; CHECK-NEXT:    vmov.u16 r1, q0[3]
-; CHECK-NEXT:    sxtb r1, r1
-; CHECK-NEXT:    sxtb r3, r3
-; CHECK-NEXT:    vmov q1[2], q1[0], r3, r1
-; CHECK-NEXT:    asrs r0, r1, #31
-; CHECK-NEXT:    asrs r3, r3, #31
-; CHECK-NEXT:    vmov q1[3], q1[1], r3, r0
-; CHECK-NEXT:    vmov r3, s4
-; CHECK-NEXT:    vmov r0, s5
-; CHECK-NEXT:    adds r2, r2, r3
-; CHECK-NEXT:    vmov r3, s6
-; CHECK-NEXT:    adc.w r0, r0, r12
-; CHECK-NEXT:    adds r2, r2, r3
-; CHECK-NEXT:    vmov.u16 r3, q0[4]
-; CHECK-NEXT:    adc.w r12, r0, r1, asr #31
-; CHECK-NEXT:    vmov.u16 r1, q0[5]
-; CHECK-NEXT:    sxtb r1, r1
-; CHECK-NEXT:    sxtb r3, r3
-; CHECK-NEXT:    vmov q1[2], q1[0], r3, r1
-; CHECK-NEXT:    asrs r0, r1, #31
-; CHECK-NEXT:    asrs r3, r3, #31
-; CHECK-NEXT:    vmov q1[3], q1[1], r3, r0
-; CHECK-NEXT:    vmov r3, s4
-; CHECK-NEXT:    vmov r0, s5
-; CHECK-NEXT:    adds r2, r2, r3
-; CHECK-NEXT:    vmov r3, s6
-; CHECK-NEXT:    adc.w r0, r0, r12
-; CHECK-NEXT:    adds r2, r2, r3
-; CHECK-NEXT:    adc.w r0, r0, r1, asr #31
-; CHECK-NEXT:    vmov.u16 r1, q0[6]
-; CHECK-NEXT:    sxtb r1, r1
-; CHECK-NEXT:    adds r2, r2, r1
-; CHECK-NEXT:    adc.w r1, r0, r1, asr #31
-; CHECK-NEXT:    vmov.u16 r0, q0[7]
-; CHECK-NEXT:    sxtb r3, r0
-; CHECK-NEXT:    adds r0, r2, r3
-; CHECK-NEXT:    adc.w r1, r1, r3, asr #31
+; CHECK-NEXT:    sxtb r2, r2
+; CHECK-NEXT:    asrs r1, r0, #31
+; CHECK-NEXT:    adds r0, r0, r2
+; CHECK-NEXT:    adc.w r1, r1, r2, asr #31
+; CHECK-NEXT:    vmov.u16 r2, q0[2]
+; CHECK-NEXT:    sxtb r2, r2
+; CHECK-NEXT:    adds r0, r0, r2
+; CHECK-NEXT:    adc.w r1, r1, r2, asr #31
+; CHECK-NEXT:    vmov.u16 r2, q0[3]
+; CHECK-NEXT:    sxtb r2, r2
+; CHECK-NEXT:    adds r0, r0, r2
+; CHECK-NEXT:    adc.w r1, r1, r2, asr #31
+; CHECK-NEXT:    vmov.u16 r2, q0[4]
+; CHECK-NEXT:    sxtb r2, r2
+; CHECK-NEXT:    adds r0, r0, r2
+; CHECK-NEXT:    adc.w r1, r1, r2, asr #31
+; CHECK-NEXT:    vmov.u16 r2, q0[5]
+; CHECK-NEXT:    sxtb r2, r2
+; CHECK-NEXT:    adds r0, r0, r2
+; CHECK-NEXT:    adc.w r1, r1, r2, asr #31
+; CHECK-NEXT:    vmov.u16 r2, q0[6]
+; CHECK-NEXT:    sxtb r2, r2
+; CHECK-NEXT:    adds r0, r0, r2
+; CHECK-NEXT:    adc.w r1, r1, r2, asr #31
+; CHECK-NEXT:    vmov.u16 r2, q0[7]
+; CHECK-NEXT:    sxtb r2, r2
+; CHECK-NEXT:    adds r0, r0, r2
+; CHECK-NEXT:    adc.w r1, r1, r2, asr #31
 ; CHECK-NEXT:    bx lr
 entry:
   %xx = sext <8 x i8> %x to <8 x i64>
   %z = call i64 @llvm.vector.reduce.add.v8i64(<8 x i64> %xx)
   ret i64 %z
 }
 
 define arm_aapcs_vfpcc i64 @add_v4i8_v4i64_zext(<4 x i8> %x) {
entry:
   %xx = sext <4 x i8> %x to <4 x i64>
   %z = call i64 @llvm.vector.reduce.add.v4i64(<4 x i64> %xx)
   ret i64 %z
 }
 
 define arm_aapcs_vfpcc i64 @add_v2i8_v2i64_zext(<2 x i8> %x) {
 ; CHECK-LABEL: add_v2i8_v2i64_zext:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.i64 q1, #0xff
+; CHECK-NEXT:    vmov.i32 q1, #0xff
 ; CHECK-NEXT:    vand q0, q0, q1
 ; CHECK-NEXT:    vmov r0, s2
 ; CHECK-NEXT:    vmov r1, s0
 ; CHECK-NEXT:    add r0, r1
-; CHECK-NEXT:    vmov r1, s3
+; CHECK-NEXT:    movs r1, #0
 ; CHECK-NEXT:    bx lr
 entry:
   %xx = zext <2 x i8> %x to <2 x i64>
   %z = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> %xx)
   ret i64 %z
 }
 
 define arm_aapcs_vfpcc i64 @add_v2i8_v2i64_sext(<2 x i8> %x) {
entry:
   %z = call i64 @llvm.vector.reduce.add.v4i64(<4 x i64> %xx)
   %r = add i64 %z, %a
   ret i64 %r
 }
 
 define arm_aapcs_vfpcc i64 @add_v2i32_v2i64_acc_zext(<2 x i32> %x, i64 %a) {
 ; CHECK-LABEL: add_v2i32_v2i64_acc_zext:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .save {r7, lr}
-; CHECK-NEXT:    push {r7, lr}
-; CHECK-NEXT:    vmov.i64 q1, #0xffffffff
-; CHECK-NEXT:    vand q0, q0, q1
 ; CHECK-NEXT:    vmov r2, s2
+; CHECK-NEXT:    mov.w r12, #0
 ; CHECK-NEXT:    vmov r3, s0
-; CHECK-NEXT:    vmov r12, s3
-; CHECK-NEXT:    vmov lr, s1
 ; CHECK-NEXT:    adds r2, r2, r3
-; CHECK-NEXT:    adc.w r3, lr, r12
+; CHECK-NEXT:    adc r3, r12, #0
 ; CHECK-NEXT:    adds r0, r0, r2
 ; CHECK-NEXT:    adcs r1, r3
-; CHECK-NEXT:    pop {r7, pc}
+; CHECK-NEXT:    bx lr
 entry:
   %xx = zext <2 x i32> %x to <2 x i64>
   %z = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> %xx)
   %r = add i64 %z, %a
   ret i64 %r
 }
 
 define arm_aapcs_vfpcc i64 @add_v2i32_v2i64_acc_sext(<2 x i32> %x, i64 %a) {
 ; CHECK-LABEL: add_v2i32_v2i64_acc_sext:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    vmov r2, s0
-; CHECK-NEXT:    vmov q1, q0
 ; CHECK-NEXT:    vmov r3, s2
-; CHECK-NEXT:    asrs r2, r2, #31
-; CHECK-NEXT:    vmov.32 q1[1], r2
-; CHECK-NEXT:    vmov r12, s4
-; CHECK-NEXT:    adds.w r12, r12, r3
-; CHECK-NEXT:    adc.w r2, r2, r3, asr #31
-; CHECK-NEXT:    adds.w r0, r0, r12
-; CHECK-NEXT:    adcs r1, r2
+; CHECK-NEXT:    asr.w r12, r2, #31
+; CHECK-NEXT:    adds r2, r2, r3
+; CHECK-NEXT:    adc.w r3, r12, r3, asr #31
+; CHECK-NEXT:    adds r0, r0, r2
+; CHECK-NEXT:    adcs r1, r3
 ; CHECK-NEXT:    bx lr
 entry:
   %xx = sext <2 x i32> %x to <2 x i64>
   %z = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> %xx)
   %r = add i64 %z, %a
   ret i64 %r
 }
 
entry:
   %z = call i16 @llvm.vector.reduce.add.v8i16(<8 x i16> %x)
   %r = add i16 %z, %a
   ret i16 %r
 }
 
 define arm_aapcs_vfpcc i64 @add_v8i16_v8i64_acc_zext(<8 x i16> %x, i64 %a) {
 ; CHECK-LABEL: add_v8i16_v8i64_acc_zext:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .save {r4, lr}
-; CHECK-NEXT:    push {r4, lr}
+; CHECK-NEXT:    .save {r7, lr}
+; CHECK-NEXT:    push {r7, lr}
 ; CHECK-NEXT:    vmov.u16 r2, q0[1]
 ; CHECK-NEXT:    vmov.u16 r3, q0[0]
 ; CHECK-NEXT:    vmov q2[2], q2[0], r3, r2
 ; CHECK-NEXT:    vmov.i64 q1, #0xffff
 ; CHECK-NEXT:    vand q2, q2, q1
 ; CHECK-NEXT:    vmov r2, s10
 ; CHECK-NEXT:    vmov r3, s8
 ; CHECK-NEXT:    add.w r12, r3, r2
 ; CHECK-NEXT:    vmov.u16 r3, q0[3]
 ; CHECK-NEXT:    vmov.u16 r2, q0[2]
-; CHECK-NEXT:    vmov q3[2], q3[0], r2, r3
-; CHECK-NEXT:    vand q3, q3, q1
-; CHECK-NEXT:    vmov r2, s12
-; CHECK-NEXT:    vmov r3, s14
+; CHECK-NEXT:    vmov q2[2], q2[0], r2, r3
+; CHECK-NEXT:    vand q2, q2, q1
+; CHECK-NEXT:    vmov r2, s8
+; CHECK-NEXT:    vmov r3, s10
 ; CHECK-NEXT:    add r2, r12
 ; CHECK-NEXT:    add.w r12, r2, r3
 ; CHECK-NEXT:    vmov.u16 r3, q0[5]
 ; CHECK-NEXT:    vmov.u16 r2, q0[4]
-; CHECK-NEXT:    vmov q3[2], q3[0], r2, r3
-; CHECK-NEXT:    vand q3, q3, q1
-; CHECK-NEXT:    vmov r2, s12
-; CHECK-NEXT:    vmov r3, s14
-; CHECK-NEXT:    vmov lr, s15
-; CHECK-NEXT:    add r12, r2
-; CHECK-NEXT:    vmov r2, s11
-; CHECK-NEXT:    adds.w r4, r12, r3
-; CHECK-NEXT:    vmov.u16 r3, q0[6]
-; CHECK-NEXT:    adc.w r12, r2, lr
+; CHECK-NEXT:    vmov q2[2], q2[0], r2, r3
+; CHECK-NEXT:    vand q2, q2, q1
+; CHECK-NEXT:    vmov r2, s8
+; CHECK-NEXT:    vmov r3, s10
+; CHECK-NEXT:    add r2, r12
+; CHECK-NEXT:    mov.w r12, #0
+; CHECK-NEXT:    adds.w lr, r2, r3
 ; CHECK-NEXT:    vmov.u16 r2, q0[7]
+; CHECK-NEXT:    vmov.u16 r3, q0[6]
+; CHECK-NEXT:    adc r12, r12, #0
 ; CHECK-NEXT:    vmov q0[2], q0[0], r3, r2
 ; CHECK-NEXT:    vand q0, q0, q1
-; CHECK-NEXT:    vmov r3, s0
-; CHECK-NEXT:    vmov r2, s1
-; CHECK-NEXT:    adds r3, r3, r4
-; CHECK-NEXT:    vmov r4, s3
-; CHECK-NEXT:    adc.w r12, r12, r2
+; CHECK-NEXT:    vmov r2, s0
+; CHECK-NEXT:    adds.w lr, lr, r2
 ; CHECK-NEXT:    vmov r2, s2
-; CHECK-NEXT:    adds r2, r2, r3
-; CHECK-NEXT:    adc.w r3, r12, r4
+; CHECK-NEXT:    adc r3, r12, #0
+; CHECK-NEXT:    adds.w r2, r2, lr
+; CHECK-NEXT:    adc r3, r3, #0
 ; CHECK-NEXT:    adds r0, r0, r2
 ; CHECK-NEXT:    adcs r1, r3
-; CHECK-NEXT:    pop {r4, pc}
+; CHECK-NEXT:    pop {r7, pc}
 entry:
   %xx = zext <8 x i16> %x to <8 x i64>
   %z = call i64 @llvm.vector.reduce.add.v8i64(<8 x i64> %xx)
   %r = add i64 %z, %a
   ret i64 %r
 }
 
 define arm_aapcs_vfpcc i64 @add_v8i16_v8i64_acc_sext(<8 x i16> %x, i64 %a) {
 ; CHECK-LABEL: add_v8i16_v8i64_acc_sext:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .save {r4, lr}
-; CHECK-NEXT:    push {r4, lr}
-; CHECK-NEXT:    vmov.s16 r2, q0[1]
-; CHECK-NEXT:    vmov.s16 r3, q0[0]
-; CHECK-NEXT:    vmov q1[2], q1[0], r3, r2
+; CHECK-NEXT:    .save {r7, lr}
+; CHECK-NEXT:    push {r7, lr}
+; CHECK-NEXT:    vmov.s16 r2, q0[0]
+; CHECK-NEXT:    vmov.s16 r3, q0[1]
 ; CHECK-NEXT:    asr.w r12, r2, #31
-; CHECK-NEXT:    asrs r3, r3, #31
-; CHECK-NEXT:    vmov q1[3], q1[1], r3, r12
-; CHECK-NEXT:    vmov lr, s6
-; CHECK-NEXT:    vmov r3, s4
-; CHECK-NEXT:    vmov r12, s5
-; CHECK-NEXT:    adds.w lr, lr, r3
-; CHECK-NEXT:    vmov.s16 r3, q0[2]
-; CHECK-NEXT:    adc.w r12, r12, r2, asr #31
-; CHECK-NEXT:    vmov.s16 r2, q0[3]
-; CHECK-NEXT:    vmov q1[2], q1[0], r3, r2
-; CHECK-NEXT:    asrs r4, r2, #31
-; CHECK-NEXT:    asrs r3, r3, #31
-; CHECK-NEXT:    vmov q1[3], q1[1], r3, r4
-; CHECK-NEXT:    vmov r4, s4
-; CHECK-NEXT:    vmov r3, s5
-; CHECK-NEXT:    adds.w r4, r4, lr
-; CHECK-NEXT:    adc.w r12, r12, r3
-; CHECK-NEXT:    vmov r3, s6
-; CHECK-NEXT:    adds.w lr, r4, r3
-; CHECK-NEXT:    vmov.s16 r4, q0[5]
-; CHECK-NEXT:    adc.w r12, r12, r2, asr #31
-; CHECK-NEXT:    vmov.s16 r2, q0[4]
-; CHECK-NEXT:    vmov q1[2], q1[0], r2, r4
-; CHECK-NEXT:    asrs r3, r4, #31
-; CHECK-NEXT:    asrs r2, r2, #31
-; CHECK-NEXT:    vmov q1[3], q1[1], r2, r3
-; CHECK-NEXT:    vmov r3, s4
-; CHECK-NEXT:    vmov r2, s5
-; CHECK-NEXT:    adds.w r3, r3, lr
-; CHECK-NEXT:    adc.w r12, r12, r2
-; CHECK-NEXT:    vmov r2, s6
-; CHECK-NEXT:    adds r2, r2, r3
-; CHECK-NEXT:    adc.w r3, r12, r4, asr #31
-; CHECK-NEXT:    vmov.s16 r4, q0[6]
-; CHECK-NEXT:    adds r2, r2, r4
-; CHECK-NEXT:    adc.w r3, r3, r4, asr #31
-; CHECK-NEXT:    vmov.s16 r4, q0[7]
-; CHECK-NEXT:    adds r2, r2, r4
-; CHECK-NEXT:    adc.w r3, r3, r4, asr #31
+; CHECK-NEXT:    adds.w lr, r2, r3
+; CHECK-NEXT:    vmov.s16 r2, q0[2]
+; CHECK-NEXT:    adc.w r3, r12, r3, asr #31
+; CHECK-NEXT:    adds.w r12, lr, r2
+; CHECK-NEXT:    adc.w r2, r3, r2, asr #31
+; CHECK-NEXT:    vmov.s16 r3, q0[3]
+; CHECK-NEXT:    adds.w r12, r12, r3
+; CHECK-NEXT:    adc.w r2, r2, r3, asr #31
+; CHECK-NEXT:    vmov.s16 r3, q0[4]
+; CHECK-NEXT:    adds.w r12, r12, r3
+; CHECK-NEXT:    adc.w r2, r2, r3, asr #31
+; CHECK-NEXT:    vmov.s16 r3, q0[5]
+; CHECK-NEXT:    adds.w r12, r12, r3
+; CHECK-NEXT:    adc.w r2, r2, r3, asr #31
+; CHECK-NEXT:    vmov.s16 r3, q0[6]
+; CHECK-NEXT:    adds.w r12, r12, r3
+; CHECK-NEXT:    adc.w lr, r2, r3, asr #31
+; CHECK-NEXT:    vmov.s16 r3, q0[7]
+; CHECK-NEXT:    adds.w r2, r12, r3
+; CHECK-NEXT:    adc.w r3, lr, r3, asr #31
 ; CHECK-NEXT:    adds r0, r0, r2
 ; CHECK-NEXT:    adcs r1, r3
-; CHECK-NEXT:    pop {r4, pc}
+; CHECK-NEXT:    pop {r7, pc}
 entry:
   %xx = sext <8 x i16> %x to <8 x i64>
   %z = call i64 @llvm.vector.reduce.add.v8i64(<8 x i64> %xx)
   %r = add i64 %z, %a
   ret i64 %r
 }
 
 define arm_aapcs_vfpcc i64 @add_v4i16_v4i64_acc_zext(<4 x i16> %x, i64 %a) {
entry:
   %z = call i64 @llvm.vector.reduce.add.v4i64(<4 x i64> %xx)
   %r = add i64 %z, %a
   ret i64 %r
 }
 
 define arm_aapcs_vfpcc i64 @add_v2i16_v2i64_acc_zext(<2 x i16> %x, i64 %a) {
 ; CHECK-LABEL: add_v2i16_v2i64_acc_zext:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.i64 q1, #0xffff
+; CHECK-NEXT:    vmov.i32 q1, #0xffff
 ; CHECK-NEXT:    vand q0, q0, q1
 ; CHECK-NEXT:    vmov r2, s2
 ; CHECK-NEXT:    vmov r3, s0
 ; CHECK-NEXT:    add r2, r3
-; CHECK-NEXT:    vmov r3, s3
 ; CHECK-NEXT:    adds r0, r0, r2
-; CHECK-NEXT:    adcs r1, r3
+; CHECK-NEXT:    adc r1, r1, #0
 ; CHECK-NEXT:    bx lr
 entry:
   %xx = zext <2 x i16> %x to <2 x i64>
   %z = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> %xx)
   %r = add i64 %z, %a
   ret i64 %r
 }
 
entry:
   %z = call i8 @llvm.vector.reduce.add.v16i8(<16 x i8> %x)
   %r = add i8 %z, %a
   ret i8 %r
 }
 
 define arm_aapcs_vfpcc i64 @add_v16i8_v16i64_acc_zext(<16 x i8> %x, i64 %a) {
 ; CHECK-LABEL: add_v16i8_v16i64_acc_zext:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .save {r4, lr}
-; CHECK-NEXT:    push {r4, lr}
+; CHECK-NEXT:    .save {r7, lr}
+; CHECK-NEXT:    push {r7, lr}
 ; CHECK-NEXT:    vmov.u8 r2, q0[1]
 ; CHECK-NEXT:    vmov.u8 r3, q0[0]
 ; CHECK-NEXT:    vmov q2[2], q2[0], r3, r2
 ; CHECK-NEXT:    vmov.i64 q1, #0xff
 ; CHECK-NEXT:    vand q2, q2, q1
 ; CHECK-NEXT:    vmov r2, s10
 ; CHECK-NEXT:    vmov r3, s8
 ; CHECK-NEXT:    add.w r12, r3, r2
 ; CHECK-NEXT:    vmov.u8 r3, q0[3]
 ; CHECK-NEXT:    vmov.u8 r2, q0[2]
-; CHECK-NEXT:    vmov q3[2], q3[0], r2, r3
-; CHECK-NEXT:    vand q3, q3, q1
-; CHECK-NEXT:    vmov r2, s12
-; CHECK-NEXT:    vmov r3, s14
+; CHECK-NEXT:    vmov q2[2], q2[0], r2, r3
+; CHECK-NEXT:    vand q2, q2, q1
+; CHECK-NEXT:    vmov r2, s8
+; CHECK-NEXT:    vmov r3, s10
 ; CHECK-NEXT:    add r2, r12
 ; CHECK-NEXT:    add.w r12, r2, r3
 ; CHECK-NEXT:    vmov.u8 r3, q0[5]
 ; CHECK-NEXT:    vmov.u8 r2, q0[4]
-; CHECK-NEXT:    vmov q3[2], q3[0], r2, r3
-; CHECK-NEXT:    vand q3, q3, q1
-; CHECK-NEXT:    vmov r2, s12
-; CHECK-NEXT:    vmov r3, s14
-; CHECK-NEXT:    vmov lr, s15
-; CHECK-NEXT:    add r12, r2
-; CHECK-NEXT:    vmov r2, s11
-; CHECK-NEXT:    adds.w r4, r12, r3
-; CHECK-NEXT:    vmov.u8 r3, q0[6]
-; CHECK-NEXT:    adc.w r12, r2, lr
+; CHECK-NEXT:    vmov q2[2], q2[0], r2, r3
+; CHECK-NEXT:    vand q2, q2, q1
+; CHECK-NEXT:    vmov r2, s8
+; CHECK-NEXT:    vmov r3, s10
+; CHECK-NEXT:    add r2, r12
+; CHECK-NEXT:    mov.w r12, #0
+; CHECK-NEXT:    adds.w lr, r2, r3
 ; CHECK-NEXT:    vmov.u8 r2, q0[7]
+; CHECK-NEXT:    vmov.u8 r3, q0[6]
+; CHECK-NEXT:    adc r12, r12, #0
 ; CHECK-NEXT:    vmov q2[2], q2[0], r3, r2
 ; CHECK-NEXT:    vand q2, q2, q1
-; CHECK-NEXT:    vmov r3, s8
-; CHECK-NEXT:    vmov r2, s9
-; CHECK-NEXT:    adds r3, r3, r4
-; CHECK-NEXT:    vmov r4, s11
-; CHECK-NEXT:    adc.w r12, r12, r2
+; CHECK-NEXT:    vmov r2, s8
+; CHECK-NEXT:    adds.w lr, lr, r2
 ; CHECK-NEXT:    vmov r2, s10
-; CHECK-NEXT:    adds.w lr, r3, r2
-; CHECK-NEXT:    vmov.u8 r2, q0[8]
-; CHECK-NEXT:    adc.w r3, r12, r4
-; CHECK-NEXT:    vmov.u8 r4, q0[9]
-; CHECK-NEXT:    vmov q2[2], q2[0], r2, r4
-; CHECK-NEXT:    vand q2, q2, q1
-; CHECK-NEXT:    vmov r4, s8
-; CHECK-NEXT:    vmov r2, s9
-; CHECK-NEXT:    adds.w r4, r4, lr
-; CHECK-NEXT:    adc.w r12, r3, r2
+; CHECK-NEXT:    adc r3, r12, #0
+; CHECK-NEXT:    adds.w r12, lr, r2
+; CHECK-NEXT:    vmov.u8 r2, q0[9]
+; CHECK-NEXT:    adc lr, r3, #0
+; CHECK-NEXT:    vmov.u8 r3, q0[8]
+; CHECK-NEXT:    vmov q2[2], q2[0], r3, r2
+; CHECK-NEXT:    vand q2, q2, q1
+; CHECK-NEXT:    vmov r2, s8
+; CHECK-NEXT:    adds.w r12, r12, r2
 ; CHECK-NEXT:    vmov r2, s10
-; CHECK-NEXT:    vmov r3, s11
-; CHECK-NEXT:    adds.w lr, r4, r2
-; CHECK-NEXT:    vmov.u8 r4, q0[11]
-; CHECK-NEXT:    vmov.u8 r2, q0[10]
-; CHECK-NEXT:    adc.w r3, r3, r12
-; CHECK-NEXT:    vmov q2[2], q2[0], r2, r4
-; CHECK-NEXT:    vand q2, q2, q1
-; CHECK-NEXT:    vmov r4, s8
-; CHECK-NEXT:    vmov r2, s9
-; CHECK-NEXT:    adds.w r4, r4, lr
-; CHECK-NEXT:    adc.w r12, r3, r2
+; CHECK-NEXT:    adc r3, lr, #0
+; CHECK-NEXT:    adds.w r12, r12, r2
+; CHECK-NEXT:    vmov.u8 r2, q0[11]
+; CHECK-NEXT:    adc lr, r3, #0
+; CHECK-NEXT:    vmov.u8 r3, q0[10]
+; CHECK-NEXT:    vmov q2[2], q2[0], r3, r2
+; CHECK-NEXT:    vand q2, q2, q1
+; CHECK-NEXT:    vmov r2, s8
+; CHECK-NEXT:    adds.w r12, r12, r2
 ; CHECK-NEXT:    vmov r2, s10
-; CHECK-NEXT:    vmov r3, s11
-; CHECK-NEXT:    adds.w lr, r4, r2
-; CHECK-NEXT:    vmov.u8 r4, q0[13]
-; CHECK-NEXT:    vmov.u8 r2, q0[12]
-; CHECK-NEXT:    adc.w r3, r3, r12
-; CHECK-NEXT:    vmov q2[2], q2[0], r2, r4
-; CHECK-NEXT:    vand q2, q2, q1
-; CHECK-NEXT:    vmov r4, s8
-; CHECK-NEXT:    vmov r2, s9
-; CHECK-NEXT:    adds.w r4, r4, lr
-; CHECK-NEXT:    adc.w r12, r3, r2
+; CHECK-NEXT:    adc r3, lr, #0
+; CHECK-NEXT:    adds.w r12, r12, r2
+; CHECK-NEXT:    vmov.u8 r2, q0[13]
+; CHECK-NEXT:    adc lr, r3, #0
+; CHECK-NEXT:    vmov.u8 r3, q0[12]
+; CHECK-NEXT:    vmov q2[2], q2[0], r3, r2
+; CHECK-NEXT:    vand q2, q2, q1
+; CHECK-NEXT:    vmov r2, s8
+; CHECK-NEXT:    adds.w r12, r12, r2
 ; CHECK-NEXT:    vmov r2, s10
-; CHECK-NEXT:    vmov r3, s11
-; CHECK-NEXT:    adds.w lr, r4, r2
-; CHECK-NEXT:    vmov.u8 r4, q0[15]
-; CHECK-NEXT:    vmov.u8 r2, q0[14]
-; CHECK-NEXT:    adc.w r3, r3, r12
-; CHECK-NEXT:    vmov q0[2], q0[0], r2, r4
+; CHECK-NEXT:    adc r3, lr, #0
+; CHECK-NEXT:    adds.w r12, r12, r2
+; CHECK-NEXT:    vmov.u8 r2, q0[15]
+; CHECK-NEXT:    adc lr, r3, #0
+; CHECK-NEXT:    vmov.u8 r3, q0[14]
+; CHECK-NEXT:    vmov q0[2], q0[0], r3, r2
 ; CHECK-NEXT:    vand q0, q0, q1
-; CHECK-NEXT:    vmov r4, s0
-; CHECK-NEXT:    vmov r2, s1
-; CHECK-NEXT:    adds.w r4, r4, lr
-; CHECK-NEXT:    adc.w r12, r3, r2
+; CHECK-NEXT:    vmov r2, s0
+; CHECK-NEXT:    adds.w r12, r12, r2
 ; CHECK-NEXT:    vmov r2, s2
-; CHECK-NEXT:    vmov r3, s3
-; CHECK-NEXT:    adds r2, r2, r4
-; CHECK-NEXT:    adc.w r3, r3, r12
+; CHECK-NEXT:    adc r3, lr, #0
+; CHECK-NEXT:    adds.w r2, r2, r12
+; CHECK-NEXT:    adc r3, r3, #0
 ; CHECK-NEXT:    adds r0, r0, r2
 ; CHECK-NEXT:    adcs r1, r3
-; CHECK-NEXT:    pop {r4, pc}
+; CHECK-NEXT:    pop {r7, pc}
 entry:
   %xx = zext <16 x i8> %x to <16 x i64>
   %z = call i64 @llvm.vector.reduce.add.v16i64(<16 x i64> %xx)
   %r = add i64 %z, %a
   ret i64 %r
 }
 
 define arm_aapcs_vfpcc i64 @add_v16i8_v16i64_acc_sext(<16 x i8> %x, i64 %a) {
 ; CHECK-LABEL: add_v16i8_v16i64_acc_sext:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .save {r4, lr}
-; CHECK-NEXT:    push {r4, lr}
-; CHECK-NEXT:    vmov.s8 r2, q0[1]
-; CHECK-NEXT:    vmov.s8 r3, q0[0]
-; CHECK-NEXT:    vmov q1[2], q1[0], r3, r2
+; CHECK-NEXT:    .save {r7, lr}
+; CHECK-NEXT:    push {r7, lr}
+; CHECK-NEXT:    vmov.s8 r2, q0[0]
+; CHECK-NEXT:    vmov.s8 r3, q0[1]
 ; CHECK-NEXT:    asr.w r12, r2, #31
-; CHECK-NEXT:    asrs r3, r3, #31
-; CHECK-NEXT:    vmov q1[3], q1[1], r3, r12
-; CHECK-NEXT:    vmov lr, s6
-; CHECK-NEXT:    vmov r3, s4
-; CHECK-NEXT:    vmov r12, s5
-; CHECK-NEXT:    adds.w lr, lr, r3
-; CHECK-NEXT:    vmov.s8 r3, q0[2]
-; CHECK-NEXT:    adc.w r12, r12, r2, asr #31
-; CHECK-NEXT:    vmov.s8 r2, q0[3]
-; CHECK-NEXT:    vmov q1[2], q1[0], r3, r2
-; CHECK-NEXT:    asrs r4, r2, #31
-; CHECK-NEXT:    asrs r3, r3, #31
-; CHECK-NEXT:    vmov q1[3], q1[1], r3, r4
-; CHECK-NEXT:    vmov r4, s4
-; CHECK-NEXT:    vmov r3, s5
-; CHECK-NEXT:    adds.w r4, r4, lr
-; CHECK-NEXT:    adc.w r12, r12, r3
-; CHECK-NEXT:    vmov r3, s6
-; CHECK-NEXT:    adds.w lr, r4, r3
-; CHECK-NEXT:    vmov.s8 r4, q0[5]
-; CHECK-NEXT:    adc.w r12, r12, r2, asr #31
-; CHECK-NEXT:    vmov.s8 r2, q0[4]
-; CHECK-NEXT:    vmov q1[2], q1[0], r2, r4
-; CHECK-NEXT:    asrs r3, r4, #31
-; CHECK-NEXT:    asrs r2, r2, #31
-; CHECK-NEXT:    vmov q1[3], q1[1], r2, r3
-; CHECK-NEXT:    vmov r3, s4
-; CHECK-NEXT:    vmov r2, s5
-; CHECK-NEXT:    adds.w r3, r3, lr
-; CHECK-NEXT:    adc.w r12, r12, r2
-; CHECK-NEXT:    vmov r2, s6
-; CHECK-NEXT:    adds.w lr, r3, r2
-; CHECK-NEXT:    vmov.s8 r2, q0[6]
-; CHECK-NEXT:    adc.w r12, r12, r4, asr #31
-; CHECK-NEXT:    vmov.s8 r4, q0[7]
-; CHECK-NEXT:    vmov q1[2], q1[0], r2, r4
-; CHECK-NEXT:    asrs r3, r4, #31
-; CHECK-NEXT:    asrs r2, r2, #31
-; CHECK-NEXT:    vmov q1[3], q1[1], r2, r3
-; CHECK-NEXT:    vmov r3, s4
-; CHECK-NEXT:    vmov r2, s5
-; CHECK-NEXT:    adds.w r3, r3, lr
-; CHECK-NEXT:    adc.w r12, r12, r2
-; CHECK-NEXT:    vmov r2, s6
-; CHECK-NEXT:    adds.w lr, r3, r2
-; CHECK-NEXT:    vmov.s8 r2, q0[8]
-; CHECK-NEXT:    adc.w r12, r12, r4, asr #31
-; CHECK-NEXT:    vmov.s8 r4, q0[9]
-; CHECK-NEXT:    vmov q1[2], q1[0], r2, r4
-; CHECK-NEXT:    asrs r3, r4, #31
-; CHECK-NEXT:    asrs r2, r2, #31
-; CHECK-NEXT:    vmov q1[3], q1[1], r2, r3
-; CHECK-NEXT:    vmov r3, s4
-; CHECK-NEXT:    vmov r2, s5
-; CHECK-NEXT:    adds.w r3, r3, lr
-; CHECK-NEXT:    adc.w r12, r12, r2
-; CHECK-NEXT:    vmov r2, s6
-; CHECK-NEXT:    adds.w lr, r3, r2
-; CHECK-NEXT:    vmov.s8 r2, q0[10]
-; CHECK-NEXT:    adc.w r12, r12, r4, asr #31
-; CHECK-NEXT:    vmov.s8 r4, q0[11]
-; CHECK-NEXT:    vmov q1[2], q1[0], r2, r4
-; CHECK-NEXT:    asrs r3, r4, #31
-; CHECK-NEXT:    asrs r2, r2, #31
-; CHECK-NEXT:    vmov q1[3], q1[1], r2, r3
-; CHECK-NEXT:    vmov r3, s4
-; CHECK-NEXT:    vmov r2, s5
-; CHECK-NEXT:    adds.w r3, r3, lr
-; CHECK-NEXT:    adc.w r12, r12, r2
-; CHECK-NEXT:    vmov r2, s6
-; CHECK-NEXT:    adds.w lr, r3, r2
-; CHECK-NEXT:    vmov.s8 r2, q0[12]
-; CHECK-NEXT:    adc.w r12, r12, r4, asr #31
-; CHECK-NEXT:    vmov.s8 r4, q0[13]
-; CHECK-NEXT:    vmov q1[2], q1[0], r2, r4
-; CHECK-NEXT:    asrs r3, r4, #31
-; CHECK-NEXT:    asrs r2, r2, #31
-; CHECK-NEXT:    vmov q1[3], q1[1], r2, r3
-; CHECK-NEXT:    vmov r3, s4
-; CHECK-NEXT:    vmov r2, s5
-; CHECK-NEXT:    adds.w r3, r3, lr
-; CHECK-NEXT:    adc.w r12, r12, r2
-; CHECK-NEXT:    vmov r2, s6
-; CHECK-NEXT:    adds r2, r2, r3
-; CHECK-NEXT:    adc.w r3, r12, r4, asr #31
-; CHECK-NEXT:    vmov.s8 r4, q0[14]
-; CHECK-NEXT:    adds r2, r2, r4
-; CHECK-NEXT:    adc.w r3, r3, r4, asr #31
-; CHECK-NEXT:    vmov.s8 r4, q0[15]
-; CHECK-NEXT:    adds r2, r2, r4
-; CHECK-NEXT:    adc.w r3, r3, r4, asr #31
+; CHECK-NEXT:    adds.w lr, r2, r3
+; CHECK-NEXT:    vmov.s8 r2, q0[2]
+; CHECK-NEXT:    adc.w r3, r12, r3, asr #31
+; CHECK-NEXT:    adds.w r12, lr, r2
+; CHECK-NEXT:    adc.w r2, r3, r2, asr #31
+; CHECK-NEXT:    vmov.s8 r3, q0[3]
+; CHECK-NEXT:    adds.w r12, r12, r3
+; CHECK-NEXT:    adc.w r2, r2, r3, asr #31
+; CHECK-NEXT:    vmov.s8 r3, q0[4]
+; CHECK-NEXT:    adds.w r12, r12, r3
+; CHECK-NEXT:    adc.w r2, r2, r3, asr #31
+; CHECK-NEXT:    vmov.s8 r3, q0[5]
+; CHECK-NEXT:    adds.w r12, r12, r3
+; CHECK-NEXT:    adc.w r2, r2, r3, asr #31
+; CHECK-NEXT:    vmov.s8 r3, q0[6]
+; CHECK-NEXT:    adds.w r12, r12, r3
+; CHECK-NEXT:    adc.w r2, r2, r3, asr #31
+; CHECK-NEXT:    vmov.s8 r3, q0[7]
+; CHECK-NEXT:    adds.w r12, r12, r3
+; CHECK-NEXT:    adc.w r2, r2, r3, asr #31
+; CHECK-NEXT:    vmov.s8 r3, q0[8]
+; CHECK-NEXT:    adds.w r12, r12, r3
+; CHECK-NEXT:    adc.w r2, r2, r3, asr #31
+; CHECK-NEXT:    vmov.s8 r3, q0[9]
+; CHECK-NEXT:    adds.w r12, r12, r3
+; CHECK-NEXT:    adc.w r2, r2, r3, asr #31
+; CHECK-NEXT:    vmov.s8 r3, q0[10]
+; CHECK-NEXT:    adds.w r12, r12, r3
+; CHECK-NEXT:    adc.w r2, r2, r3, asr #31
+; CHECK-NEXT:    vmov.s8 r3, q0[11]
+; CHECK-NEXT:    adds.w r12, r12, r3
+; CHECK-NEXT:    adc.w r2, r2, r3, asr #31
+; CHECK-NEXT:    vmov.s8 r3, q0[12]
+; CHECK-NEXT:    adds.w r12, r12, r3
+; CHECK-NEXT:    adc.w r2, r2, r3, asr #31
+; CHECK-NEXT:    vmov.s8 r3, q0[13]
+; CHECK-NEXT:    adds.w r12, r12, r3
+; CHECK-NEXT:    adc.w r2, r2, r3, asr #31
+; CHECK-NEXT:    vmov.s8 r3, q0[14]
+; CHECK-NEXT:    adds.w r12, r12, r3
+; CHECK-NEXT:    adc.w lr, r2, r3, asr #31
+; CHECK-NEXT:    vmov.s8 r3, q0[15]
+; CHECK-NEXT:    adds.w r2, r12, r3
+; CHECK-NEXT:    adc.w r3, lr, r3, asr #31
 ; CHECK-NEXT:    adds r0, r0, r2
 ; CHECK-NEXT:    adcs r1, r3
-; CHECK-NEXT:    pop {r4, pc}
+; CHECK-NEXT:    pop {r7, pc}
 entry:
   %xx = sext <16 x i8> %x to <16 x i64>
   %z = call i64 @llvm.vector.reduce.add.v16i64(<16 x i64> %xx)
   %r = add i64 %z, %a
   ret i64 %r
 }
 
 define arm_aapcs_vfpcc i64 @add_v8i8_v8i64_acc_zext(<8 x i8> %x, i64 %a) {
 ; CHECK-LABEL: add_v8i8_v8i64_acc_zext:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .save {r4, lr}
-; CHECK-NEXT:    push {r4, lr}
+; CHECK-NEXT:    .save {r7, lr}
+; CHECK-NEXT:    push {r7, lr}
 ; CHECK-NEXT:    vmovlb.u8 q0, q0
 ; CHECK-NEXT:    vmov.i64 q1, #0xffff
 ; CHECK-NEXT:    vmov.u16 r2, q0[1]
 ; CHECK-NEXT:    vmov.u16 r3, q0[0]
 ; CHECK-NEXT:    vmov q2[2], q2[0], r3, r2
 ; CHECK-NEXT:    vand q2, q2, q1
 ; CHECK-NEXT:    vmov r2, s10
 ; CHECK-NEXT:    vmov r3, s8
 ; CHECK-NEXT:    add.w r12, r3, r2
 ; CHECK-NEXT:    vmov.u16 r3, q0[3]
 ; CHECK-NEXT:    vmov.u16 r2, q0[2]
-; CHECK-NEXT:    vmov q3[2], q3[0], r2, r3
-; CHECK-NEXT:    vand q3, q3, q1
-; CHECK-NEXT:    vmov r2, s12
-; CHECK-NEXT:    vmov r3, s14
+; CHECK-NEXT:    vmov q2[2], q2[0], r2, r3
+; CHECK-NEXT:    vand q2, q2, q1
+; CHECK-NEXT:    vmov r2, s8
+; CHECK-NEXT:    vmov r3, s10
 ; CHECK-NEXT:    add r2, r12
 ; CHECK-NEXT:    add.w r12, r2, r3
 ; CHECK-NEXT:    vmov.u16 r3, q0[5]
 ; CHECK-NEXT:    vmov.u16 r2, q0[4]
-; CHECK-NEXT:    vmov q3[2], q3[0], r2, r3
-; CHECK-NEXT:    vand q3, q3, q1
-; CHECK-NEXT:    vmov r2, s12
-; CHECK-NEXT:    vmov r3, s14
-; CHECK-NEXT:    vmov lr, s15
-; CHECK-NEXT:    add r12, r2
-; CHECK-NEXT:    vmov r2, s11
-; CHECK-NEXT:    adds.w r4, r12, r3
-; CHECK-NEXT:    vmov.u16 r3, q0[6]
-; CHECK-NEXT:    adc.w r12, r2, lr
+; CHECK-NEXT:    vmov q2[2], q2[0], r2, r3
+; CHECK-NEXT:    vand q2, q2, q1
+; CHECK-NEXT:    vmov r2, s8
+; CHECK-NEXT:    vmov r3, s10
+; CHECK-NEXT:    add r2, r12
+; CHECK-NEXT:    mov.w r12, #0
+; CHECK-NEXT:    adds.w lr, r2, r3
 ; CHECK-NEXT:    vmov.u16 r2, q0[7]
+; CHECK-NEXT:    vmov.u16 r3, q0[6]
+; CHECK-NEXT:    adc r12, r12, #0
 ; CHECK-NEXT:    vmov q0[2], q0[0], r3, r2
 ; CHECK-NEXT:    vand q0, q0, q1
-; CHECK-NEXT:    vmov r3, s0
-; CHECK-NEXT:    vmov r2, s1
-; CHECK-NEXT:    adds r3, r3, r4
-; CHECK-NEXT:    vmov r4, s3
-; CHECK-NEXT:    adc.w r12, r12, r2
+; CHECK-NEXT:    vmov r2, s0
+; CHECK-NEXT:    adds.w lr, lr, r2
 ; CHECK-NEXT:    vmov r2, s2
-; CHECK-NEXT:    adds r2, r2, r3
-; CHECK-NEXT:    adc.w r3, r12, r4
+; CHECK-NEXT:    adc r3, r12, #0
+; CHECK-NEXT:    adds.w r2, r2, lr
+; CHECK-NEXT:    adc r3, r3, #0
 ; CHECK-NEXT:    adds r0, r0, r2
 ; CHECK-NEXT:    adcs r1, r3
-; CHECK-NEXT:    pop {r4, pc}
+; CHECK-NEXT:    pop {r7, pc}
 entry:
   %xx = zext <8 x i8> %x to <8 x i64>
   %z = call i64 @llvm.vector.reduce.add.v8i64(<8 x i64> %xx)
   %r = add i64 %z, %a
   ret i64 %r
 }
 
 define arm_aapcs_vfpcc i64 @add_v8i8_v8i64_acc_sext(<8 x i8> %x, i64 %a) {
 ; CHECK-LABEL: add_v8i8_v8i64_acc_sext:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .save {r4, lr}
-; CHECK-NEXT:    push {r4, lr}
-; CHECK-NEXT:    vmov.u16 r2, q0[1]
-; CHECK-NEXT:    vmov.u16 r3, q0[0]
+; CHECK-NEXT:    .save {r7, lr}
+; CHECK-NEXT:    push {r7, lr}
+; CHECK-NEXT:    vmov.u16 r2, q0[0]
+; CHECK-NEXT:    vmov.u16 r3, q0[1]
 ; CHECK-NEXT:    sxtb r2, r2
 ; CHECK-NEXT:    sxtb r3, r3
-; CHECK-NEXT:    vmov q1[2], q1[0], r3, r2
 ; CHECK-NEXT:    asr.w r12, r2, #31
-; CHECK-NEXT:    asrs r3, r3, #31
-; CHECK-NEXT:    vmov q1[3], q1[1], r3, r12
-; CHECK-NEXT:    vmov lr, s6
-; CHECK-NEXT:    vmov r3, s4
-; CHECK-NEXT:    vmov r12, s5
-; CHECK-NEXT:    adds.w lr, lr, r3
-; CHECK-NEXT:    vmov.u16 r3, q0[2]
-; CHECK-NEXT:    adc.w r12, r12, r2, asr #31
-; CHECK-NEXT:    vmov.u16 r2, q0[3]
+; CHECK-NEXT:    adds.w lr, r2, r3
+; CHECK-NEXT:    vmov.u16 r2, q0[2]
+; CHECK-NEXT:    adc.w r3, r12, r3, asr #31
 ; CHECK-NEXT:    sxtb r2, r2
+; CHECK-NEXT:    adds.w r12, lr, r2
+; CHECK-NEXT:    adc.w r2, r3, r2, asr #31
+; CHECK-NEXT:    vmov.u16 r3, q0[3]
 ; CHECK-NEXT:    sxtb r3, r3
-; CHECK-NEXT:    vmov q1[2], q1[0], r3, r2
-; CHECK-NEXT:    asrs r4, r2, #31
-; CHECK-NEXT:    asrs r3, r3, #31
-; CHECK-NEXT:    vmov q1[3], q1[1], r3, r4
-; CHECK-NEXT:    vmov r4, s4
-; CHECK-NEXT:    vmov r3, s5
-; CHECK-NEXT:    adds.w r4, r4, lr
-; CHECK-NEXT:    adc.w r12, r12, r3
-; CHECK-NEXT:    vmov r3, s6
-; CHECK-NEXT:    adds.w lr, r4, r3
-; CHECK-NEXT:    vmov.u16 r4, q0[5]
-; CHECK-NEXT:    adc.w r12, r12, r2, asr #31
-; CHECK-NEXT:    vmov.u16 r2, q0[4]
-; CHECK-NEXT:    sxtb r4, r4
-; CHECK-NEXT:    sxtb r2, r2
-; CHECK-NEXT:    vmov q1[2], q1[0], r2, r4
-; CHECK-NEXT:    asrs r3, r4, #31
-; CHECK-NEXT:    asrs r2, r2, #31
-; CHECK-NEXT:    vmov q1[3], q1[1], r2, r3
-; CHECK-NEXT:    vmov r3, s4
-; CHECK-NEXT:    vmov r2, s5
-; CHECK-NEXT:    adds.w r3, r3, lr
-; CHECK-NEXT:    adc.w r12, r12, r2
-; CHECK-NEXT:    vmov r2, s6
-; CHECK-NEXT:    adds r2, r2, r3
-; CHECK-NEXT:    adc.w r3, r12, r4, asr #31
-; CHECK-NEXT:    vmov.u16 r4, q0[6]
-; CHECK-NEXT:    sxtb r4, r4
-; CHECK-NEXT:    adds r2, r2, r4
-; CHECK-NEXT:    adc.w r3, r3, r4, asr #31
-; CHECK-NEXT:    vmov.u16 r4, q0[7]
-; CHECK-NEXT:    sxtb r4, r4
-; CHECK-NEXT:    adds r2, r2, r4
-; CHECK-NEXT:    adc.w r3, r3, r4, asr #31
+; CHECK-NEXT:    adds.w r12, r12, r3
+; CHECK-NEXT:    adc.w r2, r2, r3, asr #31
+; CHECK-NEXT:    vmov.u16 r3, q0[4]
+; CHECK-NEXT:    sxtb r3, r3
+; CHECK-NEXT:    adds.w r12, r12, r3
+; CHECK-NEXT:    adc.w r2, r2, r3, asr #31
+; CHECK-NEXT:    vmov.u16 r3, q0[5]
+; CHECK-NEXT:    sxtb r3, r3
+; CHECK-NEXT:    adds.w r12, r12, r3
+; CHECK-NEXT:    adc.w r2, r2, r3, asr #31
+; CHECK-NEXT:    vmov.u16 r3, q0[6]
+; CHECK-NEXT:    sxtb r3, r3
+; CHECK-NEXT:    adds.w r12, r12, r3
+; CHECK-NEXT:    adc.w lr, r2, r3, asr #31
+; CHECK-NEXT:    vmov.u16 r3, q0[7]
+; CHECK-NEXT:    sxtb r3, r3
+; CHECK-NEXT:    adds.w r2, r12, r3
+; CHECK-NEXT:    adc.w r3, lr, r3, asr #31
 ; CHECK-NEXT:    adds r0, r0, r2
 ; CHECK-NEXT:    adcs r1, r3
-; CHECK-NEXT:    pop {r4, pc}
+; CHECK-NEXT:    pop {r7, pc}
 entry:
   %xx = sext <8 x i8> %x to <8 x i64>
   %z = call i64 @llvm.vector.reduce.add.v8i64(<8 x i64> %xx)
   %r = add i64 %z, %a
   ret i64 %r
 }
 
 define arm_aapcs_vfpcc i64 @add_v4i8_v4i64_acc_zext(<4 x i8> %x, i64 %a) {
entry:
   %z = call i64 @llvm.vector.reduce.add.v4i64(<4 x i64> %xx)
   %r = add i64 %z, %a
   ret i64 %r
 }
 
 define arm_aapcs_vfpcc i64 @add_v2i8_v2i64_acc_zext(<2 x i8> %x, i64 %a) {
 ; CHECK-LABEL: add_v2i8_v2i64_acc_zext:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.i64 q1, #0xff
+; CHECK-NEXT:    vmov.i32 q1, #0xff
 ; CHECK-NEXT:    vand q0, q0, q1
 ; CHECK-NEXT:    vmov r2, s2
 ; CHECK-NEXT:    vmov r3, s0
 ; CHECK-NEXT:    add r2, r3
-; CHECK-NEXT:    vmov r3, s3
 ; CHECK-NEXT:    adds r0, r0, r2
-; CHECK-NEXT:    adcs r1, r3
+; CHECK-NEXT:    adc r1, r1, #0
 ; CHECK-NEXT:    bx lr
 entry:
   %xx = zext <2 x i8> %x to <2 x i64>
   %z = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> %xx)
   %r = add i64 %z, %a
   ret i64 %r
 }
 

llvm/test/CodeGen/Thumb2/mve-vecreduce-addpred.ll

entry:
   %z = call i64 @llvm.vector.reduce.add.v4i64(<4 x i64> %s)
   ret i64 %z
 }
 
 define arm_aapcs_vfpcc i64 @add_v2i32_v2i64_zext(<2 x i32> %x, <2 x i32> %b) {
 ; CHECK-LABEL: add_v2i32_v2i64_zext:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    vmov r0, s6
-; CHECK-NEXT:    vmov.i64 q2, #0xffffffff
+; CHECK-NEXT:    movs r2, #0
 ; CHECK-NEXT:    vmov r1, s4
-; CHECK-NEXT:    vand q0, q0, q2
 ; CHECK-NEXT:    cmp r0, #0
 ; CHECK-NEXT:    cset r0, eq
 ; CHECK-NEXT:    cmp r0, #0
 ; CHECK-NEXT:    csetm r0, ne
 ; CHECK-NEXT:    cmp r1, #0
 ; CHECK-NEXT:    cset r1, eq
 ; CHECK-NEXT:    cmp r1, #0
 ; CHECK-NEXT:    csetm r1, ne
 ; CHECK-NEXT:    vmov q1[2], q1[0], r1, r0
-; CHECK-NEXT:    vmov q1[3], q1[1], r1, r0
 ; CHECK-NEXT:    vand q0, q0, q1
 ; CHECK-NEXT:    vmov r0, s2
-; CHECK-NEXT:    vmov r3, s0
-; CHECK-NEXT:    vmov r1, s3
-; CHECK-NEXT:    vmov r2, s1
-; CHECK-NEXT:    adds r0, r0, r3
-; CHECK-NEXT:    adcs r1, r2
+; CHECK-NEXT:    vmov r1, s0
+; CHECK-NEXT:    adds r0, r0, r1
+; CHECK-NEXT:    adc r1, r2, #0
 ; CHECK-NEXT:    bx lr
 entry:
   %c = icmp eq <2 x i32> %b, zeroinitializer
   %xx = zext <2 x i32> %x to <2 x i64>
   %s = select <2 x i1> %c, <2 x i64> %xx, <2 x i64> zeroinitializer
   %z = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> %s)
   ret i64 %z
 }

 ; CHECK-NEXT:    vmov.i8 q3, #0xff
 ; CHECK-NEXT:    vcmp.i16 eq, q1, zr
 ; CHECK-NEXT:    vpsel q2, q3, q2
 ; CHECK-NEXT:    vmov.u16 r0, q2[2]
 ; CHECK-NEXT:    vmov.u16 r1, q2[0]
 ; CHECK-NEXT:    vmov q1[2], q1[0], r1, r0
 ; CHECK-NEXT:    vmov.u16 r0, q2[3]
 ; CHECK-NEXT:    vmov.u16 r1, q2[1]
+; CHECK-NEXT:    vmov.u16 r3, q2[4]
 ; CHECK-NEXT:    vmov q1[3], q1[1], r1, r0
 ; CHECK-NEXT:    vcmp.i32 ne, q1, zr
 ; CHECK-NEXT:    vmov.i64 q1, #0xffff
 ; CHECK-NEXT:    vmrs r0, p0
 ; CHECK-NEXT:    and r2, r0, #1
 ; CHECK-NEXT:    ubfx r1, r0, #4, #1
 ; CHECK-NEXT:    rsbs r2, r2, #0
 ; CHECK-NEXT:    rsbs r1, r1, #0
 ; CHECK-NEXT:    vmov q3[2], q3[0], r2, r1
-; CHECK-NEXT:    vmov q3[3], q3[1], r2, r1
 ; CHECK-NEXT:    vmov.u16 r1, q0[1]
 ; CHECK-NEXT:    vmov.u16 r2, q0[0]
 ; CHECK-NEXT:    vmov q4[2], q4[0], r2, r1
 ; CHECK-NEXT:    vand q4, q4, q1
 ; CHECK-NEXT:    vand q3, q4, q3
-; CHECK-NEXT:    vmov r1, s15
-; CHECK-NEXT:    vmov r2, s13
-; CHECK-NEXT:    vmov r3, s12
-; CHECK-NEXT:    orrs r1, r2
-; CHECK-NEXT:    vmov r2, s14
-; CHECK-NEXT:    add r2, r3
-; CHECK-NEXT:    ubfx r3, r0, #12, #1
+; CHECK-NEXT:    vmov r1, s14
+; CHECK-NEXT:    vmov r2, s12
+; CHECK-NEXT:    add r1, r2
+; CHECK-NEXT:    ubfx r2, r0, #12, #1
 ; CHECK-NEXT:    ubfx r0, r0, #8, #1
-; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    rsbs r2, r2, #0
 ; CHECK-NEXT:    rsbs r0, r0, #0
-; CHECK-NEXT:    vmov q3[2], q3[0], r0, r3
-; CHECK-NEXT:    vmov q3[3], q3[1], r0, r3
+; CHECK-NEXT:    vmov q3[2], q3[0], r0, r2
 ; CHECK-NEXT:    vmov.u16 r0, q0[3]
-; CHECK-NEXT:    vmov.u16 r3, q0[2]
-; CHECK-NEXT:    vmov q4[2], q4[0], r3, r0
+; CHECK-NEXT:    vmov.u16 r2, q0[2]
+; CHECK-NEXT:    vmov q4[2], q4[0], r2, r0
+; CHECK-NEXT:    movs r2, #0
 ; CHECK-NEXT:    vand q4, q4, q1
 ; CHECK-NEXT:    vand q3, q4, q3
-; CHECK-NEXT:    vmov r3, s12
-; CHECK-NEXT:    vmov r0, s13
-; CHECK-NEXT:    adds r2, r2, r3
-; CHECK-NEXT:    vmov r3, s14
-; CHECK-NEXT:    adcs r0, r1
-; CHECK-NEXT:    vmov r1, s15
-; CHECK-NEXT:    adds r2, r2, r3
-; CHECK-NEXT:    vmov.u16 r3, q2[4]
-; CHECK-NEXT:    adc.w r12, r0, r1
-; CHECK-NEXT:    vmov.u16 r1, q2[6]
-; CHECK-NEXT:    vmov q3[2], q3[0], r3, r1
-; CHECK-NEXT:    vmov.u16 r1, q2[7]
+; CHECK-NEXT:    vmov r0, s12
+; CHECK-NEXT:    adds r0, r0, r1
+; CHECK-NEXT:    adc r1, r2, #0
+; CHECK-NEXT:    vmov r2, s14
+; CHECK-NEXT:    adds.w r12, r0, r2
+; CHECK-NEXT:    vmov.u16 r2, q2[6]
+; CHECK-NEXT:    vmov q3[2], q3[0], r3, r2
+; CHECK-NEXT:    vmov.u16 r2, q2[7]
 ; CHECK-NEXT:    vmov.u16 r3, q2[5]
-; CHECK-NEXT:    vmov q3[3], q3[1], r3, r1
+; CHECK-NEXT:    adc r1, r1, #0
+; CHECK-NEXT:    vmov q3[3], q3[1], r3, r2
 ; CHECK-NEXT:    vcmp.i32 ne, q3, zr
-; CHECK-NEXT:    vmrs r1, p0
-; CHECK-NEXT:    and r0, r1, #1
-; CHECK-NEXT:    ubfx r3, r1, #4, #1
+; CHECK-NEXT:    vmrs r2, p0
+; CHECK-NEXT:    and r0, r2, #1
+; CHECK-NEXT:    ubfx r3, r2, #4, #1
 ; CHECK-NEXT:    rsbs r0, r0, #0
 ; CHECK-NEXT:    rsbs r3, r3, #0
 ; CHECK-NEXT:    vmov q2[2], q2[0], r0, r3
-; CHECK-NEXT:    vmov q2[3], q2[1], r0, r3
 ; CHECK-NEXT:    vmov.u16 r0, q0[5]
 ; CHECK-NEXT:    vmov.u16 r3, q0[4]
 ; CHECK-NEXT:    vmov q3[2], q3[0], r3, r0
 ; CHECK-NEXT:    vand q3, q3, q1
 ; CHECK-NEXT:    vand q2, q3, q2
-; CHECK-NEXT:    vmov r3, s8
-; CHECK-NEXT:    vmov r0, s9
-; CHECK-NEXT:    adds r2, r2, r3
-; CHECK-NEXT:    vmov r3, s11
-; CHECK-NEXT:    adc.w r12, r12, r0
-; CHECK-NEXT:    vmov r0, s10
-; CHECK-NEXT:    adds r0, r0, r2
-; CHECK-NEXT:    adc.w r2, r12, r3
-; CHECK-NEXT:    ubfx r3, r1, #12, #1
-; CHECK-NEXT:    ubfx r1, r1, #8, #1
-; CHECK-NEXT:    rsbs r3, r3, #0
-; CHECK-NEXT:    rsbs r1, r1, #0
-; CHECK-NEXT:    vmov q2[2], q2[0], r1, r3
-; CHECK-NEXT:    vmov q2[3], q2[1], r1, r3
-; CHECK-NEXT:    vmov.u16 r1, q0[7]
+; CHECK-NEXT:    vmov r0, s8
+; CHECK-NEXT:    vmov r3, s10
+; CHECK-NEXT:    adds.w r0, r0, r12
+; CHECK-NEXT:    adc r1, r1, #0
+; CHECK-NEXT:    adds r0, r0, r3
+; CHECK-NEXT:    ubfx r3, r2, #12, #1
+; CHECK-NEXT:    ubfx r2, r2, #8, #1
+; CHECK-NEXT:    rsb.w r3, r3, #0
+; CHECK-NEXT:    rsb.w r2, r2, #0
+; CHECK-NEXT:    vmov q2[2], q2[0], r2, r3
+; CHECK-NEXT:    vmov.u16 r2, q0[7]
 ; CHECK-NEXT:    vmov.u16 r3, q0[6]
-; CHECK-NEXT:    vmov q0[2], q0[0], r3, r1
+; CHECK-NEXT:    adc r1, r1, #0
+; CHECK-NEXT:    vmov q0[2], q0[0], r3, r2
 ; CHECK-NEXT:    vand q0, q0, q1
 ; CHECK-NEXT:    vand q0, q0, q2
-; CHECK-NEXT:    vmov r3, s0
-; CHECK-NEXT:    vmov r1, s1
-; CHECK-NEXT:    adds r0, r0, r3
-; CHECK-NEXT:    vmov r3, s2
-; CHECK-NEXT:    adcs r1, r2
-; CHECK-NEXT:    vmov r2, s3
-; CHECK-NEXT:    adds r0, r0, r3
-; CHECK-NEXT:    adcs r1, r2
+; CHECK-NEXT:    vmov r2, s0
+; CHECK-NEXT:    adds r0, r0, r2
+; CHECK-NEXT:    vmov r2, s2
+; CHECK-NEXT:    adc r1, r1, #0
+; CHECK-NEXT:    adds r0, r0, r2
+; CHECK-NEXT:    adc r1, r1, #0
 ; CHECK-NEXT:    vpop {d8, d9}
 ; CHECK-NEXT:    bx lr
 entry:
   %c = icmp eq <8 x i16> %b, zeroinitializer
   %xx = zext <8 x i16> %x to <8 x i64>
   %s = select <8 x i1> %c, <8 x i64> %xx, <8 x i64> zeroinitializer
   %z = call i64 @llvm.vector.reduce.add.v8i64(<8 x i64> %s)
   ret i64 %z

 ; CHECK-NEXT:    cset r0, eq
 ; CHECK-NEXT:    cmp r0, #0
 ; CHECK-NEXT:    csetm r0, ne
 ; CHECK-NEXT:    cmp r1, #0
 ; CHECK-NEXT:    cset r1, eq
 ; CHECK-NEXT:    cmp r1, #0
 ; CHECK-NEXT:    csetm r1, ne
 ; CHECK-NEXT:    vmov q1[2], q1[0], r1, r0
-; CHECK-NEXT:    vmov q1[3], q1[1], r1, r0
 ; CHECK-NEXT:    vand q0, q0, q1
 ; CHECK-NEXT:    vmov r0, s2
 ; CHECK-NEXT:    vmov r1, s0
-; CHECK-NEXT:    vmov r2, s1
 ; CHECK-NEXT:    add r0, r1
-; CHECK-NEXT:    vmov r1, s3
-; CHECK-NEXT:    orrs r1, r2
+; CHECK-NEXT:    movs r1, #0
 ; CHECK-NEXT:    bx lr
 entry:
   %c = icmp eq <2 x i16> %b, zeroinitializer
   %xx = zext <2 x i16> %x to <2 x i64>
   %s = select <2 x i1> %c, <2 x i64> %xx, <2 x i64> zeroinitializer
   %z = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> %s)
   ret i64 %z
 }

 ; CHECK-NEXT:    vmov.16 q1[7], r0
 ; CHECK-NEXT:    vcmp.i16 ne, q1, zr
 ; CHECK-NEXT:    vpsel q5, q3, q2
 ; CHECK-NEXT:    vmov.u16 r0, q5[2]
 ; CHECK-NEXT:    vmov.u16 r1, q5[0]
 ; CHECK-NEXT:    vmov q1[2], q1[0], r1, r0
 ; CHECK-NEXT:    vmov.u16 r0, q5[3]
 ; CHECK-NEXT:    vmov.u16 r1, q5[1]
+; CHECK-NEXT:    vmov.u16 r3, q5[4]
 ; CHECK-NEXT:    vmov q1[3], q1[1], r1, r0
 ; CHECK-NEXT:    vcmp.i32 ne, q1, zr
 ; CHECK-NEXT:    vmov.i64 q1, #0xff
 ; CHECK-NEXT:    vmrs r0, p0
 ; CHECK-NEXT:    and r2, r0, #1
 ; CHECK-NEXT:    ubfx r1, r0, #4, #1
 ; CHECK-NEXT:    rsbs r2, r2, #0
 ; CHECK-NEXT:    rsbs r1, r1, #0
 ; CHECK-NEXT:    vmov q6[2], q6[0], r2, r1
-; CHECK-NEXT:    vmov q6[3], q6[1], r2, r1
 ; CHECK-NEXT:    vmov.u8 r1, q0[1]
 ; CHECK-NEXT:    vmov.u8 r2, q0[0]
 ; CHECK-NEXT:    vmov q7[2], q7[0], r2, r1
 ; CHECK-NEXT:    vand q7, q7, q1
 ; CHECK-NEXT:    vand q6, q7, q6
-; CHECK-NEXT:    vmov r1, s27
-; CHECK-NEXT:    vmov r2, s25
-; CHECK-NEXT:    vmov r3, s24
-; CHECK-NEXT:    orrs r1, r2
-; CHECK-NEXT:    vmov r2, s26
-; CHECK-NEXT:    add r2, r3
-; CHECK-NEXT:    ubfx r3, r0, #12, #1
+; CHECK-NEXT:    vmov r1, s26
+; CHECK-NEXT:    vmov r2, s24
+; CHECK-NEXT:    add r1, r2
+; CHECK-NEXT:    ubfx r2, r0, #12, #1
 ; CHECK-NEXT:    ubfx r0, r0, #8, #1
-; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    rsbs r2, r2, #0
 ; CHECK-NEXT:    rsbs r0, r0, #0
-; CHECK-NEXT:    vmov q6[2], q6[0], r0, r3
-; CHECK-NEXT:    vmov q6[3], q6[1], r0, r3
+; CHECK-NEXT:    vmov q6[2], q6[0], r0, r2
 ; CHECK-NEXT:    vmov.u8 r0, q0[3]
-; CHECK-NEXT:    vmov.u8 r3, q0[2]
-; CHECK-NEXT:    vmov q7[2], q7[0], r3, r0
+; CHECK-NEXT:    vmov.u8 r2, q0[2]
+; CHECK-NEXT:    vmov q7[2], q7[0], r2, r0
+; CHECK-NEXT:    movs r2, #0
 ; CHECK-NEXT:    vand q7, q7, q1
 ; CHECK-NEXT:    vand q6, q7, q6
-; CHECK-NEXT:    vmov r3, s24
-; CHECK-NEXT:    vmov r0, s25
-; CHECK-NEXT:    adds r2, r2, r3
-; CHECK-NEXT:    vmov r3, s26
-; CHECK-NEXT:    adcs r0, r1
-; CHECK-NEXT:    vmov r1, s27
-; CHECK-NEXT:    adds r2, r2, r3
-; CHECK-NEXT:    vmov.u16 r3, q5[4]
-; CHECK-NEXT:    adc.w r12, r0, r1
-; CHECK-NEXT:    vmov.u16 r1, q5[6]
-; CHECK-NEXT:    vmov q6[2], q6[0], r3, r1
-; CHECK-NEXT:    vmov.u16 r1, q5[7]
+; CHECK-NEXT:    vmov r0, s24
+; CHECK-NEXT:    adds r0, r0, r1
+; CHECK-NEXT:    adc r1, r2, #0
+; CHECK-NEXT:    vmov r2, s26
+; CHECK-NEXT:    adds.w r12, r0, r2
+; CHECK-NEXT:    vmov.u16 r2, q5[6]
+; CHECK-NEXT:    vmov q6[2], q6[0], r3, r2
+; CHECK-NEXT:    vmov.u16 r2, q5[7]
 ; CHECK-NEXT:    vmov.u16 r3, q5[5]
-; CHECK-NEXT:    vmov q6[3], q6[1], r3, r1
+; CHECK-NEXT:    adc r1, r1, #0
+; CHECK-NEXT:    vmov q6[3], q6[1], r3, r2
 ; CHECK-NEXT:    vcmp.i32 ne, q6, zr
-; CHECK-NEXT:    vmrs r1, p0
-; CHECK-NEXT:    and r0, r1, #1
-; CHECK-NEXT:    ubfx r3, r1, #4, #1
+; CHECK-NEXT:    vmrs r2, p0
+; CHECK-NEXT:    and r0, r2, #1
+; CHECK-NEXT:    ubfx r3, r2, #4, #1
 ; CHECK-NEXT:    rsbs r0, r0, #0
 ; CHECK-NEXT:    rsbs r3, r3, #0
 ; CHECK-NEXT:    vmov q5[2], q5[0], r0, r3
-; CHECK-NEXT:    vmov q5[3], q5[1], r0, r3
 ; CHECK-NEXT:    vmov.u8 r0, q0[5]
 ; CHECK-NEXT:    vmov.u8 r3, q0[4]
 ; CHECK-NEXT:    vmov q6[2], q6[0], r3, r0
 ; CHECK-NEXT:    vand q6, q6, q1
 ; CHECK-NEXT:    vand q5, q6, q5
-; CHECK-NEXT:    vmov r3, s20
-; CHECK-NEXT:    vmov r0, s21
-; CHECK-NEXT:    adds r2, r2, r3
-; CHECK-NEXT:    vmov r3, s23
-; CHECK-NEXT:    adc.w r12, r12, r0
-; CHECK-NEXT:    vmov r0, s22
-; CHECK-NEXT:    adds r0, r0, r2
-; CHECK-NEXT:    adc.w r2, r12, r3
-; CHECK-NEXT:    ubfx r3, r1, #12, #1
-; CHECK-NEXT:    ubfx r1, r1, #8, #1
-; CHECK-NEXT:    rsbs r3, r3, #0
-; CHECK-NEXT:    rsbs r1, r1, #0
-; CHECK-NEXT:    vmov q5[2], q5[0], r1, r3
-; CHECK-NEXT:    vmov q5[3], q5[1], r1, r3
-; CHECK-NEXT:    vmov.u8 r1, q0[7]
+; CHECK-NEXT:    vmov r0, s20
+; CHECK-NEXT:    vmov r3, s22
+; CHECK-NEXT:    adds.w r0, r0, r12
+; CHECK-NEXT:    adc r1, r1, #0
+; CHECK-NEXT:    adds r0, r0, r3
+; CHECK-NEXT:    ubfx r3, r2, #12, #1
+; CHECK-NEXT:    ubfx r2, r2, #8, #1
+; CHECK-NEXT:    rsb.w r3, r3, #0
+; CHECK-NEXT:    rsb.w r2, r2, #0
+; CHECK-NEXT:    vmov q5[2], q5[0], r2, r3
+; CHECK-NEXT:    vmov.u8 r2, q0[7]
 ; CHECK-NEXT:    vmov.u8 r3, q0[6]
-; CHECK-NEXT:    vmov q6[2], q6[0], r3, r1
+; CHECK-NEXT:    adc r1, r1, #0
+; CHECK-NEXT:    vmov q6[2], q6[0], r3, r2
 ; CHECK-NEXT:    vand q6, q6, q1
 ; CHECK-NEXT:    vand q5, q6, q5
-; CHECK-NEXT:    vmov r3, s20
-; CHECK-NEXT:    vmov r1, s21
-; CHECK-NEXT:    adds r0, r0, r3
-; CHECK-NEXT:    vmov r3, s22
-; CHECK-NEXT:    adcs r1, r2
-; CHECK-NEXT:    vmov r2, s23
-; CHECK-NEXT:    adds.w r12, r0, r3
-; CHECK-NEXT:    adcs r1, r2
+; CHECK-NEXT:    vmov r2, s20
+; CHECK-NEXT:    adds r0, r0, r2
+; CHECK-NEXT:    vmov r2, s22
+; CHECK-NEXT:    adc r1, r1, #0
+; CHECK-NEXT:    adds.w r12, r0, r2
 ; CHECK-NEXT:    vmov.u8 r2, q4[8]
 ; CHECK-NEXT:    vmov.16 q5[0], r2
 ; CHECK-NEXT:    vmov.u8 r2, q4[9]
 ; CHECK-NEXT:    vmov.16 q5[1], r2
 ; CHECK-NEXT:    vmov.u8 r2, q4[10]
 ; CHECK-NEXT:    vmov.16 q5[2], r2
 ; CHECK-NEXT:    vmov.u8 r2, q4[11]
 ; CHECK-NEXT:    vmov.16 q5[3], r2
 ; CHECK-NEXT:    vmov.u8 r2, q4[12]
 ; CHECK-NEXT:    vmov.16 q5[4], r2
 ; CHECK-NEXT:    vmov.u8 r2, q4[13]
 ; CHECK-NEXT:    vmov.16 q5[5], r2
 ; CHECK-NEXT:    vmov.u8 r2, q4[14]
 ; CHECK-NEXT:    vmov.16 q5[6], r2
 ; CHECK-NEXT:    vmov.u8 r2, q4[15]
 ; CHECK-NEXT:    vmov.16 q5[7], r2
+; CHECK-NEXT:    adc r1, r1, #0
 ; CHECK-NEXT:    vcmp.i16 ne, q5, zr
 ; CHECK-NEXT:    vpsel q2, q3, q2
 ; CHECK-NEXT:    vmov.u16 r2, q2[2]
 ; CHECK-NEXT:    vmov.u16 r3, q2[0]
 ; CHECK-NEXT:    vmov q3[2], q3[0], r3, r2
 ; CHECK-NEXT:    vmov.u16 r2, q2[3]
 ; CHECK-NEXT:    vmov.u16 r3, q2[1]
 ; CHECK-NEXT:    vmov q3[3], q3[1], r3, r2
 ; CHECK-NEXT:    vcmp.i32 ne, q3, zr
 ; CHECK-NEXT:    vmrs r2, p0
 ; CHECK-NEXT:    and r0, r2, #1
 ; CHECK-NEXT:    ubfx r3, r2, #4, #1
 ; CHECK-NEXT:    rsbs r0, r0, #0
 ; CHECK-NEXT:    rsbs r3, r3, #0
 ; CHECK-NEXT:    vmov q3[2], q3[0], r0, r3
-; CHECK-NEXT:    vmov q3[3], q3[1], r0, r3
 ; CHECK-NEXT:    vmov.u8 r0, q0[9]
 ; CHECK-NEXT:    vmov.u8 r3, q0[8]
 ; CHECK-NEXT:    vmov q4[2], q4[0], r3, r0
 ; CHECK-NEXT:    vand q4, q4, q1
 ; CHECK-NEXT:    vand q3, q4, q3
-; CHECK-NEXT:    vmov r3, s12
-; CHECK-NEXT:    vmov r0, s13
-; CHECK-NEXT:    adds.w r3, r3, r12
-; CHECK-NEXT:    adc.w r12, r1, r0
-; CHECK-NEXT:    vmov r0, s14
-; CHECK-NEXT:    vmov r1, s15
+; CHECK-NEXT:    vmov r0, s12
+; CHECK-NEXT:    vmov r3, s14
+; CHECK-NEXT:    adds.w r0, r0, r12
+; CHECK-NEXT:    adc r1, r1, #0
 ; CHECK-NEXT:    adds r0, r0, r3
 ; CHECK-NEXT:    ubfx r3, r2, #12, #1
 ; CHECK-NEXT:    ubfx r2, r2, #8, #1
 ; CHECK-NEXT:    rsb.w r3, r3, #0
 ; CHECK-NEXT:    rsb.w r2, r2, #0
-; CHECK-NEXT:    adc.w r1, r1, r12
 ; CHECK-NEXT:    vmov q3[2], q3[0], r2, r3
-; CHECK-NEXT:    vmov q3[3], q3[1], r2, r3
 ; CHECK-NEXT:    vmov.u8 r2, q0[11]
 ; CHECK-NEXT:    vmov.u8 r3, q0[10]
+; CHECK-NEXT:    adc r1, r1, #0
 ; CHECK-NEXT:    vmov q4[2], q4[0], r3, r2
+; CHECK-NEXT:    vmov.u16 r3, q2[4]
 ; CHECK-NEXT:    vand q4, q4, q1
 ; CHECK-NEXT:    vand q3, q4, q3
-; CHECK-NEXT:    vmov r3, s12
-; CHECK-NEXT:    vmov r2, s13
-; CHECK-NEXT:    adds r0, r0, r3
-; CHECK-NEXT:    vmov r3, s14
-; CHECK-NEXT:    adcs r1, r2
-; CHECK-NEXT:    vmov r2, s15
-; CHECK-NEXT:    adds.w r12, r0, r3
-; CHECK-NEXT:    vmov.u16 r3, q2[4]
-; CHECK-NEXT:    adcs r1, r2
+; CHECK-NEXT:    vmov r2, s12
+; CHECK-NEXT:    adds r0, r0, r2
+; CHECK-NEXT:    vmov r2, s14
+; CHECK-NEXT:    adc r1, r1, #0
+; CHECK-NEXT:    adds.w r12, r0, r2
 ; CHECK-NEXT:    vmov.u16 r2, q2[6]
 ; CHECK-NEXT:    vmov q3[2], q3[0], r3, r2
 ; CHECK-NEXT:    vmov.u16 r2, q2[7]
 ; CHECK-NEXT:    vmov.u16 r3, q2[5]
+; CHECK-NEXT:    adc r1, r1, #0
 ; CHECK-NEXT:    vmov q3[3], q3[1], r3, r2
 ; CHECK-NEXT:    vcmp.i32 ne, q3, zr
 ; CHECK-NEXT:    vmrs r2, p0
 ; CHECK-NEXT:    and r0, r2, #1
 ; CHECK-NEXT:    ubfx r3, r2, #4, #1
 ; CHECK-NEXT:    rsbs r0, r0, #0
 ; CHECK-NEXT:    rsbs r3, r3, #0
 ; CHECK-NEXT:    vmov q2[2], q2[0], r0, r3
-; CHECK-NEXT:    vmov q2[3], q2[1], r0, r3
 ; CHECK-NEXT:    vmov.u8 r0, q0[13]
 ; CHECK-NEXT:    vmov.u8 r3, q0[12]
 ; CHECK-NEXT:    vmov q3[2], q3[0], r3, r0
 ; CHECK-NEXT:    vand q3, q3, q1
 ; CHECK-NEXT:    vand q2, q3, q2
-; CHECK-NEXT:    vmov r3, s8
-; CHECK-NEXT:    vmov r0, s9
-; CHECK-NEXT:    adds.w r3, r3, r12
-; CHECK-NEXT:    adc.w r12, r1, r0
-; CHECK-NEXT:    vmov r0, s10
-; CHECK-NEXT:    vmov r1, s11
+; CHECK-NEXT:    vmov r0, s8
+; CHECK-NEXT:    vmov r3, s10
+; CHECK-NEXT:    adds.w r0, r0, r12
+; CHECK-NEXT:    adc r1, r1, #0
 ; CHECK-NEXT:    adds r0, r0, r3
 ; CHECK-NEXT:    ubfx r3, r2, #12, #1
 ; CHECK-NEXT:    ubfx r2, r2, #8, #1
 ; CHECK-NEXT:    rsb.w r3, r3, #0
 ; CHECK-NEXT:    rsb.w r2, r2, #0
-; CHECK-NEXT:    adc.w r1, r1, r12
 ; CHECK-NEXT:    vmov q2[2], q2[0], r2, r3
-; CHECK-NEXT:    vmov q2[3], q2[1], r2, r3
 ; CHECK-NEXT:    vmov.u8 r2, q0[15]
 ; CHECK-NEXT:    vmov.u8 r3, q0[14]
+; CHECK-NEXT:    adc r1, r1, #0
 ; CHECK-NEXT:    vmov q0[2], q0[0], r3, r2
 ; CHECK-NEXT:    vand q0, q0, q1
 ; CHECK-NEXT:    vand q0, q0, q2
-; CHECK-NEXT:    vmov r3, s0
-; CHECK-NEXT:    vmov r2, s1
-; CHECK-NEXT:    adds r0, r0, r3
-; CHECK-NEXT:    vmov r3, s2
-; CHECK-NEXT:    adcs r1, r2
-; CHECK-NEXT:    vmov r2, s3
-; CHECK-NEXT:    adds r0, r0, r3
-; CHECK-NEXT:    adcs r1, r2
+; CHECK-NEXT:    vmov r2, s0
+; CHECK-NEXT:    adds r0, r0, r2
+; CHECK-NEXT:    vmov r2, s2
+; CHECK-NEXT:    adc r1, r1, #0
+; CHECK-NEXT:    adds r0, r0, r2
+; CHECK-NEXT:    adc r1, r1, #0
 ; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13, d14, d15}
 ; CHECK-NEXT:    bx lr
 entry:
   %c = icmp eq <16 x i8> %b, zeroinitializer
   %xx = zext <16 x i8> %x to <16 x i64>
   %s = select <16 x i1> %c, <16 x i64> %xx, <16 x i64> zeroinitializer
   %z = call i64 @llvm.vector.reduce.add.v16i64(<16 x i64> %s)
   ret i64 %z

 ; CHECK-NEXT:    vmov.i8 q1, #0x0
 ; CHECK-NEXT:    vpsel q2, q2, q1
 ; CHECK-NEXT:    vmovlb.u8 q0, q0
 ; CHECK-NEXT:    vmov.u16 r0, q2[2]
 ; CHECK-NEXT:    vmov.u16 r1, q2[0]
 ; CHECK-NEXT:    vmov q1[2], q1[0], r1, r0
 ; CHECK-NEXT:    vmov.u16 r0, q2[3]
 ; CHECK-NEXT:    vmov.u16 r1, q2[1]
+; CHECK-NEXT:    vmov.u16 r3, q2[4]
 ; CHECK-NEXT:    vmov q1[3], q1[1], r1, r0
 ; CHECK-NEXT:    vcmp.i32 ne, q1, zr
 ; CHECK-NEXT:    vmov.i64 q1, #0xffff
 ; CHECK-NEXT:    vmrs r0, p0
 ; CHECK-NEXT:    and r2, r0, #1
 ; CHECK-NEXT:    ubfx r1, r0, #4, #1
 ; CHECK-NEXT:    rsbs r2, r2, #0
 ; CHECK-NEXT:    rsbs r1, r1, #0
 ; CHECK-NEXT:    vmov q3[2], q3[0], r2, r1
-; CHECK-NEXT:    vmov q3[3], q3[1], r2, r1
 ; CHECK-NEXT:    vmov.u16 r1, q0[1]
 ; CHECK-NEXT:    vmov.u16 r2, q0[0]
 ; CHECK-NEXT:    vmov q4[2], q4[0], r2, r1
 ; CHECK-NEXT:    vand q4, q4, q1
 ; CHECK-NEXT:    vand q3, q4, q3
-; CHECK-NEXT:    vmov r1, s15
-; CHECK-NEXT:    vmov r2, s13
-; CHECK-NEXT:    vmov r3, s12
-; CHECK-NEXT:    orrs r1, r2
-; CHECK-NEXT:    vmov r2, s14
-; CHECK-NEXT:    add r2, r3
-; CHECK-NEXT:    ubfx r3, r0, #12, #1
+; CHECK-NEXT:    vmov r1, s14
+; CHECK-NEXT:    vmov r2, s12
+; CHECK-NEXT:    add r1, r2
+; CHECK-NEXT:    ubfx r2, r0, #12, #1
 ; CHECK-NEXT:    ubfx r0, r0, #8, #1
-; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    rsbs r2, r2, #0
 ; CHECK-NEXT:    rsbs r0, r0, #0
-; CHECK-NEXT:    vmov q3[2], q3[0], r0, r3
-; CHECK-NEXT:    vmov q3[3], q3[1], r0, r3
+; CHECK-NEXT:    vmov q3[2], q3[0], r0, r2
 ; CHECK-NEXT:    vmov.u16 r0, q0[3]
-; CHECK-NEXT:    vmov.u16 r3, q0[2]
-; CHECK-NEXT:    vmov q4[2], q4[0], r3, r0
+; CHECK-NEXT:    vmov.u16 r2, q0[2]
+; CHECK-NEXT:    vmov q4[2], q4[0], r2, r0
+; CHECK-NEXT:    movs r2, #0
 ; CHECK-NEXT:    vand q4, q4, q1
 ; CHECK-NEXT:    vand q3, q4, q3
-; CHECK-NEXT:    vmov r3, s12
-; CHECK-NEXT:    vmov r0, s13
-; CHECK-NEXT:    adds r2, r2, r3
-; CHECK-NEXT:    vmov r3, s14
-; CHECK-NEXT:    adcs r0, r1
-; CHECK-NEXT:    vmov r1, s15
-; CHECK-NEXT:    adds r2, r2, r3
-; CHECK-NEXT:    vmov.u16 r3, q2[4]
-; CHECK-NEXT:    adc.w r12, r0, r1
-; CHECK-NEXT:    vmov.u16 r1, q2[6]
-; CHECK-NEXT:    vmov q3[2], q3[0], r3, r1
-; CHECK-NEXT:    vmov.u16 r1, q2[7]
+; CHECK-NEXT:    vmov r0, s12
+; CHECK-NEXT:    adds r0, r0, r1
+; CHECK-NEXT:    adc r1, r2, #0
+; CHECK-NEXT:    vmov r2, s14
+; CHECK-NEXT:    adds.w r12, r0, r2
+; CHECK-NEXT:    vmov.u16 r2, q2[6]
+; CHECK-NEXT:    vmov q3[2], q3[0], r3, r2
+; CHECK-NEXT:    vmov.u16 r2, q2[7]
 ; CHECK-NEXT:    vmov.u16 r3, q2[5]
-; CHECK-NEXT:    vmov q3[3], q3[1], r3, r1
+; CHECK-NEXT:    adc r1, r1, #0
+; CHECK-NEXT:    vmov q3[3], q3[1], r3, r2
 ; CHECK-NEXT:    vcmp.i32 ne, q3, zr
-; CHECK-NEXT:    vmrs r1, p0
-; CHECK-NEXT:    and r0, r1, #1
-; CHECK-NEXT:    ubfx r3, r1, #4, #1
+; CHECK-NEXT:    vmrs r2, p0
+; CHECK-NEXT:    and r0, r2, #1
+; CHECK-NEXT:    ubfx r3, r2, #4, #1
 ; CHECK-NEXT:    rsbs r0, r0, #0
 ; CHECK-NEXT:    rsbs r3, r3, #0
 ; CHECK-NEXT:    vmov q2[2], q2[0], r0, r3
-; CHECK-NEXT:    vmov q2[3], q2[1], r0, r3
 ; CHECK-NEXT:    vmov.u16 r0, q0[5]
 ; CHECK-NEXT:    vmov.u16 r3, q0[4]
 ; CHECK-NEXT:    vmov q3[2], q3[0], r3, r0
 ; CHECK-NEXT:    vand q3, q3, q1
 ; CHECK-NEXT:    vand q2, q3, q2
-; CHECK-NEXT:    vmov r3, s8
-; CHECK-NEXT:    vmov r0, s9
-; CHECK-NEXT:    adds r2, r2, r3
-; CHECK-NEXT:    vmov r3, s11
-; CHECK-NEXT:    adc.w r12, r12, r0
-; CHECK-NEXT:    vmov r0, s10
-; CHECK-NEXT:    adds r0, r0, r2
-; CHECK-NEXT:    adc.w r2, r12, r3
-; CHECK-NEXT:    ubfx r3, r1, #12, #1
-; CHECK-NEXT:    ubfx r1, r1, #8, #1
-; CHECK-NEXT:    rsbs r3, r3, #0
-; CHECK-NEXT:    rsbs r1, r1, #0
-; CHECK-NEXT:    vmov q2[2], q2[0], r1, r3
-; CHECK-NEXT:    vmov q2[3], q2[1], r1, r3
-; CHECK-NEXT:    vmov.u16 r1, q0[7]
+; CHECK-NEXT:    vmov r0, s8
+; CHECK-NEXT:    vmov r3, s10
+; CHECK-NEXT:    adds.w r0, r0, r12
+; CHECK-NEXT:    adc r1, r1, #0
+; CHECK-NEXT:    adds r0, r0, r3
+; CHECK-NEXT:    ubfx r3, r2, #12, #1
+; CHECK-NEXT:    ubfx r2, r2, #8, #1
+; CHECK-NEXT:    rsb.w r3, r3, #0
+; CHECK-NEXT:    rsb.w r2, r2, #0
+; CHECK-NEXT:    vmov q2[2], q2[0], r2, r3
+; CHECK-NEXT:    vmov.u16 r2, q0[7]
 ; CHECK-NEXT:    vmov.u16 r3, q0[6]
-; CHECK-NEXT:    vmov q0[2], q0[0], r3, r1
+; CHECK-NEXT:    adc r1, r1, #0
+; CHECK-NEXT:    vmov q0[2], q0[0], r3, r2
 ; CHECK-NEXT:    vand q0, q0, q1
 ; CHECK-NEXT:    vand q0, q0, q2
-; CHECK-NEXT:    vmov r3, s0
-; CHECK-NEXT:    vmov r1, s1
-; CHECK-NEXT:    adds r0, r0, r3
-; CHECK-NEXT:    vmov r3, s2
-; CHECK-NEXT:    adcs r1, r2
-; CHECK-NEXT:    vmov r2, s3
-; CHECK-NEXT:    adds r0, r0, r3
-; CHECK-NEXT:    adcs r1, r2
+; CHECK-NEXT:    vmov r2, s0
+; CHECK-NEXT:    adds r0, r0, r2
+; CHECK-NEXT:    vmov r2, s2
+; CHECK-NEXT:    adc r1, r1, #0
+; CHECK-NEXT:    adds r0, r0, r2
+; CHECK-NEXT:    adc r1, r1, #0
 ; CHECK-NEXT:    vpop {d8, d9}
 ; CHECK-NEXT:    bx lr
 entry:
   %c = icmp eq <8 x i8> %b, zeroinitializer
   %xx = zext <8 x i8> %x to <8 x i64>
   %s = select <8 x i1> %c, <8 x i64> %xx, <8 x i64> zeroinitializer
   %z = call i64 @llvm.vector.reduce.add.v8i64(<8 x i64> %s)
   ret i64 %z

 ; CHECK-NEXT:    cset r0, eq
 ; CHECK-NEXT:    cmp r0, #0
 ; CHECK-NEXT:    csetm r0, ne
 ; CHECK-NEXT:    cmp r1, #0
 ; CHECK-NEXT:    cset r1, eq
 ; CHECK-NEXT:    cmp r1, #0
 ; CHECK-NEXT:    csetm r1, ne
 ; CHECK-NEXT:    vmov q1[2], q1[0], r1, r0
-; CHECK-NEXT:    vmov q1[3], q1[1], r1, r0
 ; CHECK-NEXT:    vand q0, q0, q1
 ; CHECK-NEXT:    vmov r0, s2
 ; CHECK-NEXT:    vmov r1, s0
-; CHECK-NEXT:    vmov r2, s1
 ; CHECK-NEXT:    add r0, r1
-; CHECK-NEXT:    vmov r1, s3
-; CHECK-NEXT:    orrs r1, r2
+; CHECK-NEXT:    movs r1, #0
 ; CHECK-NEXT:    bx lr
 entry:
   %c = icmp eq <2 x i8> %b, zeroinitializer
   %xx = zext <2 x i8> %x to <2 x i64>
   %s = select <2 x i1> %c, <2 x i64> %xx, <2 x i64> zeroinitializer
   %z = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> %s)
   ret i64 %z
 }
entry:
   %z = call i64 @llvm.vector.reduce.add.v4i64(<4 x i64> %s)
   %r = add i64 %z, %a
   ret i64 %r
 }
 
 define arm_aapcs_vfpcc i64 @add_v2i32_v2i64_acc_zext(<2 x i32> %x, <2 x i32> %b, i64 %a) {
 ; CHECK-LABEL: add_v2i32_v2i64_acc_zext:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .save {r7, lr}
-; CHECK-NEXT:    push {r7, lr}
 ; CHECK-NEXT:    vmov r2, s6
-; CHECK-NEXT:    vmov.i64 q2, #0xffffffff
+; CHECK-NEXT:    mov.w r12, #0
 ; CHECK-NEXT:    vmov r3, s4
-; CHECK-NEXT:    vand q0, q0, q2
 ; CHECK-NEXT:    cmp r2, #0
 ; CHECK-NEXT:    cset r2, eq
 ; CHECK-NEXT:    cmp r2, #0
 ; CHECK-NEXT:    csetm r2, ne
 ; CHECK-NEXT:    cmp r3, #0
 ; CHECK-NEXT:    cset r3, eq
 ; CHECK-NEXT:    cmp r3, #0
 ; CHECK-NEXT:    csetm r3, ne
 ; CHECK-NEXT:    vmov q1[2], q1[0], r3, r2
-; CHECK-NEXT:    vmov q1[3], q1[1], r3, r2
 ; CHECK-NEXT:    vand q0, q0, q1
 ; CHECK-NEXT:    vmov r2, s2
 ; CHECK-NEXT:    vmov r3, s0
-; CHECK-NEXT:    vmov r12, s3
-; CHECK-NEXT:    vmov lr, s1
 ; CHECK-NEXT:    adds r2, r2, r3
-; CHECK-NEXT:    adc.w r3, lr, r12
+; CHECK-NEXT:    adc r3, r12, #0
 ; CHECK-NEXT:    adds r0, r0, r2
 ; CHECK-NEXT:    adcs r1, r3
-; CHECK-NEXT:    pop {r7, pc}
+; CHECK-NEXT:    bx lr
 entry:
   %c = icmp eq <2 x i32> %b, zeroinitializer
   %xx = zext <2 x i32> %x to <2 x i64>
   %s = select <2 x i1> %c, <2 x i64> %xx, <2 x i64> zeroinitializer
   %z = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> %s)
   %r = add i64 %z, %a
   ret i64 %r
 }

 ; CHECK-NEXT:    vcmp.i32 ne, q1, zr
 ; CHECK-NEXT:    vmov.i64 q1, #0xffff
 ; CHECK-NEXT:    vmrs r2, p0
 ; CHECK-NEXT:    ubfx r3, r2, #4, #1
 ; CHECK-NEXT:    rsb.w r12, r3, #0
 ; CHECK-NEXT:    and r3, r2, #1
 ; CHECK-NEXT:    rsbs r3, r3, #0
 ; CHECK-NEXT:    vmov q3[2], q3[0], r3, r12
-; CHECK-NEXT:    vmov q3[3], q3[1], r3, r12
 ; CHECK-NEXT:    vmov.u16 r12, q0[1]
 ; CHECK-NEXT:    vmov.u16 r3, q0[0]
 ; CHECK-NEXT:    vmov q4[2], q4[0], r3, r12
 ; CHECK-NEXT:    vand q4, q4, q1
 ; CHECK-NEXT:    vand q3, q4, q3
-; CHECK-NEXT:    vmov r12, s15
-; CHECK-NEXT:    vmov r3, s13
-; CHECK-NEXT:    vmov lr, s14
-; CHECK-NEXT:    orr.w r12, r12, r3
+; CHECK-NEXT:    vmov r12, s14
 ; CHECK-NEXT:    vmov r3, s12
-; CHECK-NEXT:    add lr, r3
+; CHECK-NEXT:    add r12, r3
 ; CHECK-NEXT:    ubfx r3, r2, #12, #1
 ; CHECK-NEXT:    ubfx r2, r2, #8, #1
 ; CHECK-NEXT:    rsbs r3, r3, #0
 ; CHECK-NEXT:    rsbs r2, r2, #0
 ; CHECK-NEXT:    vmov q3[2], q3[0], r2, r3
-; CHECK-NEXT:    vmov q3[3], q3[1], r2, r3
 ; CHECK-NEXT:    vmov.u16 r2, q0[3]
 ; CHECK-NEXT:    vmov.u16 r3, q0[2]
 ; CHECK-NEXT:    vmov q4[2], q4[0], r3, r2
+; CHECK-NEXT:    movs r3, #0
 ; CHECK-NEXT:    vand q4, q4, q1
 ; CHECK-NEXT:    vand q3, q4, q3
-; CHECK-NEXT:    vmov r3, s12
-; CHECK-NEXT:    vmov r2, s13
-; CHECK-NEXT:    adds.w lr, lr, r3
-; CHECK-NEXT:    vmov r3, s14
-; CHECK-NEXT:    adc.w r12, r12, r2
-; CHECK-NEXT:    vmov r2, s15
-; CHECK-NEXT:    adds.w lr, lr, r3
-; CHECK-NEXT:    vmov.u16 r3, q2[4]
-; CHECK-NEXT:    adc.w r12, r12, r2
+; CHECK-NEXT:    vmov r2, s12
+; CHECK-NEXT:    adds.w r12, r12, r2
+; CHECK-NEXT:    vmov r2, s14
+; CHECK-NEXT:    adc r3, r3, #0
+; CHECK-NEXT:    adds.w r12, r12, r2
 ; CHECK-NEXT:    vmov.u16 r2, q2[6]
+; CHECK-NEXT:    adc lr, r3, #0
+; CHECK-NEXT:    vmov.u16 r3, q2[4]
 ; CHECK-NEXT:    vmov q3[2], q3[0], r3, r2
 ; CHECK-NEXT:    vmov.u16 r2, q2[7]
 ; CHECK-NEXT:    vmov.u16 r3, q2[5]
 ; CHECK-NEXT:    vmov q3[3], q3[1], r3, r2
 ; CHECK-NEXT:    vcmp.i32 ne, q3, zr
 ; CHECK-NEXT:    vmrs r2, p0
 ; CHECK-NEXT:    and r4, r2, #1
 ; CHECK-NEXT:    ubfx r3, r2, #4, #1
 ; CHECK-NEXT:    rsbs r4, r4, #0
 ; CHECK-NEXT:    rsbs r3, r3, #0
 ; CHECK-NEXT:    vmov q2[2], q2[0], r4, r3
-; CHECK-NEXT:    vmov q2[3], q2[1], r4, r3
 ; CHECK-NEXT:    vmov.u16 r3, q0[5]
 ; CHECK-NEXT:    vmov.u16 r4, q0[4]
 ; CHECK-NEXT:    vmov q3[2], q3[0], r4, r3
 ; CHECK-NEXT:    vand q3, q3, q1
 ; CHECK-NEXT:    vand q2, q3, q2
-; CHECK-NEXT:    vmov r4, s8
-; CHECK-NEXT:    vmov r3, s9
-; CHECK-NEXT:    adds.w lr, lr, r4
-; CHECK-NEXT:    vmov r4, s10
-; CHECK-NEXT:    adc.w r12, r12, r3
-; CHECK-NEXT:    vmov r3, s11
-; CHECK-NEXT:    adds.w r4, r4, lr
-; CHECK-NEXT:    adc.w r12, r12, r3
+; CHECK-NEXT:    vmov r3, s8
+; CHECK-NEXT:    adds.w r12, r12, r3
+; CHECK-NEXT:    vmov r3, s10
+; CHECK-NEXT:    adc r4, lr, #0
+; CHECK-NEXT:    adds.w r12, r12, r3
 ; CHECK-NEXT:    ubfx r3, r2, #12, #1
 ; CHECK-NEXT:    ubfx r2, r2, #8, #1
-; CHECK-NEXT:    rsbs r3, r3, #0
-; CHECK-NEXT:    rsbs r2, r2, #0
+; CHECK-NEXT:    rsb.w r3, r3, #0
+; CHECK-NEXT:    rsb.w r2, r2, #0
+; CHECK-NEXT:    adc r4, r4, #0
 ; CHECK-NEXT:    vmov q2[2], q2[0], r2, r3
-; CHECK-NEXT:    vmov q2[3], q2[1], r2, r3
 ; CHECK-NEXT:    vmov.u16 r2, q0[7]
 ; CHECK-NEXT:    vmov.u16 r3, q0[6]
 ; CHECK-NEXT:    vmov q0[2], q0[0], r3, r2
 ; CHECK-NEXT:    vand q0, q0, q1
 ; CHECK-NEXT:    vand q0, q0, q2
-; CHECK-NEXT:    vmov r3, s0
-; CHECK-NEXT:    vmov r2, s1
-; CHECK-NEXT:    adds r3, r3, r4
-; CHECK-NEXT:    vmov r4, s3
-; CHECK-NEXT:    adc.w r12, r12, r2
-; CHECK-NEXT:    vmov r2, s2
-; CHECK-NEXT:    adds r2, r2, r3
-; CHECK-NEXT:    adc.w r3, r12, r4
+; CHECK-NEXT:    vmov r2, s0
+; CHECK-NEXT:    adds.w r2, r2, r12
+; CHECK-NEXT:    adc r3, r4, #0
+; CHECK-NEXT:    vmov r4, s2
+; CHECK-NEXT:    adds r2, r2, r4
+; CHECK-NEXT:    adc r3, r3, #0
 ; CHECK-NEXT:    adds r0, r0, r2
 ; CHECK-NEXT:    adcs r1, r3
 ; CHECK-NEXT:    vpop {d8, d9}
 ; CHECK-NEXT:    pop {r4, pc}
 entry:
   %c = icmp eq <8 x i16> %b, zeroinitializer
   %xx = zext <8 x i16> %x to <8 x i64>
   %s = select <8 x i1> %c, <8 x i64> %xx, <8 x i64> zeroinitializer

 ; CHECK-NEXT:    cset r2, eq
 ; CHECK-NEXT:    cmp r2, #0
 ; CHECK-NEXT:    csetm r2, ne
 ; CHECK-NEXT:    cmp r3, #0
 ; CHECK-NEXT:    cset r3, eq
 ; CHECK-NEXT:    cmp r3, #0
 ; CHECK-NEXT:    csetm r3, ne
 ; CHECK-NEXT:    vmov q1[2], q1[0], r3, r2
-; CHECK-NEXT:    vmov q1[3], q1[1], r3, r2
 ; CHECK-NEXT:    vand q0, q0, q1
-; CHECK-NEXT:    vmov r2, s3
-; CHECK-NEXT:    vmov r3, s1
-; CHECK-NEXT:    orr.w r12, r3, r2
-; CHECK-NEXT:    vmov r3, s2
-; CHECK-NEXT:    vmov r2, s0
+; CHECK-NEXT:    vmov r2, s2
+; CHECK-NEXT:    vmov r3, s0
 ; CHECK-NEXT:    add r2, r3
 ; CHECK-NEXT:    adds r0, r0, r2
-; CHECK-NEXT:    adc.w r1, r1, r12
+; CHECK-NEXT:    adc r1, r1, #0
 ; CHECK-NEXT:    bx lr
 entry:
   %c = icmp eq <2 x i16> %b, zeroinitializer
   %xx = zext <2 x i16> %x to <2 x i64>
   %s = select <2 x i1> %c, <2 x i64> %xx, <2 x i64> zeroinitializer
   %z = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> %s)
   %r = add i64 %z, %a
   ret i64 %r
entry:
   %z = call i8 @llvm.vector.reduce.add.v16i8(<16 x i8> %s)
   %r = add i8 %z, %a
   ret i8 %r
 }
 
 define arm_aapcs_vfpcc i64 @add_v16i8_v16i64_acc_zext(<16 x i8> %x, <16 x i8> %b, i64 %a) {
 ; CHECK-LABEL: add_v16i8_v16i64_acc_zext:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .save {r4, r5, r7, lr}
-; CHECK-NEXT:    push {r4, r5, r7, lr}
+; CHECK-NEXT:    .save {r4, lr}
+; CHECK-NEXT:    push {r4, lr}
 ; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13, d14, d15}
 ; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13, d14, d15}
 ; CHECK-NEXT:    vcmp.i8 eq, q1, zr
 ; CHECK-NEXT:    vmov.i8 q2, #0x0
 ; CHECK-NEXT:    vmov.i8 q3, #0xff
 ; CHECK-NEXT:    vpsel q4, q3, q2
 ; CHECK-NEXT:    vmov.u8 r2, q4[0]
 ; CHECK-NEXT:    vmov.16 q1[0], r2

 ; CHECK-NEXT:    vcmp.i32 ne, q1, zr
 ; CHECK-NEXT:    vmov.i64 q1, #0xff
 ; CHECK-NEXT:    vmrs r2, p0
 ; CHECK-NEXT:    ubfx r3, r2, #4, #1
 ; CHECK-NEXT:    rsb.w r12, r3, #0
 ; CHECK-NEXT:    and r3, r2, #1
 ; CHECK-NEXT:    rsbs r3, r3, #0
 ; CHECK-NEXT:    vmov q6[2], q6[0], r3, r12
-; CHECK-NEXT:    vmov q6[3], q6[1], r3, r12
 ; CHECK-NEXT:    vmov.u8 r12, q0[1]
 ; CHECK-NEXT:    vmov.u8 r3, q0[0]
 ; CHECK-NEXT:    vmov q7[2], q7[0], r3, r12
 ; CHECK-NEXT:    vand q7, q7, q1
 ; CHECK-NEXT:    vand q6, q7, q6
-; CHECK-NEXT:    vmov r12, s27
-; CHECK-NEXT:    vmov r3, s25
-; CHECK-NEXT:    vmov lr, s26
-; CHECK-NEXT:    orr.w r12, r12, r3
+; CHECK-NEXT:    vmov r12, s26
 ; CHECK-NEXT:    vmov r3, s24
-; CHECK-NEXT:    add lr, r3
+; CHECK-NEXT:    add r12, r3
 ; CHECK-NEXT:    ubfx r3, r2, #12, #1
 ; CHECK-NEXT:    ubfx r2, r2, #8, #1
 ; CHECK-NEXT:    rsbs r3, r3, #0
 ; CHECK-NEXT:    rsbs r2, r2, #0
 ; CHECK-NEXT:    vmov q6[2], q6[0], r2, r3
-; CHECK-NEXT:    vmov q6[3], q6[1], r2, r3
 ; CHECK-NEXT:    vmov.u8 r2, q0[3]
 ; CHECK-NEXT:    vmov.u8 r3, q0[2]
 ; CHECK-NEXT:    vmov q7[2], q7[0], r3, r2
+; CHECK-NEXT:    movs r3, #0
 ; CHECK-NEXT:    vand q7, q7, q1
 ; CHECK-NEXT:    vand q6, q7, q6
-; CHECK-NEXT:    vmov r3, s24
-; CHECK-NEXT:    vmov r2, s25
-; CHECK-NEXT:    adds.w lr, lr, r3
-; CHECK-NEXT:    vmov r3, s26
-; CHECK-NEXT:    adc.w r12, r12, r2
-; CHECK-NEXT:    vmov r2, s27
-; CHECK-NEXT:    adds.w lr, lr, r3
-; CHECK-NEXT:    vmov.u16 r3, q5[4]
-; CHECK-NEXT:    adc.w r12, r12, r2
+; CHECK-NEXT:    vmov r2, s24
+; CHECK-NEXT:    adds.w r12, r12, r2
+; CHECK-NEXT:    vmov r2, s26
+; CHECK-NEXT:    adc r3, r3, #0
+; CHECK-NEXT:    adds.w r12, r12, r2
 ; CHECK-NEXT:    vmov.u16 r2, q5[6]
+; CHECK-NEXT:    adc lr, r3, #0
+; CHECK-NEXT:    vmov.u16 r3, q5[4]
 ; CHECK-NEXT:    vmov q6[2], q6[0], r3, r2
 ; CHECK-NEXT:    vmov.u16 r2, q5[7]
 ; CHECK-NEXT:    vmov.u16 r3, q5[5]
 ; CHECK-NEXT:    vmov q6[3], q6[1], r3, r2
 ; CHECK-NEXT:    vcmp.i32 ne, q6, zr
 ; CHECK-NEXT:    vmrs r2, p0
 ; CHECK-NEXT:    and r4, r2, #1
 ; CHECK-NEXT:    ubfx r3, r2, #4, #1
 ; CHECK-NEXT:    rsbs r4, r4, #0
 ; CHECK-NEXT:    rsbs r3, r3, #0
 ; CHECK-NEXT:    vmov q5[2], q5[0], r4, r3
-; CHECK-NEXT:    vmov q5[3], q5[1], r4, r3
 ; CHECK-NEXT:    vmov.u8 r3, q0[5]
 ; CHECK-NEXT:    vmov.u8 r4, q0[4]
 ; CHECK-NEXT:    vmov q6[2], q6[0], r4, r3
 ; CHECK-NEXT:    vand q6, q6, q1
 ; CHECK-NEXT:    vand q5, q6, q5
-; CHECK-NEXT:    vmov r4, s20
-; CHECK-NEXT:    vmov r3, s21
-; CHECK-NEXT:    adds.w lr, lr, r4
-; CHECK-NEXT:    vmov r4, s22
-; CHECK-NEXT:    adc.w r12, r12, r3
-; CHECK-NEXT:    vmov r3, s23
-; CHECK-NEXT:    adds.w r4, r4, lr
-; CHECK-NEXT:    adc.w r12, r12, r3
+; CHECK-NEXT:    vmov r3, s20
+; CHECK-NEXT:    adds.w r12, r12, r3
+; CHECK-NEXT:    vmov r3, s22
+; CHECK-NEXT:    adc r4, lr, #0
+; CHECK-NEXT:    adds.w r12, r12, r3
 ; CHECK-NEXT:    ubfx r3, r2, #12, #1
 ; CHECK-NEXT:    ubfx r2, r2, #8, #1
-; CHECK-NEXT:    rsbs r3, r3, #0
-; CHECK-NEXT:    rsbs r2, r2, #0
+; CHECK-NEXT:    rsb.w r3, r3, #0
+; CHECK-NEXT:    rsb.w r2, r2, #0
+; CHECK-NEXT:    adc r4, r4, #0
 ; CHECK-NEXT:    vmov q5[2], q5[0], r2, r3
-; CHECK-NEXT:    vmov q5[3], q5[1], r2, r3
 ; CHECK-NEXT:    vmov.u8 r2, q0[7]
 ; CHECK-NEXT:    vmov.u8 r3, q0[6]
 ; CHECK-NEXT:    vmov q6[2], q6[0], r3, r2
 ; CHECK-NEXT:    vand q6, q6, q1
 ; CHECK-NEXT:    vand q5, q6, q5
-; CHECK-NEXT:    vmov r3, s20
-; CHECK-NEXT:    vmov r2, s21
-; CHECK-NEXT:    adds.w lr, r4, r3
-; CHECK-NEXT:    vmov r3, s22
-; CHECK-NEXT:    adc.w r4, r12, r2
-; CHECK-NEXT:    vmov r2, s23
-; CHECK-NEXT:    adds.w r12, lr, r3
-; CHECK-NEXT:    adc.w lr, r4, r2
-; CHECK-NEXT:    vmov.u8 r2, q4[8]
-; CHECK-NEXT:    vmov.16 q5[0], r2
-; CHECK-NEXT:    vmov.u8 r2, q4[9]
-; CHECK-NEXT:    vmov.16 q5[1], r2
-; CHECK-NEXT:    vmov.u8 r2, q4[10]
-; CHECK-NEXT:    vmov.16 q5[2], r2
-; CHECK-NEXT:    vmov.u8 r2, q4[11]
-; CHECK-NEXT:    vmov.16 q5[3], r2
-; CHECK-NEXT:    vmov.u8 r2, q4[12]
-; CHECK-NEXT:    vmov.16 q5[4], r2
-; CHECK-NEXT:    vmov.u8 r2, q4[13]
-; CHECK-NEXT:    vmov.16 q5[5], r2
-; CHECK-NEXT:    vmov.u8 r2, q4[14]
-; CHECK-NEXT:    vmov.16 q5[6], r2
-; CHECK-NEXT:    vmov.u8 r2, q4[15]
-; CHECK-NEXT:    vmov.16 q5[7], r2
+; CHECK-NEXT:    vmov r2, s20
+; CHECK-NEXT:    adds.w r2, r2, r12
+; CHECK-NEXT:    adc r3, r4, #0
+; CHECK-NEXT:    vmov r4, s22
+; CHECK-NEXT:    adds.w r12, r2, r4
+; CHECK-NEXT:    vmov.u8 r4, q4[8]
+; CHECK-NEXT:    vmov.16 q5[0], r4
+; CHECK-NEXT:    vmov.u8 r4, q4[9]
+; CHECK-NEXT:    vmov.16 q5[1], r4
+; CHECK-NEXT:    vmov.u8 r4, q4[10]
+; CHECK-NEXT:    vmov.16 q5[2], r4
+; CHECK-NEXT:    vmov.u8 r4, q4[11]
+; CHECK-NEXT:    vmov.16 q5[3], r4
+; CHECK-NEXT:    vmov.u8 r4, q4[12]
+; CHECK-NEXT:    vmov.16 q5[4], r4
+; CHECK-NEXT:    vmov.u8 r4, q4[13]
+; CHECK-NEXT:    vmov.16 q5[5], r4
+; CHECK-NEXT:    vmov.u8 r4, q4[14]
+; CHECK-NEXT:    vmov.16 q5[6], r4
+; CHECK-NEXT:    vmov.u8 r4, q4[15]
+; CHECK-NEXT:    vmov.16 q5[7], r4
+; CHECK-NEXT:    adc lr, r3, #0
 ; CHECK-NEXT:    vcmp.i16 ne, q5, zr
 ; CHECK-NEXT:    vpsel q2, q3, q2
-; CHECK-NEXT:    vmov.u16 r2, q2[2]
-; CHECK-NEXT:    vmov.u16 r4, q2[0]
-; CHECK-NEXT:    vmov q3[2], q3[0], r4, r2
+; CHECK-NEXT:    vmov.u16 r4, q2[2]
+; CHECK-NEXT:    vmov.u16 r2, q2[0]
+; CHECK-NEXT:    vmov q3[2], q3[0], r2, r4
 ; CHECK-NEXT:    vmov.u16 r2, q2[3]
 ; CHECK-NEXT:    vmov.u16 r4, q2[1]
 ; CHECK-NEXT:    vmov q3[3], q3[1], r4, r2
 ; CHECK-NEXT:    vcmp.i32 ne, q3, zr
 ; CHECK-NEXT:    vmrs r2, p0
 ; CHECK-NEXT:    and r3, r2, #1
 ; CHECK-NEXT:    ubfx r4, r2, #4, #1
 ; CHECK-NEXT:    rsbs r3, r3, #0
 ; CHECK-NEXT:    rsbs r4, r4, #0
 ; CHECK-NEXT:    vmov q3[2], q3[0], r3, r4
-; CHECK-NEXT:    vmov q3[3], q3[1], r3, r4
 ; CHECK-NEXT:    vmov.u8 r3, q0[9]
 ; CHECK-NEXT:    vmov.u8 r4, q0[8]
 ; CHECK-NEXT:    vmov q4[2], q4[0], r4, r3
 ; CHECK-NEXT:    vand q4, q4, q1
 ; CHECK-NEXT:    vand q3, q4, q3
-; CHECK-NEXT:    vmov r4, s12
-; CHECK-NEXT:    vmov r3, s13
-; CHECK-NEXT:    adds.w r5, r12, r4
-; CHECK-NEXT:    vmov r4, s14
-; CHECK-NEXT:    adc.w r12, lr, r3
-; CHECK-NEXT:    vmov r3, s15
-; CHECK-NEXT:    adds r5, r5, r4
-; CHECK-NEXT:    ubfx r4, r2, #12, #1
+; CHECK-NEXT:    vmov r3, s12
+; CHECK-NEXT:    adds.w r12, r12, r3
+; CHECK-NEXT:    vmov r3, s14
+; CHECK-NEXT:    adc r4, lr, #0
+; CHECK-NEXT:    adds.w r12, r12, r3
+; CHECK-NEXT:    ubfx r3, r2, #12, #1
 ; CHECK-NEXT:    ubfx r2, r2, #8, #1
-; CHECK-NEXT:    rsb.w r4, r4, #0
+; CHECK-NEXT:    rsb.w r3, r3, #0
 ; CHECK-NEXT:    rsb.w r2, r2, #0
-; CHECK-NEXT:    adc.w r3, r3, r12
-; CHECK-NEXT:    vmov q3[2], q3[0], r2, r4
-; CHECK-NEXT:    vmov q3[3], q3[1], r2, r4
+; CHECK-NEXT:    adc r4, r4, #0
+; CHECK-NEXT:    vmov q3[2], q3[0], r2, r3
 ; CHECK-NEXT:    vmov.u8 r2, q0[11]
-; CHECK-NEXT:    vmov.u8 r4, q0[10]
-; CHECK-NEXT:    vmov q4[2], q4[0], r4, r2
+; CHECK-NEXT:    vmov.u8 r3, q0[10]
+; CHECK-NEXT:    vmov q4[2], q4[0], r3, r2
 ; CHECK-NEXT:    vand q4, q4, q1
 ; CHECK-NEXT:    vand q3, q4, q3
-; CHECK-NEXT:    vmov r4, s12
-; CHECK-NEXT:    vmov r2, s13
-; CHECK-NEXT:    adds r5, r5, r4
+; CHECK-NEXT:    vmov r2, s12
+; CHECK-NEXT:    adds.w r2, r2, r12
+; CHECK-NEXT:    adc r3, r4, #0
 ; CHECK-NEXT:    vmov r4, s14
-; CHECK-NEXT:    adcs r2, r3
-; CHECK-NEXT:    vmov r3, s15
-; CHECK-NEXT:    adds r5, r5, r4
-; CHECK-NEXT:    vmov.u16 r4, q2[4]
-; CHECK-NEXT:    adc.w r12, r2, r3
-; CHECK-NEXT:    vmov.u16 r3, q2[6]
-; CHECK-NEXT:    vmov q3[2], q3[0], r4, r3
-; CHECK-NEXT:    vmov.u16 r3, q2[7]
+; CHECK-NEXT:    adds.w r12, r2, r4
+; CHECK-NEXT:    vmov.u16 r4, q2[6]
+; CHECK-NEXT:    vmov.u16 r2, q2[4]
+; CHECK-NEXT:    adc lr, r3, #0
+; CHECK-NEXT:    vmov q3[2], q3[0], r2, r4
+; CHECK-NEXT:    vmov.u16 r2, q2[7]
 ; CHECK-NEXT:    vmov.u16 r4, q2[5]
-; CHECK-NEXT:    vmov q3[3], q3[1], r4, r3
+; CHECK-NEXT:    vmov q3[3], q3[1], r4, r2
 ; CHECK-NEXT:    vcmp.i32 ne, q3, zr
-; CHECK-NEXT:    vmrs r3, p0
-; CHECK-NEXT:    and r2, r3, #1
-; CHECK-NEXT:    ubfx r4, r3, #4, #1
-; CHECK-NEXT:    rsbs r2, r2, #0
+; CHECK-NEXT:    vmrs r2, p0
+; CHECK-NEXT:    and r3, r2, #1
+; CHECK-NEXT:    ubfx r4, r2, #4, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
 ; CHECK-NEXT:    rsbs r4, r4, #0
-; CHECK-NEXT:    vmov q2[2], q2[0], r2, r4
-; CHECK-NEXT:    vmov q2[3], q2[1], r2, r4
-; CHECK-NEXT:    vmov.u8 r2, q0[13]
+; CHECK-NEXT:    vmov q2[2], q2[0], r3, r4
+; CHECK-NEXT:    vmov.u8 r3, q0[13]
 ; CHECK-NEXT:    vmov.u8 r4, q0[12]
-; CHECK-NEXT:    vmov q3[2], q3[0], r4, r2
+; CHECK-NEXT:    vmov q3[2], q3[0], r4, r3
 ; CHECK-NEXT:    vand q3, q3, q1
 ; CHECK-NEXT:    vand q2, q3, q2
-; CHECK-NEXT:    vmov r4, s8
-; CHECK-NEXT:    vmov r2, s9
-; CHECK-NEXT:    adds r5, r5, r4
-; CHECK-NEXT:    vmov r4, s11
-; CHECK-NEXT:    adc.w r12, r12, r2
-; CHECK-NEXT:    vmov r2, s10
-; CHECK-NEXT:    adds r2, r2, r5
-; CHECK-NEXT:    adc.w r5, r12, r4
-; CHECK-NEXT:    ubfx r4, r3, #12, #1
-; CHECK-NEXT:    ubfx r3, r3, #8, #1
-; CHECK-NEXT:    rsbs r4, r4, #0
-; CHECK-NEXT:    rsbs r3, r3, #0
-; CHECK-NEXT:    vmov q2[2], q2[0], r3, r4
-; CHECK-NEXT:    vmov q2[3], q2[1], r3, r4
-; CHECK-NEXT:    vmov.u8 r3, q0[15]
-; CHECK-NEXT:    vmov.u8 r4, q0[14]
-; CHECK-NEXT:    vmov q0[2], q0[0], r4, r3
+; CHECK-NEXT:    vmov r3, s8
+; CHECK-NEXT:    adds.w r12, r12, r3
+; CHECK-NEXT:    vmov r3, s10
+; CHECK-NEXT:    adc r4, lr, #0
+; CHECK-NEXT:    adds.w r12, r12, r3
+; CHECK-NEXT:    ubfx r3, r2, #12, #1
+; CHECK-NEXT:    ubfx r2, r2, #8, #1
+; CHECK-NEXT:    rsb.w r3, r3, #0
+; CHECK-NEXT:    rsb.w r2, r2, #0
+; CHECK-NEXT:    adc r4, r4, #0
+; CHECK-NEXT:    vmov q2[2], q2[0], r2, r3
+; CHECK-NEXT:    vmov.u8 r2, q0[15]
+; CHECK-NEXT:    vmov.u8 r3, q0[14]
+; CHECK-NEXT:    vmov q0[2], q0[0], r3, r2
 ; CHECK-NEXT:    vand q0, q0, q1
 ; CHECK-NEXT:    vand q0, q0, q2
-; CHECK-NEXT:    vmov r4, s0
-; CHECK-NEXT:    vmov r3, s1
-; CHECK-NEXT:    adds r2, r2, r4
+; CHECK-NEXT:    vmov r2, s0
+; CHECK-NEXT:    adds.w r2, r2, r12
+; CHECK-NEXT:    adc r3, r4, #0
 ; CHECK-NEXT:    vmov r4, s2
-; CHECK-NEXT:    adcs r3, r5
-; CHECK-NEXT:    vmov r5, s3
 ; CHECK-NEXT:    adds r2, r2, r4
-; CHECK-NEXT:    adcs r3, r5
+; CHECK-NEXT:    adc r3, r3, #0
 ; CHECK-NEXT:    adds r0, r0, r2
 ; CHECK-NEXT:    adcs r1, r3
 ; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13, d14, d15}
-; CHECK-NEXT:    pop {r4, r5, r7, pc}
+; CHECK-NEXT:    pop {r4, pc}
 entry:
   %c = icmp eq <16 x i8> %b, zeroinitializer
   %xx = zext <16 x i8> %x to <16 x i64>
   %s = select <16 x i1> %c, <16 x i64> %xx, <16 x i64> zeroinitializer
   %z = call i64 @llvm.vector.reduce.add.v16i64(<16 x i64> %s)
   %r = add i64 %z, %a
   ret i64 %r
 }

 ; CHECK-NEXT:    cset r2, eq
 ; CHECK-NEXT:    cmp r2, #0
 ; CHECK-NEXT:    csetm r2, ne
 ; CHECK-NEXT:    cmp r3, #0
 ; CHECK-NEXT:    cset r3, eq
 ; CHECK-NEXT:    cmp r3, #0
 ; CHECK-NEXT:    csetm r3, ne
 ; CHECK-NEXT:    vmov q1[2], q1[0], r3, r2
-; CHECK-NEXT:    vmov q1[3], q1[1], r3, r2
 ; CHECK-NEXT:    vand q0, q0, q1
-; CHECK-NEXT:    vmov r2, s3
-; CHECK-NEXT:    vmov r3, s1
-; CHECK-NEXT:    orr.w r12, r3, r2
-; CHECK-NEXT:    vmov r3, s2
-; CHECK-NEXT:    vmov r2, s0
+; CHECK-NEXT:    vmov r2, s2
+; CHECK-NEXT:    vmov r3, s0
 ; CHECK-NEXT:    add r2, r3
 ; CHECK-NEXT:    adds r0, r0, r2
-; CHECK-NEXT:    adc.w r1, r1, r12
+; CHECK-NEXT:    adc r1, r1, #0
 ; CHECK-NEXT:    bx lr
 entry:
   %c = icmp eq <2 x i8> %b, zeroinitializer
   %xx = zext <2 x i8> %x to <2 x i64>
   %s = select <2 x i1> %c, <2 x i64> %xx, <2 x i64> zeroinitializer
   %z = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> %s)
   %r = add i64 %z, %a
   ret i64 %r

llvm/test/CodeGen/Thumb2/mve-vecreduce-mla.ll

entry:
   %z = call i64 @llvm.vector.reduce.add.v8i64(<8 x i64> %ma)
   ret i64 %z
 }
 
 define arm_aapcs_vfpcc i64 @add_v2i16_v2i64_zext(<2 x i16> %x, <2 x i16> %y) {
 ; CHECK-LABEL: add_v2i16_v2i64_zext:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    vmov.i64 q2, #0xffff
-; CHECK-NEXT:    vand q1, q1, q2
 ; CHECK-NEXT:    vand q0, q0, q2
-; CHECK-NEXT:    vmov r0, s4
-; CHECK-NEXT:    vmov r1, s0
-; CHECK-NEXT:    vmov r2, s6
-; CHECK-NEXT:    vmov r3, s2
-; CHECK-NEXT:    umull r0, r1, r1, r0
-; CHECK-NEXT:    umlal r0, r1, r3, r2
+; CHECK-NEXT:    vand q1, q1, q2
+; CHECK-NEXT:    vmov r0, s6
+; CHECK-NEXT:    vmov r1, s2
+; CHECK-NEXT:    vmov r2, s0
+; CHECK-NEXT:    muls r0, r1, r0
+; CHECK-NEXT:    vmov r1, s4
+; CHECK-NEXT:    muls r1, r2, r1
+; CHECK-NEXT:    movs r2, #0
+; CHECK-NEXT:    adds r0, r0, r1
+; CHECK-NEXT:    adc r1, r2, #0
 ; CHECK-NEXT:    bx lr
 entry:
   %xx = zext <2 x i16> %x to <2 x i64>
   %yy = zext <2 x i16> %y to <2 x i64>
   %m = mul <2 x i64> %xx, %yy
   %z = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> %m)
   ret i64 %z
 }
 
 define arm_aapcs_vfpcc i64 @add_v2i16_v2i64_sext(<2 x i16> %x, <2 x i16> %y) {
 ; CHECK-LABEL: add_v2i16_v2i64_sext:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov r0, s4
-; CHECK-NEXT:    vmov r1, s0
-; CHECK-NEXT:    vmov r2, s6
-; CHECK-NEXT:    vmov r3, s2
+; CHECK-NEXT:    vmov r0, s6
+; CHECK-NEXT:    vmov r1, s2
+; CHECK-NEXT:    vmov r2, s4
+; CHECK-NEXT:    vmov r3, s0
 ; CHECK-NEXT:    sxth r0, r0
 ; CHECK-NEXT:    sxth r1, r1
 ; CHECK-NEXT:    smull r0, r1, r1, r0
 ; CHECK-NEXT:    sxth r2, r2
 ; CHECK-NEXT:    sxth r3, r3
 ; CHECK-NEXT:    smlal r0, r1, r3, r2
 ; CHECK-NEXT:    bx lr
 entry:
entry:
   %m = mul <16 x i8> %x, %y
   %z = call i8 @llvm.vector.reduce.add.v16i8(<16 x i8> %m)
   ret i8 %z
 }
 
 define arm_aapcs_vfpcc i64 @add_v16i8_v16i64_zext(<16 x i8> %x, <16 x i8> %y) {
 ; CHECK-LABEL: add_v16i8_v16i64_zext:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .save {r4, lr}
-; CHECK-NEXT:    push {r4, lr}
-; CHECK-NEXT:    .vsave {d8, d9, d10, d11}
-; CHECK-NEXT:    vpush {d8, d9, d10, d11}
-; CHECK-NEXT:    vmov.u8 r0, q1[3]
-; CHECK-NEXT:    vmov.u8 r1, q1[2]
+; CHECK-NEXT:    .vsave {d8, d9}
+; CHECK-NEXT:    vpush {d8, d9}
+; CHECK-NEXT:    vmov.u8 r0, q1[1]
+; CHECK-NEXT:    vmov.u8 r1, q1[0]
 ; CHECK-NEXT:    vmov q3[2], q3[0], r1, r0
-; CHECK-NEXT:    vmov.u8 r1, q0[3]
-; CHECK-NEXT:    vmov.u8 r2, q0[2]
+; CHECK-NEXT:    vmov.u8 r1, q0[1]
+; CHECK-NEXT:    vmov.u8 r2, q0[0]
 ; CHECK-NEXT:    vmov.i64 q2, #0xff
 ; CHECK-NEXT:    vmov q4[2], q4[0], r2, r1
 ; CHECK-NEXT:    vand q3, q3, q2
 ; CHECK-NEXT:    vand q4, q4, q2
-; CHECK-NEXT:    vmov r0, s14
-; CHECK-NEXT:    vmov r1, s18
-; CHECK-NEXT:    vmov r2, s12
-; CHECK-NEXT:    vmov r3, s16
-; CHECK-NEXT:    umull r0, r1, r1, r0
-; CHECK-NEXT:    umull r2, r3, r3, r2
-; CHECK-NEXT:    vmov q3[2], q3[0], r2, r0
-; CHECK-NEXT:    vmov.u8 r0, q1[0]
-; CHECK-NEXT:    vmov q3[3], q3[1], r3, r1
-; CHECK-NEXT:    vmov.u8 r3, q1[1]
-; CHECK-NEXT:    vmov q4[2], q4[0], r0, r3
-; CHECK-NEXT:    vmov.u8 r3, q0[1]
-; CHECK-NEXT:    vmov.u8 r2, q0[0]
+; CHECK-NEXT:    vmov r0, s12
+; CHECK-NEXT:    vmov r1, s16
+; CHECK-NEXT:    vmov.u8 r3, q0[2]
+; CHECK-NEXT:    vmov r2, s18
+; CHECK-NEXT:    muls r0, r1, r0
+; CHECK-NEXT:    vmov r1, s14
+; CHECK-NEXT:    smlabb r0, r2, r1, r0
+; CHECK-NEXT:    vmov.u8 r1, q1[3]
+; CHECK-NEXT:    vmov.u8 r2, q1[2]
+; CHECK-NEXT:    vmov q3[2], q3[0], r2, r1
+; CHECK-NEXT:    vmov.u8 r2, q0[3]
+; CHECK-NEXT:    vmov q4[2], q4[0], r3, r2
+; CHECK-NEXT:    vand q3, q3, q2
 ; CHECK-NEXT:    vand q4, q4, q2
-; CHECK-NEXT:    vmov q5[2], q5[0], r2, r3
-; CHECK-NEXT:    vmov r0, s16
-; CHECK-NEXT:    vand q5, q5, q2
-; CHECK-NEXT:    vmov r3, s18
-; CHECK-NEXT:    vmov r2, s20
-; CHECK-NEXT:    vmov r4, s22
-; CHECK-NEXT:    vmov lr, s12
-; CHECK-NEXT:    vmov r12, s13
-; CHECK-NEXT:    umull r0, r2, r2, r0
-; CHECK-NEXT:    smlabb r0, r4, r3, r0
-; CHECK-NEXT:    vmov r3, s14
-; CHECK-NEXT:    vmov.u8 r4, q0[4]
-; CHECK-NEXT:    adds.w r0, r0, lr
-; CHECK-NEXT:    adc.w r2, r2, r12
-; CHECK-NEXT:    adds.w r12, r0, r3
-; CHECK-NEXT:    vmov.u8 r3, q1[4]
-; CHECK-NEXT:    adcs r1, r2
-; CHECK-NEXT:    vmov.u8 r2, q1[5]
-; CHECK-NEXT:    vmov q3[2], q3[0], r3, r2
-; CHECK-NEXT:    vmov.u8 r3, q0[5]
-; CHECK-NEXT:    vmov q4[2], q4[0], r4, r3
+; CHECK-NEXT:    vmov r1, s12
+; CHECK-NEXT:    vmov r2, s16
+; CHECK-NEXT:    vmov.u8 r3, q0[4]
+; CHECK-NEXT:    smlabb r0, r2, r1, r0
+; CHECK-NEXT:    vmov r1, s14
+; CHECK-NEXT:    vmov r2, s18
+; CHECK-NEXT:    smlabb r0, r2, r1, r0
+; CHECK-NEXT:    vmov.u8 r1, q1[5]
+; CHECK-NEXT:    vmov.u8 r2, q1[4]
+; CHECK-NEXT:    vmov q3[2], q3[0], r2, r1
+; CHECK-NEXT:    vmov.u8 r2, q0[5]
+; CHECK-NEXT:    vmov q4[2], q4[0], r3, r2
 ; CHECK-NEXT:    vand q3, q3, q2
 ; CHECK-NEXT:    vand q4, q4, q2
-; CHECK-NEXT:    vmov r2, s14
+; CHECK-NEXT:    vmov r1, s12
+; CHECK-NEXT:    vmov r2, s16
 ; CHECK-NEXT:    vmov r3, s18
-; CHECK-NEXT:    vmov r4, s12
-; CHECK-NEXT:    vmov r0, s16
-; CHECK-NEXT:    umull r2, r3, r3, r2
-; CHECK-NEXT:    umull r0, r4, r0, r4
-; CHECK-NEXT:    vmov q3[2], q3[0], r0, r2
-; CHECK-NEXT:    vmov q3[3], q3[1], r4, r3
-; CHECK-NEXT:    vmov.u8 r4, q0[6]
-; CHECK-NEXT:    vmov r2, s12
-; CHECK-NEXT:    vmov r0, s13
-; CHECK-NEXT:    adds.w r2, r2, r12
-; CHECK-NEXT:    adcs r0, r1
-; CHECK-NEXT:    vmov r1, s14
-; CHECK-NEXT:    adds r1, r1, r2
-; CHECK-NEXT:    vmov.u8 r2, q1[7]
-; CHECK-NEXT:    adc.w r12, r0, r3
+; CHECK-NEXT:    muls r1, r2, r1
+; CHECK-NEXT:    movs r2, #0
+; CHECK-NEXT:    adds r0, r0, r1
+; CHECK-NEXT:    adc r1, r2, #0
+; CHECK-NEXT:    vmov r2, s14
+; CHECK-NEXT:    muls r2, r3, r2
 ; CHECK-NEXT:    vmov.u8 r3, q1[6]
+; CHECK-NEXT:    adds.w r12, r0, r2
+; CHECK-NEXT:    vmov.u8 r2, q1[7]
 ; CHECK-NEXT:    vmov q3[2], q3[0], r3, r2
 ; CHECK-NEXT:    vmov.u8 r3, q0[7]
-; CHECK-NEXT:    vmov q4[2], q4[0], r4, r3
+; CHECK-NEXT:    vmov.u8 r0, q0[6]
 ; CHECK-NEXT:    vand q3, q3, q2
+; CHECK-NEXT:    vmov q4[2], q4[0], r0, r3
+; CHECK-NEXT:    vmov r2, s12
 ; CHECK-NEXT:    vand q4, q4, q2
-; CHECK-NEXT:    vmov r2, s14
-; CHECK-NEXT:    vmov r3, s18
-; CHECK-NEXT:    vmov r4, s12
+; CHECK-NEXT:    adc r1, r1, #0
 ; CHECK-NEXT:    vmov r0, s16
-; CHECK-NEXT:    umull r2, r3, r3, r2
-; CHECK-NEXT:    umull r0, r4, r0, r4
-; CHECK-NEXT:    vmov q3[2], q3[0], r0, r2
-; CHECK-NEXT:    vmov q3[3], q3[1], r4, r3
-; CHECK-NEXT:    vmov.u8 r4, q0[8]
-; CHECK-NEXT:    vmov r2, s12
-; CHECK-NEXT:    vmov r0, s13
-; CHECK-NEXT:    adds r1, r1, r2
+; CHECK-NEXT:    vmov r3, s18
+; CHECK-NEXT:    muls r0, r2, r0
 ; CHECK-NEXT:    vmov r2, s14
-; CHECK-NEXT:    adc.w r0, r0, r12
-; CHECK-NEXT:    adds r1, r1, r2
-; CHECK-NEXT:    vmov.u8 r2, q1[9]
-; CHECK-NEXT:    adc.w r12, r0, r3
+; CHECK-NEXT:    adds.w r0, r0, r12
+; CHECK-NEXT:    adc r1, r1, #0
+; CHECK-NEXT:    muls r2, r3, r2
 ; CHECK-NEXT:    vmov.u8 r3, q1[8]
+; CHECK-NEXT:    adds.w r12, r0, r2
+; CHECK-NEXT:    vmov.u8 r2, q1[9]
 ; CHECK-NEXT:    vmov q3[2], q3[0], r3, r2
 ; CHECK-NEXT:    vmov.u8 r3, q0[9]
-; CHECK-NEXT:    vmov q4[2], q4[0], r4, r3
+; CHECK-NEXT:    vmov.u8 r0, q0[8]
 ; CHECK-NEXT:    vand q3, q3, q2
+; CHECK-NEXT:    vmov q4[2], q4[0], r0, r3
+; CHECK-NEXT:    vmov r2, s12
 ; CHECK-NEXT:    vand q4, q4, q2
-; CHECK-NEXT:    vmov r2, s14
-; CHECK-NEXT:    vmov r3, s18
-; CHECK-NEXT:    vmov r4, s12
+; CHECK-NEXT:    adc r1, r1, #0
 ; CHECK-NEXT:    vmov r0, s16
-; CHECK-NEXT:    umull r2, r3, r3, r2
-; CHECK-NEXT:    umull r0, r4, r0, r4
-; CHECK-NEXT:    vmov q3[2], q3[0], r0, r2
-; CHECK-NEXT:    vmov q3[3], q3[1], r4, r3
-; CHECK-NEXT:    vmov.u8 r4, q0[10]
-; CHECK-NEXT:    vmov r2, s12
-; CHECK-NEXT:    vmov r0, s13
-; CHECK-NEXT:    adds r1, r1, r2
+; CHECK-NEXT:    vmov r3, s18
+; CHECK-NEXT:    muls r0, r2, r0
 ; CHECK-NEXT:    vmov r2, s14
-; CHECK-NEXT:    adc.w r0, r0, r12
-; CHECK-NEXT:    adds r1, r1, r2
-; CHECK-NEXT:    vmov.u8 r2, q1[11]
-; CHECK-NEXT:    adc.w r12, r0, r3
+; CHECK-NEXT:    adds.w r0, r0, r12
+; CHECK-NEXT:    adc r1, r1, #0
+; CHECK-NEXT:    muls r2, r3, r2
 ; CHECK-NEXT:    vmov.u8 r3, q1[10]
+; CHECK-NEXT:    adds.w r12, r0, r2
+; CHECK-NEXT:    vmov.u8 r2, q1[11]
 ; CHECK-NEXT:    vmov q3[2], q3[0], r3, r2
 ; CHECK-NEXT:    vmov.u8 r3, q0[11]
-; CHECK-NEXT:    vmov q4[2], q4[0], r4, r3
+; CHECK-NEXT:    vmov.u8 r0, q0[10]
 ; CHECK-NEXT:    vand q3, q3, q2
+; CHECK-NEXT:    vmov q4[2], q4[0], r0, r3
+; CHECK-NEXT:    vmov r2, s12
 ; CHECK-NEXT:    vand q4, q4, q2
-; CHECK-NEXT:    vmov r2, s14
-; CHECK-NEXT:    vmov r3, s18
-; CHECK-NEXT:    vmov r4, s12
+; CHECK-NEXT:    adc r1, r1, #0
 ; CHECK-NEXT:    vmov r0, s16
-; CHECK-NEXT:    umull r2, r3, r3, r2
-; CHECK-NEXT:    umull r0, r4, r0, r4
-; CHECK-NEXT:    vmov q3[2], q3[0], r0, r2
-; CHECK-NEXT:    vmov q3[3], q3[1], r4, r3
-; CHECK-NEXT:    vmov.u8 r4, q0[12]
-; CHECK-NEXT:    vmov r2, s12
-; CHECK-NEXT:    vmov r0, s13
-; CHECK-NEXT:    adds r1, r1, r2
+; CHECK-NEXT:    vmov r3, s18
+; CHECK-NEXT:    muls r0, r2, r0
 ; CHECK-NEXT:    vmov r2, s14
-; CHECK-NEXT:    adc.w r0, r0, r12
-; CHECK-NEXT:    adds r1, r1, r2
-; CHECK-NEXT:    vmov.u8 r2, q1[13]
-; CHECK-NEXT:    adc.w r12, r0, r3
+; CHECK-NEXT:    adds.w r0, r0, r12
+; CHECK-NEXT:    adc r1, r1, #0
+; CHECK-NEXT:    muls r2, r3, r2
 ; CHECK-NEXT:    vmov.u8 r3, q1[12]
+; CHECK-NEXT:    adds.w r12, r0, r2
+; CHECK-NEXT:    vmov.u8 r2, q1[13]
 ; CHECK-NEXT:    vmov q3[2], q3[0], r3, r2
 ; CHECK-NEXT:    vmov.u8 r3, q0[13]
-; CHECK-NEXT:    vmov q4[2], q4[0], r4, r3
+; CHECK-NEXT:    vmov.u8 r0, q0[12]
 ; CHECK-NEXT:    vand q3, q3, q2
+; CHECK-NEXT:    vmov q4[2], q4[0], r0, r3
+; CHECK-NEXT:    vmov r2, s12
 ; CHECK-NEXT:    vand q4, q4, q2
-; CHECK-NEXT:    vmov r2, s14
-; CHECK-NEXT:    vmov r3, s18
-; CHECK-NEXT:    vmov r4, s12
+; CHECK-NEXT:    adc r1, r1, #0
 ; CHECK-NEXT:    vmov r0, s16
-; CHECK-NEXT:    umull r2, r3, r3, r2
-; CHECK-NEXT:    umull r0, r4, r0, r4
-; CHECK-NEXT:    vmov q3[2], q3[0], r0, r2
-; CHECK-NEXT:    vmov q3[3], q3[1], r4, r3
-; CHECK-NEXT:    vmov.u8 r4, q0[14]
-; CHECK-NEXT:    vmov r2, s12
-; CHECK-NEXT:    vmov r0, s13
-; CHECK-NEXT:    adds r1, r1, r2
-; CHECK-NEXT:    adc.w r2, r12, r0
-; CHECK-NEXT:    vmov r0, s14
-; CHECK-NEXT:    adds r0, r0, r1
-; CHECK-NEXT:    adc.w r1, r2, r3
-; CHECK-NEXT:    vmov.u8 r2, q1[15]
+; CHECK-NEXT:    vmov r3, s18
+; CHECK-NEXT:    muls r0, r2, r0
+; CHECK-NEXT:    vmov r2, s14
+; CHECK-NEXT:    adds.w r0, r0, r12
+; CHECK-NEXT:    adc r1, r1, #0
+; CHECK-NEXT:    muls r2, r3, r2
 ; CHECK-NEXT:    vmov.u8 r3, q1[14]
+; CHECK-NEXT:    adds.w r12, r0, r2
+; CHECK-NEXT:    vmov.u8 r2, q1[15]
 ; CHECK-NEXT:    vmov q1[2], q1[0], r3, r2
 ; CHECK-NEXT:    vmov.u8 r3, q0[15]
-; CHECK-NEXT:    vmov q0[2], q0[0], r4, r3
+; CHECK-NEXT:    vmov.u8 r0, q0[14]
 ; CHECK-NEXT:    vand q1, q1, q2
-; CHECK-NEXT:    vand q0, q0, q2
+; CHECK-NEXT:    vmov q0[2], q0[0], r0, r3
 ; CHECK-NEXT:    vmov r2, s4
-; CHECK-NEXT:    vmov r3, s0
-; CHECK-NEXT:    umlal r0, r1, r3, r2
-; CHECK-NEXT:    vmov r2, s6
+; CHECK-NEXT:    vand q0, q0, q2
+; CHECK-NEXT:    adc r1, r1, #0
+; CHECK-NEXT:    vmov r0, s0
 ; CHECK-NEXT:    vmov r3, s2
-; CHECK-NEXT:    umlal r0, r1, r3, r2
-; CHECK-NEXT:    vpop {d8, d9, d10, d11}
-; CHECK-NEXT:    pop {r4, pc}
+; CHECK-NEXT:    muls r0, r2, r0
+; CHECK-NEXT:    vmov r2, s6
+; CHECK-NEXT:    adds.w r0, r0, r12
+; CHECK-NEXT:    adc r1, r1, #0
+; CHECK-NEXT:    muls r2, r3, r2
+; CHECK-NEXT:    adds r0, r0, r2
+; CHECK-NEXT:    adc r1, r1, #0
+; CHECK-NEXT:    vpop {d8, d9}
+; CHECK-NEXT:    bx lr
 entry:
   %xx = zext <16 x i8> %x to <16 x i64>
   %yy = zext <16 x i8> %y to <16 x i64>
   %m = mul <16 x i64> %xx, %yy
   %z = call i64 @llvm.vector.reduce.add.v16i64(<16 x i64> %m)
   ret i64 %z
 }
 
 define arm_aapcs_vfpcc i64 @add_v16i8_v16i64_sext(<16 x i8> %x, <16 x i8> %y) {
 ; CHECK-LABEL: add_v16i8_v16i64_sext:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .save {r7, lr}
-; CHECK-NEXT:    push {r7, lr}
 ; CHECK-NEXT:    vmov.s8 r0, q1[1]
 ; CHECK-NEXT:    vmov.s8 r1, q0[1]
+; CHECK-NEXT:    smull r0, r1, r1, r0
 ; CHECK-NEXT:    vmov.s8 r2, q1[0]
 ; CHECK-NEXT:    vmov.s8 r3, q0[0]
-; CHECK-NEXT:    smull r0, r1, r1, r0
-; CHECK-NEXT:    smull r2, r3, r3, r2
-; CHECK-NEXT:    vmov q2[2], q2[0], r2, r0
-; CHECK-NEXT:    vmov q2[3], q2[1], r3, r1
-; CHECK-NEXT:    vmov r2, s10
-; CHECK-NEXT:    vmov r3, s8
-; CHECK-NEXT:    vmov r0, s9
-; CHECK-NEXT:    adds.w lr, r3, r2
+; CHECK-NEXT:    smlal r0, r1, r3, r2
+; CHECK-NEXT:    vmov.s8 r2, q1[2]
+; CHECK-NEXT:    vmov.s8 r3, q0[2]
+; CHECK-NEXT:    smlal r0, r1, r3, r2
+; CHECK-NEXT:    vmov.s8 r2, q1[3]
 ; CHECK-NEXT:    vmov.s8 r3, q0[3]
-; CHECK-NEXT:    adc.w r12, r0, r1
-; CHECK-NEXT:    vmov.s8 r1, q1[3]
-; CHECK-NEXT:    vmov.s8 r0, q1[2]
-; CHECK-NEXT:    vmov.s8 r2, q0[2]
-; CHECK-NEXT:    smull r1, r3, r3, r1
-; CHECK-NEXT:    smull r0, r2, r2, r0
-; CHECK-NEXT:    vmov q2[2], q2[0], r0, r1
-; CHECK-NEXT:    vmov q2[3], q2[1], r2, r3
-; CHECK-NEXT:    vmov r1, s8
-; CHECK-NEXT:    vmov r0, s9
-; CHECK-NEXT:    vmov r2, s10
-; CHECK-NEXT:    adds.w r1, r1, lr
-; CHECK-NEXT:    adc.w r0, r0, r12
-; CHECK-NEXT:    adds.w lr, r1, r2
+; CHECK-NEXT:    smlal r0, r1, r3, r2
+; CHECK-NEXT:    vmov.s8 r2, q1[4]
+; CHECK-NEXT:    vmov.s8 r3, q0[4]
+; CHECK-NEXT:    smlal r0, r1, r3, r2
 ; CHECK-NEXT:    vmov.s8 r2, q1[5]
-; CHECK-NEXT:    adc.w r12, r0, r3
 ; CHECK-NEXT:    vmov.s8 r3, q0[5]
-; CHECK-NEXT:    vmov.s8 r0, q1[4]
-; CHECK-NEXT:    vmov.s8 r1, q0[4]
-; CHECK-NEXT:    smull r2, r3, r3, r2
-; CHECK-NEXT:    smull r0, r1, r1, r0
-; CHECK-NEXT:    vmov q2[2], q2[0], r0, r2
-; CHECK-NEXT:    vmov q2[3], q2[1], r1, r3
-; CHECK-NEXT:    vmov r1, s8
-; CHECK-NEXT:    vmov r0, s9
-; CHECK-NEXT:    vmov r2, s10
-; CHECK-NEXT:    adds.w r1, r1, lr
-; CHECK-NEXT:    adc.w r0, r0, r12
-; CHECK-NEXT:    adds.w lr, r1, r2
+; CHECK-NEXT:    smlal r0, r1, r3, r2
+; CHECK-NEXT:    vmov.s8 r2, q1[6]
+; CHECK-NEXT:    vmov.s8 r3, q0[6]
+; CHECK-NEXT:    smlal r0, r1, r3, r2
 ; CHECK-NEXT:    vmov.s8 r2, q1[7]
-; CHECK-NEXT:    adc.w r12, r0, r3
 ; CHECK-NEXT:    vmov.s8 r3, q0[7]
-; CHECK-NEXT:    vmov.s8 r0, q1[6]
-; CHECK-NEXT:    vmov.s8 r1, q0[6]
-; CHECK-NEXT:    smull r2, r3, r3, r2
-; CHECK-NEXT:    smull r0, r1, r1, r0
-; CHECK-NEXT:    vmov q2[2], q2[0], r0, r2
-; CHECK-NEXT:    vmov q2[3], q2[1], r1, r3
-; CHECK-NEXT:    vmov r1, s8
-; CHECK-NEXT:    vmov r0, s9
-; CHECK-NEXT:    vmov r2, s10
-; CHECK-NEXT:    adds.w r1, r1, lr
-; CHECK-NEXT:    adc.w r0, r0, r12
-; CHECK-NEXT:    adds.w lr, r1, r2
+; CHECK-NEXT:    smlal r0, r1, r3, r2
+; CHECK-NEXT:    vmov.s8 r2, q1[8]
+; CHECK-NEXT:    vmov.s8 r3, q0[8]
+; CHECK-NEXT:    smlal r0, r1, r3, r2
 ; CHECK-NEXT:    vmov.s8 r2, q1[9]
-; CHECK-NEXT:    adc.w r12, r0, r3
 ; CHECK-NEXT:    vmov.s8 r3, q0[9]
-; CHECK-NEXT:    vmov.s8 r0, q1[8]
-; CHECK-NEXT:    vmov.s8 r1, q0[8]
-; CHECK-NEXT:    smull r2, r3, r3, r2
-; CHECK-NEXT:    smull r0, r1, r1, r0
-; CHECK-NEXT:    vmov q2[2], q2[0], r0, r2
-; CHECK-NEXT:    vmov q2[3], q2[1], r1, r3
-; CHECK-NEXT:    vmov r1, s8
-; CHECK-NEXT:    vmov r0, s9
-; CHECK-NEXT:    vmov r2, s10
-; CHECK-NEXT:    adds.w r1, r1, lr
-; CHECK-NEXT:    adc.w r0, r0, r12
-; CHECK-NEXT:    adds.w lr, r1, r2
+; CHECK-NEXT:    smlal r0, r1, r3, r2
+; CHECK-NEXT:    vmov.s8 r2, q1[10]
+; CHECK-NEXT:    vmov.s8 r3, q0[10]
+; CHECK-NEXT:    smlal r0, r1, r3, r2
 ; CHECK-NEXT:    vmov.s8 r2, q1[11]
-; CHECK-NEXT:    adc.w r12, r0, r3
 ; CHECK-NEXT:    vmov.s8 r3, q0[11]
-; CHECK-NEXT:    vmov.s8 r0, q1[10]
-; CHECK-NEXT:    vmov.s8 r1, q0[10]
-; CHECK-NEXT:    smull r2, r3, r3, r2
-; CHECK-NEXT:    smull r0, r1, r1, r0
-; CHECK-NEXT:    vmov q2[2], q2[0], r0, r2
-; CHECK-NEXT:    vmov q2[3], q2[1], r1, r3
-; CHECK-NEXT:    vmov r1, s8
-; CHECK-NEXT:    vmov r0, s9
-; CHECK-NEXT:    vmov r2, s10
-; CHECK-NEXT:    adds.w r1, r1, lr
-; CHECK-NEXT:    adc.w r0, r0, r12
-; CHECK-NEXT:    adds.w lr, r1, r2
+; CHECK-NEXT:    smlal r0, r1, r3, r2
+; CHECK-NEXT:    vmov.s8 r2, q1[12]
+; CHECK-NEXT:    vmov.s8 r3, q0[12]
+; CHECK-NEXT:    smlal r0, r1, r3, r2
 ; CHECK-NEXT:    vmov.s8 r2, q1[13]
-; CHECK-NEXT:    adc.w r12, r0, r3
 ; CHECK-NEXT:    vmov.s8 r3, q0[13]
-; CHECK-NEXT:    vmov.s8 r0, q1[12]
-; CHECK-NEXT:    vmov.s8 r1, q0[12]
-; CHECK-NEXT:    smull r2, r3, r3, r2
-; CHECK-NEXT:    smull r0, r1, r1, r0
-; CHECK-NEXT:    vmov q2[2], q2[0], r0, r2
-; CHECK-NEXT:    vmov q2[3], q2[1], r1, r3
-; CHECK-NEXT:    vmov r1, s8
-; CHECK-NEXT:    vmov r0, s9
-; CHECK-NEXT:    adds.w r1, r1, lr
-; CHECK-NEXT:    adc.w r2, r12, r0
-; CHECK-NEXT:    vmov r0, s10
-; CHECK-NEXT:    adds r0, r0, r1
-; CHECK-NEXT:    adc.w r1, r2, r3
+; CHECK-NEXT:    smlal r0, r1, r3, r2
 ; CHECK-NEXT:    vmov.s8 r2, q1[14]
 ; CHECK-NEXT:    vmov.s8 r3, q0[14]
 ; CHECK-NEXT:    smlal r0, r1, r3, r2
 ; CHECK-NEXT:    vmov.s8 r2, q1[15]
 ; CHECK-NEXT:    vmov.s8 r3, q0[15]
 ; CHECK-NEXT:    smlal r0, r1, r3, r2
-; CHECK-NEXT:    pop {r7, pc}
+; CHECK-NEXT:    bx lr
 entry:
   %xx = sext <16 x i8> %x to <16 x i64>
   %yy = sext <16 x i8> %y to <16 x i64>
   %m = mul <16 x i64> %xx, %yy
   %z = call i64 @llvm.vector.reduce.add.v16i64(<16 x i64> %m)
   ret i64 %z
 }
 
entry:
   %z = call i64 @llvm.vector.reduce.add.v4i64(<4 x i64> %m)
   ret i64 %z
 }
 
 define arm_aapcs_vfpcc i64 @add_v2i8_v2i64_zext(<2 x i8> %x, <2 x i8> %y) {
 ; CHECK-LABEL: add_v2i8_v2i64_zext:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    vmov.i64 q2, #0xff
-; CHECK-NEXT:    vand q1, q1, q2
 ; CHECK-NEXT:    vand q0, q0, q2
+; CHECK-NEXT:    vand q1, q1, q2
 ; CHECK-NEXT:    vmov r0, s6
 ; CHECK-NEXT:    vmov r1, s2
-; CHECK-NEXT:    vmov r2, s4
-; CHECK-NEXT:    vmov r3, s0
-; CHECK-NEXT:    umull r0, r1, r1, r0
-; CHECK-NEXT:    umull r2, r3, r3, r2
-; CHECK-NEXT:    add r0, r2
-; CHECK-NEXT:    orrs r1, r3
+; CHECK-NEXT:    vmov r2, s0
+; CHECK-NEXT:    muls r0, r1, r0
+; CHECK-NEXT:    vmov r1, s4
+; CHECK-NEXT:    mla r0, r2, r1, r0
+; CHECK-NEXT:    movs r1, #0
 ; CHECK-NEXT:    bx lr
 entry:
   %xx = zext <2 x i8> %x to <2 x i64>
   %yy = zext <2 x i8> %y to <2 x i64>
   %m = mul <2 x i64> %xx, %yy
   %z = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> %m)
   ret i64 %z
 }
 
 define arm_aapcs_vfpcc i64 @add_v2i8_v2i64_sext(<2 x i8> %x, <2 x i8> %y) {
 ; CHECK-LABEL: add_v2i8_v2i64_sext:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov r0, s4
-; CHECK-NEXT:    vmov r1, s0
-; CHECK-NEXT:    vmov r2, s6
-; CHECK-NEXT:    vmov r3, s2
+; CHECK-NEXT:    vmov r0, s6
+; CHECK-NEXT:    vmov r1, s2
+; CHECK-NEXT:    vmov r2, s4
+; CHECK-NEXT:    vmov r3, s0
 ; CHECK-NEXT:    sxtb r0, r0
 ; CHECK-NEXT:    sxtb r1, r1
 ; CHECK-NEXT:    smull r0, r1, r1, r0
 ; CHECK-NEXT:    sxtb r2, r2
 ; CHECK-NEXT:    sxtb r3, r3
 ; CHECK-NEXT:    smlal r0, r1, r3, r2
 ; CHECK-NEXT:    bx lr
 entry:
   %xx = sext <2 x i8> %x to <2 x i64>
   %yy = sext <2 x i8> %y to <2 x i64>
   %m = mul <2 x i64> %xx, %yy
   %z = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> %m)
   ret i64 %z
 }
 
 define arm_aapcs_vfpcc i64 @add_v2i64_v2i64(<2 x i64> %x, <2 x i64> %y) {
 ; CHECK-LABEL: add_v2i64_v2i64:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .save {r4, lr}
 ; CHECK-NEXT:    push {r4, lr}
-; CHECK-NEXT:    vmov r0, s4
-; CHECK-NEXT:    vmov r1, s0
-; CHECK-NEXT:    vmov r3, s5
-; CHECK-NEXT:    vmov r4, s7
-; CHECK-NEXT:    umull r12, r2, r1, r0
-; CHECK-NEXT:    mla r1, r1, r3, r2
-; CHECK-NEXT:    vmov r2, s1
-; CHECK-NEXT:    mla lr, r2, r0, r1
 ; CHECK-NEXT:    vmov r0, s6
-; CHECK-NEXT:    vmov r2, s2
+; CHECK-NEXT:    vmov r1, s2
+; CHECK-NEXT:    vmov r2, s7
+; CHECK-NEXT:    vmov r4, s5
+; CHECK-NEXT:    umull r12, r3, r1, r0
+; CHECK-NEXT:    mla r1, r1, r2, r3
+; CHECK-NEXT:    vmov r2, s3
+; CHECK-NEXT:    mla lr, r2, r0, r1
+; CHECK-NEXT:    vmov r0, s4
+; CHECK-NEXT:    vmov r2, s0
 ; CHECK-NEXT:    umull r3, r1, r2, r0
 ; CHECK-NEXT:    mla r1, r2, r4, r1
-; CHECK-NEXT:    vmov r2, s3
+; CHECK-NEXT:    vmov r2, s1
 ; CHECK-NEXT:    mla r1, r2, r0, r1
-; CHECK-NEXT:    adds.w r0, r12, r3
+; CHECK-NEXT:    adds.w r0, r3, r12
 ; CHECK-NEXT:    adc.w r1, r1, lr
 ; CHECK-NEXT:    pop {r4, pc}
 entry:
   %m = mul <2 x i64> %x, %y
   %z = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> %m)
   ret i64 %z
 }
 
entry:
   %z = call i64 @llvm.vector.reduce.add.v8i64(<8 x i64> %ma)
   %r = add i64 %z, %a
   ret i64 %r
 }
 
 define arm_aapcs_vfpcc i64 @add_v2i16_v2i64_acc_zext(<2 x i16> %x, <2 x i16> %y, i64 %a) {
 ; CHECK-LABEL: add_v2i16_v2i64_acc_zext:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .save {r7, lr}
-; CHECK-NEXT:    push {r7, lr}
 ; CHECK-NEXT:    vmov.i64 q2, #0xffff
 ; CHECK-NEXT:    vand q1, q1, q2
 ; CHECK-NEXT:    vand q0, q0, q2
-; CHECK-NEXT:    vmov r2, s4
-; CHECK-NEXT:    vmov r3, s0
-; CHECK-NEXT:    vmov r12, s6
-; CHECK-NEXT:    umull r2, lr, r3, r2
+; CHECK-NEXT:    vmov r2, s6
 ; CHECK-NEXT:    vmov r3, s2
-; CHECK-NEXT:    umlal r2, lr, r3, r12
+; CHECK-NEXT:    mul r12, r3, r2
+; CHECK-NEXT:    vmov r3, s4
+; CHECK-NEXT:    vmov r2, s0
+; CHECK-NEXT:    muls r2, r3, r2
+; CHECK-NEXT:    movs r3, #0
+; CHECK-NEXT:    adds.w r2, r2, r12
+; CHECK-NEXT:    adc r3, r3, #0
 ; CHECK-NEXT:    adds r0, r0, r2
-; CHECK-NEXT:    adc.w r1, r1, lr
-; CHECK-NEXT:    pop {r7, pc}
+; CHECK-NEXT:    adcs r1, r3
+; CHECK-NEXT:    bx lr
 entry:
   %xx = zext <2 x i16> %x to <2 x i64>
   %yy = zext <2 x i16> %y to <2 x i64>
   %m = mul <2 x i64> %xx, %yy
   %z = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> %m)
   %r = add i64 %z, %a
   ret i64 %r
 }
 
 define arm_aapcs_vfpcc i64 @add_v2i16_v2i64_acc_sext(<2 x i16> %x, <2 x i16> %y, i64 %a) {
 ; CHECK-LABEL: add_v2i16_v2i64_acc_sext:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .save {r7, lr}
 ; CHECK-NEXT:    push {r7, lr}
-; CHECK-NEXT:    vmov r2, s4
-; CHECK-NEXT:    vmov r3, s0
+; CHECK-NEXT:    vmov r2, s6
+; CHECK-NEXT:    vmov r3, s2
 ; CHECK-NEXT:    sxth r2, r2
 ; CHECK-NEXT:    sxth r3, r3
 ; CHECK-NEXT:    smull r2, r12, r3, r2
-; CHECK-NEXT:    vmov r3, s6
+; CHECK-NEXT:    vmov r3, s4
 ; CHECK-NEXT:    sxth.w lr, r3
-; CHECK-NEXT:    vmov r3, s2
+; CHECK-NEXT:    vmov r3, s0
 ; CHECK-NEXT:    sxth r3, r3
 ; CHECK-NEXT:    smlal r2, r12, r3, lr
 ; CHECK-NEXT:    adds r0, r0, r2
 ; CHECK-NEXT:    adc.w r1, r1, r12
 ; CHECK-NEXT:    pop {r7, pc}
 entry:
   %xx = sext <2 x i16> %x to <2 x i64>
   %yy = sext <2 x i16> %y to <2 x i64>
entry:
   %z = call i8 @llvm.vector.reduce.add.v16i8(<16 x i8> %m)
   %r = add i8 %z, %a
   ret i8 %r
 }
 
 define arm_aapcs_vfpcc i64 @add_v16i8_v16i64_acc_zext(<16 x i8> %x, <16 x i8> %y, i64 %a) {
 ; CHECK-LABEL: add_v16i8_v16i64_acc_zext:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .save {r4, r5, r6, lr}
-; CHECK-NEXT:    push {r4, r5, r6, lr}
-; CHECK-NEXT:    .vsave {d8, d9, d10, d11}
-; CHECK-NEXT:    vpush {d8, d9, d10, d11}
-; CHECK-NEXT:    vmov.u8 r2, q1[3]
-; CHECK-NEXT:    vmov.u8 r3, q1[2]
+; CHECK-NEXT:    .save {r4, lr}
+; CHECK-NEXT:    push {r4, lr}
+; CHECK-NEXT:    .vsave {d8, d9}
+; CHECK-NEXT:    vpush {d8, d9}
+; CHECK-NEXT:    vmov.u8 r2, q1[1]
+; CHECK-NEXT:    vmov.u8 r3, q1[0]
 ; CHECK-NEXT:    vmov q3[2], q3[0], r3, r2
-; CHECK-NEXT:    vmov.u8 r3, q0[3]
-; CHECK-NEXT:    vmov.u8 r2, q0[2]
+; CHECK-NEXT:    vmov.u8 r3, q0[1]
+; CHECK-NEXT:    vmov.u8 r2, q0[0]
 ; CHECK-NEXT:    vmov.i64 q2, #0xff
 ; CHECK-NEXT:    vmov q4[2], q4[0], r2, r3
 ; CHECK-NEXT:    vand q3, q3, q2
 ; CHECK-NEXT:    vand q4, q4, q2
-; CHECK-NEXT:    vmov r12, s14
-; CHECK-NEXT:    vmov r2, s18
-; CHECK-NEXT:    vmov.u8 r4, q1[0]
-; CHECK-NEXT:    vmov r3, s12
-; CHECK-NEXT:    vmov.u8 r5, q0[0]
-; CHECK-NEXT:    umull lr, r12, r2, r12
+; CHECK-NEXT:    vmov r12, s12
 ; CHECK-NEXT:    vmov r2, s16
-; CHECK-NEXT:    umull r2, r3, r2, r3
-; CHECK-NEXT:    vmov q3[2], q3[0], r2, lr
-; CHECK-NEXT:    vmov.u8 r2, q1[1]
-; CHECK-NEXT:    vmov q4[2], q4[0], r4, r2
-; CHECK-NEXT:    vmov.u8 r4, q0[1]
-; CHECK-NEXT:    vmov q5[2], q5[0], r5, r4
+; CHECK-NEXT:    vmov.u8 r4, q0[6]
+; CHECK-NEXT:    vmov r3, s14
+; CHECK-NEXT:    mul r12, r2, r12
+; CHECK-NEXT:    vmov r2, s18
+; CHECK-NEXT:    smlabb r12, r2, r3, r12
+; CHECK-NEXT:    vmov.u8 r3, q1[3]
+; CHECK-NEXT:    vmov.u8 r2, q1[2]
+; CHECK-NEXT:    vmov q3[2], q3[0], r2, r3
+; CHECK-NEXT:    vmov.u8 r3, q0[3]
+; CHECK-NEXT:    vmov.u8 r2, q0[2]
+; CHECK-NEXT:    vand q3, q3, q2
+; CHECK-NEXT:    vmov q4[2], q4[0], r2, r3
+; CHECK-NEXT:    vmov lr, s12
 ; CHECK-NEXT:    vand q4, q4, q2
-; CHECK-NEXT:    vand q5, q5, q2
+; CHECK-NEXT:    vmov r3, s14
 ; CHECK-NEXT:    vmov r2, s16
-; CHECK-NEXT:    vmov r4, s20
-; CHECK-NEXT:    vmov q3[3], q3[1], r3, r12
-; CHECK-NEXT:    vmov r5, s18
-; CHECK-NEXT:    vmov r6, s22
-; CHECK-NEXT:    vmov r3, s12
-; CHECK-NEXT:    vmov lr, s13
-; CHECK-NEXT:    umull r2, r4, r4, r2
-; CHECK-NEXT:    smlabb r2, r6, r5, r2
-; CHECK-NEXT:    vmov r6, s14
-; CHECK-NEXT:    vmov.u8 r5, q1[4]
-; CHECK-NEXT:    adds r2, r2, r3
-; CHECK-NEXT:    adc.w r3, r4, lr
-; CHECK-NEXT:    vmov.u8 r4, q0[4]
-; CHECK-NEXT:    adds.w lr, r2, r6
-; CHECK-NEXT:    vmov.u8 r6, q1[5]
-; CHECK-NEXT:    vmov q3[2], q3[0], r5, r6
-; CHECK-NEXT:    vmov.u8 r5, q0[5]
-; CHECK-NEXT:    vmov q4[2], q4[0], r4, r5
+; CHECK-NEXT:    smlabb r12, r2, lr, r12
+; CHECK-NEXT:    vmov r2, s18
+; CHECK-NEXT:    smlabb r12, r2, r3, r12
+; CHECK-NEXT:    vmov.u8 r3, q1[5]
+; CHECK-NEXT:    vmov.u8 r2, q1[4]
+; CHECK-NEXT:    vmov q3[2], q3[0], r2, r3
+; CHECK-NEXT:    vmov.u8 r3, q0[5]
+; CHECK-NEXT:    vmov.u8 r2, q0[4]
 ; CHECK-NEXT:    vand q3, q3, q2
+; CHECK-NEXT:    vmov q4[2], q4[0], r2, r3
+; CHECK-NEXT:    vmov lr, s12
 ; CHECK-NEXT:    vand q4, q4, q2
-; CHECK-NEXT:    vmov r6, s14
-; CHECK-NEXT:    vmov r5, s18
-; CHECK-NEXT:    adc.w r3, r3, r12
-; CHECK-NEXT:    vmov r4, s12
+; CHECK-NEXT:    movs r3, #0
 ; CHECK-NEXT:    vmov r2, s16
-; CHECK-NEXT:    umull r6, r5, r5, r6
-; CHECK-NEXT:    umull r2, r4, r2, r4
-; CHECK-NEXT:    vmov q3[2], q3[0], r2, r6
-; CHECK-NEXT:    vmov q3[3], q3[1], r4, r5
-; CHECK-NEXT:    vmov.u8 r4, q0[6]
-; CHECK-NEXT:    vmov r6, s12
-; CHECK-NEXT:    vmov r2, s13
-; CHECK-NEXT:    adds.w r6, r6, lr
-; CHECK-NEXT:    adcs r2, r3
-; CHECK-NEXT:    vmov r3, s14
-; CHECK-NEXT:    adds r3, r3, r6
-; CHECK-NEXT:    vmov.u8 r6, q1[7]
-; CHECK-NEXT:    adc.w r12, r2, r5
-; CHECK-NEXT:    vmov.u8 r5, q1[6]
-; CHECK-NEXT:    vmov q3[2], q3[0], r5, r6
-; CHECK-NEXT:    vmov.u8 r5, q0[7]
-; CHECK-NEXT:    vmov q4[2], q4[0], r4, r5
+; CHECK-NEXT:    mul r2, r2, lr
+; CHECK-NEXT:    adds.w r12, r12, r2
+; CHECK-NEXT:    vmov r2, s14
+; CHECK-NEXT:    adc lr, r3, #0
+; CHECK-NEXT:    vmov r3, s18
+; CHECK-NEXT:    muls r2, r3, r2
+; CHECK-NEXT:    vmov.u8 r3, q1[6]
+; CHECK-NEXT:    adds.w r12, r12, r2
+; CHECK-NEXT:    vmov.u8 r2, q1[7]
+; CHECK-NEXT:    vmov q3[2], q3[0], r3, r2
+; CHECK-NEXT:    vmov.u8 r3, q0[7]
+; CHECK-NEXT:    vmov q4[2], q4[0], r4, r3
 ; CHECK-NEXT:    vand q3, q3, q2
 ; CHECK-NEXT:    vand q4, q4, q2
-; CHECK-NEXT:    vmov r6, s14
-; CHECK-NEXT:    vmov r5, s18
-; CHECK-NEXT:    vmov r4, s12
-; CHECK-NEXT:    vmov r2, s16
-; CHECK-NEXT:    umull r6, r5, r5, r6
-; CHECK-NEXT:    umull r2, r4, r2, r4
-; CHECK-NEXT:    vmov q3[2], q3[0], r2, r6
-; CHECK-NEXT:    vmov q3[3], q3[1], r4, r5
-; CHECK-NEXT:    vmov.u8 r4, q0[8]
-; CHECK-NEXT:    vmov r6, s12
-; CHECK-NEXT:    vmov r2, s13
-; CHECK-NEXT:    adds r3, r3, r6
-; CHECK-NEXT:    vmov r6, s14
-; CHECK-NEXT:    adc.w r2, r2, r12
-; CHECK-NEXT:    adds r3, r3, r6
-; CHECK-NEXT:    vmov.u8 r6, q1[9]
-; CHECK-NEXT:    adc.w r12, r2, r5
-; CHECK-NEXT:    vmov.u8 r5, q1[8]
-; CHECK-NEXT:    vmov q3[2], q3[0], r5, r6
-; CHECK-NEXT:    vmov.u8 r5, q0[9]
-; CHECK-NEXT:    vmov q4[2], q4[0], r4, r5
+; CHECK-NEXT:    vmov r2, s12
+; CHECK-NEXT:    vmov r3, s16
+; CHECK-NEXT:    adc lr, lr, #0
+; CHECK-NEXT:    vmov r4, s14
+; CHECK-NEXT:    muls r2, r3, r2
+; CHECK-NEXT:    adds.w r12, r12, r2
+; CHECK-NEXT:    vmov r2, s18
+; CHECK-NEXT:    adc r3, lr, #0
+; CHECK-NEXT:    muls r2, r4, r2
+; CHECK-NEXT:    vmov.u8 r4, q1[9]
+; CHECK-NEXT:    adds.w r12, r12, r2
+; CHECK-NEXT:    vmov.u8 r2, q1[8]
+; CHECK-NEXT:    vmov q3[2], q3[0], r2, r4
+; CHECK-NEXT:    adc lr, r3, #0
+; CHECK-NEXT:    vmov.u8 r4, q0[9]
+; CHECK-NEXT:    vmov.u8 r3, q0[8]
+; CHECK-NEXT:    vmov q4[2], q4[0], r3, r4
 ; CHECK-NEXT:    vand q3, q3, q2
 ; CHECK-NEXT:    vand q4, q4, q2
-; CHECK-NEXT:    vmov r6, s14
-; CHECK-NEXT:    vmov r5, s18
-; CHECK-NEXT:    vmov r4, s12
-; CHECK-NEXT:    vmov r2, s16
-; CHECK-NEXT:    umull r6, r5, r5, r6
-; CHECK-NEXT:    umull r2, r4, r2, r4
-; CHECK-NEXT:    vmov q3[2], q3[0], r2, r6
-; CHECK-NEXT:    vmov q3[3], q3[1], r4, r5
-; CHECK-NEXT:    vmov.u8 r4, q0[10]
-; CHECK-NEXT:    vmov r6, s12
-; CHECK-NEXT:    vmov r2, s13
-; CHECK-NEXT:    adds r3, r3, r6
-; CHECK-NEXT:    vmov r6, s14
-; CHECK-NEXT:    adc.w r2, r2, r12
-; CHECK-NEXT:    adds r3, r3, r6
-; CHECK-NEXT:    vmov.u8 r6, q1[11]
-; CHECK-NEXT:    adc.w r12, r2, r5
-; CHECK-NEXT:    vmov.u8 r5, q1[10]
-; CHECK-NEXT:    vmov q3[2], q3[0], r5, r6
-; CHECK-NEXT:    vmov.u8 r5, q0[11]
-; CHECK-NEXT:    vmov q4[2], q4[0], r4, r5
+; CHECK-NEXT:    vmov r2, s12
+; CHECK-NEXT:    vmov r3, s16
+; CHECK-NEXT:    vmov r4, s14
+; CHECK-NEXT:    muls r2, r3, r2
+; CHECK-NEXT:    adds.w r12, r12, r2
+; CHECK-NEXT:    vmov r2, s18
+; CHECK-NEXT:    adc r3, lr, #0
+; CHECK-NEXT:    muls r2, r4, r2
+; CHECK-NEXT:    vmov.u8 r4, q1[11]
+; CHECK-NEXT:    adds.w r12, r12, r2
+; CHECK-NEXT:    vmov.u8 r2, q1[10]
+; CHECK-NEXT:    vmov q3[2], q3[0], r2, r4
+; CHECK-NEXT:    adc lr, r3, #0
+; CHECK-NEXT:    vmov.u8 r4, q0[11]
+; CHECK-NEXT:    vmov.u8 r3, q0[10]
+; CHECK-NEXT:    vmov q4[2], q4[0], r3, r4
 ; CHECK-NEXT:    vand q3, q3, q2
 ; CHECK-NEXT:    vand q4, q4, q2
-; CHECK-NEXT:    vmov r6, s14
-; CHECK-NEXT:    vmov r5, s18
-; CHECK-NEXT:    vmov r4, s12
-; CHECK-NEXT:    vmov r2, s16
-; CHECK-NEXT:    umull r6, r5, r5, r6
-; CHECK-NEXT:    umull r2, r4, r2, r4
-; CHECK-NEXT:    vmov q3[2], q3[0], r2, r6
-; CHECK-NEXT:    vmov q3[3], q3[1], r4, r5
-; CHECK-NEXT:    vmov.u8 r4, q0[12]
-; CHECK-NEXT:    vmov r6, s12
-; CHECK-NEXT:    vmov r2, s13
-; CHECK-NEXT:    adds r3, r3, r6
-; CHECK-NEXT:    vmov r6, s14
-; CHECK-NEXT:    adc.w r2, r2, r12
-; CHECK-NEXT:    adds r3, r3, r6
-; CHECK-NEXT:    vmov.u8 r6, q1[13]
-; CHECK-NEXT:    adc.w r12, r2, r5
-; CHECK-NEXT:    vmov.u8 r5, q1[12]
-; CHECK-NEXT:    vmov q3[2], q3[0], r5, r6
-; CHECK-NEXT:    vmov.u8 r5, q0[13]
-; CHECK-NEXT:    vmov q4[2], q4[0], r4, r5
+; CHECK-NEXT:    vmov r2, s12
+; CHECK-NEXT:    vmov r3, s16
+; CHECK-NEXT:    vmov r4, s14
+; CHECK-NEXT:    muls r2, r3, r2
+; CHECK-NEXT:    adds.w r12, r12, r2
+; CHECK-NEXT:    vmov r2, s18
+; CHECK-NEXT:    adc r3, lr, #0
+; CHECK-NEXT:    muls r2, r4, r2
+; CHECK-NEXT:    vmov.u8 r4, q1[13]
+; CHECK-NEXT:    adds.w r12, r12, r2
+; CHECK-NEXT:    vmov.u8 r2, q1[12]
+; CHECK-NEXT:    vmov q3[2], q3[0], r2, r4
+; CHECK-NEXT:    adc lr, r3, #0
+; CHECK-NEXT:    vmov.u8 r4, q0[13]
+; CHECK-NEXT:    vmov.u8 r3, q0[12]
+; CHECK-NEXT:    vmov q4[2], q4[0], r3, r4
 ; CHECK-NEXT:    vand q3, q3, q2
 ; CHECK-NEXT:    vand q4, q4, q2
-; CHECK-NEXT:    vmov r6, s14
-; CHECK-NEXT:    vmov r5, s18
-; CHECK-NEXT:    vmov r4, s12
-; CHECK-NEXT:    vmov r2, s16
-; CHECK-NEXT:    umull r6, r5, r5, r6
-; CHECK-NEXT:    umull r2, r4, r2, r4
-; CHECK-NEXT:    vmov q3[2], q3[0], r2, r6
-; CHECK-NEXT:    vmov q3[3], q3[1], r4, r5
-; CHECK-NEXT:    vmov.u8 r4, q0[14]
-; CHECK-NEXT:    vmov r6, s12
-; CHECK-NEXT:    vmov r2, s13
-; CHECK-NEXT:    adds r3, r3, r6
-; CHECK-NEXT:    vmov r6, s14
-; CHECK-NEXT:    adc.w r2, r2, r12
-; CHECK-NEXT:    adds r3, r3, r6
-; CHECK-NEXT:    vmov.u8 r6, q1[15]
-; CHECK-NEXT:    adcs r2, r5
-; CHECK-NEXT:    vmov.u8 r5, q1[14]
-; CHECK-NEXT:    vmov q1[2], q1[0], r5, r6
-; CHECK-NEXT:    vmov.u8 r5, q0[15]
-; CHECK-NEXT:    vmov q0[2], q0[0], r4, r5
+; CHECK-NEXT:    vmov r2, s12
+; CHECK-NEXT:    vmov r3, s16
+; CHECK-NEXT:    vmov r4, s14
+; CHECK-NEXT:    muls r2, r3, r2
+; CHECK-NEXT:    adds.w r12, r12, r2
+; CHECK-NEXT:    vmov r2, s18
+; CHECK-NEXT:    adc r3, lr, #0
+; CHECK-NEXT:    muls r2, r4, r2
+; CHECK-NEXT:    vmov.u8 r4, q1[15]
+; CHECK-NEXT:    adds.w r12, r12, r2
+; CHECK-NEXT:    vmov.u8 r2, q1[14]
+; CHECK-NEXT:    vmov q1[2], q1[0], r2, r4
+; CHECK-NEXT:    adc lr, r3, #0
+; CHECK-NEXT:    vmov.u8 r4, q0[15]
+; CHECK-NEXT:    vmov.u8 r3, q0[14]
+; CHECK-NEXT:    vmov q0[2], q0[0], r3, r4
 ; CHECK-NEXT:    vand q1, q1, q2
 ; CHECK-NEXT:    vand q0, q0, q2
-; CHECK-NEXT:    vmov r6, s4
-; CHECK-NEXT:    vmov r5, s0
-; CHECK-NEXT:    umlal r3, r2, r5, r6
-; CHECK-NEXT:    vmov r6, s6
-; CHECK-NEXT:    vmov r5, s2
-; CHECK-NEXT:    umlal r3, r2, r5, r6
-; CHECK-NEXT:    adds r0, r0, r3
-; CHECK-NEXT:    adcs r1, r2
-; CHECK-NEXT:    vpop {d8, d9, d10, d11}
-; CHECK-NEXT:    pop {r4, r5, r6, pc}
+; CHECK-NEXT:    vmov r2, s4
+; CHECK-NEXT:    vmov r3, s0
+; CHECK-NEXT:    vmov r4, s6
+; CHECK-NEXT:    muls r2, r3, r2
+; CHECK-NEXT:    adds.w r12, r12, r2
+; CHECK-NEXT:    vmov r2, s2
+; CHECK-NEXT:    adc r3, lr, #0
+; CHECK-NEXT:    muls r2, r4, r2
+; CHECK-NEXT:    adds.w r2, r2, r12
+; CHECK-NEXT:    adc r3, r3, #0
+; CHECK-NEXT:    adds r0, r0, r2
+; CHECK-NEXT:    adcs r1, r3
+; CHECK-NEXT:    vpop {d8, d9}
+; CHECK-NEXT:    pop {r4, pc}
 entry:
   %xx = zext <16 x i8> %x to <16 x i64>
   %yy = zext <16 x i8> %y to <16 x i64>
   %m = mul <16 x i64> %xx, %yy
   %z = call i64 @llvm.vector.reduce.add.v16i64(<16 x i64> %m)
   %r = add i64 %z, %a
   ret i64 %r
 }
 
 define arm_aapcs_vfpcc i64 @add_v16i8_v16i64_acc_sext(<16 x i8> %x, <16 x i8> %y, i64 %a) {
 ; CHECK-LABEL: add_v16i8_v16i64_acc_sext:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .save {r4, r5, r7, lr}
-; CHECK-NEXT:    push {r4, r5, r7, lr}
+; CHECK-NEXT:    .save {r7, lr}
+; CHECK-NEXT:    push {r7, lr}
 ; CHECK-NEXT:    vmov.s8 r2, q1[1]
 ; CHECK-NEXT:    vmov.s8 r3, q0[1]
-; CHECK-NEXT:    smull r12, r3, r3, r2
-; CHECK-NEXT:    vmov.s8 lr, q1[0]
+; CHECK-NEXT:    smull r3, lr, r3, r2
+; CHECK-NEXT:    vmov.s8 r12, q1[0]
 ; CHECK-NEXT:    vmov.s8 r2, q0[0]
-; CHECK-NEXT:    vmov.s8 r4, q1[2]
-; CHECK-NEXT:    vmov.s8 r5, q0[2]
-; CHECK-NEXT:    smull r2, lr, r2, lr
-; CHECK-NEXT:    vmov q2[2], q2[0], r2, r12
-; CHECK-NEXT:    smull r4, r5, r5, r4
-; CHECK-NEXT:    vmov q2[3], q2[1], lr, r3
-; CHECK-NEXT:    vmov lr, s10
-; CHECK-NEXT:    vmov r2, s8
-; CHECK-NEXT:    vmov r12, s9
-; CHECK-NEXT:    adds.w lr, lr, r2
-; CHECK-NEXT:    vmov.s8 r2, q1[3]
-; CHECK-NEXT:    adc.w r12, r12, r3
-; CHECK-NEXT:    vmov.s8 r3, q0[3]
-; CHECK-NEXT:    smull r2, r3, r3, r2
-; CHECK-NEXT:    vmov q2[2], q2[0], r4, r2
-; CHECK-NEXT:    vmov q2[3], q2[1], r5, r3
-; CHECK-NEXT:    vmov r4, s8
-; CHECK-NEXT:    vmov r2, s9
-; CHECK-NEXT:    adds.w r5, lr, r4
-; CHECK-NEXT:    vmov r4, s10
-; CHECK-NEXT:    adc.w r2, r2, r12
-; CHECK-NEXT:    adds.w lr, r5, r4
-; CHECK-NEXT:    vmov.s8 r4, q0[5]
-; CHECK-NEXT:    adc.w r12, r2, r3
-; CHECK-NEXT:    vmov.s8 r3, q1[5]
-; CHECK-NEXT:    vmov.s8 r2, q1[4]
-; CHECK-NEXT:    vmov.s8 r5, q0[4]
-; CHECK-NEXT:    smull r3, r4, r4, r3
-; CHECK-NEXT:    smull r2, r5, r5, r2
-; CHECK-NEXT:    vmov q2[2], q2[0], r2, r3
-; CHECK-NEXT:    vmov q2[3], q2[1], r5, r4
-; CHECK-NEXT:    vmov r3, s8
-; CHECK-NEXT:    vmov r2, s9
-; CHECK-NEXT:    vmov r5, s10
-; CHECK-NEXT:    adds.w r3, r3, lr
-; CHECK-NEXT:    adc.w r2, r2, r12
-; CHECK-NEXT:    adds.w lr, r3, r5
-; CHECK-NEXT:    vmov.s8 r5, q1[7]
-; CHECK-NEXT:    adc.w r12, r2, r4
-; CHECK-NEXT:    vmov.s8 r4, q0[7]
-; CHECK-NEXT:    vmov.s8 r2, q1[6]
-; CHECK-NEXT:    vmov.s8 r3, q0[6]
-; CHECK-NEXT:    smull r5, r4, r4, r5
-; CHECK-NEXT:    smull r2, r3, r3, r2
-; CHECK-NEXT:    vmov q2[2], q2[0], r2, r5
-; CHECK-NEXT:    vmov q2[3], q2[1], r3, r4
-; CHECK-NEXT:    vmov r3, s8
-; CHECK-NEXT:    vmov r2, s9
-; CHECK-NEXT:    vmov r5, s10
-; CHECK-NEXT:    adds.w r3, r3, lr
-; CHECK-NEXT:    adc.w r2, r2, r12
-; CHECK-NEXT:    adds.w lr, r3, r5
-; CHECK-NEXT:    vmov.s8 r5, q1[9]
-; CHECK-NEXT:    adc.w r12, r2, r4
-; CHECK-NEXT:    vmov.s8 r4, q0[9]
-; CHECK-NEXT:    vmov.s8 r2, q1[8]
-; CHECK-NEXT:    vmov.s8 r3, q0[8]
-; CHECK-NEXT:    smull r5, r4, r4, r5
-; CHECK-NEXT:    smull r2, r3, r3, r2
-; CHECK-NEXT:    vmov q2[2], q2[0], r2, r5
-; CHECK-NEXT:    vmov q2[3], q2[1], r3, r4
-; CHECK-NEXT:    vmov r3, s8
-; CHECK-NEXT:    vmov r2, s9
-; CHECK-NEXT:    vmov r5, s10
-; CHECK-NEXT:    adds.w r3, r3, lr
-; CHECK-NEXT:    adc.w r2, r2, r12
-; CHECK-NEXT:    adds.w lr, r3, r5
-; CHECK-NEXT:    vmov.s8 r5, q1[11]
-; CHECK-NEXT:    adc.w r12, r2, r4
-; CHECK-NEXT:    vmov.s8 r4, q0[11]
-; CHECK-NEXT:    vmov.s8 r2, q1[10]
-; CHECK-NEXT:    vmov.s8 r3, q0[10]
-; CHECK-NEXT:    smull r5, r4, r4, r5
-; CHECK-NEXT:    smull r2, r3, r3, r2
-; CHECK-NEXT:    vmov q2[2], q2[0], r2, r5
-; CHECK-NEXT:    vmov q2[3], q2[1], r3, r4
-; CHECK-NEXT:    vmov r3, s8
-; CHECK-NEXT:    vmov r2, s9
-; CHECK-NEXT:    vmov r5, s10
-; CHECK-NEXT:    adds.w r3, r3, lr
-; CHECK-NEXT:    adc.w r2, r2, r12
-; CHECK-NEXT:    adds.w lr, r3, r5
-; CHECK-NEXT:    vmov.s8 r5, q1[13]
-; CHECK-NEXT:    adc.w r12, r2, r4
-; CHECK-NEXT:    vmov.s8 r4, q0[13]
-; CHECK-NEXT:    vmov.s8 r2, q1[12]
-; CHECK-NEXT:    vmov.s8 r3, q0[12]
-; CHECK-NEXT:    smull r5, r4, r4, r5
-; CHECK-NEXT:    smull r2, r3, r3, r2
-; CHECK-NEXT:    vmov q2[2], q2[0], r2, r5
-; CHECK-NEXT:    vmov q2[3], q2[1], r3, r4
-; CHECK-NEXT:    vmov r3, s8
-; CHECK-NEXT:    vmov r2, s9
-; CHECK-NEXT:    vmov r5, s10
-; CHECK-NEXT:    adds.w r3, r3, lr
-; CHECK-NEXT:    adc.w r2, r2, r12
-; CHECK-NEXT:    adds r3, r3, r5
-; CHECK-NEXT:    vmov.s8 r5, q1[14]
-; CHECK-NEXT:    adcs r2, r4
-; CHECK-NEXT:    vmov.s8 r4, q0[14]
-; CHECK-NEXT:    smlal r3, r2, r4, r5
-; CHECK-NEXT:    vmov.s8 r5, q1[15]
-; CHECK-NEXT:    vmov.s8 r4, q0[15]
-; CHECK-NEXT:    smlal r3, r2, r4, r5
+; CHECK-NEXT:    smlal r3, lr, r2, r12
+; CHECK-NEXT:    vmov.s8 r12, q1[2]
+; CHECK-NEXT:    vmov.s8 r2, q0[2]
+; CHECK-NEXT:    smlal r3, lr, r2, r12
+; CHECK-NEXT:    vmov.s8 r12, q1[3]
+; CHECK-NEXT:    vmov.s8 r2, q0[3]
+; CHECK-NEXT:    smlal r3, lr, r2, r12
+; CHECK-NEXT:    vmov.s8 r12, q1[4]
+; CHECK-NEXT:    vmov.s8 r2, q0[4]
+; CHECK-NEXT:    smlal r3, lr, r2, r12
+; CHECK-NEXT:    vmov.s8 r12, q1[5]
+; CHECK-NEXT:    vmov.s8 r2, q0[5]
+; CHECK-NEXT:    smlal r3, lr, r2, r12
+; CHECK-NEXT:    vmov.s8 r12, q1[6]
+; CHECK-NEXT:    vmov.s8 r2, q0[6]
+; CHECK-NEXT:    smlal r3, lr, r2, r12
+; CHECK-NEXT:    vmov.s8 r12, q1[7]
+; CHECK-NEXT:    vmov.s8 r2, q0[7]
+; CHECK-NEXT:    smlal r3, lr, r2, r12
+; CHECK-NEXT:    vmov.s8 r12, q1[8]
+; CHECK-NEXT:    vmov.s8 r2, q0[8]
+; CHECK-NEXT:    smlal r3, lr, r2, r12
+; CHECK-NEXT:    vmov.s8 r12, q1[9]
+; CHECK-NEXT:    vmov.s8 r2, q0[9]
+; CHECK-NEXT:    smlal r3, lr, r2, r12
+; CHECK-NEXT:    vmov.s8 r12, q1[10]
+; CHECK-NEXT:    vmov.s8 r2, q0[10]
+; CHECK-NEXT:    smlal r3, lr, r2, r12
+; CHECK-NEXT:    vmov.s8 r12, q1[11]
+; CHECK-NEXT:    vmov.s8 r2, q0[11]
+; CHECK-NEXT:    smlal r3, lr, r2, r12
+; CHECK-NEXT:    vmov.s8 r12, q1[12]
+; CHECK-NEXT:    vmov.s8 r2, q0[12]
+; CHECK-NEXT:    smlal r3, lr, r2, r12
+; CHECK-NEXT:    vmov.s8 r12, q1[13]
+; CHECK-NEXT:    vmov.s8 r2, q0[13]
+; CHECK-NEXT:    smlal r3, lr, r2, r12
+; CHECK-NEXT:    vmov.s8 r12, q1[14]
+; CHECK-NEXT:    vmov.s8 r2, q0[14]
+; CHECK-NEXT:    smlal r3, lr, r2, r12
+; CHECK-NEXT:    vmov.s8 r12, q1[15]
+; CHECK-NEXT:    vmov.s8 r2, q0[15]
+; CHECK-NEXT:    smlal r3, lr, r2, r12
 ; CHECK-NEXT:    adds r0, r0, r3
-; CHECK-NEXT:    adcs r1, r2
-; CHECK-NEXT:    pop {r4, r5, r7, pc}
+; CHECK-NEXT:    adc.w r1, r1, lr
+; CHECK-NEXT:    pop {r7, pc}
 entry:
   %xx = sext <16 x i8> %x to <16 x i64>
   %yy = sext <16 x i8> %y to <16 x i64>
   %m = mul <16 x i64> %xx, %yy
   %z = call i64 @llvm.vector.reduce.add.v16i64(<16 x i64> %m)
   %r = add i64 %z, %a
   ret i64 %r
 }
 
 define arm_aapcs_vfpcc i64 @add_v2i8_v2i64_acc_zext(<2 x i8> %x, <2 x i8> %y, i64 %a) {
 ; CHECK-LABEL: add_v2i8_v2i64_acc_zext:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .save {r7, lr}
-; CHECK-NEXT:    push {r7, lr}
 ; CHECK-NEXT:    vmov.i64 q2, #0xff
 ; CHECK-NEXT:    vand q1, q1, q2
 ; CHECK-NEXT:    vand q0, q0, q2
 ; CHECK-NEXT:    vmov r2, s6
 ; CHECK-NEXT:    vmov r3, s2
-; CHECK-NEXT:    umull r12, lr, r3, r2
-; CHECK-NEXT:    vmov r2, s4
-; CHECK-NEXT:    vmov r3, s0
-; CHECK-NEXT:    umull r2, r3, r3, r2
-; CHECK-NEXT:    add r2, r12
-; CHECK-NEXT:    orr.w r3, r3, lr
+; CHECK-NEXT:    mul r12, r3, r2
+; CHECK-NEXT:    vmov r3, s4
+; CHECK-NEXT:    vmov r2, s0
+; CHECK-NEXT:    mla r2, r2, r3, r12
 ; CHECK-NEXT:    adds r0, r0, r2
-; CHECK-NEXT:    adcs r1, r3
-; CHECK-NEXT:    pop {r7, pc}
+; CHECK-NEXT:    adc r1, r1, #0
+; CHECK-NEXT:    bx lr
 entry:
   %xx = zext <2 x i8> %x to <2 x i64>
   %yy = zext <2 x i8> %y to <2 x i64>
   %m = mul <2 x i64> %xx, %yy
   %z = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> %m)
   %r = add i64 %z, %a
   ret i64 %r
 }
 
 define arm_aapcs_vfpcc i64 @add_v2i8_v2i64_acc_sext(<2 x i8> %x, <2 x i8> %y, i64 %a) {
 ; CHECK-LABEL: add_v2i8_v2i64_acc_sext:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .save {r7, lr}
 ; CHECK-NEXT:    push {r7, lr}
-; CHECK-NEXT:    vmov r2, s4
-; CHECK-NEXT:    vmov r3, s0
+; CHECK-NEXT:    vmov r2, s6
+; CHECK-NEXT:    vmov r3, s2
 ; CHECK-NEXT:    sxtb r2, r2
 ; CHECK-NEXT:    sxtb r3, r3
 ; CHECK-NEXT:    smull r2, r12, r3, r2
-; CHECK-NEXT:    vmov r3, s6
+; CHECK-NEXT:    vmov r3, s4
 ; CHECK-NEXT:    sxtb.w lr, r3
-; CHECK-NEXT:    vmov r3, s2
+; CHECK-NEXT:    vmov r3, s0
 ; CHECK-NEXT:    sxtb r3, r3
 ; CHECK-NEXT:    smlal r2, r12, r3, lr
 ; CHECK-NEXT:    adds r0, r0, r2
 ; CHECK-NEXT:    adc.w r1, r1, r12
 ; CHECK-NEXT:    pop {r7, pc}
 entry:
   %xx = sext <2 x i8> %x to <2 x i64>
   %yy = sext <2 x i8> %y to <2 x i64>
   %m = mul <2 x i64> %xx, %yy
   %z = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> %m)
   %r = add i64 %z, %a
   ret i64 %r
 }
 
 define arm_aapcs_vfpcc i64 @add_v2i64_v2i64_acc(<2 x i64> %x, <2 x i64> %y, i64 %a) {
 ; CHECK-LABEL: add_v2i64_v2i64_acc:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .save {r4, r5, r6, lr}
 ; CHECK-NEXT:    push {r4, r5, r6, lr}
-; CHECK-NEXT:    vmov r2, s4
-; CHECK-NEXT:    vmov r3, s0
-; CHECK-NEXT:    vmov r4, s5
-; CHECK-NEXT:    vmov r6, s7
-; CHECK-NEXT:    umull lr, r12, r3, r2
-; CHECK-NEXT:    mla r3, r3, r4, r12
-; CHECK-NEXT:    vmov r4, s1
-; CHECK-NEXT:    mla r12, r4, r2, r3
-; CHECK-NEXT:    vmov r3, s6
-; CHECK-NEXT:    vmov r4, s2
+; CHECK-NEXT:    vmov r2, s6
+; CHECK-NEXT:    vmov r3, s2
+; CHECK-NEXT:    vmov r4, s7
+; CHECK-NEXT:    vmov r6, s5
+; CHECK-NEXT:    umull r12, lr, r3, r2
+; CHECK-NEXT:    mla r3, r3, r4, lr
+; CHECK-NEXT:    vmov r4, s3
+; CHECK-NEXT:    mla lr, r4, r2, r3
+; CHECK-NEXT:    vmov r3, s4
+; CHECK-NEXT:    vmov r4, s0
 ; CHECK-NEXT:    umull r2, r5, r4, r3
 ; CHECK-NEXT:    mla r4, r4, r6, r5
-; CHECK-NEXT:    vmov r5, s3
-; CHECK-NEXT:    adds.w r2, r2, lr
+; CHECK-NEXT:    vmov r5, s1
+; CHECK-NEXT:    adds.w r2, r2, r12
 ; CHECK-NEXT:    mla r3, r5, r3, r4
-; CHECK-NEXT:    adc.w r3, r3, r12
+; CHECK-NEXT:    adc.w r3, r3, lr
 ; CHECK-NEXT:    adds r0, r0, r2
 ; CHECK-NEXT:    adcs r1, r3
 ; CHECK-NEXT:    pop {r4, r5, r6, pc}
 entry:
   %m = mul <2 x i64> %x, %y
   %z = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> %m)
   %r = add i64 %z, %a
   ret i64 %r

llvm/test/CodeGen/Thumb2/mve-vld2-post.ll

entry:
   ret <32 x i8> *%ret
 }
 
 ; i64
 
 define <4 x i64> *@vld2_v2i64(<4 x i64> *%src, <2 x i64> *%dst) {
 ; CHECK-LABEL: vld2_v2i64:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .save {r4, r5, r6, lr}
-; CHECK-NEXT:    push {r4, r5, r6, lr}
-; CHECK-NEXT:    vldrw.u32 q2, [r0, #16]
+; CHECK-NEXT:    .save {r4, r5, r7, lr}
+; CHECK-NEXT:    push {r4, r5, r7, lr}
+; CHECK-NEXT:    vldrw.u32 q1, [r0, #16]
 ; CHECK-NEXT:    vldrw.u32 q0, [r0], #32
-; CHECK-NEXT:    vmov.f64 d2, d1
-; CHECK-NEXT:    vmov.f32 s5, s3
-; CHECK-NEXT:    vmov.f32 s6, s10
-; CHECK-NEXT:    vmov.f32 s2, s8
-; CHECK-NEXT:    vmov.f32 s7, s11
-; CHECK-NEXT:    vmov.f32 s3, s9
-; CHECK-NEXT:    vmov r2, s6
-; CHECK-NEXT:    vmov r3, s2
-; CHECK-NEXT:    vmov r4, s4
+; CHECK-NEXT:    vmov.f64 d4, d0
+; CHECK-NEXT:    vmov.f32 s9, s1
+; CHECK-NEXT:    vmov.f32 s10, s4
+; CHECK-NEXT:    vmov.f32 s11, s5
+; CHECK-NEXT:    vmov lr, s10
+; CHECK-NEXT:    vmov r3, s6
+; CHECK-NEXT:    vmov r2, s7
 ; CHECK-NEXT:    vmov r5, s0
-; CHECK-NEXT:    vmov r12, s7
-; CHECK-NEXT:    vmov lr, s3
-; CHECK-NEXT:    adds r6, r3, r2
-; CHECK-NEXT:    vmov r3, s5
-; CHECK-NEXT:    vmov r2, s1
-; CHECK-NEXT:    adc.w r12, r12, lr
-; CHECK-NEXT:    adds r5, r5, r4
-; CHECK-NEXT:    vmov q0[2], q0[0], r5, r6
-; CHECK-NEXT:    adcs r2, r3
+; CHECK-NEXT:    vmov r4, s1
+; CHECK-NEXT:    vmov r12, s11
+; CHECK-NEXT:    vmov.f64 d4, d1
+; CHECK-NEXT:    vmov.f32 s9, s3
+; CHECK-NEXT:    adds.w lr, lr, r3
+; CHECK-NEXT:    vmov r3, s8
+; CHECK-NEXT:    adc.w r12, r12, r2
+; CHECK-NEXT:    vmov r2, s9
+; CHECK-NEXT:    adds r3, r3, r5
+; CHECK-NEXT:    vmov q0[2], q0[0], r3, lr
+; CHECK-NEXT:    adcs r2, r4
 ; CHECK-NEXT:    vmov q0[3], q0[1], r2, r12
 ; CHECK-NEXT:    vstrw.32 q0, [r1]
-; CHECK-NEXT:    pop {r4, r5, r6, pc}
+; CHECK-NEXT:    pop {r4, r5, r7, pc}
 entry:
   %l1 = load <4 x i64>, <4 x i64>* %src, align 4
   %s1 = shufflevector <4 x i64> %l1, <4 x i64> undef, <2 x i32> <i32 0, i32 2>
   %s2 = shufflevector <4 x i64> %l1, <4 x i64> undef, <2 x i32> <i32 1, i32 3>
   %a = add <2 x i64> %s1, %s2
   store <2 x i64> %a, <2 x i64> *%dst
   %ret = getelementptr inbounds <4 x i64>, <4 x i64>* %src, i32 1
   ret <4 x i64> *%ret

llvm/test/CodeGen/Thumb2/mve-vld2.ll

 ; i64
 
 define void @vld2_v2i64(<4 x i64> *%src, <2 x i64> *%dst) {
 ; CHECK-LABEL: vld2_v2i64:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .save {r4, lr}
 ; CHECK-NEXT:    push {r4, lr}
 ; CHECK-NEXT:    vldrw.u32 q0, [r0]
-; CHECK-NEXT:    vldrw.u32 q2, [r0, #16]
-; CHECK-NEXT:    vmov.f64 d2, d1
-; CHECK-NEXT:    vmov.f32 s5, s3
-; CHECK-NEXT:    vmov.f32 s6, s10
-; CHECK-NEXT:    vmov.f32 s2, s8
-; CHECK-NEXT:    vmov.f32 s3, s9
-; CHECK-NEXT:    vmov.f32 s7, s11
-; CHECK-NEXT:    vmov r3, s6
-; CHECK-NEXT:    vmov r0, s2
+; CHECK-NEXT:    vldrw.u32 q1, [r0, #16]
+; CHECK-NEXT:    vmov.f64 d4, d0
+; CHECK-NEXT:    vmov.f32 s9, s1
+; CHECK-NEXT:    vmov.f32 s10, s4
+; CHECK-NEXT:    vmov.f32 s11, s5
+; CHECK-NEXT:    vmov r2, s10
+; CHECK-NEXT:    vmov r0, s6
+; CHECK-NEXT:    vmov r3, s7
 ; CHECK-NEXT:    vmov r4, s0
-; CHECK-NEXT:    vmov r2, s3
-; CHECK-NEXT:    vmov r12, s7
-; CHECK-NEXT:    adds.w lr, r0, r3
-; CHECK-NEXT:    vmov r0, s4
-; CHECK-NEXT:    vmov r3, s5
-; CHECK-NEXT:    adc.w r12, r12, r2
-; CHECK-NEXT:    vmov r2, s1
-; CHECK-NEXT:    adds r0, r0, r4
-; CHECK-NEXT:    vmov q0[2], q0[0], r0, lr
-; CHECK-NEXT:    adcs r2, r3
-; CHECK-NEXT:    vmov q0[3], q0[1], r2, r12
+; CHECK-NEXT:    vmov r12, s11
+; CHECK-NEXT:    vmov.f64 d4, d1
+; CHECK-NEXT:    vmov.f32 s9, s3
+; CHECK-NEXT:    adds.w lr, r2, r0
+; CHECK-NEXT:    vmov r2, s8
+; CHECK-NEXT:    vmov r0, s1
+; CHECK-NEXT:    adc.w r12, r12, r3
+; CHECK-NEXT:    vmov r3, s9
+; CHECK-NEXT:    adds r2, r2, r4
+; CHECK-NEXT:    vmov q0[2], q0[0], r2, lr
+; CHECK-NEXT:    adcs r0, r3
+; CHECK-NEXT:    vmov q0[3], q0[1], r0, r12
 ; CHECK-NEXT:    vstrw.32 q0, [r1]
 ; CHECK-NEXT:    pop {r4, pc}
 entry:
   %l1 = load <4 x i64>, <4 x i64>* %src, align 8
   %s1 = shufflevector <4 x i64> %l1, <4 x i64> undef, <2 x i32> <i32 0, i32 2>
   %s2 = shufflevector <4 x i64> %l1, <4 x i64> undef, <2 x i32> <i32 1, i32 3>
   %a = add <2 x i64> %s1, %s2
   store <2 x i64> %a, <2 x i64> *%dst
   ret void
 }
 
 define void @vld2_v4i64(<8 x i64> *%src, <4 x i64> *%dst) {
 ; CHECK-LABEL: vld2_v4i64:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .save {r4, r5, r6, lr}
 ; CHECK-NEXT:    push {r4, r5, r6, lr}
 ; CHECK-NEXT:    .vsave {d8, d9, d10, d11}
 ; CHECK-NEXT:    vpush {d8, d9, d10, d11}
 ; CHECK-NEXT:    vldrw.u32 q0, [r0]
-; CHECK-NEXT:    vldrw.u32 q5, [r0, #16]
+; CHECK-NEXT:    vldrw.u32 q4, [r0, #16]
 ; CHECK-NEXT:    vldrw.u32 q1, [r0, #32]
-; CHECK-NEXT:    vldrw.u32 q4, [r0, #48]
-; CHECK-NEXT:    vmov.f64 d4, d1
-; CHECK-NEXT:    vmov.f32 s9, s3
-; CHECK-NEXT:    vmov.f32 s10, s22
-; CHECK-NEXT:    vmov.f32 s2, s20
-; CHECK-NEXT:    vmov.f32 s11, s23
-; CHECK-NEXT:    vmov.f32 s3, s21
-; CHECK-NEXT:    vmov r3, s10
-; CHECK-NEXT:    vmov r0, s2
-; CHECK-NEXT:    vmov.f64 d6, d3
-; CHECK-NEXT:    vmov r12, s11
-; CHECK-NEXT:    vmov r2, s3
-; CHECK-NEXT:    vmov.f32 s13, s7
-; CHECK-NEXT:    vmov.f32 s14, s18
-; CHECK-NEXT:    vmov.f32 s6, s16
-; CHECK-NEXT:    vmov.f32 s7, s17
-; CHECK-NEXT:    vmov.f32 s15, s19
-; CHECK-NEXT:    vmov r4, s6
-; CHECK-NEXT:    vmov r5, s12
+; CHECK-NEXT:    vldrw.u32 q2, [r0, #48]
+; CHECK-NEXT:    vmov.f64 d10, d0
+; CHECK-NEXT:    vmov.f32 s21, s1
+; CHECK-NEXT:    vmov.f32 s22, s16
+; CHECK-NEXT:    vmov.f32 s23, s17
+; CHECK-NEXT:    vmov r2, s22
+; CHECK-NEXT:    vmov r0, s18
+; CHECK-NEXT:    vmov r3, s19
+; CHECK-NEXT:    vmov.f64 d6, d2
+; CHECK-NEXT:    vmov r12, s23
+; CHECK-NEXT:    vmov.f32 s13, s5
+; CHECK-NEXT:    vmov.f32 s14, s8
+; CHECK-NEXT:    vmov.f32 s15, s9
+; CHECK-NEXT:    vmov r4, s10
+; CHECK-NEXT:    vmov.f64 d10, d3
+; CHECK-NEXT:    vmov.f32 s21, s7
 ; CHECK-NEXT:    vmov r6, s4
-; CHECK-NEXT:    adds.w lr, r0, r3
-; CHECK-NEXT:    vmov r3, s14
-; CHECK-NEXT:    vmov r0, s7
-; CHECK-NEXT:    adc.w r12, r12, r2
-; CHECK-NEXT:    vmov r2, s15
-; CHECK-NEXT:    adds r3, r3, r4
-; CHECK-NEXT:    vmov r4, s5
-; CHECK-NEXT:    adcs r0, r2
-; CHECK-NEXT:    vmov r2, s13
-; CHECK-NEXT:    adds r5, r5, r6
+; CHECK-NEXT:    vmov r5, s5
+; CHECK-NEXT:    vmov.f64 d2, d1
+; CHECK-NEXT:    adds.w lr, r2, r0
+; CHECK-NEXT:    vmov r0, s14
+; CHECK-NEXT:    vmov r2, s11
+; CHECK-NEXT:    vmov.f32 s5, s3
+; CHECK-NEXT:    adc.w r12, r12, r3
+; CHECK-NEXT:    vmov r3, s15
+; CHECK-NEXT:    adds r0, r0, r4
+; CHECK-NEXT:    vmov r4, s20
+; CHECK-NEXT:    adcs r2, r3
+; CHECK-NEXT:    vmov r3, s21
+; CHECK-NEXT:    adds r4, r4, r6
 ; CHECK-NEXT:    vmov r6, s0
-; CHECK-NEXT:    vmov q3[2], q3[0], r5, r3
-; CHECK-NEXT:    adcs r2, r4
-; CHECK-NEXT:    vmov r4, s8
-; CHECK-NEXT:    vmov q3[3], q3[1], r2, r0
-; CHECK-NEXT:    vmov r0, s9
+; CHECK-NEXT:    vmov q3[2], q3[0], r4, r0
+; CHECK-NEXT:    vmov r0, s5
+; CHECK-NEXT:    adcs r3, r5
+; CHECK-NEXT:    vmov r5, s4
+; CHECK-NEXT:    vmov q3[3], q3[1], r3, r2
 ; CHECK-NEXT:    vmov r2, s1
 ; CHECK-NEXT:    vstrw.32 q3, [r1, #16]
-; CHECK-NEXT:    adds r4, r4, r6
-; CHECK-NEXT:    vmov q1[2], q1[0], r4, lr
+; CHECK-NEXT:    adds r5, r5, r6
+; CHECK-NEXT:    vmov q2[2], q2[0], r5, lr
 ; CHECK-NEXT:    adcs r0, r2
-; CHECK-NEXT:    vmov q1[3], q1[1], r0, r12
-; CHECK-NEXT:    vstrw.32 q1, [r1]
+; CHECK-NEXT:    vmov q2[3], q2[1], r0, r12
+; CHECK-NEXT:    vstrw.32 q2, [r1]
 ; CHECK-NEXT:    vpop {d8, d9, d10, d11}
 ; CHECK-NEXT:    pop {r4, r5, r6, pc}
 entry:
   %l1 = load <8 x i64>, <8 x i64>* %src, align 8
   %s1 = shufflevector <8 x i64> %l1, <8 x i64> undef, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
   %s2 = shufflevector <8 x i64> %l1, <8 x i64> undef, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
   %a = add <4 x i64> %s1, %s2
   store <4 x i64> %a, <4 x i64> *%dst

 ; f16
 
 define void @vld2_v2f16(<4 x half> *%src, <2 x half> *%dst) {
 ; CHECK-LABEL: vld2_v2f16:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    ldr r2, [r0]
 ; CHECK-NEXT:    ldr r0, [r0, #4]
 ; CHECK-NEXT:    vmov.32 q0[0], r2
-; CHECK-NEXT:    vmov.32 q0[1], r0
-; CHECK-NEXT:    vmovx.f16 s4, s1
-; CHECK-NEXT:    vmovx.f16 s8, s0
-; CHECK-NEXT:    vins.f16 s8, s4
-; CHECK-NEXT:    vins.f16 s0, s1
-; CHECK-NEXT:    vadd.f16 q0, q0, q2
+; CHECK-NEXT:    vmov q1, q0
+; CHECK-NEXT:    vmovx.f16 s12, s0
+; CHECK-NEXT:    vmov.32 q1[1], r0
+; CHECK-NEXT:    vmovx.f16 s8, s5
+; CHECK-NEXT:    vins.f16 s0, s5
+; CHECK-NEXT:    vins.f16 s12, s8
+; CHECK-NEXT:    vadd.f16 q0, q0, q3
 ; CHECK-NEXT:    vmov r0, s0
 ; CHECK-NEXT:    str r0, [r1]
 ; CHECK-NEXT:    bx lr
 entry:
   %l1 = load <4 x half>, <4 x half>* %src, align 2
   %s1 = shufflevector <4 x half> %l1, <4 x half> undef, <2 x i32> <i32 0, i32 2>
   %s2 = shufflevector <4 x half> %l1, <4 x half> undef, <2 x i32> <i32 1, i32 3>
   %a = fadd <2 x half> %s1, %s2

llvm/test/CodeGen/Thumb2/mve-vld3.ll

entry:
   %a = add <4 x i16> %a1, %s3
   store <4 x i16> %a, <4 x i16> *%dst
   ret void
 }
 
 define void @vld3_v8i16(<24 x i16> *%src, <8 x i16> *%dst) {
 ; CHECK-LABEL: vld3_v8i16:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .vsave {d8, d9, d10, d11}
-; CHECK-NEXT:    vpush {d8, d9, d10, d11}
-; CHECK-NEXT:    vldrw.u32 q1, [r0]
-; CHECK-NEXT:    vmovx.f16 s8, s6
-; CHECK-NEXT:    vmov.f32 s0, s5
-; CHECK-NEXT:    vins.f16 s0, s8
+; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13}
+; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13}
+; CHECK-NEXT:    vldrw.u32 q0, [r0]
+; CHECK-NEXT:    vmovx.f16 s8, s2
+; CHECK-NEXT:    vmov.f32 s4, s1
+; CHECK-NEXT:    vins.f16 s4, s8
 ; CHECK-NEXT:    vldrw.u32 q2, [r0, #16]
 ; CHECK-NEXT:    vmovx.f16 s12, s9
-; CHECK-NEXT:    vmov.f32 s1, s8
-; CHECK-NEXT:    vins.f16 s1, s12
+; CHECK-NEXT:    vmov.f32 s5, s8
+; CHECK-NEXT:    vins.f16 s5, s12
 ; CHECK-NEXT:    vldrw.u32 q3, [r0, #32]
-; CHECK-NEXT:    vmov.f32 s2, s11
+; CHECK-NEXT:    vmov.f32 s6, s11
 ; CHECK-NEXT:    vmov.u16 r0, q2[5]
-; CHECK-NEXT:    vmovx.f16 s20, s15
-; CHECK-NEXT:    vmov.f32 s19, s14
-; CHECK-NEXT:    vins.f16 s19, s20
-; CHECK-NEXT:    vmov.f32 s18, s12
+; CHECK-NEXT:    vmovx.f16 s16, s15
+; CHECK-NEXT:    vmov.f32 s23, s14
+; CHECK-NEXT:    vins.f16 s23, s16
+; CHECK-NEXT:    vmov.f32 s22, s12
+; CHECK-NEXT:    vmov q4, q5
+; CHECK-NEXT:    vmovnb.i32 q4, q1
+; CHECK-NEXT:    vmov.f32 s6, s18
+; CHECK-NEXT:    vmovx.f16 s16, s0
+; CHECK-NEXT:    vins.f16 s16, s2
+; CHECK-NEXT:    vmov.f32 s7, s23
+; CHECK-NEXT:    vmovx.f16 s17, s3
+; CHECK-NEXT:    vins.f16 s17, s9
+; CHECK-NEXT:    vins.f16 s18, s12
 ; CHECK-NEXT:    vmov q5, q4
-; CHECK-NEXT:    vmovnb.i32 q5, q0
-; CHECK-NEXT:    vmov.f32 s2, s22
-; CHECK-NEXT:    vmovx.f16 s20, s5
-; CHECK-NEXT:    vmov.f32 s3, s19
-; CHECK-NEXT:    vmov.f64 d8, d2
-; CHECK-NEXT:    vins.f16 s16, s20
+; CHECK-NEXT:    vmovx.f16 s19, s13
+; CHECK-NEXT:    vins.f16 s19, s15
+; CHECK-NEXT:    vmov.16 q5[4], r0
+; CHECK-NEXT:    vmov q6, q4
+; CHECK-NEXT:    vmovnb.i32 q6, q5
+; CHECK-NEXT:    vmovx.f16 s20, s1
+; CHECK-NEXT:    vins.f16 s0, s20
 ; CHECK-NEXT:    vmovx.f16 s20, s8
-; CHECK-NEXT:    vmov.f32 s17, s7
-; CHECK-NEXT:    vins.f16 s17, s20
+; CHECK-NEXT:    vins.f16 s3, s20
 ; CHECK-NEXT:    vmovx.f16 s20, s11
-; CHECK-NEXT:    vmov.f32 s18, s10
-; CHECK-NEXT:    vins.f16 s18, s20
-; CHECK-NEXT:    vmovx.f16 s20, s14
-; CHECK-NEXT:    vmov.f32 s19, s13
-; CHECK-NEXT:    vins.f16 s19, s20
-; CHECK-NEXT:    vmovx.f16 s20, s4
-; CHECK-NEXT:    vins.f16 s20, s6
-; CHECK-NEXT:    vmovx.f16 s21, s7
-; CHECK-NEXT:    vins.f16 s6, s12
-; CHECK-NEXT:    vmovx.f16 s7, s13
-; CHECK-NEXT:    vins.f16 s21, s9
-; CHECK-NEXT:    vins.f16 s7, s15
-; CHECK-NEXT:    vmov.16 q5[4], r0
-; CHECK-NEXT:    vmov q2, q1
-; CHECK-NEXT:    vmovnb.i32 q2, q5
-; CHECK-NEXT:    vmov.f32 s22, s10
-; CHECK-NEXT:    vmov.f32 s23, s7
-; CHECK-NEXT:    vadd.i16 q1, q4, q5
-; CHECK-NEXT:    vadd.i16 q0, q1, q0
+; CHECK-NEXT:    vmov.f32 s1, s3
+; CHECK-NEXT:    vins.f16 s10, s20
+; CHECK-NEXT:    vmov.f32 s2, s10
+; CHECK-NEXT:    vmovx.f16 s8, s14
+; CHECK-NEXT:    vins.f16 s13, s8
+; CHECK-NEXT:    vmov.f32 s18, s26
+; CHECK-NEXT:    vmov.f32 s3, s13
+; CHECK-NEXT:    vadd.i16 q0, q0, q4
+; CHECK-NEXT:    vadd.i16 q0, q0, q1
 ; CHECK-NEXT:    vstrw.32 q0, [r1]
-; CHECK-NEXT:    vpop {d8, d9, d10, d11}
+; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13}
 ; CHECK-NEXT:    bx lr
 entry:
   %l1 = load <24 x i16>, <24 x i16>* %src, align 4
   %s1 = shufflevector <24 x i16> %l1, <24 x i16> undef, <8 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21>
   %s2 = shufflevector <24 x i16> %l1, <24 x i16> undef, <8 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22>
   %s3 = shufflevector <24 x i16> %l1, <24 x i16> undef, <8 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23>
   %a1 = add <8 x i16> %s1, %s2
   %a = add <8 x i16> %a1, %s3
   store <8 x i16> %a, <8 x i16> *%dst
   ret void
 }
 
 define void @vld3_v16i16(<48 x i16> *%src, <16 x i16> *%dst) {
 ; CHECK-LABEL: vld3_v16i16:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13, d14, d15}
 ; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13, d14, d15}
 ; CHECK-NEXT:    .pad #16
 ; CHECK-NEXT:    sub sp, #16
-; CHECK-NEXT:    vldrw.u32 q1, [r0, #48]
-; CHECK-NEXT:    vmovx.f16 s8, s6
-; CHECK-NEXT:    vmov.f32 s0, s5
-; CHECK-NEXT:    vins.f16 s0, s8
-; CHECK-NEXT:    vldrw.u32 q2, [r0, #64]
-; CHECK-NEXT:    vmovx.f16 s12, s9
-; CHECK-NEXT:    vmov.f32 s1, s8
-; CHECK-NEXT:    vins.f16 s1, s12
-; CHECK-NEXT:    vldrw.u32 q3, [r0, #80]
-; CHECK-NEXT:    vmov.f32 s2, s11
-; CHECK-NEXT:    vmov.u16 r2, q2[5]
-; CHECK-NEXT:    vmovx.f16 s20, s15
-; CHECK-NEXT:    vmov.f32 s19, s14
-; CHECK-NEXT:    vins.f16 s19, s20
-; CHECK-NEXT:    vmov.f32 s18, s12
-; CHECK-NEXT:    vmov q5, q4
-; CHECK-NEXT:    vmovnb.i32 q5, q0
+; CHECK-NEXT:    vldrw.u32 q1, [r0, #80]
+; CHECK-NEXT:    vldrw.u32 q2, [r0, #48]
+; CHECK-NEXT:    vldrw.u32 q3, [r0, #64]
+; CHECK-NEXT:    vins.f16 s2, s4
+; CHECK-NEXT:    vmovx.f16 s24, s10
+; CHECK-NEXT:    vmovx.f16 s3, s5
+; CHECK-NEXT:    vmov.u16 r2, q3[5]
+; CHECK-NEXT:    vins.f16 s3, s7
+; CHECK-NEXT:    vmov.f32 s31, s6
+; CHECK-NEXT:    vmovx.f16 s0, s8
+; CHECK-NEXT:    vins.f16 s0, s10
+; CHECK-NEXT:    vmovx.f16 s1, s11
+; CHECK-NEXT:    vins.f16 s1, s13
+; CHECK-NEXT:    vmov q4, q0
+; CHECK-NEXT:    vmov q5, q0
+; CHECK-NEXT:    vmov.16 q4[4], r2
+; CHECK-NEXT:    vmovnb.i32 q5, q4
+; CHECK-NEXT:    vmov.f32 s16, s9
+; CHECK-NEXT:    vins.f16 s16, s24
 ; CHECK-NEXT:    vmov.f32 s2, s22
-; CHECK-NEXT:    vmovx.f16 s20, s5
-; CHECK-NEXT:    vmov.f32 s3, s19
-; CHECK-NEXT:    vmov.f64 d8, d2
-; CHECK-NEXT:    vins.f16 s16, s20
-; CHECK-NEXT:    vmovx.f16 s20, s8
-; CHECK-NEXT:    vmov.f32 s17, s7
-; CHECK-NEXT:    vins.f16 s17, s20
-; CHECK-NEXT:    vmovx.f16 s20, s11
-; CHECK-NEXT:    vmov.f32 s18, s10
-; CHECK-NEXT:    vins.f16 s18, s20
-; CHECK-NEXT:    vmovx.f16 s20, s14
-; CHECK-NEXT:    vmov.f32 s19, s13
-; CHECK-NEXT:    vins.f16 s19, s20
-; CHECK-NEXT:    vmovx.f16 s20, s4
-; CHECK-NEXT:    vins.f16 s20, s6
-; CHECK-NEXT:    vmovx.f16 s21, s7
-; CHECK-NEXT:    vins.f16 s6, s12
-; CHECK-NEXT:    vmovx.f16 s7, s13
-; CHECK-NEXT:    vins.f16 s21, s9
-; CHECK-NEXT:    vins.f16 s7, s15
-; CHECK-NEXT:    vmov.16 q5[4], r2
-; CHECK-NEXT:    vmov q2, q1
-; CHECK-NEXT:    vmovnb.i32 q2, q5
-; CHECK-NEXT:    vmov.f32 s22, s10
-; CHECK-NEXT:    vldrw.u32 q2, [r0]
-; CHECK-NEXT:    vmov.f32 s23, s7
-; CHECK-NEXT:    vadd.i16 q1, q4, q5
-; CHECK-NEXT:    vmovx.f16 s12, s10
-; CHECK-NEXT:    vadd.i16 q0, q1, q0
-; CHECK-NEXT:    vmov.f32 s4, s9
-; CHECK-NEXT:    vins.f16 s4, s12
-; CHECK-NEXT:    vldrw.u32 q3, [r0, #16]
+; CHECK-NEXT:    vmovx.f16 s20, s9
+; CHECK-NEXT:    vmovx.f16 s24, s13
+; CHECK-NEXT:    vmov.f32 s17, s12
+; CHECK-NEXT:    vins.f16 s8, s20
+; CHECK-NEXT:    vmovx.f16 s20, s12
+; CHECK-NEXT:    vins.f16 s17, s24
+; CHECK-NEXT:    vins.f16 s11, s20
+; CHECK-NEXT:    vmovx.f16 s20, s15
+; CHECK-NEXT:    vmov.f32 s18, s15
+; CHECK-NEXT:    vmovx.f16 s24, s7
+; CHECK-NEXT:    vmov.f32 s9, s11
+; CHECK-NEXT:    vins.f16 s14, s20
+; CHECK-NEXT:    vins.f16 s31, s24
+; CHECK-NEXT:    vmov.f32 s10, s14
+; CHECK-NEXT:    vmovx.f16 s12, s6
+; CHECK-NEXT:    vmov.f32 s30, s4
+; CHECK-NEXT:    vins.f16 s5, s12
+; CHECK-NEXT:    vmov q6, q7
+; CHECK-NEXT:    vmov.f32 s11, s5
+; CHECK-NEXT:    vmovnb.i32 q6, q4
+; CHECK-NEXT:    vadd.i16 q0, q2, q0
+; CHECK-NEXT:    vldrw.u32 q2, [r0, #32]
+; CHECK-NEXT:    vldrw.u32 q3, [r0]
+; CHECK-NEXT:    vmov.f32 s18, s26
+; CHECK-NEXT:    vins.f16 s6, s8
+; CHECK-NEXT:    vmov.f32 s19, s31
+; CHECK-NEXT:    vmovx.f16 s7, s9
+; CHECK-NEXT:    vadd.i16 q0, q0, q4
+; CHECK-NEXT:    vins.f16 s7, s11
+; CHECK-NEXT:    vldrw.u32 q4, [r0, #16]
+; CHECK-NEXT:    vmovx.f16 s4, s12
 ; CHECK-NEXT:    vstrw.32 q0, [sp] @ 16-byte Spill
-; CHECK-NEXT:    vmovx.f16 s0, s9
-; CHECK-NEXT:    vmovx.f16 s16, s13
-; CHECK-NEXT:    vmov.f32 s5, s12
-; CHECK-NEXT:    vins.f16 s5, s16
-; CHECK-NEXT:    vldrw.u32 q4, [r0, #32]
-; CHECK-NEXT:    vmov.f32 s6, s15
-; CHECK-NEXT:    vmov.u16 r0, q3[5]
+; CHECK-NEXT:    vins.f16 s4, s14
+; CHECK-NEXT:    vmovx.f16 s28, s14
+; CHECK-NEXT:    vmovx.f16 s5, s15
+; CHECK-NEXT:    vmov.f32 s0, s13
+; CHECK-NEXT:    vins.f16 s5, s17
+; CHECK-NEXT:    vins.f16 s0, s28
+; CHECK-NEXT:    vmov.u16 r0, q4[5]
+; CHECK-NEXT:    vmov q5, q1
+; CHECK-NEXT:    vmov.16 q5[4], r0
+; CHECK-NEXT:    vmov q6, q1
+; CHECK-NEXT:    vmovx.f16 s28, s17
+; CHECK-NEXT:    vmov.f32 s1, s16
+; CHECK-NEXT:    vmovnb.i32 q6, q5
+; CHECK-NEXT:    vins.f16 s1, s28
+; CHECK-NEXT:    vmovx.f16 s28, s11
+; CHECK-NEXT:    vmov.f32 s23, s10
+; CHECK-NEXT:    vins.f16 s23, s28
+; CHECK-NEXT:    vmov.f32 s2, s19
+; CHECK-NEXT:    vmov.f32 s22, s8
+; CHECK-NEXT:    vmov q7, q5
+; CHECK-NEXT:    vmov.f32 s6, s26
+; CHECK-NEXT:    vmovnb.i32 q7, q0
+; CHECK-NEXT:    vmov.f32 s2, s30
+; CHECK-NEXT:    vmov.f32 s3, s23
+; CHECK-NEXT:    vmovx.f16 s20, s13
+; CHECK-NEXT:    vins.f16 s12, s20
+; CHECK-NEXT:    vmovx.f16 s20, s16
+; CHECK-NEXT:    vins.f16 s15, s20
 ; CHECK-NEXT:    vmovx.f16 s20, s19
-; CHECK-NEXT:    vmov.f32 s27, s18
-; CHECK-NEXT:    vins.f16 s27, s20
-; CHECK-NEXT:    vmov.f64 d10, d4
-; CHECK-NEXT:    vins.f16 s20, s0
-; CHECK-NEXT:    vmov.f32 s26, s16
-; CHECK-NEXT:    vmovx.f16 s0, s12
-; CHECK-NEXT:    vmov.f32 s21, s11
-; CHECK-NEXT:    vins.f16 s21, s0
-; CHECK-NEXT:    vmov q7, q6
-; CHECK-NEXT:    vmovnb.i32 q7, q1
-; CHECK-NEXT:    vmovx.f16 s0, s15
-; CHECK-NEXT:    vmov.f32 s22, s14
-; CHECK-NEXT:    vins.f16 s22, s0
-; CHECK-NEXT:    vmov.f32 s6, s30
-; CHECK-NEXT:    vmov.f32 s7, s27
-; CHECK-NEXT:    vmovx.f16 s24, s8
-; CHECK-NEXT:    vmovx.f16 s0, s18
-; CHECK-NEXT:    vmov.f32 s23, s17
-; CHECK-NEXT:    vins.f16 s24, s10
-; CHECK-NEXT:    vins.f16 s23, s0
-; CHECK-NEXT:    vins.f16 s2, s16
-; CHECK-NEXT:    vmovx.f16 s25, s11
-; CHECK-NEXT:    vmovx.f16 s3, s17
-; CHECK-NEXT:    vins.f16 s25, s13
-; CHECK-NEXT:    vins.f16 s3, s19
-; CHECK-NEXT:    vmov.16 q6[4], r0
-; CHECK-NEXT:    vmov q2, q0
-; CHECK-NEXT:    vmovnb.i32 q2, q6
-; CHECK-NEXT:    vmov.f32 s26, s10
-; CHECK-NEXT:    vmov.f32 s27, s3
-; CHECK-NEXT:    vadd.i16 q0, q5, q6
-; CHECK-NEXT:    vadd.i16 q0, q0, q1
+; CHECK-NEXT:    vmov.f32 s13, s15
+; CHECK-NEXT:    vins.f16 s18, s20
+; CHECK-NEXT:    vmov.f32 s14, s18
+; CHECK-NEXT:    vmovx.f16 s16, s10
+; CHECK-NEXT:    vins.f16 s9, s16
+; CHECK-NEXT:    vmov.f32 s15, s9
+; CHECK-NEXT:    vadd.i16 q1, q3, q1
+; CHECK-NEXT:    vadd.i16 q0, q1, q0
 ; CHECK-NEXT:    vldrw.u32 q1, [sp] @ 16-byte Reload
 ; CHECK-NEXT:    vstrw.32 q0, [r1]
 ; CHECK-NEXT:    vstrw.32 q1, [r1, #16]
 ; CHECK-NEXT:    add sp, #16
 ; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13, d14, d15}
 ; CHECK-NEXT:    bx lr
 entry:
   %l1 = load <48 x i16>, <48 x i16>* %src, align 4
entry:
   %a = add <8 x i8> %a1, %s3
   store <8 x i8> %a, <8 x i8> *%dst
   ret void
 }
 
 define void @vld3_v16i8(<48 x i8> *%src, <16 x i8> *%dst) {
 ; CHECK-LABEL: vld3_v16i8:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13}
-; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13}
+; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13, d14, d15}
+; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13, d14, d15}
 ; CHECK-NEXT:    vldrw.u32 q1, [r0]
 ; CHECK-NEXT:    vldrw.u32 q0, [r0, #16]
 ; CHECK-NEXT:    vldrw.u32 q2, [r0, #32]
 ; CHECK-NEXT:    vmov.u8 r2, q1[1]
 ; CHECK-NEXT:    vmov.8 q3[0], r2
 ; CHECK-NEXT:    vmov.u8 r2, q1[4]
 ; CHECK-NEXT:    vmov.8 q3[1], r2
 ; CHECK-NEXT:    vmov.u8 r2, q1[7]
 ; CHECK-NEXT:    vmov.8 q3[2], r2
 ; CHECK-NEXT:    vmov.u8 r2, q1[10]
 ; CHECK-NEXT:    vmov.8 q3[3], r2
 ; CHECK-NEXT:    vmov.u8 r2, q1[13]
-; CHECK-NEXT:    vmov.8 q3[4], r2
+; CHECK-NEXT:    vmov q4, q3
+; CHECK-NEXT:    vmov.u8 r0, q2[5]
+; CHECK-NEXT:    vmov.8 q4[4], r2
 ; CHECK-NEXT:    vmov.u8 r2, q0[0]
-; CHECK-NEXT:    vmov.8 q3[5], r2
+; CHECK-NEXT:    vmov.8 q4[5], r2
 ; CHECK-NEXT:    vmov.u8 r2, q0[3]
-; CHECK-NEXT:    vmov.8 q3[6], r2
+; CHECK-NEXT:    vmov.8 q4[6], r2
 ; CHECK-NEXT:    vmov.u8 r2, q0[6]
-; CHECK-NEXT:    vmov.8 q3[7], r2
+; CHECK-NEXT:    vmov.8 q4[7], r2
+; CHECK-NEXT:    vmov.8 q6[12], r0
+; CHECK-NEXT:    vmov.u8 r0, q2[8]
 ; CHECK-NEXT:    vmov.u8 r2, q0[9]
-; CHECK-NEXT:    vmov.u8 r0, q2[5]
-; CHECK-NEXT:    vmov.8 q3[8], r2
+; CHECK-NEXT:    vmov q5, q4
+; CHECK-NEXT:    vmov.8 q6[13], r0
+; CHECK-NEXT:    vmov.8 q5[8], r2
 ; CHECK-NEXT:    vmov.u8 r2, q0[12]
-; CHECK-NEXT:    vmov.8 q4[12], r0
-; CHECK-NEXT:    vmov.u8 r0, q2[8]
-; CHECK-NEXT:    vmov.8 q3[9], r2
-; CHECK-NEXT:    vmov.u8 r2, q0[15]
-; CHECK-NEXT:    vmov.8 q4[13], r0
 ; CHECK-NEXT:    vmov.u8 r0, q2[11]
-; CHECK-NEXT:    vmov.8 q3[10], r2
-; CHECK-NEXT:    vmov.8 q4[14], r0
+; CHECK-NEXT:    vmov.8 q5[9], r2
+; CHECK-NEXT:    vmov.u8 r2, q0[15]
+; CHECK-NEXT:    vmov.8 q6[14], r0
 ; CHECK-NEXT:    vmov.u8 r0, q2[14]
-; CHECK-NEXT:    vmov.8 q4[15], r0
+; CHECK-NEXT:    vmov.8 q5[10], r2
+; CHECK-NEXT:    vmov.8 q6[15], r0
 ; CHECK-NEXT:    vmov.u8 r0, q2[2]
-; CHECK-NEXT:    vmov q5, q3
 ; CHECK-NEXT:    vmov.8 q5[11], r0
 ; CHECK-NEXT:    vmov.u8 r0, q1[0]
-; CHECK-NEXT:    vmov.f32 s14, s22
-; CHECK-NEXT:    vmov.f32 s15, s19
+; CHECK-NEXT:    vmov.f32 s13, s17
 ; CHECK-NEXT:    vmov.8 q4[0], r0
 ; CHECK-NEXT:    vmov.u8 r0, q1[3]
+; CHECK-NEXT:    vmov.f32 s14, s22
 ; CHECK-NEXT:    vmov.8 q4[1], r0
 ; CHECK-NEXT:    vmov.u8 r0, q1[6]
 ; CHECK-NEXT:    vmov.8 q4[2], r0
 ; CHECK-NEXT:    vmov.u8 r0, q1[9]
 ; CHECK-NEXT:    vmov.8 q4[3], r0
 ; CHECK-NEXT:    vmov.u8 r0, q1[12]
-; CHECK-NEXT:    vmov.8 q4[4], r0
+; CHECK-NEXT:    vmov q5, q4
+; CHECK-NEXT:    vmov.f32 s15, s27
+; CHECK-NEXT:    vmov.8 q5[4], r0
 ; CHECK-NEXT:    vmov.u8 r0, q1[15]
-; CHECK-NEXT:    vmov.8 q4[5], r0
+; CHECK-NEXT:    vmov.8 q5[5], r0
 ; CHECK-NEXT:    vmov.u8 r0, q0[2]
-; CHECK-NEXT:    vmov.8 q4[6], r0
+; CHECK-NEXT:    vmov.8 q5[6], r0
 ; CHECK-NEXT:    vmov.u8 r0, q0[5]
-; CHECK-NEXT:    vmov.8 q4[7], r0
+; CHECK-NEXT:    vmov.8 q5[7], r0
 ; CHECK-NEXT:    vmov.u8 r0, q0[8]
-; CHECK-NEXT:    vmov.8 q4[8], r0
+; CHECK-NEXT:    vmov q6, q5
+; CHECK-NEXT:    vmov.f32 s17, s21
+; CHECK-NEXT:    vmov.8 q6[8], r0
 ; CHECK-NEXT:    vmov.u8 r0, q0[11]
-; CHECK-NEXT:    vmov.8 q4[9], r0
+; CHECK-NEXT:    vmov.8 q6[9], r0
 ; CHECK-NEXT:    vmov.u8 r0, q0[14]
-; CHECK-NEXT:    vmov.8 q4[10], r0
+; CHECK-NEXT:    vmov.8 q6[10], r0
 ; CHECK-NEXT:    vmov.u8 r0, q2[4]
-; CHECK-NEXT:    vmov.8 q5[12], r0
+; CHECK-NEXT:    vmov.8 q7[12], r0
 ; CHECK-NEXT:    vmov.u8 r0, q2[7]
-; CHECK-NEXT:    vmov.8 q5[13], r0
+; CHECK-NEXT:    vmov.8 q7[13], r0
 ; CHECK-NEXT:    vmov.u8 r0, q2[10]
-; CHECK-NEXT:    vmov.8 q5[14], r0
+; CHECK-NEXT:    vmov.8 q7[14], r0
 ; CHECK-NEXT:    vmov.u8 r0, q2[13]
-; CHECK-NEXT:    vmov.8 q5[15], r0
+; CHECK-NEXT:    vmov.8 q7[15], r0
 ; CHECK-NEXT:    vmov.u8 r0, q2[1]
-; CHECK-NEXT:    vmov q6, q4
 ; CHECK-NEXT:    vmov.8 q6[11], r0
 ; CHECK-NEXT:    vmov.u8 r0, q1[2]
 ; CHECK-NEXT:    vmov.f32 s18, s26
-; CHECK-NEXT:    vmov.f32 s19, s23
+; CHECK-NEXT:    vmov.f32 s19, s31
 ; CHECK-NEXT:    vadd.i8 q3, q4, q3
 ; CHECK-NEXT:    vmov.8 q4[0], r0
 ; CHECK-NEXT:    vmov.u8 r0, q1[5]
 ; CHECK-NEXT:    vmov.8 q4[1], r0
 ; CHECK-NEXT:    vmov.u8 r0, q1[8]
 ; CHECK-NEXT:    vmov.8 q4[2], r0
 ; CHECK-NEXT:    vmov.u8 r0, q1[11]
 ; CHECK-NEXT:    vmov.8 q4[3], r0
 ; CHECK-NEXT:    vmov.u8 r0, q1[14]
-; CHECK-NEXT:    vmov.8 q4[4], r0
+; CHECK-NEXT:    vmov q1, q4
+; CHECK-NEXT:    vmov.8 q1[4], r0
 ; CHECK-NEXT:    vmov.u8 r0, q0[1]
-; CHECK-NEXT:    vmov.8 q4[5], r0
+; CHECK-NEXT:    vmov.8 q1[5], r0
 ; CHECK-NEXT:    vmov.u8 r0, q0[4]
-; CHECK-NEXT:    vmov.8 q4[6], r0
+; CHECK-NEXT:    vmov.8 q1[6], r0
 ; CHECK-NEXT:    vmov.u8 r0, q2[6]
-; CHECK-NEXT:    vmov.8 q1[12], r0
+; CHECK-NEXT:    vmov.8 q5[12], r0
 ; CHECK-NEXT:    vmov.u8 r0, q2[9]
-; CHECK-NEXT:    vmov.8 q1[13], r0
+; CHECK-NEXT:    vmov.8 q5[13], r0
 ; CHECK-NEXT:    vmov.u8 r0, q2[12]
-; CHECK-NEXT:    vmov.8 q1[14], r0
+; CHECK-NEXT:    vmov.8 q5[14], r0
 ; CHECK-NEXT:    vmov.u8 r0, q2[15]
-; CHECK-NEXT:    vmov.8 q1[15], r0
+; CHECK-NEXT:    vmov.8 q5[15], r0
 ; CHECK-NEXT:    vmov.u8 r0, q0[10]
-; CHECK-NEXT:    vmov.8 q5[8], r0
+; CHECK-NEXT:    vmov.8 q6[8], r0
 ; CHECK-NEXT:    vmov.u8 r0, q0[13]
-; CHECK-NEXT:    vmov.8 q5[9], r0
+; CHECK-NEXT:    vmov.8 q6[9], r0
 ; CHECK-NEXT:    vmov.u8 r0, q2[0]
-; CHECK-NEXT:    vmov.8 q5[10], r0
+; CHECK-NEXT:    vmov.8 q6[10], r0
 ; CHECK-NEXT:    vmov.u8 r0, q2[3]
-; CHECK-NEXT:    vmov.8 q5[11], r0
+; CHECK-NEXT:    vmov.8 q6[11], r0
 ; CHECK-NEXT:    vmov.u8 r0, q0[7]
-; CHECK-NEXT:    vmov.8 q4[7], r0
-; CHECK-NEXT:    vmov.f32 s18, s22
-; CHECK-NEXT:    vmov.f32 s19, s7
+; CHECK-NEXT:    vmov.8 q1[7], r0
+; CHECK-NEXT:    vmov.f32 s17, s5
+; CHECK-NEXT:    vmov.f32 s18, s26
+; CHECK-NEXT:    vmov.f32 s19, s23
 ; CHECK-NEXT:    vadd.i8 q0, q3, q4
 ; CHECK-NEXT:    vstrw.32 q0, [r1]
-; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13}
+; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13, d14, d15}
 ; CHECK-NEXT:    bx lr
 entry:
   %l1 = load <48 x i8>, <48 x i8>* %src, align 4
   %s1 = shufflevector <48 x i8> %l1, <48 x i8> undef, <16 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21, i32 24, i32 27, i32 30, i32 33, i32 36, i32 39, i32 42, i32 45>
   %s2 = shufflevector <48 x i8> %l1, <48 x i8> undef, <16 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22, i32 25, i32 28, i32 31, i32 34, i32 37, i32 40, i32 43, i32 46>
   %s3 = shufflevector <48 x i8> %l1, <48 x i8> undef, <16 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23, i32 26, i32 29, i32 32, i32 35, i32 38, i32 41, i32 44, i32 47>
   %a1 = add <16 x i8> %s1, %s2
   %a = add <16 x i8> %a1, %s3
   store <16 x i8> %a, <16 x i8> *%dst
   ret void
 }
 
 ; i64
 
 define void @vld3_v2i64(<6 x i64> *%src, <2 x i64> *%dst) {
 ; CHECK-LABEL: vld3_v2i64:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .save {r4, lr}
 ; CHECK-NEXT:    push {r4, lr}
-; CHECK-NEXT:    .vsave {d8, d9}
-; CHECK-NEXT:    vpush {d8, d9}
-; CHECK-NEXT:    vldrw.u32 q1, [r0]
-; CHECK-NEXT:    vldrw.u32 q2, [r0, #16]
-; CHECK-NEXT:    vldrw.u32 q4, [r0, #32]
-; CHECK-NEXT:    vmov.f64 d6, d3
-; CHECK-NEXT:    vmov.f32 s13, s7
-; CHECK-NEXT:    vmov.f32 s14, s16
-; CHECK-NEXT:    vmov.f32 s6, s10
-; CHECK-NEXT:    vmov.f32 s7, s11
-; CHECK-NEXT:    vmov.f32 s15, s17
-; CHECK-NEXT:    vmov r3, s14
+; CHECK-NEXT:    vldrw.u32 q0, [r0]
+; CHECK-NEXT:    vldrw.u32 q1, [r0, #32]
+; CHECK-NEXT:    vldrw.u32 q3, [r0, #16]
+; CHECK-NEXT:    vmov.f64 d4, d1
+; CHECK-NEXT:    vmov.f32 s9, s3
+; CHECK-NEXT:    vmov.f32 s10, s4
+; CHECK-NEXT:    vmov.f32 s11, s5
+; CHECK-NEXT:    vmov r3, s10
+; CHECK-NEXT:    vmov r2, s14
+; CHECK-NEXT:    vmov r0, s15
+; CHECK-NEXT:    vmov r4, s0
+; CHECK-NEXT:    vmov r12, s11
+; CHECK-NEXT:    adds r2, r2, r3
+; CHECK-NEXT:    vmov r3, s7
+; CHECK-NEXT:    adc.w r12, r12, r0
 ; CHECK-NEXT:    vmov r0, s6
-; CHECK-NEXT:    vmov.f64 d0, d4
-; CHECK-NEXT:    vmov.f32 s1, s9
-; CHECK-NEXT:    vmov.f32 s2, s18
-; CHECK-NEXT:    vmov.f32 s3, s19
-; CHECK-NEXT:    vmov r12, s15
-; CHECK-NEXT:    vmov r2, s7
-; CHECK-NEXT:    vmov r4, s4
-; CHECK-NEXT:    adds.w lr, r0, r3
-; CHECK-NEXT:    vmov r0, s2
-; CHECK-NEXT:    vmov r3, s3
-; CHECK-NEXT:    adc.w r2, r2, r12
-; CHECK-NEXT:    adds.w lr, lr, r0
-; CHECK-NEXT:    vmov r0, s12
-; CHECK-NEXT:    adc.w r12, r2, r3
+; CHECK-NEXT:    adds.w lr, r2, r0
+; CHECK-NEXT:    vmov r2, s8
+; CHECK-NEXT:    adc.w r12, r12, r3
+; CHECK-NEXT:    vmov r3, s9
+; CHECK-NEXT:    vmov r0, s1
+; CHECK-NEXT:    adds r2, r2, r4
+; CHECK-NEXT:    vmov r4, s12
+; CHECK-NEXT:    adcs r0, r3
 ; CHECK-NEXT:    vmov r3, s13
-; CHECK-NEXT:    vmov r2, s5
-; CHECK-NEXT:    adds r0, r0, r4
-; CHECK-NEXT:    vmov r4, s0
-; CHECK-NEXT:    adcs r2, r3
-; CHECK-NEXT:    vmov r3, s1
-; CHECK-NEXT:    adds r0, r0, r4
-; CHECK-NEXT:    vmov q0[2], q0[0], r0, lr
-; CHECK-NEXT:    adcs r2, r3
-; CHECK-NEXT:    vmov q0[3], q0[1], r2, r12
+; CHECK-NEXT:    adds r2, r2, r4
+; CHECK-NEXT:    vmov q0[2], q0[0], r2, lr
+; CHECK-NEXT:    adcs r0, r3
+; CHECK-NEXT:    vmov q0[3], q0[1], r0, r12
 ; CHECK-NEXT:    vstrw.32 q0, [r1]
-; CHECK-NEXT:    vpop {d8, d9}
 ; CHECK-NEXT:    pop {r4, pc}
 entry:
   %l1 = load <6 x i64>, <6 x i64>* %src, align 4
   %s1 = shufflevector <6 x i64> %l1, <6 x i64> undef, <2 x i32> <i32 0, i32 3>
   %s2 = shufflevector <6 x i64> %l1, <6 x i64> undef, <2 x i32> <i32 1, i32 4>
   %s3 = shufflevector <6 x i64> %l1, <6 x i64> undef, <2 x i32> <i32 2, i32 5>
   %a1 = add <2 x i64> %s1, %s2
   %a = add <2 x i64> %a1, %s3
   store <2 x i64> %a, <2 x i64> *%dst
   ret void
 }
 
 define void @vld3_v4i64(<12 x i64> *%src, <4 x i64> *%dst) {
 ; CHECK-LABEL: vld3_v4i64:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .save {r4, r5, r6, r7, r8, lr}
 ; CHECK-NEXT:    push.w {r4, r5, r6, r7, r8, lr}
-; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13, d14, d15}
-; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13, d14, d15}
-; CHECK-NEXT:    .pad #24
-; CHECK-NEXT:    sub sp, #24
-; CHECK-NEXT:    vldrw.u32 q2, [r0, #64]
-; CHECK-NEXT:    vldrw.u32 q1, [r0, #80]
-; CHECK-NEXT:    vldrw.u32 q5, [r0]
+; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13}
+; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13}
+; CHECK-NEXT:    vldrw.u32 q4, [r0]
+; CHECK-NEXT:    vldrw.u32 q5, [r0, #32]
 ; CHECK-NEXT:    vldrw.u32 q3, [r0, #16]
-; CHECK-NEXT:    vmov.f64 d0, d4
-; CHECK-NEXT:    vstrw.32 q2, [sp] @ 16-byte Spill
-; CHECK-NEXT:    vmov.f32 s1, s9
-; CHECK-NEXT:    vldrw.u32 q2, [r0, #48]
-; CHECK-NEXT:    vmov.f32 s2, s6
-; CHECK-NEXT:    vmov.f64 d8, d5
-; CHECK-NEXT:    vmov.f32 s17, s11
-; CHECK-NEXT:    vmov.f32 s18, s4
-; CHECK-NEXT:    vmov.f32 s19, s5
-; CHECK-NEXT:    vmov.f64 d12, d11
-; CHECK-NEXT:    vmov.f32 s3, s7
-; CHECK-NEXT:    vldrw.u32 q1, [r0, #32]
-; CHECK-NEXT:    vmov.f32 s25, s23
-; CHECK-NEXT:    vmov.f32 s26, s4
-; CHECK-NEXT:    vmov.f32 s22, s14
-; CHECK-NEXT:    vmov.f32 s27, s5
-; CHECK-NEXT:    vmov.f32 s23, s15
-; CHECK-NEXT:    vmov r3, s26
-; CHECK-NEXT:    vmov r0, s22
-; CHECK-NEXT:    vmov.f64 d14, d6
+; CHECK-NEXT:    vldrw.u32 q0, [r0, #48]
+; CHECK-NEXT:    vmov.f64 d12, d9
+; CHECK-NEXT:    vldrw.u32 q1, [r0, #80]
+; CHECK-NEXT:    vmov.f32 s25, s19
+; CHECK-NEXT:    vmov.f32 s26, s20
+; CHECK-NEXT:    vmov.f32 s27, s21
+; CHECK-NEXT:    vmov lr, s26
+; CHECK-NEXT:    vmov r3, s14
+; CHECK-NEXT:    vmov r2, s15
+; CHECK-NEXT:    vmov r5, s16
+; CHECK-NEXT:    vmov r4, s17
+; CHECK-NEXT:    vldrw.u32 q4, [r0, #64]
+; CHECK-NEXT:    vmov.f64 d4, d1
 ; CHECK-NEXT:    vmov r12, s27
+; CHECK-NEXT:    vmov.f32 s9, s3
+; CHECK-NEXT:    vmov.f32 s10, s4
+; CHECK-NEXT:    vmov.f32 s11, s5
+; CHECK-NEXT:    vmov r6, s18
+; CHECK-NEXT:    vmov r0, s19
+; CHECK-NEXT:    vmov r7, s1
+; CHECK-NEXT:    adds.w lr, lr, r3
+; CHECK-NEXT:    vmov r3, s22
+; CHECK-NEXT:    adc.w r12, r12, r2
 ; CHECK-NEXT:    vmov r2, s23
-; CHECK-NEXT:    vmov.f32 s29, s13
-; CHECK-NEXT:    vmov.f32 s30, s6
-; CHECK-NEXT:    vmov.f32 s31, s7
-; CHECK-NEXT:    vldrw.u32 q1, [sp] @ 16-byte Reload
-; CHECK-NEXT:    vmov r5, s16
-; CHECK-NEXT:    vmov.f32 s10, s6
-; CHECK-NEXT:    vmov.f32 s11, s7
-; CHECK-NEXT:    vmov r4, s10
-; CHECK-NEXT:    vmov r6, s8
-; CHECK-NEXT:    vmov r7, s24
-; CHECK-NEXT:    adds.w lr, r0, r3
-; CHECK-NEXT:    vmov r0, s30
-; CHECK-NEXT:    adc.w r3, r2, r12
-; CHECK-NEXT:    vmov r2, s31
-; CHECK-NEXT:    adds.w lr, lr, r0
-; CHECK-NEXT:    vmov r0, s18
-; CHECK-NEXT:    adc.w r12, r3, r2
-; CHECK-NEXT:    vmov r3, s19
-; CHECK-NEXT:    vmov r2, s11
-; CHECK-NEXT:    adds r0, r0, r4
-; CHECK-NEXT:    vmov r4, s2
-; CHECK-NEXT:    adcs r2, r3
-; CHECK-NEXT:    vmov r3, s3
-; CHECK-NEXT:    adds r0, r0, r4
-; CHECK-NEXT:    vmov r4, s9
-; CHECK-NEXT:    adc.w r8, r2, r3
-; CHECK-NEXT:    vmov r3, s17
+; CHECK-NEXT:    adds.w lr, lr, r3
+; CHECK-NEXT:    vmov r3, s24
+; CHECK-NEXT:    adc.w r12, r12, r2
+; CHECK-NEXT:    vmov r2, s25
+; CHECK-NEXT:    adds r3, r3, r5
+; CHECK-NEXT:    vmov r5, s10
+; CHECK-NEXT:    adc.w r8, r4, r2
+; CHECK-NEXT:    vmov r4, s11
+; CHECK-NEXT:    vmov r2, s0
 ; CHECK-NEXT:    adds r5, r5, r6
-; CHECK-NEXT:    vmov r6, s0
-; CHECK-NEXT:    vmov r2, s20
-; CHECK-NEXT:    adcs r3, r4
-; CHECK-NEXT:    vmov r4, s1
+; CHECK-NEXT:    vmov r6, s6
+; CHECK-NEXT:    adcs r0, r4
+; CHECK-NEXT:    vmov r4, s7
 ; CHECK-NEXT:    adds r5, r5, r6
-; CHECK-NEXT:    vmov r6, s21
-; CHECK-NEXT:    vmov q1[2], q1[0], r5, r0
-; CHECK-NEXT:    vmov r0, s29
-; CHECK-NEXT:    adcs r3, r4
-; CHECK-NEXT:    vmov r4, s25
+; CHECK-NEXT:    vmov r6, s8
+; CHECK-NEXT:    adcs r0, r4
+; CHECK-NEXT:    vmov r4, s9
+; CHECK-NEXT:    adds r2, r2, r6
+; CHECK-NEXT:    vmov r6, s17
+; CHECK-NEXT:    adcs r4, r7
+; CHECK-NEXT:    vmov r7, s16
 ; CHECK-NEXT:    adds r2, r2, r7
-; CHECK-NEXT:    vmov q1[3], q1[1], r3, r8
-; CHECK-NEXT:    vstrw.32 q1, [r1, #16]
 ; CHECK-NEXT:    adcs r4, r6
-; CHECK-NEXT:    vmov r6, s28
-; CHECK-NEXT:    adds r2, r2, r6
-; CHECK-NEXT:    vmov q0[2], q0[0], r2, lr
-; CHECK-NEXT:    adcs r0, r4
+; CHECK-NEXT:    vmov q1[2], q1[0], r2, r5
+; CHECK-NEXT:    vmov r6, s12
+; CHECK-NEXT:    vmov q1[3], q1[1], r4, r0
+; CHECK-NEXT:    vmov r0, s13
+; CHECK-NEXT:    vstrw.32 q1, [r1, #16]
+; CHECK-NEXT:    adds r3, r3, r6
+; CHECK-NEXT:    vmov q0[2], q0[0], r3, lr
+; CHECK-NEXT:    adc.w r0, r0, r8
 ; CHECK-NEXT:    vmov q0[3], q0[1], r0, r12
 ; CHECK-NEXT:    vstrw.32 q0, [r1]
-; CHECK-NEXT:    add sp, #24
-; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13, d14, d15}
+; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13}
 ; CHECK-NEXT:    pop.w {r4, r5, r6, r7, r8, pc}
 entry:
   %l1 = load <12 x i64>, <12 x i64>* %src, align 4
   %s1 = shufflevector <12 x i64> %l1, <12 x i64> undef, <4 x i32> <i32 0, i32 3, i32 6, i32 9>
   %s2 = shufflevector <12 x i64> %l1, <12 x i64> undef, <4 x i32> <i32 1, i32 4, i32 7, i32 10>
   %s3 = shufflevector <12 x i64> %l1, <12 x i64> undef, <4 x i32> <i32 2, i32 5, i32 8, i32 11>
   %a1 = add <4 x i64> %s1, %s2
   %a = add <4 x i64> %a1, %s3
entry:
   ret void
 }
 
 ; f16
 
 define void @vld3_v2f16(<6 x half> *%src, <2 x half> *%dst) {
 ; CHECK-LABEL: vld3_v2f16:
 ; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    .vsave {d8, d9}
+; CHECK-NEXT:    vpush {d8, d9}
 ; CHECK-NEXT:    ldrd r2, r3, [r0]
 ; CHECK-NEXT:    ldr r0, [r0, #8]
-; CHECK-NEXT:    vmov.32 q0[0], r2
+; CHECK-NEXT:    vmov.32 q1[0], r2
+; CHECK-NEXT:    vmov q0, q1
+; CHECK-NEXT:    vmovx.f16 s16, s4
 ; CHECK-NEXT:    vmov.32 q0[1], r3
-; CHECK-NEXT:    vmov.32 q0[2], r0
-; CHECK-NEXT:    vmovx.f16 s8, s0
-; CHECK-NEXT:    vmovx.f16 s4, s2
-; CHECK-NEXT:    vins.f16 s8, s2
-; CHECK-NEXT:    vmovx.f16 s6, s1
-; CHECK-NEXT:    vins.f16 s1, s4
-; CHECK-NEXT:    vins.f16 s0, s6
-; CHECK-NEXT:    vadd.f16 q1, q0, q2
+; CHECK-NEXT:    vmov q2, q0
+; CHECK-NEXT:    vmov.32 q2[2], r0
+; CHECK-NEXT:    vins.f16 s16, s10
+; CHECK-NEXT:    vmovx.f16 s8, s1
+; CHECK-NEXT:    vmovx.f16 s12, s10
+; CHECK-NEXT:    vins.f16 s4, s8
+; CHECK-NEXT:    vins.f16 s1, s12
+; CHECK-NEXT:    vadd.f16 q1, q1, q4
 ; CHECK-NEXT:    vmov.f32 s0, s1
 ; CHECK-NEXT:    vadd.f16 q0, q1, q0
 ; CHECK-NEXT:    vmov r0, s0
 ; CHECK-NEXT:    str r0, [r1]
+; CHECK-NEXT:    vpop {d8, d9}
 ; CHECK-NEXT:    bx lr
 entry:
   %l1 = load <6 x half>, <6 x half>* %src, align 4
   %s1 = shufflevector <6 x half> %l1, <6 x half> undef, <2 x i32> <i32 0, i32 3>
   %s2 = shufflevector <6 x half> %l1, <6 x half> undef, <2 x i32> <i32 1, i32 4>
   %s3 = shufflevector <6 x half> %l1, <6 x half> undef, <2 x i32> <i32 2, i32 5>
   %a1 = fadd <2 x half> %s1, %s2
   %a = fadd <2 x half> %a1, %s3
   store <2 x half> %a, <2 x half> *%dst
   ret void
 }
 
 define void @vld3_v4f16(<12 x half> *%src, <4 x half> *%dst) {
 ; CHECK-LABEL: vld3_v4f16:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .vsave {d8, d9}
-; CHECK-NEXT:    vpush {d8, d9}
+; CHECK-NEXT:    .vsave {d8, d9, d10, d11}
+; CHECK-NEXT:    vpush {d8, d9, d10, d11}
 ; CHECK-NEXT:    ldrd r2, r3, [r0, #16]
 ; CHECK-NEXT:    vmov.32 q2[0], r2
-; CHECK-NEXT:    vmov.32 q2[1], r3
+; CHECK-NEXT:    vmov q3, q2
 ; CHECK-NEXT:    vmov.f32 s1, s8
-; CHECK-NEXT:    vmovx.f16 s4, s9
+; CHECK-NEXT:    vmov.32 q3[1], r3
+; CHECK-NEXT:    vmovx.f16 s8, s8
+; CHECK-NEXT:    vmovx.f16 s4, s13
 ; CHECK-NEXT:    vins.f16 s1, s4
 ; CHECK-NEXT:    vldrw.u32 q1, [r0]
-; CHECK-NEXT:    vmovx.f16 s8, s8
-; CHECK-NEXT:    vmovx.f16 s12, s4
-; CHECK-NEXT:    vmovx.f16 s16, s6
-; CHECK-NEXT:    vins.f16 s12, s6
-; CHECK-NEXT:    vmovx.f16 s18, s5
-; CHECK-NEXT:    vins.f16 s5, s16
-; CHECK-NEXT:    vins.f16 s4, s18
-; CHECK-NEXT:    vmovx.f16 s13, s7
+; CHECK-NEXT:    vmovx.f16 s16, s4
+; CHECK-NEXT:    vmovx.f16 s20, s6
+; CHECK-NEXT:    vins.f16 s16, s6
+; CHECK-NEXT:    vmovx.f16 s22, s5
+; CHECK-NEXT:    vins.f16 s5, s20
+; CHECK-NEXT:    vins.f16 s4, s22
+; CHECK-NEXT:    vmovx.f16 s17, s7
 ; CHECK-NEXT:    vins.f16 s7, s8
 ; CHECK-NEXT:    vmov.f32 s0, s5
-; CHECK-NEXT:    vins.f16 s13, s9
+; CHECK-NEXT:    vins.f16 s17, s13
 ; CHECK-NEXT:    vmov.f32 s5, s7
-; CHECK-NEXT:    vadd.f16 q1, q1, q3
+; CHECK-NEXT:    vadd.f16 q1, q1, q4
 ; CHECK-NEXT:    vadd.f16 q0, q1, q0
 ; CHECK-NEXT:    vmov r2, s1
 ; CHECK-NEXT:    vmov r0, s0
 ; CHECK-NEXT:    strd r0, r2, [r1]
-; CHECK-NEXT:    vpop {d8, d9}
+; CHECK-NEXT:    vpop {d8, d9, d10, d11}
 ; CHECK-NEXT:    bx lr
 entry:
   %l1 = load <12 x half>, <12 x half>* %src, align 4
   %s1 = shufflevector <12 x half> %l1, <12 x half> undef, <4 x i32> <i32 0, i32 3, i32 6, i32 9>
   %s2 = shufflevector <12 x half> %l1, <12 x half> undef, <4 x i32> <i32 1, i32 4, i32 7, i32 10>
   %s3 = shufflevector <12 x half> %l1, <12 x half> undef, <4 x i32> <i32 2, i32 5, i32 8, i32 11>
   %a1 = fadd <4 x half> %s1, %s2
   %a = fadd <4 x half> %a1, %s3

llvm/test/CodeGen/Thumb2/mve-vld4-post.ll

 
 define <8 x i64> *@vld4_v2i64(<8 x i64> *%src, <2 x i64> *%dst) {
 ; CHECK-LABEL: vld4_v2i64:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .save {r4, r5, r6, r7, lr}
 ; CHECK-NEXT:    push {r4, r5, r6, r7, lr}
 ; CHECK-NEXT:    .pad #4
 ; CHECK-NEXT:    sub sp, #4
-; CHECK-NEXT:    .vsave {d8, d9, d10, d11}
-; CHECK-NEXT:    vpush {d8, d9, d10, d11}
-; CHECK-NEXT:    vldrw.u32 q2, [r0, #32]
+; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13}
+; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13}
+; CHECK-NEXT:    vldrw.u32 q1, [r0, #32]
 ; CHECK-NEXT:    vldrw.u32 q0, [r0], #64
-; CHECK-NEXT:    vldrw.u32 q3, [r0, #-48]
+; CHECK-NEXT:    vldrw.u32 q2, [r0, #-48]
 ; CHECK-NEXT:    vldrw.u32 q5, [r0, #-16]
-; CHECK-NEXT:    vmov.f64 d2, d1
-; CHECK-NEXT:    vmov.f64 d8, d7
-; CHECK-NEXT:    vmov.f32 s17, s15
-; CHECK-NEXT:    vmov.f32 s18, s22
-; CHECK-NEXT:    vmov.f32 s14, s20
-; CHECK-NEXT:    vmov.f32 s19, s23
-; CHECK-NEXT:    vmov.f32 s15, s21
-; CHECK-NEXT:    vmov r2, s18
-; CHECK-NEXT:    vmov r3, s14
-; CHECK-NEXT:    vmov.f32 s5, s3
-; CHECK-NEXT:    vmov.f32 s6, s10
-; CHECK-NEXT:    vmov.f32 s2, s8
-; CHECK-NEXT:    vmov.f32 s3, s9
-; CHECK-NEXT:    vmov.f32 s7, s11
+; CHECK-NEXT:    vmov.f64 d6, d0
+; CHECK-NEXT:    vmov.f64 d8, d4
+; CHECK-NEXT:    vmov.f32 s17, s9
+; CHECK-NEXT:    vmov.f32 s18, s20
+; CHECK-NEXT:    vmov.f32 s19, s21
+; CHECK-NEXT:    vmov lr, s18
+; CHECK-NEXT:    vmov r3, s22
+; CHECK-NEXT:    vmov r2, s23
+; CHECK-NEXT:    vmov.f32 s13, s1
+; CHECK-NEXT:    vmov.f32 s14, s4
+; CHECK-NEXT:    vmov.f32 s15, s5
 ; CHECK-NEXT:    vmov r12, s19
-; CHECK-NEXT:    vmov lr, s15
-; CHECK-NEXT:    vmov r4, s6
-; CHECK-NEXT:    vmov r5, s2
+; CHECK-NEXT:    vmov r5, s6
+; CHECK-NEXT:    vmov r4, s7
+; CHECK-NEXT:    vmov.f64 d12, d5
+; CHECK-NEXT:    vmov.f32 s25, s11
+; CHECK-NEXT:    vmov.f64 d8, d1
+; CHECK-NEXT:    vmov.f32 s17, s3
 ; CHECK-NEXT:    vmov r7, s0
-; CHECK-NEXT:    adds r6, r3, r2
-; CHECK-NEXT:    vmov r2, s7
-; CHECK-NEXT:    vmov r3, s3
-; CHECK-NEXT:    adc.w r12, r12, lr
-; CHECK-NEXT:    adds r5, r5, r4
-; CHECK-NEXT:    vmov r4, s16
-; CHECK-NEXT:    adcs r2, r3
-; CHECK-NEXT:    adds.w lr, r5, r6
+; CHECK-NEXT:    adds.w lr, lr, r3
+; CHECK-NEXT:    vmov r3, s14
+; CHECK-NEXT:    vmov r6, s1
+; CHECK-NEXT:    adc.w r12, r12, r2
+; CHECK-NEXT:    vmov r2, s15
+; CHECK-NEXT:    adds r3, r3, r5
+; CHECK-NEXT:    vmov r5, s25
+; CHECK-NEXT:    adcs r2, r4
+; CHECK-NEXT:    adds.w lr, lr, r3
+; CHECK-NEXT:    vmov r4, s24
 ; CHECK-NEXT:    adc.w r12, r12, r2
-; CHECK-NEXT:    vmov r2, s12
-; CHECK-NEXT:    vmov r6, s17
-; CHECK-NEXT:    vmov r5, s13
-; CHECK-NEXT:    vmov r3, s4
-; CHECK-NEXT:    adds r2, r2, r4
-; CHECK-NEXT:    vmov r4, s1
-; CHECK-NEXT:    adcs r6, r5
-; CHECK-NEXT:    vmov r5, s5
-; CHECK-NEXT:    adds r3, r3, r7
-; CHECK-NEXT:    adcs r4, r5
-; CHECK-NEXT:    adds r2, r2, r3
-; CHECK-NEXT:    adc.w r3, r4, r6
-; CHECK-NEXT:    vmov q0[2], q0[0], r2, lr
-; CHECK-NEXT:    vmov q0[3], q0[1], r3, r12
+; CHECK-NEXT:    vmov r3, s8
+; CHECK-NEXT:    vmov r2, s9
+; CHECK-NEXT:    adds r3, r3, r4
+; CHECK-NEXT:    vmov r4, s17
+; CHECK-NEXT:    adcs r2, r5
+; CHECK-NEXT:    vmov r5, s16
+; CHECK-NEXT:    adds r5, r5, r7
+; CHECK-NEXT:    adcs r4, r6
+; CHECK-NEXT:    adds r3, r3, r5
+; CHECK-NEXT:    adcs r2, r4
+; CHECK-NEXT:    vmov q0[2], q0[0], r3, lr
+; CHECK-NEXT:    vmov q0[3], q0[1], r2, r12
 ; CHECK-NEXT:    vstrw.32 q0, [r1]
-; CHECK-NEXT:    vpop {d8, d9, d10, d11}
+; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13}
 ; CHECK-NEXT:    add sp, #4
 ; CHECK-NEXT:    pop {r4, r5, r6, r7, pc}
 entry:
   %l1 = load <8 x i64>, <8 x i64>* %src, align 4
   %s1 = shufflevector <8 x i64> %l1, <8 x i64> undef, <2 x i32> <i32 0, i32 4>
   %s2 = shufflevector <8 x i64> %l1, <8 x i64> undef, <2 x i32> <i32 1, i32 5>
   %s3 = shufflevector <8 x i64> %l1, <8 x i64> undef, <2 x i32> <i32 2, i32 6>
   %s4 = shufflevector <8 x i64> %l1, <8 x i64> undef, <2 x i32> <i32 3, i32 7>

llvm/test/CodeGen/Thumb2/mve-vld4.ll

 
 ; i64
 
 define void @vld4_v2i64(<8 x i64> *%src, <2 x i64> *%dst) {
 ; CHECK-LABEL: vld4_v2i64:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .save {r4, r5, r6, lr}
 ; CHECK-NEXT:    push {r4, r5, r6, lr}
-; CHECK-NEXT:    .vsave {d8, d9, d10, d11}
-; CHECK-NEXT:    vpush {d8, d9, d10, d11}
-; CHECK-NEXT:    vldrw.u32 q3, [r0, #16]
+; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13}
+; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13}
+; CHECK-NEXT:    vldrw.u32 q2, [r0, #16]
 ; CHECK-NEXT:    vldrw.u32 q5, [r0, #48]
 ; CHECK-NEXT:    vldrw.u32 q0, [r0]
-; CHECK-NEXT:    vldrw.u32 q2, [r0, #32]
-; CHECK-NEXT:    vmov.f64 d8, d7
-; CHECK-NEXT:    vmov.f32 s17, s15
-; CHECK-NEXT:    vmov.f32 s18, s22
-; CHECK-NEXT:    vmov.f32 s14, s20
-; CHECK-NEXT:    vmov.f32 s19, s23
-; CHECK-NEXT:    vmov.f32 s15, s21
-; CHECK-NEXT:    vmov r3, s18
-; CHECK-NEXT:    vmov r0, s14
-; CHECK-NEXT:    vmov.f64 d2, d1
+; CHECK-NEXT:    vldrw.u32 q1, [r0, #32]
+; CHECK-NEXT:    vmov.f64 d8, d4
+; CHECK-NEXT:    vmov.f32 s17, s9
+; CHECK-NEXT:    vmov.f32 s18, s20
+; CHECK-NEXT:    vmov.f32 s19, s21
+; CHECK-NEXT:    vmov r2, s18
+; CHECK-NEXT:    vmov r0, s22
+; CHECK-NEXT:    vmov r3, s23
+; CHECK-NEXT:    vmov.f64 d6, d0
 ; CHECK-NEXT:    vmov r12, s19
-; CHECK-NEXT:    vmov r2, s15
-; CHECK-NEXT:    vmov.f32 s5, s3
-; CHECK-NEXT:    vmov.f32 s6, s10
-; CHECK-NEXT:    vmov.f32 s2, s8
-; CHECK-NEXT:    vmov.f32 s3, s9
-; CHECK-NEXT:    vmov.f32 s7, s11
-; CHECK-NEXT:    vmov r4, s2
-; CHECK-NEXT:    vmov r5, s4
+; CHECK-NEXT:    vmov.f32 s13, s1
+; CHECK-NEXT:    vmov.f32 s14, s4
+; CHECK-NEXT:    vmov.f32 s15, s5
+; CHECK-NEXT:    vmov r4, s6
+; CHECK-NEXT:    vmov.f64 d12, d5
+; CHECK-NEXT:    vmov.f32 s25, s11
+; CHECK-NEXT:    vmov.f64 d8, d1
+; CHECK-NEXT:    adds.w lr, r2, r0
+; CHECK-NEXT:    vmov r0, s14
+; CHECK-NEXT:    vmov r2, s7
+; CHECK-NEXT:    vmov.f32 s17, s3
 ; CHECK-NEXT:    vmov r6, s0
-; CHECK-NEXT:    adds.w lr, r0, r3
-; CHECK-NEXT:    vmov r3, s7
-; CHECK-NEXT:    vmov r0, s3
+; CHECK-NEXT:    vmov r5, s1
+; CHECK-NEXT:    adc.w r12, r12, r3
+; CHECK-NEXT:    vmov r3, s15
+; CHECK-NEXT:    adds r0, r0, r4
+; CHECK-NEXT:    vmov r4, s24
+; CHECK-NEXT:    adcs r2, r3
+; CHECK-NEXT:    adds.w lr, lr, r0
+; CHECK-NEXT:    vmov r0, s8
 ; CHECK-NEXT:    adc.w r12, r12, r2
-; CHECK-NEXT:    vmov r2, s6
-; CHECK-NEXT:    adds r2, r2, r4
-; CHECK-NEXT:    vmov r4, s13
-; CHECK-NEXT:    adcs r0, r3
-; CHECK-NEXT:    adds.w lr, lr, r2
-; CHECK-NEXT:    adc.w r12, r12, r0
-; CHECK-NEXT:    vmov r0, s16
-; CHECK-NEXT:    vmov r2, s12
+; CHECK-NEXT:    vmov r3, s25
+; CHECK-NEXT:    vmov r2, s9
+; CHECK-NEXT:    adds r0, r0, r4
+; CHECK-NEXT:    vmov r4, s16
+; CHECK-NEXT:    adcs r2, r3
 ; CHECK-NEXT:    vmov r3, s17
-; CHECK-NEXT:    adds r0, r0, r2
-; CHECK-NEXT:    adc.w r2, r4, r3
-; CHECK-NEXT:    vmov r3, s5
-; CHECK-NEXT:    vmov r4, s1
-; CHECK-NEXT:    adds r5, r5, r6
-; CHECK-NEXT:    adcs r3, r4
-; CHECK-NEXT:    adds r0, r0, r5
+; CHECK-NEXT:    adds r4, r4, r6
+; CHECK-NEXT:    adcs r3, r5
+; CHECK-NEXT:    adds r0, r0, r4
 ; CHECK-NEXT:    adcs r2, r3
 ; CHECK-NEXT:    vmov q0[2], q0[0], r0, lr
 ; CHECK-NEXT:    vmov q0[3], q0[1], r2, r12
 ; CHECK-NEXT:    vstrw.32 q0, [r1]
-; CHECK-NEXT:    vpop {d8, d9, d10, d11}
+; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13}
 ; CHECK-NEXT:    pop {r4, r5, r6, pc}
 entry:
   %l1 = load <8 x i64>, <8 x i64>* %src, align 8
   %s1 = shufflevector <8 x i64> %l1, <8 x i64> undef, <2 x i32> <i32 0, i32 4>
   %s2 = shufflevector <8 x i64> %l1, <8 x i64> undef, <2 x i32> <i32 1, i32 5>
   %s3 = shufflevector <8 x i64> %l1, <8 x i64> undef, <2 x i32> <i32 2, i32 6>
   %s4 = shufflevector <8 x i64> %l1, <8 x i64> undef, <2 x i32> <i32 3, i32 7>
   %a1 = add <2 x i64> %s1, %s2
   %a2 = add <2 x i64> %s3, %s4
   %a3 = add <2 x i64> %a1, %a2
   store <2 x i64> %a3, <2 x i64> *%dst
   ret void
 }
 
 define void @vld4_v4i64(<16 x i64> *%src, <4 x i64> *%dst) {
 ; CHECK-LABEL: vld4_v4i64:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .save {r4, r5, r6, r7, r8, r9, r10, lr}
 ; CHECK-NEXT:    push.w {r4, r5, r6, r7, r8, r9, r10, lr}
 ; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13, d14, d15}
 ; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13, d14, d15}
-; CHECK-NEXT:    .pad #64
-; CHECK-NEXT:    sub sp, #64
-; CHECK-NEXT:    vldrw.u32 q1, [r0, #64]
+; CHECK-NEXT:    .pad #32
+; CHECK-NEXT:    sub sp, #32
+; CHECK-NEXT:    vldrw.u32 q3, [r0, #64]
 ; CHECK-NEXT:    vldrw.u32 q0, [r0, #96]
-; CHECK-NEXT:    vldrw.u32 q4, [r0, #80]
-; CHECK-NEXT:    vldrw.u32 q6, [r0, #16]
-; CHECK-NEXT:    vmov.f64 d4, d3
-; CHECK-NEXT:    vstrw.32 q0, [sp, #16] @ 16-byte Spill
-; CHECK-NEXT:    vldrw.u32 q3, [r0]
-; CHECK-NEXT:    vstrw.32 q1, [sp, #32] @ 16-byte Spill
-; CHECK-NEXT:    vmov.f32 s9, s7
-; CHECK-NEXT:    vldrw.u32 q1, [r0, #32]
-; CHECK-NEXT:    vmov.f32 s10, s2
-; CHECK-NEXT:    vmov.f32 s11, s3
-; CHECK-NEXT:    vldrw.u32 q0, [r0, #112]
-; CHECK-NEXT:    vmov.f64 d14, d9
-; CHECK-NEXT:    vstrw.32 q2, [sp, #48] @ 16-byte Spill
+; CHECK-NEXT:    vldrw.u32 q4, [r0, #16]
+; CHECK-NEXT:    vldrw.u32 q1, [r0, #48]
+; CHECK-NEXT:    vmov.f64 d10, d6
 ; CHECK-NEXT:    vstrw.32 q0, [sp] @ 16-byte Spill
-; CHECK-NEXT:    vmov.f32 s29, s19
-; CHECK-NEXT:    vmov.f32 s30, s2
-; CHECK-NEXT:    vmov.f64 d4, d13
-; CHECK-NEXT:    vmov.f32 s31, s3
-; CHECK-NEXT:    vldrw.u32 q0, [r0, #48]
-; CHECK-NEXT:    vmov.f32 s9, s27
-; CHECK-NEXT:    vmov.f32 s10, s2
-; CHECK-NEXT:    vmov.f32 s26, s0
-; CHECK-NEXT:    vmov.f32 s11, s3
-; CHECK-NEXT:    vmov.f32 s27, s1
-; CHECK-NEXT:    vldrw.u32 q0, [sp] @ 16-byte Reload
-; CHECK-NEXT:    vmov r3, s10
-; CHECK-NEXT:    vmov r0, s26
-; CHECK-NEXT:    vmov.f64 d10, d7
-; CHECK-NEXT:    vmov r12, s11
-; CHECK-NEXT:    vmov r2, s27
-; CHECK-NEXT:    vmov.f32 s21, s15
-; CHECK-NEXT:    vmov.f32 s22, s6
-; CHECK-NEXT:    vmov.f32 s14, s4
-; CHECK-NEXT:    vmov.f32 s15, s5
-; CHECK-NEXT:    vmov.f32 s23, s7
-; CHECK-NEXT:    vldrw.u32 q1, [sp, #16] @ 16-byte Reload
-; CHECK-NEXT:    vmov r4, s14
-; CHECK-NEXT:    vmov.f32 s18, s0
-; CHECK-NEXT:    vmov.f32 s19, s1
-; CHECK-NEXT:    vldrw.u32 q0, [sp, #48] @ 16-byte Reload
-; CHECK-NEXT:    adds.w lr, r0, r3
-; CHECK-NEXT:    vmov r3, s22
-; CHECK-NEXT:    vmov r0, s15
-; CHECK-NEXT:    adc.w r12, r12, r2
-; CHECK-NEXT:    vmov r2, s23
-; CHECK-NEXT:    adds r3, r3, r4
-; CHECK-NEXT:    vmov r4, s28
-; CHECK-NEXT:    adcs r0, r2
-; CHECK-NEXT:    adds.w lr, lr, r3
-; CHECK-NEXT:    vmov r2, s16
-; CHECK-NEXT:    adc.w r12, r12, r0
-; CHECK-NEXT:    vmov r0, s29
-; CHECK-NEXT:    vmov r3, s17
-; CHECK-NEXT:    adds r2, r2, r4
-; CHECK-NEXT:    adcs r3, r0
-; CHECK-NEXT:    vmov r0, s1
-; CHECK-NEXT:    vldrw.u32 q0, [sp, #32] @ 16-byte Reload
+; CHECK-NEXT:    vldrw.u32 q2, [r0, #80]
+; CHECK-NEXT:    vldrw.u32 q6, [r0, #112]
+; CHECK-NEXT:    vstrw.32 q4, [sp, #16] @ 16-byte Spill
+; CHECK-NEXT:    vmov.f32 s21, s13
+; CHECK-NEXT:    vmov.f32 s22, s0
+; CHECK-NEXT:    vmov.f32 s23, s1
+; CHECK-NEXT:    vmov.f64 d0, d8
+; CHECK-NEXT:    vmov.f32 s1, s17
+; CHECK-NEXT:    vldrw.u32 q4, [r0]
 ; CHECK-NEXT:    vmov.f32 s2, s4
 ; CHECK-NEXT:    vmov.f32 s3, s5
-; CHECK-NEXT:    vldrw.u32 q1, [sp, #48] @ 16-byte Reload
+; CHECK-NEXT:    vmov r3, s2
+; CHECK-NEXT:    vmov r2, s6
+; CHECK-NEXT:    vmov.f64 d14, d4
+; CHECK-NEXT:    vmov r12, s3
+; CHECK-NEXT:    vldrw.u32 q0, [r0, #32]
+; CHECK-NEXT:    vmov r0, s7
+; CHECK-NEXT:    vmov.f64 d2, d8
+; CHECK-NEXT:    vmov.f32 s5, s17
+; CHECK-NEXT:    vmov.f32 s6, s0
+; CHECK-NEXT:    vmov.f32 s7, s1
+; CHECK-NEXT:    vmov r4, s2
+; CHECK-NEXT:    vmov.f32 s29, s9
+; CHECK-NEXT:    vmov.f32 s30, s24
+; CHECK-NEXT:    vmov.f32 s31, s25
+; CHECK-NEXT:    vmov r7, s13
+; CHECK-NEXT:    adds.w lr, r3, r2
+; CHECK-NEXT:    vmov r2, s3
+; CHECK-NEXT:    vmov r3, s7
+; CHECK-NEXT:    vldrw.u32 q0, [sp] @ 16-byte Reload
+; CHECK-NEXT:    vmov r6, s2
+; CHECK-NEXT:    vmov r5, s3
+; CHECK-NEXT:    vmov.f64 d0, d5
+; CHECK-NEXT:    adc.w r12, r12, r0
+; CHECK-NEXT:    vmov r0, s6
+; CHECK-NEXT:    vldrw.u32 q1, [sp, #16] @ 16-byte Reload
+; CHECK-NEXT:    vmov.f32 s1, s11
+; CHECK-NEXT:    adds r0, r0, r4
+; CHECK-NEXT:    vmov r4, s27
+; CHECK-NEXT:    adcs r3, r2
+; CHECK-NEXT:    adds.w lr, lr, r0
+; CHECK-NEXT:    adc.w r12, r12, r3
+; CHECK-NEXT:    vmov r3, s30
+; CHECK-NEXT:    vmov r2, s26
+; CHECK-NEXT:    vmov r0, s31
+; CHECK-NEXT:    adds r2, r2, r3
+; CHECK-NEXT:    adc.w r3, r0, r4
+; CHECK-NEXT:    vmov r4, s22
+; CHECK-NEXT:    vmov r0, s23
+; CHECK-NEXT:    adds r4, r4, r6
+; CHECK-NEXT:    vmov r6, s8
+; CHECK-NEXT:    adcs r5, r0
+; CHECK-NEXT:    adds.w r9, r4, r2
+; CHECK-NEXT:    vmov r4, s0
+; CHECK-NEXT:    adc.w r8, r5, r3
+; CHECK-NEXT:    vmov r2, s1
+; CHECK-NEXT:    vmov r5, s9
+; CHECK-NEXT:    vmov.f64 d0, d7
+; CHECK-NEXT:    vmov.f32 s1, s15
+; CHECK-NEXT:    vmov r3, s12
+; CHECK-NEXT:    vmov r0, s5
+; CHECK-NEXT:    adds r4, r4, r6
 ; CHECK-NEXT:    vmov r6, s0
-; CHECK-NEXT:    vmov r5, s4
+; CHECK-NEXT:    adcs r2, r5
+; CHECK-NEXT:    vmov r5, s1
+; CHECK-NEXT:    vmov.f64 d0, d9
+; CHECK-NEXT:    vmov.f32 s1, s19
+; CHECK-NEXT:    adds r3, r3, r6
+; CHECK-NEXT:    vmov r6, s17
+; CHECK-NEXT:    adcs r5, r7
+; CHECK-NEXT:    adds r3, r3, r4
+; CHECK-NEXT:    adc.w r10, r5, r2
+; CHECK-NEXT:    vmov r5, s0
+; CHECK-NEXT:    vmov r7, s16
 ; CHECK-NEXT:    vmov r4, s1
-; CHECK-NEXT:    vmov r7, s6
-; CHECK-NEXT:    adds r5, r5, r6
-; CHECK-NEXT:    vmov r6, s18
-; CHECK-NEXT:    adcs r4, r0
-; CHECK-NEXT:    adds.w r9, r5, r2
-; CHECK-NEXT:    vmov r5, s30
-; CHECK-NEXT:    adc.w r8, r4, r3
-; CHECK-NEXT:    vmov r2, s31
-; CHECK-NEXT:    vmov r4, s19
-; CHECK-NEXT:    vmov r3, s2
-; CHECK-NEXT:    vmov r0, s8
-; CHECK-NEXT:    adds r5, r5, r6
-; CHECK-NEXT:    vmov r6, s3
-; CHECK-NEXT:    adcs r2, r4
-; CHECK-NEXT:    vmov r4, s7
-; CHECK-NEXT:    adds r3, r3, r7
-; CHECK-NEXT:    vmov r7, s12
-; CHECK-NEXT:    adcs r4, r6
-; CHECK-NEXT:    adds r3, r3, r5
-; CHECK-NEXT:    vmov r6, s20
-; CHECK-NEXT:    adc.w r10, r4, r2
-; CHECK-NEXT:    vmov r4, s21
-; CHECK-NEXT:    vmov q1[2], q1[0], r9, r3
-; CHECK-NEXT:    vmov r5, s13
-; CHECK-NEXT:    vmov q1[3], q1[1], r8, r10
-; CHECK-NEXT:    vmov r2, s24
+; CHECK-NEXT:    vmov.f64 d0, d3
+; CHECK-NEXT:    vmov.f32 s1, s7
+; CHECK-NEXT:    vmov r2, s4
+; CHECK-NEXT:    vmov q1[2], q1[0], r3, r9
+; CHECK-NEXT:    vmov q1[3], q1[1], r10, r8
 ; CHECK-NEXT:    vstrw.32 q1, [r1, #16]
-; CHECK-NEXT:    adds r6, r6, r7
-; CHECK-NEXT:    vmov r7, s25
-; CHECK-NEXT:    adcs r4, r5
-; CHECK-NEXT:    vmov r5, s9
-; CHECK-NEXT:    adds r0, r0, r2
-; CHECK-NEXT:    adc.w r2, r7, r5
-; CHECK-NEXT:    adds r0, r0, r6
-; CHECK-NEXT:    vmov q0[2], q0[0], r0, lr
-; CHECK-NEXT:    adc.w r0, r4, r2
+; CHECK-NEXT:    adds r5, r5, r7
+; CHECK-NEXT:    vmov r7, s0
+; CHECK-NEXT:    adcs r4, r6
+; CHECK-NEXT:    vmov r6, s1
+; CHECK-NEXT:    adds r2, r2, r7
+; CHECK-NEXT:    adcs r0, r6
+; CHECK-NEXT:    adds r2, r2, r5
+; CHECK-NEXT:    vmov q0[2], q0[0], r2, lr
+; CHECK-NEXT:    adcs r0, r4
 ; CHECK-NEXT:    vmov q0[3], q0[1], r0, r12
 ; CHECK-NEXT:    vstrw.32 q0, [r1]
-; CHECK-NEXT:    add sp, #64
+; CHECK-NEXT:    add sp, #32
 ; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13, d14, d15}
 ; CHECK-NEXT:    pop.w {r4, r5, r6, r7, r8, r9, r10, pc}
 entry:
   %l1 = load <16 x i64>, <16 x i64>* %src, align 8
   %s1 = shufflevector <16 x i64> %l1, <16 x i64> undef, <4 x i32> <i32 0, i32 4, i32 8, i32 12>
   %s2 = shufflevector <16 x i64> %l1, <16 x i64> undef, <4 x i32> <i32 1, i32 5, i32 9, i32 13>
   %s3 = shufflevector <16 x i64> %l1, <16 x i64> undef, <4 x i32> <i32 2, i32 6, i32 10, i32 14>
   %s4 = shufflevector <16 x i64> %l1, <16 x i64> undef, <4 x i32> <i32 3, i32 7, i32 11, i32 15>

llvm/test/CodeGen/Thumb2/mve-vldst4.ll

 ; CHECK-NEXT:    vmovx.f16 s8, s15
 ; CHECK-NEXT:    vmovx.f16 s4, s13
 ; CHECK-NEXT:    vins.f16 s25, s27
 ; CHECK-NEXT:    vins.f16 s4, s8
 ; CHECK-NEXT:    vmovx.f16 s8, s27
 ; CHECK-NEXT:    vins.f16 s5, s8
 ; CHECK-NEXT:    vmov.f32 s1, s25
 ; CHECK-NEXT:    vmul.f16 q2, q1, r2
-; CHECK-NEXT:    vmul.f16 q0, q0, r2
-; CHECK-NEXT:    vmovx.f16 s7, s0
+; CHECK-NEXT:    vmul.f16 q1, q0, r2
+; CHECK-NEXT:    vmovx.f16 s3, s4
 ; CHECK-NEXT:    vmovx.f16 s28, s8
-; CHECK-NEXT:    vins.f16 s7, s28
+; CHECK-NEXT:    vins.f16 s3, s28
 ; CHECK-NEXT:    vmovx.f16 s30, s16
 ; CHECK-NEXT:    vmovx.f16 s31, s20
-; CHECK-NEXT:    vstrw.32 q1, [sp] @ 16-byte Spill
+; CHECK-NEXT:    vstrw.32 q0, [sp] @ 16-byte Spill
 ; CHECK-NEXT:    vmovx.f16 s28, s12
 ; CHECK-NEXT:    vins.f16 s16, s18
 ; CHECK-NEXT:    vmovx.f16 s29, s24
-; CHECK-NEXT:    vmovx.f16 s4, s18
+; CHECK-NEXT:    vmovx.f16 s0, s18
 ; CHECK-NEXT:    vins.f16 s20, s22
-; CHECK-NEXT:    vins.f16 s30, s4
-; CHECK-NEXT:    vmovx.f16 s4, s22
+; CHECK-NEXT:    vins.f16 s30, s0
+; CHECK-NEXT:    vmovx.f16 s0, s22
 ; CHECK-NEXT:    vins.f16 s12, s14
-; CHECK-NEXT:    vins.f16 s31, s4
-; CHECK-NEXT:    vmovx.f16 s4, s14
+; CHECK-NEXT:    vins.f16 s31, s0
+; CHECK-NEXT:    vmovx.f16 s0, s14
 ; CHECK-NEXT:    vmov.f32 s14, s16
 ; CHECK-NEXT:    vins.f16 s24, s26
 ; CHECK-NEXT:    vmov.f32 s15, s20
-; CHECK-NEXT:    vins.f16 s28, s4
-; CHECK-NEXT:    vmovx.f16 s4, s26
+; CHECK-NEXT:    vins.f16 s28, s0
+; CHECK-NEXT:    vmovx.f16 s0, s26
 ; CHECK-NEXT:    vmov.f32 s13, s24
-; CHECK-NEXT:    vins.f16 s29, s4
+; CHECK-NEXT:    vins.f16 s29, s0
 ; CHECK-NEXT:    vmul.f16 q3, q3, r2
 ; CHECK-NEXT:    vmul.f16 q7, q7, r2
-; CHECK-NEXT:    vmovx.f16 s4, s12
-; CHECK-NEXT:    vmovx.f16 s6, s28
-; CHECK-NEXT:    vins.f16 s0, s8
-; CHECK-NEXT:    vins.f16 s4, s6
-; CHECK-NEXT:    vmovx.f16 s6, s1
-; CHECK-NEXT:    vmovx.f16 s5, s9
+; CHECK-NEXT:    vmovx.f16 s0, s12
+; CHECK-NEXT:    vmovx.f16 s2, s28
+; CHECK-NEXT:    vins.f16 s4, s8
+; CHECK-NEXT:    vins.f16 s0, s2
+; CHECK-NEXT:    vmovx.f16 s2, s5
+; CHECK-NEXT:    vmovx.f16 s1, s9
 ; CHECK-NEXT:    vins.f16 s12, s28
-; CHECK-NEXT:    vins.f16 s6, s5
+; CHECK-NEXT:    vins.f16 s2, s1
 ; CHECK-NEXT:    vmovx.f16 s18, s13
-; CHECK-NEXT:    vmovx.f16 s5, s29
-; CHECK-NEXT:    vins.f16 s1, s9
-; CHECK-NEXT:    vins.f16 s18, s5
-; CHECK-NEXT:    vmovx.f16 s23, s2
-; CHECK-NEXT:    vmovx.f16 s5, s10
-; CHECK-NEXT:    vins.f16 s2, s10
-; CHECK-NEXT:    vins.f16 s23, s5
+; CHECK-NEXT:    vmovx.f16 s1, s29
+; CHECK-NEXT:    vins.f16 s5, s9
+; CHECK-NEXT:    vins.f16 s18, s1
+; CHECK-NEXT:    vmovx.f16 s23, s6
+; CHECK-NEXT:    vmovx.f16 s1, s10
+; CHECK-NEXT:    vins.f16 s6, s10
+; CHECK-NEXT:    vins.f16 s23, s1
 ; CHECK-NEXT:    vins.f16 s13, s29
-; CHECK-NEXT:    vmovx.f16 s27, s3
+; CHECK-NEXT:    vmovx.f16 s27, s7
 ; CHECK-NEXT:    vmovx.f16 s8, s11
 ; CHECK-NEXT:    vmovx.f16 s22, s14
 ; CHECK-NEXT:    vins.f16 s27, s8
 ; CHECK-NEXT:    vins.f16 s14, s30
 ; CHECK-NEXT:    vmovx.f16 s26, s15
 ; CHECK-NEXT:    vins.f16 s15, s31
 ; CHECK-NEXT:    vmovx.f16 s8, s31
-; CHECK-NEXT:    vins.f16 s3, s11
+; CHECK-NEXT:    vins.f16 s7, s11
 ; CHECK-NEXT:    vins.f16 s26, s8
 ; CHECK-NEXT:    vmov q2, q3
-; CHECK-NEXT:    vmovx.f16 s5, s30
+; CHECK-NEXT:    vmovx.f16 s1, s30
+; CHECK-NEXT:    vmov.f32 s10, s0
 ; CHECK-NEXT:    vldrw.u32 q7, [sp] @ 16-byte Reload
-; CHECK-NEXT:    vmov.f32 s10, s4
-; CHECK-NEXT:    vmov.f32 s29, s0
-; CHECK-NEXT:    vins.f16 s22, s5
-; CHECK-NEXT:    vmov.f32 s9, s0
-; CHECK-NEXT:    vmov.f32 s11, s31
-; CHECK-NEXT:    vmov q7, q0
-; CHECK-NEXT:    vmov.f32 s31, s6
+; CHECK-NEXT:    vmov.f32 s29, s4
+; CHECK-NEXT:    vins.f16 s22, s1
+; CHECK-NEXT:    vmov.f32 s28, s12
 ; CHECK-NEXT:    vmov.f32 s16, s13
-; CHECK-NEXT:    vmov.f32 s21, s2
-; CHECK-NEXT:    vmov.f32 s25, s3
-; CHECK-NEXT:    vmov.f32 s17, s29
+; CHECK-NEXT:    vmov.f32 s21, s6
+; CHECK-NEXT:    vmov.f32 s25, s7
 ; CHECK-NEXT:    vmov.f32 s20, s14
-; CHECK-NEXT:    vmov.f32 s24, s15
+; CHECK-NEXT:    vmov.f32 s30, s10
+; CHECK-NEXT:    vmov q2, q1
+; CHECK-NEXT:    vmov.f32 s11, s2
 ; CHECK-NEXT:    vstrh.16 q5, [r1, #32]
+; CHECK-NEXT:    vmov.f32 s17, s5
+; CHECK-NEXT:    vmov.f32 s24, s15
 ; CHECK-NEXT:    vstrh.16 q6, [r1, #48]
-; CHECK-NEXT:    vstrh.16 q2, [r1], #64
-; CHECK-NEXT:    vmov.f32 s19, s31
+; CHECK-NEXT:    vstrh.16 q7, [r1], #64
+; CHECK-NEXT:    vmov.f32 s19, s11
 ; CHECK-NEXT:    vstrh.16 q4, [r1, #-48]
 ; CHECK-NEXT:    le lr, .LBB0_2
 ; CHECK-NEXT:  .LBB0_3: @ %while.end
 ; CHECK-NEXT:    add sp, #24
 ; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13, d14, d15}
 ; CHECK-NEXT:    pop {r7, pc}
 entry:
   %tmp.0.extract.trunc = trunc i32 %scale.coerce to i16

llvm/test/CodeGen/Thumb2/mve-vmulh.ll

entry:
   %s = lshr <2 x i64> %m, <i64 32, i64 32>
   %s2 = trunc <2 x i64> %s to <2 x i32>
   ret <2 x i32> %s2
 }
 
 define arm_aapcs_vfpcc <4 x i32> @vmulhs_v4i32(<4 x i32> %s0, <4 x i32> %s1) {
 ; CHECK-LABEL: vmulhs_v4i32:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .vsave {d8, d9, d10, d11}
-; CHECK-NEXT:    vpush {d8, d9, d10, d11}
-; CHECK-NEXT:    vmov.f32 s8, s4
-; CHECK-NEXT:    vmov.f32 s12, s0
-; CHECK-NEXT:    vmov.f32 s10, s5
-; CHECK-NEXT:    vmov.f32 s14, s1
-; CHECK-NEXT:    vmov r0, s8
-; CHECK-NEXT:    vmov r1, s12
-; CHECK-NEXT:    vmov.f32 s16, s6
-; CHECK-NEXT:    vmov.f32 s18, s7
-; CHECK-NEXT:    vmov.f32 s4, s2
-; CHECK-NEXT:    vmov.f32 s6, s3
-; CHECK-NEXT:    vmullb.s32 q5, q1, q4
-; CHECK-NEXT:    smmul r0, r1, r0
-; CHECK-NEXT:    vmov r1, s21
-; CHECK-NEXT:    vmov q0[2], q0[0], r0, r1
-; CHECK-NEXT:    vmov r0, s10
-; CHECK-NEXT:    vmov r1, s14
+; CHECK-NEXT:    .vsave {d8, d9}
+; CHECK-NEXT:    vpush {d8, d9}
+; CHECK-NEXT:    vmov r1, s4
+; CHECK-NEXT:    vmov r2, s0
+; CHECK-NEXT:    vmov.f32 s8, s6
+; CHECK-NEXT:    vmov.f32 s16, s2
+; CHECK-NEXT:    vmov.f32 s10, s7
+; CHECK-NEXT:    vmov.f32 s18, s3
+; CHECK-NEXT:    vmov.f32 s6, s5
+; CHECK-NEXT:    vmullb.s32 q3, q4, q2
+; CHECK-NEXT:    vmov.f32 s2, s1
+; CHECK-NEXT:    vmov r0, s13
+; CHECK-NEXT:    smmul r1, r2, r1
+; CHECK-NEXT:    vmov q2[2], q2[0], r1, r0
+; CHECK-NEXT:    vmov r0, s6
+; CHECK-NEXT:    vmov r1, s2
 ; CHECK-NEXT:    smmul r0, r1, r0
-; CHECK-NEXT:    vmov r1, s23
-; CHECK-NEXT:    vmov q0[3], q0[1], r0, r1
-; CHECK-NEXT:    vpop {d8, d9, d10, d11}
+; CHECK-NEXT:    vmov r1, s15
+; CHECK-NEXT:    vmov q2[3], q2[1], r0, r1
+; CHECK-NEXT:    vmov q0, q2
+; CHECK-NEXT:    vpop {d8, d9}
 ; CHECK-NEXT:    bx lr
 entry:
   %s0s = sext <4 x i32> %s0 to <4 x i64>
   %s1s = sext <4 x i32> %s1 to <4 x i64>
   %m = mul <4 x i64> %s0s, %s1s
   %s = ashr <4 x i64> %m, <i64 32, i64 32, i64 32, i64 32>
   %s2 = trunc <4 x i64> %s to <4 x i32>
   ret <4 x i32> %s2

llvm/test/CodeGen/Thumb2/mve-vmull-loop.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=thumbv8.1m.main-none-none-eabi -mattr=+mve -verify-machineinstrs %s -o - | FileCheck %s --check-prefix=CHECK
 
 define arm_aapcs_vfpcc void @test32(i32* noalias nocapture readonly %x, i32* noalias nocapture readonly %y, i32* nocapture %z, i32 %n) {
 ; CHECK-LABEL: test32:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .save {r4, r5, r7, lr}
-; CHECK-NEXT:    push {r4, r5, r7, lr}
+; CHECK-NEXT:    .save {r4, r5, r6, lr}
+; CHECK-NEXT:    push {r4, r5, r6, lr}
 ; CHECK-NEXT:    cmp r3, #1
 ; CHECK-NEXT:    blt .LBB0_2
 ; CHECK-NEXT:  .LBB0_1: @ %vector.body
 ; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
-; CHECK-NEXT:    vldrw.u32 q0, [r0], #16
-; CHECK-NEXT:    vldrw.u32 q1, [r1], #16
+; CHECK-NEXT:    vldrw.u32 q1, [r0], #16
+; CHECK-NEXT:    vldrw.u32 q2, [r1], #16
 ; CHECK-NEXT:    subs r3, #4
-; CHECK-NEXT:    vmullb.s32 q2, q1, q0
-; CHECK-NEXT:    vmullt.s32 q3, q1, q0
-; CHECK-NEXT:    vmov r5, s11
-; CHECK-NEXT:    vmov r12, s10
-; CHECK-NEXT:    lsrl r12, r5, #31
-; CHECK-NEXT:    vmov r4, s8
-; CHECK-NEXT:    vmov r5, s9
+; CHECK-NEXT:    vmullt.s32 q0, q2, q1
+; CHECK-NEXT:    vmullb.s32 q3, q2, q1
+; CHECK-NEXT:    vmov r5, s3
+; CHECK-NEXT:    vmov r4, s2
 ; CHECK-NEXT:    lsrl r4, r5, #31
-; CHECK-NEXT:    vmov q2[2], q2[0], r4, r12
-; CHECK-NEXT:    vmov r5, s15
-; CHECK-NEXT:    vmov r12, s14
+; CHECK-NEXT:    vmov r12, s0
+; CHECK-NEXT:    vmov r5, s1
 ; CHECK-NEXT:    lsrl r12, r5, #31
-; CHECK-NEXT:    vmov r4, s12
-; CHECK-NEXT:    vmov r5, s13
+; CHECK-NEXT:    vmov r6, s12
+; CHECK-NEXT:    vmov.32 q0[1], r5
+; CHECK-NEXT:    vmov r5, s15
+; CHECK-NEXT:    vmov q0[2], q0[0], r12, r4
+; CHECK-NEXT:    vmov r4, s14
 ; CHECK-NEXT:    lsrl r4, r5, #31
-; CHECK-NEXT:    vmov q2[3], q2[1], r4, r12
+; CHECK-NEXT:    vmov r5, s13
+; CHECK-NEXT:    lsrl r6, r5, #31
+; CHECK-NEXT:    vmov.32 q1[1], r5
+; CHECK-NEXT:    vmov.32 q2[0], r6
+; CHECK-NEXT:    vmov q1[2], q1[0], r6, r4
+; CHECK-NEXT:    vmov.f32 s7, s2
+; CHECK-NEXT:    vmov.32 q0[0], r12
+; CHECK-NEXT:    vmov.f32 s9, s0
+; CHECK-NEXT:    vmov.f32 s10, s6
+; CHECK-NEXT:    vmov.f32 s11, s7
 ; CHECK-NEXT:    vstrb.8 q2, [r2], #16
 ; CHECK-NEXT:    bne .LBB0_1
 ; CHECK-NEXT:  .LBB0_2: @ %for.cond.cleanup
-; CHECK-NEXT:    pop {r4, r5, r7, pc}
+; CHECK-NEXT:    pop {r4, r5, r6, pc}
 entry:
   %0 = and i32 %n, 3
   %cmp = icmp eq i32 %0, 0
   %cmp113 = icmp sgt i32 %n, 0
   br i1 %cmp113, label %vector.body, label %for.cond.cleanup
 
 vector.body:                                      ; preds = %vector.body, %entry
   %index = phi i32 [ %index.next, %vector.body ], [ 0, %entry ]

llvm/test/CodeGen/Thumb2/mve-vqmovn.ll

 ; CHECK-NEXT:    it lt
 ; CHECK-NEXT:    movlt r2, #1
 ; CHECK-NEXT:    cmp r2, #0
 ; CHECK-NEXT:    csetm r2, ne
 ; CHECK-NEXT:    vmov q1[2], q1[0], r2, r1
 ; CHECK-NEXT:    vmov q1[3], q1[1], r2, r1
 ; CHECK-NEXT:    adr r1, .LCPI12_0
 ; CHECK-NEXT:    vldrw.u32 q2, [r1]
-; CHECK-NEXT:    vand q0, q0, q1
 ; CHECK-NEXT:    vbic q2, q2, q1
-; CHECK-NEXT:    vorr q0, q0, q2
-; CHECK-NEXT:    vmov r2, s2
-; CHECK-NEXT:    vmov r1, s3
-; CHECK-NEXT:    vmov r3, s0
-; CHECK-NEXT:    rsbs.w r2, r2, #-2147483648
-; CHECK-NEXT:    vmov r2, s1
-; CHECK-NEXT:    sbcs.w r1, r12, r1
+; CHECK-NEXT:    vand q1, q0, q1
+; CHECK-NEXT:    vorr q0, q1, q2
+; CHECK-NEXT:    vmov r2, s7
+; CHECK-NEXT:    vmov r1, s2
+; CHECK-NEXT:    vmov r3, s5
+; CHECK-NEXT:    rsbs.w r1, r1, #-2147483648
+; CHECK-NEXT:    sbcs.w r1, r12, r2
+; CHECK-NEXT:    vmov r2, s0
 ; CHECK-NEXT:    mov.w r1, #0
 ; CHECK-NEXT:    it lt
 ; CHECK-NEXT:    movlt r1, #1
 ; CHECK-NEXT:    cmp r1, #0
 ; CHECK-NEXT:    csetm r1, ne
-; CHECK-NEXT:    rsbs.w r3, r3, #-2147483648
-; CHECK-NEXT:    sbcs.w r2, r12, r2
+; CHECK-NEXT:    rsbs.w r2, r2, #-2147483648
+; CHECK-NEXT:    sbcs.w r2, r12, r3
 ; CHECK-NEXT:    it lt
 ; CHECK-NEXT:    movlt r0, #1
 ; CHECK-NEXT:    cmp r0, #0
 ; CHECK-NEXT:    csetm r0, ne
 ; CHECK-NEXT:    vmov q1[2], q1[0], r0, r1
 ; CHECK-NEXT:    vmov q1[3], q1[1], r0, r1
 ; CHECK-NEXT:    adr r0, .LCPI12_1
 ; CHECK-NEXT:    vldrw.u32 q2, [r0]

llvm/test/CodeGen/Thumb2/mve-vqshrn.ll

 ; CHECK-NEXT:    it lt
 ; CHECK-NEXT:    movlt r0, #1
 ; CHECK-NEXT:    cmp r0, #0
 ; CHECK-NEXT:    csetm r0, ne
 ; CHECK-NEXT:    vmov q0[2], q0[0], r0, r5
 ; CHECK-NEXT:    vmov q0[3], q0[1], r0, r5
 ; CHECK-NEXT:    adr r0, .LCPI12_0
 ; CHECK-NEXT:    vldrw.u32 q1, [r0]
+; CHECK-NEXT:    vand q2, q2, q0
+; CHECK-NEXT:    vmov r1, s11
 ; CHECK-NEXT:    vbic q1, q1, q0
-; CHECK-NEXT:    vand q0, q2, q0
-; CHECK-NEXT:    vorr q0, q0, q1
-; CHECK-NEXT:    vmov r1, s2
-; CHECK-NEXT:    vmov r0, s3
-; CHECK-NEXT:    vmov r3, s0
-; CHECK-NEXT:    rsbs.w r1, r1, #-2147483648
-; CHECK-NEXT:    vmov r1, s1
-; CHECK-NEXT:    sbcs.w r0, r2, r0
+; CHECK-NEXT:    vmov r3, s9
+; CHECK-NEXT:    vorr q0, q2, q1
+; CHECK-NEXT:    vmov r0, s2
+; CHECK-NEXT:    rsbs.w r0, r0, #-2147483648
+; CHECK-NEXT:    sbcs.w r0, r2, r1
+; CHECK-NEXT:    vmov r1, s0
 ; CHECK-NEXT:    mov.w r0, #0
 ; CHECK-NEXT:    it lt
 ; CHECK-NEXT:    movlt r0, #1
 ; CHECK-NEXT:    cmp r0, #0
 ; CHECK-NEXT:    csetm r0, ne
-; CHECK-NEXT:    rsbs.w r3, r3, #-2147483648
-; CHECK-NEXT:    sbcs.w r1, r2, r1
+; CHECK-NEXT:    rsbs.w r1, r1, #-2147483648
+; CHECK-NEXT:    sbcs.w r1, r2, r3
 ; CHECK-NEXT:    it lt
 ; CHECK-NEXT:    movlt.w lr, #1
 ; CHECK-NEXT:    cmp.w lr, #0
 ; CHECK-NEXT:    csetm r1, ne
 ; CHECK-NEXT:    vmov q1[2], q1[0], r1, r0
 ; CHECK-NEXT:    vmov q1[3], q1[1], r1, r0
 ; CHECK-NEXT:    adr r0, .LCPI12_1
 ; CHECK-NEXT:    vldrw.u32 q2, [r0]

llvm/test/CodeGen/Thumb2/mve-vst2.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=thumbv8.1m.main-none-none-eabi -mattr=+mve.fp,+fp64 -verify-machineinstrs %s -o - | FileCheck %s
 
 ; i32
 
 define void @vst2_v2i32(<2 x i32> *%src, <4 x i32> *%dst) {
 ; CHECK-LABEL: vst2_v2i32:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    ldm.w r0, {r2, r3, r12}
-; CHECK-NEXT:    ldr r0, [r0, #12]
-; CHECK-NEXT:    vmov q0[2], q0[0], r2, r3
-; CHECK-NEXT:    vmov q0[3], q0[1], r12, r0
-; CHECK-NEXT:    vstrw.32 q0, [r1]
+; CHECK-NEXT:    ldrd r2, r12, [r0]
+; CHECK-NEXT:    ldrd r3, r0, [r0, #8]
+; CHECK-NEXT:    vmov q0[2], q0[0], r2, r12
+; CHECK-NEXT:    vmov.32 q2[0], r2
+; CHECK-NEXT:    vmov q1[2], q1[0], r3, r0
+; CHECK-NEXT:    vmov.f32 s3, s6
+; CHECK-NEXT:    vmov.32 q1[0], r3
+; CHECK-NEXT:    vmov.f32 s9, s4
+; CHECK-NEXT:    vmov.f32 s10, s2
+; CHECK-NEXT:    vmov.f32 s11, s3
+; CHECK-NEXT:    vstrw.32 q2, [r1]
 ; CHECK-NEXT:    bx lr
 entry:
   %s1 = getelementptr <2 x i32>, <2 x i32>* %src, i32 0
   %l1 = load <2 x i32>, <2 x i32>* %s1, align 4
   %s2 = getelementptr <2 x i32>, <2 x i32>* %src, i32 1
   %l2 = load <2 x i32>, <2 x i32>* %s2, align 4
   %s = shufflevector <2 x i32> %l1, <2 x i32> %l2, <4 x i32> <i32 0, i32 2, i32 1, i32 3>
   store <4 x i32> %s, <4 x i32> *%dst, align 4
entry:
   ret void
 }
 
 define void @vst2_v4f16(<4 x half> *%src, <8 x half> *%dst) {
 ; CHECK-LABEL: vst2_v4f16:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    ldrd r2, r12, [r0]
 ; CHECK-NEXT:    ldrd r3, r0, [r0, #8]
-; CHECK-NEXT:    vmov.32 q0[0], r2
-; CHECK-NEXT:    vmov.32 q1[0], r3
-; CHECK-NEXT:    vmov.32 q0[1], r12
-; CHECK-NEXT:    vmov.32 q1[1], r0
-; CHECK-NEXT:    vmovx.f16 s8, s0
-; CHECK-NEXT:    vmovx.f16 s10, s4
-; CHECK-NEXT:    vins.f16 s0, s4
-; CHECK-NEXT:    vins.f16 s8, s10
-; CHECK-NEXT:    vmovx.f16 s10, s1
-; CHECK-NEXT:    vmovx.f16 s4, s5
-; CHECK-NEXT:    vins.f16 s1, s5
-; CHECK-NEXT:    vins.f16 s10, s4
-; CHECK-NEXT:    vmov q1, q0
-; CHECK-NEXT:    vmov.f32 s5, s8
-; CHECK-NEXT:    vmov.f32 s6, s1
-; CHECK-NEXT:    vmov.f32 s7, s10
-; CHECK-NEXT:    vstrh.16 q1, [r1]
+; CHECK-NEXT:    vmov.32 q1[0], r2
+; CHECK-NEXT:    vmovx.f16 s1, s4
+; CHECK-NEXT:    vmov.32 q2[0], r3
+; CHECK-NEXT:    vmov.f32 s0, s4
+; CHECK-NEXT:    vmovx.f16 s12, s8
+; CHECK-NEXT:    vins.f16 s0, s8
+; CHECK-NEXT:    vmov.32 q2[1], r0
+; CHECK-NEXT:    vins.f16 s1, s12
+; CHECK-NEXT:    vmov.32 q1[1], r12
+; CHECK-NEXT:    vmovx.f16 s3, s5
+; CHECK-NEXT:    vmovx.f16 s8, s9
+; CHECK-NEXT:    vins.f16 s5, s9
+; CHECK-NEXT:    vins.f16 s3, s8
+; CHECK-NEXT:    vmov.f32 s2, s5
+; CHECK-NEXT:    vstrh.16 q0, [r1]
 ; CHECK-NEXT:    bx lr
 entry:
   %s1 = getelementptr <4 x half>, <4 x half>* %src, i32 0
   %l1 = load <4 x half>, <4 x half>* %s1, align 4
   %s2 = getelementptr <4 x half>, <4 x half>* %src, i32 1
   %l2 = load <4 x half>, <4 x half>* %s2, align 4
   %s = shufflevector <4 x half> %l1, <4 x half> %l2, <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
   store <8 x half> %s, <8 x half> *%dst, align 2

llvm/test/CodeGen/Thumb2/mve-vst3.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=thumbv8.1m.main-none-none-eabi -mattr=+mve.fp,+fp64 -verify-machineinstrs %s -o - | FileCheck %s
 
 ; i32
 
 define void @vst3_v2i32(<2 x i32> *%src, <6 x i32> *%dst) {
 ; CHECK-LABEL: vst3_v2i32:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .save {r4, lr}
 ; CHECK-NEXT:    push {r4, lr}
-; CHECK-NEXT:    ldrd lr, r12, [r0]
-; CHECK-NEXT:    ldrd r3, r2, [r0, #8]
+; CHECK-NEXT:    ldm.w r0, {r2, r3, r12, lr}
 ; CHECK-NEXT:    ldrd r4, r0, [r0, #16]
-; CHECK-NEXT:    vmov q1[2], q1[0], lr, r3
+; CHECK-NEXT:    vmov q2[2], q2[0], r2, r12
 ; CHECK-NEXT:    vmov.32 q0[0], r4
-; CHECK-NEXT:    vmov q1[3], q1[1], r12, r2
-; CHECK-NEXT:    vmov.32 q0[1], r0
-; CHECK-NEXT:    vmov.f32 s8, s7
-; CHECK-NEXT:    vmov.f32 s10, s1
-; CHECK-NEXT:    vmov r2, s8
-; CHECK-NEXT:    vmov.f64 d4, d2
+; CHECK-NEXT:    vmov q2[3], q2[1], r3, lr
+; CHECK-NEXT:    vmov q1, q0
+; CHECK-NEXT:    vmov.32 q1[1], r0
+; CHECK-NEXT:    vmov.f32 s4, s11
+; CHECK-NEXT:    vmov.32 q2[0], r2
+; CHECK-NEXT:    vmov q3, q2
+; CHECK-NEXT:    vmov.32 q3[1], r3
+; CHECK-NEXT:    vmov r4, s4
+; CHECK-NEXT:    vmov.32 q1[1], r3
+; CHECK-NEXT:    vmov q1[2], q1[0], r2, r12
 ; CHECK-NEXT:    vmov.f32 s9, s6
 ; CHECK-NEXT:    vmov.f32 s10, s0
-; CHECK-NEXT:    vmov.f32 s11, s5
+; CHECK-NEXT:    vmov.f32 s11, s13
 ; CHECK-NEXT:    vstrw.32 q2, [r1]
-; CHECK-NEXT:    strd r2, r0, [r1, #16]
+; CHECK-NEXT:    strd r4, r0, [r1, #16]
 ; CHECK-NEXT:    pop {r4, pc}
 entry:
   %s1 = getelementptr <2 x i32>, <2 x i32>* %src, i32 0
   %l1 = load <2 x i32>, <2 x i32>* %s1, align 4
   %s2 = getelementptr <2 x i32>, <2 x i32>* %src, i32 1
   %l2 = load <2 x i32>, <2 x i32>* %s2, align 4
   %s3 = getelementptr <2 x i32>, <2 x i32>* %src, i32 2
   %l3 = load <2 x i32>, <2 x i32>* %s3, align 4

 
 ; i16
 
 define void @vst3_v2i16(<2 x i16> *%src, <6 x i16> *%dst) {
 ; CHECK-LABEL: vst3_v2i16:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .save {r4, lr}
 ; CHECK-NEXT:    push {r4, lr}
-; CHECK-NEXT:    ldrh r2, [r0, #6]
+; CHECK-NEXT:    ldrh r4, [r0, #6]
 ; CHECK-NEXT:    ldrh.w lr, [r0, #4]
+; CHECK-NEXT:    ldrh r3, [r0]
+; CHECK-NEXT:    vmov.16 q0[4], r4
 ; CHECK-NEXT:    ldrh.w r12, [r0, #8]
-; CHECK-NEXT:    vmov.16 q0[4], r2
-; CHECK-NEXT:    ldrh r3, [r0, #2]
-; CHECK-NEXT:    vmov q1[2], q1[0], lr, r2
-; CHECK-NEXT:    ldrh r4, [r0]
+; CHECK-NEXT:    vmov.32 q1[0], lr
+; CHECK-NEXT:    ldrh r2, [r0, #2]
+; CHECK-NEXT:    vmov.32 q2[0], r3
 ; CHECK-NEXT:    ldrh r0, [r0, #10]
+; CHECK-NEXT:    vmov.f32 s9, s4
 ; CHECK-NEXT:    vmov.16 q0[5], r0
 ; CHECK-NEXT:    vmov r0, s2
-; CHECK-NEXT:    vmov q0[2], q0[0], r4, r3
-; CHECK-NEXT:    vmov.f32 s1, s4
-; CHECK-NEXT:    vmov.f32 s3, s2
-; CHECK-NEXT:    vmov.32 q0[2], r12
-; CHECK-NEXT:    vstrh.32 q0, [r1]
+; CHECK-NEXT:    vmov q0[2], q0[0], r3, r2
+; CHECK-NEXT:    vmov.f32 s11, s2
+; CHECK-NEXT:    vmov.32 q2[2], r12
+; CHECK-NEXT:    vstrh.32 q2, [r1]
 ; CHECK-NEXT:    str r0, [r1, #8]
 ; CHECK-NEXT:    pop {r4, pc}
 entry:
   %s1 = getelementptr <2 x i16>, <2 x i16>* %src, i32 0
   %l1 = load <2 x i16>, <2 x i16>* %s1, align 4
   %s2 = getelementptr <2 x i16>, <2 x i16>* %src, i32 1
   %l2 = load <2 x i16>, <2 x i16>* %s2, align 4
   %s3 = getelementptr <2 x i16>, <2 x i16>* %src, i32 2
entry:
   ret void
 }
 
 define void @vst3_v8i16(<8 x i16> *%src, <24 x i16> *%dst) {
 ; CHECK-LABEL: vst3_v8i16:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12}
 ; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12}
-; CHECK-NEXT:    vldrw.u32 q3, [r0]
-; CHECK-NEXT:    vldrw.u32 q1, [r0, #16]
-; CHECK-NEXT:    vmov.f64 d0, d6
-; CHECK-NEXT:    vmov.u16 r2, q1[1]
-; CHECK-NEXT:    vmovx.f16 s20, s12
-; CHECK-NEXT:    vins.f16 s0, s4
-; CHECK-NEXT:    vmov.f32 s8, s13
-; CHECK-NEXT:    vins.f16 s8, s5
-; CHECK-NEXT:    vmov.16 q0[4], r2
-; CHECK-NEXT:    vmov.f32 s3, s8
-; CHECK-NEXT:    vldrw.u32 q2, [r0, #32]
-; CHECK-NEXT:    vmov.f32 s1, s12
-; CHECK-NEXT:    vmov.f32 s17, s8
-; CHECK-NEXT:    vmov.f32 s18, s8
-; CHECK-NEXT:    vins.f16 s17, s20
-; CHECK-NEXT:    vmovx.f16 s20, s18
+; CHECK-NEXT:    vldrw.u32 q1, [r0, #32]
+; CHECK-NEXT:    vldrw.u32 q2, [r0, #16]
+; CHECK-NEXT:    vldrw.u32 q4, [r0]
+; CHECK-NEXT:    vmovx.f16 s12, s6
+; CHECK-NEXT:    vmovx.f16 s0, s10
+; CHECK-NEXT:    vins.f16 s0, s12
+; CHECK-NEXT:    vmovx.f16 s12, s7
+; CHECK-NEXT:    vins.f16 s1, s11
+; CHECK-NEXT:    vmov.u16 r0, q2[1]
+; CHECK-NEXT:    vmovx.f16 s3, s11
+; CHECK-NEXT:    vins.f16 s3, s12
+; CHECK-NEXT:    vmov.f32 s13, s19
+; CHECK-NEXT:    vmov.f32 s2, s7
+; CHECK-NEXT:    vmovx.f16 s20, s1
+; CHECK-NEXT:    vmov.f32 s14, s19
+; CHECK-NEXT:    vins.f16 s13, s20
+; CHECK-NEXT:    vmovx.f16 s20, s14
 ; CHECK-NEXT:    vins.f16 s2, s20
-; CHECK-NEXT:    vmovx.f16 s20, s10
-; CHECK-NEXT:    vmov.f32 s18, s2
-; CHECK-NEXT:    vmov.f32 s1, s17
-; CHECK-NEXT:    vmov.f32 s2, s18
-; CHECK-NEXT:    vmovx.f16 s16, s6
-; CHECK-NEXT:    vins.f16 s16, s20
-; CHECK-NEXT:    vmovx.f16 s20, s11
-; CHECK-NEXT:    vins.f16 s17, s7
-; CHECK-NEXT:    vstrw.32 q0, [r1]
-; CHECK-NEXT:    vmovx.f16 s19, s7
-; CHECK-NEXT:    vrev32.16 q1, q1
-; CHECK-NEXT:    vins.f16 s19, s20
-; CHECK-NEXT:    vmov.f32 s21, s15
-; CHECK-NEXT:    vmov.f32 s18, s11
-; CHECK-NEXT:    vmovx.f16 s24, s17
-; CHECK-NEXT:    vmov.f32 s22, s15
+; CHECK-NEXT:    vmovx.f16 s20, s17
+; CHECK-NEXT:    vmov.f32 s14, s2
+; CHECK-NEXT:    vmov.f32 s1, s13
+; CHECK-NEXT:    vmov.f32 s2, s14
+; CHECK-NEXT:    vmov.f32 s12, s5
+; CHECK-NEXT:    vstrw.32 q0, [r1, #32]
+; CHECK-NEXT:    vins.f16 s12, s20
+; CHECK-NEXT:    vmovx.f16 s20, s18
+; CHECK-NEXT:    vins.f16 s6, s20
+; CHECK-NEXT:    vrev32.16 q5, q2
+; CHECK-NEXT:    vmov.f32 s15, s6
+; CHECK-NEXT:    vmov.f32 s13, s5
+; CHECK-NEXT:    vmov.f32 s14, s18
+; CHECK-NEXT:    vmovx.f16 s24, s13
 ; CHECK-NEXT:    vins.f16 s21, s24
 ; CHECK-NEXT:    vmovx.f16 s24, s22
-; CHECK-NEXT:    vins.f16 s18, s24
-; CHECK-NEXT:    vmov.f32 s8, s9
-; CHECK-NEXT:    vmov.f32 s22, s18
-; CHECK-NEXT:    vmov.f32 s17, s21
-; CHECK-NEXT:    vmov.f32 s18, s22
-; CHECK-NEXT:    vmovx.f16 s20, s13
-; CHECK-NEXT:    vins.f16 s8, s20
-; CHECK-NEXT:    vmovx.f16 s20, s14
-; CHECK-NEXT:    vins.f16 s10, s20
-; CHECK-NEXT:    vstrw.32 q4, [r1, #32]
-; CHECK-NEXT:    vmov.f32 s11, s10
-; CHECK-NEXT:    vmov.f32 s10, s14
-; CHECK-NEXT:    vmovx.f16 s12, s9
-; CHECK-NEXT:    vins.f16 s5, s12
-; CHECK-NEXT:    vmovx.f16 s12, s6
-; CHECK-NEXT:    vins.f16 s10, s12
-; CHECK-NEXT:    vmov.f32 s6, s10
-; CHECK-NEXT:    vmov.f32 s9, s5
-; CHECK-NEXT:    vmov.f32 s10, s6
-; CHECK-NEXT:    vstrw.32 q2, [r1, #16]
+; CHECK-NEXT:    vins.f16 s14, s24
+; CHECK-NEXT:    vmov.f32 s22, s14
+; CHECK-NEXT:    vmov.f32 s13, s21
+; CHECK-NEXT:    vmov.f32 s14, s22
+; CHECK-NEXT:    vmov.f64 d10, d8
+; CHECK-NEXT:    vins.f16 s17, s9
+; CHECK-NEXT:    vstrw.32 q3, [r1, #16]
+; CHECK-NEXT:    vins.f16 s20, s8
+; CHECK-NEXT:    vmov q2, q5
+; CHECK-NEXT:    vmov.f32 s21, s16
+; CHECK-NEXT:    vmov.16 q2[4], r0
+; CHECK-NEXT:    vmov.f32 s11, s17
+; CHECK-NEXT:    vmov.f32 s22, s10
+; CHECK-NEXT:    vmov.f32 s23, s11
+; CHECK-NEXT:    vmovx.f16 s16, s21
+; CHECK-NEXT:    vmov.f32 s9, s4
+; CHECK-NEXT:    vmov.f32 s10, s4
+; CHECK-NEXT:    vins.f16 s9, s16
+; CHECK-NEXT:    vmovx.f16 s4, s10
+; CHECK-NEXT:    vins.f16 s22, s4
+; CHECK-NEXT:    vmov.f32 s10, s22
+; CHECK-NEXT:    vmov.f32 s21, s9
+; CHECK-NEXT:    vmov.f32 s22, s10
+; CHECK-NEXT:    vstrw.32 q5, [r1]
 ; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12}
 ; CHECK-NEXT:    bx lr
 entry:
   %s1 = getelementptr <8 x i16>, <8 x i16>* %src, i32 0
   %l1 = load <8 x i16>, <8 x i16>* %s1, align 4
   %s2 = getelementptr <8 x i16>, <8 x i16>* %src, i32 1
   %l2 = load <8 x i16>, <8 x i16>* %s2, align 4
   %s3 = getelementptr <8 x i16>, <8 x i16>* %src, i32 2
   %l3 = load <8 x i16>, <8 x i16>* %s3, align 4
   %t1 = shufflevector <8 x i16> %l1, <8 x i16> %l2, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
   %t2 = shufflevector <8 x i16> %l3, <8 x i16> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
   %s = shufflevector <16 x i16> %t1, <16 x i16> %t2, <24 x i32> <i32 0, i32 8, i32 16, i32 1, i32 9, i32 17, i32 2, i32 10, i32 18, i32 3, i32 11, i32 19, i32 4, i32 12, i32 20, i32 5, i32 13, i32 21, i32 6, i32 14, i32 22, i32 7, i32 15, i32 23>
   store <24 x i16> %s, <24 x i16> *%dst
   ret void
 }
 
 define void @vst3_v16i16(<16 x i16> *%src, <48 x i16> *%dst) {
 ; CHECK-LABEL: vst3_v16i16:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13, d14, d15}
 ; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13, d14, d15}
 ; CHECK-NEXT:    .pad #80
 ; CHECK-NEXT:    sub sp, #80
-; CHECK-NEXT:    vldrw.u32 q1, [r0, #48]
-; CHECK-NEXT:    vldrw.u32 q3, [r0, #80]
-; CHECK-NEXT:    vldrw.u32 q6, [r0, #32]
-; CHECK-NEXT:    vldrw.u32 q5, [r0, #64]
-; CHECK-NEXT:    vmovx.f16 s0, s14
-; CHECK-NEXT:    vmovx.f16 s8, s6
+; CHECK-NEXT:    vldrw.u32 q3, [r0, #48]
+; CHECK-NEXT:    vldrw.u32 q1, [r0, #80]
+; CHECK-NEXT:    vldrw.u32 q5, [r0, #32]
+; CHECK-NEXT:    vldrw.u32 q7, [r0]
+; CHECK-NEXT:    vmovx.f16 s0, s6
+; CHECK-NEXT:    vmovx.f16 s8, s14
 ; CHECK-NEXT:    vins.f16 s8, s0
-; CHECK-NEXT:    vmovx.f16 s0, s15
-; CHECK-NEXT:    vins.f16 s9, s7
-; CHECK-NEXT:    vstrw.32 q1, [sp, #48] @ 16-byte Spill
-; CHECK-NEXT:    vmovx.f16 s11, s7
-; CHECK-NEXT:    vmov.u16 r2, q6[1]
+; CHECK-NEXT:    vmovx.f16 s0, s7
+; CHECK-NEXT:    vins.f16 s9, s15
+; CHECK-NEXT:    vstrw.32 q3, [sp, #64] @ 16-byte Spill
+; CHECK-NEXT:    vmovx.f16 s11, s15
+; CHECK-NEXT:    vmovx.f16 s12, s22
 ; CHECK-NEXT:    vins.f16 s11, s0
-; CHECK-NEXT:    vstrw.32 q6, [sp] @ 16-byte Spill
-; CHECK-NEXT:    vmov.f32 s10, s15
-; CHECK-NEXT:    vmovx.f16 s4, s9
+; CHECK-NEXT:    vstrw.32 q5, [sp, #48] @ 16-byte Spill
+; CHECK-NEXT:    vmov.f32 s10, s7
+; CHECK-NEXT:    vmovx.f16 s0, s9
 ; CHECK-NEXT:    vmov q4, q2
 ; CHECK-NEXT:    vldrw.u32 q2, [r0, #16]
-; CHECK-NEXT:    vmov.f32 s1, s11
+; CHECK-NEXT:    vmov.f32 s25, s11
 ; CHECK-NEXT:    vstrw.32 q2, [sp, #32] @ 16-byte Spill
-; CHECK-NEXT:    vmov.f32 s2, s11
-; CHECK-NEXT:    vins.f16 s1, s4
-; CHECK-NEXT:    vmovx.f16 s4, s2
-; CHECK-NEXT:    vins.f16 s18, s4
-; CHECK-NEXT:    vldrw.u32 q1, [r0]
-; CHECK-NEXT:    vmov.f32 s2, s18
-; CHECK-NEXT:    vmov.f64 d4, d2
-; CHECK-NEXT:    vstrw.32 q1, [sp, #64] @ 16-byte Spill
-; CHECK-NEXT:    vmovx.f16 s28, s4
-; CHECK-NEXT:    vins.f16 s8, s24
-; CHECK-NEXT:    vmov.f32 s17, s1
-; CHECK-NEXT:    vmov.16 q2[4], r2
-; CHECK-NEXT:    vmov.f32 s11, s5
-; CHECK-NEXT:    vins.f16 s11, s25
-; CHECK-NEXT:    vmov.f32 s18, s2
-; CHECK-NEXT:    vmov.f32 s9, s4
+; CHECK-NEXT:    vmov.f32 s26, s11
+; CHECK-NEXT:    vldrw.u32 q2, [r0, #64]
+; CHECK-NEXT:    vins.f16 s25, s0
+; CHECK-NEXT:    vmovx.f16 s0, s26
+; CHECK-NEXT:    vins.f16 s18, s0
+; CHECK-NEXT:    vmovx.f16 s0, s10
+; CHECK-NEXT:    vins.f16 s12, s0
+; CHECK-NEXT:    vmovx.f16 s0, s11
+; CHECK-NEXT:    vins.f16 s13, s23
+; CHECK-NEXT:    vmov.f32 s26, s18
+; CHECK-NEXT:    vmovx.f16 s15, s23
+; CHECK-NEXT:    vmov.f32 s17, s25
+; CHECK-NEXT:    vins.f16 s15, s0
+; CHECK-NEXT:    vmov.f32 s1, s31
+; CHECK-NEXT:    vmov.f32 s14, s11
+; CHECK-NEXT:    vmovx.f16 s20, s13
+; CHECK-NEXT:    vmov.f32 s2, s31
+; CHECK-NEXT:    vins.f16 s1, s20
+; CHECK-NEXT:    vmovx.f16 s20, s2
+; CHECK-NEXT:    vins.f16 s14, s20
+; CHECK-NEXT:    vmov q5, q7
+; CHECK-NEXT:    vmov.f32 s2, s14
+; CHECK-NEXT:    vstrw.32 q5, [sp] @ 16-byte Spill
+; CHECK-NEXT:    vmov.f32 s18, s26
+; CHECK-NEXT:    vmov q6, q5
 ; CHECK-NEXT:    vstrw.32 q4, [sp, #16] @ 16-byte Spill
-; CHECK-NEXT:    vmov.f32 s5, s20
-; CHECK-NEXT:    vldrw.u32 q4, [sp, #32] @ 16-byte Reload
-; CHECK-NEXT:    vmov.f32 s6, s20
-; CHECK-NEXT:    vins.f16 s5, s28
-; CHECK-NEXT:    vmovx.f16 s28, s6
-; CHECK-NEXT:    vins.f16 s10, s28
-; CHECK-NEXT:    vmov.f64 d14, d8
-; CHECK-NEXT:    vmov.f32 s6, s10
-; CHECK-NEXT:    vmov.f32 s9, s5
-; CHECK-NEXT:    vmov.f32 s0, s17
-; CHECK-NEXT:    vmov.f32 s10, s6
-; CHECK-NEXT:    vldrw.u32 q1, [sp, #48] @ 16-byte Reload
-; CHECK-NEXT:    vstrw.32 q2, [r1]
-; CHECK-NEXT:    vins.f16 s28, s4
-; CHECK-NEXT:    vmov.u16 r0, q1[1]
-; CHECK-NEXT:    vins.f16 s0, s5
-; CHECK-NEXT:    vmov.16 q7[4], r0
-; CHECK-NEXT:    vmov.f32 s31, s0
-; CHECK-NEXT:    vmovx.f16 s4, s16
-; CHECK-NEXT:    vmov.f32 s1, s12
-; CHECK-NEXT:    vmov.f32 s2, s12
-; CHECK-NEXT:    vins.f16 s1, s4
-; CHECK-NEXT:    vmov.f32 s29, s16
-; CHECK-NEXT:    vmovx.f16 s4, s2
+; CHECK-NEXT:    vldrw.u32 q4, [sp, #48] @ 16-byte Reload
+; CHECK-NEXT:    vmov.f32 s13, s1
+; CHECK-NEXT:    vins.f16 s28, s16
+; CHECK-NEXT:    vmov.u16 r0, q4[1]
+; CHECK-NEXT:    vmov.f32 s14, s2
+; CHECK-NEXT:    vmov.f32 s0, s21
+; CHECK-NEXT:    vmov q5, q7
+; CHECK-NEXT:    vins.f16 s0, s17
+; CHECK-NEXT:    vmov.16 q5[4], r0
+; CHECK-NEXT:    vmov.f32 s23, s0
 ; CHECK-NEXT:    vldrw.u32 q4, [sp, #64] @ 16-byte Reload
-; CHECK-NEXT:    vins.f16 s30, s4
-; CHECK-NEXT:    vmovx.f16 s4, s22
+; CHECK-NEXT:    vmov.f32 s29, s24
+; CHECK-NEXT:    vstrw.32 q3, [r1, #32]
+; CHECK-NEXT:    vmov.f32 s30, s22
+; CHECK-NEXT:    vmov.u16 r0, q4[1]
+; CHECK-NEXT:    vmov.f32 s1, s8
+; CHECK-NEXT:    vmov.f32 s2, s8
+; CHECK-NEXT:    vmov.f32 s8, s9
+; CHECK-NEXT:    vmov.f32 s31, s23
+; CHECK-NEXT:    vmovx.f16 s20, s29
+; CHECK-NEXT:    vins.f16 s1, s20
+; CHECK-NEXT:    vmovx.f16 s20, s2
+; CHECK-NEXT:    vins.f16 s30, s20
 ; CHECK-NEXT:    vmov.f32 s2, s30
 ; CHECK-NEXT:    vmov.f32 s29, s1
-; CHECK-NEXT:    vmov.f32 s12, s13
 ; CHECK-NEXT:    vmov.f32 s30, s2
-; CHECK-NEXT:    vmovx.f16 s0, s26
-; CHECK-NEXT:    vins.f16 s0, s4
-; CHECK-NEXT:    vmovx.f16 s4, s23
-; CHECK-NEXT:    vins.f16 s1, s27
-; CHECK-NEXT:    vstrw.32 q7, [r1, #48]
-; CHECK-NEXT:    vmovx.f16 s3, s27
-; CHECK-NEXT:    vins.f16 s3, s4
-; CHECK-NEXT:    vmov.f32 s5, s19
-; CHECK-NEXT:    vmov.f32 s2, s23
-; CHECK-NEXT:    vmovx.f16 s24, s1
-; CHECK-NEXT:    vmov.f32 s6, s19
+; CHECK-NEXT:    vldrw.u32 q0, [sp, #32] @ 16-byte Reload
+; CHECK-NEXT:    vstrw.32 q7, [r1]
+; CHECK-NEXT:    vmov.f64 d12, d0
+; CHECK-NEXT:    vins.f16 s24, s16
+; CHECK-NEXT:    vmov.f32 s0, s1
+; CHECK-NEXT:    vmov q5, q6
+; CHECK-NEXT:    vins.f16 s0, s17
+; CHECK-NEXT:    vmov.16 q5[4], r0
 ; CHECK-NEXT:    vldrw.u32 q4, [sp, #32] @ 16-byte Reload
-; CHECK-NEXT:    vins.f16 s5, s24
-; CHECK-NEXT:    vmovx.f16 s24, s6
-; CHECK-NEXT:    vins.f16 s2, s24
-; CHECK-NEXT:    vmovx.f16 s24, s17
-; CHECK-NEXT:    vmov.f32 s6, s2
-; CHECK-NEXT:    vins.f16 s12, s24
-; CHECK-NEXT:    vmovx.f16 s24, s18
-; CHECK-NEXT:    vmov.f32 s1, s5
-; CHECK-NEXT:    vins.f16 s14, s24
-; CHECK-NEXT:    vldrw.u32 q6, [sp, #48] @ 16-byte Reload
-; CHECK-NEXT:    vmov.f32 s15, s14
-; CHECK-NEXT:    vmov.f32 s14, s18
-; CHECK-NEXT:    vmovx.f16 s16, s13
-; CHECK-NEXT:    vrev32.16 q6, q6
-; CHECK-NEXT:    vmov.f32 s20, s21
-; CHECK-NEXT:    vins.f16 s25, s16
-; CHECK-NEXT:    vmovx.f16 s16, s26
-; CHECK-NEXT:    vins.f16 s14, s16
-; CHECK-NEXT:    vldrw.u32 q4, [sp, #64] @ 16-byte Reload
-; CHECK-NEXT:    vmov.f32 s2, s6
-; CHECK-NEXT:    vmovx.f16 s4, s17
-; CHECK-NEXT:    vmov.f32 s26, s14
-; CHECK-NEXT:    vins.f16 s20, s4
-; CHECK-NEXT:    vmovx.f16 s4, s18
-; CHECK-NEXT:    vins.f16 s22, s4
-; CHECK-NEXT:    vldrw.u32 q1, [sp] @ 16-byte Reload
-; CHECK-NEXT:    vmov.f32 s23, s22
-; CHECK-NEXT:    vstrw.32 q0, [r1, #32]
-; CHECK-NEXT:    vmov.f32 s22, s18
-; CHECK-NEXT:    vmovx.f16 s16, s21
-; CHECK-NEXT:    vrev32.16 q1, q1
-; CHECK-NEXT:    vmov.f32 s13, s25
-; CHECK-NEXT:    vins.f16 s5, s16
-; CHECK-NEXT:    vmovx.f16 s16, s6
-; CHECK-NEXT:    vins.f16 s22, s16
-; CHECK-NEXT:    vldrw.u32 q0, [sp, #16] @ 16-byte Reload
+; CHECK-NEXT:    vmov.f32 s23, s0
+; CHECK-NEXT:    vmov.f32 s25, s16
+; CHECK-NEXT:    vmov.f32 s26, s22
+; CHECK-NEXT:    vmov.f32 s1, s4
+; CHECK-NEXT:    vmov.f32 s2, s4
+; CHECK-NEXT:    vmov.f32 s4, s5
+; CHECK-NEXT:    vmov.f32 s27, s23
+; CHECK-NEXT:    vmovx.f16 s20, s25
+; CHECK-NEXT:    vins.f16 s1, s20
+; CHECK-NEXT:    vmovx.f16 s20, s2
+; CHECK-NEXT:    vins.f16 s26, s20
+; CHECK-NEXT:    vmovx.f16 s20, s17
+; CHECK-NEXT:    vmov.f32 s2, s26
+; CHECK-NEXT:    vins.f16 s4, s20
+; CHECK-NEXT:    vmovx.f16 s20, s18
+; CHECK-NEXT:    vmov.f32 s25, s1
+; CHECK-NEXT:    vins.f16 s6, s20
+; CHECK-NEXT:    vldrw.u32 q5, [sp, #64] @ 16-byte Reload
+; CHECK-NEXT:    vmov.f32 s7, s6
+; CHECK-NEXT:    vmov.f32 s6, s18
+; CHECK-NEXT:    vmovx.f16 s16, s5
+; CHECK-NEXT:    vrev32.16 q5, q5
+; CHECK-NEXT:    vins.f16 s21, s16
+; CHECK-NEXT:    vmovx.f16 s16, s22
+; CHECK-NEXT:    vins.f16 s6, s16
+; CHECK-NEXT:    vldrw.u32 q4, [sp] @ 16-byte Reload
+; CHECK-NEXT:    vmov.f32 s26, s2
+; CHECK-NEXT:    vmovx.f16 s0, s17
+; CHECK-NEXT:    vmov.f32 s22, s6
+; CHECK-NEXT:    vins.f16 s8, s0
+; CHECK-NEXT:    vmovx.f16 s0, s18
+; CHECK-NEXT:    vins.f16 s10, s0
+; CHECK-NEXT:    vldrw.u32 q0, [sp, #48] @ 16-byte Reload
+; CHECK-NEXT:    vmov.f32 s11, s10
+; CHECK-NEXT:    vstrw.32 q6, [r1, #48]
+; CHECK-NEXT:    vmov.f32 s10, s18
+; CHECK-NEXT:    vmovx.f16 s16, s9
+; CHECK-NEXT:    vrev32.16 q0, q0
+; CHECK-NEXT:    vmov.f32 s5, s21
+; CHECK-NEXT:    vins.f16 s1, s16
+; CHECK-NEXT:    vmovx.f16 s16, s2
+; CHECK-NEXT:    vins.f16 s10, s16
 ; CHECK-NEXT:    vmov.f32 s6, s22
-; CHECK-NEXT:    vmov.f32 s21, s5
+; CHECK-NEXT:    vmov.f32 s2, s10
+; CHECK-NEXT:    vstrw.32 q1, [r1, #64]
+; CHECK-NEXT:    vmov.f32 s9, s1
+; CHECK-NEXT:    vmov.f32 s10, s2
+; CHECK-NEXT:    vldrw.u32 q0, [sp, #16] @ 16-byte Reload
+; CHECK-NEXT:    vstrw.32 q2, [r1, #16]
 ; CHECK-NEXT:    vstrw.32 q0, [r1, #80]
-; CHECK-NEXT:    vmov.f32 s14, s26
-; CHECK-NEXT:    vstrw.32 q3, [r1, #64]
-; CHECK-NEXT:    vmov.f32 s22, s6
-; CHECK-NEXT:    vstrw.32 q5, [r1, #16]
 ; CHECK-NEXT:    add sp, #80
 ; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13, d14, d15}
 ; CHECK-NEXT:    bx lr
 entry:
   %s1 = getelementptr <16 x i16>, <16 x i16>* %src, i32 0
   %l1 = load <16 x i16>, <16 x i16>* %s1, align 4
   %s2 = getelementptr <16 x i16>, <16 x i16>* %src, i32 1
   %l2 = load <16 x i16>, <16 x i16>* %s2, align 4
   %s3 = getelementptr <16 x i16>, <16 x i16>* %src, i32 2
   %l3 = load <16 x i16>, <16 x i16>* %s3, align 4
   %t1 = shufflevector <16 x i16> %l1, <16 x i16> %l2, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
   %t2 = shufflevector <16 x i16> %l3, <16 x i16> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
   %s = shufflevector <32 x i16> %t1, <32 x i16> %t2, <48 x i32> <i32 0, i32 16, i32 32, i32 1, i32 17, i32 33, i32 2, i32 18, i32 34, i32 3, i32 19, i32 35, i32 4, i32 20, i32 36, i32 5, i32 21, i32 37, i32 6, i32 22, i32 38, i32 7, i32 23, i32 39, i32 8, i32 24, i32 40, i32 9, i32 25, i32 41, i32 10, i32 26, i32 42, i32 11, i32 27, i32 43, i32 12, i32 28, i32 44, i32 13, i32 29, i32 45, i32 14, i32 30, i32 46, i32 15, i32 31, i32 47>
   store <48 x i16> %s, <48 x i16> *%dst
   ret void
 }
 
 ; i8
 
 define void @vst3_v2i8(<2 x i8> *%src, <6 x i8> *%dst) {
 ; CHECK-LABEL: vst3_v2i8:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .save {r4, r5, r6, lr}
-; CHECK-NEXT:    push {r4, r5, r6, lr}
-; CHECK-NEXT:    .pad #16
-; CHECK-NEXT:    sub sp, #16
-; CHECK-NEXT:    ldrb r2, [r0]
-; CHECK-NEXT:    movs r6, #0
-; CHECK-NEXT:    ldrb r3, [r0, #1]
-; CHECK-NEXT:    ldrb.w r12, [r0, #2]
-; CHECK-NEXT:    vmov q0[2], q0[0], r2, r3
-; CHECK-NEXT:    ldrb.w lr, [r0, #3]
-; CHECK-NEXT:    vmov r4, s0
-; CHECK-NEXT:    ldrb r5, [r0, #5]
-; CHECK-NEXT:    vmov.16 q0[0], r4
+; CHECK-NEXT:    .save {r4, r5, r6, r7, lr}
+; CHECK-NEXT:    push {r4, r5, r6, r7, lr}
+; CHECK-NEXT:    .pad #20
+; CHECK-NEXT:    sub sp, #20
+; CHECK-NEXT:    ldrb r3, [r0]
+; CHECK-NEXT:    mov.w r12, #0
+; CHECK-NEXT:    ldrb r4, [r0, #2]
+; CHECK-NEXT:    mov r7, sp
+; CHECK-NEXT:    vmov.16 q0[0], r3
+; CHECK-NEXT:    ldrb r2, [r0, #1]
+; CHECK-NEXT:    ldrb r5, [r0, #3]
+; CHECK-NEXT:    vmov.16 q0[1], r4
+; CHECK-NEXT:    ldrb r6, [r0, #5]
 ; CHECK-NEXT:    ldrb r0, [r0, #4]
-; CHECK-NEXT:    vmov.16 q0[1], r12
-; CHECK-NEXT:    mov r2, sp
 ; CHECK-NEXT:    vmov.16 q0[2], r0
 ; CHECK-NEXT:    add r0, sp, #8
-; CHECK-NEXT:    vmov.16 q0[3], r3
-; CHECK-NEXT:    vmov.16 q0[4], lr
-; CHECK-NEXT:    vmov.16 q0[5], r5
-; CHECK-NEXT:    vmov.16 q0[6], r6
-; CHECK-NEXT:    vmov.16 q0[7], r6
-; CHECK-NEXT:    vstrb.16 q0, [r2]
+; CHECK-NEXT:    vmov.16 q0[3], r2
+; CHECK-NEXT:    vmov.16 q0[4], r5
+; CHECK-NEXT:    vmov.16 q0[5], r6
+; CHECK-NEXT:    vmov.16 q0[6], r12
+; CHECK-NEXT:    vmov.16 q0[7], r12
+; CHECK-NEXT:    vstrb.16 q0, [r7]
 ; CHECK-NEXT:    vstrb.16 q0, [r0]
 ; CHECK-NEXT:    vldrh.u32 q0, [r0]
 ; CHECK-NEXT:    ldr r2, [sp]
 ; CHECK-NEXT:    str r2, [r1]
 ; CHECK-NEXT:    vmov r0, s2
 ; CHECK-NEXT:    strh r0, [r1, #4]
-; CHECK-NEXT:    add sp, #16
-; CHECK-NEXT:    pop {r4, r5, r6, pc}
+; CHECK-NEXT:    add sp, #20
+; CHECK-NEXT:    pop {r4, r5, r6, r7, pc}
 entry:
   %s1 = getelementptr <2 x i8>, <2 x i8>* %src, i32 0
   %l1 = load <2 x i8>, <2 x i8>* %s1, align 4
   %s2 = getelementptr <2 x i8>, <2 x i8>* %src, i32 1
   %l2 = load <2 x i8>, <2 x i8>* %s2, align 4
   %s3 = getelementptr <2 x i8>, <2 x i8>* %src, i32 2
   %l3 = load <2 x i8>, <2 x i8>* %s3, align 4
   %t1 = shufflevector <2 x i8> %l1, <2 x i8> %l2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
entry:
   %s = shufflevector <16 x i8> %t1, <16 x i8> %t2, <24 x i32> <i32 0, i32 8, i32 16, i32 1, i32 9, i32 17, i32 2, i32 10, i32 18, i32 3, i32 11, i32 19, i32 4, i32 12, i32 20, i32 5, i32 13, i32 21, i32 6, i32 14, i32 22, i32 7, i32 15, i32 23>
   store <24 x i8> %s, <24 x i8> *%dst
   ret void
 }
 
 define void @vst3_v16i8(<16 x i8> *%src, <48 x i8> *%dst) {
 ; CHECK-LABEL: vst3_v16i8:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13}
-; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13}
-; CHECK-NEXT:    vldrw.u32 q3, [r0]
-; CHECK-NEXT:    vldrw.u32 q2, [r0, #16]
+; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13, d14, d15}
+; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13, d14, d15}
 ; CHECK-NEXT:    vldrw.u32 q1, [r0, #32]
+; CHECK-NEXT:    vldrw.u32 q3, [r0]
+; CHECK-NEXT:    vmov.u8 r2, q1[0]
 ; CHECK-NEXT:    vmov.u8 r3, q3[0]
-; CHECK-NEXT:    vmov.u8 r0, q2[0]
+; CHECK-NEXT:    vmov.8 q4[2], r2
+; CHECK-NEXT:    vmov.u8 r2, q1[2]
+; CHECK-NEXT:    vmov q0, q4
+; CHECK-NEXT:    vmov.f32 s17, s4
+; CHECK-NEXT:    vmov.8 q0[8], r2
+; CHECK-NEXT:    vmov.u8 r2, q1[3]
+; CHECK-NEXT:    vmov.8 q0[11], r2
+; CHECK-NEXT:    vmov.u8 r2, q1[4]
+; CHECK-NEXT:    vmov q2, q0
+; CHECK-NEXT:    vmov.f32 s18, s2
+; CHECK-NEXT:    vmov.8 q2[14], r2
 ; CHECK-NEXT:    vmov.8 q5[0], r3
-; CHECK-NEXT:    vmov.u8 r2, q1[0]
+; CHECK-NEXT:    vmov.f32 s19, s11
+; CHECK-NEXT:    vldrw.u32 q2, [r0, #16]
+; CHECK-NEXT:    vmov.u8 r2, q4[2]
+; CHECK-NEXT:    vmov.u8 r0, q2[0]
 ; CHECK-NEXT:    vmov.8 q5[1], r0
 ; CHECK-NEXT:    vmov.u8 r0, q3[1]
 ; CHECK-NEXT:    vmov.8 q5[3], r0
 ; CHECK-NEXT:    vmov.u8 r0, q2[1]
 ; CHECK-NEXT:    vmov.8 q5[4], r0
 ; CHECK-NEXT:    vmov.u8 r0, q3[2]
 ; CHECK-NEXT:    vmov.8 q5[6], r0
 ; CHECK-NEXT:    vmov.u8 r0, q2[2]
 ; CHECK-NEXT:    vmov.8 q5[7], r0
 ; CHECK-NEXT:    vmov.u8 r0, q3[3]
 ; CHECK-NEXT:    vmov.8 q5[9], r0
 ; CHECK-NEXT:    vmov.u8 r0, q2[3]
 ; CHECK-NEXT:    vmov.8 q5[10], r0
 ; CHECK-NEXT:    vmov.u8 r0, q3[4]
-; CHECK-NEXT:    vmov.8 q4[2], r2
-; CHECK-NEXT:    vmov.u8 r2, q1[2]
 ; CHECK-NEXT:    vmov.8 q5[12], r0
 ; CHECK-NEXT:    vmov.u8 r0, q2[4]
-; CHECK-NEXT:    vmov.8 q4[8], r2
-; CHECK-NEXT:    vmov.u8 r2, q1[3]
 ; CHECK-NEXT:    vmov.8 q5[13], r0
 ; CHECK-NEXT:    vmov.u8 r0, q3[5]
 ; CHECK-NEXT:    vmov.8 q5[15], r0
-; CHECK-NEXT:    vmov.8 q4[11], r2
-; CHECK-NEXT:    vmov.u8 r2, q1[4]
 ; CHECK-NEXT:    vmov.u8 r0, q5[0]
-; CHECK-NEXT:    vmov.8 q4[14], r2
 ; CHECK-NEXT:    vmov.8 q0[0], r0
-; CHECK-NEXT:    vmov.f32 s17, s4
 ; CHECK-NEXT:    vmov.u8 r0, q5[1]
 ; CHECK-NEXT:    vmov.8 q0[1], r0
-; CHECK-NEXT:    vmov.u8 r2, q4[2]
-; CHECK-NEXT:    vmov.8 q0[2], r2
 ; CHECK-NEXT:    vmov.u8 r0, q5[3]
+; CHECK-NEXT:    vmov.8 q0[2], r2
 ; CHECK-NEXT:    vmov.8 q0[3], r0
 ; CHECK-NEXT:    vmov.u8 r0, q5[4]
 ; CHECK-NEXT:    vmov.8 q0[4], r0
 ; CHECK-NEXT:    vmov.u8 r0, q4[5]
 ; CHECK-NEXT:    vmov.8 q0[5], r0
 ; CHECK-NEXT:    vmov.u8 r0, q5[6]
 ; CHECK-NEXT:    vmov.8 q0[6], r0
 ; CHECK-NEXT:    vmov.u8 r0, q5[7]

 ; CHECK-NEXT:    vstrw.32 q0, [r1]
 ; CHECK-NEXT:    vmov.u8 r0, q5[0]
 ; CHECK-NEXT:    vmov.8 q4[0], r0
 ; CHECK-NEXT:    vmov.u8 r0, q5[1]
 ; CHECK-NEXT:    vmov.8 q4[1], r0
 ; CHECK-NEXT:    vmov.u8 r0, q3[7]
 ; CHECK-NEXT:    vmov.8 q6[5], r0
 ; CHECK-NEXT:    vmov.u8 r0, q3[8]
-; CHECK-NEXT:    vmov.8 q6[8], r0
-; CHECK-NEXT:    vmov.u8 r0, q3[9]
-; CHECK-NEXT:    vmov.8 q6[11], r0
+; CHECK-NEXT:    vmov q7, q6
 ; CHECK-NEXT:    vmov.f32 s24, s13
+; CHECK-NEXT:    vmov.8 q7[8], r0
+; CHECK-NEXT:    vmov.u8 r0, q3[9]
+; CHECK-NEXT:    vmov.8 q7[11], r0
+; CHECK-NEXT:    vmov.f32 s26, s30
 ; CHECK-NEXT:    vmov.f32 s27, s14
 ; CHECK-NEXT:    vmov.u8 r0, q6[2]
 ; CHECK-NEXT:    vmov.8 q4[2], r0
 ; CHECK-NEXT:    vmov.u8 r0, q5[3]
 ; CHECK-NEXT:    vmov.8 q4[3], r0
 ; CHECK-NEXT:    vmov.u8 r0, q5[4]
 ; CHECK-NEXT:    vmov.8 q4[4], r0
 ; CHECK-NEXT:    vmov.u8 r0, q6[5]

 ; CHECK-NEXT:    vmov.8 q1[12], r0
 ; CHECK-NEXT:    vmov.u8 r0, q5[13]
 ; CHECK-NEXT:    vmov.8 q1[13], r0
 ; CHECK-NEXT:    vmov.u8 r0, q3[14]
 ; CHECK-NEXT:    vmov.8 q1[14], r0
 ; CHECK-NEXT:    vmov.u8 r0, q5[15]
 ; CHECK-NEXT:    vmov.8 q1[15], r0
 ; CHECK-NEXT:    vstrw.32 q1, [r1, #32]
-; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13}
+; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13, d14, d15}
 ; CHECK-NEXT:    bx lr
 entry:
   %s1 = getelementptr <16 x i8>, <16 x i8>* %src, i32 0
   %l1 = load <16 x i8>, <16 x i8>* %s1, align 4
   %s2 = getelementptr <16 x i8>, <16 x i8>* %src, i32 1
   %l2 = load <16 x i8>, <16 x i8>* %s2, align 4
   %s3 = getelementptr <16 x i8>, <16 x i8>* %src, i32 2
   %l3 = load <16 x i8>, <16 x i8>* %s3, align 4

 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .save {r7, lr}
 ; CHECK-NEXT:    push {r7, lr}
 ; CHECK-NEXT:    .vsave {d8, d9}
 ; CHECK-NEXT:    vpush {d8, d9}
 ; CHECK-NEXT:    ldrd r2, r12, [r0]
 ; CHECK-NEXT:    ldrd r3, lr, [r0, #8]
 ; CHECK-NEXT:    vmov.32 q0[0], r2
-; CHECK-NEXT:    ldrd r2, r0, [r0, #16]
+; CHECK-NEXT:    ldr r2, [r0, #16]
+; CHECK-NEXT:    vmov q3, q0
 ; CHECK-NEXT:    vmov.32 q1[0], r3
-; CHECK-NEXT:    vmov.32 q0[1], r12
-; CHECK-NEXT:    vmov.32 q1[1], lr
+; CHECK-NEXT:    vmov.32 q3[1], r12
+; CHECK-NEXT:    vmov q2, q1
+; CHECK-NEXT:    vmov.f32 s1, s13
+; CHECK-NEXT:    vmov.32 q2[1], lr
 ; CHECK-NEXT:    vmov.f32 s2, s4
-; CHECK-NEXT:    vmov.f32 s3, s5
-; CHECK-NEXT:    vmovx.f16 s10, s0
+; CHECK-NEXT:    ldr r0, [r0, #20]
+; CHECK-NEXT:    vmov.f32 s3, s9
+; CHECK-NEXT:    vmov.32 q1[0], r2
 ; CHECK-NEXT:    vmov.f32 s8, s1
+; CHECK-NEXT:    vmovx.f16 s10, s0
 ; CHECK-NEXT:    vins.f16 s0, s2
-; CHECK-NEXT:    vins.f16 s8, s5
-; CHECK-NEXT:    vmov.32 q1[0], r2
-; CHECK-NEXT:    vmov.32 q1[1], r0
-; CHECK-NEXT:    vmovx.f16 s2, s2
+; CHECK-NEXT:    vmov.f32 s14, s4
 ; CHECK-NEXT:    vmovx.f16 s12, s4
-; CHECK-NEXT:    vins.f16 s4, s10
+; CHECK-NEXT:    vmovx.f16 s2, s2
+; CHECK-NEXT:    vmov.32 q1[1], r0
+; CHECK-NEXT:    vins.f16 s14, s10
 ; CHECK-NEXT:    vins.f16 s2, s12
-; CHECK-NEXT:    vmovx.f16 s10, s1
-; CHECK-NEXT:    vmovx.f16 s12, s5
+; CHECK-NEXT:    vmovx.f16 s10, s5
 ; CHECK-NEXT:    vmovx.f16 s17, s3
+; CHECK-NEXT:    vins.f16 s8, s9
+; CHECK-NEXT:    vins.f16 s17, s10
+; CHECK-NEXT:    vmovx.f16 s10, s1
 ; CHECK-NEXT:    vins.f16 s5, s10
-; CHECK-NEXT:    vins.f16 s17, s12
+; CHECK-NEXT:    vmov.f32 s1, s14
 ; CHECK-NEXT:    vmov.f32 s16, s5
 ; CHECK-NEXT:    vmov r2, s17
-; CHECK-NEXT:    vmov.f32 s1, s4
 ; CHECK-NEXT:    vmov.f32 s3, s8
 ; CHECK-NEXT:    vstrw.32 q0, [r1]
 ; CHECK-NEXT:    vmov r0, s16
 ; CHECK-NEXT:    strd r0, r2, [r1, #16]
 ; CHECK-NEXT:    vpop {d8, d9}
 ; CHECK-NEXT:    pop {r7, pc}
 entry:
   %s1 = getelementptr <4 x half>, <4 x half>* %src, i32 0
   %l1 = load <4 x half>, <4 x half>* %s1, align 4
   %s2 = getelementptr <4 x half>, <4 x half>* %src, i32 1
   %l2 = load <4 x half>, <4 x half>* %s2, align 4
   %s3 = getelementptr <4 x half>, <4 x half>* %src, i32 2
   %l3 = load <4 x half>, <4 x half>* %s3, align 4
   %t1 = shufflevector <4 x half> %l1, <4 x half> %l2, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
   %t2 = shufflevector <4 x half> %l3, <4 x half> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef>
   %s = shufflevector <8 x half> %t1, <8 x half> %t2, <12 x i32> <i32 0, i32 4, i32 8, i32 1, i32 5, i32 9, i32 2, i32 6, i32 10, i32 3, i32 7, i32 11>
   store <12 x half> %s, <12 x half> *%dst
   ret void
 }
 
 define void @vst3_v8f16(<8 x half> *%src, <24 x half> *%dst) {
 ; CHECK-LABEL: vst3_v8f16:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13, d14}
-; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13, d14}
-; CHECK-NEXT:    vldrw.u32 q3, [r0]
-; CHECK-NEXT:    vldrw.u32 q5, [r0, #16]
-; CHECK-NEXT:    vmov.f64 d0, d6
-; CHECK-NEXT:    vmovx.f16 s6, s20
-; CHECK-NEXT:    vmovx.f16 s8, s12
-; CHECK-NEXT:    vmov.f32 s4, s13
-; CHECK-NEXT:    vins.f16 s0, s20
-; CHECK-NEXT:    vmov r2, s6
-; CHECK-NEXT:    vins.f16 s4, s21
-; CHECK-NEXT:    vmov.16 q0[4], r2
-; CHECK-NEXT:    vmov.f32 s3, s4
+; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13, d14, d15}
+; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13, d14, d15}
+; CHECK-NEXT:    vldrw.u32 q4, [r0, #16]
 ; CHECK-NEXT:    vldrw.u32 q1, [r0, #32]
-; CHECK-NEXT:    vmov.f32 s1, s12
-; CHECK-NEXT:    vmov.f32 s17, s4
-; CHECK-NEXT:    vmovx.f16 s24, s7
-; CHECK-NEXT:    vmov.f32 s18, s4
-; CHECK-NEXT:    vins.f16 s17, s8
-; CHECK-NEXT:    vmovx.f16 s8, s18
+; CHECK-NEXT:    vldrw.u32 q7, [r0]
+; CHECK-NEXT:    vmovx.f16 s8, s6
+; CHECK-NEXT:    vmovx.f16 s0, s18
+; CHECK-NEXT:    vins.f16 s0, s8
+; CHECK-NEXT:    vmovx.f16 s8, s7
+; CHECK-NEXT:    vmovx.f16 s3, s19
+; CHECK-NEXT:    vmov.f32 s13, s31
+; CHECK-NEXT:    vins.f16 s3, s8
+; CHECK-NEXT:    vmov.f32 s14, s31
+; CHECK-NEXT:    vins.f16 s1, s19
+; CHECK-NEXT:    vmovx.f16 s20, s29
+; CHECK-NEXT:    vmov.f32 s2, s7
+; CHECK-NEXT:    vmovx.f16 s8, s1
+; CHECK-NEXT:    vins.f16 s13, s8
+; CHECK-NEXT:    vmovx.f16 s8, s14
 ; CHECK-NEXT:    vins.f16 s2, s8
-; CHECK-NEXT:    vmovx.f16 s11, s23
-; CHECK-NEXT:    vins.f16 s11, s24
-; CHECK-NEXT:    vmovx.f16 s24, s6
-; CHECK-NEXT:    vmovx.f16 s8, s22
-; CHECK-NEXT:    vmov.f32 s18, s2
-; CHECK-NEXT:    vins.f16 s8, s24
-; CHECK-NEXT:    vmov.f32 s25, s15
-; CHECK-NEXT:    vins.f16 s9, s23
-; CHECK-NEXT:    vmov.f32 s26, s15
-; CHECK-NEXT:    vmov.f32 s10, s7
-; CHECK-NEXT:    vmovx.f16 s28, s9
-; CHECK-NEXT:    vins.f16 s25, s28
-; CHECK-NEXT:    vmovx.f16 s28, s26
-; CHECK-NEXT:    vins.f16 s10, s28
-; CHECK-NEXT:    vmovx.f16 s28, s13
-; CHECK-NEXT:    vmov.f32 s4, s5
-; CHECK-NEXT:    vrev32.16 q5, q5
-; CHECK-NEXT:    vins.f16 s4, s28
-; CHECK-NEXT:    vmovx.f16 s28, s14
-; CHECK-NEXT:    vins.f16 s6, s28
+; CHECK-NEXT:    vmov.f32 s8, s5
+; CHECK-NEXT:    vins.f16 s8, s20
+; CHECK-NEXT:    vmovx.f16 s20, s30
+; CHECK-NEXT:    vins.f16 s6, s20
+; CHECK-NEXT:    vrev32.16 q6, q4
+; CHECK-NEXT:    vmov.f32 s11, s6
+; CHECK-NEXT:    vmov.f32 s9, s5
+; CHECK-NEXT:    vmov.f32 s10, s30
+; CHECK-NEXT:    vmovx.f16 s20, s9
+; CHECK-NEXT:    vins.f16 s25, s20
+; CHECK-NEXT:    vmovx.f16 s20, s26
+; CHECK-NEXT:    vins.f16 s10, s20
+; CHECK-NEXT:    vmov.f64 d10, d14
+; CHECK-NEXT:    vins.f16 s29, s17
+; CHECK-NEXT:    vins.f16 s20, s16
+; CHECK-NEXT:    vmovx.f16 s16, s16
+; CHECK-NEXT:    vmov r0, s16
+; CHECK-NEXT:    vmov q4, q5
+; CHECK-NEXT:    vmov.16 q4[4], r0
+; CHECK-NEXT:    vmov.f32 s21, s28
+; CHECK-NEXT:    vmov.f32 s19, s29
+; CHECK-NEXT:    vmov.f32 s22, s18
+; CHECK-NEXT:    vmov.f32 s14, s2
 ; CHECK-NEXT:    vmov.f32 s26, s10
-; CHECK-NEXT:    vmov.f32 s7, s6
-; CHECK-NEXT:    vmov.f32 s6, s14
-; CHECK-NEXT:    vmovx.f16 s12, s5
-; CHECK-NEXT:    vins.f16 s21, s12
-; CHECK-NEXT:    vmovx.f16 s12, s22
-; CHECK-NEXT:    vins.f16 s6, s12
-; CHECK-NEXT:    vmov.f32 s1, s17
-; CHECK-NEXT:    vmov.f32 s22, s6
+; CHECK-NEXT:    vmov.f32 s1, s13
 ; CHECK-NEXT:    vmov.f32 s9, s25
-; CHECK-NEXT:    vmov.f32 s5, s21
-; CHECK-NEXT:    vmov.f32 s2, s18
+; CHECK-NEXT:    vmov.f32 s23, s19
+; CHECK-NEXT:    vmovx.f16 s28, s21
+; CHECK-NEXT:    vmov.f32 s17, s4
+; CHECK-NEXT:    vmov.f32 s18, s4
+; CHECK-NEXT:    vins.f16 s17, s28
+; CHECK-NEXT:    vmovx.f16 s4, s18
+; CHECK-NEXT:    vins.f16 s22, s4
+; CHECK-NEXT:    vmov.f32 s2, s14
+; CHECK-NEXT:    vmov.f32 s18, s22
+; CHECK-NEXT:    vstrw.32 q0, [r1, #32]
+; CHECK-NEXT:    vmov.f32 s21, s17
 ; CHECK-NEXT:    vmov.f32 s10, s26
-; CHECK-NEXT:    vstrw.32 q0, [r1]
-; CHECK-NEXT:    vstrw.32 q2, [r1, #32]
-; CHECK-NEXT:    vmov.f32 s6, s22
-; CHECK-NEXT:    vstrw.32 q1, [r1, #16]
-; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13, d14}
+; CHECK-NEXT:    vstrw.32 q2, [r1, #16]
+; CHECK-NEXT:    vmov.f32 s22, s18
+; CHECK-NEXT:    vstrw.32 q5, [r1]
+; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13, d14, d15}
 ; CHECK-NEXT:    bx lr
 entry:
   %s1 = getelementptr <8 x half>, <8 x half>* %src, i32 0
   %l1 = load <8 x half>, <8 x half>* %s1, align 4
   %s2 = getelementptr <8 x half>, <8 x half>* %src, i32 1
   %l2 = load <8 x half>, <8 x half>* %s2, align 4
   %s3 = getelementptr <8 x half>, <8 x half>* %src, i32 2
   %l3 = load <8 x half>, <8 x half>* %s3, align 4
   %t1 = shufflevector <8 x half> %l1, <8 x half> %l2, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
   %t2 = shufflevector <8 x half> %l3, <8 x half> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
   %s = shufflevector <16 x half> %t1, <16 x half> %t2, <24 x i32> <i32 0, i32 8, i32 16, i32 1, i32 9, i32 17, i32 2, i32 10, i32 18, i32 3, i32 11, i32 19, i32 4, i32 12, i32 20, i32 5, i32 13, i32 21, i32 6, i32 14, i32 22, i32 7, i32 15, i32 23>
   store <24 x half> %s, <24 x half> *%dst
   ret void
 }
 
 define void @vst3_v16f16(<16 x half> *%src, <48 x half> *%dst) {
 ; CHECK-LABEL: vst3_v16f16:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13, d14, d15}
 ; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13, d14, d15}
-; CHECK-NEXT:    .pad #128
-; CHECK-NEXT:    sub sp, #128
-; CHECK-NEXT:    vldrw.u32 q1, [r0, #32]
-; CHECK-NEXT:    vldrw.u32 q7, [r0, #64]
-; CHECK-NEXT:    vldrw.u32 q5, [r0, #80]
-; CHECK-NEXT:    vmovx.f16 s0, s31
-; CHECK-NEXT:    vmovx.f16 s11, s7
-; CHECK-NEXT:    vins.f16 s11, s0
-; CHECK-NEXT:    vmovx.f16 s0, s30
-; CHECK-NEXT:    vmovx.f16 s8, s6
-; CHECK-NEXT:    vmov q4, q1
-; CHECK-NEXT:    vins.f16 s8, s0
+; CHECK-NEXT:    .pad #144
+; CHECK-NEXT:    sub sp, #144
+; CHECK-NEXT:    vldrw.u32 q2, [r0, #64]
+; CHECK-NEXT:    vldrw.u32 q4, [r0, #32]
+; CHECK-NEXT:    vldrw.u32 q3, [r0]
+; CHECK-NEXT:    vldrw.u32 q5, [r0, #16]
+; CHECK-NEXT:    vmovx.f16 s0, s11
+; CHECK-NEXT:    vmovx.f16 s7, s19
+; CHECK-NEXT:    vins.f16 s7, s0
+; CHECK-NEXT:    vmovx.f16 s0, s10
+; CHECK-NEXT:    vmovx.f16 s4, s18
+; CHECK-NEXT:    vstrw.32 q2, [sp, #128] @ 16-byte Spill
+; CHECK-NEXT:    vins.f16 s4, s0
+; CHECK-NEXT:    vmov q7, q3
+; CHECK-NEXT:    vins.f16 s5, s19
 ; CHECK-NEXT:    vstrw.32 q4, [sp, #48] @ 16-byte Spill
-; CHECK-NEXT:    vins.f16 s9, s7
-; CHECK-NEXT:    vmov.f32 s10, s31
-; CHECK-NEXT:    vmovx.f16 s0, s9
-; CHECK-NEXT:    vmov q3, q2
-; CHECK-NEXT:    vldrw.u32 q2, [r0]
-; CHECK-NEXT:    vmov.f32 s5, s11
-; CHECK-NEXT:    vmov q6, q2
 ; CHECK-NEXT:    vmov.f32 s6, s11
+; CHECK-NEXT:    vmovx.f16 s0, s5
+; CHECK-NEXT:    vmov.f32 s9, s15
+; CHECK-NEXT:    vmov.f32 s10, s15
+; CHECK-NEXT:    vldrw.u32 q3, [r0, #80]
+; CHECK-NEXT:    vins.f16 s9, s0
+; CHECK-NEXT:    vmovx.f16 s0, s10
+; CHECK-NEXT:    vstrw.32 q2, [sp, #96] @ 16-byte Spill
 ; CHECK-NEXT:    vldrw.u32 q2, [r0, #48]
-; CHECK-NEXT:    vins.f16 s5, s0
-; CHECK-NEXT:    vmovx.f16 s0, s6
-; CHECK-NEXT:    vstrw.32 q1, [sp, #64] @ 16-byte Spill
-; CHECK-NEXT:    vldrw.u32 q1, [r0, #16]
-; CHECK-NEXT:    vins.f16 s14, s0
-; CHECK-NEXT:    vmovx.f16 s2, s8
-; CHECK-NEXT:    vstrw.32 q3, [sp, #32] @ 16-byte Spill
-; CHECK-NEXT:    vmov.f64 d6, d2
+; CHECK-NEXT:    vins.f16 s6, s0
+; CHECK-NEXT:    vmovx.f16 s0, s15
 ; CHECK-NEXT:    vstrw.32 q1, [sp, #80] @ 16-byte Spill
-; CHECK-NEXT:    vstrw.32 q6, [sp, #16] @ 16-byte Spill
-; CHECK-NEXT:    vstrw.32 q2, [sp] @ 16-byte Spill
-; CHECK-NEXT:    vmov.f32 s0, s5
-; CHECK-NEXT:    vins.f16 s12, s8
-; CHECK-NEXT:    vmov r2, s2
-; CHECK-NEXT:    vins.f16 s0, s9
-; CHECK-NEXT:    vmov.16 q3[4], r2
+; CHECK-NEXT:    vmovx.f16 s7, s11
+; CHECK-NEXT:    vins.f16 s7, s0
+; CHECK-NEXT:    vmovx.f16 s0, s14
+; CHECK-NEXT:    vmovx.f16 s4, s10
+; CHECK-NEXT:    vmovx.f16 s2, s8
+; CHECK-NEXT:    vins.f16 s4, s0
+; CHECK-NEXT:    vmov r0, s2
+; CHECK-NEXT:    vins.f16 s5, s11
+; CHECK-NEXT:    vstrw.32 q2, [sp, #32] @ 16-byte Spill
+; CHECK-NEXT:    vmov.f32 s6, s15
+; CHECK-NEXT:    vmovx.f16 s0, s5
+; CHECK-NEXT:    vmov q6, q1
+; CHECK-NEXT:    vmov.f32 s5, s23
+; CHECK-NEXT:    vmov.f32 s6, s23
 ; CHECK-NEXT:    vmovx.f16 s2, s16
-; CHECK-NEXT:    vmov.f32 s15, s0
-; CHECK-NEXT:    vmovx.f16 s0, s4
-; CHECK-NEXT:    vmov.f32 s13, s4
-; CHECK-NEXT:    vmov.f32 s5, s20
-; CHECK-NEXT:    vmov.f32 s6, s20
 ; CHECK-NEXT:    vins.f16 s5, s0
 ; CHECK-NEXT:    vmovx.f16 s0, s6
-; CHECK-NEXT:    vstrw.32 q1, [sp, #112] @ 16-byte Spill
+; CHECK-NEXT:    vins.f16 s26, s0
+; CHECK-NEXT:    vmov.f32 s0, s21
+; CHECK-NEXT:    vstrw.32 q6, [sp, #112] @ 16-byte Spill
+; CHECK-NEXT:    vmov.f64 d12, d10
+; CHECK-NEXT:    vstrw.32 q1, [sp, #16] @ 16-byte Spill
+; CHECK-NEXT:    vins.f16 s0, s9
+; CHECK-NEXT:    vins.f16 s24, s8
+; CHECK-NEXT:    vmov q2, q7
 ; CHECK-NEXT:    vmov q1, q6
-; CHECK-NEXT:    vins.f16 s14, s0
-; CHECK-NEXT:    vmov.f32 s0, s5
-; CHECK-NEXT:    vins.f16 s24, s16
+; CHECK-NEXT:    vmov.f32 s25, s20
+; CHECK-NEXT:    vmov.16 q1[4], r0
+; CHECK-NEXT:    vins.f16 s28, s16
+; CHECK-NEXT:    vmov.f32 s7, s0
+; CHECK-NEXT:    vstrw.32 q2, [sp] @ 16-byte Spill
+; CHECK-NEXT:    vmov.f32 s26, s6
 ; CHECK-NEXT:    vmov r0, s2
-; CHECK-NEXT:    vins.f16 s0, s17
-; CHECK-NEXT:    vmov.16 q6[4], r0
-; CHECK-NEXT:    vmov.f32 s27, s0
-; CHECK-NEXT:    vmovx.f16 s0, s4
-; CHECK-NEXT:    vmov.f32 s25, s4
-; CHECK-NEXT:    vmov.f32 s5, s28
-; CHECK-NEXT:    vmov.f32 s6, s28
+; CHECK-NEXT:    vmov.f32 s27, s7
+; CHECK-NEXT:    vmovx.f16 s0, s25
+; CHECK-NEXT:    vmov.f32 s5, s12
+; CHECK-NEXT:    vmov.f32 s6, s12
 ; CHECK-NEXT:    vins.f16 s5, s0
 ; CHECK-NEXT:    vmovx.f16 s0, s6
-; CHECK-NEXT:    vstrw.32 q1, [sp, #96] @ 16-byte Spill
+; CHECK-NEXT:    vstrw.32 q1, [sp, #64] @ 16-byte Spill
+; CHECK-NEXT:    vmov q1, q7
 ; CHECK-NEXT:    vins.f16 s26, s0
-; CHECK-NEXT:    vmovx.f16 s0, s22
-; CHECK-NEXT:    vmovx.f16 s4, s10
-; CHECK-NEXT:    vins.f16 s4, s0
-; CHECK-NEXT:    vmovx.f16 s0, s23
-; CHECK-NEXT:    vmovx.f16 s7, s11
-; CHECK-NEXT:    vmov.f32 s28, s29
-; CHECK-NEXT:    vins.f16 s7, s0
-; CHECK-NEXT:    vins.f16 s5, s11
-; CHECK-NEXT:    vldrw.u32 q2, [sp, #80] @ 16-byte Reload
-; CHECK-NEXT:    vmov.f32 s6, s23
-; CHECK-NEXT:    vmovx.f16 s16, s5
-; CHECK-NEXT:    vmov.f32 s1, s11
-; CHECK-NEXT:    vmov.f32 s2, s11
-; CHECK-NEXT:    vins.f16 s1, s16
-; CHECK-NEXT:    vmovx.f16 s16, s2
-; CHECK-NEXT:    vins.f16 s6, s16
-; CHECK-NEXT:    vmovx.f16 s16, s9
+; CHECK-NEXT:    vmov.f32 s0, s9
+; CHECK-NEXT:    vins.f16 s0, s17
+; CHECK-NEXT:    vmov.16 q1[4], r0
+; CHECK-NEXT:    vmov.f32 s7, s0
+; CHECK-NEXT:    vldrw.u32 q0, [sp, #128] @ 16-byte Reload
+; CHECK-NEXT:    vmov.f32 s29, s8
+; CHECK-NEXT:    vldrw.u32 q2, [sp, #96] @ 16-byte Reload
+; CHECK-NEXT:    vmov.f32 s30, s6
+; CHECK-NEXT:    vmov.f32 s17, s0
+; CHECK-NEXT:    vmov.f32 s18, s0
+; CHECK-NEXT:    vmov.f32 s12, s13
+; CHECK-NEXT:    vmov.f32 s31, s7
+; CHECK-NEXT:    vmovx.f16 s4, s29
+; CHECK-NEXT:    vins.f16 s17, s4
+; CHECK-NEXT:    vmovx.f16 s4, s18
+; CHECK-NEXT:    vins.f16 s30, s4
+; CHECK-NEXT:    vldrw.u32 q1, [sp, #80] @ 16-byte Reload
+; CHECK-NEXT:    vmov.f32 s18, s30
+; CHECK-NEXT:    vmov.f32 s10, s6
+; CHECK-NEXT:    vldrw.u32 q1, [sp, #112] @ 16-byte Reload
+; CHECK-NEXT:    vstrw.32 q2, [sp, #96] @ 16-byte Spill
+; CHECK-NEXT:    vldrw.u32 q2, [sp, #16] @ 16-byte Reload
+; CHECK-NEXT:    vmov.f32 s10, s6
+; CHECK-NEXT:    vmovx.f16 s4, s21
+; CHECK-NEXT:    vins.f16 s12, s4
+; CHECK-NEXT:    vmovx.f16 s4, s22
+; CHECK-NEXT:    vins.f16 s14, s4
+; CHECK-NEXT:    vldrw.u32 q1, [sp, #32] @ 16-byte Reload
+; CHECK-NEXT:    vmov.f32 s15, s14
+; CHECK-NEXT:    vmov.f32 s14, s22
+; CHECK-NEXT:    vrev32.16 q0, q1
+; CHECK-NEXT:    vmovx.f16 s4, s13
+; CHECK-NEXT:    vldrw.u32 q5, [sp, #128] @ 16-byte Reload
+; CHECK-NEXT:    vins.f16 s1, s4
+; CHECK-NEXT:    vmovx.f16 s4, s2
+; CHECK-NEXT:    vstrw.32 q0, [sp, #32] @ 16-byte Spill
+; CHECK-NEXT:    vldrw.u32 q0, [sp] @ 16-byte Reload
+; CHECK-NEXT:    vins.f16 s14, s4
 ; CHECK-NEXT:    vmov.f32 s20, s21
-; CHECK-NEXT:    vins.f16 s20, s16
-; CHECK-NEXT:    vmovx.f16 s16, s10
-; CHECK-NEXT:    vins.f16 s22, s16
-; CHECK-NEXT:    vldrw.u32 q2, [sp] @ 16-byte Reload
-; CHECK-NEXT:    vldrw.u32 q4, [sp, #80] @ 16-byte Reload
-; CHECK-NEXT:    vmov.f32 s23, s22
-; CHECK-NEXT:    vrev32.16 q2, q2
-; CHECK-NEXT:    vmov.f32 s2, s6
-; CHECK-NEXT:    vmov.f32 s22, s18
-; CHECK-NEXT:    vmovx.f16 s16, s21
-; CHECK-NEXT:    vins.f16 s9, s16
-; CHECK-NEXT:    vldrw.u32 q4, [sp, #112] @ 16-byte Reload
-; CHECK-NEXT:    vmov.f32 s18, s14
-; CHECK-NEXT:    vstrw.32 q2, [sp, #80] @ 16-byte Spill
-; CHECK-NEXT:    vstrw.32 q4, [sp, #112] @ 16-byte Spill
-; CHECK-NEXT:    vldrw.u32 q4, [sp, #96] @ 16-byte Reload
-; CHECK-NEXT:    vmov.f32 s18, s26
-; CHECK-NEXT:    vstrw.32 q4, [sp, #96] @ 16-byte Spill
-; CHECK-NEXT:    vmovx.f16 s16, s10
-; CHECK-NEXT:    vins.f16 s22, s16
-; CHECK-NEXT:    vldrw.u32 q4, [sp, #16] @ 16-byte Reload
-; CHECK-NEXT:    vmov.f32 s5, s1
-; CHECK-NEXT:    vmovx.f16 s8, s17
-; CHECK-NEXT:    vmov.f32 s6, s2
-; CHECK-NEXT:    vins.f16 s28, s8
-; CHECK-NEXT:    vmovx.f16 s0, s18
-; CHECK-NEXT:    vins.f16 s30, s0
-; CHECK-NEXT:    vldrw.u32 q0, [sp, #48] @ 16-byte Reload
-; CHECK-NEXT:    vmov.f32 s31, s30
-; CHECK-NEXT:    vldrw.u32 q2, [sp, #32] @ 16-byte Reload
+; CHECK-NEXT:    vmovx.f16 s4, s1
+; CHECK-NEXT:    vmov.f32 s29, s17
+; CHECK-NEXT:    vins.f16 s20, s4
+; CHECK-NEXT:    vmov q1, q2
+; CHECK-NEXT:    vstrw.32 q5, [sp, #128] @ 16-byte Spill
+; CHECK-NEXT:    vldrw.u32 q5, [sp, #112] @ 16-byte Reload
+; CHECK-NEXT:    vmov.f32 s21, s5
+; CHECK-NEXT:    vmov q2, q0
+; CHECK-NEXT:    vldrw.u32 q0, [sp, #128] @ 16-byte Reload
+; CHECK-NEXT:    vmov.f32 s22, s6
+; CHECK-NEXT:    vmovx.f16 s4, s10
+; CHECK-NEXT:    vstrw.32 q5, [sp, #112] @ 16-byte Spill
+; CHECK-NEXT:    vins.f16 s2, s4
 ; CHECK-NEXT:    vmov.f32 s30, s18
-; CHECK-NEXT:    vmovx.f16 s16, s29
-; CHECK-NEXT:    vrev32.16 q0, q0
-; CHECK-NEXT:    vstrw.32 q1, [r1, #80]
-; CHECK-NEXT:    vins.f16 s1, s16
-; CHECK-NEXT:    vmovx.f16 s16, s2
-; CHECK-NEXT:    vins.f16 s30, s16
-; CHECK-NEXT:    vldrw.u32 q4, [sp, #96] @ 16-byte Reload
-; CHECK-NEXT:    vmov.f32 s2, s30
-; CHECK-NEXT:    vmov.f32 s25, s17
-; CHECK-NEXT:    vmov.f32 s26, s18
-; CHECK-NEXT:    vldrw.u32 q4, [sp, #112] @ 16-byte Reload
-; CHECK-NEXT:    vmov.f32 s29, s1
-; CHECK-NEXT:    vstrw.32 q6, [r1]
-; CHECK-NEXT:    vmov.f32 s13, s17
-; CHECK-NEXT:    vmov.f32 s14, s18
-; CHECK-NEXT:    vldrw.u32 q4, [sp, #64] @ 16-byte Reload
-; CHECK-NEXT:    vmov.f32 s18, s10
-; CHECK-NEXT:    vstrw.32 q3, [r1, #48]
-; CHECK-NEXT:    vmov.f32 s9, s17
-; CHECK-NEXT:    vmov.f32 s30, s2
-; CHECK-NEXT:    vstrw.32 q7, [r1, #16]
-; CHECK-NEXT:    vmov.f32 s10, s18
-; CHECK-NEXT:    vldrw.u32 q4, [sp, #80] @ 16-byte Reload
-; CHECK-NEXT:    vmov.f32 s18, s22
-; CHECK-NEXT:    vstrw.32 q2, [r1, #32]
-; CHECK-NEXT:    vmov.f32 s21, s17
-; CHECK-NEXT:    vmov.f32 s22, s18
-; CHECK-NEXT:    vstrw.32 q5, [r1, #64]
-; CHECK-NEXT:    add sp, #128
+; CHECK-NEXT:    vmov.f32 s3, s2
+; CHECK-NEXT:    vstrw.32 q7, [r1]
+; CHECK-NEXT:    vmov.f32 s2, s10
+; CHECK-NEXT:    vmovx.f16 s20, s1
+; CHECK-NEXT:    vmov q1, q0
+; CHECK-NEXT:    vldrw.u32 q0, [sp, #48] @ 16-byte Reload
+; CHECK-NEXT:    vrev32.16 q2, q0
+; CHECK-NEXT:    vldrw.u32 q0, [sp, #96] @ 16-byte Reload
+; CHECK-NEXT:    vins.f16 s9, s20
+; CHECK-NEXT:    vmovx.f16 s20, s10
+; CHECK-NEXT:    vins.f16 s6, s20
+; CHECK-NEXT:    vldrw.u32 q5, [sp, #80] @ 16-byte Reload
+; CHECK-NEXT:    vmov.f32 s21, s1
+; CHECK-NEXT:    vmov.f32 s22, s2
+; CHECK-NEXT:    vldrw.u32 q0, [sp, #64] @ 16-byte Reload
+; CHECK-NEXT:    vmov.f32 s2, s26
+; CHECK-NEXT:    vstrw.32 q5, [r1, #32]
+; CHECK-NEXT:    vmov.f32 s25, s1
+; CHECK-NEXT:    vmov.f32 s10, s6
+; CHECK-NEXT:    vmov.f32 s5, s9
+; CHECK-NEXT:    vmov.f32 s26, s2
+; CHECK-NEXT:    vldrw.u32 q0, [sp, #32] @ 16-byte Reload
+; CHECK-NEXT:    vmov.f32 s2, s14
+; CHECK-NEXT:    vstrw.32 q6, [r1, #48]
+; CHECK-NEXT:    vmov.f32 s13, s1
+; CHECK-NEXT:    vmov.f32 s6, s10
+; CHECK-NEXT:    vstrw.32 q1, [r1, #16]
+; CHECK-NEXT:    vmov.f32 s14, s2
+; CHECK-NEXT:    vldrw.u32 q0, [sp, #112] @ 16-byte Reload
+; CHECK-NEXT:    vstrw.32 q3, [r1, #64]
+; CHECK-NEXT:    vstrw.32 q0, [r1, #80]
+; CHECK-NEXT:    add sp, #144
 ; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13, d14, d15}
 ; CHECK-NEXT:    bx lr
 entry:
   %s1 = getelementptr <16 x half>, <16 x half>* %src, i32 0
   %l1 = load <16 x half>, <16 x half>* %s1, align 4
   %s2 = getelementptr <16 x half>, <16 x half>* %src, i32 1
   %l2 = load <16 x half>, <16 x half>* %s2, align 4
   %s3 = getelementptr <16 x half>, <16 x half>* %src, i32 2

llvm/test/CodeGen/Thumb2/mve-vst4.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=thumbv8.1m.main-none-none-eabi -mattr=+mve.fp,+fp64 -mve-max-interleave-factor=4 -verify-machineinstrs %s -o - | FileCheck %s
 
 ; i32
 
 define void @vst4_v2i32(<2 x i32> *%src, <8 x i32> *%dst) {
 ; CHECK-LABEL: vst4_v2i32:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .save {r4, lr}
 ; CHECK-NEXT:    push {r4, lr}
-; CHECK-NEXT:    ldrd lr, r12, [r0]
-; CHECK-NEXT:    ldrd r3, r2, [r0, #8]
+; CHECK-NEXT:    ldrd r3, r2, [r0]
+; CHECK-NEXT:    ldrd r12, lr, [r0, #8]
 ; CHECK-NEXT:    ldrd r4, r0, [r0, #16]
+; CHECK-NEXT:    vmov.32 q2[0], r3
+; CHECK-NEXT:    vmov.f64 d6, d4
+; CHECK-NEXT:    vmov.32 q2[1], r2
 ; CHECK-NEXT:    vmov q1[2], q1[0], r4, r0
-; CHECK-NEXT:    vmov.f64 d0, d2
+; CHECK-NEXT:    vmov.32 q0[0], r4
 ; CHECK-NEXT:    vmov.f32 s1, s6
-; CHECK-NEXT:    vmov.f32 s2, s4
+; CHECK-NEXT:    vmov.f32 s2, s0
 ; CHECK-NEXT:    vmov.f32 s3, s6
-; CHECK-NEXT:    vmov q1[2], q1[0], lr, r3
-; CHECK-NEXT:    vmov q1[3], q1[1], r12, r2
-; CHECK-NEXT:    vmov.f64 d4, d2
-; CHECK-NEXT:    vmov.f32 s9, s6
-; CHECK-NEXT:    vmov.f32 s10, s0
-; CHECK-NEXT:    vmov.f32 s11, s2
-; CHECK-NEXT:    vstrw.32 q2, [r1]
-; CHECK-NEXT:    vmov.f32 s8, s5
+; CHECK-NEXT:    vmov.32 q1[1], r2
+; CHECK-NEXT:    vmov q1[2], q1[0], r3, r12
+; CHECK-NEXT:    vmov.f32 s8, s9
+; CHECK-NEXT:    vmov.f32 s13, s6
+; CHECK-NEXT:    vmov q1[2], q1[0], r3, r12
+; CHECK-NEXT:    vmov q1[3], q1[1], r2, lr
+; CHECK-NEXT:    vmov.f32 s14, s0
 ; CHECK-NEXT:    vmov.f32 s9, s7
 ; CHECK-NEXT:    vmov.f32 s10, s1
+; CHECK-NEXT:    vmov.f32 s15, s2
 ; CHECK-NEXT:    vmov.f32 s11, s3
+; CHECK-NEXT:    vstrw.32 q3, [r1]
 ; CHECK-NEXT:    vstrw.32 q2, [r1, #16]
 ; CHECK-NEXT:    pop {r4, pc}
 entry:
   %s1 = getelementptr <2 x i32>, <2 x i32>* %src, i32 0
   %l1 = load <2 x i32>, <2 x i32>* %s1, align 4
   %s2 = getelementptr <2 x i32>, <2 x i32>* %src, i32 1
   %l2 = load <2 x i32>, <2 x i32>* %s2, align 4
   %s3 = getelementptr <2 x i32>, <2 x i32>* %src, i32 2

 
 ; i16
 
 define void @vst4_v2i16(<2 x i16> *%src, <8 x i16> *%dst) {
 ; CHECK-LABEL: vst4_v2i16:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .save {r4, lr}
 ; CHECK-NEXT:    push {r4, lr}
-; CHECK-NEXT:    ldrh r2, [r0, #2]
-; CHECK-NEXT:    ldrh.w r12, [r0, #4]
-; CHECK-NEXT:    ldrh r3, [r0, #8]
-; CHECK-NEXT:    ldrh.w lr, [r0, #6]
-; CHECK-NEXT:    ldrh r4, [r0, #10]
-; CHECK-NEXT:    ldrh r0, [r0]
-; CHECK-NEXT:    vmov q0[2], q0[0], r0, r2
-; CHECK-NEXT:    vmov r0, s0
-; CHECK-NEXT:    vmov.16 q0[0], r0
-; CHECK-NEXT:    vmov.16 q0[1], r12
-; CHECK-NEXT:    vmov.16 q0[2], r3
-; CHECK-NEXT:    vmov.16 q0[3], r3
-; CHECK-NEXT:    vmov.16 q0[4], r2
-; CHECK-NEXT:    vmov.16 q0[5], lr
-; CHECK-NEXT:    vmov.16 q0[6], r4
-; CHECK-NEXT:    vmov.16 q0[7], r4
+; CHECK-NEXT:    ldrh r4, [r0]
+; CHECK-NEXT:    ldrh r2, [r0, #4]
+; CHECK-NEXT:    vmov.16 q0[0], r4
+; CHECK-NEXT:    ldrh.w lr, [r0, #2]
+; CHECK-NEXT:    ldrh.w r12, [r0, #6]
+; CHECK-NEXT:    vmov.16 q0[1], r2
+; CHECK-NEXT:    ldrh r3, [r0, #10]
+; CHECK-NEXT:    ldrh r0, [r0, #8]
+; CHECK-NEXT:    vmov.16 q0[2], r0
+; CHECK-NEXT:    vmov.16 q0[3], r0
+; CHECK-NEXT:    vmov.16 q0[4], lr
+; CHECK-NEXT:    vmov.16 q0[5], r12
+; CHECK-NEXT:    vmov.16 q0[6], r3
+; CHECK-NEXT:    vmov.16 q0[7], r3
 ; CHECK-NEXT:    vstrh.16 q0, [r1]
 ; CHECK-NEXT:    pop {r4, pc}
 entry:
   %s1 = getelementptr <2 x i16>, <2 x i16>* %src, i32 0
   %l1 = load <2 x i16>, <2 x i16>* %s1, align 4
   %s2 = getelementptr <2 x i16>, <2 x i16>* %src, i32 1
   %l2 = load <2 x i16>, <2 x i16>* %s2, align 4
   %s3 = getelementptr <2 x i16>, <2 x i16>* %src, i32 2

 ; CHECK-NEXT:    vmovx.f16 s24, s11
 ; CHECK-NEXT:    vmovx.f16 s16, s13
 ; CHECK-NEXT:    vins.f16 s13, s13
 ; CHECK-NEXT:    vins.f16 s16, s16
 ; CHECK-NEXT:    vmov q5, q3
 ; CHECK-NEXT:    vmov.f32 s23, s16
 ; CHECK-NEXT:    vmovx.f16 s16, s8
 ; CHECK-NEXT:    vmov.f32 s0, s5
-; CHECK-NEXT:    vmov.f32 s1, s21
+; CHECK-NEXT:    vmov.f32 s1, s13
 ; CHECK-NEXT:    vmov.f32 s3, s23
 ; CHECK-NEXT:    vmovx.f16 s20, s4
 ; CHECK-NEXT:    vins.f16 s4, s8
 ; CHECK-NEXT:    vins.f16 s20, s16
 ; CHECK-NEXT:    vmov q4, q1
 ; CHECK-NEXT:    vmovx.f16 s8, s10
 ; CHECK-NEXT:    vmov.f32 s18, s20
 ; CHECK-NEXT:    vmovx.f16 s22, s7
-; CHECK-NEXT:    vins.f16 s7, s11
-; CHECK-NEXT:    vins.f16 s22, s24
-; CHECK-NEXT:    vmovx.f16 s26, s6
 ; CHECK-NEXT:    vmovx.f16 s19, s12
 ; CHECK-NEXT:    vins.f16 s12, s12
+; CHECK-NEXT:    vmov.f32 s17, s12
+; CHECK-NEXT:    vins.f16 s22, s24
+; CHECK-NEXT:    vins.f16 s19, s19
+; CHECK-NEXT:    vstrb.8 q0, [r1, #16]
+; CHECK-NEXT:    vmov.f32 s16, s4
+; CHECK-NEXT:    vins.f16 s7, s11
+; CHECK-NEXT:    vmovx.f16 s26, s6
 ; CHECK-NEXT:    vmov.f32 s20, s7
 ; CHECK-NEXT:    vins.f16 s6, s10
 ; CHECK-NEXT:    vins.f16 s26, s8
-; CHECK-NEXT:    vmov.f32 s17, s12
 ; CHECK-NEXT:    vmovx.f16 s23, s15
 ; CHECK-NEXT:    vins.f16 s15, s15
 ; CHECK-NEXT:    vmov.f32 s24, s6
+; CHECK-NEXT:    vstrb.8 q4, [r1]
 ; CHECK-NEXT:    vmov.f32 s21, s15
 ; CHECK-NEXT:    vmovx.f16 s27, s14
 ; CHECK-NEXT:    vins.f16 s14, s14
-; CHECK-NEXT:    vins.f16 s19, s19
-; CHECK-NEXT:    vmov.f32 s25, s14
 ; CHECK-NEXT:    vins.f16 s23, s23
-; CHECK-NEXT:    vins.f16 s27, s27
+; CHECK-NEXT:    vmov.f32 s25, s14
 ; CHECK-NEXT:    vstrb.8 q5, [r1, #48]
+; CHECK-NEXT:    vins.f16 s27, s27
 ; CHECK-NEXT:    vstrb.8 q6, [r1, #32]
-; CHECK-NEXT:    vstrb.8 q0, [r1, #16]
-; CHECK-NEXT:    vstrb.8 q4, [r1]
 ; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13}
 ; CHECK-NEXT:    bx lr
 entry:
   %s1 = getelementptr <8 x i16>, <8 x i16>* %src, i32 0
   %l1 = load <8 x i16>, <8 x i16>* %s1, align 4
   %s2 = getelementptr <8 x i16>, <8 x i16>* %src, i32 1
   %l2 = load <8 x i16>, <8 x i16>* %s2, align 4
   %s3 = getelementptr <8 x i16>, <8 x i16>* %src, i32 2

 ; i8
 
 define void @vst4_v2i8(<2 x i8> *%src, <8 x i8> *%dst) {
 ; CHECK-LABEL: vst4_v2i8:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .save {r4, lr}
 ; CHECK-NEXT:    push {r4, lr}
 ; CHECK-NEXT:    ldrb r2, [r0]
-; CHECK-NEXT:    ldrb r3, [r0, #1]
-; CHECK-NEXT:    ldrb.w r12, [r0, #2]
-; CHECK-NEXT:    vmov q0[2], q0[0], r2, r3
+; CHECK-NEXT:    ldrb r3, [r0, #2]
+; CHECK-NEXT:    vmov.16 q0[0], r2
+; CHECK-NEXT:    ldrb.w r12, [r0, #1]
 ; CHECK-NEXT:    ldrb.w lr, [r0, #3]
-; CHECK-NEXT:    vmov r2, s0
+; CHECK-NEXT:    vmov.16 q0[1], r3
 ; CHECK-NEXT:    ldrb r4, [r0, #5]
-; CHECK-NEXT:    vmov.16 q0[0], r2
 ; CHECK-NEXT:    ldrb r0, [r0, #4]
-; CHECK-NEXT:    vmov.16 q0[1], r12
 ; CHECK-NEXT:    vmov.16 q0[2], r0
 ; CHECK-NEXT:    vmov.16 q0[3], r0
-; CHECK-NEXT:    vmov.16 q0[4], r3
+; CHECK-NEXT:    vmov.16 q0[4], r12
 ; CHECK-NEXT:    vmov.16 q0[5], lr
 ; CHECK-NEXT:    vmov.16 q0[6], r4
 ; CHECK-NEXT:    vmov.16 q0[7], r4
 ; CHECK-NEXT:    vstrb.16 q0, [r1]
 ; CHECK-NEXT:    pop {r4, pc}
 entry:
   %s1 = getelementptr <2 x i8>, <2 x i8>* %src, i32 0
   %l1 = load <2 x i8>, <2 x i8>* %s1, align 4
entry:
   ret void
 }
 
 define void @vst4_v4f16(<4 x half> *%src, <16 x half> *%dst) {
 ; CHECK-LABEL: vst4_v4f16:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .save {r7, lr}
 ; CHECK-NEXT:    push {r7, lr}
+; CHECK-NEXT:    .vsave {d8, d9}
+; CHECK-NEXT:    vpush {d8, d9}
 ; CHECK-NEXT:    ldrd r2, r12, [r0]
 ; CHECK-NEXT:    ldrd r3, lr, [r0, #8]
 ; CHECK-NEXT:    vmov.32 q0[0], r2
 ; CHECK-NEXT:    ldrd r2, r0, [r0, #16]
+; CHECK-NEXT:    vmov q3, q0
 ; CHECK-NEXT:    vmov.32 q1[0], r3
-; CHECK-NEXT:    vmov.32 q0[1], r12
-; CHECK-NEXT:    vmov.32 q1[1], lr
+; CHECK-NEXT:    vmov.32 q3[1], r12
+; CHECK-NEXT:    vmov q2, q1
+; CHECK-NEXT:    vmov.f32 s1, s13
+; CHECK-NEXT:    vmov.32 q2[1], lr
 ; CHECK-NEXT:    vmov.f32 s2, s4
-; CHECK-NEXT:    vmov.f32 s3, s5
+; CHECK-NEXT:    vmov.f32 s3, s9
 ; CHECK-NEXT:    vmovx.f16 s12, s0
 ; CHECK-NEXT:    vins.f16 s0, s2
 ; CHECK-NEXT:    vmovx.f16 s4, s2
 ; CHECK-NEXT:    vins.f16 s12, s4
 ; CHECK-NEXT:    vmovx.f16 s6, s1
 ; CHECK-NEXT:    vins.f16 s1, s3
 ; CHECK-NEXT:    vmovx.f16 s8, s3
 ; CHECK-NEXT:    vins.f16 s6, s8
 ; CHECK-NEXT:    vmov.32 q2[0], r2
-; CHECK-NEXT:    vmov.32 q2[1], r0
+; CHECK-NEXT:    vmov q4, q2
+; CHECK-NEXT:    vmov.32 q4[1], r0
+; CHECK-NEXT:    vmov.f32 s9, s17
 ; CHECK-NEXT:    vmov.f32 s10, s8
-; CHECK-NEXT:    vmov.f32 s11, s9
+; CHECK-NEXT:    vmov.f32 s11, s17
 ; CHECK-NEXT:    vmovx.f16 s3, s8
 ; CHECK-NEXT:    vins.f16 s8, s10
 ; CHECK-NEXT:    vmovx.f16 s14, s10
 ; CHECK-NEXT:    vmovx.f16 s7, s9
 ; CHECK-NEXT:    vins.f16 s3, s14
 ; CHECK-NEXT:    vins.f16 s9, s11
 ; CHECK-NEXT:    vmovx.f16 s14, s11
 ; CHECK-NEXT:    vins.f16 s7, s14
 ; CHECK-NEXT:    vmov.f32 s4, s1
 ; CHECK-NEXT:    vmov.f32 s1, s8
 ; CHECK-NEXT:    vmov.f32 s5, s9
 ; CHECK-NEXT:    vmov.f32 s2, s12
 ; CHECK-NEXT:    vstrh.16 q1, [r1, #16]
 ; CHECK-NEXT:    vstrh.16 q0, [r1]
+; CHECK-NEXT:    vpop {d8, d9}
 ; CHECK-NEXT:    pop {r7, pc}
 entry:
   %s1 = getelementptr <4 x half>, <4 x half>* %src, i32 0
   %l1 = load <4 x half>, <4 x half>* %s1, align 4
   %s2 = getelementptr <4 x half>, <4 x half>* %src, i32 1
   %l2 = load <4 x half>, <4 x half>* %s2, align 4
   %s3 = getelementptr <4 x half>, <4 x half>* %src, i32 2
   %l3 = load <4 x half>, <4 x half>* %s3, align 4
entry:
   ret void
 }
 
 define void @vst4_v8f16_align1(<8 x half> *%src, <32 x half> *%dst) {
 ; CHECK-LABEL: vst4_v8f16_align1:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13, d14, d15}
 ; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13, d14, d15}
-; CHECK-NEXT:    .pad #8
-; CHECK-NEXT:    sub sp, #8
 ; CHECK-NEXT:    vldrw.u32 q4, [r0, #16]
 ; CHECK-NEXT:    vldrw.u32 q1, [r0]
-; CHECK-NEXT:    vmovx.f16 s30, s5
+; CHECK-NEXT:    vldrw.u32 q5, [r0, #32]
+; CHECK-NEXT:    vmovx.f16 s2, s5
 ; CHECK-NEXT:    vmovx.f16 s8, s17
 ; CHECK-NEXT:    vins.f16 s5, s17
-; CHECK-NEXT:    vins.f16 s30, s8
-; CHECK-NEXT:    vmovx.f16 s0, s4
+; CHECK-NEXT:    vins.f16 s2, s8
+; CHECK-NEXT:    vmovx.f16 s24, s4
 ; CHECK-NEXT:    vmovx.f16 s8, s16
 ; CHECK-NEXT:    vins.f16 s4, s16
-; CHECK-NEXT:    vins.f16 s0, s8
+; CHECK-NEXT:    vins.f16 s24, s8
 ; CHECK-NEXT:    vmovx.f16 s10, s7
 ; CHECK-NEXT:    vmovx.f16 s12, s19
+; CHECK-NEXT:    vmovx.f16 s26, s21
 ; CHECK-NEXT:    vins.f16 s7, s19
 ; CHECK-NEXT:    vins.f16 s10, s12
 ; CHECK-NEXT:    vmovx.f16 s14, s6
 ; CHECK-NEXT:    vmovx.f16 s16, s18
+; CHECK-NEXT:    vins.f16 s21, s21
 ; CHECK-NEXT:    vins.f16 s6, s18
 ; CHECK-NEXT:    vins.f16 s14, s16
-; CHECK-NEXT:    vldrw.u32 q4, [r0, #32]
-; CHECK-NEXT:    vstr s0, [sp, #4] @ 4-byte Spill
-; CHECK-NEXT:    vmov.f32 s28, s5
-; CHECK-NEXT:    vmovx.f16 s24, s17
-; CHECK-NEXT:    vins.f16 s17, s17
-; CHECK-NEXT:    vmovx.f16 s23, s16
-; CHECK-NEXT:    vins.f16 s16, s16
-; CHECK-NEXT:    vmovx.f16 s11, s19
+; CHECK-NEXT:    vmovx.f16 s19, s20
+; CHECK-NEXT:    vins.f16 s20, s20
+; CHECK-NEXT:    vmovx.f16 s11, s23
+; CHECK-NEXT:    vins.f16 s23, s23
+; CHECK-NEXT:    vmovx.f16 s15, s22
+; CHECK-NEXT:    vins.f16 s22, s22
+; CHECK-NEXT:    vins.f16 s26, s26
+; CHECK-NEXT:    vmov q7, q5
+; CHECK-NEXT:    vmov.f32 s31, s26
 ; CHECK-NEXT:    vins.f16 s19, s19
-; CHECK-NEXT:    vmovx.f16 s15, s18
-; CHECK-NEXT:    vins.f16 s18, s18
-; CHECK-NEXT:    vins.f16 s24, s24
-; CHECK-NEXT:    vmov q0, q4
-; CHECK-NEXT:    vmov.f32 s3, s24
-; CHECK-NEXT:    vmov q6, q1
-; CHECK-NEXT:    vmov.f32 s29, s1
+; CHECK-NEXT:    vmov.f32 s0, s5
 ; CHECK-NEXT:    vins.f16 s11, s11
+; CHECK-NEXT:    vmov.f32 s1, s21
 ; CHECK-NEXT:    vins.f16 s15, s15
-; CHECK-NEXT:    vins.f16 s23, s23
+; CHECK-NEXT:    vmov.f32 s17, s20
 ; CHECK-NEXT:    vmov.f32 s8, s7
 ; CHECK-NEXT:    vmov.f32 s12, s6
-; CHECK-NEXT:    vmov.f32 s9, s19
-; CHECK-NEXT:    vmov.f32 s13, s18
+; CHECK-NEXT:    vmov.f32 s16, s4
+; CHECK-NEXT:    vmov.f32 s3, s31
+; CHECK-NEXT:    vmov q7, q1
+; CHECK-NEXT:    vmov.f32 s30, s24
+; CHECK-NEXT:    vstrb.8 q0, [r1, #16]
+; CHECK-NEXT:    vmov.f32 s9, s23
+; CHECK-NEXT:    vmov.f32 s13, s22
 ; CHECK-NEXT:    vstrb.8 q2, [r1, #48]
-; CHECK-NEXT:    vmov.f32 s31, s3
-; CHECK-NEXT:    vldr s0, [sp, #4] @ 4-byte Reload
-; CHECK-NEXT:    vmov.f32 s21, s16
 ; CHECK-NEXT:    vstrb.8 q3, [r1, #32]
-; CHECK-NEXT:    vmov.f32 s26, s0
-; CHECK-NEXT:    vstrb.8 q7, [r1, #16]
-; CHECK-NEXT:    vmov.f32 s25, s16
-; CHECK-NEXT:    vmov.f32 s27, s23
-; CHECK-NEXT:    vstrb.8 q6, [r1]
-; CHECK-NEXT:    add sp, #8
+; CHECK-NEXT:    vmov.f32 s18, s30
+; CHECK-NEXT:    vstrb.8 q4, [r1]
 ; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13, d14, d15}
 ; CHECK-NEXT:    bx lr
 entry:
   %s1 = getelementptr <8 x half>, <8 x half>* %src, i32 0
   %l1 = load <8 x half>, <8 x half>* %s1, align 4
   %s2 = getelementptr <8 x half>, <8 x half>* %src, i32 1
   %l2 = load <8 x half>, <8 x half>* %s2, align 4
   %s3 = getelementptr <8 x half>, <8 x half>* %src, i32 2

llvm/test/CodeGen/Thumb2/mve-zext-masked-load.ll

entry:
   %wide.masked.load = call <4 x i8> @llvm.masked.load.v4i8.p0v4i8(<4 x i8>* %pSrc, i32 1, <4 x i1> %active.lane.mask, <4 x i8> undef)
   %0 = uitofp <4 x i8> %wide.masked.load to <4 x float>
   ret <4 x float> %0
 }
 
 define arm_aapcs_vfpcc <4 x double> @foo_v4i32(<4 x i32>* nocapture readonly %pSrc, i32 %blockSize, <4 x i32> %a) {
 ; CHECK-LABEL: foo_v4i32:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .save {r4, r5, r6, r7, lr}
-; CHECK-NEXT:    push {r4, r5, r6, r7, lr}
-; CHECK-NEXT:    .pad #4
-; CHECK-NEXT:    sub sp, #4
-; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13}
-; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13}
+; CHECK-NEXT:    .save {r4, r5, r6, lr}
+; CHECK-NEXT:    push {r4, r5, r6, lr}
+; CHECK-NEXT:    .vsave {d8, d9, d10, d11}
+; CHECK-NEXT:    vpush {d8, d9, d10, d11}
 ; CHECK-NEXT:    vpt.s32 lt, q0, zr
 ; CHECK-NEXT:    vldrwt.u32 q4, [r0]
-; CHECK-NEXT:    vmov.f64 d0, d8
-; CHECK-NEXT:    vmov.i64 q5, #0xffffffff
-; CHECK-NEXT:    vmov.f32 s2, s17
-; CHECK-NEXT:    vand q6, q0, q5
-; CHECK-NEXT:    vmov r0, s24
-; CHECK-NEXT:    vmov r1, s25
+; CHECK-NEXT:    vmov.f64 d10, d9
+; CHECK-NEXT:    movs r1, #0
+; CHECK-NEXT:    vmov.f32 s22, s19
+; CHECK-NEXT:    vmov r0, s20
 ; CHECK-NEXT:    bl __aeabi_ul2d
 ; CHECK-NEXT:    mov r4, r0
+; CHECK-NEXT:    vmov r0, s22
 ; CHECK-NEXT:    mov r5, r1
-; CHECK-NEXT:    vmov r0, s26
-; CHECK-NEXT:    vmov r1, s27
+; CHECK-NEXT:    movs r1, #0
 ; CHECK-NEXT:    bl __aeabi_ul2d
-; CHECK-NEXT:    vmov.f64 d0, d9
-; CHECK-NEXT:    vmov.f32 s2, s19
-; CHECK-NEXT:    vand q0, q0, q5
+; CHECK-NEXT:    vmov.f32 s18, s17
+; CHECK-NEXT:    vmov r6, s16
+; CHECK-NEXT:    vmov r2, s18
 ; CHECK-NEXT:    vmov d9, r0, r1
-; CHECK-NEXT:    vmov r2, s2
-; CHECK-NEXT:    vmov r3, s3
-; CHECK-NEXT:    vmov r6, s0
-; CHECK-NEXT:    vmov r7, s1
+; CHECK-NEXT:    movs r1, #0
 ; CHECK-NEXT:    vmov d8, r4, r5
 ; CHECK-NEXT:    mov r0, r2
-; CHECK-NEXT:    mov r1, r3
 ; CHECK-NEXT:    bl __aeabi_ul2d
 ; CHECK-NEXT:    vmov d11, r0, r1
 ; CHECK-NEXT:    mov r0, r6
-; CHECK-NEXT:    mov r1, r7
+; CHECK-NEXT:    movs r1, #0
 ; CHECK-NEXT:    bl __aeabi_ul2d
 ; CHECK-NEXT:    vmov d10, r0, r1
-; CHECK-NEXT:    vmov q0, q4
-; CHECK-NEXT:    vmov q1, q5
-; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13}
-; CHECK-NEXT:    add sp, #4
-; CHECK-NEXT:    pop {r4, r5, r6, r7, pc}
+; CHECK-NEXT:    vmov q1, q4
+; CHECK-NEXT:    vmov q0, q5
+; CHECK-NEXT:    vpop {d8, d9, d10, d11}
+; CHECK-NEXT:    pop {r4, r5, r6, pc}
 entry:
   %active.lane.mask = icmp slt <4 x i32> %a, zeroinitializer
   %wide.masked.load = call <4 x i32> @llvm.masked.load.v4i32.p0v4i32(<4 x i32>* %pSrc, i32 4, <4 x i1> %active.lane.mask, <4 x i32> undef)
   %0 = uitofp <4 x i32> %wide.masked.load to <4 x double>
   ret <4 x double> %0
 }
 
 declare <4 x i16> @llvm.masked.load.v4i16.p0v4i16(<4 x i16>*, i32 immarg, <4 x i1>, <4 x i16>)
 
 declare <8 x i8> @llvm.masked.load.v8i8.p0v8i8(<8 x i8>*, i32 immarg, <8 x i1>, <8 x i8>)
 
 declare <4 x i8> @llvm.masked.load.v4i8.p0v4i8(<4 x i8>*, i32 immarg, <4 x i1>, <4 x i8>)
 
 declare <4 x i32> @llvm.masked.load.v4i32.p0v4i32(<4 x i32>*, i32 immarg, <4 x i1>, <4 x i32>)

llvm/test/CodeGen/X86/2011-10-19-widen_vselect.ll

entry:
   %1 = select <2 x i1> %0, <2 x float> %a, <2 x float> %b
   store <2 x float> %1, <2 x float>* undef
   ret void
 }
 
 define void @zero_test() {
 ; X86-LABEL: zero_test:
 ; X86:       # %bb.0: # %entry
-; X86-NEXT:    xorps %xmm0, %xmm0
-; X86-NEXT:    movlps %xmm0, (%eax)
+; X86-NEXT:    movl $0, (%eax)
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: zero_test:
 ; X64:       # %bb.0: # %entry
-; X64-NEXT:    xorps %xmm0, %xmm0
-; X64-NEXT:    movlps %xmm0, (%rax)
+; X64-NEXT:    movq $0, (%rax)
 ; X64-NEXT:    retq
 entry:
   %0 = select <2 x i1> undef, <2 x float> undef, <2 x float> zeroinitializer
   store <2 x float> %0, <2 x float>* undef
   ret void
 }
 
 define void @full_test() {

llvm/test/CodeGen/X86/2012-07-10-extload64.ll

entry:
   ret void
 }
 
 ; Make sure that we store a 64bit value, even on 32bit systems.
 define void @store_64(<2 x i32>* %ptr) {
 ; X86-LABEL: store_64:
 ; X86:       # %bb.0: # %BB
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT:    xorps %xmm0, %xmm0
-; X86-NEXT:    movlps %xmm0, (%eax)
+; X86-NEXT:    movl $0, 4(%eax)
+; X86-NEXT:    movl $0, (%eax)
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: store_64:
 ; X64:       # %bb.0: # %BB
 ; X64-NEXT:    movq $0, (%rcx)
 ; X64-NEXT:    retq
 BB:
   store <2 x i32> zeroinitializer, <2 x i32>* %ptr

llvm/test/CodeGen/X86/bitcast-vector-bool.ll

 ; SSE2-SSSE3-NEXT:    pmovmskb %xmm1, %ecx
 ; SSE2-SSSE3-NEXT:    shll $16, %ecx
 ; SSE2-SSSE3-NEXT:    orl %eax, %ecx
 ; SSE2-SSSE3-NEXT:    pmovmskb %xmm2, %eax
 ; SSE2-SSSE3-NEXT:    pmovmskb %xmm3, %edx
 ; SSE2-SSSE3-NEXT:    shll $16, %edx
 ; SSE2-SSSE3-NEXT:    orl %eax, %edx
 ; SSE2-SSSE3-NEXT:    shlq $32, %rdx
-; SSE2-SSSE3-NEXT:    orq %rcx, %rdx
 ; SSE2-SSSE3-NEXT:    movq %rdx, %xmm0
 ; SSE2-SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
 ; SSE2-SSSE3-NEXT:    movd %xmm0, %eax
 ; SSE2-SSSE3-NEXT:    addl %ecx, %eax
 ; SSE2-SSSE3-NEXT:    retq
 ;
 ; AVX1-LABEL: bitcast_v64i8_to_v2i32:
 ; AVX1:       # %bb.0:

llvm/test/CodeGen/X86/buildvec-insertvec.ll

; AVX2-NEXT:    retq
   ret void
 }
 
 ; OSS-Fuzz #5688
 ; https://bugs.chromium.org/p/oss-fuzz/issues/detail?id=5688
 define <4 x i32> @ossfuzz5688(i32 %a0) {
 ; CHECK-LABEL: ossfuzz5688:
 ; CHECK:       # %bb.0:
+; CHECK-NEXT:    movl $0, (%rax)
 ; CHECK-NEXT:    retq
   %1 = insertelement <4 x i32> zeroinitializer, i32 -2147483648, i32 %a0
   %2 = extractelement <4 x i32> %1, i32 %a0
   %3 = extractelement <4 x i32> <i32 30, i32 53, i32 42, i32 12>, i32 %2
   %4 = extractelement <4 x i32> zeroinitializer, i32 %2
   %5 = insertelement <4 x i32> undef, i32 %3, i32 undef
   store i32 %4, i32* undef
   ret <4 x i32> %5
 }
 
 ; If we do not define all bytes that are extracted, this is a miscompile.
 
 define i32 @PR46586(i8* %p, <4 x i32> %v) {
 ; SSE2-LABEL: PR46586:
 ; SSE2:       # %bb.0:

llvm/test/CodeGen/X86/fold-load-vec.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s -mtriple=x86_64-unknown -mcpu=corei7 -mattr=+sse4.1 | FileCheck %s
 
 ; rdar://12721174
 ; We should not fold movss into pshufd since pshufd expects m128 while movss
 ; loads from m32.
 define void @sample_test(<4 x float>* %source, <2 x float>* %dest) nounwind {
 ; CHECK-LABEL: sample_test:
 ; CHECK:       # %bb.0: # %entry
 ; CHECK-NEXT:    subq $24, %rsp
 ; CHECK-NEXT:    movq %rdi, {{[0-9]+}}(%rsp)
 ; CHECK-NEXT:    movq %rsi, {{[0-9]+}}(%rsp)
+; CHECK-NEXT:    movq $0, (%rsp)
 ; CHECK-NEXT:    xorps %xmm0, %xmm0
-; CHECK-NEXT:    movlps %xmm0, (%rsp)
 ; CHECK-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
 ; CHECK-NEXT:    movlps %xmm0, (%rsp)
 ; CHECK-NEXT:    movlps %xmm0, (%rsi)
 ; CHECK-NEXT:    movq {{[0-9]+}}(%rsp), %rax
 ; CHECK-NEXT:    movsd {{.*#+}} xmm0 = mem[0],zero
 ; CHECK-NEXT:    movshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
 ; CHECK-NEXT:    callq ext
 ; CHECK-NEXT:    addq $24, %rsp

llvm/test/CodeGen/X86/madd.ll

 }
 
 ; Do not select if constant is too large
 define <4 x i32> @pmaddwd_negative2(<8 x i16> %A) {
 ; SSE2-LABEL: pmaddwd_negative2:
 ; SSE2:       # %bb.0:
 ; SSE2-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
 ; SSE2-NEXT:    psrad $16, %xmm1
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    psrad $16, %xmm0
-; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[1,1,3,3]
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm0[4],xmm2[5],xmm0[5],xmm2[6],xmm0[6],xmm2[7],xmm0[7]
+; SSE2-NEXT:    psrad $16, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
 ; SSE2-NEXT:    movdqa {{.*#+}} xmm3 = [1,7,42,32]
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm3[1,1,3,3]
-; SSE2-NEXT:    pmuludq %xmm2, %xmm4
-; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[1,1,3,3]
-; SSE2-NEXT:    movdqa {{.*#+}} xmm5 = [32768,4294934528,0,0]
-; SSE2-NEXT:    pshufd {{.*#+}} xmm6 = xmm5[1,1,3,3]
-; SSE2-NEXT:    pmuludq %xmm2, %xmm6
-; SSE2-NEXT:    shufps {{.*#+}} xmm6 = xmm6[0,2],xmm4[0,2]
-; SSE2-NEXT:    pmuludq %xmm3, %xmm0
-; SSE2-NEXT:    pmuludq %xmm5, %xmm1
-; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,2],xmm0[0,2]
-; SSE2-NEXT:    paddd %xmm6, %xmm1
-; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    pmuludq %xmm0, %xmm4
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[1,1,3,3]
+; SSE2-NEXT:    pmuludq {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,1],xmm4[0,2]
+; SSE2-NEXT:    pmuludq %xmm3, %xmm2
+; SSE2-NEXT:    pmuludq {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm2[0,2]
+; SSE2-NEXT:    movaps {{.*#+}} xmm2 = [4294967295,0,4294967295,4294967295]
+; SSE2-NEXT:    andps %xmm2, %xmm1
+; SSE2-NEXT:    andps %xmm2, %xmm0
+; SSE2-NEXT:    paddd %xmm1, %xmm0
 ; SSE2-NEXT:    retq
 ;
 ; AVX1-LABEL: pmaddwd_negative2:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
 ; AVX1-NEXT:    vpmovsxwd %xmm1, %xmm1
 ; AVX1-NEXT:    vpmovsxwd %xmm0, %xmm0
 ; AVX1-NEXT:    vpmulld {{.*}}(%rip), %xmm0, %xmm0

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

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown | FileCheck %s --check-prefixes=SSE,SSE2
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse4a | FileCheck %s --check-prefixes=SSE,SSE4A
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse4.1 | FileCheck %s --check-prefixes=SSE,SSE41
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx | FileCheck %s --check-prefix=AVX
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefix=AVX
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512dq,+avx512vl | FileCheck %s --check-prefix=AVX512
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512bw,+avx512vl | FileCheck %s --check-prefix=AVX512
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown | FileCheck %s --check-prefixes=CHECK,SSE
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse4a | FileCheck %s --check-prefixes=CHECK,SSE
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse4.1 | FileCheck %s --check-prefixes=CHECK,SSE
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx | FileCheck %s --check-prefixes=CHECK,AVX
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefixes=CHECK,AVX
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512dq,+avx512vl | FileCheck %s --check-prefixes=CHECK,AVX512
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512bw,+avx512vl | FileCheck %s --check-prefixes=CHECK,AVX512
 
 ; Test codegen for under aligned nontemporal vector stores
 
 ; XMM versions.
 
 define void @test_zero_v2f64_align1(<2 x double>* %dst) nounwind {
-; SSE-LABEL: test_zero_v2f64_align1:
-; SSE:       # %bb.0:
-; SSE-NEXT:    xorl %eax, %eax
-; SSE-NEXT:    movntiq %rax, 8(%rdi)
-; SSE-NEXT:    movntiq %rax, (%rdi)
-; SSE-NEXT:    retq
-;
-; AVX-LABEL: test_zero_v2f64_align1:
-; AVX:       # %bb.0:
-; AVX-NEXT:    xorl %eax, %eax
-; AVX-NEXT:    movntiq %rax, 8(%rdi)
-; AVX-NEXT:    movntiq %rax, (%rdi)
-; AVX-NEXT:    retq
-;
-; AVX512-LABEL: test_zero_v2f64_align1:
-; AVX512:       # %bb.0:
-; AVX512-NEXT:    xorl %eax, %eax
-; AVX512-NEXT:    movntiq %rax, 8(%rdi)
-; AVX512-NEXT:    movntiq %rax, (%rdi)
-; AVX512-NEXT:    retq
+; CHECK-LABEL: test_zero_v2f64_align1:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xorl %eax, %eax
+; CHECK-NEXT:    movntiq %rax, 8(%rdi)
+; CHECK-NEXT:    movntiq %rax, (%rdi)
+; CHECK-NEXT:    retq
   store <2 x double> zeroinitializer, <2 x double>* %dst, align 1, !nontemporal !1
   ret void
 }
 
 define void @test_zero_v4f32_align1(<4 x float>* %dst) nounwind {
-; SSE-LABEL: test_zero_v4f32_align1:
-; SSE:       # %bb.0:
-; SSE-NEXT:    xorl %eax, %eax
-; SSE-NEXT:    movntiq %rax, 8(%rdi)
-; SSE-NEXT:    movntiq %rax, (%rdi)
-; SSE-NEXT:    retq
-;
-; AVX-LABEL: test_zero_v4f32_align1:
-; AVX:       # %bb.0:
-; AVX-NEXT:    xorl %eax, %eax
-; AVX-NEXT:    movntiq %rax, 8(%rdi)
-; AVX-NEXT:    movntiq %rax, (%rdi)
-; AVX-NEXT:    retq
-;
-; AVX512-LABEL: test_zero_v4f32_align1:
-; AVX512:       # %bb.0:
-; AVX512-NEXT:    xorl %eax, %eax
-; AVX512-NEXT:    movntiq %rax, 8(%rdi)
-; AVX512-NEXT:    movntiq %rax, (%rdi)
-; AVX512-NEXT:    retq
+; CHECK-LABEL: test_zero_v4f32_align1:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xorl %eax, %eax
+; CHECK-NEXT:    movntiq %rax, 8(%rdi)
+; CHECK-NEXT:    movntiq %rax, (%rdi)
+; CHECK-NEXT:    retq
   store <4 x float> zeroinitializer, <4 x float>* %dst, align 1, !nontemporal !1
   ret void
 }
 
 define void @test_zero_v2i64_align1(<2 x i64>* %dst) nounwind {
-; SSE-LABEL: test_zero_v2i64_align1:
-; SSE:       # %bb.0:
-; SSE-NEXT:    xorl %eax, %eax
-; SSE-NEXT:    movntiq %rax, 8(%rdi)
-; SSE-NEXT:    movntiq %rax, (%rdi)
-; SSE-NEXT:    retq
-;
-; AVX-LABEL: test_zero_v2i64_align1:
-; AVX:       # %bb.0:
-; AVX-NEXT:    xorl %eax, %eax
-; AVX-NEXT:    movntiq %rax, 8(%rdi)
-; AVX-NEXT:    movntiq %rax, (%rdi)
-; AVX-NEXT:    retq
-;
-; AVX512-LABEL: test_zero_v2i64_align1:
-; AVX512:       # %bb.0:
-; AVX512-NEXT:    xorl %eax, %eax
-; AVX512-NEXT:    movntiq %rax, 8(%rdi)
-; AVX512-NEXT:    movntiq %rax, (%rdi)
-; AVX512-NEXT:    retq
+; CHECK-LABEL: test_zero_v2i64_align1:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xorl %eax, %eax
+; CHECK-NEXT:    movntiq %rax, 8(%rdi)
+; CHECK-NEXT:    movntiq %rax, (%rdi)
+; CHECK-NEXT:    retq
   store <2 x i64> zeroinitializer, <2 x i64>* %dst, align 1, !nontemporal !1
   ret void
 }
 
 define void @test_zero_v4i32_align1(<4 x i32>* %dst) nounwind {
-; SSE-LABEL: test_zero_v4i32_align1:
-; SSE:       # %bb.0:
-; SSE-NEXT:    xorl %eax, %eax
-; SSE-NEXT:    movntiq %rax, 8(%rdi)
-; SSE-NEXT:    movntiq %rax, (%rdi)
-; SSE-NEXT:    retq
-;
-; AVX-LABEL: test_zero_v4i32_align1:
-; AVX:       # %bb.0:
-; AVX-NEXT:    xorl %eax, %eax
-; AVX-NEXT:    movntiq %rax, 8(%rdi)
-; AVX-NEXT:    movntiq %rax, (%rdi)
-; AVX-NEXT:    retq
-;
-; AVX512-LABEL: test_zero_v4i32_align1:
-; AVX512:       # %bb.0:
-; AVX512-NEXT:    xorl %eax, %eax
-; AVX512-NEXT:    movntiq %rax, 8(%rdi)
-; AVX512-NEXT:    movntiq %rax, (%rdi)
-; AVX512-NEXT:    retq
+; CHECK-LABEL: test_zero_v4i32_align1:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xorl %eax, %eax
+; CHECK-NEXT:    movntiq %rax, 8(%rdi)
+; CHECK-NEXT:    movntiq %rax, (%rdi)
+; CHECK-NEXT:    retq
   store <4 x i32> zeroinitializer, <4 x i32>* %dst, align 1, !nontemporal !1
   ret void
 }
 
 define void @test_zero_v8i16_align1(<8 x i16>* %dst) nounwind {
-; SSE-LABEL: test_zero_v8i16_align1:
-; SSE:       # %bb.0:
-; SSE-NEXT:    xorl %eax, %eax
-; SSE-NEXT:    movntiq %rax, 8(%rdi)
-; SSE-NEXT:    movntiq %rax, (%rdi)
-; SSE-NEXT:    retq
-;
-; AVX-LABEL: test_zero_v8i16_align1:
-; AVX:       # %bb.0:
-; AVX-NEXT:    xorl %eax, %eax
-; AVX-NEXT:    movntiq %rax, 8(%rdi)
-; AVX-NEXT:    movntiq %rax, (%rdi)
-; AVX-NEXT:    retq
-;
-; AVX512-LABEL: test_zero_v8i16_align1:
-; AVX512:       # %bb.0:
-; AVX512-NEXT:    xorl %eax, %eax
-; AVX512-NEXT:    movntiq %rax, 8(%rdi)
-; AVX512-NEXT:    movntiq %rax, (%rdi)
-; AVX512-NEXT:    retq
+; CHECK-LABEL: test_zero_v8i16_align1:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xorl %eax, %eax
+; CHECK-NEXT:    movntiq %rax, 8(%rdi)
+; CHECK-NEXT:    movntiq %rax, (%rdi)
+; CHECK-NEXT:    retq
   store <8 x i16> zeroinitializer, <8 x i16>* %dst, align 1, !nontemporal !1
   ret void
 }
 
 define void @test_zero_v16i8_align1(<16 x i8>* %dst) nounwind {
-; SSE-LABEL: test_zero_v16i8_align1:
-; SSE:       # %bb.0:
-; SSE-NEXT:    xorl %eax, %eax
-; SSE-NEXT:    movntiq %rax, 8(%rdi)
-; SSE-NEXT:    movntiq %rax, (%rdi)
-; SSE-NEXT:    retq
-;
-; AVX-LABEL: test_zero_v16i8_align1:
-; AVX:       # %bb.0:
-; AVX-NEXT:    xorl %eax, %eax
-; AVX-NEXT:    movntiq %rax, 8(%rdi)
-; AVX-NEXT:    movntiq %rax, (%rdi)
-; AVX-NEXT:    retq
-;
-; AVX512-LABEL: test_zero_v16i8_align1:
-; AVX512:       # %bb.0:
-; AVX512-NEXT:    xorl %eax, %eax
-; AVX512-NEXT:    movntiq %rax, 8(%rdi)
-; AVX512-NEXT:    movntiq %rax, (%rdi)
-; AVX512-NEXT:    retq
+; CHECK-LABEL: test_zero_v16i8_align1:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xorl %eax, %eax
+; CHECK-NEXT:    movntiq %rax, 8(%rdi)
+; CHECK-NEXT:    movntiq %rax, (%rdi)
+; CHECK-NEXT:    retq
   store <16 x i8> zeroinitializer, <16 x i8>* %dst, align 1, !nontemporal !1
   ret void
 }
 
 ; YMM versions.
 
 define void @test_zero_v4f64_align1(<4 x double>* %dst) nounwind {
-; SSE-LABEL: test_zero_v4f64_align1:
-; SSE:       # %bb.0:
-; SSE-NEXT:    xorl %eax, %eax
-; SSE-NEXT:    movntiq %rax, 8(%rdi)
-; SSE-NEXT:    movntiq %rax, (%rdi)
-; SSE-NEXT:    movntiq %rax, 24(%rdi)
-; SSE-NEXT:    movntiq %rax, 16(%rdi)
-; SSE-NEXT:    retq
-;
-; AVX-LABEL: test_zero_v4f64_align1:
-; AVX:       # %bb.0:
-; AVX-NEXT:    xorl %eax, %eax
-; AVX-NEXT:    movntiq %rax, 8(%rdi)
-; AVX-NEXT:    movntiq %rax, (%rdi)
-; AVX-NEXT:    movntiq %rax, 24(%rdi)
-; AVX-NEXT:    movntiq %rax, 16(%rdi)
-; AVX-NEXT:    retq
-;
-; AVX512-LABEL: test_zero_v4f64_align1:
-; AVX512:       # %bb.0:
-; AVX512-NEXT:    xorl %eax, %eax
-; AVX512-NEXT:    movntiq %rax, 8(%rdi)
-; AVX512-NEXT:    movntiq %rax, (%rdi)
-; AVX512-NEXT:    movntiq %rax, 24(%rdi)
-; AVX512-NEXT:    movntiq %rax, 16(%rdi)
-; AVX512-NEXT:    retq
+; CHECK-LABEL: test_zero_v4f64_align1:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xorl %eax, %eax
+; CHECK-NEXT:    movntiq %rax, 8(%rdi)
+; CHECK-NEXT:    movntiq %rax, (%rdi)
+; CHECK-NEXT:    movntiq %rax, 24(%rdi)
+; CHECK-NEXT:    movntiq %rax, 16(%rdi)
+; CHECK-NEXT:    retq
   store <4 x double> zeroinitializer, <4 x double>* %dst, align 1, !nontemporal !1
   ret void
 }
 
 define void @test_zero_v8f32_align1(<8 x float>* %dst) nounwind {
-; SSE2-LABEL: test_zero_v8f32_align1:
-; SSE2:       # %bb.0:
-; SSE2-NEXT:    xorl %eax, %eax
-; SSE2-NEXT:    movntiq %rax, 8(%rdi)
-; SSE2-NEXT:    movntiq %rax, (%rdi)
-; SSE2-NEXT:    movntiq %rax, 24(%rdi)
-; SSE2-NEXT:    movntiq %rax, 16(%rdi)
-; SSE2-NEXT:    retq
-;
-; SSE4A-LABEL: test_zero_v8f32_align1:
-; SSE4A:       # %bb.0:
-; SSE4A-NEXT:    xorl %eax, %eax
-; SSE4A-NEXT:    movntiq %rax, 8(%rdi)
-; SSE4A-NEXT:    movntiq %rax, 24(%rdi)
-; SSE4A-NEXT:    xorps %xmm0, %xmm0
-; SSE4A-NEXT:    movntsd %xmm0, (%rdi)
-; SSE4A-NEXT:    movntsd %xmm0, 16(%rdi)
-; SSE4A-NEXT:    retq
-;
-; SSE41-LABEL: test_zero_v8f32_align1:
-; SSE41:       # %bb.0:
-; SSE41-NEXT:    xorl %eax, %eax
-; SSE41-NEXT:    movntiq %rax, 8(%rdi)
-; SSE41-NEXT:    movntiq %rax, (%rdi)
-; SSE41-NEXT:    movntiq %rax, 24(%rdi)
-; SSE41-NEXT:    movntiq %rax, 16(%rdi)
-; SSE41-NEXT:    retq
-;
-; AVX-LABEL: test_zero_v8f32_align1:
-; AVX:       # %bb.0:
-; AVX-NEXT:    xorl %eax, %eax
-; AVX-NEXT:    movntiq %rax, 8(%rdi)
-; AVX-NEXT:    movntiq %rax, (%rdi)
-; AVX-NEXT:    movntiq %rax, 24(%rdi)
-; AVX-NEXT:    movntiq %rax, 16(%rdi)
-; AVX-NEXT:    retq
-;
-; AVX512-LABEL: test_zero_v8f32_align1:
-; AVX512:       # %bb.0:
-; AVX512-NEXT:    xorl %eax, %eax
-; AVX512-NEXT:    movntiq %rax, 8(%rdi)
-; AVX512-NEXT:    movntiq %rax, (%rdi)
-; AVX512-NEXT:    movntiq %rax, 24(%rdi)
-; AVX512-NEXT:    movntiq %rax, 16(%rdi)
-; AVX512-NEXT:    retq
+; CHECK-LABEL: test_zero_v8f32_align1:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xorl %eax, %eax
+; CHECK-NEXT:    movntiq %rax, 8(%rdi)
+; CHECK-NEXT:    movntiq %rax, (%rdi)
+; CHECK-NEXT:    movntiq %rax, 24(%rdi)
+; CHECK-NEXT:    movntiq %rax, 16(%rdi)
+; CHECK-NEXT:    retq
   store <8 x float> zeroinitializer, <8 x float>* %dst, align 1, !nontemporal !1
   ret void
 }
 
 define void @test_zero_v4i64_align1(<4 x i64>* %dst) nounwind {
-; SSE2-LABEL: test_zero_v4i64_align1:
-; SSE2:       # %bb.0:
-; SSE2-NEXT:    xorl %eax, %eax
-; SSE2-NEXT:    movntiq %rax, 8(%rdi)
-; SSE2-NEXT:    movntiq %rax, (%rdi)
-; SSE2-NEXT:    movntiq %rax, 24(%rdi)
-; SSE2-NEXT:    movntiq %rax, 16(%rdi)
-; SSE2-NEXT:    retq
-;
-; SSE4A-LABEL: test_zero_v4i64_align1:
-; SSE4A:       # %bb.0:
-; SSE4A-NEXT:    xorps %xmm0, %xmm0
-; SSE4A-NEXT:    movntsd %xmm0, 8(%rdi)
-; SSE4A-NEXT:    movntsd %xmm0, (%rdi)
-; SSE4A-NEXT:    movntsd %xmm0, 24(%rdi)
-; SSE4A-NEXT:    movntsd %xmm0, 16(%rdi)
-; SSE4A-NEXT:    retq
-;
-; SSE41-LABEL: test_zero_v4i64_align1:
-; SSE41:       # %bb.0:
-; SSE41-NEXT:    xorl %eax, %eax
-; SSE41-NEXT:    movntiq %rax, 8(%rdi)
-; SSE41-NEXT:    movntiq %rax, (%rdi)
-; SSE41-NEXT:    movntiq %rax, 24(%rdi)
-; SSE41-NEXT:    movntiq %rax, 16(%rdi)
-; SSE41-NEXT:    retq
-;
-; AVX-LABEL: test_zero_v4i64_align1:
-; AVX:       # %bb.0:
-; AVX-NEXT:    xorl %eax, %eax
-; AVX-NEXT:    movntiq %rax, 8(%rdi)
-; AVX-NEXT:    movntiq %rax, (%rdi)
-; AVX-NEXT:    movntiq %rax, 24(%rdi)
-; AVX-NEXT:    movntiq %rax, 16(%rdi)
-; AVX-NEXT:    retq
-;
-; AVX512-LABEL: test_zero_v4i64_align1:
-; AVX512:       # %bb.0:
-; AVX512-NEXT:    xorl %eax, %eax
-; AVX512-NEXT:    movntiq %rax, 8(%rdi)
-; AVX512-NEXT:    movntiq %rax, (%rdi)
-; AVX512-NEXT:    movntiq %rax, 24(%rdi)
-; AVX512-NEXT:    movntiq %rax, 16(%rdi)
-; AVX512-NEXT:    retq
+; CHECK-LABEL: test_zero_v4i64_align1:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xorl %eax, %eax
+; CHECK-NEXT:    movntiq %rax, 8(%rdi)
+; CHECK-NEXT:    movntiq %rax, (%rdi)
+; CHECK-NEXT:    movntiq %rax, 24(%rdi)
+; CHECK-NEXT:    movntiq %rax, 16(%rdi)
+; CHECK-NEXT:    retq
   store <4 x i64> zeroinitializer, <4 x i64>* %dst, align 1, !nontemporal !1
   ret void
 }
 
 define void @test_zero_v8i32_align1(<8 x i32>* %dst) nounwind {
-; SSE2-LABEL: test_zero_v8i32_align1:
-; SSE2:       # %bb.0:
-; SSE2-NEXT:    xorl %eax, %eax
-; SSE2-NEXT:    movntiq %rax, 8(%rdi)
-; SSE2-NEXT:    movntiq %rax, (%rdi)
-; SSE2-NEXT:    movntiq %rax, 24(%rdi)
-; SSE2-NEXT:    movntiq %rax, 16(%rdi)
-; SSE2-NEXT:    retq
-;
-; SSE4A-LABEL: test_zero_v8i32_align1:
-; SSE4A:       # %bb.0:
-; SSE4A-NEXT:    xorps %xmm0, %xmm0
-; SSE4A-NEXT:    movntsd %xmm0, 8(%rdi)
-; SSE4A-NEXT:    movntsd %xmm0, (%rdi)
-; SSE4A-NEXT:    movntsd %xmm0, 24(%rdi)
-; SSE4A-NEXT:    movntsd %xmm0, 16(%rdi)
-; SSE4A-NEXT:    retq
-;
-; SSE41-LABEL: test_zero_v8i32_align1:
-; SSE41:       # %bb.0:
-; SSE41-NEXT:    xorl %eax, %eax
-; SSE41-NEXT:    movntiq %rax, 8(%rdi)
-; SSE41-NEXT:    movntiq %rax, (%rdi)
-; SSE41-NEXT:    movntiq %rax, 24(%rdi)
-; SSE41-NEXT:    movntiq %rax, 16(%rdi)
-; SSE41-NEXT:    retq
-;
-; AVX-LABEL: test_zero_v8i32_align1:
-; AVX:       # %bb.0:
-; AVX-NEXT:    xorl %eax, %eax
-; AVX-NEXT:    movntiq %rax, 8(%rdi)
-; AVX-NEXT:    movntiq %rax, (%rdi)
-; AVX-NEXT:    movntiq %rax, 24(%rdi)
-; AVX-NEXT:    movntiq %rax, 16(%rdi)
-; AVX-NEXT:    retq
-;
-; AVX512-LABEL: test_zero_v8i32_align1:
-; AVX512:       # %bb.0:
-; AVX512-NEXT:    xorl %eax, %eax
-; AVX512-NEXT:    movntiq %rax, 8(%rdi)
-; AVX512-NEXT:    movntiq %rax, (%rdi)
-; AVX512-NEXT:    movntiq %rax, 24(%rdi)
-; AVX512-NEXT:    movntiq %rax, 16(%rdi)
-; AVX512-NEXT:    retq
+; CHECK-LABEL: test_zero_v8i32_align1:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xorl %eax, %eax
+; CHECK-NEXT:    movntiq %rax, 8(%rdi)
+; CHECK-NEXT:    movntiq %rax, (%rdi)
+; CHECK-NEXT:    movntiq %rax, 24(%rdi)
+; CHECK-NEXT:    movntiq %rax, 16(%rdi)
+; CHECK-NEXT:    retq
   store <8 x i32> zeroinitializer, <8 x i32>* %dst, align 1, !nontemporal !1
   ret void
 }
 
 define void @test_zero_v16i16_align1(<16 x i16>* %dst) nounwind {
-; SSE2-LABEL: test_zero_v16i16_align1:
-; SSE2:       # %bb.0:
-; SSE2-NEXT:    xorl %eax, %eax
-; SSE2-NEXT:    movntiq %rax, 8(%rdi)
-; SSE2-NEXT:    movntiq %rax, (%rdi)
-; SSE2-NEXT:    movntiq %rax, 24(%rdi)
-; SSE2-NEXT:    movntiq %rax, 16(%rdi)
-; SSE2-NEXT:    retq
-;
-; SSE4A-LABEL: test_zero_v16i16_align1:
-; SSE4A:       # %bb.0:
-; SSE4A-NEXT:    xorps %xmm0, %xmm0
-; SSE4A-NEXT:    movntsd %xmm0, 8(%rdi)
-; SSE4A-NEXT:    movntsd %xmm0, (%rdi)
-; SSE4A-NEXT:    movntsd %xmm0, 24(%rdi)
-; SSE4A-NEXT:    movntsd %xmm0, 16(%rdi)
-; SSE4A-NEXT:    retq
-;
-; SSE41-LABEL: test_zero_v16i16_align1:
-; SSE41:       # %bb.0:
-; SSE41-NEXT:    xorl %eax, %eax
-; SSE41-NEXT:    movntiq %rax, 8(%rdi)
-; SSE41-NEXT:    movntiq %rax, (%rdi)
-; SSE41-NEXT:    movntiq %rax, 24(%rdi)
-; SSE41-NEXT:    movntiq %rax, 16(%rdi)
-; SSE41-NEXT:    retq
-;
-; AVX-LABEL: test_zero_v16i16_align1:
-; AVX:       # %bb.0:
-; AVX-NEXT:    xorl %eax, %eax
-; AVX-NEXT:    movntiq %rax, 8(%rdi)
-; AVX-NEXT:    movntiq %rax, (%rdi)
-; AVX-NEXT:    movntiq %rax, 24(%rdi)
-; AVX-NEXT:    movntiq %rax, 16(%rdi)
-; AVX-NEXT:    retq
-;
-; AVX512-LABEL: test_zero_v16i16_align1:
-; AVX512:       # %bb.0:
-; AVX512-NEXT:    xorl %eax, %eax
-; AVX512-NEXT:    movntiq %rax, 8(%rdi)
-; AVX512-NEXT:    movntiq %rax, (%rdi)
-; AVX512-NEXT:    movntiq %rax, 24(%rdi)
-; AVX512-NEXT:    movntiq %rax, 16(%rdi)
-; AVX512-NEXT:    retq
+; CHECK-LABEL: test_zero_v16i16_align1:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xorl %eax, %eax
+; CHECK-NEXT:    movntiq %rax, 8(%rdi)
+; CHECK-NEXT:    movntiq %rax, (%rdi)
+; CHECK-NEXT:    movntiq %rax, 24(%rdi)
+; CHECK-NEXT:    movntiq %rax, 16(%rdi)
+; CHECK-NEXT:    retq
   store <16 x i16> zeroinitializer, <16 x i16>* %dst, align 1, !nontemporal !1
   ret void
 }
 
 define void @test_zero_v32i8_align1(<32 x i8>* %dst) nounwind {
-; SSE2-LABEL: test_zero_v32i8_align1:
-; SSE2:       # %bb.0:
-; SSE2-NEXT:    xorl %eax, %eax
-; SSE2-NEXT:    movntiq %rax, 8(%rdi)
-; SSE2-NEXT:    movntiq %rax, (%rdi)
-; SSE2-NEXT:    movntiq %rax, 24(%rdi)
-; SSE2-NEXT:    movntiq %rax, 16(%rdi)
-; SSE2-NEXT:    retq
-;
-; SSE4A-LABEL: test_zero_v32i8_align1:
-; SSE4A:       # %bb.0:
-; SSE4A-NEXT:    xorps %xmm0, %xmm0
-; SSE4A-NEXT:    movntsd %xmm0, 8(%rdi)
-; SSE4A-NEXT:    movntsd %xmm0, (%rdi)
-; SSE4A-NEXT:    movntsd %xmm0, 24(%rdi)
-; SSE4A-NEXT:    movntsd %xmm0, 16(%rdi)
-; SSE4A-NEXT:    retq
-;
-; SSE41-LABEL: test_zero_v32i8_align1:
-; SSE41:       # %bb.0:
-; SSE41-NEXT:    xorl %eax, %eax
-; SSE41-NEXT:    movntiq %rax, 8(%rdi)
-; SSE41-NEXT:    movntiq %rax, (%rdi)
-; SSE41-NEXT:    movntiq %rax, 24(%rdi)
-; SSE41-NEXT:    movntiq %rax, 16(%rdi)
-; SSE41-NEXT:    retq
-;
-; AVX-LABEL: test_zero_v32i8_align1:
-; AVX:       # %bb.0:
-; AVX-NEXT:    xorl %eax, %eax
-; AVX-NEXT:    movntiq %rax, 8(%rdi)
-; AVX-NEXT:    movntiq %rax, (%rdi)
-; AVX-NEXT:    movntiq %rax, 24(%rdi)
-; AVX-NEXT:    movntiq %rax, 16(%rdi)
-; AVX-NEXT:    retq
-;
-; AVX512-LABEL: test_zero_v32i8_align1:
-; AVX512:       # %bb.0:
-; AVX512-NEXT:    xorl %eax, %eax
-; AVX512-NEXT:    movntiq %rax, 8(%rdi)
-; AVX512-NEXT:    movntiq %rax, (%rdi)
-; AVX512-NEXT:    movntiq %rax, 24(%rdi)
-; AVX512-NEXT:    movntiq %rax, 16(%rdi)
-; AVX512-NEXT:    retq
+; CHECK-LABEL: test_zero_v32i8_align1:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xorl %eax, %eax
+; CHECK-NEXT:    movntiq %rax, 8(%rdi)
+; CHECK-NEXT:    movntiq %rax, (%rdi)
+; CHECK-NEXT:    movntiq %rax, 24(%rdi)
+; CHECK-NEXT:    movntiq %rax, 16(%rdi)
+; CHECK-NEXT:    retq
   store <32 x i8> zeroinitializer, <32 x i8>* %dst, align 1, !nontemporal !1
   ret void
 }
 
 define void @test_zero_v4f64_align16(<4 x double>* %dst) nounwind {
 ; SSE-LABEL: test_zero_v4f64_align16:
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    xorps %xmm0, %xmm0

 ; AVX512-NEXT:    retq
   store <32 x i8> zeroinitializer, <32 x i8>* %dst, align 16, !nontemporal !1
   ret void
 }
 
 ; ZMM versions.
 
 define void @test_zero_v8f64_align1(<8 x double>* %dst) nounwind {
-; SSE-LABEL: test_zero_v8f64_align1:
-; SSE:       # %bb.0:
-; SSE-NEXT:    xorl %eax, %eax
-; SSE-NEXT:    movntiq %rax, 24(%rdi)
-; SSE-NEXT:    movntiq %rax, 16(%rdi)
-; SSE-NEXT:    movntiq %rax, 8(%rdi)
-; SSE-NEXT:    movntiq %rax, (%rdi)
-; SSE-NEXT:    movntiq %rax, 56(%rdi)
-; SSE-NEXT:    movntiq %rax, 48(%rdi)
-; SSE-NEXT:    movntiq %rax, 40(%rdi)
-; SSE-NEXT:    movntiq %rax, 32(%rdi)
-; SSE-NEXT:    retq
-;
-; AVX-LABEL: test_zero_v8f64_align1:
-; AVX:       # %bb.0:
-; AVX-NEXT:    xorl %eax, %eax
-; AVX-NEXT:    movntiq %rax, 24(%rdi)
-; AVX-NEXT:    movntiq %rax, 16(%rdi)
-; AVX-NEXT:    movntiq %rax, 8(%rdi)
-; AVX-NEXT:    movntiq %rax, (%rdi)
-; AVX-NEXT:    movntiq %rax, 56(%rdi)
-; AVX-NEXT:    movntiq %rax, 48(%rdi)
-; AVX-NEXT:    movntiq %rax, 40(%rdi)
-; AVX-NEXT:    movntiq %rax, 32(%rdi)
-; AVX-NEXT:    retq
-;
-; AVX512-LABEL: test_zero_v8f64_align1:
-; AVX512:       # %bb.0:
-; AVX512-NEXT:    xorl %eax, %eax
-; AVX512-NEXT:    movntiq %rax, 24(%rdi)
-; AVX512-NEXT:    movntiq %rax, 16(%rdi)
-; AVX512-NEXT:    movntiq %rax, 8(%rdi)
-; AVX512-NEXT:    movntiq %rax, (%rdi)
-; AVX512-NEXT:    movntiq %rax, 56(%rdi)
-; AVX512-NEXT:    movntiq %rax, 48(%rdi)
-; AVX512-NEXT:    movntiq %rax, 40(%rdi)
-; AVX512-NEXT:    movntiq %rax, 32(%rdi)
-; AVX512-NEXT:    retq
+; CHECK-LABEL: test_zero_v8f64_align1:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xorl %eax, %eax
+; CHECK-NEXT:    movntiq %rax, 24(%rdi)
+; CHECK-NEXT:    movntiq %rax, 16(%rdi)
+; CHECK-NEXT:    movntiq %rax, 8(%rdi)
+; CHECK-NEXT:    movntiq %rax, (%rdi)
+; CHECK-NEXT:    movntiq %rax, 56(%rdi)
+; CHECK-NEXT:    movntiq %rax, 48(%rdi)
+; CHECK-NEXT:    movntiq %rax, 40(%rdi)
+; CHECK-NEXT:    movntiq %rax, 32(%rdi)
+; CHECK-NEXT:    retq
   store <8 x double> zeroinitializer, <8 x double>* %dst, align 1, !nontemporal !1
   ret void
 }
 
 define void @test_zero_v16f32_align1(<16 x float>* %dst) nounwind {
-; SSE2-LABEL: test_zero_v16f32_align1:
-; SSE2:       # %bb.0:
-; SSE2-NEXT:    xorl %eax, %eax
-; SSE2-NEXT:    movntiq %rax, 24(%rdi)
-; SSE2-NEXT:    movntiq %rax, 16(%rdi)
-; SSE2-NEXT:    movntiq %rax, 8(%rdi)
-; SSE2-NEXT:    movntiq %rax, (%rdi)
-; SSE2-NEXT:    movntiq %rax, 56(%rdi)
-; SSE2-NEXT:    movntiq %rax, 48(%rdi)
-; SSE2-NEXT:    movntiq %rax, 40(%rdi)
-; SSE2-NEXT:    movntiq %rax, 32(%rdi)
-; SSE2-NEXT:    retq
-;
-; SSE4A-LABEL: test_zero_v16f32_align1:
-; SSE4A:       # %bb.0:
-; SSE4A-NEXT:    xorl %eax, %eax
-; SSE4A-NEXT:    movntiq %rax, 24(%rdi)
-; SSE4A-NEXT:    movntiq %rax, 8(%rdi)
-; SSE4A-NEXT:    movntiq %rax, 56(%rdi)
-; SSE4A-NEXT:    movntiq %rax, 40(%rdi)
-; SSE4A-NEXT:    xorps %xmm0, %xmm0
-; SSE4A-NEXT:    movntsd %xmm0, 16(%rdi)
-; SSE4A-NEXT:    movntsd %xmm0, (%rdi)
-; SSE4A-NEXT:    movntsd %xmm0, 48(%rdi)
-; SSE4A-NEXT:    movntsd %xmm0, 32(%rdi)
-; SSE4A-NEXT:    retq
-;
-; SSE41-LABEL: test_zero_v16f32_align1:
-; SSE41:       # %bb.0:
-; SSE41-NEXT:    xorl %eax, %eax
-; SSE41-NEXT:    movntiq %rax, 24(%rdi)
-; SSE41-NEXT:    movntiq %rax, 16(%rdi)
-; SSE41-NEXT:    movntiq %rax, 8(%rdi)
-; SSE41-NEXT:    movntiq %rax, (%rdi)
-; SSE41-NEXT:    movntiq %rax, 56(%rdi)
-; SSE41-NEXT:    movntiq %rax, 48(%rdi)
-; SSE41-NEXT:    movntiq %rax, 40(%rdi)
-; SSE41-NEXT:    movntiq %rax, 32(%rdi)
-; SSE41-NEXT:    retq
-;
-; AVX-LABEL: test_zero_v16f32_align1:
-; AVX:       # %bb.0:
-; AVX-NEXT:    xorl %eax, %eax
-; AVX-NEXT:    movntiq %rax, 24(%rdi)
-; AVX-NEXT:    movntiq %rax, 16(%rdi)
-; AVX-NEXT:    movntiq %rax, 8(%rdi)
-; AVX-NEXT:    movntiq %rax, (%rdi)
-; AVX-NEXT:    movntiq %rax, 56(%rdi)
-; AVX-NEXT:    movntiq %rax, 48(%rdi)
-; AVX-NEXT:    movntiq %rax, 40(%rdi)
-; AVX-NEXT:    movntiq %rax, 32(%rdi)
-; AVX-NEXT:    retq
-;
-; AVX512-LABEL: test_zero_v16f32_align1:
-; AVX512:       # %bb.0:
-; AVX512-NEXT:    xorl %eax, %eax
-; AVX512-NEXT:    movntiq %rax, 24(%rdi)
-; AVX512-NEXT:    movntiq %rax, 16(%rdi)
-; AVX512-NEXT:    movntiq %rax, 8(%rdi)
-; AVX512-NEXT:    movntiq %rax, (%rdi)
-; AVX512-NEXT:    movntiq %rax, 56(%rdi)
-; AVX512-NEXT:    movntiq %rax, 48(%rdi)
-; AVX512-NEXT:    movntiq %rax, 40(%rdi)
-; AVX512-NEXT:    movntiq %rax, 32(%rdi)
-; AVX512-NEXT:    retq
+; CHECK-LABEL: test_zero_v16f32_align1:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xorl %eax, %eax
+; CHECK-NEXT:    movntiq %rax, 24(%rdi)
+; CHECK-NEXT:    movntiq %rax, 16(%rdi)
+; CHECK-NEXT:    movntiq %rax, 8(%rdi)
+; CHECK-NEXT:    movntiq %rax, (%rdi)
+; CHECK-NEXT:    movntiq %rax, 56(%rdi)
+; CHECK-NEXT:    movntiq %rax, 48(%rdi)
+; CHECK-NEXT:    movntiq %rax, 40(%rdi)
+; CHECK-NEXT:    movntiq %rax, 32(%rdi)
+; CHECK-NEXT:    retq
   store <16 x float> zeroinitializer, <16 x float>* %dst, align 1, !nontemporal !1
   ret void
 }
 
 define void @test_zero_v8i64_align1(<8 x i64>* %dst) nounwind {
-; SSE2-LABEL: test_zero_v8i64_align1:
-; SSE2:       # %bb.0:
-; SSE2-NEXT:    xorl %eax, %eax
-; SSE2-NEXT:    movntiq %rax, 24(%rdi)
-; SSE2-NEXT:    movntiq %rax, 16(%rdi)
-; SSE2-NEXT:    movntiq %rax, 8(%rdi)
-; SSE2-NEXT:    movntiq %rax, (%rdi)
-; SSE2-NEXT:    movntiq %rax, 56(%rdi)
-; SSE2-NEXT:    movntiq %rax, 48(%rdi)
-; SSE2-NEXT:    movntiq %rax, 40(%rdi)
-; SSE2-NEXT:    movntiq %rax, 32(%rdi)
-; SSE2-NEXT:    retq
-;
-; SSE4A-LABEL: test_zero_v8i64_align1:
-; SSE4A:       # %bb.0:
-; SSE4A-NEXT:    xorps %xmm0, %xmm0
-; SSE4A-NEXT:    movntsd %xmm0, 24(%rdi)
-; SSE4A-NEXT:    movntsd %xmm0, 16(%rdi)
-; SSE4A-NEXT:    movntsd %xmm0, 8(%rdi)
-; SSE4A-NEXT:    movntsd %xmm0, (%rdi)
-; SSE4A-NEXT:    movntsd %xmm0, 56(%rdi)
-; SSE4A-NEXT:    movntsd %xmm0, 48(%rdi)
-; SSE4A-NEXT:    movntsd %xmm0, 40(%rdi)
-; SSE4A-NEXT:    movntsd %xmm0, 32(%rdi)
-; SSE4A-NEXT:    retq
-;
-; SSE41-LABEL: test_zero_v8i64_align1:
-; SSE41:       # %bb.0:
-; SSE41-NEXT:    xorl %eax, %eax
-; SSE41-NEXT:    movntiq %rax, 24(%rdi)
-; SSE41-NEXT:    movntiq %rax, 16(%rdi)
-; SSE41-NEXT:    movntiq %rax, 8(%rdi)
-; SSE41-NEXT:    movntiq %rax, (%rdi)
-; SSE41-NEXT:    movntiq %rax, 56(%rdi)
-; SSE41-NEXT:    movntiq %rax, 48(%rdi)
-; SSE41-NEXT:    movntiq %rax, 40(%rdi)
-; SSE41-NEXT:    movntiq %rax, 32(%rdi)
-; SSE41-NEXT:    retq
-;
-; AVX-LABEL: test_zero_v8i64_align1:
-; AVX:       # %bb.0:
-; AVX-NEXT:    xorl %eax, %eax
-; AVX-NEXT:    movntiq %rax, 24(%rdi)
-; AVX-NEXT:    movntiq %rax, 16(%rdi)
-; AVX-NEXT:    movntiq %rax, 8(%rdi)
-; AVX-NEXT:    movntiq %rax, (%rdi)
-; AVX-NEXT:    movntiq %rax, 56(%rdi)
-; AVX-NEXT:    movntiq %rax, 48(%rdi)
-; AVX-NEXT:    movntiq %rax, 40(%rdi)
-; AVX-NEXT:    movntiq %rax, 32(%rdi)
-; AVX-NEXT:    retq
-;
-; AVX512-LABEL: test_zero_v8i64_align1:
-; AVX512:       # %bb.0:
-; AVX512-NEXT:    xorl %eax, %eax
-; AVX512-NEXT:    movntiq %rax, 24(%rdi)
-; AVX512-NEXT:    movntiq %rax, 16(%rdi)
-; AVX512-NEXT:    movntiq %rax, 8(%rdi)
-; AVX512-NEXT:    movntiq %rax, (%rdi)
-; AVX512-NEXT:    movntiq %rax, 56(%rdi)
-; AVX512-NEXT:    movntiq %rax, 48(%rdi)
-; AVX512-NEXT:    movntiq %rax, 40(%rdi)
-; AVX512-NEXT:    movntiq %rax, 32(%rdi)
-; AVX512-NEXT:    retq
+; CHECK-LABEL: test_zero_v8i64_align1:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xorl %eax, %eax
+; CHECK-NEXT:    movntiq %rax, 24(%rdi)
+; CHECK-NEXT:    movntiq %rax, 16(%rdi)
+; CHECK-NEXT:    movntiq %rax, 8(%rdi)
+; CHECK-NEXT:    movntiq %rax, (%rdi)
+; CHECK-NEXT:    movntiq %rax, 56(%rdi)
+; CHECK-NEXT:    movntiq %rax, 48(%rdi)
+; CHECK-NEXT:    movntiq %rax, 40(%rdi)
+; CHECK-NEXT:    movntiq %rax, 32(%rdi)
+; CHECK-NEXT:    retq
   store <8 x i64> zeroinitializer, <8 x i64>* %dst, align 1, !nontemporal !1
   ret void
 }
 
 define void @test_zero_v16i32_align1(<16 x i32>* %dst) nounwind {
-; SSE2-LABEL: test_zero_v16i32_align1:
-; SSE2:       # %bb.0:
-; SSE2-NEXT:    xorl %eax, %eax
-; SSE2-NEXT:    movntiq %rax, 24(%rdi)
-; SSE2-NEXT:    movntiq %rax, 16(%rdi)
-; SSE2-NEXT:    movntiq %rax, 8(%rdi)
-; SSE2-NEXT:    movntiq %rax, (%rdi)
-; SSE2-NEXT:    movntiq %rax, 56(%rdi)
-; SSE2-NEXT:    movntiq %rax, 48(%rdi)
-; SSE2-NEXT:    movntiq %rax, 40(%rdi)
-; SSE2-NEXT:    movntiq %rax, 32(%rdi)
-; SSE2-NEXT:    retq
-;
-; SSE4A-LABEL: test_zero_v16i32_align1:
-; SSE4A:       # %bb.0:
-; SSE4A-NEXT:    xorps %xmm0, %xmm0
-; SSE4A-NEXT:    movntsd %xmm0, 24(%rdi)
-; SSE4A-NEXT:    movntsd %xmm0, 16(%rdi)
-; SSE4A-NEXT:    movntsd %xmm0, 8(%rdi)
-; SSE4A-NEXT:    movntsd %xmm0, (%rdi)
-; SSE4A-NEXT:    movntsd %xmm0, 56(%rdi)
-; SSE4A-NEXT:    movntsd %xmm0, 48(%rdi)
-; SSE4A-NEXT:    movntsd %xmm0, 40(%rdi)
-; SSE4A-NEXT:    movntsd %xmm0, 32(%rdi)
-; SSE4A-NEXT:    retq
-;
-; SSE41-LABEL: test_zero_v16i32_align1:
-; SSE41:       # %bb.0:
-; SSE41-NEXT:    xorl %eax, %eax
-; SSE41-NEXT:    movntiq %rax, 24(%rdi)
-; SSE41-NEXT:    movntiq %rax, 16(%rdi)
-; SSE41-NEXT:    movntiq %rax, 8(%rdi)
-; SSE41-NEXT:    movntiq %rax, (%rdi)
-; SSE41-NEXT:    movntiq %rax, 56(%rdi)
-; SSE41-NEXT:    movntiq %rax, 48(%rdi)
-; SSE41-NEXT:    movntiq %rax, 40(%rdi)
-; SSE41-NEXT:    movntiq %rax, 32(%rdi)
-; SSE41-NEXT:    retq
-;
-; AVX-LABEL: test_zero_v16i32_align1:
-; AVX:       # %bb.0:
-; AVX-NEXT:    xorl %eax, %eax
-; AVX-NEXT:    movntiq %rax, 24(%rdi)
-; AVX-NEXT:    movntiq %rax, 16(%rdi)
-; AVX-NEXT:    movntiq %rax, 8(%rdi)
-; AVX-NEXT:    movntiq %rax, (%rdi)
-; AVX-NEXT:    movntiq %rax, 56(%rdi)
-; AVX-NEXT:    movntiq %rax, 48(%rdi)
-; AVX-NEXT:    movntiq %rax, 40(%rdi)
-; AVX-NEXT:    movntiq %rax, 32(%rdi)
-; AVX-NEXT:    retq
-;
-; AVX512-LABEL: test_zero_v16i32_align1:
-; AVX512:       # %bb.0:
-; AVX512-NEXT:    xorl %eax, %eax
-; AVX512-NEXT:    movntiq %rax, 24(%rdi)
-; AVX512-NEXT:    movntiq %rax, 16(%rdi)
-; AVX512-NEXT:    movntiq %rax, 8(%rdi)
-; AVX512-NEXT:    movntiq %rax, (%rdi)
-; AVX512-NEXT:    movntiq %rax, 56(%rdi)
-; AVX512-NEXT:    movntiq %rax, 48(%rdi)
-; AVX512-NEXT:    movntiq %rax, 40(%rdi)
-; AVX512-NEXT:    movntiq %rax, 32(%rdi)
-; AVX512-NEXT:    retq
+; CHECK-LABEL: test_zero_v16i32_align1:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xorl %eax, %eax
+; CHECK-NEXT:    movntiq %rax, 24(%rdi)
+; CHECK-NEXT:    movntiq %rax, 16(%rdi)
+; CHECK-NEXT:    movntiq %rax, 8(%rdi)
+; CHECK-NEXT:    movntiq %rax, (%rdi)
+; CHECK-NEXT:    movntiq %rax, 56(%rdi)
+; CHECK-NEXT:    movntiq %rax, 48(%rdi)
+; CHECK-NEXT:    movntiq %rax, 40(%rdi)
+; CHECK-NEXT:    movntiq %rax, 32(%rdi)
+; CHECK-NEXT:    retq
   store <16 x i32> zeroinitializer, <16 x i32>* %dst, align 1, !nontemporal !1
   ret void
 }
 
 define void @test_zero_v32i16_align1(<32 x i16>* %dst) nounwind {
-; SSE2-LABEL: test_zero_v32i16_align1:
-; SSE2:       # %bb.0:
-; SSE2-NEXT:    xorl %eax, %eax
-; SSE2-NEXT:    movntiq %rax, 24(%rdi)
-; SSE2-NEXT:    movntiq %rax, 16(%rdi)
-; SSE2-NEXT:    movntiq %rax, 8(%rdi)
-; SSE2-NEXT:    movntiq %rax, (%rdi)
-; SSE2-NEXT:    movntiq %rax, 56(%rdi)
-; SSE2-NEXT:    movntiq %rax, 48(%rdi)
-; SSE2-NEXT:    movntiq %rax, 40(%rdi)
-; SSE2-NEXT:    movntiq %rax, 32(%rdi)
-; SSE2-NEXT:    retq
-;
-; SSE4A-LABEL: test_zero_v32i16_align1:
-; SSE4A:       # %bb.0:
-; SSE4A-NEXT:    xorps %xmm0, %xmm0
-; SSE4A-NEXT:    movntsd %xmm0, 24(%rdi)
-; SSE4A-NEXT:    movntsd %xmm0, 16(%rdi)
-; SSE4A-NEXT:    movntsd %xmm0, 8(%rdi)
-; SSE4A-NEXT:    movntsd %xmm0, (%rdi)
-; SSE4A-NEXT:    movntsd %xmm0, 56(%rdi)
-; SSE4A-NEXT:    movntsd %xmm0, 48(%rdi)
-; SSE4A-NEXT:    movntsd %xmm0, 40(%rdi)
-; SSE4A-NEXT:    movntsd %xmm0, 32(%rdi)
-; SSE4A-NEXT:    retq
-;
-; SSE41-LABEL: test_zero_v32i16_align1:
-; SSE41:       # %bb.0:
-; SSE41-NEXT:    xorl %eax, %eax
-; SSE41-NEXT:    movntiq %rax, 24(%rdi)
-; SSE41-NEXT:    movntiq %rax, 16(%rdi)
-; SSE41-NEXT:    movntiq %rax, 8(%rdi)
-; SSE41-NEXT:    movntiq %rax, (%rdi)
-; SSE41-NEXT:    movntiq %rax, 56(%rdi)
-; SSE41-NEXT:    movntiq %rax, 48(%rdi)
-; SSE41-NEXT:    movntiq %rax, 40(%rdi)
-; SSE41-NEXT:    movntiq %rax, 32(%rdi)
-; SSE41-NEXT:    retq
-;
-; AVX-LABEL: test_zero_v32i16_align1:
-; AVX:       # %bb.0:
-; AVX-NEXT:    xorl %eax, %eax
-; AVX-NEXT:    movntiq %rax, 24(%rdi)
-; AVX-NEXT:    movntiq %rax, 16(%rdi)
-; AVX-NEXT:    movntiq %rax, 8(%rdi)
-; AVX-NEXT:    movntiq %rax, (%rdi)
-; AVX-NEXT:    movntiq %rax, 56(%rdi)
-; AVX-NEXT:    movntiq %rax, 48(%rdi)
-; AVX-NEXT:    movntiq %rax, 40(%rdi)
-; AVX-NEXT:    movntiq %rax, 32(%rdi)
-; AVX-NEXT:    retq
-;
-; AVX512-LABEL: test_zero_v32i16_align1:
-; AVX512:       # %bb.0:
-; AVX512-NEXT:    xorl %eax, %eax
-; AVX512-NEXT:    movntiq %rax, 24(%rdi)
-; AVX512-NEXT:    movntiq %rax, 16(%rdi)
-; AVX512-NEXT:    movntiq %rax, 8(%rdi)
-; AVX512-NEXT:    movntiq %rax, (%rdi)
-; AVX512-NEXT:    movntiq %rax, 56(%rdi)
-; AVX512-NEXT:    movntiq %rax, 48(%rdi)
-; AVX512-NEXT:    movntiq %rax, 40(%rdi)
-; AVX512-NEXT:    movntiq %rax, 32(%rdi)
-; AVX512-NEXT:    retq
+; CHECK-LABEL: test_zero_v32i16_align1:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xorl %eax, %eax
+; CHECK-NEXT:    movntiq %rax, 24(%rdi)
+; CHECK-NEXT:    movntiq %rax, 16(%rdi)
+; CHECK-NEXT:    movntiq %rax, 8(%rdi)
+; CHECK-NEXT:    movntiq %rax, (%rdi)
+; CHECK-NEXT:    movntiq %rax, 56(%rdi)
+; CHECK-NEXT:    movntiq %rax, 48(%rdi)
+; CHECK-NEXT:    movntiq %rax, 40(%rdi)
+; CHECK-NEXT:    movntiq %rax, 32(%rdi)
+; CHECK-NEXT:    retq
   store <32 x i16> zeroinitializer, <32 x i16>* %dst, align 1, !nontemporal !1
   ret void
 }
 
 define void @test_zero_v64i8_align1(<64 x i8>* %dst) nounwind {
-; SSE2-LABEL: test_zero_v64i8_align1:
-; SSE2:       # %bb.0:
-; SSE2-NEXT:    xorl %eax, %eax
-; SSE2-NEXT:    movntiq %rax, 24(%rdi)
-; SSE2-NEXT:    movntiq %rax, 16(%rdi)
-; SSE2-NEXT:    movntiq %rax, 8(%rdi)
-; SSE2-NEXT:    movntiq %rax, (%rdi)
-; SSE2-NEXT:    movntiq %rax, 56(%rdi)
-; SSE2-NEXT:    movntiq %rax, 48(%rdi)
-; SSE2-NEXT:    movntiq %rax, 40(%rdi)
-; SSE2-NEXT:    movntiq %rax, 32(%rdi)
-; SSE2-NEXT:    retq
-;
-; SSE4A-LABEL: test_zero_v64i8_align1:
-; SSE4A:       # %bb.0:
-; SSE4A-NEXT:    xorps %xmm0, %xmm0
-; SSE4A-NEXT:    movntsd %xmm0, 24(%rdi)
-; SSE4A-NEXT:    movntsd %xmm0, 16(%rdi)
-; SSE4A-NEXT:    movntsd %xmm0, 8(%rdi)
-; SSE4A-NEXT:    movntsd %xmm0, (%rdi)
-; SSE4A-NEXT:    movntsd %xmm0, 56(%rdi)
-; SSE4A-NEXT:    movntsd %xmm0, 48(%rdi)
-; SSE4A-NEXT:    movntsd %xmm0, 40(%rdi)
-; SSE4A-NEXT:    movntsd %xmm0, 32(%rdi)
-; SSE4A-NEXT:    retq
-;
-; SSE41-LABEL: test_zero_v64i8_align1:
-; SSE41:       # %bb.0:
-; SSE41-NEXT:    xorl %eax, %eax
-; SSE41-NEXT:    movntiq %rax, 24(%rdi)
-; SSE41-NEXT:    movntiq %rax, 16(%rdi)
-; SSE41-NEXT:    movntiq %rax, 8(%rdi)
-; SSE41-NEXT:    movntiq %rax, (%rdi)
-; SSE41-NEXT:    movntiq %rax, 56(%rdi)
-; SSE41-NEXT:    movntiq %rax, 48(%rdi)
-; SSE41-NEXT:    movntiq %rax, 40(%rdi)
-; SSE41-NEXT:    movntiq %rax, 32(%rdi)
-; SSE41-NEXT:    retq
-;
-; AVX-LABEL: test_zero_v64i8_align1:
-; AVX:       # %bb.0:
-; AVX-NEXT:    xorl %eax, %eax
-; AVX-NEXT:    movntiq %rax, 24(%rdi)
-; AVX-NEXT:    movntiq %rax, 16(%rdi)
-; AVX-NEXT:    movntiq %rax, 8(%rdi)
-; AVX-NEXT:    movntiq %rax, (%rdi)
-; AVX-NEXT:    movntiq %rax, 56(%rdi)
-; AVX-NEXT:    movntiq %rax, 48(%rdi)
-; AVX-NEXT:    movntiq %rax, 40(%rdi)
-; AVX-NEXT:    movntiq %rax, 32(%rdi)
-; AVX-NEXT:    retq
-;
-; AVX512-LABEL: test_zero_v64i8_align1:
-; AVX512:       # %bb.0:
-; AVX512-NEXT:    xorl %eax, %eax
-; AVX512-NEXT:    movntiq %rax, 24(%rdi)
-; AVX512-NEXT:    movntiq %rax, 16(%rdi)
-; AVX512-NEXT:    movntiq %rax, 8(%rdi)
-; AVX512-NEXT:    movntiq %rax, (%rdi)
-; AVX512-NEXT:    movntiq %rax, 56(%rdi)
-; AVX512-NEXT:    movntiq %rax, 48(%rdi)
-; AVX512-NEXT:    movntiq %rax, 40(%rdi)
-; AVX512-NEXT:    movntiq %rax, 32(%rdi)
-; AVX512-NEXT:    retq
+; CHECK-LABEL: test_zero_v64i8_align1:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    xorl %eax, %eax
+; CHECK-NEXT:    movntiq %rax, 24(%rdi)
+; CHECK-NEXT:    movntiq %rax, 16(%rdi)
+; CHECK-NEXT:    movntiq %rax, 8(%rdi)
+; CHECK-NEXT:    movntiq %rax, (%rdi)
+; CHECK-NEXT:    movntiq %rax, 56(%rdi)
+; CHECK-NEXT:    movntiq %rax, 48(%rdi)
+; CHECK-NEXT:    movntiq %rax, 40(%rdi)
+; CHECK-NEXT:    movntiq %rax, 32(%rdi)
+; CHECK-NEXT:    retq
   store <64 x i8> zeroinitializer, <64 x i8>* %dst, align 1, !nontemporal !1
   ret void
 }
 
 define void @test_zero_v8f64_align16(<8 x double>* %dst) nounwind {
 ; SSE-LABEL: test_zero_v8f64_align16:
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    xorps %xmm0, %xmm0

llvm/test/CodeGen/X86/pr41619.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s -mtriple=x86_64-apple-macosx10.14.0 -mattr=avx2 | FileCheck %s --check-prefixes=CHECK,AVX
 ; RUN: llc < %s -mtriple=x86_64-apple-macosx10.14.0 -mattr=avx512bw | FileCheck %s --check-prefixes=CHECK,AVX512
 
 define void @foo(double %arg) {
 ; CHECK-LABEL: foo:
 ; CHECK:       ## %bb.0: ## %bb
 ; CHECK-NEXT:    vmovq %xmm0, %rax
 ; CHECK-NEXT:    vmovd %eax, %xmm0
-; CHECK-NEXT:    vxorps %xmm1, %xmm1, %xmm1
 ; CHECK-NEXT:    vmovq %xmm0, %rax
 ; CHECK-NEXT:    movl %eax, (%rax)
-; CHECK-NEXT:    vmovlps %xmm1, (%rax)
+; CHECK-NEXT:    movq $0, (%rax)
 ; CHECK-NEXT:    retq
 bb:
   %tmp = bitcast double %arg to i64
   %tmp1 = trunc i64 %tmp to i32
   %tmp2 = bitcast i32 %tmp1 to float
   %tmp3 = insertelement <4 x float> zeroinitializer, float %tmp2, i32 2
   %tmp4 = bitcast <4 x float> %tmp3 to <2 x double>
   %tmp5 = extractelement <2 x double> %tmp4, i32 0

llvm/test/CodeGen/X86/promote-cmp.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s -mtriple=x86_64-- -mattr=+sse2   | FileCheck %s --check-prefix=SSE2
 ; RUN: llc < %s -mtriple=x86_64-- -mattr=+sse4.2 | FileCheck %s --check-prefix=SSE4
 ; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx    | FileCheck %s --check-prefix=AVX1
 ; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx2   | FileCheck %s --check-prefix=AVX2
 
 define <4 x i64> @PR45808(<4 x i64> %0, <4 x i64> %1) {
 ; SSE2-LABEL: PR45808:
 ; SSE2:       # %bb.0:
 ; SSE2-NEXT:    movdqa {{.*#+}} xmm4 = [2147483648,2147483648]
-; SSE2-NEXT:    movdqa %xmm3, %xmm9
-; SSE2-NEXT:    pxor %xmm4, %xmm9
-; SSE2-NEXT:    movdqa %xmm1, %xmm6
+; SSE2-NEXT:    movdqa %xmm2, %xmm5
+; SSE2-NEXT:    pxor %xmm4, %xmm5
+; SSE2-NEXT:    movdqa %xmm0, %xmm6
 ; SSE2-NEXT:    pxor %xmm4, %xmm6
-; SSE2-NEXT:    movdqa %xmm6, %xmm8
-; SSE2-NEXT:    pcmpgtd %xmm9, %xmm8
-; SSE2-NEXT:    movdqa %xmm2, %xmm7
-; SSE2-NEXT:    pxor %xmm4, %xmm7
-; SSE2-NEXT:    pxor %xmm0, %xmm4
-; SSE2-NEXT:    movdqa %xmm4, %xmm5
-; SSE2-NEXT:    pcmpgtd %xmm7, %xmm5
-; SSE2-NEXT:    movdqa %xmm5, %xmm10
-; SSE2-NEXT:    shufps {{.*#+}} xmm10 = xmm10[0,2],xmm8[0,2]
-; SSE2-NEXT:    pcmpeqd %xmm9, %xmm6
-; SSE2-NEXT:    pcmpeqd %xmm7, %xmm4
-; SSE2-NEXT:    shufps {{.*#+}} xmm4 = xmm4[1,3],xmm6[1,3]
-; SSE2-NEXT:    andps %xmm10, %xmm4
-; SSE2-NEXT:    shufps {{.*#+}} xmm5 = xmm5[1,3],xmm8[1,3]
-; SSE2-NEXT:    orps %xmm4, %xmm5
-; SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm5[2,1,3,3]
-; SSE2-NEXT:    psllq $63, %xmm4
-; SSE2-NEXT:    psrad $31, %xmm4
+; SSE2-NEXT:    movdqa %xmm6, %xmm7
+; SSE2-NEXT:    pcmpgtd %xmm5, %xmm7
+; SSE2-NEXT:    pcmpeqd %xmm5, %xmm6
+; SSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm6[1,1,3,3]
+; SSE2-NEXT:    pand %xmm7, %xmm5
+; SSE2-NEXT:    pshufd {{.*#+}} xmm6 = xmm7[1,1,3,3]
+; SSE2-NEXT:    por %xmm5, %xmm6
+; SSE2-NEXT:    movdqa %xmm3, %xmm5
+; SSE2-NEXT:    pxor %xmm4, %xmm5
+; SSE2-NEXT:    pxor %xmm1, %xmm4
+; SSE2-NEXT:    movdqa %xmm4, %xmm7
+; SSE2-NEXT:    pcmpgtd %xmm5, %xmm7
+; SSE2-NEXT:    pshufd {{.*#+}} xmm8 = xmm7[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm5, %xmm4
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[1,1,3,3]
-; SSE2-NEXT:    pand %xmm4, %xmm1
-; SSE2-NEXT:    pandn %xmm3, %xmm4
-; SSE2-NEXT:    por %xmm4, %xmm1
-; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm5[0,1,1,3]
-; SSE2-NEXT:    pxor {{.*}}(%rip), %xmm3
-; SSE2-NEXT:    psllq $63, %xmm3
-; SSE2-NEXT:    psrad $31, %xmm3
-; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[1,1,3,3]
+; SSE2-NEXT:    pand %xmm8, %xmm4
+; SSE2-NEXT:    pshufd {{.*#+}} xmm5 = xmm7[1,1,3,3]
+; SSE2-NEXT:    por %xmm4, %xmm5
+; SSE2-NEXT:    pand %xmm5, %xmm1
+; SSE2-NEXT:    pandn %xmm3, %xmm5
+; SSE2-NEXT:    por %xmm5, %xmm1
+; SSE2-NEXT:    pxor {{.*}}(%rip), %xmm6
+; SSE2-NEXT:    psllq $63, %xmm6
+; SSE2-NEXT:    psrad $31, %xmm6
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm6[1,1,3,3]
 ; SSE2-NEXT:    pand %xmm3, %xmm0
 ; SSE2-NEXT:    pandn %xmm2, %xmm3
 ; SSE2-NEXT:    por %xmm3, %xmm0
 ; SSE2-NEXT:    retq
 ;
 ; SSE4-LABEL: PR45808:
 ; SSE4:       # %bb.0:
 ; SSE4-NEXT:    movdqa %xmm0, %xmm4
 ; SSE4-NEXT:    movdqa %xmm1, %xmm0
 ; SSE4-NEXT:    pcmpgtq %xmm3, %xmm0
 ; SSE4-NEXT:    movdqa %xmm4, %xmm5
 ; SSE4-NEXT:    pcmpgtq %xmm2, %xmm5
 ; SSE4-NEXT:    pshufd {{.*#+}} xmm5 = xmm5[0,2,2,3]
 ; SSE4-NEXT:    pcmpeqd %xmm6, %xmm6
 ; SSE4-NEXT:    pxor %xmm5, %xmm6
-; SSE4-NEXT:    psllq $63, %xmm0
 ; SSE4-NEXT:    blendvpd %xmm0, %xmm1, %xmm3
 ; SSE4-NEXT:    pmovzxdq {{.*#+}} xmm0 = xmm6[0],zero,xmm6[1],zero
 ; SSE4-NEXT:    psllq $63, %xmm0
 ; SSE4-NEXT:    blendvpd %xmm0, %xmm4, %xmm2
 ; SSE4-NEXT:    movapd %xmm2, %xmm0
 ; SSE4-NEXT:    movapd %xmm3, %xmm1
 ; SSE4-NEXT:    retq
 ;

llvm/test/CodeGen/X86/srem-seteq-illegal-types.ll

 ; SSE2-NEXT:    movabsq $2049638230412172402, %r8 # imm = 0x1C71C71C71C71C72
 ; SSE2-NEXT:    movq %rsi, %rax
 ; SSE2-NEXT:    imulq %r8
 ; SSE2-NEXT:    movq %rdx, %rax
 ; SSE2-NEXT:    shrq $63, %rax
 ; SSE2-NEXT:    addq %rdx, %rax
 ; SSE2-NEXT:    leaq (%rax,%rax,8), %rax
 ; SSE2-NEXT:    subq %rax, %rsi
-; SSE2-NEXT:    movq %rsi, %xmm1
+; SSE2-NEXT:    movq %rsi, %xmm0
 ; SSE2-NEXT:    movq %rdi, %rax
 ; SSE2-NEXT:    imulq %r8
 ; SSE2-NEXT:    movq %rdx, %rax
 ; SSE2-NEXT:    shrq $63, %rax
 ; SSE2-NEXT:    addq %rdx, %rax
 ; SSE2-NEXT:    leaq (%rax,%rax,8), %rax
 ; SSE2-NEXT:    subq %rax, %rdi
-; SSE2-NEXT:    movq %rdi, %xmm0
-; SSE2-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
-; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [8589934591,8589934591]
-; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    movq %rdi, %xmm1
+; SSE2-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm0[0]
+; SSE2-NEXT:    movdqa {{.*#+}} xmm0 = [8589934591,8589934591]
+; SSE2-NEXT:    pand %xmm0, %xmm1
 ; SSE2-NEXT:    movabsq $2049638230412172401, %rdx # imm = 0x1C71C71C71C71C71
 ; SSE2-NEXT:    movq %rcx, %rax
 ; SSE2-NEXT:    imulq %rdx
 ; SSE2-NEXT:    subq %rcx, %rdx
 ; SSE2-NEXT:    movq %rdx, %rax
 ; SSE2-NEXT:    shrq $63, %rax
 ; SSE2-NEXT:    sarq $3, %rdx
 ; SSE2-NEXT:    addq %rax, %rdx
 ; SSE2-NEXT:    leaq (%rdx,%rdx,8), %rax
 ; SSE2-NEXT:    addq %rcx, %rax
 ; SSE2-NEXT:    movq %rax, %xmm2
-; SSE2-NEXT:    pand %xmm1, %xmm2
-; SSE2-NEXT:    pcmpeqd {{.*}}(%rip), %xmm0
-; SSE2-NEXT:    pcmpeqd {{.*}}(%rip), %xmm2
-; SSE2-NEXT:    movdqa %xmm0, %xmm1
-; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,3],xmm2[1,2]
-; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,3]
-; SSE2-NEXT:    andps %xmm1, %xmm0
+; SSE2-NEXT:    pand %xmm0, %xmm2
+; SSE2-NEXT:    pcmpeqd {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[1,0,3,2]
+; SSE2-NEXT:    pand %xmm1, %xmm0
 ; SSE2-NEXT:    pcmpeqd %xmm1, %xmm1
-; SSE2-NEXT:    pxor %xmm0, %xmm1
-; SSE2-NEXT:    movdqa %xmm1, -{{[0-9]+}}(%rsp)
-; SSE2-NEXT:    movb -{{[0-9]+}}(%rsp), %al
+; SSE2-NEXT:    pxor %xmm1, %xmm0
+; SSE2-NEXT:    pcmpeqd {{.*}}(%rip), %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[1,0,3,2]
+; SSE2-NEXT:    pand %xmm2, %xmm3
+; SSE2-NEXT:    pxor %xmm1, %xmm3
+; SSE2-NEXT:    movd %xmm0, %eax
+; SSE2-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,2],xmm3[0,2]
+; SSE2-NEXT:    movaps %xmm0, -{{[0-9]+}}(%rsp)
 ; SSE2-NEXT:    movb -{{[0-9]+}}(%rsp), %dl
 ; SSE2-NEXT:    movb -{{[0-9]+}}(%rsp), %cl
+; SSE2-NEXT:    # kill: def $al killed $al killed $eax
 ; SSE2-NEXT:    retq
 ;
 ; SSE41-LABEL: test_srem_vec:
 ; SSE41:       # %bb.0:
 ; SSE41-NEXT:    movq %rdx, %rcx
 ; SSE41-NEXT:    shlq $31, %rcx
 ; SSE41-NEXT:    sarq $31, %rcx
 ; SSE41-NEXT:    shlq $31, %rdi

 ; SSE41-NEXT:    addq %rcx, %rax
 ; SSE41-NEXT:    movq %rax, %xmm2
 ; SSE41-NEXT:    pand %xmm1, %xmm2
 ; SSE41-NEXT:    pcmpeqq {{.*}}(%rip), %xmm0
 ; SSE41-NEXT:    pcmpeqd %xmm1, %xmm1
 ; SSE41-NEXT:    pxor %xmm1, %xmm0
 ; SSE41-NEXT:    pcmpeqq {{.*}}(%rip), %xmm2
 ; SSE41-NEXT:    pxor %xmm1, %xmm2
-; SSE41-NEXT:    pextrb $0, %xmm0, %eax
+; SSE41-NEXT:    movd %xmm0, %eax
 ; SSE41-NEXT:    pextrb $8, %xmm0, %edx
 ; SSE41-NEXT:    pextrb $0, %xmm2, %ecx
 ; SSE41-NEXT:    # kill: def $al killed $al killed $eax
 ; SSE41-NEXT:    # kill: def $dl killed $dl killed $edx
 ; SSE41-NEXT:    # kill: def $cl killed $cl killed $ecx
 ; SSE41-NEXT:    retq
 ;
 ; AVX1-LABEL: test_srem_vec:

 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX1-NEXT:    vandps {{.*}}(%rip), %ymm0, %ymm0
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vpcmpeqq {{.*}}(%rip), %xmm1, %xmm1
 ; AVX1-NEXT:    vpcmpeqd %xmm2, %xmm2, %xmm2
 ; AVX1-NEXT:    vpxor %xmm2, %xmm1, %xmm1
 ; AVX1-NEXT:    vpcmpeqq {{.*}}(%rip), %xmm0, %xmm0
 ; AVX1-NEXT:    vpxor %xmm2, %xmm0, %xmm0
-; AVX1-NEXT:    vpextrb $0, %xmm0, %eax
+; AVX1-NEXT:    vmovd %xmm0, %eax
 ; AVX1-NEXT:    vpextrb $8, %xmm0, %edx
 ; AVX1-NEXT:    vpextrb $0, %xmm1, %ecx
 ; AVX1-NEXT:    # kill: def $al killed $al killed $eax
 ; AVX1-NEXT:    # kill: def $dl killed $dl killed $edx
 ; AVX1-NEXT:    # kill: def $cl killed $cl killed $ecx
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;

 ; AVX2-NEXT:    vmovq %rax, %xmm1
 ; AVX2-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [8589934591,8589934591,8589934591,8589934591]
 ; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
 ; AVX2-NEXT:    vpcmpeqq {{.*}}(%rip), %ymm0, %ymm0
 ; AVX2-NEXT:    vpcmpeqd %ymm1, %ymm1, %ymm1
 ; AVX2-NEXT:    vpxor %ymm1, %ymm0, %ymm0
 ; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vpextrb $0, %xmm0, %eax
+; AVX2-NEXT:    vmovd %xmm0, %eax
 ; AVX2-NEXT:    vpextrb $8, %xmm0, %edx
 ; AVX2-NEXT:    vpextrb $0, %xmm1, %ecx
 ; AVX2-NEXT:    # kill: def $al killed $al killed $eax
 ; AVX2-NEXT:    # kill: def $dl killed $dl killed $edx
 ; AVX2-NEXT:    # kill: def $cl killed $cl killed $ecx
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
 ;

llvm/test/CodeGen/X86/vec_setcc.ll

 ; This asserted in type legalization for v3i1 setcc after v3i16 was made
 ; a simple value type.
 define <3 x i1> @test_setcc_v3i1_v3i16(<3 x i16>* %a) nounwind {
 ; SSE2-LABEL: test_setcc_v3i1_v3i16:
 ; SSE2:       # %bb.0:
 ; SSE2-NEXT:    movq {{.*#+}} xmm0 = mem[0],zero
 ; SSE2-NEXT:    pxor %xmm1, %xmm1
 ; SSE2-NEXT:    pcmpeqw %xmm0, %xmm1
+; SSE2-NEXT:    movd %xmm1, %eax
 ; SSE2-NEXT:    punpcklwd {{.*#+}} xmm1 = xmm1[0,0,1,1,2,2,3,3]
 ; SSE2-NEXT:    movdqa %xmm1, -{{[0-9]+}}(%rsp)
-; SSE2-NEXT:    movb -{{[0-9]+}}(%rsp), %al
 ; SSE2-NEXT:    movb -{{[0-9]+}}(%rsp), %dl
 ; SSE2-NEXT:    movb -{{[0-9]+}}(%rsp), %cl
+; SSE2-NEXT:    # kill: def $al killed $al killed $eax
 ; SSE2-NEXT:    retq
 ;
 ; SSE41-LABEL: test_setcc_v3i1_v3i16:
 ; SSE41:       # %bb.0:
 ; SSE41-NEXT:    movq {{.*#+}} xmm0 = mem[0],zero
 ; SSE41-NEXT:    pxor %xmm1, %xmm1
 ; SSE41-NEXT:    pcmpeqw %xmm0, %xmm1
-; SSE41-NEXT:    pextrb $0, %xmm1, %eax
-; SSE41-NEXT:    pextrb $2, %xmm1, %edx
-; SSE41-NEXT:    pextrb $4, %xmm1, %ecx
+; SSE41-NEXT:    pmovsxwd %xmm1, %xmm0
+; SSE41-NEXT:    movd %xmm1, %eax
+; SSE41-NEXT:    pextrb $4, %xmm0, %edx
+; SSE41-NEXT:    pextrb $8, %xmm0, %ecx
 ; SSE41-NEXT:    # kill: def $al killed $al killed $eax
 ; SSE41-NEXT:    # kill: def $dl killed $dl killed $edx
 ; SSE41-NEXT:    # kill: def $cl killed $cl killed $ecx
 ; SSE41-NEXT:    retq
 ;
 ; AVX-LABEL: test_setcc_v3i1_v3i16:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vmovq {{.*#+}} xmm0 = mem[0],zero
 ; AVX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
 ; AVX-NEXT:    vpcmpeqw %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpextrb $0, %xmm0, %eax
-; AVX-NEXT:    vpextrb $2, %xmm0, %edx
-; AVX-NEXT:    vpextrb $4, %xmm0, %ecx
+; AVX-NEXT:    vpmovsxwd %xmm0, %xmm1
+; AVX-NEXT:    vmovd %xmm0, %eax
+; AVX-NEXT:    vpextrb $4, %xmm1, %edx
+; AVX-NEXT:    vpextrb $8, %xmm1, %ecx
 ; AVX-NEXT:    # kill: def $al killed $al killed $eax
 ; AVX-NEXT:    # kill: def $dl killed $dl killed $edx
 ; AVX-NEXT:    # kill: def $cl killed $cl killed $ecx
 ; AVX-NEXT:    retq
   %b = load <3 x i16>, <3 x i16>* %a
   %cmp = icmp eq <3 x i16> %b, <i16 0, i16 0, i16 0>
   ret <3 x i1> %cmp
 }

llvm/test/CodeGen/X86/vec_zero_cse.ll

 @S1 = external dso_local global <2 x i64>
 @S2 = external dso_local global <4 x i32>
 
 define void @test1() {
 ; X32-LABEL: test1:
 ; X32:       # %bb.0:
 ; X32-NEXT:    movl $0, M1+4
 ; X32-NEXT:    movl $0, M1
-; X32-NEXT:    xorps %xmm0, %xmm0
-; X32-NEXT:    movlps %xmm0, M2
+; X32-NEXT:    movl $0, M2+4
+; X32-NEXT:    movl $0, M2
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: test1:
 ; X64:       # %bb.0:
 ; X64-NEXT:    movq $0, {{.*}}(%rip)
 ; X64-NEXT:    movq $0, {{.*}}(%rip)
 ; X64-NEXT:    retq
   store <1 x i64> zeroinitializer, <1 x i64>* @M1
   store <2 x i32> zeroinitializer, <2 x i32>* @M2
   ret void
 }
 
 define void @test2() {
 ; X32-LABEL: test2:
 ; X32:       # %bb.0:
 ; X32-NEXT:    movl $-1, M1+4
 ; X32-NEXT:    movl $-1, M1
-; X32-NEXT:    pcmpeqd %xmm0, %xmm0
-; X32-NEXT:    movq %xmm0, M2
+; X32-NEXT:    movl $-1, M2+4
+; X32-NEXT:    movl $-1, M2
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: test2:
 ; X64:       # %bb.0:
 ; X64-NEXT:    movq $-1, {{.*}}(%rip)
 ; X64-NEXT:    movq $-1, {{.*}}(%rip)
 ; X64-NEXT:    retq
   store <1 x i64> < i64 -1 >, <1 x i64>* @M1

llvm/test/CodeGen/X86/vector-fshl-128.ll

 ; XOPAVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [63,63]
 ; XOPAVX2-NEXT:    vpandn %xmm3, %xmm2, %xmm4
 ; XOPAVX2-NEXT:    vpsrlq $1, %xmm1, %xmm1
 ; XOPAVX2-NEXT:    vpsrlvq %xmm4, %xmm1, %xmm1
 ; XOPAVX2-NEXT:    vpand %xmm3, %xmm2, %xmm2
 ; XOPAVX2-NEXT:    vpsllvq %xmm2, %xmm0, %xmm0
 ; XOPAVX2-NEXT:    vpor %xmm1, %xmm0, %xmm0
 ; XOPAVX2-NEXT:    retq
 ;
 ; X86-SSE2-LABEL: var_funnnel_v2i64:
 ; X86-SSE2:       # %bb.0:
 ; X86-SSE2-NEXT:    movdqa {{.*#+}} xmm3 = [63,0,63,0]
 ; X86-SSE2-NEXT:    movdqa %xmm2, %xmm4
 ; X86-SSE2-NEXT:    pandn %xmm3, %xmm4
 ; X86-SSE2-NEXT:    psrlq $1, %xmm1
 ; X86-SSE2-NEXT:    movdqa %xmm1, %xmm5
 ; X86-SSE2-NEXT:    psrlq %xmm4, %xmm5
-; X86-SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[2,3,2,3]
+; X86-SSE2-NEXT:    pxor %xmm6, %xmm6
+; X86-SSE2-NEXT:    punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm6[2],xmm4[3],xmm6[3]
 ; X86-SSE2-NEXT:    psrlq %xmm4, %xmm1
 ; X86-SSE2-NEXT:    movsd {{.*#+}} xmm1 = xmm5[0],xmm1[1]
 ; X86-SSE2-NEXT:    pand %xmm3, %xmm2
 ; X86-SSE2-NEXT:    movdqa %xmm0, %xmm3
 ; X86-SSE2-NEXT:    psllq %xmm2, %xmm3
-; X86-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,3,2,3]
+; X86-SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm6[2],xmm2[3],xmm6[3]
 ; X86-SSE2-NEXT:    psllq %xmm2, %xmm0
 ; X86-SSE2-NEXT:    movsd {{.*#+}} xmm0 = xmm3[0],xmm0[1]
 ; X86-SSE2-NEXT:    orpd %xmm1, %xmm0
 ; X86-SSE2-NEXT:    retl
   %res = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %x, <2 x i64> %y, <2 x i64> %amt)
   ret <2 x i64> %res
 }
 

 ; X86-SSE2:       # %bb.0:
 ; X86-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,0,1]
 ; X86-SSE2-NEXT:    movdqa {{.*#+}} xmm3 = [63,0,63,0]
 ; X86-SSE2-NEXT:    movdqa %xmm2, %xmm4
 ; X86-SSE2-NEXT:    pandn %xmm3, %xmm4
 ; X86-SSE2-NEXT:    psrlq $1, %xmm1
 ; X86-SSE2-NEXT:    movdqa %xmm1, %xmm5
 ; X86-SSE2-NEXT:    psrlq %xmm4, %xmm5
-; X86-SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[2,3,2,3]
+; X86-SSE2-NEXT:    pxor %xmm6, %xmm6
+; X86-SSE2-NEXT:    punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm6[2],xmm4[3],xmm6[3]
 ; X86-SSE2-NEXT:    psrlq %xmm4, %xmm1
 ; X86-SSE2-NEXT:    movsd {{.*#+}} xmm1 = xmm5[0],xmm1[1]
 ; X86-SSE2-NEXT:    pand %xmm3, %xmm2
 ; X86-SSE2-NEXT:    movdqa %xmm0, %xmm3
 ; X86-SSE2-NEXT:    psllq %xmm2, %xmm3
-; X86-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,3,2,3]
+; X86-SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm6[2],xmm2[3],xmm6[3]
 ; X86-SSE2-NEXT:    psllq %xmm2, %xmm0
 ; X86-SSE2-NEXT:    movsd {{.*#+}} xmm0 = xmm3[0],xmm0[1]
 ; X86-SSE2-NEXT:    orpd %xmm1, %xmm0
 ; X86-SSE2-NEXT:    retl
   %splat = shufflevector <2 x i64> %amt, <2 x i64> undef, <2 x i32> zeroinitializer
   %res = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %x, <2 x i64> %y, <2 x i64> %splat)
   ret <2 x i64> %res
 }

 ; XOPAVX2-NEXT:    retq
 ;
 ; X86-SSE2-LABEL: constant_funnnel_v2i64:
 ; X86-SSE2:       # %bb.0:
 ; X86-SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [63,0,63,0]
 ; X86-SSE2-NEXT:    movdqa {{.*#+}} xmm3 = <4,u,14,u>
 ; X86-SSE2-NEXT:    movdqa %xmm3, %xmm4
 ; X86-SSE2-NEXT:    pandn %xmm2, %xmm4
-; X86-SSE2-NEXT:    psrlq $1, %xmm1
 ; X86-SSE2-NEXT:    movdqa %xmm1, %xmm5
+; X86-SSE2-NEXT:    psrlq $1, %xmm5
 ; X86-SSE2-NEXT:    psrlq %xmm4, %xmm5
-; X86-SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[2,3,2,3]
-; X86-SSE2-NEXT:    psrlq %xmm4, %xmm1
+; X86-SSE2-NEXT:    psrlq $50, %xmm1
 ; X86-SSE2-NEXT:    movsd {{.*#+}} xmm1 = xmm5[0],xmm1[1]
 ; X86-SSE2-NEXT:    pand %xmm2, %xmm3
 ; X86-SSE2-NEXT:    movdqa %xmm0, %xmm2
 ; X86-SSE2-NEXT:    psllq %xmm3, %xmm2
-; X86-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,3,2,3]
-; X86-SSE2-NEXT:    psllq %xmm3, %xmm0
+; X86-SSE2-NEXT:    psllq $14, %xmm0
 ; X86-SSE2-NEXT:    movsd {{.*#+}} xmm0 = xmm2[0],xmm0[1]
 ; X86-SSE2-NEXT:    orpd %xmm1, %xmm0
 ; X86-SSE2-NEXT:    retl
   %res = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %x, <2 x i64> %y, <2 x i64> <i64 4, i64 14>)
   ret <2 x i64> %res
 }
 
 define <4 x i32> @constant_funnnel_v4i32(<4 x i32> %x, <4 x i32> %y) nounwind {

 ; XOP-NEXT:    vpsrlq $50, %xmm1, %xmm1
 ; XOP-NEXT:    vpsllq $14, %xmm0, %xmm0
 ; XOP-NEXT:    vpor %xmm1, %xmm0, %xmm0
 ; XOP-NEXT:    retq
 ;
 ; X86-SSE2-LABEL: splatconstant_funnnel_v2i64:
 ; X86-SSE2:       # %bb.0:
 ; X86-SSE2-NEXT:    psrlq $50, %xmm1
-; X86-SSE2-NEXT:    movsd {{.*#+}} xmm1 = xmm1[0,1]
 ; X86-SSE2-NEXT:    psllq $14, %xmm0
-; X86-SSE2-NEXT:    movsd {{.*#+}} xmm0 = xmm0[0,1]
-; X86-SSE2-NEXT:    orpd %xmm1, %xmm0
+; X86-SSE2-NEXT:    por %xmm1, %xmm0
 ; X86-SSE2-NEXT:    retl
   %res = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %x, <2 x i64> %y, <2 x i64> <i64 14, i64 14>)
   ret <2 x i64> %res
 }
 
 define <4 x i32> @splatconstant_funnnel_v4i32(<4 x i32> %x, <4 x i32> %y) nounwind {
 ; SSE-LABEL: splatconstant_funnnel_v4i32:
 ; SSE:       # %bb.0:

llvm/test/CodeGen/X86/vector-fshl-rot-128.ll

 ; X86-SSE2-LABEL: var_funnnel_v2i64:
 ; X86-SSE2:       # %bb.0:
 ; X86-SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [63,0,63,0]
 ; X86-SSE2-NEXT:    pxor %xmm3, %xmm3
 ; X86-SSE2-NEXT:    psubq %xmm1, %xmm3
 ; X86-SSE2-NEXT:    pand %xmm2, %xmm1
 ; X86-SSE2-NEXT:    movdqa %xmm0, %xmm4
 ; X86-SSE2-NEXT:    psllq %xmm1, %xmm4
-; X86-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
-; X86-SSE2-NEXT:    movdqa %xmm0, %xmm5
-; X86-SSE2-NEXT:    psllq %xmm1, %xmm5
-; X86-SSE2-NEXT:    movsd {{.*#+}} xmm5 = xmm4[0],xmm5[1]
+; X86-SSE2-NEXT:    pxor %xmm5, %xmm5
+; X86-SSE2-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm5[2],xmm1[3],xmm5[3]
+; X86-SSE2-NEXT:    movdqa %xmm0, %xmm6
+; X86-SSE2-NEXT:    psllq %xmm1, %xmm6
+; X86-SSE2-NEXT:    movsd {{.*#+}} xmm6 = xmm4[0],xmm6[1]
 ; X86-SSE2-NEXT:    pand %xmm2, %xmm3
 ; X86-SSE2-NEXT:    movdqa %xmm0, %xmm1
 ; X86-SSE2-NEXT:    psrlq %xmm3, %xmm1
-; X86-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[2,3,2,3]
-; X86-SSE2-NEXT:    psrlq %xmm2, %xmm0
+; X86-SSE2-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm5[2],xmm3[3],xmm5[3]
+; X86-SSE2-NEXT:    psrlq %xmm3, %xmm0
 ; X86-SSE2-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
-; X86-SSE2-NEXT:    orpd %xmm5, %xmm0
+; X86-SSE2-NEXT:    orpd %xmm6, %xmm0
 ; X86-SSE2-NEXT:    retl
   %res = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %x, <2 x i64> %x, <2 x i64> %amt)
   ret <2 x i64> %res
 }
 
 define <4 x i32> @var_funnnel_v4i32(<4 x i32> %x, <4 x i32> %amt) nounwind {
 ; SSE2-LABEL: var_funnnel_v4i32:
 ; SSE2:       # %bb.0:

 ; X86-SSE2:       # %bb.0:
 ; X86-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,0,1]
 ; X86-SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [63,0,63,0]
 ; X86-SSE2-NEXT:    pxor %xmm3, %xmm3
 ; X86-SSE2-NEXT:    psubq %xmm1, %xmm3
 ; X86-SSE2-NEXT:    pand %xmm2, %xmm1
 ; X86-SSE2-NEXT:    movdqa %xmm0, %xmm4
 ; X86-SSE2-NEXT:    psllq %xmm1, %xmm4
-; X86-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
-; X86-SSE2-NEXT:    movdqa %xmm0, %xmm5
-; X86-SSE2-NEXT:    psllq %xmm1, %xmm5
-; X86-SSE2-NEXT:    movsd {{.*#+}} xmm5 = xmm4[0],xmm5[1]
+; X86-SSE2-NEXT:    pxor %xmm5, %xmm5
+; X86-SSE2-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm5[2],xmm1[3],xmm5[3]
+; X86-SSE2-NEXT:    movdqa %xmm0, %xmm6
+; X86-SSE2-NEXT:    psllq %xmm1, %xmm6
+; X86-SSE2-NEXT:    movsd {{.*#+}} xmm6 = xmm4[0],xmm6[1]
 ; X86-SSE2-NEXT:    pand %xmm2, %xmm3
 ; X86-SSE2-NEXT:    movdqa %xmm0, %xmm1
 ; X86-SSE2-NEXT:    psrlq %xmm3, %xmm1
-; X86-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[2,3,2,3]
-; X86-SSE2-NEXT:    psrlq %xmm2, %xmm0
+; X86-SSE2-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm5[2],xmm3[3],xmm5[3]
+; X86-SSE2-NEXT:    psrlq %xmm3, %xmm0
 ; X86-SSE2-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
-; X86-SSE2-NEXT:    orpd %xmm5, %xmm0
+; X86-SSE2-NEXT:    orpd %xmm6, %xmm0
 ; X86-SSE2-NEXT:    retl
   %splat = shufflevector <2 x i64> %amt, <2 x i64> undef, <2 x i32> zeroinitializer
   %res = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %x, <2 x i64> %x, <2 x i64> %splat)
   ret <2 x i64> %res
 }
 
 define <4 x i32> @splatvar_funnnel_v4i32(<4 x i32> %x, <4 x i32> %amt) nounwind {
 ; SSE2-LABEL: splatvar_funnnel_v4i32:

 ; X86-SSE2:       # %bb.0:
 ; X86-SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [63,0,63,0]
 ; X86-SSE2-NEXT:    movdqa {{.*#+}} xmm2 = <4,u,14,u>
 ; X86-SSE2-NEXT:    pxor %xmm3, %xmm3
 ; X86-SSE2-NEXT:    psubq %xmm2, %xmm3
 ; X86-SSE2-NEXT:    pand %xmm1, %xmm2
 ; X86-SSE2-NEXT:    movdqa %xmm0, %xmm4
 ; X86-SSE2-NEXT:    psllq %xmm2, %xmm4
-; X86-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,3,2,3]
-; X86-SSE2-NEXT:    movdqa %xmm0, %xmm5
-; X86-SSE2-NEXT:    psllq %xmm2, %xmm5
-; X86-SSE2-NEXT:    movsd {{.*#+}} xmm5 = xmm4[0],xmm5[1]
+; X86-SSE2-NEXT:    movdqa %xmm0, %xmm2
+; X86-SSE2-NEXT:    psllq $14, %xmm2
+; X86-SSE2-NEXT:    movsd {{.*#+}} xmm2 = xmm4[0],xmm2[1]
 ; X86-SSE2-NEXT:    pand %xmm1, %xmm3
 ; X86-SSE2-NEXT:    movdqa %xmm0, %xmm1
 ; X86-SSE2-NEXT:    psrlq %xmm3, %xmm1
-; X86-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[2,3,2,3]
-; X86-SSE2-NEXT:    psrlq %xmm2, %xmm0
+; X86-SSE2-NEXT:    psrlq $50, %xmm0
 ; X86-SSE2-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
-; X86-SSE2-NEXT:    orpd %xmm5, %xmm0
+; X86-SSE2-NEXT:    orpd %xmm2, %xmm0
 ; X86-SSE2-NEXT:    retl
   %res = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %x, <2 x i64> %x, <2 x i64> <i64 4, i64 14>)
   ret <2 x i64> %res
 }
 
 define <4 x i32> @constant_funnnel_v4i32(<4 x i32> %x) nounwind {
 ; SSE2-LABEL: constant_funnnel_v4i32:
 ; SSE2:       # %bb.0:

 ; XOP:       # %bb.0:
 ; XOP-NEXT:    vprotq $14, %xmm0, %xmm0
 ; XOP-NEXT:    retq
 ;
 ; X86-SSE2-LABEL: splatconstant_funnnel_v2i64:
 ; X86-SSE2:       # %bb.0:
 ; X86-SSE2-NEXT:    movdqa %xmm0, %xmm1
 ; X86-SSE2-NEXT:    psrlq $50, %xmm1
-; X86-SSE2-NEXT:    movsd {{.*#+}} xmm1 = xmm1[0,1]
 ; X86-SSE2-NEXT:    psllq $14, %xmm0
-; X86-SSE2-NEXT:    movsd {{.*#+}} xmm0 = xmm0[0,1]
-; X86-SSE2-NEXT:    orpd %xmm1, %xmm0
+; X86-SSE2-NEXT:    por %xmm1, %xmm0
 ; X86-SSE2-NEXT:    retl
   %res = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %x, <2 x i64> %x, <2 x i64> <i64 14, i64 14>)
   ret <2 x i64> %res
 }
 
 define <4 x i32> @splatconstant_funnnel_v4i32(<4 x i32> %x) nounwind {
 ; SSE-LABEL: splatconstant_funnnel_v4i32:
 ; SSE:       # %bb.0:

llvm/test/CodeGen/X86/vector-fshr-128.ll

 ; XOPAVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [63,63]
 ; XOPAVX2-NEXT:    vpand %xmm3, %xmm2, %xmm4
 ; XOPAVX2-NEXT:    vpsrlvq %xmm4, %xmm1, %xmm1
 ; XOPAVX2-NEXT:    vpandn %xmm3, %xmm2, %xmm2
 ; XOPAVX2-NEXT:    vpsllq $1, %xmm0, %xmm0
 ; XOPAVX2-NEXT:    vpsllvq %xmm2, %xmm0, %xmm0
 ; XOPAVX2-NEXT:    vpor %xmm1, %xmm0, %xmm0
 ; XOPAVX2-NEXT:    retq
 ;
 ; X86-SSE2-LABEL: var_funnnel_v2i64:
 ; X86-SSE2:       # %bb.0:
 ; X86-SSE2-NEXT:    movdqa {{.*#+}} xmm3 = [63,0,63,0]
 ; X86-SSE2-NEXT:    movdqa %xmm2, %xmm4
 ; X86-SSE2-NEXT:    pand %xmm3, %xmm4
 ; X86-SSE2-NEXT:    movdqa %xmm1, %xmm5
 ; X86-SSE2-NEXT:    psrlq %xmm4, %xmm5
-; X86-SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[2,3,2,3]
+; X86-SSE2-NEXT:    pxor %xmm6, %xmm6
+; X86-SSE2-NEXT:    punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm6[2],xmm4[3],xmm6[3]
 ; X86-SSE2-NEXT:    psrlq %xmm4, %xmm1
 ; X86-SSE2-NEXT:    movsd {{.*#+}} xmm1 = xmm5[0],xmm1[1]
 ; X86-SSE2-NEXT:    pandn %xmm3, %xmm2
 ; X86-SSE2-NEXT:    psllq $1, %xmm0
 ; X86-SSE2-NEXT:    movdqa %xmm0, %xmm3
 ; X86-SSE2-NEXT:    psllq %xmm2, %xmm3
-; X86-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,3,2,3]
+; X86-SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm6[2],xmm2[3],xmm6[3]
 ; X86-SSE2-NEXT:    psllq %xmm2, %xmm0
 ; X86-SSE2-NEXT:    movsd {{.*#+}} xmm0 = xmm3[0],xmm0[1]
 ; X86-SSE2-NEXT:    orpd %xmm1, %xmm0
 ; X86-SSE2-NEXT:    retl
   %res = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %x, <2 x i64> %y, <2 x i64> %amt)
   ret <2 x i64> %res
 }
 

 ; X86-SSE2-LABEL: splatvar_funnnel_v2i64:
 ; X86-SSE2:       # %bb.0:
 ; X86-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,0,1]
 ; X86-SSE2-NEXT:    movdqa {{.*#+}} xmm3 = [63,0,63,0]
 ; X86-SSE2-NEXT:    movdqa %xmm2, %xmm4
 ; X86-SSE2-NEXT:    pand %xmm3, %xmm4
 ; X86-SSE2-NEXT:    movdqa %xmm1, %xmm5
 ; X86-SSE2-NEXT:    psrlq %xmm4, %xmm5
-; X86-SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[2,3,2,3]
+; X86-SSE2-NEXT:    pxor %xmm6, %xmm6
+; X86-SSE2-NEXT:    punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm6[2],xmm4[3],xmm6[3]
 ; X86-SSE2-NEXT:    psrlq %xmm4, %xmm1
 ; X86-SSE2-NEXT:    movsd {{.*#+}} xmm1 = xmm5[0],xmm1[1]
 ; X86-SSE2-NEXT:    pandn %xmm3, %xmm2
 ; X86-SSE2-NEXT:    psllq $1, %xmm0
 ; X86-SSE2-NEXT:    movdqa %xmm0, %xmm3
 ; X86-SSE2-NEXT:    psllq %xmm2, %xmm3
-; X86-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,3,2,3]
+; X86-SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm6[2],xmm2[3],xmm6[3]
 ; X86-SSE2-NEXT:    psllq %xmm2, %xmm0
 ; X86-SSE2-NEXT:    movsd {{.*#+}} xmm0 = xmm3[0],xmm0[1]
 ; X86-SSE2-NEXT:    orpd %xmm1, %xmm0
 ; X86-SSE2-NEXT:    retl
   %splat = shufflevector <2 x i64> %amt, <2 x i64> undef, <2 x i32> zeroinitializer
   %res = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %x, <2 x i64> %y, <2 x i64> %splat)
   ret <2 x i64> %res
 }

 ; X86-SSE2-LABEL: constant_funnnel_v2i64:
 ; X86-SSE2:       # %bb.0:
 ; X86-SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [63,0,63,0]
 ; X86-SSE2-NEXT:    movdqa {{.*#+}} xmm3 = <4,u,14,u>
 ; X86-SSE2-NEXT:    movdqa %xmm3, %xmm4
 ; X86-SSE2-NEXT:    pand %xmm2, %xmm4
 ; X86-SSE2-NEXT:    movdqa %xmm1, %xmm5
 ; X86-SSE2-NEXT:    psrlq %xmm4, %xmm5
-; X86-SSE2-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[2,3,2,3]
-; X86-SSE2-NEXT:    psrlq %xmm4, %xmm1
+; X86-SSE2-NEXT:    psrlq $14, %xmm1
 ; X86-SSE2-NEXT:    movsd {{.*#+}} xmm1 = xmm5[0],xmm1[1]
 ; X86-SSE2-NEXT:    pandn %xmm2, %xmm3
-; X86-SSE2-NEXT:    psllq $1, %xmm0
 ; X86-SSE2-NEXT:    movdqa %xmm0, %xmm2
+; X86-SSE2-NEXT:    psllq $1, %xmm2
 ; X86-SSE2-NEXT:    psllq %xmm3, %xmm2
-; X86-SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,3,2,3]
-; X86-SSE2-NEXT:    psllq %xmm3, %xmm0
+; X86-SSE2-NEXT:    psllq $50, %xmm0
 ; X86-SSE2-NEXT:    movsd {{.*#+}} xmm0 = xmm2[0],xmm0[1]
 ; X86-SSE2-NEXT:    orpd %xmm1, %xmm0
 ; X86-SSE2-NEXT:    retl
   %res = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %x, <2 x i64> %y, <2 x i64> <i64 4, i64 14>)
   ret <2 x i64> %res
 }
 
 define <4 x i32> @constant_funnnel_v4i32(<4 x i32> %x, <4 x i32> %y) nounwind {

 ; XOP-NEXT:    vpsrlq $14, %xmm1, %xmm1
 ; XOP-NEXT:    vpsllq $50, %xmm0, %xmm0
 ; XOP-NEXT:    vpor %xmm1, %xmm0, %xmm0
 ; XOP-NEXT:    retq
 ;
 ; X86-SSE2-LABEL: splatconstant_funnnel_v2i64:
 ; X86-SSE2:       # %bb.0:
 ; X86-SSE2-NEXT:    psrlq $14, %xmm1
-; X86-SSE2-NEXT:    movsd {{.*#+}} xmm1 = xmm1[0,1]
 ; X86-SSE2-NEXT:    psllq $50, %xmm0
-; X86-SSE2-NEXT:    movsd {{.*#+}} xmm0 = xmm0[0,1]
-; X86-SSE2-NEXT:    orpd %xmm1, %xmm0
+; X86-SSE2-NEXT:    por %xmm1, %xmm0
 ; X86-SSE2-NEXT:    retl
   %res = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %x, <2 x i64> %y, <2 x i64> <i64 14, i64 14>)
   ret <2 x i64> %res
 }
 
 define <4 x i32> @splatconstant_funnnel_v4i32(<4 x i32> %x, <4 x i32> %y) nounwind {
 ; SSE-LABEL: splatconstant_funnnel_v4i32:
 ; SSE:       # %bb.0:

llvm/test/CodeGen/X86/vector-fshr-rot-128.ll

 ; X86-SSE2-LABEL: var_funnnel_v2i64:
 ; X86-SSE2:       # %bb.0:
 ; X86-SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [63,0,63,0]
 ; X86-SSE2-NEXT:    pxor %xmm3, %xmm3
 ; X86-SSE2-NEXT:    psubq %xmm1, %xmm3
 ; X86-SSE2-NEXT:    pand %xmm2, %xmm1
 ; X86-SSE2-NEXT:    movdqa %xmm0, %xmm4
 ; X86-SSE2-NEXT:    psrlq %xmm1, %xmm4
-; X86-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
-; X86-SSE2-NEXT:    movdqa %xmm0, %xmm5
-; X86-SSE2-NEXT:    psrlq %xmm1, %xmm5
-; X86-SSE2-NEXT:    movsd {{.*#+}} xmm5 = xmm4[0],xmm5[1]
+; X86-SSE2-NEXT:    pxor %xmm5, %xmm5
+; X86-SSE2-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm5[2],xmm1[3],xmm5[3]
+; X86-SSE2-NEXT:    movdqa %xmm0, %xmm6
+; X86-SSE2-NEXT:    psrlq %xmm1, %xmm6
+; X86-SSE2-NEXT:    movsd {{.*#+}} xmm6 = xmm4[0],xmm6[1]
 ; X86-SSE2-NEXT:    pand %xmm2, %xmm3
 ; X86-SSE2-NEXT:    movdqa %xmm0, %xmm1
 ; X86-SSE2-NEXT:    psllq %xmm3, %xmm1
-; X86-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[2,3,2,3]
-; X86-SSE2-NEXT:    psllq %xmm2, %xmm0
+; X86-SSE2-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm5[2],xmm3[3],xmm5[3]
+; X86-SSE2-NEXT:    psllq %xmm3, %xmm0
 ; X86-SSE2-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
-; X86-SSE2-NEXT:    orpd %xmm5, %xmm0
+; X86-SSE2-NEXT:    orpd %xmm6, %xmm0
 ; X86-SSE2-NEXT:    retl
   %res = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %x, <2 x i64> %x, <2 x i64> %amt)
   ret <2 x i64> %res
 }
 
 define <4 x i32> @var_funnnel_v4i32(<4 x i32> %x, <4 x i32> %amt) nounwind {
 ; SSE2-LABEL: var_funnnel_v4i32:
 ; SSE2:       # %bb.0:

 ; X86-SSE2:       # %bb.0:
 ; X86-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,0,1]
 ; X86-SSE2-NEXT:    movdqa {{.*#+}} xmm2 = [63,0,63,0]
 ; X86-SSE2-NEXT:    pxor %xmm3, %xmm3
 ; X86-SSE2-NEXT:    psubq %xmm1, %xmm3
 ; X86-SSE2-NEXT:    pand %xmm2, %xmm1
 ; X86-SSE2-NEXT:    movdqa %xmm0, %xmm4
 ; X86-SSE2-NEXT:    psrlq %xmm1, %xmm4
-; X86-SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
-; X86-SSE2-NEXT:    movdqa %xmm0, %xmm5
-; X86-SSE2-NEXT:    psrlq %xmm1, %xmm5
-; X86-SSE2-NEXT:    movsd {{.*#+}} xmm5 = xmm4[0],xmm5[1]
+; X86-SSE2-NEXT:    pxor %xmm5, %xmm5
+; X86-SSE2-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm5[2],xmm1[3],xmm5[3]
+; X86-SSE2-NEXT:    movdqa %xmm0, %xmm6
+; X86-SSE2-NEXT:    psrlq %xmm1, %xmm6
+; X86-SSE2-NEXT:    movsd {{.*#+}} xmm6 = xmm4[0],xmm6[1]
 ; X86-SSE2-NEXT:    pand %xmm2, %xmm3
 ; X86-SSE2-NEXT:    movdqa %xmm0, %xmm1
 ; X86-SSE2-NEXT:    psllq %xmm3, %xmm1
-; X86-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[2,3,2,3]
-; X86-SSE2-NEXT:    psllq %xmm2, %xmm0
+; X86-SSE2-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm5[2],xmm3[3],xmm5[3]
+; X86-SSE2-NEXT:    psllq %xmm3, %xmm0
 ; X86-SSE2-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
-; X86-SSE2-NEXT:    orpd %xmm5, %xmm0
+; X86-SSE2-NEXT:    orpd %xmm6, %xmm0
 ; X86-SSE2-NEXT:    retl
   %splat = shufflevector <2 x i64> %amt, <2 x i64> undef, <2 x i32> zeroinitializer
   %res = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %x, <2 x i64> %x, <2 x i64> %splat)
   ret <2 x i64> %res
 }
 
 define <4 x i32> @splatvar_funnnel_v4i32(<4 x i32> %x, <4 x i32> %amt) nounwind {
 ; SSE2-LABEL: splatvar_funnnel_v4i32:

 ; X86-SSE2:       # %bb.0:
 ; X86-SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [63,0,63,0]
 ; X86-SSE2-NEXT:    movdqa {{.*#+}} xmm2 = <4,u,14,u>
 ; X86-SSE2-NEXT:    pxor %xmm3, %xmm3
 ; X86-SSE2-NEXT:    psubq %xmm2, %xmm3
 ; X86-SSE2-NEXT:    pand %xmm1, %xmm2
 ; X86-SSE2-NEXT:    movdqa %xmm0, %xmm4
 ; X86-SSE2-NEXT:    psrlq %xmm2, %xmm4
-; X86-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,3,2,3]
-; X86-SSE2-NEXT:    movdqa %xmm0, %xmm5
-; X86-SSE2-NEXT:    psrlq %xmm2, %xmm5
-; X86-SSE2-NEXT:    movsd {{.*#+}} xmm5 = xmm4[0],xmm5[1]
+; X86-SSE2-NEXT:    movdqa %xmm0, %xmm2
+; X86-SSE2-NEXT:    psrlq $14, %xmm2
+; X86-SSE2-NEXT:    movsd {{.*#+}} xmm2 = xmm4[0],xmm2[1]
 ; X86-SSE2-NEXT:    pand %xmm1, %xmm3
 ; X86-SSE2-NEXT:    movdqa %xmm0, %xmm1
 ; X86-SSE2-NEXT:    psllq %xmm3, %xmm1
-; X86-SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[2,3,2,3]
-; X86-SSE2-NEXT:    psllq %xmm2, %xmm0
+; X86-SSE2-NEXT:    psllq $50, %xmm0
 ; X86-SSE2-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
-; X86-SSE2-NEXT:    orpd %xmm5, %xmm0
+; X86-SSE2-NEXT:    orpd %xmm2, %xmm0
 ; X86-SSE2-NEXT:    retl
   %res = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %x, <2 x i64> %x, <2 x i64> <i64 4, i64 14>)
   ret <2 x i64> %res
 }
 
 define <4 x i32> @constant_funnnel_v4i32(<4 x i32> %x) nounwind {
 ; SSE2-LABEL: constant_funnnel_v4i32:
 ; SSE2:       # %bb.0:

 ; XOP:       # %bb.0:
 ; XOP-NEXT:    vprotq $50, %xmm0, %xmm0
 ; XOP-NEXT:    retq
 ;
 ; X86-SSE2-LABEL: splatconstant_funnnel_v2i64:
 ; X86-SSE2:       # %bb.0:
 ; X86-SSE2-NEXT:    movdqa %xmm0, %xmm1
 ; X86-SSE2-NEXT:    psllq $50, %xmm1
-; X86-SSE2-NEXT:    movsd {{.*#+}} xmm1 = xmm1[0,1]
 ; X86-SSE2-NEXT:    psrlq $14, %xmm0
-; X86-SSE2-NEXT:    movsd {{.*#+}} xmm0 = xmm0[0,1]
-; X86-SSE2-NEXT:    orpd %xmm1, %xmm0
+; X86-SSE2-NEXT:    por %xmm1, %xmm0
 ; X86-SSE2-NEXT:    retl
   %res = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %x, <2 x i64> %x, <2 x i64> <i64 14, i64 14>)
   ret <2 x i64> %res
 }
 
 define <4 x i32> @splatconstant_funnnel_v4i32(<4 x i32> %x) nounwind {
 ; SSE-LABEL: splatconstant_funnnel_v4i32:
 ; SSE:       # %bb.0:

llvm/test/CodeGen/X86/vector-idiv-v2i32.ll

; X86-NEXT:    retl
   %b = udiv <2 x i32> %a, <i32 8, i32 8>
   store <2 x i32> %b, <2 x i32>* %y
   ret void
 }
 
 define void @test_urem_pow2_v2i32(<2 x i32>* %x, <2 x i32>* %y) nounwind {
 ; X64-LABEL: test_urem_pow2_v2i32:
 ; X64:       # %bb.0:
-; X64-NEXT:    movabsq $30064771079, %rax # imm = 0x700000007
-; X64-NEXT:    andq (%rdi), %rax
-; X64-NEXT:    movq %rax, (%rsi)
+; X64-NEXT:    movsd {{.*#+}} xmm0 = mem[0],zero
+; X64-NEXT:    andps {{.*}}(%rip), %xmm0
+; X64-NEXT:    movlps %xmm0, (%rsi)
 ; X64-NEXT:    retq
 ;
 ; X86-LABEL: test_urem_pow2_v2i32:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
 ; X86-NEXT:    movsd {{.*#+}} xmm0 = mem[0],zero
 ; X86-NEXT:    andps {{\.LCPI.*}}, %xmm0

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

 }
 declare <8 x i16> @llvm.x86.sse2.psrli.w(<8 x i16>, i32)
 
 define void @PR43024() {
 ; SSE-LABEL: PR43024:
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    movaps {{.*#+}} xmm0 = [NaN,NaN,0.0E+0,0.0E+0]
 ; SSE-NEXT:    movaps %xmm0, (%rax)
-; SSE-NEXT:    addss {{.*}}(%rip), %xmm0
-; SSE-NEXT:    xorps %xmm1, %xmm1
-; SSE-NEXT:    addss %xmm1, %xmm0
-; SSE-NEXT:    addss %xmm1, %xmm0
-; SSE-NEXT:    movss %xmm0, (%rax)
+; SSE-NEXT:    movl $2143289344, (%rax) # imm = 0x7FC00000
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: PR43024:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vmovaps {{.*#+}} xmm0 = [NaN,NaN,0.0E+0,0.0E+0]
 ; AVX-NEXT:    vmovaps %xmm0, (%rax)
-; AVX-NEXT:    vaddss {{\.LCPI.*}}+{{.*}}(%rip), %xmm0, %xmm0
-; AVX-NEXT:    vxorps %xmm1, %xmm1, %xmm1
-; AVX-NEXT:    vaddss %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vaddss {{\.LCPI.*}}+{{.*}}(%rip), %xmm0, %xmm0
-; AVX-NEXT:    vmovss %xmm0, (%rax)
+; AVX-NEXT:    movl $2143289344, (%rax) # imm = 0x7FC00000
 ; AVX-NEXT:    retq
   store <4 x float> <float 0x7FF8000000000000, float 0x7FF8000000000000, float 0x0, float 0x0>, <4 x float>* undef, align 16
   %1 = load <4 x float>, <4 x float>* undef, align 16
   %2 = fmul <4 x float> %1, <float 0x0, float 0x0, float 0x0, float 0x0>
   %3 = shufflevector <4 x float> %2, <4 x float> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef>
   %4 = fadd <4 x float> %2, %3
   %5 = fadd <4 x float> zeroinitializer, %4
   %6 = shufflevector <4 x float> %2, <4 x float> undef, <4 x i32> <i32 3, i32 undef, i32 undef, i32 undef>

llvm/test/CodeGen/X86/vector-trunc-math.ll

 
 ;
 ; mul to constant
 ;
 
 define <4 x i32> @trunc_mul_const_v4i64_v4i32(<4 x i64> %a0) nounwind {
 ; SSE-LABEL: trunc_mul_const_v4i64_v4i32:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    xorps %xmm2, %xmm2
-; SSE-NEXT:    unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm0[1]
 ; SSE-NEXT:    pmuludq {{.*}}(%rip), %xmm1
-; SSE-NEXT:    shufps {{.*#+}} xmm2 = xmm2[0,2],xmm1[0,2]
-; SSE-NEXT:    movaps %xmm2, %xmm0
+; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
+; SSE-NEXT:    andps {{.*}}(%rip), %xmm0
 ; SSE-NEXT:    retq
 ;
 ; AVX1-LABEL: trunc_mul_const_v4i64_v4i32:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
 ; AVX1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
 ; AVX1-NEXT:    vpmulld {{.*}}(%rip), %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper

llvm/test/CodeGen/X86/vselect.ll

 ; SSE2-LABEL: simplify_select:
 ; SSE2:       # %bb.0:
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; SSE2-NEXT:    pslld $31, %xmm0
 ; SSE2-NEXT:    psrad $31, %xmm0
 ; SSE2-NEXT:    movd %edi, %xmm1
 ; SSE2-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[1,0,1,1]
 ; SSE2-NEXT:    por %xmm1, %xmm2
-; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[0,1],xmm2[1,3]
-; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[2,0],xmm2[1,1]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
 ; SSE2-NEXT:    pand %xmm0, %xmm2
 ; SSE2-NEXT:    pandn %xmm1, %xmm0
 ; SSE2-NEXT:    por %xmm2, %xmm0
 ; SSE2-NEXT:    retq
 ;
 ; SSE41-LABEL: simplify_select:
 ; SSE41:       # %bb.0:
 ; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
 ; SSE41-NEXT:    pslld $31, %xmm0
 ; SSE41-NEXT:    movd %edi, %xmm1
 ; SSE41-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[1,0,1,1]
 ; SSE41-NEXT:    por %xmm1, %xmm2
-; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[1,1,1,1]
-; SSE41-NEXT:    pinsrd $1, %edi, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
 ; SSE41-NEXT:    blendvps %xmm0, %xmm2, %xmm1
 ; SSE41-NEXT:    movaps %xmm1, %xmm0
 ; SSE41-NEXT:    retq
 ;
-; AVX-LABEL: simplify_select:
-; AVX:       # %bb.0:
-; AVX-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
-; AVX-NEXT:    vpslld $31, %xmm0, %xmm0
-; AVX-NEXT:    vmovd %edi, %xmm1
-; AVX-NEXT:    vpshufd {{.*#+}} xmm2 = xmm1[1,0,1,1]
-; AVX-NEXT:    vpor %xmm1, %xmm2, %xmm1
-; AVX-NEXT:    vpshufd {{.*#+}} xmm2 = xmm1[1,1,1,1]
-; AVX-NEXT:    vpinsrd $1, %edi, %xmm2, %xmm2
-; AVX-NEXT:    vblendvps %xmm0, %xmm1, %xmm2, %xmm0
-; AVX-NEXT:    retq
+; AVX1-LABEL: simplify_select:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; AVX1-NEXT:    vpslld $31, %xmm0, %xmm0
+; AVX1-NEXT:    vmovd %edi, %xmm1
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm1[1,0,1,1]
+; AVX1-NEXT:    vpor %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,0,1,1]
+; AVX1-NEXT:    vblendvps %xmm0, %xmm2, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: simplify_select:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
+; AVX2-NEXT:    vpslld $31, %xmm0, %xmm0
+; AVX2-NEXT:    vmovd %edi, %xmm1
+; AVX2-NEXT:    vpshufd {{.*#+}} xmm2 = xmm1[1,0,1,1]
+; AVX2-NEXT:    vpor %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpbroadcastd %xmm1, %xmm1
+; AVX2-NEXT:    vblendvps %xmm0, %xmm2, %xmm1, %xmm0
+; AVX2-NEXT:    retq
   %a = insertelement <2 x i32> <i32 0, i32 undef>, i32 %x, i32 1
   %b = insertelement <2 x i32> <i32 undef, i32 0>, i32 %x, i32 0
   %y = or <2 x i32> %a, %b
   %p16 = extractelement <2 x i32> %y, i32 1
   %p17 = insertelement <2 x i32> undef, i32 %p16, i32 0
   %p18 = insertelement <2 x i32> %p17, i32 %x, i32 1
   %r = select <2 x i1> %z, <2 x i32> %y, <2 x i32> %p18
   ret <2 x i32> %r

llvm/test/CodeGen/X86/widen_cast-5.ll

 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; X86-NEXT:    movsd {{.*#+}} xmm0 = mem[0],zero
 ; X86-NEXT:    xorps LCPI0_0, %xmm0
 ; X86-NEXT:    movlps %xmm0, (%eax)
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: convert:
 ; X64:       ## %bb.0: ## %entry
-; X64-NEXT:    movabsq $140733193388287, %rax ## imm = 0x7FFF000000FF
-; X64-NEXT:    xorq %rsi, %rax
-; X64-NEXT:    movq %rax, (%rdi)
+; X64-NEXT:    movq %rsi, %xmm0
+; X64-NEXT:    pxor {{.*}}(%rip), %xmm0
+; X64-NEXT:    movq %xmm0, (%rdi)
 ; X64-NEXT:    retq
 entry:
   %conv = bitcast i64 %src to <2 x i32>
   %xor = xor <2 x i32> %conv, < i32 255, i32 32767 >
   store <2 x i32> %xor, <2 x i32>* %dst.addr
   ret void
 }

llvm/test/CodeGen/X86/widen_shuffle-1.ll

; X64-NEXT:    retq
   ret <8 x i8> %vshuf
 }
 
 ; PR11389: another CONCAT_VECTORS case
 define void @shuf5(<8 x i8>* %p) nounwind {
 ; X86-LABEL: shuf5:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT:    movsd {{.*#+}} xmm0 = mem[0],zero
-; X86-NEXT:    movsd %xmm0, (%eax)
+; X86-NEXT:    movl $555819297, 4(%eax) # imm = 0x21212121
+; X86-NEXT:    movl $555819297, (%eax) # imm = 0x21212121
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: shuf5:
 ; X64:       # %bb.0:
 ; X64-NEXT:    movabsq $2387225703656530209, %rax # imm = 0x2121212121212121
 ; X64-NEXT:    movq %rax, (%rdi)
 ; X64-NEXT:    retq
   %v = shufflevector <2 x i8> <i8 4, i8 33>, <2 x i8> undef, <8 x i32> <i32 1, i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
   store <8 x i8> %v, <8 x i8>* %p, align 8
   ret void
 }