[DAGCombiner] Use known bits to fold extract_vector_elt with const index

[foad]

This isn't ready to commit yet:

1. there are a few code quality regressions in the test cases
2. it provokes a crash on test/CodeGen/AMDGPU/global-extload-i16.ll which I haven't been able to fix yet

[foad]

In D87502#2267572, @foad wrote:

2. it provokes a crash on test/CodeGen/AMDGPU/global-extload-i16.ll which I haven't been able to fix yet

https://bugs.llvm.org/show_bug.cgi?id=47492

[foad]

Rebase.

[foad]

In D87502#2267572, @foad wrote:

This isn't ready to commit yet:

1. there are a few code quality regressions in the test cases
2. it provokes a crash on test/CodeGen/AMDGPU/global-extload-i16.ll which I haven't been able to fix yet

Update:

1. See inline comments on the regressions.
2. The crash is fixed. See D87757.

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

4206–4248

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

177–191

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

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

188–190

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.

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

2713

Regression. Perhaps we need better known bits analysis.

[foad]

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

2713

D87912 is my attempt at fixing this.

[RKSimon]

@foad please can you rebase?

[foad]

Rebase after D88225.

[RKSimon]

llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp

17848

Would this be better handled by SimplifyDemandedBits?

[foad]

llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp

17848

Good idea. D88569, D88570.

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

2713

This regression got fixed by D88225.

[foad]

Abandoning in favour of D88569 + D88570.

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();
 
+  // If all bits of the extracted element are known, return it as a constant.
+  if (IndexC && ScalarVT == VecVT.getVectorElementType()) {
+    APInt DemandedElts = APInt::getOneBitSet(NumElts, IndexC->getZExtValue());
+    KnownBits Known = DAG.computeKnownBits(VecOp, DemandedElts);
+    if (Known.isConstant()) {
+      APInt KnownVal = Known.getConstant();
+      if (ScalarVT.isInteger())
+        return DAG.getConstant(KnownVal, DL, ScalarVT);
+      if (ScalarVT.isFloatingPoint())
+        return DAG.getConstantFP(
+            APFloat(DAG.EVTToAPFloatSemantics(ScalarVT), KnownVal), DL,
+            ScalarVT);
+    }
+  }
+
   // 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();

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-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/dag-numsignbits.ll

 ; RUN: llc < %s -mtriple=aarch64-unknown | FileCheck %s
 
 ; PR32273
 
 define void @signbits_vXi1(<4 x i16> %a1) {
 ; CHECK-LABEL: signbits_vXi1
-; CHECK: 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: mov w1, wzr
+; CHECK: 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/popcount.ll

 ; CHECK-NEXT:    fmov d0, x0
 ; CHECK-NEXT:    mov v0.d[1], x1
 ; CHECK-NEXT:    cnt v0.16b, v0.16b
 ; CHECK-NEXT:    uaddlv h0, v0.16b
 ; CHECK-NEXT:    // implicit-def: $q1
 ; CHECK-NEXT:    mov v1.16b, v0.16b
 ; CHECK-NEXT:    fmov w0, s1
 ; 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/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>
   store <2 x i64> %r.val, <2 x i64> addrspace(1)* %r
   ret void
 }

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/sub.v2i16.ll

 ; GFX9-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX9-NEXT:    v_mov_b32_e32 v3, v1
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT:    global_load_dword v2, v0, s[6:7]
 ; GFX9-NEXT:    global_load_dword v0, v0, s[0:1]
 ; GFX9-NEXT:    s_mov_b32 s7, 0xf000
 ; GFX9-NEXT:    s_mov_b32 s6, -1
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-NEXT:    v_pk_sub_i16 v2, v2, v0
-; 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, v0
+; GFX9-NEXT:    v_lshrrev_b32_e32 v2, 16, v0
+; GFX9-NEXT:    v_and_b32_e32 v0, 0xffff, v0
 ; 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
 ; VI-NEXT:    v_lshlrev_b32_e32 v2, 2, v0

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

 ; CHECK-BE-NEXT:    vmov r1, r0, d16
 ; CHECK-BE-NEXT:    bx lr
     ret double 1.0
 }
 
 define <4 x i32> @return_v4i32() {
 ; CHECK-LE-LABEL: return_v4i32:
 ; CHECK-LE:       @ %bb.0:
-; CHECK-LE-NEXT:    adr r0, .LCPI6_0
-; CHECK-LE-NEXT:    vld1.64 {d16, d17}, [r0:128]
+; CHECK-LE-NEXT:    vldr d16, .LCPI6_0
+; CHECK-LE-NEXT:    vldr d17, .LCPI6_1
 ; CHECK-LE-NEXT:    vmov r0, r1, d16
 ; CHECK-LE-NEXT:    vmov r2, r3, d17
 ; CHECK-LE-NEXT:    bx lr
-; CHECK-LE-NEXT:    .p2align 4
+; CHECK-LE-NEXT:    .p2align 3
 ; CHECK-LE-NEXT:  @ %bb.1:
 ; CHECK-LE-NEXT:  .LCPI6_0:
 ; CHECK-LE-NEXT:    .long 42 @ double 9.1245819032257467E-313
 ; CHECK-LE-NEXT:    .long 43
+; CHECK-LE-NEXT:  .LCPI6_1:
 ; CHECK-LE-NEXT:    .long 44 @ double 9.5489810615176143E-313
 ; CHECK-LE-NEXT:    .long 45
 ;
 ; CHECK-BE-LABEL: return_v4i32:
 ; CHECK-BE:       @ %bb.0:
-; CHECK-BE-NEXT:    adr r0, .LCPI6_0
-; CHECK-BE-NEXT:    vld1.64 {d16, d17}, [r0:128]
+; CHECK-BE-NEXT:    vldr d16, .LCPI6_0
+; CHECK-BE-NEXT:    vldr d17, .LCPI6_1
 ; CHECK-BE-NEXT:    vmov r1, r0, d16
 ; CHECK-BE-NEXT:    vmov r3, r2, d17
 ; CHECK-BE-NEXT:    bx lr
-; CHECK-BE-NEXT:    .p2align 4
+; CHECK-BE-NEXT:    .p2align 3
 ; CHECK-BE-NEXT:  @ %bb.1:
 ; CHECK-BE-NEXT:  .LCPI6_0:
 ; CHECK-BE-NEXT:    .long 42 @ double 8.912382324178626E-313
 ; CHECK-BE-NEXT:    .long 43
+; CHECK-BE-NEXT:  .LCPI6_1:
 ; CHECK-BE-NEXT:    .long 44 @ double 9.3367814824704935E-313
 ; CHECK-BE-NEXT:    .long 45
    ret < 4 x i32> < i32 42, i32 43, i32 44, i32 45 >
 }
 
 define <2 x double> @return_v2f64() {
 ; CHECK-LE-LABEL: return_v2f64:
 ; CHECK-LE:       @ %bb.0:

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

 ; 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
 ; MIPS32EL-NEXT:    ori $2, $1, 2060

 ; 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/mve-vecreduce-add.ll

entry:
   %xx = sext <4 x i32> %x to <4 x i64>
   %z = call i64 @llvm.experimental.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.experimental.vector.reduce.add.v2i64(<2 x i64> %xx)
   ret i64 %z
 }
 
 define arm_aapcs_vfpcc i64 @add_v2i32_v2i64_sext(<2 x i32> %x) {

 define arm_aapcs_vfpcc i64 @add_v8i16_v8i64_zext(<8 x i16> %x) {
 ; CHECK-LABEL: add_v8i16_v8i64_zext:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    vmov.u16 r0, q0[0]
 ; CHECK-NEXT:    vmov.i64 q1, #0xffff
 ; CHECK-NEXT:    vmov.32 q2[0], r0
 ; CHECK-NEXT:    vmov.u16 r0, q0[1]
 ; CHECK-NEXT:    vmov.32 q2[2], r0
+; CHECK-NEXT:    movs r2, #0
 ; CHECK-NEXT:    vand q2, q2, q1
 ; CHECK-NEXT:    vmov r0, s10
 ; CHECK-NEXT:    vmov r1, s8
-; CHECK-NEXT:    vmov r2, s11
 ; CHECK-NEXT:    add r0, r1
 ; CHECK-NEXT:    vmov.u16 r1, q0[2]
-; CHECK-NEXT:    vmov.32 q3[0], r1
+; CHECK-NEXT:    vmov.32 q2[0], r1
 ; CHECK-NEXT:    vmov.u16 r1, q0[3]
-; CHECK-NEXT:    vmov.32 q3[2], r1
-; CHECK-NEXT:    vand q3, q3, q1
-; CHECK-NEXT:    vmov r1, s12
+; CHECK-NEXT:    vmov.32 q2[2], r1
+; 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[4]
-; CHECK-NEXT:    vmov.32 q3[0], r1
+; CHECK-NEXT:    vmov.32 q2[0], r1
 ; CHECK-NEXT:    vmov.u16 r1, q0[5]
-; CHECK-NEXT:    vmov.32 q3[2], r1
-; CHECK-NEXT:    vand q3, q3, q1
-; CHECK-NEXT:    vmov r1, s12
-; CHECK-NEXT:    vmov r3, s14
+; CHECK-NEXT:    vmov.32 q2[2], r1
+; 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:    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[6]
 ; CHECK-NEXT:    vmov.32 q2[0], r2
 ; CHECK-NEXT:    vmov.u16 r2, q0[7]
 ; CHECK-NEXT:    vmov.32 q2[2], r2
 ; CHECK-NEXT:    vand q0, q2, 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.experimental.vector.reduce.add.v8i64(<8 x i64> %xx)
   ret i64 %z
 }
 
 define arm_aapcs_vfpcc i64 @add_v8i16_v8i64_sext(<8 x i16> %x) {
entry:
   %xx = sext <8 x i16> %x to <8 x i64>
   %z = call i64 @llvm.experimental.vector.reduce.add.v8i64(<8 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.experimental.vector.reduce.add.v2i64(<2 x i64> %xx)
   ret i64 %z
 }
 
 define arm_aapcs_vfpcc i64 @add_v2i16_v2i64_sext(<2 x i16> %x) {

 define arm_aapcs_vfpcc i64 @add_v16i8_v16i64_zext(<16 x i8> %x) {
 ; CHECK-LABEL: add_v16i8_v16i64_zext:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    vmov.u8 r0, q0[0]
 ; CHECK-NEXT:    vmov.i64 q1, #0xff
 ; CHECK-NEXT:    vmov.32 q2[0], r0
 ; CHECK-NEXT:    vmov.u8 r0, q0[1]
 ; CHECK-NEXT:    vmov.32 q2[2], r0
+; CHECK-NEXT:    movs r2, #0
 ; CHECK-NEXT:    vand q2, q2, q1
 ; CHECK-NEXT:    vmov r0, s10
 ; CHECK-NEXT:    vmov r1, s8
-; CHECK-NEXT:    vmov r2, s11
 ; CHECK-NEXT:    add r0, r1
 ; CHECK-NEXT:    vmov.u8 r1, q0[2]
-; CHECK-NEXT:    vmov.32 q3[0], r1
+; CHECK-NEXT:    vmov.32 q2[0], r1
 ; CHECK-NEXT:    vmov.u8 r1, q0[3]
-; CHECK-NEXT:    vmov.32 q3[2], r1
-; CHECK-NEXT:    vand q3, q3, q1
-; CHECK-NEXT:    vmov r1, s12
+; CHECK-NEXT:    vmov.32 q2[2], r1
+; 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[4]
-; CHECK-NEXT:    vmov.32 q3[0], r1
+; CHECK-NEXT:    vmov.32 q2[0], r1
 ; CHECK-NEXT:    vmov.u8 r1, q0[5]
-; CHECK-NEXT:    vmov.32 q3[2], r1
-; CHECK-NEXT:    vand q3, q3, q1
-; CHECK-NEXT:    vmov r1, s12
-; CHECK-NEXT:    vmov r3, s14
+; CHECK-NEXT:    vmov.32 q2[2], r1
+; 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:    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[6]
 ; CHECK-NEXT:    vmov.32 q2[0], r2
 ; CHECK-NEXT:    vmov.u8 r2, q0[7]
 ; CHECK-NEXT:    vmov.32 q2[2], 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:    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[8]
 ; CHECK-NEXT:    vmov.32 q2[0], r2
 ; CHECK-NEXT:    vmov.u8 r2, q0[9]
 ; CHECK-NEXT:    vmov.32 q2[2], 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:    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[10]
 ; CHECK-NEXT:    vmov.32 q2[0], r2
 ; CHECK-NEXT:    vmov.u8 r2, q0[11]
 ; CHECK-NEXT:    vmov.32 q2[2], 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:    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[12]
 ; CHECK-NEXT:    vmov.32 q2[0], r2
 ; CHECK-NEXT:    vmov.u8 r2, q0[13]
 ; CHECK-NEXT:    vmov.32 q2[2], 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:    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[14]
 ; CHECK-NEXT:    vmov.32 q2[0], r2
 ; CHECK-NEXT:    vmov.u8 r2, q0[15]
 ; CHECK-NEXT:    vmov.32 q2[2], r2
+; CHECK-NEXT:    adc r1, r1, #0
 ; CHECK-NEXT:    vand q0, q2, 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.experimental.vector.reduce.add.v16i64(<16 x i64> %xx)
   ret i64 %z
 }
 
 define arm_aapcs_vfpcc i64 @add_v16i8_v16i64_sext(<16 x i8> %x) {
entry:
   %xx = sext <16 x i8> %x to <16 x i64>
   %z = call i64 @llvm.experimental.vector.reduce.add.v16i64(<16 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.experimental.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.experimental.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.experimental.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) {
entry:
   %z = call i16 @llvm.experimental.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[0]
 ; CHECK-NEXT:    vmov.i64 q1, #0xffff
 ; CHECK-NEXT:    vmov.32 q2[0], r2
 ; CHECK-NEXT:    vmov.u16 r2, q0[1]
 ; CHECK-NEXT:    vmov.32 q2[2], r2
+; CHECK-NEXT:    mov.w r12, #0
 ; CHECK-NEXT:    vand q2, q2, q1
 ; CHECK-NEXT:    vmov r2, s10
 ; CHECK-NEXT:    vmov r3, s8
 ; CHECK-NEXT:    add r2, r3
 ; CHECK-NEXT:    vmov.u16 r3, q0[2]
-; CHECK-NEXT:    vmov.32 q3[0], r3
+; CHECK-NEXT:    vmov.32 q2[0], r3
 ; CHECK-NEXT:    vmov.u16 r3, q0[3]
-; CHECK-NEXT:    vmov.32 q3[2], r3
-; CHECK-NEXT:    vand q3, q3, q1
-; CHECK-NEXT:    vmov r3, s12
+; CHECK-NEXT:    vmov.32 q2[2], r3
+; CHECK-NEXT:    vand q2, q2, q1
+; CHECK-NEXT:    vmov r3, s8
 ; CHECK-NEXT:    add r2, r3
-; CHECK-NEXT:    vmov r3, s14
+; CHECK-NEXT:    vmov r3, s10
 ; CHECK-NEXT:    add r2, r3
 ; CHECK-NEXT:    vmov.u16 r3, q0[4]
-; CHECK-NEXT:    vmov.32 q3[0], r3
+; CHECK-NEXT:    vmov.32 q2[0], r3
 ; CHECK-NEXT:    vmov.u16 r3, q0[5]
-; CHECK-NEXT:    vmov.32 q3[2], r3
-; CHECK-NEXT:    vand q3, q3, q1
-; CHECK-NEXT:    vmov r3, s12
-; CHECK-NEXT:    vmov lr, s15
-; CHECK-NEXT:    add.w r12, r2, r3
-; CHECK-NEXT:    vmov r3, s14
-; CHECK-NEXT:    vmov r2, s11
-; CHECK-NEXT:    adds.w r4, r12, r3
-; CHECK-NEXT:    adc.w r12, r2, lr
+; CHECK-NEXT:    vmov.32 q2[2], r3
+; CHECK-NEXT:    vand q2, q2, q1
+; CHECK-NEXT:    vmov r3, s8
+; CHECK-NEXT:    add r2, r3
+; CHECK-NEXT:    vmov r3, s10
+; CHECK-NEXT:    adds.w lr, r2, r3
 ; CHECK-NEXT:    vmov.u16 r2, q0[6]
 ; CHECK-NEXT:    vmov.32 q2[0], r2
 ; CHECK-NEXT:    vmov.u16 r2, q0[7]
 ; CHECK-NEXT:    vmov.32 q2[2], r2
+; CHECK-NEXT:    adc r3, r12, #0
 ; CHECK-NEXT:    vand q0, q2, 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 r12, lr, r2
 ; CHECK-NEXT:    vmov r2, s2
-; CHECK-NEXT:    adds r2, r2, r3
-; CHECK-NEXT:    adc.w r3, r12, r4
+; CHECK-NEXT:    adc r3, r3, #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 <8 x i16> %x to <8 x i64>
   %z = call i64 @llvm.experimental.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) {
entry:
   %z = call i64 @llvm.experimental.vector.reduce.add.v8i64(<8 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.experimental.vector.reduce.add.v2i64(<2 x i64> %xx)
   %r = add i64 %z, %a
   ret i64 %r
 }
 
entry:
   %z = call i8 @llvm.experimental.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[0]
 ; CHECK-NEXT:    vmov.i64 q1, #0xff
 ; CHECK-NEXT:    vmov.32 q2[0], r2
 ; CHECK-NEXT:    vmov.u8 r2, q0[1]
 ; CHECK-NEXT:    vmov.32 q2[2], r2
+; CHECK-NEXT:    mov.w r12, #0
 ; CHECK-NEXT:    vand q2, q2, q1
 ; CHECK-NEXT:    vmov r2, s10
 ; CHECK-NEXT:    vmov r3, s8
 ; CHECK-NEXT:    add r2, r3
 ; CHECK-NEXT:    vmov.u8 r3, q0[2]
-; CHECK-NEXT:    vmov.32 q3[0], r3
+; CHECK-NEXT:    vmov.32 q2[0], r3
 ; CHECK-NEXT:    vmov.u8 r3, q0[3]
-; CHECK-NEXT:    vmov.32 q3[2], r3
-; CHECK-NEXT:    vand q3, q3, q1
-; CHECK-NEXT:    vmov r3, s12
+; CHECK-NEXT:    vmov.32 q2[2], r3
+; CHECK-NEXT:    vand q2, q2, q1
+; CHECK-NEXT:    vmov r3, s8
 ; CHECK-NEXT:    add r2, r3
-; CHECK-NEXT:    vmov r3, s14
+; CHECK-NEXT:    vmov r3, s10
 ; CHECK-NEXT:    add r2, r3
 ; CHECK-NEXT:    vmov.u8 r3, q0[4]
-; CHECK-NEXT:    vmov.32 q3[0], r3
+; CHECK-NEXT:    vmov.32 q2[0], r3
 ; CHECK-NEXT:    vmov.u8 r3, q0[5]
-; CHECK-NEXT:    vmov.32 q3[2], r3
-; CHECK-NEXT:    vand q3, q3, q1
-; CHECK-NEXT:    vmov r3, s12
-; CHECK-NEXT:    vmov lr, s15
-; CHECK-NEXT:    add.w r12, r2, r3
-; CHECK-NEXT:    vmov r3, s14
-; CHECK-NEXT:    vmov r2, s11
-; CHECK-NEXT:    adds.w r4, r12, r3
-; CHECK-NEXT:    adc.w r12, r2, lr
+; CHECK-NEXT:    vmov.32 q2[2], r3
+; CHECK-NEXT:    vand q2, q2, q1
+; CHECK-NEXT:    vmov r3, s8
+; CHECK-NEXT:    add r2, r3
+; CHECK-NEXT:    vmov r3, s10
+; CHECK-NEXT:    adds.w lr, r2, r3
 ; CHECK-NEXT:    vmov.u8 r2, q0[6]
 ; CHECK-NEXT:    vmov.32 q2[0], r2
 ; CHECK-NEXT:    vmov.u8 r2, q0[7]
 ; CHECK-NEXT:    vmov.32 q2[2], r2
+; CHECK-NEXT:    adc r3, r12, #0
 ; 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 r12, lr, r2
 ; CHECK-NEXT:    vmov r2, s10
-; CHECK-NEXT:    adds.w lr, r3, r2
-; CHECK-NEXT:    adc.w r3, r12, r4
-; CHECK-NEXT:    vmov.u8 r4, q0[8]
-; CHECK-NEXT:    vmov.32 q2[0], r4
-; CHECK-NEXT:    vmov.u8 r4, q0[9]
-; CHECK-NEXT:    vmov.32 q2[2], r4
+; CHECK-NEXT:    adc r3, r3, #0
+; CHECK-NEXT:    adds.w r12, r12, r2
+; CHECK-NEXT:    vmov.u8 r2, q0[8]
+; CHECK-NEXT:    vmov.32 q2[0], r2
+; CHECK-NEXT:    vmov.u8 r2, q0[9]
+; CHECK-NEXT:    vmov.32 q2[2], r2
+; CHECK-NEXT:    adc r3, r3, #0
 ; CHECK-NEXT:    vand q2, q2, q1
 ; CHECK-NEXT:    vmov r2, s8
-; CHECK-NEXT:    vmov r4, s9
-; CHECK-NEXT:    adds.w r12, lr, r2
+; CHECK-NEXT:    adds.w r12, r12, r2
 ; CHECK-NEXT:    vmov r2, s10
-; CHECK-NEXT:    adcs r3, r4
-; CHECK-NEXT:    vmov r4, s11
+; CHECK-NEXT:    adc r3, r3, #0
 ; CHECK-NEXT:    adds.w r12, r12, r2
-; CHECK-NEXT:    adcs r3, r4
-; CHECK-NEXT:    vmov.u8 r4, q0[10]
-; CHECK-NEXT:    vmov.32 q2[0], r4
-; CHECK-NEXT:    vmov.u8 r4, q0[11]
-; CHECK-NEXT:    vmov.32 q2[2], r4
+; CHECK-NEXT:    vmov.u8 r2, q0[10]
+; CHECK-NEXT:    vmov.32 q2[0], r2
+; CHECK-NEXT:    vmov.u8 r2, q0[11]
+; CHECK-NEXT:    vmov.32 q2[2], r2
+; CHECK-NEXT:    adc r3, r3, #0
 ; CHECK-NEXT:    vand q2, q2, q1
 ; CHECK-NEXT:    vmov r2, s8
-; CHECK-NEXT:    vmov r4, s9
 ; CHECK-NEXT:    adds.w r12, r12, r2
 ; CHECK-NEXT:    vmov r2, s10
-; CHECK-NEXT:    adcs r3, r4
-; CHECK-NEXT:    vmov r4, s11
+; CHECK-NEXT:    adc r3, r3, #0
 ; CHECK-NEXT:    adds.w r12, r12, r2
-; CHECK-NEXT:    adcs r3, r4
-; CHECK-NEXT:    vmov.u8 r4, q0[12]
-; CHECK-NEXT:    vmov.32 q2[0], r4
-; CHECK-NEXT:    vmov.u8 r4, q0[13]
-; CHECK-NEXT:    vmov.32 q2[2], r4
+; CHECK-NEXT:    vmov.u8 r2, q0[12]
+; CHECK-NEXT:    vmov.32 q2[0], r2
+; CHECK-NEXT:    vmov.u8 r2, q0[13]
+; CHECK-NEXT:    vmov.32 q2[2], r2
+; CHECK-NEXT:    adc r3, r3, #0
 ; CHECK-NEXT:    vand q2, q2, q1
 ; CHECK-NEXT:    vmov r2, s8
-; CHECK-NEXT:    vmov r4, s9
 ; CHECK-NEXT:    adds.w r12, r12, r2
 ; CHECK-NEXT:    vmov r2, s10
-; CHECK-NEXT:    adcs r3, r4
-; CHECK-NEXT:    vmov r4, s11
+; CHECK-NEXT:    adc r3, r3, #0
 ; CHECK-NEXT:    adds.w r12, r12, r2
-; CHECK-NEXT:    adcs r3, r4
-; CHECK-NEXT:    vmov.u8 r4, q0[14]
-; CHECK-NEXT:    vmov.32 q2[0], r4
-; CHECK-NEXT:    vmov.u8 r4, q0[15]
-; CHECK-NEXT:    vmov.32 q2[2], r4
+; CHECK-NEXT:    vmov.u8 r2, q0[14]
+; CHECK-NEXT:    vmov.32 q2[0], r2
+; CHECK-NEXT:    vmov.u8 r2, q0[15]
+; CHECK-NEXT:    vmov.32 q2[2], r2
+; CHECK-NEXT:    adc r3, r3, #0
 ; CHECK-NEXT:    vand q0, q2, q1
 ; CHECK-NEXT:    vmov r2, s0
-; CHECK-NEXT:    vmov r4, s1
 ; CHECK-NEXT:    adds.w r12, r12, r2
 ; CHECK-NEXT:    vmov r2, s2
-; CHECK-NEXT:    adcs r3, r4
-; CHECK-NEXT:    vmov r4, s3
+; CHECK-NEXT:    adc r3, r3, #0
 ; CHECK-NEXT:    adds.w r2, r2, r12
-; CHECK-NEXT:    adcs r3, r4
+; 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.experimental.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) {
entry:
   %z = call i64 @llvm.experimental.vector.reduce.add.v16i64(<16 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.experimental.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.experimental.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, s4
-; CHECK-NEXT:    vmov.i64 q2, #0xffffffff
-; CHECK-NEXT:    vand q0, q0, q2
+; CHECK-NEXT:    movs r2, #0
 ; CHECK-NEXT:    cmp r0, #0
 ; CHECK-NEXT:    cset r0, eq
 ; CHECK-NEXT:    tst.w r0, #1
 ; CHECK-NEXT:    csetm r0, ne
 ; CHECK-NEXT:    vmov.32 q2[0], r0
-; CHECK-NEXT:    vmov.32 q2[1], r0
 ; CHECK-NEXT:    vmov r0, s6
 ; CHECK-NEXT:    cmp r0, #0
 ; CHECK-NEXT:    cset r0, eq
 ; CHECK-NEXT:    tst.w r0, #1
 ; CHECK-NEXT:    csetm r0, ne
 ; CHECK-NEXT:    vmov.32 q2[2], r0
-; CHECK-NEXT:    vmov.32 q2[3], r0
 ; CHECK-NEXT:    vand q0, q0, q2
 ; 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.experimental.vector.reduce.add.v2i64(<2 x i64> %s)
   ret i64 %z
 }

 ; CHECK-NEXT:    vmov.u16 r0, q2[3]
 ; CHECK-NEXT:    vmov.32 q1[3], r0
 ; CHECK-NEXT:    vcmp.i32 ne, q1, zr
 ; CHECK-NEXT:    vmov.i64 q1, #0xffff
 ; CHECK-NEXT:    vmrs r0, p0
 ; CHECK-NEXT:    and r1, r0, #1
 ; CHECK-NEXT:    rsbs r1, r1, #0
 ; CHECK-NEXT:    vmov.32 q3[0], r1
-; CHECK-NEXT:    vmov.32 q3[1], r1
 ; CHECK-NEXT:    ubfx r1, r0, #4, #1
 ; CHECK-NEXT:    rsbs r1, r1, #0
 ; CHECK-NEXT:    vmov.32 q3[2], r1
-; CHECK-NEXT:    vmov.32 q3[3], r1
 ; CHECK-NEXT:    vmov.u16 r1, q0[0]
 ; CHECK-NEXT:    vmov.32 q4[0], r1
 ; CHECK-NEXT:    vmov.u16 r1, q0[1]
 ; CHECK-NEXT:    vmov.32 q4[2], 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, #8, #1
-; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    vmov r1, s14
+; CHECK-NEXT:    vmov r2, s12
+; CHECK-NEXT:    add r1, r2
+; CHECK-NEXT:    ubfx r2, r0, #8, #1
 ; CHECK-NEXT:    ubfx r0, r0, #12, #1
-; CHECK-NEXT:    vmov.32 q3[0], r3
+; CHECK-NEXT:    rsbs r2, r2, #0
+; CHECK-NEXT:    vmov.32 q3[0], r2
 ; CHECK-NEXT:    rsbs r0, r0, #0
-; CHECK-NEXT:    vmov.32 q3[1], r3
 ; CHECK-NEXT:    vmov.32 q3[2], r0
-; CHECK-NEXT:    vmov.32 q3[3], r0
 ; CHECK-NEXT:    vmov.u16 r0, q0[2]
 ; CHECK-NEXT:    vmov.32 q4[0], r0
 ; CHECK-NEXT:    vmov.u16 r0, q0[3]
 ; CHECK-NEXT:    vmov.32 q4[2], 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, s15
-; CHECK-NEXT:    adcs r1, r0
-; CHECK-NEXT:    vmov r0, s14
-; CHECK-NEXT:    adds.w r12, r2, r0
+; 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 r0, r0, r2
 ; CHECK-NEXT:    vmov.u16 r2, q2[4]
 ; CHECK-NEXT:    vmov.32 q3[0], r2
 ; CHECK-NEXT:    vmov.u16 r2, q2[5]
 ; CHECK-NEXT:    vmov.32 q3[1], r2
 ; CHECK-NEXT:    vmov.u16 r2, q2[6]
 ; CHECK-NEXT:    vmov.32 q3[2], r2
 ; CHECK-NEXT:    vmov.u16 r2, q2[7]
 ; CHECK-NEXT:    vmov.32 q3[3], r2
-; CHECK-NEXT:    adcs r1, r3
+; CHECK-NEXT:    adc r1, r1, #0
 ; CHECK-NEXT:    vcmp.i32 ne, q3, zr
 ; CHECK-NEXT:    vmrs r2, p0
 ; CHECK-NEXT:    and r3, r2, #1
 ; CHECK-NEXT:    rsbs r3, r3, #0
 ; CHECK-NEXT:    vmov.32 q2[0], r3
-; CHECK-NEXT:    vmov.32 q2[1], r3
 ; CHECK-NEXT:    ubfx r3, r2, #4, #1
 ; CHECK-NEXT:    rsbs r3, r3, #0
 ; CHECK-NEXT:    vmov.32 q2[2], r3
-; CHECK-NEXT:    vmov.32 q2[3], r3
 ; CHECK-NEXT:    vmov.u16 r3, q0[4]
 ; CHECK-NEXT:    vmov.32 q3[0], r3
 ; CHECK-NEXT:    vmov.u16 r3, q0[5]
 ; CHECK-NEXT:    vmov.32 q3[2], r3
 ; CHECK-NEXT:    vand q3, q3, q1
 ; CHECK-NEXT:    vand q2, q3, q2
-; CHECK-NEXT:    vmov r0, s8
-; CHECK-NEXT:    vmov r3, s9
-; CHECK-NEXT:    adds.w r12, r12, r0
-; CHECK-NEXT:    vmov r0, s10
-; CHECK-NEXT:    adcs r1, r3
-; CHECK-NEXT:    vmov r3, s11
-; CHECK-NEXT:    adds.w r0, r0, r12
-; CHECK-NEXT:    adcs r1, r3
+; CHECK-NEXT:    vmov r3, s8
+; CHECK-NEXT:    adds r0, r0, r3
+; CHECK-NEXT:    vmov r3, s10
+; CHECK-NEXT:    adc r1, r1, #0
+; CHECK-NEXT:    adds r0, r0, r3
 ; CHECK-NEXT:    ubfx r3, r2, #8, #1
-; CHECK-NEXT:    rsbs r3, r3, #0
 ; CHECK-NEXT:    ubfx r2, r2, #12, #1
+; CHECK-NEXT:    rsb.w r3, r3, #0
 ; CHECK-NEXT:    vmov.32 q2[0], r3
-; CHECK-NEXT:    rsbs r2, r2, #0
-; CHECK-NEXT:    vmov.32 q2[1], r3
+; CHECK-NEXT:    rsb.w r2, r2, #0
 ; CHECK-NEXT:    vmov.32 q2[2], r2
-; CHECK-NEXT:    vmov.32 q2[3], r2
 ; CHECK-NEXT:    vmov.u16 r2, q0[6]
 ; CHECK-NEXT:    vmov.32 q3[0], r2
 ; CHECK-NEXT:    vmov.u16 r2, q0[7]
 ; CHECK-NEXT:    vmov.32 q3[2], r2
+; CHECK-NEXT:    adc r1, r1, #0
 ; CHECK-NEXT:    vand q0, q3, 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}
 ; 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.experimental.vector.reduce.add.v8i64(<8 x i64> %s)
   ret i64 %z

 ; CHECK-NEXT:    vand q1, q1, q2
 ; CHECK-NEXT:    vand q0, q0, q2
 ; CHECK-NEXT:    vmov r0, s4
 ; CHECK-NEXT:    cmp r0, #0
 ; CHECK-NEXT:    cset r0, eq
 ; CHECK-NEXT:    tst.w r0, #1
 ; CHECK-NEXT:    csetm r0, ne
 ; CHECK-NEXT:    vmov.32 q3[0], r0
-; CHECK-NEXT:    vmov.32 q3[1], r0
 ; CHECK-NEXT:    vmov r0, s6
 ; CHECK-NEXT:    cmp r0, #0
 ; CHECK-NEXT:    cset r0, eq
 ; CHECK-NEXT:    tst.w r0, #1
 ; CHECK-NEXT:    csetm r0, ne
 ; CHECK-NEXT:    vmov.32 q3[2], r0
-; CHECK-NEXT:    vmov.32 q3[3], r0
 ; CHECK-NEXT:    vand q0, q0, q3
 ; 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.experimental.vector.reduce.add.v2i64(<2 x i64> %s)
   ret i64 %z
 }

 ; CHECK-NEXT:    vmov.u16 r0, q5[3]
 ; CHECK-NEXT:    vmov.32 q1[3], r0
 ; CHECK-NEXT:    vcmp.i32 ne, q1, zr
 ; CHECK-NEXT:    vmov.i64 q1, #0xff
 ; CHECK-NEXT:    vmrs r0, p0
 ; CHECK-NEXT:    and r1, r0, #1
 ; CHECK-NEXT:    rsbs r1, r1, #0
 ; CHECK-NEXT:    vmov.32 q6[0], r1
-; CHECK-NEXT:    vmov.32 q6[1], r1
 ; CHECK-NEXT:    ubfx r1, r0, #4, #1
 ; CHECK-NEXT:    rsbs r1, r1, #0
 ; CHECK-NEXT:    vmov.32 q6[2], r1
-; CHECK-NEXT:    vmov.32 q6[3], r1
 ; CHECK-NEXT:    vmov.u8 r1, q0[0]
 ; CHECK-NEXT:    vmov.32 q7[0], r1
 ; CHECK-NEXT:    vmov.u8 r1, q0[1]
 ; CHECK-NEXT:    vmov.32 q7[2], 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, #8, #1
-; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    vmov r1, s26
+; CHECK-NEXT:    vmov r2, s24
+; CHECK-NEXT:    add r1, r2
+; CHECK-NEXT:    ubfx r2, r0, #8, #1
 ; CHECK-NEXT:    ubfx r0, r0, #12, #1
-; CHECK-NEXT:    vmov.32 q6[0], r3
+; CHECK-NEXT:    rsbs r2, r2, #0
+; CHECK-NEXT:    vmov.32 q6[0], r2
 ; CHECK-NEXT:    rsbs r0, r0, #0
-; CHECK-NEXT:    vmov.32 q6[1], r3
 ; CHECK-NEXT:    vmov.32 q6[2], r0
-; CHECK-NEXT:    vmov.32 q6[3], r0
 ; CHECK-NEXT:    vmov.u8 r0, q0[2]
 ; CHECK-NEXT:    vmov.32 q7[0], r0
 ; CHECK-NEXT:    vmov.u8 r0, q0[3]
 ; CHECK-NEXT:    vmov.32 q7[2], 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, s27
-; CHECK-NEXT:    adcs r1, r0
-; CHECK-NEXT:    vmov r0, s26
-; CHECK-NEXT:    adds.w r12, r2, r0
+; 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 r0, r0, r2
 ; CHECK-NEXT:    vmov.u16 r2, q5[4]
 ; CHECK-NEXT:    vmov.32 q6[0], r2
 ; CHECK-NEXT:    vmov.u16 r2, q5[5]
 ; CHECK-NEXT:    vmov.32 q6[1], r2
 ; CHECK-NEXT:    vmov.u16 r2, q5[6]
 ; CHECK-NEXT:    vmov.32 q6[2], r2
 ; CHECK-NEXT:    vmov.u16 r2, q5[7]
 ; CHECK-NEXT:    vmov.32 q6[3], r2
-; CHECK-NEXT:    adcs r1, r3
+; CHECK-NEXT:    adc r1, r1, #0
 ; CHECK-NEXT:    vcmp.i32 ne, q6, zr
 ; CHECK-NEXT:    vmrs r2, p0
 ; CHECK-NEXT:    and r3, r2, #1
 ; CHECK-NEXT:    rsbs r3, r3, #0
 ; CHECK-NEXT:    vmov.32 q5[0], r3
-; CHECK-NEXT:    vmov.32 q5[1], r3
 ; CHECK-NEXT:    ubfx r3, r2, #4, #1
 ; CHECK-NEXT:    rsbs r3, r3, #0
 ; CHECK-NEXT:    vmov.32 q5[2], r3
-; CHECK-NEXT:    vmov.32 q5[3], r3
 ; CHECK-NEXT:    vmov.u8 r3, q0[4]
 ; CHECK-NEXT:    vmov.32 q6[0], r3
 ; CHECK-NEXT:    vmov.u8 r3, q0[5]
 ; CHECK-NEXT:    vmov.32 q6[2], r3
 ; CHECK-NEXT:    vand q6, q6, q1
 ; CHECK-NEXT:    vand q5, q6, q5
-; CHECK-NEXT:    vmov r0, s20
-; CHECK-NEXT:    vmov r3, s21
-; CHECK-NEXT:    adds.w r12, r12, r0
-; CHECK-NEXT:    vmov r0, s22
-; CHECK-NEXT:    adcs r1, r3
-; CHECK-NEXT:    vmov r3, s23
-; CHECK-NEXT:    adds.w r0, r0, r12
-; CHECK-NEXT:    adcs r1, r3
+; CHECK-NEXT:    vmov r3, s20
+; CHECK-NEXT:    adds r0, r0, r3
+; CHECK-NEXT:    vmov r3, s22
+; CHECK-NEXT:    adc r1, r1, #0
+; CHECK-NEXT:    adds r0, r0, r3
 ; CHECK-NEXT:    ubfx r3, r2, #8, #1
-; CHECK-NEXT:    rsbs r3, r3, #0
 ; CHECK-NEXT:    ubfx r2, r2, #12, #1
+; CHECK-NEXT:    rsb.w r3, r3, #0
 ; CHECK-NEXT:    vmov.32 q5[0], r3
-; CHECK-NEXT:    rsbs r2, r2, #0
-; CHECK-NEXT:    vmov.32 q5[1], r3
+; CHECK-NEXT:    rsb.w r2, r2, #0
 ; CHECK-NEXT:    vmov.32 q5[2], r2
-; CHECK-NEXT:    vmov.32 q5[3], r2
 ; CHECK-NEXT:    vmov.u8 r2, q0[6]
 ; CHECK-NEXT:    vmov.32 q6[0], r2
 ; CHECK-NEXT:    vmov.u8 r2, q0[7]
 ; CHECK-NEXT:    vmov.32 q6[2], r2
+; CHECK-NEXT:    adc r1, r1, #0
 ; 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 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 r0, 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[0]
 ; CHECK-NEXT:    vmov.32 q3[0], r2
 ; CHECK-NEXT:    vmov.u16 r2, q2[1]
 ; CHECK-NEXT:    vmov.32 q3[1], r2
 ; CHECK-NEXT:    vmov.u16 r2, q2[2]
 ; CHECK-NEXT:    vmov.32 q3[2], r2
 ; CHECK-NEXT:    vmov.u16 r2, q2[3]
 ; CHECK-NEXT:    vmov.32 q3[3], r2
 ; CHECK-NEXT:    vcmp.i32 ne, q3, zr
 ; CHECK-NEXT:    vmrs r2, p0
 ; CHECK-NEXT:    and r3, r2, #1
 ; CHECK-NEXT:    rsbs r3, r3, #0
 ; CHECK-NEXT:    vmov.32 q3[0], r3
-; CHECK-NEXT:    vmov.32 q3[1], r3
 ; CHECK-NEXT:    ubfx r3, r2, #4, #1
 ; CHECK-NEXT:    rsbs r3, r3, #0
 ; CHECK-NEXT:    vmov.32 q3[2], r3
-; CHECK-NEXT:    vmov.32 q3[3], r3
 ; CHECK-NEXT:    vmov.u8 r3, q0[8]
 ; CHECK-NEXT:    vmov.32 q4[0], r3
 ; CHECK-NEXT:    vmov.u8 r3, q0[9]
 ; CHECK-NEXT:    vmov.32 q4[2], r3
 ; CHECK-NEXT:    vand q4, q4, q1
 ; CHECK-NEXT:    vand q3, q4, q3
-; CHECK-NEXT:    vmov r0, s12
-; CHECK-NEXT:    vmov r3, s13
-; CHECK-NEXT:    adds.w r12, r12, r0
-; CHECK-NEXT:    vmov r0, s14
-; CHECK-NEXT:    adcs r1, r3
-; CHECK-NEXT:    vmov r3, s15
-; CHECK-NEXT:    adds.w r0, r0, r12
-; CHECK-NEXT:    adcs r1, r3
+; CHECK-NEXT:    vmov r3, s12
+; CHECK-NEXT:    adds r0, r0, r3
+; CHECK-NEXT:    vmov r3, s14
+; CHECK-NEXT:    adc r1, r1, #0
+; CHECK-NEXT:    adds r0, r0, r3
 ; CHECK-NEXT:    ubfx r3, r2, #8, #1
-; CHECK-NEXT:    rsbs r3, r3, #0
 ; CHECK-NEXT:    ubfx r2, r2, #12, #1
+; CHECK-NEXT:    rsb.w r3, r3, #0
 ; CHECK-NEXT:    vmov.32 q3[0], r3
-; CHECK-NEXT:    rsbs r2, r2, #0
-; CHECK-NEXT:    vmov.32 q3[1], r3
+; CHECK-NEXT:    rsb.w r2, r2, #0
 ; CHECK-NEXT:    vmov.32 q3[2], r2
-; CHECK-NEXT:    vmov.32 q3[3], r2
 ; CHECK-NEXT:    vmov.u8 r2, q0[10]
 ; CHECK-NEXT:    vmov.32 q4[0], r2
 ; CHECK-NEXT:    vmov.u8 r2, q0[11]
 ; CHECK-NEXT:    vmov.32 q4[2], r2
+; CHECK-NEXT:    adc r1, r1, #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 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:    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 r0, r0, r2
 ; CHECK-NEXT:    vmov.u16 r2, q2[4]
 ; CHECK-NEXT:    vmov.32 q3[0], r2
 ; CHECK-NEXT:    vmov.u16 r2, q2[5]
 ; CHECK-NEXT:    vmov.32 q3[1], r2
 ; CHECK-NEXT:    vmov.u16 r2, q2[6]
 ; CHECK-NEXT:    vmov.32 q3[2], r2
 ; CHECK-NEXT:    vmov.u16 r2, q2[7]
 ; CHECK-NEXT:    vmov.32 q3[3], r2
+; CHECK-NEXT:    adc r1, r1, #0
 ; CHECK-NEXT:    vcmp.i32 ne, q3, zr
 ; CHECK-NEXT:    vmrs r2, p0
 ; CHECK-NEXT:    and r3, r2, #1
 ; CHECK-NEXT:    rsbs r3, r3, #0
 ; CHECK-NEXT:    vmov.32 q2[0], r3
-; CHECK-NEXT:    vmov.32 q2[1], r3
 ; CHECK-NEXT:    ubfx r3, r2, #4, #1
 ; CHECK-NEXT:    rsbs r3, r3, #0
 ; CHECK-NEXT:    vmov.32 q2[2], r3
-; CHECK-NEXT:    vmov.32 q2[3], r3
 ; CHECK-NEXT:    vmov.u8 r3, q0[12]
 ; CHECK-NEXT:    vmov.32 q3[0], r3
 ; CHECK-NEXT:    vmov.u8 r3, q0[13]
 ; CHECK-NEXT:    vmov.32 q3[2], r3
 ; CHECK-NEXT:    vand q3, q3, q1
 ; CHECK-NEXT:    vand q2, q3, q2
-; CHECK-NEXT:    vmov r0, s8
-; CHECK-NEXT:    vmov r3, s9
-; CHECK-NEXT:    adds.w r12, r12, r0
-; CHECK-NEXT:    vmov r0, s10
-; CHECK-NEXT:    adcs r1, r3
-; CHECK-NEXT:    vmov r3, s11
-; CHECK-NEXT:    adds.w r0, r0, r12
-; CHECK-NEXT:    adcs r1, r3
+; CHECK-NEXT:    vmov r3, s8
+; CHECK-NEXT:    adds r0, r0, r3
+; CHECK-NEXT:    vmov r3, s10
+; CHECK-NEXT:    adc r1, r1, #0
+; CHECK-NEXT:    adds r0, r0, r3
 ; CHECK-NEXT:    ubfx r3, r2, #8, #1
-; CHECK-NEXT:    rsbs r3, r3, #0
 ; CHECK-NEXT:    ubfx r2, r2, #12, #1
+; CHECK-NEXT:    rsb.w r3, r3, #0
 ; CHECK-NEXT:    vmov.32 q2[0], r3
-; CHECK-NEXT:    rsbs r2, r2, #0
-; CHECK-NEXT:    vmov.32 q2[1], r3
+; CHECK-NEXT:    rsb.w r2, r2, #0
 ; CHECK-NEXT:    vmov.32 q2[2], r2
-; CHECK-NEXT:    vmov.32 q2[3], r2
 ; CHECK-NEXT:    vmov.u8 r2, q0[14]
 ; CHECK-NEXT:    vmov.32 q3[0], r2
 ; CHECK-NEXT:    vmov.u8 r2, q0[15]
 ; CHECK-NEXT:    vmov.32 q3[2], r2
+; CHECK-NEXT:    adc r1, r1, #0
 ; CHECK-NEXT:    vand q0, q3, 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.experimental.vector.reduce.add.v16i64(<16 x i64> %s)
   ret i64 %z

 ; CHECK-NEXT:    vand q1, q1, q2
 ; CHECK-NEXT:    vand q0, q0, q2
 ; CHECK-NEXT:    vmov r0, s4
 ; CHECK-NEXT:    cmp r0, #0
 ; CHECK-NEXT:    cset r0, eq
 ; CHECK-NEXT:    tst.w r0, #1
 ; CHECK-NEXT:    csetm r0, ne
 ; CHECK-NEXT:    vmov.32 q3[0], r0
-; CHECK-NEXT:    vmov.32 q3[1], r0
 ; CHECK-NEXT:    vmov r0, s6
 ; CHECK-NEXT:    cmp r0, #0
 ; CHECK-NEXT:    cset r0, eq
 ; CHECK-NEXT:    tst.w r0, #1
 ; CHECK-NEXT:    csetm r0, ne
 ; CHECK-NEXT:    vmov.32 q3[2], r0
-; CHECK-NEXT:    vmov.32 q3[3], r0
 ; CHECK-NEXT:    vand q0, q0, q3
 ; 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.experimental.vector.reduce.add.v2i64(<2 x i64> %s)
   ret i64 %z
 }
entry:
   %z = call i64 @llvm.experimental.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, s4
-; CHECK-NEXT:    vmov.i64 q2, #0xffffffff
-; CHECK-NEXT:    vand q0, q0, q2
+; CHECK-NEXT:    mov.w r12, #0
 ; CHECK-NEXT:    cmp r2, #0
 ; CHECK-NEXT:    cset r2, eq
 ; CHECK-NEXT:    tst.w r2, #1
 ; CHECK-NEXT:    csetm r2, ne
 ; CHECK-NEXT:    vmov.32 q2[0], r2
-; CHECK-NEXT:    vmov.32 q2[1], r2
 ; CHECK-NEXT:    vmov r2, s6
 ; CHECK-NEXT:    cmp r2, #0
 ; CHECK-NEXT:    cset r2, eq
 ; CHECK-NEXT:    tst.w r2, #1
 ; CHECK-NEXT:    csetm r2, ne
 ; CHECK-NEXT:    vmov.32 q2[2], r2
-; CHECK-NEXT:    vmov.32 q2[3], r2
 ; CHECK-NEXT:    vand q0, q0, q2
 ; 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.experimental.vector.reduce.add.v2i64(<2 x i64> %s)
   %r = add i64 %z, %a
   ret i64 %r
 }

 ; CHECK-NEXT:    vmov.u16 r2, q2[3]
 ; CHECK-NEXT:    vmov.32 q1[3], r2
 ; CHECK-NEXT:    vcmp.i32 ne, q1, zr
 ; CHECK-NEXT:    vmov.i64 q1, #0xffff
 ; CHECK-NEXT:    vmrs r2, p0
 ; CHECK-NEXT:    and r3, r2, #1
 ; CHECK-NEXT:    rsbs r3, r3, #0
 ; CHECK-NEXT:    vmov.32 q3[0], r3
-; CHECK-NEXT:    vmov.32 q3[1], r3
 ; CHECK-NEXT:    ubfx r3, r2, #4, #1
 ; CHECK-NEXT:    rsbs r3, r3, #0
 ; CHECK-NEXT:    vmov.32 q3[2], r3
-; CHECK-NEXT:    vmov.32 q3[3], r3
 ; CHECK-NEXT:    vmov.u16 r3, q0[0]
 ; CHECK-NEXT:    vmov.32 q4[0], r3
 ; CHECK-NEXT:    vmov.u16 r3, q0[1]
 ; CHECK-NEXT:    vmov.32 q4[2], r3
 ; 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, #8, #1
-; CHECK-NEXT:    rsbs r3, r3, #0
 ; CHECK-NEXT:    ubfx r2, r2, #12, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
 ; CHECK-NEXT:    vmov.32 q3[0], r3
 ; CHECK-NEXT:    rsbs r2, r2, #0
-; CHECK-NEXT:    vmov.32 q3[1], r3
 ; CHECK-NEXT:    vmov.32 q3[2], r2
-; CHECK-NEXT:    vmov.32 q3[3], r2
 ; CHECK-NEXT:    vmov.u16 r2, q0[2]
 ; CHECK-NEXT:    vmov.32 q4[0], r2
 ; CHECK-NEXT:    vmov.u16 r2, q0[3]
 ; CHECK-NEXT:    vmov.32 q4[2], 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 r4, lr, r3
-; CHECK-NEXT:    vmov r3, s14
-; CHECK-NEXT:    adc.w lr, r12, r2
-; CHECK-NEXT:    vmov r2, s15
-; CHECK-NEXT:    adds.w r12, r4, r3
-; CHECK-NEXT:    adc.w lr, lr, 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[4]
 ; CHECK-NEXT:    vmov.32 q3[0], r2
 ; CHECK-NEXT:    vmov.u16 r2, q2[5]
 ; CHECK-NEXT:    vmov.32 q3[1], r2
 ; CHECK-NEXT:    vmov.u16 r2, q2[6]
 ; CHECK-NEXT:    vmov.32 q3[2], r2
 ; CHECK-NEXT:    vmov.u16 r2, q2[7]
 ; CHECK-NEXT:    vmov.32 q3[3], r2
+; CHECK-NEXT:    adc lr, r3, #0
 ; CHECK-NEXT:    vcmp.i32 ne, q3, zr
 ; CHECK-NEXT:    vmrs r2, p0
-; CHECK-NEXT:    and r4, r2, #1
-; CHECK-NEXT:    rsbs r4, r4, #0
-; CHECK-NEXT:    vmov.32 q2[0], r4
-; CHECK-NEXT:    vmov.32 q2[1], r4
-; CHECK-NEXT:    ubfx r4, r2, #4, #1
-; CHECK-NEXT:    rsbs r4, r4, #0
-; CHECK-NEXT:    vmov.32 q2[2], r4
-; CHECK-NEXT:    vmov.32 q2[3], r4
-; CHECK-NEXT:    vmov.u16 r4, q0[4]
-; CHECK-NEXT:    vmov.32 q3[0], r4
-; CHECK-NEXT:    vmov.u16 r4, q0[5]
-; CHECK-NEXT:    vmov.32 q3[2], r4
+; CHECK-NEXT:    and r3, r2, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    vmov.32 q2[0], r3
+; CHECK-NEXT:    ubfx r3, r2, #4, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    vmov.32 q2[2], r3
+; CHECK-NEXT:    vmov.u16 r3, q0[4]
+; CHECK-NEXT:    vmov.32 q3[0], r3
+; CHECK-NEXT:    vmov.u16 r3, q0[5]
+; CHECK-NEXT:    vmov.32 q3[2], r3
 ; CHECK-NEXT:    vand q3, q3, q1
 ; CHECK-NEXT:    vand q2, q3, q2
 ; CHECK-NEXT:    vmov r3, s8
-; CHECK-NEXT:    vmov r4, s9
-; CHECK-NEXT:    adds.w r12, r12, r3
-; CHECK-NEXT:    vmov r3, s11
-; CHECK-NEXT:    adc.w lr, lr, r4
-; CHECK-NEXT:    vmov r4, s10
-; CHECK-NEXT:    adds.w r4, r4, r12
-; CHECK-NEXT:    adc.w r12, lr, r3
+; CHECK-NEXT:    adds.w r4, r12, r3
+; CHECK-NEXT:    vmov r3, s10
+; CHECK-NEXT:    adc r12, lr, #0
+; CHECK-NEXT:    adds.w lr, r4, r3
 ; CHECK-NEXT:    ubfx r3, r2, #8, #1
-; CHECK-NEXT:    rsbs r3, r3, #0
 ; CHECK-NEXT:    ubfx r2, r2, #12, #1
+; CHECK-NEXT:    rsb.w r3, r3, #0
 ; CHECK-NEXT:    vmov.32 q2[0], r3
-; CHECK-NEXT:    rsbs r2, r2, #0
-; CHECK-NEXT:    vmov.32 q2[1], r3
+; CHECK-NEXT:    rsb.w r2, r2, #0
 ; CHECK-NEXT:    vmov.32 q2[2], r2
-; CHECK-NEXT:    vmov.32 q2[3], r2
 ; CHECK-NEXT:    vmov.u16 r2, q0[6]
 ; CHECK-NEXT:    vmov.32 q3[0], r2
 ; CHECK-NEXT:    vmov.u16 r2, q0[7]
 ; CHECK-NEXT:    vmov.32 q3[2], r2
+; CHECK-NEXT:    adc r4, r12, #0
 ; CHECK-NEXT:    vand q0, q3, 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, lr
+; 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:    vand q1, q1, q2
 ; CHECK-NEXT:    vand q0, q0, q2
 ; CHECK-NEXT:    vmov r2, s4
 ; CHECK-NEXT:    cmp r2, #0
 ; CHECK-NEXT:    cset r2, eq
 ; CHECK-NEXT:    tst.w r2, #1
 ; CHECK-NEXT:    csetm r2, ne
 ; CHECK-NEXT:    vmov.32 q3[0], r2
-; CHECK-NEXT:    vmov.32 q3[1], r2
 ; CHECK-NEXT:    vmov r2, s6
 ; CHECK-NEXT:    cmp r2, #0
 ; CHECK-NEXT:    cset r2, eq
 ; CHECK-NEXT:    tst.w r2, #1
 ; CHECK-NEXT:    csetm r2, ne
 ; CHECK-NEXT:    vmov.32 q3[2], r2
-; CHECK-NEXT:    vmov.32 q3[3], r2
 ; CHECK-NEXT:    vand q0, q0, q3
-; 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.experimental.vector.reduce.add.v2i64(<2 x i64> %s)
   %r = add i64 %z, %a
   ret i64 %r

 ; CHECK-NEXT:    vmov.u16 r2, q5[3]
 ; CHECK-NEXT:    vmov.32 q1[3], r2
 ; CHECK-NEXT:    vcmp.i32 ne, q1, zr
 ; CHECK-NEXT:    vmov.i64 q1, #0xff
 ; CHECK-NEXT:    vmrs r2, p0
 ; CHECK-NEXT:    and r3, r2, #1
 ; CHECK-NEXT:    rsbs r3, r3, #0
 ; CHECK-NEXT:    vmov.32 q6[0], r3
-; CHECK-NEXT:    vmov.32 q6[1], r3
 ; CHECK-NEXT:    ubfx r3, r2, #4, #1
 ; CHECK-NEXT:    rsbs r3, r3, #0
 ; CHECK-NEXT:    vmov.32 q6[2], r3
-; CHECK-NEXT:    vmov.32 q6[3], r3
 ; CHECK-NEXT:    vmov.u8 r3, q0[0]
 ; CHECK-NEXT:    vmov.32 q7[0], r3
 ; CHECK-NEXT:    vmov.u8 r3, q0[1]
 ; CHECK-NEXT:    vmov.32 q7[2], r3
 ; 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, #8, #1
-; CHECK-NEXT:    rsbs r3, r3, #0
 ; CHECK-NEXT:    ubfx r2, r2, #12, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
 ; CHECK-NEXT:    vmov.32 q6[0], r3
 ; CHECK-NEXT:    rsbs r2, r2, #0
-; CHECK-NEXT:    vmov.32 q6[1], r3
 ; CHECK-NEXT:    vmov.32 q6[2], r2
-; CHECK-NEXT:    vmov.32 q6[3], r2
 ; CHECK-NEXT:    vmov.u8 r2, q0[2]
 ; CHECK-NEXT:    vmov.32 q7[0], r2
 ; CHECK-NEXT:    vmov.u8 r2, q0[3]
 ; CHECK-NEXT:    vmov.32 q7[2], 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 r4, lr, r3
-; CHECK-NEXT:    vmov r3, s26
-; CHECK-NEXT:    adc.w lr, r12, r2
-; CHECK-NEXT:    vmov r2, s27
-; CHECK-NEXT:    adds.w r12, r4, r3
-; CHECK-NEXT:    adc.w lr, lr, 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[4]
 ; CHECK-NEXT:    vmov.32 q6[0], r2
 ; CHECK-NEXT:    vmov.u16 r2, q5[5]
 ; CHECK-NEXT:    vmov.32 q6[1], r2
 ; CHECK-NEXT:    vmov.u16 r2, q5[6]
 ; CHECK-NEXT:    vmov.32 q6[2], r2
 ; CHECK-NEXT:    vmov.u16 r2, q5[7]
 ; CHECK-NEXT:    vmov.32 q6[3], r2
+; CHECK-NEXT:    adc lr, r3, #0
 ; CHECK-NEXT:    vcmp.i32 ne, q6, zr
 ; CHECK-NEXT:    vmrs r2, p0
-; CHECK-NEXT:    and r4, r2, #1
-; CHECK-NEXT:    rsbs r4, r4, #0
-; CHECK-NEXT:    vmov.32 q5[0], r4
-; CHECK-NEXT:    vmov.32 q5[1], r4
-; CHECK-NEXT:    ubfx r4, r2, #4, #1
-; CHECK-NEXT:    rsbs r4, r4, #0
-; CHECK-NEXT:    vmov.32 q5[2], r4
-; CHECK-NEXT:    vmov.32 q5[3], r4
-; CHECK-NEXT:    vmov.u8 r4, q0[4]
-; CHECK-NEXT:    vmov.32 q6[0], r4
-; CHECK-NEXT:    vmov.u8 r4, q0[5]
-; CHECK-NEXT:    vmov.32 q6[2], r4
+; CHECK-NEXT:    and r3, r2, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    vmov.32 q5[0], r3
+; CHECK-NEXT:    ubfx r3, r2, #4, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
+; CHECK-NEXT:    vmov.32 q5[2], r3
+; CHECK-NEXT:    vmov.u8 r3, q0[4]
+; CHECK-NEXT:    vmov.32 q6[0], r3
+; CHECK-NEXT:    vmov.u8 r3, q0[5]
+; CHECK-NEXT:    vmov.32 q6[2], r3
 ; CHECK-NEXT:    vand q6, q6, q1
 ; CHECK-NEXT:    vand q5, q6, q5
 ; CHECK-NEXT:    vmov r3, s20
-; CHECK-NEXT:    vmov r4, s21
-; CHECK-NEXT:    adds.w r12, r12, r3
-; CHECK-NEXT:    vmov r3, s23
-; CHECK-NEXT:    adc.w lr, lr, r4
-; CHECK-NEXT:    vmov r4, s22
-; CHECK-NEXT:    adds.w r4, r4, r12
-; CHECK-NEXT:    adc.w r12, lr, r3
+; CHECK-NEXT:    adds.w r4, r12, r3
+; CHECK-NEXT:    vmov r3, s22
+; CHECK-NEXT:    adc r12, lr, #0
+; CHECK-NEXT:    adds.w lr, r4, r3
 ; CHECK-NEXT:    ubfx r3, r2, #8, #1
-; CHECK-NEXT:    rsbs r3, r3, #0
 ; CHECK-NEXT:    ubfx r2, r2, #12, #1
+; CHECK-NEXT:    rsb.w r3, r3, #0
 ; CHECK-NEXT:    vmov.32 q5[0], r3
-; CHECK-NEXT:    rsbs r2, r2, #0
-; CHECK-NEXT:    vmov.32 q5[1], r3
+; CHECK-NEXT:    rsb.w r2, r2, #0
 ; CHECK-NEXT:    vmov.32 q5[2], r2
-; CHECK-NEXT:    vmov.32 q5[3], r2
 ; CHECK-NEXT:    vmov.u8 r2, q0[6]
 ; CHECK-NEXT:    vmov.32 q6[0], r2
 ; CHECK-NEXT:    vmov.u8 r2, q0[7]
 ; CHECK-NEXT:    vmov.32 q6[2], r2
+; CHECK-NEXT:    adc r4, r12, #0
 ; 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 r3, r3, r4
-; CHECK-NEXT:    vmov r4, s23
-; CHECK-NEXT:    adc.w lr, r12, r2
-; CHECK-NEXT:    vmov r2, s22
-; CHECK-NEXT:    adds.w r12, r3, 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.w lr, lr, r4
+; CHECK-NEXT:    vmov r2, s20
+; CHECK-NEXT:    adds.w r2, r2, lr
+; 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 r3, r3, #0
 ; CHECK-NEXT:    vcmp.i16 ne, q5, zr
 ; CHECK-NEXT:    vpsel q2, q3, q2
-; CHECK-NEXT:    vmov.u16 r2, q2[0]
-; CHECK-NEXT:    vmov.32 q3[0], r2
-; CHECK-NEXT:    vmov.u16 r2, q2[1]
-; CHECK-NEXT:    vmov.32 q3[1], r2
-; CHECK-NEXT:    vmov.u16 r2, q2[2]
-; CHECK-NEXT:    vmov.32 q3[2], r2
-; CHECK-NEXT:    vmov.u16 r2, q2[3]
-; CHECK-NEXT:    vmov.32 q3[3], r2
-; CHECK-NEXT:    vcmp.i32 ne, q3, zr
-; CHECK-NEXT:    vmrs r2, p0
-; CHECK-NEXT:    and r4, r2, #1
-; CHECK-NEXT:    rsbs r4, r4, #0
+; CHECK-NEXT:    vmov.u16 r4, q2[0]
 ; CHECK-NEXT:    vmov.32 q3[0], r4
+; CHECK-NEXT:    vmov.u16 r4, q2[1]
 ; CHECK-NEXT:    vmov.32 q3[1], r4
-; CHECK-NEXT:    ubfx r4, r2, #4, #1
-; CHECK-NEXT:    rsbs r4, r4, #0
+; CHECK-NEXT:    vmov.u16 r4, q2[2]
 ; CHECK-NEXT:    vmov.32 q3[2], r4
+; CHECK-NEXT:    vmov.u16 r4, q2[3]
 ; CHECK-NEXT:    vmov.32 q3[3], r4
-; CHECK-NEXT:    vmov.u8 r4, q0[8]
-; CHECK-NEXT:    vmov.32 q4[0], r4
-; CHECK-NEXT:    vmov.u8 r4, q0[9]
-; CHECK-NEXT:    vmov.32 q4[2], r4
+; CHECK-NEXT:    vcmp.i32 ne, q3, zr
+; CHECK-NEXT:    vmrs r4, p0
+; CHECK-NEXT:    and r2, r4, #1
+; CHECK-NEXT:    rsbs r2, r2, #0
+; CHECK-NEXT:    vmov.32 q3[0], r2
+; CHECK-NEXT:    ubfx r2, r4, #4, #1
+; CHECK-NEXT:    rsbs r2, r2, #0
+; CHECK-NEXT:    vmov.32 q3[2], r2
+; CHECK-NEXT:    vmov.u8 r2, q0[8]
+; CHECK-NEXT:    vmov.32 q4[0], r2
+; CHECK-NEXT:    vmov.u8 r2, q0[9]
+; CHECK-NEXT:    vmov.32 q4[2], r2
 ; CHECK-NEXT:    vand q4, q4, q1
 ; CHECK-NEXT:    vand q3, q4, q3
-; CHECK-NEXT:    vmov r3, s12
-; CHECK-NEXT:    vmov r4, s13
-; CHECK-NEXT:    adds.w r12, r12, r3
-; CHECK-NEXT:    vmov r3, s15
-; CHECK-NEXT:    adc.w lr, lr, r4
-; CHECK-NEXT:    vmov r4, s14
-; CHECK-NEXT:    adds.w r4, r4, r12
-; CHECK-NEXT:    adc.w r12, lr, r3
-; CHECK-NEXT:    ubfx r3, r2, #8, #1
-; CHECK-NEXT:    rsbs r3, r3, #0
-; CHECK-NEXT:    ubfx r2, r2, #12, #1
-; CHECK-NEXT:    vmov.32 q3[0], r3
+; 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:    ubfx r2, r4, #8, #1
+; CHECK-NEXT:    rsb.w r2, r2, #0
+; CHECK-NEXT:    adc r3, r3, #0
+; CHECK-NEXT:    vmov.32 q3[0], r2
+; CHECK-NEXT:    ubfx r2, r4, #12, #1
 ; CHECK-NEXT:    rsbs r2, r2, #0
-; CHECK-NEXT:    vmov.32 q3[1], r3
 ; CHECK-NEXT:    vmov.32 q3[2], r2
-; CHECK-NEXT:    vmov.32 q3[3], r2
 ; CHECK-NEXT:    vmov.u8 r2, q0[10]
 ; CHECK-NEXT:    vmov.32 q4[0], r2
 ; CHECK-NEXT:    vmov.u8 r2, q0[11]
 ; CHECK-NEXT:    vmov.32 q4[2], r2
 ; 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 r3, r3, r4
-; CHECK-NEXT:    vmov r4, s15
-; CHECK-NEXT:    adc.w lr, r12, r2
-; CHECK-NEXT:    vmov r2, s14
-; CHECK-NEXT:    adds.w r12, r3, r2
-; CHECK-NEXT:    vmov.u16 r2, q2[4]
-; CHECK-NEXT:    vmov.32 q3[0], r2
-; CHECK-NEXT:    vmov.u16 r2, q2[5]
-; CHECK-NEXT:    vmov.32 q3[1], r2
-; CHECK-NEXT:    vmov.u16 r2, q2[6]
-; CHECK-NEXT:    vmov.32 q3[2], r2
-; CHECK-NEXT:    vmov.u16 r2, q2[7]
-; CHECK-NEXT:    vmov.32 q3[3], r2
-; CHECK-NEXT:    adc.w lr, lr, r4
-; CHECK-NEXT:    vcmp.i32 ne, q3, zr
-; CHECK-NEXT:    vmrs r2, p0
-; CHECK-NEXT:    and r4, r2, #1
-; CHECK-NEXT:    rsbs r4, r4, #0
-; CHECK-NEXT:    vmov.32 q2[0], r4
-; CHECK-NEXT:    vmov.32 q2[1], r4
-; CHECK-NEXT:    ubfx r4, r2, #4, #1
-; CHECK-NEXT:    rsbs r4, r4, #0
-; CHECK-NEXT:    vmov.32 q2[2], r4
-; CHECK-NEXT:    vmov.32 q2[3], r4
-; CHECK-NEXT:    vmov.u8 r4, q0[12]
+; CHECK-NEXT:    vmov r2, s12
+; CHECK-NEXT:    vmov r4, s14
+; CHECK-NEXT:    adds.w r2, r2, r12
+; CHECK-NEXT:    adc r3, r3, #0
+; CHECK-NEXT:    adds.w r12, r2, r4
+; CHECK-NEXT:    vmov.u16 r4, q2[4]
+; CHECK-NEXT:    adc r3, r3, #0
 ; CHECK-NEXT:    vmov.32 q3[0], r4
-; CHECK-NEXT:    vmov.u8 r4, q0[13]
+; CHECK-NEXT:    vmov.u16 r4, q2[5]
+; CHECK-NEXT:    vmov.32 q3[1], r4
+; CHECK-NEXT:    vmov.u16 r4, q2[6]
 ; CHECK-NEXT:    vmov.32 q3[2], r4
+; CHECK-NEXT:    vmov.u16 r4, q2[7]
+; CHECK-NEXT:    vmov.32 q3[3], r4
+; CHECK-NEXT:    vcmp.i32 ne, q3, zr
+; CHECK-NEXT:    vmrs r4, p0
+; CHECK-NEXT:    and r2, r4, #1
+; CHECK-NEXT:    rsbs r2, r2, #0
+; CHECK-NEXT:    vmov.32 q2[0], r2
+; CHECK-NEXT:    ubfx r2, r4, #4, #1
+; CHECK-NEXT:    rsbs r2, r2, #0
+; CHECK-NEXT:    vmov.32 q2[2], r2
+; CHECK-NEXT:    vmov.u8 r2, q0[12]
+; CHECK-NEXT:    vmov.32 q3[0], r2
+; CHECK-NEXT:    vmov.u8 r2, q0[13]
+; CHECK-NEXT:    vmov.32 q3[2], r2
 ; CHECK-NEXT:    vand q3, q3, q1
 ; CHECK-NEXT:    vand q2, q3, q2
-; CHECK-NEXT:    vmov r3, s8
-; CHECK-NEXT:    vmov r4, s9
-; CHECK-NEXT:    adds.w r12, r12, r3
-; CHECK-NEXT:    vmov r3, s11
-; CHECK-NEXT:    adc.w lr, lr, r4
-; CHECK-NEXT:    vmov r4, s10
-; CHECK-NEXT:    adds.w r4, r4, r12
-; CHECK-NEXT:    adc.w r12, lr, r3
-; CHECK-NEXT:    ubfx r3, r2, #8, #1
-; CHECK-NEXT:    rsbs r3, r3, #0
-; CHECK-NEXT:    ubfx r2, r2, #12, #1
-; CHECK-NEXT:    vmov.32 q2[0], r3
+; CHECK-NEXT:    vmov r2, s8
+; CHECK-NEXT:    adds.w r12, r12, r2
+; CHECK-NEXT:    vmov r2, s10
+; CHECK-NEXT:    adc r3, r3, #0
+; CHECK-NEXT:    adds.w r12, r12, r2
+; CHECK-NEXT:    ubfx r2, r4, #8, #1
+; CHECK-NEXT:    rsb.w r2, r2, #0
+; CHECK-NEXT:    adc r3, r3, #0
+; CHECK-NEXT:    vmov.32 q2[0], r2
+; CHECK-NEXT:    ubfx r2, r4, #12, #1
 ; CHECK-NEXT:    rsbs r2, r2, #0
-; CHECK-NEXT:    vmov.32 q2[1], r3
 ; CHECK-NEXT:    vmov.32 q2[2], r2
-; CHECK-NEXT:    vmov.32 q2[3], r2
 ; CHECK-NEXT:    vmov.u8 r2, q0[14]
 ; CHECK-NEXT:    vmov.32 q3[0], r2
 ; CHECK-NEXT:    vmov.u8 r2, q0[15]
 ; CHECK-NEXT:    vmov.32 q3[2], r2
 ; CHECK-NEXT:    vand q0, q3, 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:    vmov r4, s2
+; CHECK-NEXT:    adds.w r2, r2, r12
+; CHECK-NEXT:    adc r3, r3, #0
+; 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, d10, d11, d12, d13, d14, d15}
 ; 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

 ; CHECK-NEXT:    vand q1, q1, q2
 ; CHECK-NEXT:    vand q0, q0, q2
 ; CHECK-NEXT:    vmov r2, s4
 ; CHECK-NEXT:    cmp r2, #0
 ; CHECK-NEXT:    cset r2, eq
 ; CHECK-NEXT:    tst.w r2, #1
 ; CHECK-NEXT:    csetm r2, ne
 ; CHECK-NEXT:    vmov.32 q3[0], r2
-; CHECK-NEXT:    vmov.32 q3[1], r2
 ; CHECK-NEXT:    vmov r2, s6
 ; CHECK-NEXT:    cmp r2, #0
 ; CHECK-NEXT:    cset r2, eq
 ; CHECK-NEXT:    tst.w r2, #1
 ; CHECK-NEXT:    csetm r2, ne
 ; CHECK-NEXT:    vmov.32 q3[2], r2
-; CHECK-NEXT:    vmov.32 q3[3], r2
 ; CHECK-NEXT:    vand q0, q0, q3
-; 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.experimental.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.experimental.vector.reduce.add.v8i64(<8 x i64> %m)
   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:    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:    vmov.32 q0[0], r1
+; CHECK-NEXT:    vmov.32 q0[2], r0
 ; 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:    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.experimental.vector.reduce.add.v2i64(<2 x i64> %m)
   ret i64 %z
 }
 
entry:
   %m = mul <16 x i8> %x, %y
   %z = call i8 @llvm.experimental.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 {r7, lr}
-; CHECK-NEXT:    push {r7, lr}
-; CHECK-NEXT:    .vsave {d8, d9, d10, d11}
-; CHECK-NEXT:    vpush {d8, d9, d10, d11}
-; CHECK-NEXT:    vmov.u8 r0, q1[0]
-; CHECK-NEXT:    vmov.u8 r1, q0[0]
+; CHECK-NEXT:    .vsave {d8, d9}
+; CHECK-NEXT:    vpush {d8, d9}
+; CHECK-NEXT:    vmov.u8 r0, q1[2]
+; CHECK-NEXT:    vmov.u8 r1, q0[2]
 ; CHECK-NEXT:    vmov.32 q3[0], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[1]
+; CHECK-NEXT:    vmov.u8 r0, q1[3]
 ; CHECK-NEXT:    vmov.32 q4[0], r1
-; CHECK-NEXT:    vmov.u8 r1, q0[1]
+; CHECK-NEXT:    vmov.u8 r1, q0[3]
 ; CHECK-NEXT:    vmov.32 q3[2], r0
 ; CHECK-NEXT:    vmov.i64 q2, #0xff
 ; CHECK-NEXT:    vmov.32 q4[2], 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 r12, r1, r1, r0
-; CHECK-NEXT:    vmov.u8 r0, q0[2]
-; CHECK-NEXT:    vmov.32 q4[0], r0
-; CHECK-NEXT:    vmov.u8 r0, q0[3]
-; CHECK-NEXT:    vmov.32 q4[2], r0
-; CHECK-NEXT:    umull r2, r3, r3, r2
+; CHECK-NEXT:    vmov.u8 r3, q0[0]
+; CHECK-NEXT:    vmov r2, s16
+; CHECK-NEXT:    vmov.32 q4[0], r3
+; CHECK-NEXT:    vmov.u8 r3, q0[1]
+; CHECK-NEXT:    vmov.32 q4[2], r3
 ; CHECK-NEXT:    vand q4, q4, q2
-; CHECK-NEXT:    vmov r0, s16
-; CHECK-NEXT:    orr.w lr, r3, r1
-; CHECK-NEXT:    vmov.u8 r3, q1[2]
-; CHECK-NEXT:    vmov.32 q3[0], r3
-; CHECK-NEXT:    vmov.u8 r3, q1[3]
-; CHECK-NEXT:    vmov.32 q3[2], r3
-; CHECK-NEXT:    add r2, r12
+; CHECK-NEXT:    vmov r3, s16
+; CHECK-NEXT:    muls r0, r1, r0
+; CHECK-NEXT:    vmov r1, s12
+; CHECK-NEXT:    muls r1, r2, r1
+; CHECK-NEXT:    vmov.u8 r2, q1[0]
+; CHECK-NEXT:    vmov.32 q3[0], r1
+; CHECK-NEXT:    vmov r1, s18
+; CHECK-NEXT:    vmov.32 q3[2], r0
+; CHECK-NEXT:    vmov r12, s12
+; CHECK-NEXT:    vmov.32 q3[0], r2
+; CHECK-NEXT:    vmov.u8 r2, q1[1]
+; CHECK-NEXT:    vmov.32 q3[2], r2
 ; CHECK-NEXT:    vand q3, q3, q2
-; CHECK-NEXT:    vmov r3, s12
-; CHECK-NEXT:    umull r0, r3, r0, r3
-; CHECK-NEXT:    vmov.32 q5[0], r0
-; CHECK-NEXT:    vmov r0, s14
-; CHECK-NEXT:    vmov.32 q5[1], r3
-; CHECK-NEXT:    vmov r3, s18
-; CHECK-NEXT:    umull r0, r3, r3, r0
-; CHECK-NEXT:    vmov.32 q5[2], r0
-; CHECK-NEXT:    vmov.32 q5[3], r3
-; CHECK-NEXT:    vmov r1, s20
-; CHECK-NEXT:    vmov r0, s21
-; CHECK-NEXT:    adds r1, r1, r2
-; CHECK-NEXT:    adc.w r2, lr, r0
-; CHECK-NEXT:    vmov r0, s22
+; CHECK-NEXT:    vmov r2, s12
+; CHECK-NEXT:    muls r2, r3, r2
+; CHECK-NEXT:    vmov r3, s14
+; CHECK-NEXT:    smlabb r1, r1, r3, r2
+; CHECK-NEXT:    movs r2, #0
+; CHECK-NEXT:    vmov.u8 r3, q0[4]
+; CHECK-NEXT:    vmov.32 q4[0], r3
+; CHECK-NEXT:    vmov.u8 r3, q0[5]
+; CHECK-NEXT:    vmov.32 q4[2], r3
+; CHECK-NEXT:    adds.w r1, r1, r12
+; CHECK-NEXT:    vand q4, q4, q2
+; CHECK-NEXT:    adc r2, r2, #0
 ; CHECK-NEXT:    adds.w r12, r1, r0
-; CHECK-NEXT:    adc.w r1, r2, r3
+; CHECK-NEXT:    adc r1, r2, #0
 ; CHECK-NEXT:    vmov.u8 r2, q1[4]
-; CHECK-NEXT:    vmov.u8 r3, q0[4]
 ; CHECK-NEXT:    vmov.32 q3[0], r2
 ; CHECK-NEXT:    vmov.u8 r2, q1[5]
-; CHECK-NEXT:    vmov.32 q4[0], r3
-; CHECK-NEXT:    vmov.u8 r3, q0[5]
 ; CHECK-NEXT:    vmov.32 q3[2], r2
-; CHECK-NEXT:    vmov.32 q4[2], r3
+; CHECK-NEXT:    vmov r3, s18
 ; CHECK-NEXT:    vand q3, q3, q2
-; CHECK-NEXT:    vand q4, q4, q2
-; CHECK-NEXT:    vmov r2, s12
-; CHECK-NEXT:    vmov r3, s16
-; CHECK-NEXT:    umull r2, r3, r3, r2
-; CHECK-NEXT:    vmov.32 q5[0], r2
+; CHECK-NEXT:    vmov r0, s16
 ; CHECK-NEXT:    vmov r2, s14
-; CHECK-NEXT:    vmov.32 q5[1], r3
+; CHECK-NEXT:    muls r2, r3, r2
+; CHECK-NEXT:    vmov r3, s12
+; CHECK-NEXT:    muls r0, r3, r0
+; CHECK-NEXT:    vmov.u8 r3, q0[6]
+; CHECK-NEXT:    vmov.32 q4[0], r3
+; CHECK-NEXT:    vmov.u8 r3, q0[7]
+; CHECK-NEXT:    vmov.32 q3[0], r0
+; CHECK-NEXT:    vmov.32 q4[2], r3
+; CHECK-NEXT:    vmov.32 q3[2], r2
+; CHECK-NEXT:    vand q4, q4, q2
+; CHECK-NEXT:    vmov r0, s12
 ; CHECK-NEXT:    vmov r3, s18
-; CHECK-NEXT:    umull r2, r3, r3, r2
-; CHECK-NEXT:    vmov.32 q5[2], r2
-; CHECK-NEXT:    vmov.32 q5[3], r3
-; CHECK-NEXT:    vmov r0, s20
-; CHECK-NEXT:    vmov r2, s21
 ; CHECK-NEXT:    adds.w r0, r0, r12
-; CHECK-NEXT:    adcs r1, r2
-; CHECK-NEXT:    vmov r2, s22
+; CHECK-NEXT:    adc r1, r1, #0
 ; CHECK-NEXT:    adds.w r12, r0, r2
 ; CHECK-NEXT:    vmov.u8 r2, q1[6]
-; CHECK-NEXT:    adcs r1, r3
-; CHECK-NEXT:    vmov.u8 r3, q0[6]
+; CHECK-NEXT:    vmov r0, s16
 ; CHECK-NEXT:    vmov.32 q3[0], r2
 ; CHECK-NEXT:    vmov.u8 r2, q1[7]
-; CHECK-NEXT:    vmov.32 q4[0], r3
-; CHECK-NEXT:    vmov.u8 r3, q0[7]
 ; CHECK-NEXT:    vmov.32 q3[2], r2
-; CHECK-NEXT:    vmov.32 q4[2], r3
+; CHECK-NEXT:    adc r1, r1, #0
 ; CHECK-NEXT:    vand q3, q3, q2
-; CHECK-NEXT:    vand q4, q4, q2
-; CHECK-NEXT:    vmov r2, s12
-; CHECK-NEXT:    vmov r3, s16
-; CHECK-NEXT:    umull r2, r3, r3, r2
-; CHECK-NEXT:    vmov.32 q5[0], r2
 ; CHECK-NEXT:    vmov r2, s14
-; CHECK-NEXT:    vmov.32 q5[1], r3
+; CHECK-NEXT:    muls r2, r3, r2
+; CHECK-NEXT:    vmov r3, s12
+; CHECK-NEXT:    muls r0, r3, r0
+; CHECK-NEXT:    vmov.u8 r3, q0[8]
+; CHECK-NEXT:    vmov.32 q4[0], r3
+; CHECK-NEXT:    vmov.u8 r3, q0[9]
+; CHECK-NEXT:    vmov.32 q3[0], r0
+; CHECK-NEXT:    vmov.32 q4[2], r3
+; CHECK-NEXT:    vmov.32 q3[2], r2
+; CHECK-NEXT:    vand q4, q4, q2
+; CHECK-NEXT:    vmov r0, s12
 ; CHECK-NEXT:    vmov r3, s18
-; CHECK-NEXT:    umull r2, r3, r3, r2
-; CHECK-NEXT:    vmov.32 q5[2], r2
-; CHECK-NEXT:    vmov.32 q5[3], r3
-; CHECK-NEXT:    vmov r0, s20
-; CHECK-NEXT:    vmov r2, s21
 ; CHECK-NEXT:    adds.w r0, r0, r12
-; CHECK-NEXT:    adcs r1, r2
-; CHECK-NEXT:    vmov r2, s22
+; CHECK-NEXT:    adc r1, r1, #0
 ; CHECK-NEXT:    adds.w r12, r0, r2
 ; CHECK-NEXT:    vmov.u8 r2, q1[8]
-; CHECK-NEXT:    adcs r1, r3
-; CHECK-NEXT:    vmov.u8 r3, q0[8]
+; CHECK-NEXT:    vmov r0, s16
 ; CHECK-NEXT:    vmov.32 q3[0], r2
 ; CHECK-NEXT:    vmov.u8 r2, q1[9]
-; CHECK-NEXT:    vmov.32 q4[0], r3
-; CHECK-NEXT:    vmov.u8 r3, q0[9]
 ; CHECK-NEXT:    vmov.32 q3[2], r2
-; CHECK-NEXT:    vmov.32 q4[2], r3
+; CHECK-NEXT:    adc r1, r1, #0
 ; CHECK-NEXT:    vand q3, q3, q2
-; CHECK-NEXT:    vand q4, q4, q2
-; CHECK-NEXT:    vmov r2, s12
-; CHECK-NEXT:    vmov r3, s16
-; CHECK-NEXT:    umull r2, r3, r3, r2
-; CHECK-NEXT:    vmov.32 q5[0], r2
 ; CHECK-NEXT:    vmov r2, s14
-; CHECK-NEXT:    vmov.32 q5[1], r3
+; CHECK-NEXT:    muls r2, r3, r2
+; CHECK-NEXT:    vmov r3, s12
+; CHECK-NEXT:    muls r0, r3, r0
+; CHECK-NEXT:    vmov.u8 r3, q0[10]
+; CHECK-NEXT:    vmov.32 q4[0], r3
+; CHECK-NEXT:    vmov.u8 r3, q0[11]
+; CHECK-NEXT:    vmov.32 q3[0], r0
+; CHECK-NEXT:    vmov.32 q4[2], r3
+; CHECK-NEXT:    vmov.32 q3[2], r2
+; CHECK-NEXT:    vand q4, q4, q2
+; CHECK-NEXT:    vmov r0, s12
 ; CHECK-NEXT:    vmov r3, s18
-; CHECK-NEXT:    umull r2, r3, r3, r2
-; CHECK-NEXT:    vmov.32 q5[2], r2
-; CHECK-NEXT:    vmov.32 q5[3], r3
-; CHECK-NEXT:    vmov r0, s20
-; CHECK-NEXT:    vmov r2, s21
 ; CHECK-NEXT:    adds.w r0, r0, r12
-; CHECK-NEXT:    adcs r1, r2
-; CHECK-NEXT:    vmov r2, s22
+; CHECK-NEXT:    adc r1, r1, #0
 ; CHECK-NEXT:    adds.w r12, r0, r2
 ; CHECK-NEXT:    vmov.u8 r2, q1[10]
-; CHECK-NEXT:    adcs r1, r3
-; CHECK-NEXT:    vmov.u8 r3, q0[10]
+; CHECK-NEXT:    vmov r0, s16
 ; CHECK-NEXT:    vmov.32 q3[0], r2
 ; CHECK-NEXT:    vmov.u8 r2, q1[11]
-; CHECK-NEXT:    vmov.32 q4[0], r3
-; CHECK-NEXT:    vmov.u8 r3, q0[11]
 ; CHECK-NEXT:    vmov.32 q3[2], r2
-; CHECK-NEXT:    vmov.32 q4[2], r3
+; CHECK-NEXT:    adc r1, r1, #0
 ; CHECK-NEXT:    vand q3, q3, q2
-; CHECK-NEXT:    vand q4, q4, q2
-; CHECK-NEXT:    vmov r2, s12
-; CHECK-NEXT:    vmov r3, s16
-; CHECK-NEXT:    umull r2, r3, r3, r2
-; CHECK-NEXT:    vmov.32 q5[0], r2
 ; CHECK-NEXT:    vmov r2, s14
-; CHECK-NEXT:    vmov.32 q5[1], r3
+; CHECK-NEXT:    muls r2, r3, r2
+; CHECK-NEXT:    vmov r3, s12
+; CHECK-NEXT:    muls r0, r3, r0
+; CHECK-NEXT:    vmov.u8 r3, q0[12]
+; CHECK-NEXT:    vmov.32 q4[0], r3
+; CHECK-NEXT:    vmov.u8 r3, q0[13]
+; CHECK-NEXT:    vmov.32 q3[0], r0
+; CHECK-NEXT:    vmov.32 q4[2], r3
+; CHECK-NEXT:    vmov.32 q3[2], r2
+; CHECK-NEXT:    vand q4, q4, q2
+; CHECK-NEXT:    vmov r0, s12
 ; CHECK-NEXT:    vmov r3, s18
-; CHECK-NEXT:    umull r2, r3, r3, r2
-; CHECK-NEXT:    vmov.32 q5[2], r2
-; CHECK-NEXT:    vmov.32 q5[3], r3
-; CHECK-NEXT:    vmov r0, s20
-; CHECK-NEXT:    vmov r2, s21
 ; CHECK-NEXT:    adds.w r0, r0, r12
-; CHECK-NEXT:    adcs r1, r2
-; CHECK-NEXT:    vmov r2, s22
+; CHECK-NEXT:    adc r1, r1, #0
 ; CHECK-NEXT:    adds.w r12, r0, r2
 ; CHECK-NEXT:    vmov.u8 r2, q1[12]
-; CHECK-NEXT:    adcs r1, r3
-; CHECK-NEXT:    vmov.u8 r3, q0[12]
+; CHECK-NEXT:    vmov r0, s16
 ; CHECK-NEXT:    vmov.32 q3[0], r2
 ; CHECK-NEXT:    vmov.u8 r2, q1[13]
-; CHECK-NEXT:    vmov.32 q4[0], r3
-; CHECK-NEXT:    vmov.u8 r3, q0[13]
 ; CHECK-NEXT:    vmov.32 q3[2], r2
-; CHECK-NEXT:    vmov.32 q4[2], r3
+; CHECK-NEXT:    adc r1, r1, #0
 ; CHECK-NEXT:    vand q3, q3, q2
-; CHECK-NEXT:    vand q4, q4, q2
-; CHECK-NEXT:    vmov r2, s12
-; CHECK-NEXT:    vmov r3, s16
-; CHECK-NEXT:    umull r2, r3, r3, r2
-; CHECK-NEXT:    vmov.32 q5[0], r2
 ; CHECK-NEXT:    vmov r2, s14
-; CHECK-NEXT:    vmov.32 q5[1], r3
-; CHECK-NEXT:    vmov r3, s18
-; CHECK-NEXT:    umull r2, r3, r3, r2
-; CHECK-NEXT:    vmov.32 q5[2], r2
-; CHECK-NEXT:    vmov.32 q5[3], r3
-; CHECK-NEXT:    vmov r0, s20
-; CHECK-NEXT:    vmov r2, s21
+; CHECK-NEXT:    muls r2, r3, r2
+; CHECK-NEXT:    vmov r3, s12
+; CHECK-NEXT:    muls r0, r3, r0
+; CHECK-NEXT:    vmov.u8 r3, q0[14]
+; CHECK-NEXT:    vmov.32 q3[0], r0
+; CHECK-NEXT:    vmov.32 q3[2], r2
+; CHECK-NEXT:    vmov r0, s12
 ; CHECK-NEXT:    adds.w r0, r0, r12
-; CHECK-NEXT:    adcs r1, r2
-; CHECK-NEXT:    vmov r2, s22
-; CHECK-NEXT:    adds r0, r0, r2
+; CHECK-NEXT:    adc r1, r1, #0
+; CHECK-NEXT:    adds.w r12, r0, r2
 ; CHECK-NEXT:    vmov.u8 r2, q1[14]
+; CHECK-NEXT:    adc r1, r1, #0
 ; CHECK-NEXT:    vmov.32 q3[0], r2
 ; CHECK-NEXT:    vmov.u8 r2, q1[15]
-; CHECK-NEXT:    adcs r1, r3
 ; CHECK-NEXT:    vmov.32 q3[2], r2
-; CHECK-NEXT:    vmov.u8 r3, q0[14]
 ; CHECK-NEXT:    vand q1, q3, q2
 ; CHECK-NEXT:    vmov.32 q3[0], r3
 ; CHECK-NEXT:    vmov.u8 r3, q0[15]
+; CHECK-NEXT:    vmov r2, s6
 ; CHECK-NEXT:    vmov.32 q3[2], r3
-; CHECK-NEXT:    vmov r2, s4
 ; CHECK-NEXT:    vand q0, q3, q2
-; CHECK-NEXT:    vmov r3, s0
-; CHECK-NEXT:    umlal r0, r1, r3, r2
-; CHECK-NEXT:    vmov r2, s6
 ; CHECK-NEXT:    vmov r3, s2
-; CHECK-NEXT:    umlal r0, r1, r3, r2
-; CHECK-NEXT:    vpop {d8, d9, d10, d11}
-; CHECK-NEXT:    pop {r7, pc}
+; CHECK-NEXT:    vmov r0, s0
+; CHECK-NEXT:    muls r2, r3, r2
+; CHECK-NEXT:    vmov r3, s4
+; CHECK-NEXT:    muls r0, r3, r0
+; CHECK-NEXT:    vmov.32 q0[0], r0
+; CHECK-NEXT:    vmov.32 q0[2], r2
+; CHECK-NEXT:    vmov r0, s0
+; CHECK-NEXT:    adds.w r0, r0, r12
+; 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:
   %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.experimental.vector.reduce.add.v16i64(<16 x i64> %m)
   ret i64 %z
 }
 
entry:
   %z = call i64 @llvm.experimental.vector.reduce.add.v16i64(<16 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.experimental.vector.reduce.add.v2i64(<2 x i64> %m)
   ret i64 %z
 }
entry:
   %z = call i64 @llvm.experimental.vector.reduce.add.v8i64(<8 x i64> %m)
   %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:    vmov r12, s4
+; CHECK-NEXT:    muls r2, r3, r2
+; CHECK-NEXT:    vmov r3, s0
+; CHECK-NEXT:    mul r3, r3, r12
+; CHECK-NEXT:    mov.w r12, #0
+; CHECK-NEXT:    vmov.32 q0[0], r3
+; CHECK-NEXT:    vmov.32 q0[2], r2
+; CHECK-NEXT:    vmov r3, s0
+; CHECK-NEXT:    adds r2, r2, r3
+; CHECK-NEXT:    adc r3, r12, #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.experimental.vector.reduce.add.v2i64(<2 x i64> %m)
   %r = add i64 %z, %a
   ret i64 %r
 }
entry:
   %z = call i8 @llvm.experimental.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, r7, lr}
-; CHECK-NEXT:    push {r4, r5, r7, lr}
-; CHECK-NEXT:    .vsave {d8, d9, d10, d11}
-; CHECK-NEXT:    vpush {d8, d9, d10, d11}
-; CHECK-NEXT:    vmov.u8 r2, q1[0]
-; CHECK-NEXT:    vmov.u8 r3, q0[0]
+; 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[2]
+; CHECK-NEXT:    vmov.u8 r3, q0[2]
 ; CHECK-NEXT:    vmov.32 q3[0], r2
-; CHECK-NEXT:    vmov.u8 r2, q1[1]
+; CHECK-NEXT:    vmov.u8 r2, q1[3]
 ; CHECK-NEXT:    vmov.32 q4[0], r3
-; CHECK-NEXT:    vmov.u8 r3, q0[1]
+; CHECK-NEXT:    vmov.u8 r3, q0[3]
 ; CHECK-NEXT:    vmov.32 q3[2], r2
 ; CHECK-NEXT:    vmov.i64 q2, #0xff
 ; CHECK-NEXT:    vmov.32 q4[2], r3
 ; CHECK-NEXT:    vand q3, q3, q2
 ; CHECK-NEXT:    vand q4, q4, q2
 ; CHECK-NEXT:    vmov r2, s14
 ; CHECK-NEXT:    vmov r3, s18
-; CHECK-NEXT:    vmov.u8 r4, q0[2]
-; CHECK-NEXT:    umull r12, lr, r3, r2
-; CHECK-NEXT:    vmov r3, s16
-; CHECK-NEXT:    vmov r2, s12
-; CHECK-NEXT:    vmov.32 q4[0], r4
-; CHECK-NEXT:    vmov.u8 r4, q0[3]
-; CHECK-NEXT:    vmov.32 q4[2], r4
-; CHECK-NEXT:    vand q4, q4, q2
-; CHECK-NEXT:    vmov r4, s16
-; CHECK-NEXT:    umull r2, r3, r3, r2
-; CHECK-NEXT:    orr.w lr, lr, r3
-; CHECK-NEXT:    vmov.u8 r3, q1[2]
+; CHECK-NEXT:    mul r12, r3, r2
+; CHECK-NEXT:    vmov r3, s12
+; CHECK-NEXT:    vmov r2, s16
+; CHECK-NEXT:    muls r2, r3, r2
+; CHECK-NEXT:    vmov.u8 r3, q1[0]
+; CHECK-NEXT:    vmov.32 q3[0], r2
+; CHECK-NEXT:    vmov.u8 r2, q0[0]
+; CHECK-NEXT:    vmov.32 q3[2], r12
+; CHECK-NEXT:    vmov.32 q4[0], r2
+; CHECK-NEXT:    vmov lr, s12
 ; CHECK-NEXT:    vmov.32 q3[0], r3
-; CHECK-NEXT:    vmov.u8 r3, q1[3]
+; CHECK-NEXT:    vmov.u8 r3, q1[1]
+; CHECK-NEXT:    vmov.u8 r2, q0[1]
 ; CHECK-NEXT:    vmov.32 q3[2], r3
-; CHECK-NEXT:    add r2, r12
+; CHECK-NEXT:    vmov.32 q4[2], r2
+; CHECK-NEXT:    vand q4, q4, q2
 ; CHECK-NEXT:    vand q3, q3, q2
 ; CHECK-NEXT:    vmov r3, s12
-; CHECK-NEXT:    umull r3, r4, r4, r3
-; CHECK-NEXT:    vmov.32 q5[0], r3
-; CHECK-NEXT:    vmov r3, s14
-; CHECK-NEXT:    vmov.32 q5[1], r4
+; CHECK-NEXT:    vmov r2, s16
 ; CHECK-NEXT:    vmov r4, s18
-; CHECK-NEXT:    umull r3, r4, r4, r3
-; CHECK-NEXT:    vmov.32 q5[2], r3
-; CHECK-NEXT:    vmov.32 q5[3], r4
-; CHECK-NEXT:    vmov r3, s20
-; CHECK-NEXT:    vmov r5, s21
-; CHECK-NEXT:    adds r2, r2, r3
-; CHECK-NEXT:    adc.w r3, lr, r5
-; CHECK-NEXT:    vmov r5, s22
-; CHECK-NEXT:    adds.w r12, r2, r5
-; CHECK-NEXT:    vmov.u8 r5, q1[4]
-; CHECK-NEXT:    adcs r3, r4
-; CHECK-NEXT:    vmov.u8 r4, q0[4]
-; CHECK-NEXT:    vmov.32 q3[0], r5
-; CHECK-NEXT:    vmov.u8 r5, q1[5]
-; CHECK-NEXT:    vmov.32 q4[0], r4
-; CHECK-NEXT:    vmov.u8 r4, q0[5]
-; CHECK-NEXT:    vmov.32 q3[2], r5
-; CHECK-NEXT:    vmov.32 q4[2], r4
+; CHECK-NEXT:    muls r2, r3, r2
+; CHECK-NEXT:    vmov r3, s14
+; CHECK-NEXT:    smlabb r2, r4, r3, r2
+; CHECK-NEXT:    movs r3, #0
+; CHECK-NEXT:    vmov.u8 r4, q1[4]
+; CHECK-NEXT:    vmov.32 q3[0], r4
+; CHECK-NEXT:    vmov.u8 r4, q1[5]
+; CHECK-NEXT:    vmov.32 q3[2], r4
+; CHECK-NEXT:    adds.w r2, r2, lr
 ; CHECK-NEXT:    vand q3, q3, q2
+; CHECK-NEXT:    adc r3, r3, #0
+; CHECK-NEXT:    adds.w r12, r12, r2
+; CHECK-NEXT:    vmov.u8 r2, q0[4]
+; CHECK-NEXT:    vmov r4, s14
+; CHECK-NEXT:    vmov.32 q4[0], r2
+; CHECK-NEXT:    vmov.u8 r2, q0[5]
+; CHECK-NEXT:    vmov.32 q4[2], r2
+; CHECK-NEXT:    adc lr, r3, #0
 ; CHECK-NEXT:    vand q4, q4, q2
-; CHECK-NEXT:    vmov r5, s12
-; CHECK-NEXT:    vmov r4, s16
-; CHECK-NEXT:    umull r5, r4, r4, r5
-; CHECK-NEXT:    vmov.32 q5[0], r5
-; CHECK-NEXT:    vmov r5, s14
-; CHECK-NEXT:    vmov.32 q5[1], r4
-; CHECK-NEXT:    vmov r4, s18
-; CHECK-NEXT:    umull r5, r4, r4, r5
-; CHECK-NEXT:    vmov.32 q5[2], r5
-; CHECK-NEXT:    vmov.32 q5[3], r4
-; CHECK-NEXT:    vmov r2, s20
-; CHECK-NEXT:    vmov r5, s21
-; CHECK-NEXT:    adds.w r2, r2, r12
-; CHECK-NEXT:    adcs r3, r5
-; CHECK-NEXT:    vmov r5, s22
-; CHECK-NEXT:    adds.w r12, r2, r5
-; CHECK-NEXT:    vmov.u8 r5, q1[6]
-; CHECK-NEXT:    adcs r3, r4
-; CHECK-NEXT:    vmov.u8 r4, q0[6]
-; CHECK-NEXT:    vmov.32 q3[0], r5
-; CHECK-NEXT:    vmov.u8 r5, q1[7]
-; CHECK-NEXT:    vmov.32 q4[0], r4
-; CHECK-NEXT:    vmov.u8 r4, q0[7]
-; CHECK-NEXT:    vmov.32 q3[2], r5
-; CHECK-NEXT:    vmov.32 q4[2], r4
+; CHECK-NEXT:    vmov r2, s18
+; CHECK-NEXT:    vmov r3, s16
+; CHECK-NEXT:    muls r2, r4, r2
+; CHECK-NEXT:    vmov r4, s12
+; CHECK-NEXT:    muls r3, r4, r3
+; CHECK-NEXT:    vmov.32 q3[0], r3
+; CHECK-NEXT:    vmov.32 q3[2], r2
+; CHECK-NEXT:    vmov r3, s12
+; CHECK-NEXT:    adds.w r3, r3, r12
+; CHECK-NEXT:    adc r4, lr, #0
+; CHECK-NEXT:    adds.w r12, r3, r2
+; CHECK-NEXT:    adc lr, r4, #0
+; CHECK-NEXT:    vmov.u8 r4, q1[6]
+; CHECK-NEXT:    vmov.u8 r2, q0[6]
+; CHECK-NEXT:    vmov.32 q3[0], r4
+; CHECK-NEXT:    vmov.u8 r4, q1[7]
+; CHECK-NEXT:    vmov.32 q4[0], r2
+; CHECK-NEXT:    vmov.u8 r2, q0[7]
+; CHECK-NEXT:    vmov.32 q3[2], r4
+; CHECK-NEXT:    vmov.32 q4[2], r2
 ; CHECK-NEXT:    vand q3, q3, q2
 ; CHECK-NEXT:    vand q4, q4, q2
-; CHECK-NEXT:    vmov r5, s12
-; CHECK-NEXT:    vmov r4, s16
-; CHECK-NEXT:    umull r5, r4, r4, r5
-; CHECK-NEXT:    vmov.32 q5[0], r5
-; CHECK-NEXT:    vmov r5, s14
-; CHECK-NEXT:    vmov.32 q5[1], r4
-; CHECK-NEXT:    vmov r4, s18
-; CHECK-NEXT:    umull r5, r4, r4, r5
-; CHECK-NEXT:    vmov.32 q5[2], r5
-; CHECK-NEXT:    vmov.32 q5[3], r4
-; CHECK-NEXT:    vmov r2, s20
-; CHECK-NEXT:    vmov r5, s21
-; CHECK-NEXT:    adds.w r2, r2, r12
-; CHECK-NEXT:    adcs r3, r5
-; CHECK-NEXT:    vmov r5, s22
-; CHECK-NEXT:    adds.w r12, r2, r5
-; CHECK-NEXT:    vmov.u8 r5, q1[8]
-; CHECK-NEXT:    adcs r3, r4
-; CHECK-NEXT:    vmov.u8 r4, q0[8]
-; CHECK-NEXT:    vmov.32 q3[0], r5
-; CHECK-NEXT:    vmov.u8 r5, q1[9]
-; CHECK-NEXT:    vmov.32 q4[0], r4
-; CHECK-NEXT:    vmov.u8 r4, q0[9]
-; CHECK-NEXT:    vmov.32 q3[2], r5
-; CHECK-NEXT:    vmov.32 q4[2], r4
+; CHECK-NEXT:    vmov r4, s14
+; CHECK-NEXT:    vmov r2, s18
+; CHECK-NEXT:    vmov r3, s16
+; CHECK-NEXT:    muls r2, r4, r2
+; CHECK-NEXT:    vmov r4, s12
+; CHECK-NEXT:    muls r3, r4, r3
+; CHECK-NEXT:    vmov.32 q3[0], r3
+; CHECK-NEXT:    vmov.32 q3[2], r2
+; CHECK-NEXT:    vmov r3, s12
+; CHECK-NEXT:    adds.w r3, r3, r12
+; CHECK-NEXT:    adc r4, lr, #0
+; CHECK-NEXT:    adds.w r12, r3, r2
+; CHECK-NEXT:    adc lr, r4, #0
+; CHECK-NEXT:    vmov.u8 r4, q1[8]
+; CHECK-NEXT:    vmov.u8 r2, q0[8]
+; CHECK-NEXT:    vmov.32 q3[0], r4
+; CHECK-NEXT:    vmov.u8 r4, q1[9]
+; CHECK-NEXT:    vmov.32 q4[0], r2
+; CHECK-NEXT:    vmov.u8 r2, q0[9]
+; CHECK-NEXT:    vmov.32 q3[2], r4
+; CHECK-NEXT:    vmov.32 q4[2], r2
 ; CHECK-NEXT:    vand q3, q3, q2
 ; CHECK-NEXT:    vand q4, q4, q2
-; CHECK-NEXT:    vmov r5, s12
-; CHECK-NEXT:    vmov r4, s16
-; CHECK-NEXT:    umull r5, r4, r4, r5
-; CHECK-NEXT:    vmov.32 q5[0], r5
-; CHECK-NEXT:    vmov r5, s14
-; CHECK-NEXT:    vmov.32 q5[1], r4
-; CHECK-NEXT:    vmov r4, s18
-; CHECK-NEXT:    umull r5, r4, r4, r5
-; CHECK-NEXT:    vmov.32 q5[2], r5
-; CHECK-NEXT:    vmov.32 q5[3], r4
-; CHECK-NEXT:    vmov r2, s20
-; CHECK-NEXT:    vmov r5, s21
-; CHECK-NEXT:    adds.w r2, r2, r12
-; CHECK-NEXT:    adcs r3, r5
-; CHECK-NEXT:    vmov r5, s22
-; CHECK-NEXT:    adds.w r12, r2, r5
-; CHECK-NEXT:    vmov.u8 r5, q1[10]
-; CHECK-NEXT:    adcs r3, r4
-; CHECK-NEXT:    vmov.u8 r4, q0[10]
-; CHECK-NEXT:    vmov.32 q3[0], r5
-; CHECK-NEXT:    vmov.u8 r5, q1[11]
-; CHECK-NEXT:    vmov.32 q4[0], r4
-; CHECK-NEXT:    vmov.u8 r4, q0[11]
-; CHECK-NEXT:    vmov.32 q3[2], r5
-; CHECK-NEXT:    vmov.32 q4[2], r4
+; CHECK-NEXT:    vmov r4, s14
+; CHECK-NEXT:    vmov r2, s18
+; CHECK-NEXT:    vmov r3, s16
+; CHECK-NEXT:    muls r2, r4, r2
+; CHECK-NEXT:    vmov r4, s12
+; CHECK-NEXT:    muls r3, r4, r3
+; CHECK-NEXT:    vmov.32 q3[0], r3
+; CHECK-NEXT:    vmov.32 q3[2], r2
+; CHECK-NEXT:    vmov r3, s12
+; CHECK-NEXT:    adds.w r3, r3, r12
+; CHECK-NEXT:    adc r4, lr, #0
+; CHECK-NEXT:    adds.w r12, r3, r2
+; CHECK-NEXT:    adc lr, r4, #0
+; CHECK-NEXT:    vmov.u8 r4, q1[10]
+; CHECK-NEXT:    vmov.u8 r2, q0[10]
+; CHECK-NEXT:    vmov.32 q3[0], r4
+; CHECK-NEXT:    vmov.u8 r4, q1[11]
+; CHECK-NEXT:    vmov.32 q4[0], r2
+; CHECK-NEXT:    vmov.u8 r2, q0[11]
+; CHECK-NEXT:    vmov.32 q3[2], r4
+; CHECK-NEXT:    vmov.32 q4[2], r2
 ; CHECK-NEXT:    vand q3, q3, q2
 ; CHECK-NEXT:    vand q4, q4, q2
-; CHECK-NEXT:    vmov r5, s12
-; CHECK-NEXT:    vmov r4, s16
-; CHECK-NEXT:    umull r5, r4, r4, r5
-; CHECK-NEXT:    vmov.32 q5[0], r5
-; CHECK-NEXT:    vmov r5, s14
-; CHECK-NEXT:    vmov.32 q5[1], r4
-; CHECK-NEXT:    vmov r4, s18
-; CHECK-NEXT:    umull r5, r4, r4, r5
-; CHECK-NEXT:    vmov.32 q5[2], r5
-; CHECK-NEXT:    vmov.32 q5[3], r4
-; CHECK-NEXT:    vmov r2, s20
-; CHECK-NEXT:    vmov r5, s21
-; CHECK-NEXT:    adds.w r2, r2, r12
-; CHECK-NEXT:    adcs r3, r5
-; CHECK-NEXT:    vmov r5, s22
-; CHECK-NEXT:    adds.w r12, r2, r5
-; CHECK-NEXT:    vmov.u8 r5, q1[12]
-; CHECK-NEXT:    adcs r3, r4
-; CHECK-NEXT:    vmov.u8 r4, q0[12]
-; CHECK-NEXT:    vmov.32 q3[0], r5
-; CHECK-NEXT:    vmov.u8 r5, q1[13]
-; CHECK-NEXT:    vmov.32 q4[0], r4
-; CHECK-NEXT:    vmov.u8 r4, q0[13]
-; CHECK-NEXT:    vmov.32 q3[2], r5
-; CHECK-NEXT:    vmov.32 q4[2], r4
+; CHECK-NEXT:    vmov r4, s14
+; CHECK-NEXT:    vmov r2, s18
+; CHECK-NEXT:    vmov r3, s16
+; CHECK-NEXT:    muls r2, r4, r2
+; CHECK-NEXT:    vmov r4, s12
+; CHECK-NEXT:    muls r3, r4, r3
+; CHECK-NEXT:    vmov.32 q3[0], r3
+; CHECK-NEXT:    vmov.32 q3[2], r2
+; CHECK-NEXT:    vmov r3, s12
+; CHECK-NEXT:    adds.w r3, r3, r12
+; CHECK-NEXT:    adc r4, lr, #0
+; CHECK-NEXT:    adds.w r12, r3, r2
+; CHECK-NEXT:    adc lr, r4, #0
+; CHECK-NEXT:    vmov.u8 r4, q1[12]
+; CHECK-NEXT:    vmov.u8 r2, q0[12]
+; CHECK-NEXT:    vmov.32 q3[0], r4
+; CHECK-NEXT:    vmov.u8 r4, q1[13]
+; CHECK-NEXT:    vmov.32 q4[0], r2
+; CHECK-NEXT:    vmov.u8 r2, q0[13]
+; CHECK-NEXT:    vmov.32 q3[2], r4
+; CHECK-NEXT:    vmov.32 q4[2], r2
 ; CHECK-NEXT:    vand q3, q3, q2
 ; CHECK-NEXT:    vand q4, q4, q2
-; CHECK-NEXT:    vmov r5, s12
-; CHECK-NEXT:    vmov r4, s16
-; CHECK-NEXT:    umull r5, r4, r4, r5
-; CHECK-NEXT:    vmov.32 q5[0], r5
-; CHECK-NEXT:    vmov r5, s14
-; CHECK-NEXT:    vmov.32 q5[1], r4
-; CHECK-NEXT:    vmov r4, s18
-; CHECK-NEXT:    umull r5, r4, r4, r5
-; CHECK-NEXT:    vmov.32 q5[2], r5
-; CHECK-NEXT:    vmov.32 q5[3], r4
-; CHECK-NEXT:    vmov r2, s20
-; CHECK-NEXT:    vmov r5, s21
-; CHECK-NEXT:    adds.w r2, r2, r12
-; CHECK-NEXT:    adcs r3, r5
-; CHECK-NEXT:    vmov r5, s22
-; CHECK-NEXT:    adds r2, r2, r5
-; CHECK-NEXT:    vmov.u8 r5, q1[14]
-; CHECK-NEXT:    vmov.32 q3[0], r5
-; CHECK-NEXT:    vmov.u8 r5, q1[15]
-; CHECK-NEXT:    adcs r3, r4
-; CHECK-NEXT:    vmov.32 q3[2], r5
-; CHECK-NEXT:    vmov.u8 r4, q0[14]
-; CHECK-NEXT:    vand q1, q3, q2
+; CHECK-NEXT:    vmov r4, s14
+; CHECK-NEXT:    vmov r2, s18
+; CHECK-NEXT:    vmov r3, s16
+; CHECK-NEXT:    muls r2, r4, r2
+; CHECK-NEXT:    vmov r4, s12
+; CHECK-NEXT:    muls r3, r4, r3
+; CHECK-NEXT:    vmov.32 q3[0], r3
+; CHECK-NEXT:    vmov.32 q3[2], r2
+; CHECK-NEXT:    vmov r3, s12
+; CHECK-NEXT:    adds.w r3, r3, r12
+; CHECK-NEXT:    adc r4, lr, #0
+; CHECK-NEXT:    adds.w r12, r3, r2
+; CHECK-NEXT:    adc lr, r4, #0
+; CHECK-NEXT:    vmov.u8 r4, q1[14]
 ; CHECK-NEXT:    vmov.32 q3[0], r4
-; CHECK-NEXT:    vmov.u8 r4, q0[15]
+; CHECK-NEXT:    vmov.u8 r4, q1[15]
 ; CHECK-NEXT:    vmov.32 q3[2], r4
-; CHECK-NEXT:    vmov r5, s4
+; CHECK-NEXT:    vmov.u8 r2, q0[14]
+; CHECK-NEXT:    vand q1, q3, q2
+; CHECK-NEXT:    vmov.32 q3[0], r2
+; CHECK-NEXT:    vmov.u8 r2, q0[15]
+; CHECK-NEXT:    vmov r4, s6
+; CHECK-NEXT:    vmov.32 q3[2], r2
 ; CHECK-NEXT:    vand q0, q3, q2
-; CHECK-NEXT:    vmov r4, s0
-; CHECK-NEXT:    umlal r2, r3, r4, r5
-; CHECK-NEXT:    vmov r5, s6
-; CHECK-NEXT:    vmov r4, s2
-; CHECK-NEXT:    umlal r2, r3, r4, r5
+; CHECK-NEXT:    vmov r2, s2
+; CHECK-NEXT:    vmov r3, s0
+; CHECK-NEXT:    muls r2, r4, r2
+; CHECK-NEXT:    vmov r4, s4
+; CHECK-NEXT:    muls r3, r4, r3
+; CHECK-NEXT:    vmov.32 q0[0], r3
+; CHECK-NEXT:    vmov.32 q0[2], r2
+; CHECK-NEXT:    vmov r3, s0
+; CHECK-NEXT:    adds.w r3, r3, r12
+; CHECK-NEXT:    adc r4, lr, #0
+; CHECK-NEXT:    adds r2, r2, r3
+; CHECK-NEXT:    adc r3, r4, #0
 ; CHECK-NEXT:    adds r0, r0, r2
 ; CHECK-NEXT:    adcs r1, r3
-; CHECK-NEXT:    vpop {d8, d9, d10, d11}
-; CHECK-NEXT:    pop {r4, r5, r7, pc}
+; 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.experimental.vector.reduce.add.v16i64(<16 x i64> %m)
   %r = add i64 %z, %a
   ret i64 %r
 }
entry:
   %z = call i64 @llvm.experimental.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.experimental.vector.reduce.add.v2i64(<2 x i64> %m)
   %r = add i64 %z, %a
   ret i64 %r
 }

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

entry:
   %s = select <8 x i1> %c, <8 x i64> %m, <8 x i64> zeroinitializer
   %z = call i64 @llvm.experimental.vector.reduce.add.v8i64(<8 x i64> %s)
   ret i64 %z
 }
 
 define arm_aapcs_vfpcc i64 @add_v2i16_v2i64_zext(<2 x i16> %x, <2 x i16> %y, <2 x i16> %b) {
 ; CHECK-LABEL: add_v2i16_v2i64_zext:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .vsave {d8, d9}
-; CHECK-NEXT:    vpush {d8, d9}
 ; CHECK-NEXT:    vmov.i64 q3, #0xffff
+; CHECK-NEXT:    vand q0, q0, q3
 ; CHECK-NEXT:    vand q1, q1, q3
-; CHECK-NEXT:    vand q4, q0, q3
-; CHECK-NEXT:    vmov r0, s4
-; CHECK-NEXT:    vmov r1, s16
-; CHECK-NEXT:    umull r0, r1, r1, r0
-; CHECK-NEXT:    vmov.32 q0[0], r0
 ; CHECK-NEXT:    vmov r0, s6
-; CHECK-NEXT:    vmov.32 q0[1], r1
-; CHECK-NEXT:    vmov r1, s18
+; CHECK-NEXT:    vmov r1, s2
+; CHECK-NEXT:    vmov r2, s0
+; CHECK-NEXT:    muls r0, r1, r0
+; CHECK-NEXT:    vmov r1, s4
 ; CHECK-NEXT:    vand q1, q2, q3
-; CHECK-NEXT:    umull r0, r1, r1, r0
+; CHECK-NEXT:    muls r1, r2, r1
+; CHECK-NEXT:    movs r2, #0
+; CHECK-NEXT:    vmov.32 q0[0], r1
 ; CHECK-NEXT:    vmov.32 q0[2], r0
 ; CHECK-NEXT:    vmov r0, s4
-; CHECK-NEXT:    vmov.32 q0[3], r1
 ; CHECK-NEXT:    cmp r0, #0
 ; CHECK-NEXT:    cset r0, eq
 ; CHECK-NEXT:    tst.w r0, #1
 ; CHECK-NEXT:    csetm r0, ne
 ; CHECK-NEXT:    vmov.32 q2[0], r0
-; CHECK-NEXT:    vmov.32 q2[1], r0
 ; CHECK-NEXT:    vmov r0, s6
 ; CHECK-NEXT:    cmp r0, #0
 ; CHECK-NEXT:    cset r0, eq
 ; CHECK-NEXT:    tst.w r0, #1
 ; CHECK-NEXT:    csetm r0, ne
 ; CHECK-NEXT:    vmov.32 q2[2], r0
-; CHECK-NEXT:    vmov.32 q2[3], r0
 ; CHECK-NEXT:    vand q0, q0, q2
 ; 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:    vpop {d8, d9}
+; 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 i16> %b, zeroinitializer
   %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
   %s = select <2 x i1> %c, <2 x i64> %m, <2 x i64> zeroinitializer
   %z = call i64 @llvm.experimental.vector.reduce.add.v2i64(<2 x i64> %s)

 
 define arm_aapcs_vfpcc i64 @add_v16i8_v16i64_zext(<16 x i8> %x, <16 x i8> %y, <16 x i8> %b) {
 ; CHECK-LABEL: add_v16i8_v16i64_zext:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .save {r7, lr}
 ; CHECK-NEXT:    push {r7, 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 #88
-; CHECK-NEXT:    sub sp, #88
+; CHECK-NEXT:    .pad #56
+; CHECK-NEXT:    sub sp, #56
+; CHECK-NEXT:    vmov q4, q0
 ; CHECK-NEXT:    vmov q3, q1
-; CHECK-NEXT:    vstrw.32 q0, [sp, #64] @ 16-byte Spill
 ; CHECK-NEXT:    vmov.i8 q0, #0x0
 ; CHECK-NEXT:    vmov.i8 q1, #0xff
 ; CHECK-NEXT:    vcmp.i8 eq, q2, zr
-; CHECK-NEXT:    vstrw.32 q1, [sp, #16] @ 16-byte Spill
+; CHECK-NEXT:    vstrw.32 q1, [sp] @ 16-byte Spill
 ; CHECK-NEXT:    vpsel q5, q1, q0
-; CHECK-NEXT:    vstrw.32 q0, [sp, #32] @ 16-byte Spill
+; CHECK-NEXT:    vstrw.32 q0, [sp, #16] @ 16-byte Spill
 ; CHECK-NEXT:    vmov.u8 r0, q5[0]
-; CHECK-NEXT:    vmov.i64 q4, #0xff
+; CHECK-NEXT:    vmov.u8 r2, q4[0]
 ; CHECK-NEXT:    vmov.16 q2[0], r0
 ; CHECK-NEXT:    vmov.u8 r0, q5[1]
 ; CHECK-NEXT:    vmov.16 q2[1], r0
 ; CHECK-NEXT:    vmov.u8 r0, q5[2]
 ; CHECK-NEXT:    vmov.16 q2[2], r0
 ; CHECK-NEXT:    vmov.u8 r0, q5[3]
 ; CHECK-NEXT:    vmov.16 q2[3], r0
 ; CHECK-NEXT:    vmov.u8 r0, q5[4]

 ; CHECK-NEXT:    vmov.32 q2[0], r0
 ; CHECK-NEXT:    vmov.u16 r0, q6[1]
 ; CHECK-NEXT:    vmov.32 q2[1], r0
 ; CHECK-NEXT:    vmov.u16 r0, q6[2]
 ; CHECK-NEXT:    vmov.32 q2[2], r0
 ; CHECK-NEXT:    vmov.u16 r0, q6[3]
 ; CHECK-NEXT:    vmov.32 q2[3], r0
 ; CHECK-NEXT:    vcmp.i32 ne, q2, zr
+; CHECK-NEXT:    vmov.i64 q2, #0xff
 ; CHECK-NEXT:    vmrs r0, p0
-; CHECK-NEXT:    vldrw.u32 q1, [sp, #64] @ 16-byte Reload
-; CHECK-NEXT:    vmov.u8 r2, q1[0]
 ; CHECK-NEXT:    and r1, r0, #1
 ; CHECK-NEXT:    rsbs r1, r1, #0
 ; CHECK-NEXT:    vmov.32 q7[0], r1
-; CHECK-NEXT:    vmov.32 q7[1], r1
 ; CHECK-NEXT:    ubfx r1, r0, #4, #1
 ; CHECK-NEXT:    rsbs r1, r1, #0
 ; CHECK-NEXT:    vmov.32 q7[2], r1
-; CHECK-NEXT:    vmov.32 q7[3], r1
 ; CHECK-NEXT:    vmov.u8 r1, q3[0]
 ; CHECK-NEXT:    vmov.32 q0[0], r1
 ; CHECK-NEXT:    vmov.u8 r1, q3[1]
 ; CHECK-NEXT:    vmov.32 q0[2], r1
-; CHECK-NEXT:    vand q2, q0, q4
+; CHECK-NEXT:    vand q1, q0, q2
 ; CHECK-NEXT:    vmov.32 q0[0], r2
-; CHECK-NEXT:    vmov.u8 r2, q1[1]
-; CHECK-NEXT:    vmov r1, s8
+; CHECK-NEXT:    vmov.u8 r2, q4[1]
+; CHECK-NEXT:    vmov r1, s6
 ; CHECK-NEXT:    vmov.32 q0[2], r2
-; CHECK-NEXT:    vand q1, q0, q4
+; CHECK-NEXT:    vand q0, q0, q2
+; CHECK-NEXT:    vmov r2, s2
+; CHECK-NEXT:    vmov r3, s0
+; CHECK-NEXT:    muls r1, r2, r1
 ; CHECK-NEXT:    vmov r2, s4
-; CHECK-NEXT:    umull r1, r2, r2, r1
-; CHECK-NEXT:    vmov.32 q0[0], r1
-; CHECK-NEXT:    vmov r1, s10
-; CHECK-NEXT:    vmov.32 q0[1], r2
-; CHECK-NEXT:    vmov r2, s6
-; CHECK-NEXT:    vldrw.u32 q2, [sp, #64] @ 16-byte Reload
-; CHECK-NEXT:    umull r1, r2, r2, r1
+; CHECK-NEXT:    muls r2, r3, r2
+; CHECK-NEXT:    movs r3, #0
+; CHECK-NEXT:    vmov.32 q0[0], r2
 ; CHECK-NEXT:    vmov.32 q0[2], r1
-; CHECK-NEXT:    vmov.32 q0[3], r2
 ; CHECK-NEXT:    vand q0, q0, q7
 ; CHECK-NEXT:    vmov r1, s2
 ; CHECK-NEXT:    vmov r2, s0
-; CHECK-NEXT:    vmov r12, s3
-; CHECK-NEXT:    vmov r3, s1
-; CHECK-NEXT:    adds r1, r1, r2
-; CHECK-NEXT:    adc.w r2, r3, r12
+; CHECK-NEXT:    adds.w r12, r2, r1
+; CHECK-NEXT:    adc r2, r3, #0
 ; CHECK-NEXT:    ubfx r3, r0, #8, #1
-; CHECK-NEXT:    rsbs r3, r3, #0
 ; CHECK-NEXT:    ubfx r0, r0, #12, #1
+; CHECK-NEXT:    rsbs r3, r3, #0
 ; CHECK-NEXT:    vmov.32 q7[0], r3
 ; CHECK-NEXT:    rsbs r0, r0, #0
-; CHECK-NEXT:    vmov.32 q7[1], r3
-; CHECK-NEXT:    vmov.u8 r3, q2[2]
 ; CHECK-NEXT:    vmov.32 q7[2], r0
-; CHECK-NEXT:    vmov.32 q1[0], r3
-; CHECK-NEXT:    vmov.32 q7[3], r0
 ; CHECK-NEXT:    vmov.u8 r0, q3[2]
+; CHECK-NEXT:    vmov.u8 r3, q4[2]
 ; CHECK-NEXT:    vmov.32 q0[0], r0
 ; CHECK-NEXT:    vmov.u8 r0, q3[3]
-; CHECK-NEXT:    vmov.u8 r3, q2[3]
+; CHECK-NEXT:    vmov.32 q1[0], r3
+; CHECK-NEXT:    vmov.u8 r3, q4[3]
 ; CHECK-NEXT:    vmov.32 q0[2], r0
 ; CHECK-NEXT:    vmov.32 q1[2], r3
-; CHECK-NEXT:    vand q0, q0, q4
-; CHECK-NEXT:    vand q1, q1, q4
-; CHECK-NEXT:    vmov r0, s0
-; CHECK-NEXT:    vmov r3, s4
-; CHECK-NEXT:    umull r0, r3, r3, r0
-; CHECK-NEXT:    vmov.32 q2[0], r0
+; CHECK-NEXT:    vand q0, q0, q2
+; CHECK-NEXT:    vand q1, q1, q2
 ; CHECK-NEXT:    vmov r0, s2
-; CHECK-NEXT:    vmov.32 q2[1], r3
 ; CHECK-NEXT:    vmov r3, s6
-; CHECK-NEXT:    umull r0, r3, r3, r0
-; CHECK-NEXT:    vmov.32 q2[2], r0
-; CHECK-NEXT:    vmov.32 q2[3], r3
-; CHECK-NEXT:    vand q0, q2, q7
+; CHECK-NEXT:    vmov r1, s4
+; CHECK-NEXT:    muls r0, r3, r0
 ; CHECK-NEXT:    vmov r3, s0
-; CHECK-NEXT:    vmov r0, s1
-; CHECK-NEXT:    adds r1, r1, r3
-; CHECK-NEXT:    vmov r3, s3
-; CHECK-NEXT:    adcs r2, r0
-; CHECK-NEXT:    vmov r0, s2
-; CHECK-NEXT:    adds.w r12, r1, r0
-; CHECK-NEXT:    adc.w r1, r2, r3
+; CHECK-NEXT:    muls r1, r3, r1
+; CHECK-NEXT:    vmov.32 q0[0], r1
+; CHECK-NEXT:    vmov.32 q0[2], r0
+; CHECK-NEXT:    vand q0, q0, q7
+; CHECK-NEXT:    vmov q7, q4
+; CHECK-NEXT:    vmov r0, s0
+; CHECK-NEXT:    adds.w r0, r0, r12
+; CHECK-NEXT:    adc r1, r2, #0
+; CHECK-NEXT:    vmov r2, s2
+; CHECK-NEXT:    adds.w r12, r0, r2
 ; CHECK-NEXT:    vmov.u16 r2, q6[4]
 ; CHECK-NEXT:    vmov.32 q0[0], r2
 ; CHECK-NEXT:    vmov.u16 r2, q6[5]
 ; CHECK-NEXT:    vmov.32 q0[1], r2
 ; CHECK-NEXT:    vmov.u16 r2, q6[6]
 ; CHECK-NEXT:    vmov.32 q0[2], r2
 ; CHECK-NEXT:    vmov.u16 r2, q6[7]
 ; CHECK-NEXT:    vmov.32 q0[3], r2
+; CHECK-NEXT:    vmov.u8 r0, q4[4]
 ; CHECK-NEXT:    vcmp.i32 ne, q0, zr
-; CHECK-NEXT:    vmrs lr, p0
-; CHECK-NEXT:    vstrw.32 q3, [sp, #48] @ 16-byte Spill
-; CHECK-NEXT:    vstrw.32 q4, [sp] @ 16-byte Spill
-; CHECK-NEXT:    and r3, lr, #1
+; CHECK-NEXT:    vmov.32 q1[0], r0
+; CHECK-NEXT:    vmrs r2, p0
+; CHECK-NEXT:    vmov.u8 r0, q4[5]
+; CHECK-NEXT:    vmov.32 q1[2], r0
+; CHECK-NEXT:    adc lr, r1, #0
+; CHECK-NEXT:    vand q1, q1, q2
+; CHECK-NEXT:    vstrw.32 q3, [sp, #32] @ 16-byte Spill
+; CHECK-NEXT:    vmov r0, s6
+; CHECK-NEXT:    vmov r1, s4
+; CHECK-NEXT:    vldrw.u32 q1, [sp, #32] @ 16-byte Reload
+; CHECK-NEXT:    and r3, r2, #1
 ; CHECK-NEXT:    rsbs r3, r3, #0
 ; CHECK-NEXT:    vmov.32 q6[0], r3
-; CHECK-NEXT:    vmov.32 q6[1], r3
-; CHECK-NEXT:    ubfx r3, lr, #4, #1
+; CHECK-NEXT:    ubfx r3, r2, #4, #1
 ; CHECK-NEXT:    rsbs r3, r3, #0
 ; CHECK-NEXT:    vmov.32 q6[2], r3
-; CHECK-NEXT:    vmov.32 q6[3], r3
 ; CHECK-NEXT:    vmov.u8 r3, q3[4]
 ; CHECK-NEXT:    vmov.32 q0[0], r3
 ; CHECK-NEXT:    vmov.u8 r3, q3[5]
-; CHECK-NEXT:    vldrw.u32 q3, [sp, #64] @ 16-byte Reload
 ; CHECK-NEXT:    vmov.32 q0[2], r3
-; CHECK-NEXT:    vand q0, q0, q4
-; CHECK-NEXT:    vmov.u8 r0, q3[4]
+; CHECK-NEXT:    vand q0, q0, q2
+; CHECK-NEXT:    vmov r3, s2
+; CHECK-NEXT:    muls r0, r3, r0
 ; CHECK-NEXT:    vmov r3, s0
-; CHECK-NEXT:    vmov.32 q1[0], r0
-; CHECK-NEXT:    vmov.u8 r0, q3[5]
-; CHECK-NEXT:    vmov.32 q1[2], r0
-; CHECK-NEXT:    vmov q7, q3
-; CHECK-NEXT:    vand q1, q1, q4
-; CHECK-NEXT:    vmov r0, s4
-; CHECK-NEXT:    umull r0, r3, r0, r3
-; CHECK-NEXT:    vmov.32 q2[0], r0
-; CHECK-NEXT:    vmov r0, s2
-; CHECK-NEXT:    vmov.32 q2[1], r3
+; CHECK-NEXT:    muls r1, r3, r1
+; CHECK-NEXT:    vmov.32 q0[0], r1
+; CHECK-NEXT:    vmov.32 q0[2], r0
+; CHECK-NEXT:    vand q0, q0, q6
+; CHECK-NEXT:    vmov r0, s0
+; CHECK-NEXT:    vmov r3, s2
+; CHECK-NEXT:    adds.w r0, r0, r12
+; CHECK-NEXT:    adc r1, lr, #0
+; CHECK-NEXT:    adds.w r12, r0, r3
+; CHECK-NEXT:    ubfx r3, r2, #8, #1
+; CHECK-NEXT:    ubfx r2, r2, #12, #1
+; CHECK-NEXT:    rsb.w r3, r3, #0
+; CHECK-NEXT:    rsb.w r2, r2, #0
+; CHECK-NEXT:    vmov.32 q6[0], r3
+; CHECK-NEXT:    vmov.u8 r3, q4[6]
+; CHECK-NEXT:    vmov.32 q6[2], r2
+; CHECK-NEXT:    vmov.u8 r2, q1[6]
+; CHECK-NEXT:    vmov.32 q0[0], r2
+; CHECK-NEXT:    vmov.u8 r2, q1[7]
+; CHECK-NEXT:    vmov.32 q1[0], r3
+; CHECK-NEXT:    vmov.u8 r3, q4[7]
+; CHECK-NEXT:    vmov.32 q0[2], r2
+; CHECK-NEXT:    vmov.32 q1[2], r3
+; CHECK-NEXT:    vand q1, q1, q2
+; CHECK-NEXT:    vand q0, q0, q2
+; CHECK-NEXT:    vmov r2, s2
+; CHECK-NEXT:    adc r1, r1, #0
 ; CHECK-NEXT:    vmov r3, s6
-; CHECK-NEXT:    vldrw.u32 q1, [sp, #48] @ 16-byte Reload
-; CHECK-NEXT:    umull r0, r3, r3, r0
-; CHECK-NEXT:    vmov.32 q2[2], r0
-; CHECK-NEXT:    vmov.32 q2[3], r3
-; CHECK-NEXT:    vand q0, q2, q6
+; CHECK-NEXT:    vmov r0, s4
+; CHECK-NEXT:    vldrw.u32 q1, [sp] @ 16-byte Reload
+; CHECK-NEXT:    muls r2, r3, r2
 ; CHECK-NEXT:    vmov r3, s0
-; CHECK-NEXT:    vmov r0, s1
-; CHECK-NEXT:    vmov r2, s2
-; CHECK-NEXT:    adds.w r3, r3, r12
-; CHECK-NEXT:    adcs r1, r0
-; CHECK-NEXT:    vmov r0, s3
-; CHECK-NEXT:    adds r3, r3, r2
-; CHECK-NEXT:    vmov.u8 r2, q3[6]
-; CHECK-NEXT:    adcs r1, r0
-; CHECK-NEXT:    ubfx r0, lr, #8, #1
-; CHECK-NEXT:    rsbs r0, r0, #0
-; CHECK-NEXT:    vmov.32 q6[0], r0
-; CHECK-NEXT:    vmov.32 q6[1], r0
-; CHECK-NEXT:    ubfx r0, lr, #12, #1
-; CHECK-NEXT:    rsbs r0, r0, #0
-; CHECK-NEXT:    vmov.32 q6[2], r0
-; CHECK-NEXT:    vmov.32 q6[3], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[6]
+; CHECK-NEXT:    muls r0, r3, r0
 ; CHECK-NEXT:    vmov.32 q0[0], r0
-; CHECK-NEXT:    vmov.u8 r0, q1[7]
-; CHECK-NEXT:    vmov.32 q1[0], r2
-; CHECK-NEXT:    vmov.u8 r2, q3[7]
-; CHECK-NEXT:    vmov.32 q0[2], r0
-; CHECK-NEXT:    vmov.32 q1[2], r2
-; CHECK-NEXT:    vand q0, q0, q4
-; CHECK-NEXT:    vand q1, q1, q4
+; CHECK-NEXT:    vmov.32 q0[2], r2
+; CHECK-NEXT:    vand q0, q0, q6
 ; CHECK-NEXT:    vmov r0, s0
-; CHECK-NEXT:    vmov r2, s4
-; CHECK-NEXT:    umull r0, r2, r2, r0
-; CHECK-NEXT:    vmov.32 q2[0], r0
-; CHECK-NEXT:    vmov r0, s2
-; CHECK-NEXT:    vmov.32 q2[1], r2
-; CHECK-NEXT:    vmov r2, s6
-; CHECK-NEXT:    vldrw.u32 q1, [sp, #16] @ 16-byte Reload
-; CHECK-NEXT:    umull r0, r2, r2, r0
-; CHECK-NEXT:    vmov.32 q2[2], r0
-; CHECK-NEXT:    vmov.32 q2[3], r2
-; CHECK-NEXT:    vand q0, q2, q6
-; CHECK-NEXT:    vmov r2, s0
-; CHECK-NEXT:    vmov r0, s1
-; CHECK-NEXT:    adds r2, r2, r3
-; CHECK-NEXT:    vmov r3, s3
-; CHECK-NEXT:    adcs r1, r0
-; CHECK-NEXT:    vmov r0, s2
-; CHECK-NEXT:    vldrw.u32 q0, [sp, #32] @ 16-byte Reload
-; CHECK-NEXT:    adds.w r12, r2, r0
+; CHECK-NEXT:    vmov r2, s2
+; CHECK-NEXT:    vldrw.u32 q0, [sp, #16] @ 16-byte Reload
+; CHECK-NEXT:    adds.w r0, r0, r12
+; CHECK-NEXT:    adc r1, r1, #0
+; CHECK-NEXT:    adds.w r12, r0, r2
 ; CHECK-NEXT:    vmov.u8 r2, q5[8]
+; CHECK-NEXT:    vmov.u8 r0, q7[8]
 ; CHECK-NEXT:    vmov.16 q6[0], r2
 ; CHECK-NEXT:    vmov.u8 r2, q5[9]
 ; CHECK-NEXT:    vmov.16 q6[1], r2
 ; CHECK-NEXT:    vmov.u8 r2, q5[10]
 ; CHECK-NEXT:    vmov.16 q6[2], r2
 ; CHECK-NEXT:    vmov.u8 r2, q5[11]
 ; CHECK-NEXT:    vmov.16 q6[3], r2
 ; CHECK-NEXT:    vmov.u8 r2, q5[12]
 ; CHECK-NEXT:    vmov.16 q6[4], r2
 ; CHECK-NEXT:    vmov.u8 r2, q5[13]
 ; CHECK-NEXT:    vmov.16 q6[5], r2
 ; CHECK-NEXT:    vmov.u8 r2, q5[14]
 ; CHECK-NEXT:    vmov.16 q6[6], r2
 ; CHECK-NEXT:    vmov.u8 r2, q5[15]
 ; CHECK-NEXT:    vmov.16 q6[7], r2
-; CHECK-NEXT:    adcs r1, r3
+; CHECK-NEXT:    adc lr, r1, #0
 ; CHECK-NEXT:    vcmp.i16 ne, q6, zr
-; CHECK-NEXT:    vmov.u8 r0, q7[8]
 ; CHECK-NEXT:    vpsel q3, q1, q0
 ; CHECK-NEXT:    vmov.32 q1[0], r0
 ; CHECK-NEXT:    vmov.u16 r2, q3[0]
 ; CHECK-NEXT:    vmov.u8 r0, q7[9]
 ; CHECK-NEXT:    vmov.32 q0[0], r2
 ; CHECK-NEXT:    vmov.u16 r2, q3[1]
 ; CHECK-NEXT:    vmov.32 q0[1], r2
 ; CHECK-NEXT:    vmov.u16 r2, q3[2]
 ; CHECK-NEXT:    vmov.32 q0[2], r2
 ; CHECK-NEXT:    vmov.u16 r2, q3[3]
 ; CHECK-NEXT:    vmov.32 q0[3], r2
 ; CHECK-NEXT:    vmov.32 q1[2], r0
 ; CHECK-NEXT:    vcmp.i32 ne, q0, zr
-; CHECK-NEXT:    vmrs lr, p0
-; CHECK-NEXT:    vldrw.u32 q6, [sp, #48] @ 16-byte Reload
-; CHECK-NEXT:    vldrw.u32 q5, [sp] @ 16-byte Reload
-; CHECK-NEXT:    vand q1, q1, q5
-; CHECK-NEXT:    vmov r0, s4
-; CHECK-NEXT:    and r3, lr, #1
+; CHECK-NEXT:    vand q1, q1, q2
+; CHECK-NEXT:    vmrs r2, p0
+; CHECK-NEXT:    vldrw.u32 q5, [sp, #32] @ 16-byte Reload
+; CHECK-NEXT:    vmov r0, s6
+; CHECK-NEXT:    vmov r1, s4
+; CHECK-NEXT:    and r3, r2, #1
 ; CHECK-NEXT:    rsbs r3, r3, #0
 ; CHECK-NEXT:    vmov.32 q4[0], r3
-; CHECK-NEXT:    vmov.32 q4[1], r3
-; CHECK-NEXT:    ubfx r3, lr, #4, #1
+; CHECK-NEXT:    ubfx r3, r2, #4, #1
 ; CHECK-NEXT:    rsbs r3, r3, #0
 ; CHECK-NEXT:    vmov.32 q4[2], r3
-; CHECK-NEXT:    vmov.32 q4[3], r3
-; CHECK-NEXT:    vmov.u8 r3, q6[8]
+; CHECK-NEXT:    vmov.u8 r3, q5[8]
 ; CHECK-NEXT:    vmov.32 q0[0], r3
-; CHECK-NEXT:    vmov.u8 r3, q6[9]
+; CHECK-NEXT:    vmov.u8 r3, q5[9]
 ; CHECK-NEXT:    vmov.32 q0[2], r3
-; CHECK-NEXT:    vand q0, q0, q5
+; CHECK-NEXT:    vand q0, q0, q2
+; CHECK-NEXT:    vmov r3, s2
+; CHECK-NEXT:    muls r0, r3, r0
 ; CHECK-NEXT:    vmov r3, s0
-; CHECK-NEXT:    umull r0, r3, r0, r3
-; CHECK-NEXT:    vmov.32 q2[0], r0
-; CHECK-NEXT:    vmov r0, s2
-; CHECK-NEXT:    vmov.32 q2[1], r3
+; CHECK-NEXT:    muls r1, r3, r1
+; CHECK-NEXT:    vmov.32 q0[0], r1
+; CHECK-NEXT:    vmov.32 q0[2], r0
+; CHECK-NEXT:    vand q0, q0, q4
+; CHECK-NEXT:    vmov r0, s0
+; CHECK-NEXT:    vmov r3, s2
+; CHECK-NEXT:    adds.w r0, r0, r12
+; CHECK-NEXT:    adc r1, lr, #0
+; CHECK-NEXT:    adds.w r12, r0, r3
+; CHECK-NEXT:    ubfx r3, r2, #8, #1
+; CHECK-NEXT:    ubfx r2, r2, #12, #1
+; CHECK-NEXT:    rsb.w r3, r3, #0
+; CHECK-NEXT:    rsb.w r2, r2, #0
+; CHECK-NEXT:    vmov.32 q4[0], r3
+; CHECK-NEXT:    vmov.u8 r3, q7[10]
+; CHECK-NEXT:    vmov.32 q4[2], r2
+; CHECK-NEXT:    vmov.u8 r2, q5[10]
+; CHECK-NEXT:    vmov.32 q0[0], r2
+; CHECK-NEXT:    vmov.u8 r2, q5[11]
+; CHECK-NEXT:    vmov.32 q1[0], r3
+; CHECK-NEXT:    vmov.u8 r3, q7[11]
+; CHECK-NEXT:    vmov.32 q0[2], r2
+; CHECK-NEXT:    vmov.32 q1[2], r3
+; CHECK-NEXT:    vand q1, q1, q2
+; CHECK-NEXT:    vand q0, q0, q2
+; CHECK-NEXT:    vmov r2, s2
+; CHECK-NEXT:    adc r1, r1, #0
 ; CHECK-NEXT:    vmov r3, s6
-; CHECK-NEXT:    umull r0, r3, r3, r0
-; CHECK-NEXT:    vmov.32 q2[2], r0
-; CHECK-NEXT:    vmov.32 q2[3], r3
-; CHECK-NEXT:    vand q0, q2, q4
+; CHECK-NEXT:    vmov r0, s4
+; CHECK-NEXT:    muls r2, r3, r2
 ; CHECK-NEXT:    vmov r3, s0
-; CHECK-NEXT:    vmov r0, s1
-; CHECK-NEXT:    vmov r2, s2
-; CHECK-NEXT:    adds.w r3, r3, r12
-; CHECK-NEXT:    adcs r1, r0
-; CHECK-NEXT:    vmov r0, s3
-; CHECK-NEXT:    adds r3, r3, r2
-; CHECK-NEXT:    vmov.u8 r2, q7[10]
-; CHECK-NEXT:    vmov.32 q1[0], r2
-; CHECK-NEXT:    vmov.u8 r2, q7[11]
-; CHECK-NEXT:    vmov.32 q1[2], r2
-; CHECK-NEXT:    vand q1, q1, q5
-; CHECK-NEXT:    vmov r2, s4
-; CHECK-NEXT:    adcs r1, r0
-; CHECK-NEXT:    ubfx r0, lr, #8, #1
-; CHECK-NEXT:    rsbs r0, r0, #0
-; CHECK-NEXT:    vmov.32 q4[0], r0
-; CHECK-NEXT:    vmov.32 q4[1], r0
-; CHECK-NEXT:    ubfx r0, lr, #12, #1
-; CHECK-NEXT:    rsbs r0, r0, #0
-; CHECK-NEXT:    vmov.32 q4[2], r0
-; CHECK-NEXT:    vmov.32 q4[3], r0
-; CHECK-NEXT:    vmov.u8 r0, q6[10]
+; CHECK-NEXT:    muls r0, r3, r0
 ; CHECK-NEXT:    vmov.32 q0[0], r0
-; CHECK-NEXT:    vmov.u8 r0, q6[11]
-; CHECK-NEXT:    vmov.32 q0[2], r0
-; CHECK-NEXT:    vand q0, q0, q5
+; CHECK-NEXT:    vmov.32 q0[2], r2
+; CHECK-NEXT:    vand q0, q0, q4
 ; CHECK-NEXT:    vmov r0, s0
-; CHECK-NEXT:    umull r0, r2, r2, r0
-; CHECK-NEXT:    vmov.32 q2[0], r0
-; CHECK-NEXT:    vmov r0, s2
-; CHECK-NEXT:    vmov.32 q2[1], r2
-; CHECK-NEXT:    vmov r2, s6
-; CHECK-NEXT:    umull r0, r2, r2, r0
-; CHECK-NEXT:    vmov.32 q2[2], r0
-; CHECK-NEXT:    vmov.32 q2[3], r2
-; CHECK-NEXT:    vand q0, q2, q4
-; CHECK-NEXT:    vmov r2, s0
-; CHECK-NEXT:    vmov r0, s1
-; CHECK-NEXT:    adds r2, r2, r3
-; CHECK-NEXT:    vmov r3, s3
-; CHECK-NEXT:    adcs r1, r0
-; CHECK-NEXT:    vmov r0, s2
-; CHECK-NEXT:    adds.w r12, r2, r0
+; CHECK-NEXT:    vmov r2, s2
+; CHECK-NEXT:    adds.w r0, r0, r12
+; CHECK-NEXT:    adc r1, r1, #0
+; CHECK-NEXT:    adds.w r12, r0, r2
 ; CHECK-NEXT:    vmov.u16 r2, q3[4]
+; CHECK-NEXT:    vmov.u8 r0, q7[12]
 ; CHECK-NEXT:    vmov.32 q0[0], r2
 ; CHECK-NEXT:    vmov.u16 r2, q3[5]
 ; CHECK-NEXT:    vmov.32 q0[1], r2
 ; CHECK-NEXT:    vmov.u16 r2, q3[6]
 ; CHECK-NEXT:    vmov.32 q0[2], r2
 ; CHECK-NEXT:    vmov.u16 r2, q3[7]
 ; CHECK-NEXT:    vmov.32 q0[3], r2
-; CHECK-NEXT:    adcs r1, r3
-; CHECK-NEXT:    vcmp.i32 ne, q0, zr
-; CHECK-NEXT:    vmov.u8 r0, q7[12]
-; CHECK-NEXT:    vmrs lr, p0
 ; CHECK-NEXT:    vmov.32 q1[0], r0
+; CHECK-NEXT:    vcmp.i32 ne, q0, zr
 ; CHECK-NEXT:    vmov.u8 r0, q7[13]
+; CHECK-NEXT:    vmrs r2, p0
 ; CHECK-NEXT:    vmov.32 q1[2], r0
-; CHECK-NEXT:    vand q1, q1, q5
-; CHECK-NEXT:    vmov r0, s4
-; CHECK-NEXT:    and r3, lr, #1
+; CHECK-NEXT:    vand q1, q1, q2
+; CHECK-NEXT:    adc lr, r1, #0
+; CHECK-NEXT:    vmov r0, s6
+; CHECK-NEXT:    vmov r1, s4
+; CHECK-NEXT:    and r3, r2, #1
 ; CHECK-NEXT:    rsbs r3, r3, #0
 ; CHECK-NEXT:    vmov.32 q3[0], r3
-; CHECK-NEXT:    vmov.32 q3[1], r3
-; CHECK-NEXT:    ubfx r3, lr, #4, #1
+; CHECK-NEXT:    ubfx r3, r2, #4, #1
 ; CHECK-NEXT:    rsbs r3, r3, #0
 ; CHECK-NEXT:    vmov.32 q3[2], r3
-; CHECK-NEXT:    vmov.32 q3[3], r3
-; CHECK-NEXT:    vmov.u8 r3, q6[12]
+; CHECK-NEXT:    vmov.u8 r3, q5[12]
 ; CHECK-NEXT:    vmov.32 q0[0], r3
-; CHECK-NEXT:    vmov.u8 r3, q6[13]
+; CHECK-NEXT:    vmov.u8 r3, q5[13]
 ; CHECK-NEXT:    vmov.32 q0[2], r3
-; CHECK-NEXT:    vand q0, q0, q5
+; CHECK-NEXT:    vand q0, q0, q2
+; CHECK-NEXT:    vmov r3, s2
+; CHECK-NEXT:    muls r0, r3, r0
 ; CHECK-NEXT:    vmov r3, s0
-; CHECK-NEXT:    umull r0, r3, r0, r3
-; CHECK-NEXT:    vmov.32 q2[0], r0
-; CHECK-NEXT:    vmov r0, s2
-; CHECK-NEXT:    vmov.32 q2[1], r3
+; CHECK-NEXT:    muls r1, r3, r1
+; CHECK-NEXT:    vmov.32 q0[0], r1
+; CHECK-NEXT:    vmov.32 q0[2], r0
+; CHECK-NEXT:    vand q0, q0, q3
+; CHECK-NEXT:    vmov r0, s0
+; CHECK-NEXT:    vmov r3, s2
+; CHECK-NEXT:    adds.w r0, r0, r12
+; CHECK-NEXT:    adc r1, lr, #0
+; CHECK-NEXT:    adds.w r12, r0, r3
+; CHECK-NEXT:    ubfx r3, r2, #8, #1
+; CHECK-NEXT:    ubfx r2, r2, #12, #1
+; CHECK-NEXT:    rsb.w r3, r3, #0
+; CHECK-NEXT:    rsb.w r2, r2, #0
+; CHECK-NEXT:    vmov.32 q3[0], r3
+; CHECK-NEXT:    vmov.u8 r3, q7[14]
+; CHECK-NEXT:    vmov.32 q3[2], r2
+; CHECK-NEXT:    vmov.u8 r2, q5[14]
+; CHECK-NEXT:    vmov.32 q0[0], r2
+; CHECK-NEXT:    vmov.u8 r2, q5[15]
+; CHECK-NEXT:    vmov.32 q1[0], r3
+; CHECK-NEXT:    vmov.u8 r3, q7[15]
+; CHECK-NEXT:    vmov.32 q0[2], r2
+; CHECK-NEXT:    vmov.32 q1[2], r3
+; CHECK-NEXT:    vand q1, q1, q2
+; CHECK-NEXT:    vand q0, q0, q2
+; CHECK-NEXT:    vmov r2, s2
+; CHECK-NEXT:    adc r1, r1, #0
 ; CHECK-NEXT:    vmov r3, s6
-; CHECK-NEXT:    umull r0, r3, r3, r0
-; CHECK-NEXT:    vmov.32 q2[2], r0
-; CHECK-NEXT:    vmov.32 q2[3], r3
-; CHECK-NEXT:    vand q0, q2, q3
+; CHECK-NEXT:    vmov r0, s4
+; CHECK-NEXT:    muls r2, r3, r2
 ; CHECK-NEXT:    vmov r3, s0
-; CHECK-NEXT:    vmov r0, s1
-; CHECK-NEXT:    vmov r2, s2
-; CHECK-NEXT:    adds.w r3, r3, r12
-; CHECK-NEXT:    adcs r1, r0
-; CHECK-NEXT:    vmov r0, s3
-; CHECK-NEXT:    adds r3, r3, r2
-; CHECK-NEXT:    vmov.u8 r2, q7[14]
-; CHECK-NEXT:    vmov.32 q1[0], r2
-; CHECK-NEXT:    vmov.u8 r2, q7[15]
-; CHECK-NEXT:    vmov.32 q1[2], r2
-; CHECK-NEXT:    vand q1, q1, q5
-; CHECK-NEXT:    vmov r2, s4
-; CHECK-NEXT:    adcs r1, r0
-; CHECK-NEXT:    ubfx r0, lr, #8, #1
-; CHECK-NEXT:    rsbs r0, r0, #0
-; CHECK-NEXT:    vmov.32 q3[0], r0
-; CHECK-NEXT:    vmov.32 q3[1], r0
-; CHECK-NEXT:    ubfx r0, lr, #12, #1
-; CHECK-NEXT:    rsbs r0, r0, #0
-; CHECK-NEXT:    vmov.32 q3[2], r0
-; CHECK-NEXT:    vmov.32 q3[3], r0
-; CHECK-NEXT:    vmov.u8 r0, q6[14]
+; CHECK-NEXT:    muls r0, r3, r0
 ; CHECK-NEXT:    vmov.32 q0[0], r0
-; CHECK-NEXT:    vmov.u8 r0, q6[15]
-; CHECK-NEXT:    vmov.32 q0[2], r0
-; CHECK-NEXT:    vand q0, q0, q5
+; CHECK-NEXT:    vmov.32 q0[2], r2
+; CHECK-NEXT:    vand q0, q0, q3
 ; CHECK-NEXT:    vmov r0, s0
-; CHECK-NEXT:    umull r0, r2, r2, r0
-; CHECK-NEXT:    vmov.32 q2[0], r0
-; CHECK-NEXT:    vmov r0, s2
-; CHECK-NEXT:    vmov.32 q2[1], r2
-; CHECK-NEXT:    vmov r2, s6
-; CHECK-NEXT:    umull r0, r2, r2, r0
-; CHECK-NEXT:    vmov.32 q2[2], r0
-; CHECK-NEXT:    vmov.32 q2[3], r2
-; CHECK-NEXT:    vand q0, q2, q3
-; CHECK-NEXT:    vmov r2, s0
-; CHECK-NEXT:    vmov r0, s1
-; CHECK-NEXT:    adds r2, r2, r3
-; CHECK-NEXT:    vmov r3, s3
-; CHECK-NEXT:    adcs r1, r0
-; CHECK-NEXT:    vmov r0, s2
+; CHECK-NEXT:    vmov r2, s2
+; CHECK-NEXT:    adds.w r0, r0, r12
+; CHECK-NEXT:    adc r1, r1, #0
 ; CHECK-NEXT:    adds r0, r0, r2
-; CHECK-NEXT:    adcs r1, r3
-; CHECK-NEXT:    add sp, #88
+; CHECK-NEXT:    adc r1, r1, #0
+; CHECK-NEXT:    add sp, #56
 ; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13, d14, d15}
 ; CHECK-NEXT:    pop {r7, pc}
 entry:
   %c = icmp eq <16 x i8> %b, zeroinitializer
   %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
   %s = select <16 x i1> %c, <16 x i64> %m, <16 x i64> zeroinitializer
entry:
   %s = select <16 x i1> %c, <16 x i64> %m, <16 x i64> zeroinitializer
   %z = call i64 @llvm.experimental.vector.reduce.add.v16i64(<16 x i64> %s)
   ret i64 %z
 }
 
 define arm_aapcs_vfpcc i64 @add_v2i8_v2i64_zext(<2 x i8> %x, <2 x i8> %y, <2 x i8> %b) {
 ; CHECK-LABEL: add_v2i8_v2i64_zext:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    .vsave {d8, d9}
-; CHECK-NEXT:    vpush {d8, d9}
 ; CHECK-NEXT:    vmov.i64 q3, #0xff
+; CHECK-NEXT:    vand q0, q0, q3
 ; CHECK-NEXT:    vand q1, q1, q3
-; CHECK-NEXT:    vand q4, q0, q3
-; CHECK-NEXT:    vmov r0, s4
-; CHECK-NEXT:    vmov r1, s16
-; CHECK-NEXT:    umull r0, r1, r1, r0
-; CHECK-NEXT:    vmov.32 q0[0], r0
 ; CHECK-NEXT:    vmov r0, s6
-; CHECK-NEXT:    vmov.32 q0[1], r1
-; CHECK-NEXT:    vmov r1, s18
+; CHECK-NEXT:    vmov r1, s2
+; CHECK-NEXT:    vmov r2, s0
+; CHECK-NEXT:    muls r0, r1, r0
+; CHECK-NEXT:    vmov r1, s4
 ; CHECK-NEXT:    vand q1, q2, q3
-; CHECK-NEXT:    umull r0, r1, r1, r0
+; CHECK-NEXT:    muls r1, r2, r1
+; CHECK-NEXT:    movs r2, #0
+; CHECK-NEXT:    vmov.32 q0[0], r1
 ; CHECK-NEXT:    vmov.32 q0[2], r0
 ; CHECK-NEXT:    vmov r0, s4
-; CHECK-NEXT:    vmov.32 q0[3], r1
 ; CHECK-NEXT:    cmp r0, #0
 ; CHECK-NEXT:    cset r0, eq
 ; CHECK-NEXT:    tst.w r0, #1
 ; CHECK-NEXT:    csetm r0, ne
 ; CHECK-NEXT:    vmov.32 q2[0], r0
-; CHECK-NEXT:    vmov.32 q2[1], r0
 ; CHECK-NEXT:    vmov r0, s6
 ; CHECK-NEXT:    cmp r0, #0
 ; CHECK-NEXT:    cset r0, eq
 ; CHECK-NEXT:    tst.w r0, #1
 ; CHECK-NEXT:    csetm r0, ne
 ; CHECK-NEXT:    vmov.32 q2[2], r0
-; CHECK-NEXT:    vmov.32 q2[3], r0
 ; CHECK-NEXT:    vand q0, q0, q2
 ; 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:    vpop {d8, d9}
+; 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 i8> %b, zeroinitializer
   %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
   %s = select <2 x i1> %c, <2 x i64> %m, <2 x i64> zeroinitializer
   %z = call i64 @llvm.experimental.vector.reduce.add.v2i64(<2 x i64> %s)
entry:
   %z = call i64 @llvm.experimental.vector.reduce.add.v8i64(<8 x i64> %s)
   %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, <2 x i16> %b, 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:    .vsave {d8, d9}
-; CHECK-NEXT:    vpush {d8, d9}
 ; CHECK-NEXT:    vmov.i64 q3, #0xffff
 ; CHECK-NEXT:    vand q1, q1, q3
-; CHECK-NEXT:    vand q4, q0, q3
-; CHECK-NEXT:    vmov r2, s4
-; CHECK-NEXT:    vmov r3, s16
-; CHECK-NEXT:    umull r2, r3, r3, r2
-; CHECK-NEXT:    vmov.32 q0[0], r2
+; CHECK-NEXT:    vand q0, q0, q3
 ; CHECK-NEXT:    vmov r2, s6
-; CHECK-NEXT:    vmov.32 q0[1], r3
-; CHECK-NEXT:    vmov r3, s18
+; CHECK-NEXT:    vmov r3, s2
+; CHECK-NEXT:    mul r12, r3, r2
+; CHECK-NEXT:    vmov r3, s4
+; CHECK-NEXT:    vmov r2, s0
 ; CHECK-NEXT:    vand q1, q2, q3
-; CHECK-NEXT:    umull r2, r3, r3, r2
-; CHECK-NEXT:    vmov.32 q0[2], r2
+; CHECK-NEXT:    muls r2, r3, r2
+; CHECK-NEXT:    vmov.32 q0[0], r2
 ; CHECK-NEXT:    vmov r2, s4
-; CHECK-NEXT:    vmov.32 q0[3], r3
+; CHECK-NEXT:    vmov.32 q0[2], r12
+; CHECK-NEXT:    mov.w r12, #0
 ; CHECK-NEXT:    cmp r2, #0
 ; CHECK-NEXT:    cset r2, eq
 ; CHECK-NEXT:    tst.w r2, #1
 ; CHECK-NEXT:    csetm r2, ne
 ; CHECK-NEXT:    vmov.32 q2[0], r2
-; CHECK-NEXT:    vmov.32 q2[1], r2
 ; CHECK-NEXT:    vmov r2, s6
 ; CHECK-NEXT:    cmp r2, #0
 ; CHECK-NEXT:    cset r2, eq
 ; CHECK-NEXT:    tst.w r2, #1
 ; CHECK-NEXT:    csetm r2, ne
 ; CHECK-NEXT:    vmov.32 q2[2], r2
-; CHECK-NEXT:    vmov.32 q2[3], r2
 ; CHECK-NEXT:    vand q0, q0, q2
 ; 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:    vpop {d8, d9}
-; CHECK-NEXT:    pop {r7, pc}
+; CHECK-NEXT:    bx lr
 entry:
   %c = icmp eq <2 x i16> %b, zeroinitializer
   %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
   %s = select <2 x i1> %c, <2 x i64> %m, <2 x i64> zeroinitializer
   %z = call i64 @llvm.experimental.vector.reduce.add.v2i64(<2 x i64> %s)
   %r = add i64 %z, %a
entry:
   %z = call i8 @llvm.experimental.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> %y, <16 x i8> %b, 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:    .save {r4, r5, r7, lr}
+; CHECK-NEXT:    push {r4, r5, r7, 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 #80
-; CHECK-NEXT:    sub sp, #80
+; CHECK-NEXT:    .pad #48
+; CHECK-NEXT:    sub sp, #48
 ; CHECK-NEXT:    vmov q3, q1
-; CHECK-NEXT:    vstrw.32 q0, [sp, #64] @ 16-byte Spill
+; CHECK-NEXT:    vmov q4, q0
 ; CHECK-NEXT:    vmov.i8 q0, #0x0
 ; CHECK-NEXT:    vmov.i8 q1, #0xff
 ; CHECK-NEXT:    vcmp.i8 eq, q2, zr
-; CHECK-NEXT:    vstrw.32 q1, [sp, #16] @ 16-byte Spill
+; CHECK-NEXT:    vstrw.32 q1, [sp] @ 16-byte Spill
 ; CHECK-NEXT:    vpsel q5, q1, q0
-; CHECK-NEXT:    vstrw.32 q0, [sp, #32] @ 16-byte Spill
+; CHECK-NEXT:    vstrw.32 q0, [sp, #16] @ 16-byte Spill
 ; CHECK-NEXT:    vmov.u8 r2, q5[0]
-; CHECK-NEXT:    vmov.i64 q4, #0xff
 ; CHECK-NEXT:    vmov.16 q2[0], r2
 ; CHECK-NEXT:    vmov.u8 r2, q5[1]
 ; CHECK-NEXT:    vmov.16 q2[1], r2
 ; CHECK-NEXT:    vmov.u8 r2, q5[2]
 ; CHECK-NEXT:    vmov.16 q2[2], r2
 ; CHECK-NEXT:    vmov.u8 r2, q5[3]
 ; CHECK-NEXT:    vmov.16 q2[3], r2
 ; CHECK-NEXT:    vmov.u8 r2, q5[4]

 ; CHECK-NEXT:    vmov.32 q2[0], r2
 ; CHECK-NEXT:    vmov.u16 r2, q6[1]
 ; CHECK-NEXT:    vmov.32 q2[1], r2
 ; CHECK-NEXT:    vmov.u16 r2, q6[2]
 ; CHECK-NEXT:    vmov.32 q2[2], r2
 ; CHECK-NEXT:    vmov.u16 r2, q6[3]
 ; CHECK-NEXT:    vmov.32 q2[3], r2
 ; CHECK-NEXT:    vcmp.i32 ne, q2, zr
-; CHECK-NEXT:    vmrs lr, p0
-; CHECK-NEXT:    vldrw.u32 q1, [sp, #64] @ 16-byte Reload
-; CHECK-NEXT:    and r3, lr, #1
+; CHECK-NEXT:    vmov.i64 q2, #0xff
+; CHECK-NEXT:    vmrs r2, p0
+; CHECK-NEXT:    and r3, r2, #1
 ; CHECK-NEXT:    rsbs r3, r3, #0
 ; CHECK-NEXT:    vmov.32 q7[0], r3
-; CHECK-NEXT:    vmov.32 q7[1], r3
-; CHECK-NEXT:    ubfx r3, lr, #4, #1
+; CHECK-NEXT:    ubfx r3, r2, #4, #1
 ; CHECK-NEXT:    rsbs r3, r3, #0
 ; CHECK-NEXT:    vmov.32 q7[2], r3
-; CHECK-NEXT:    vmov.32 q7[3], r3
 ; CHECK-NEXT:    vmov.u8 r3, q3[0]
 ; CHECK-NEXT:    vmov.32 q0[0], r3
 ; CHECK-NEXT:    vmov.u8 r3, q3[1]
 ; CHECK-NEXT:    vmov.32 q0[2], r3
-; CHECK-NEXT:    vmov.u8 r3, q1[0]
-; CHECK-NEXT:    vand q2, q0, q4
+; CHECK-NEXT:    vmov.u8 r3, q4[0]
+; CHECK-NEXT:    vand q1, q0, q2
 ; CHECK-NEXT:    vmov.32 q0[0], r3
-; CHECK-NEXT:    vmov.u8 r3, q1[1]
-; CHECK-NEXT:    vmov r12, s8
+; CHECK-NEXT:    vmov.u8 r3, q4[1]
+; CHECK-NEXT:    vmov r12, s6
 ; CHECK-NEXT:    vmov.32 q0[2], r3
-; CHECK-NEXT:    vand q1, q0, q4
-; CHECK-NEXT:    vmov r3, s4
-; CHECK-NEXT:    umull r3, r2, r3, r12
+; CHECK-NEXT:    vmov lr, s4
+; CHECK-NEXT:    vand q0, q0, q2
+; CHECK-NEXT:    vmov r3, s2
+; CHECK-NEXT:    mul r12, r3, r12
+; CHECK-NEXT:    vmov r3, s0
+; CHECK-NEXT:    mul r3, r3, lr
+; CHECK-NEXT:    mov.w lr, #0
 ; CHECK-NEXT:    vmov.32 q0[0], r3
-; CHECK-NEXT:    vmov r3, s6
-; CHECK-NEXT:    vmov.32 q0[1], r2
-; CHECK-NEXT:    vmov r2, s10
-; CHECK-NEXT:    vldrw.u32 q2, [sp, #64] @ 16-byte Reload
-; CHECK-NEXT:    umull r2, r3, r3, r2
-; CHECK-NEXT:    vmov.32 q0[2], r2
-; CHECK-NEXT:    vmov.32 q0[3], r3
+; CHECK-NEXT:    vmov.32 q0[2], r12
 ; CHECK-NEXT:    vand q0, q0, q7
-; CHECK-NEXT:    vmov r4, s2
-; CHECK-NEXT:    vmov r2, s0
-; CHECK-NEXT:    vmov r12, s3
-; CHECK-NEXT:    vmov r3, s1
-; CHECK-NEXT:    adds r4, r4, r2
-; CHECK-NEXT:    ubfx r2, lr, #8, #1
+; CHECK-NEXT:    vmov r12, s2
+; CHECK-NEXT:    vmov r3, s0
+; CHECK-NEXT:    adds.w r5, r3, r12
+; CHECK-NEXT:    ubfx r3, r2, #8, #1
+; CHECK-NEXT:    ubfx r2, r2, #12, #1
+; CHECK-NEXT:    rsb.w r3, r3, #0
+; CHECK-NEXT:    vmov.32 q7[0], r3
 ; CHECK-NEXT:    rsb.w r2, r2, #0
-; CHECK-NEXT:    vmov.32 q7[0], r2
-; CHECK-NEXT:    adc.w r12, r12, r3
-; CHECK-NEXT:    vmov.32 q7[1], r2
-; CHECK-NEXT:    ubfx r2, lr, #12, #1
-; CHECK-NEXT:    rsbs r2, r2, #0
-; CHECK-NEXT:    vmov.u8 r3, q2[2]
 ; CHECK-NEXT:    vmov.32 q7[2], r2
-; CHECK-NEXT:    vmov.32 q1[0], r3
-; CHECK-NEXT:    vmov.32 q7[3], r2
 ; CHECK-NEXT:    vmov.u8 r2, q3[2]
+; CHECK-NEXT:    vmov.u8 r3, q4[2]
 ; CHECK-NEXT:    vmov.32 q0[0], r2
 ; CHECK-NEXT:    vmov.u8 r2, q3[3]
-; CHECK-NEXT:    vmov.u8 r3, q2[3]
+; CHECK-NEXT:    vmov.32 q1[0], r3
+; CHECK-NEXT:    vmov.u8 r3, q4[3]
 ; CHECK-NEXT:    vmov.32 q0[2], r2
 ; CHECK-NEXT:    vmov.32 q1[2], r3
-; CHECK-NEXT:    vand q0, q0, q4
-; CHECK-NEXT:    vand q1, q1, q4
-; CHECK-NEXT:    vmov r2, s0
-; CHECK-NEXT:    vmov r3, s4
-; CHECK-NEXT:    umull r2, r3, r3, r2
-; CHECK-NEXT:    vmov.32 q2[0], r2
+; CHECK-NEXT:    vand q0, q0, q2
+; CHECK-NEXT:    vand q1, q1, q2
 ; CHECK-NEXT:    vmov r2, s2
-; CHECK-NEXT:    vmov.32 q2[1], r3
 ; CHECK-NEXT:    vmov r3, s6
-; CHECK-NEXT:    umull r2, r3, r3, r2
-; CHECK-NEXT:    vmov.32 q2[2], r2
-; CHECK-NEXT:    vmov.32 q2[3], r3
-; CHECK-NEXT:    vand q0, q2, q7
-; 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 lr, r12, r2
-; CHECK-NEXT:    vmov r2, s2
-; CHECK-NEXT:    adds.w r12, r3, r2
+; CHECK-NEXT:    adc r12, lr, #0
+; CHECK-NEXT:    vmov r4, s4
+; CHECK-NEXT:    muls r2, r3, r2
+; CHECK-NEXT:    vmov r3, s0
+; CHECK-NEXT:    muls r3, r4, r3
+; CHECK-NEXT:    vmov.u8 r4, q4[4]
+; CHECK-NEXT:    vmov.32 q1[0], r4
+; CHECK-NEXT:    vmov.u8 r4, q4[5]
+; CHECK-NEXT:    vmov.32 q0[0], r3
+; CHECK-NEXT:    vmov.32 q1[2], r4
+; CHECK-NEXT:    vmov.32 q0[2], r2
+; CHECK-NEXT:    vand q1, q1, q2
+; CHECK-NEXT:    vand q0, q0, q7
+; CHECK-NEXT:    vmov r4, s6
+; CHECK-NEXT:    vmov r2, s0
+; CHECK-NEXT:    vmov q7, q4
+; CHECK-NEXT:    adds r2, r2, r5
+; CHECK-NEXT:    vmov r5, s2
+; CHECK-NEXT:    adc r3, r12, #0
+; CHECK-NEXT:    adds.w r12, r2, r5
 ; CHECK-NEXT:    vmov.u16 r2, q6[4]
 ; CHECK-NEXT:    vmov.32 q0[0], r2
 ; CHECK-NEXT:    vmov.u16 r2, q6[5]
 ; CHECK-NEXT:    vmov.32 q0[1], r2
 ; CHECK-NEXT:    vmov.u16 r2, q6[6]
 ; CHECK-NEXT:    vmov.32 q0[2], r2
 ; CHECK-NEXT:    vmov.u16 r2, q6[7]
 ; CHECK-NEXT:    vmov.32 q0[3], r2
-; CHECK-NEXT:    adc.w lr, lr, r4
+; CHECK-NEXT:    adc lr, r3, #0
 ; CHECK-NEXT:    vcmp.i32 ne, q0, zr
-; CHECK-NEXT:    vmrs r6, p0
-; CHECK-NEXT:    vstrw.32 q3, [sp, #48] @ 16-byte Spill
-; CHECK-NEXT:    vstrw.32 q4, [sp] @ 16-byte Spill
-; CHECK-NEXT:    and r4, r6, #1
-; CHECK-NEXT:    rsbs r4, r4, #0
-; CHECK-NEXT:    vmov.32 q6[0], r4
-; CHECK-NEXT:    vmov.32 q6[1], r4
-; CHECK-NEXT:    ubfx r4, r6, #4, #1
-; CHECK-NEXT:    rsbs r4, r4, #0
-; CHECK-NEXT:    vmov.32 q6[2], r4
-; CHECK-NEXT:    vmov.32 q6[3], r4
-; CHECK-NEXT:    vmov.u8 r4, q3[4]
-; CHECK-NEXT:    vmov.32 q0[0], r4
-; CHECK-NEXT:    vmov.u8 r4, q3[5]
-; CHECK-NEXT:    vldrw.u32 q3, [sp, #64] @ 16-byte Reload
-; CHECK-NEXT:    vmov.32 q0[2], r4
-; CHECK-NEXT:    vand q0, q0, q4
-; CHECK-NEXT:    vmov.u8 r3, q3[4]
-; CHECK-NEXT:    vmov r4, s0
-; CHECK-NEXT:    vmov.32 q1[0], r3
-; CHECK-NEXT:    vmov.u8 r3, q3[5]
-; CHECK-NEXT:    vmov.32 q1[2], r3
-; CHECK-NEXT:    vmov q7, q3
-; CHECK-NEXT:    vand q1, q1, q4
 ; CHECK-NEXT:    vmov r3, s4
-; CHECK-NEXT:    umull r3, r4, r3, r4
-; CHECK-NEXT:    vmov.32 q2[0], r3
-; CHECK-NEXT:    vmov r3, s2
-; CHECK-NEXT:    vmov.32 q2[1], r4
-; CHECK-NEXT:    vmov r4, s6
-; CHECK-NEXT:    vldrw.u32 q1, [sp, #48] @ 16-byte Reload
-; CHECK-NEXT:    umull r3, r4, r4, r3
-; CHECK-NEXT:    vmov.32 q2[2], r3
-; CHECK-NEXT:    vmov.32 q2[3], r4
-; CHECK-NEXT:    vand q0, q2, q6
-; CHECK-NEXT:    vmov r4, s0
-; CHECK-NEXT:    vmov r3, s1
+; CHECK-NEXT:    vmrs r2, p0
+; CHECK-NEXT:    vstrw.32 q3, [sp, #32] @ 16-byte Spill
+; CHECK-NEXT:    vldrw.u32 q1, [sp, #32] @ 16-byte Reload
+; CHECK-NEXT:    and r5, r2, #1
+; CHECK-NEXT:    rsbs r5, r5, #0
+; CHECK-NEXT:    vmov.32 q6[0], r5
+; CHECK-NEXT:    ubfx r5, r2, #4, #1
+; CHECK-NEXT:    rsbs r5, r5, #0
+; CHECK-NEXT:    vmov.32 q6[2], r5
+; CHECK-NEXT:    vmov.u8 r5, q3[4]
+; CHECK-NEXT:    vmov.32 q0[0], r5
+; CHECK-NEXT:    vmov.u8 r5, q3[5]
+; CHECK-NEXT:    vmov.32 q0[2], r5
+; CHECK-NEXT:    vand q0, q0, q2
 ; CHECK-NEXT:    vmov r5, s2
-; CHECK-NEXT:    vmov r2, s3
-; CHECK-NEXT:    adds.w r4, r4, r12
-; CHECK-NEXT:    adc.w r3, r3, lr
-; CHECK-NEXT:    adds r4, r4, r5
-; CHECK-NEXT:    adc.w r12, r3, r2
-; CHECK-NEXT:    ubfx r2, r6, #8, #1
-; CHECK-NEXT:    rsbs r2, r2, #0
-; CHECK-NEXT:    vmov.u8 r3, q3[6]
-; CHECK-NEXT:    vmov.32 q6[0], r2
-; CHECK-NEXT:    vmov.32 q6[1], r2
-; CHECK-NEXT:    ubfx r2, r6, #12, #1
+; CHECK-NEXT:    muls r5, r4, r5
+; CHECK-NEXT:    vmov r4, s0
+; CHECK-NEXT:    muls r3, r4, r3
+; CHECK-NEXT:    vmov.32 q0[0], r3
+; CHECK-NEXT:    vmov.32 q0[2], r5
+; CHECK-NEXT:    vand q0, q0, q6
+; CHECK-NEXT:    vmov r3, s0
+; CHECK-NEXT:    vmov r4, s2
+; CHECK-NEXT:    adds.w r3, r3, r12
+; CHECK-NEXT:    adc r5, lr, #0
+; CHECK-NEXT:    adds r3, r3, r4
+; CHECK-NEXT:    adc r12, r5, #0
+; CHECK-NEXT:    ubfx r5, r2, #8, #1
+; CHECK-NEXT:    ubfx r2, r2, #12, #1
+; CHECK-NEXT:    rsbs r5, r5, #0
+; CHECK-NEXT:    vmov.32 q6[0], r5
 ; CHECK-NEXT:    rsbs r2, r2, #0
 ; CHECK-NEXT:    vmov.32 q6[2], r2
-; CHECK-NEXT:    vmov.32 q6[3], r2
 ; CHECK-NEXT:    vmov.u8 r2, q1[6]
+; CHECK-NEXT:    vmov.u8 r5, q4[6]
 ; CHECK-NEXT:    vmov.32 q0[0], r2
 ; CHECK-NEXT:    vmov.u8 r2, q1[7]
-; CHECK-NEXT:    vmov.32 q1[0], r3
-; CHECK-NEXT:    vmov.u8 r3, q3[7]
+; CHECK-NEXT:    vmov.32 q1[0], r5
+; CHECK-NEXT:    vmov.u8 r5, q4[7]
 ; CHECK-NEXT:    vmov.32 q0[2], r2
-; CHECK-NEXT:    vmov.32 q1[2], r3
-; CHECK-NEXT:    vand q0, q0, q4
-; CHECK-NEXT:    vand q1, q1, q4
-; CHECK-NEXT:    vmov r2, s0
-; CHECK-NEXT:    vmov r3, s4
-; CHECK-NEXT:    umull r2, r3, r3, r2
-; CHECK-NEXT:    vmov.32 q2[0], r2
+; CHECK-NEXT:    vmov.32 q1[2], r5
+; CHECK-NEXT:    vand q0, q0, q2
+; CHECK-NEXT:    vand q1, q1, q2
 ; CHECK-NEXT:    vmov r2, s2
-; CHECK-NEXT:    vmov.32 q2[1], r3
-; CHECK-NEXT:    vmov r3, s6
-; CHECK-NEXT:    vldrw.u32 q1, [sp, #16] @ 16-byte Reload
-; CHECK-NEXT:    umull r2, r3, r3, r2
-; CHECK-NEXT:    vmov.32 q2[2], r2
-; CHECK-NEXT:    vmov.32 q2[3], r3
-; CHECK-NEXT:    vand q0, q2, q6
-; CHECK-NEXT:    vmov r3, s0
-; CHECK-NEXT:    vmov r2, s1
+; CHECK-NEXT:    vmov r5, s6
+; CHECK-NEXT:    vmov r4, s4
+; CHECK-NEXT:    vldrw.u32 q1, [sp] @ 16-byte Reload
+; CHECK-NEXT:    muls r2, r5, r2
+; CHECK-NEXT:    vmov r5, s0
+; CHECK-NEXT:    muls r5, r4, r5
+; CHECK-NEXT:    vmov.u8 r4, q7[8]
+; CHECK-NEXT:    vmov.32 q0[0], r5
+; CHECK-NEXT:    vmov.32 q0[2], r2
+; CHECK-NEXT:    vand q0, q0, q6
+; CHECK-NEXT:    vmov r2, s0
 ; CHECK-NEXT:    vmov r5, s2
-; CHECK-NEXT:    vmov r6, s3
-; CHECK-NEXT:    vldrw.u32 q0, [sp, #32] @ 16-byte Reload
-; CHECK-NEXT:    adds r3, r3, r4
-; CHECK-NEXT:    adc.w r2, r2, r12
-; CHECK-NEXT:    adds.w r12, r3, r5
-; CHECK-NEXT:    vmov.u8 r5, q7[8]
-; CHECK-NEXT:    adc.w r3, r2, r6
+; CHECK-NEXT:    vldrw.u32 q0, [sp, #16] @ 16-byte Reload
+; CHECK-NEXT:    adds r2, r2, r3
+; CHECK-NEXT:    adc r3, r12, #0
+; CHECK-NEXT:    adds.w r12, r2, r5
 ; CHECK-NEXT:    vmov.u8 r2, q5[8]
+; CHECK-NEXT:    adc lr, r3, #0
 ; CHECK-NEXT:    vmov.16 q6[0], r2
 ; CHECK-NEXT:    vmov.u8 r2, q5[9]
 ; CHECK-NEXT:    vmov.16 q6[1], r2
 ; CHECK-NEXT:    vmov.u8 r2, q5[10]
 ; CHECK-NEXT:    vmov.16 q6[2], r2
 ; CHECK-NEXT:    vmov.u8 r2, q5[11]
 ; CHECK-NEXT:    vmov.16 q6[3], r2
 ; CHECK-NEXT:    vmov.u8 r2, q5[12]
 ; CHECK-NEXT:    vmov.16 q6[4], r2
 ; CHECK-NEXT:    vmov.u8 r2, q5[13]
 ; CHECK-NEXT:    vmov.16 q6[5], r2
 ; CHECK-NEXT:    vmov.u8 r2, q5[14]
 ; CHECK-NEXT:    vmov.16 q6[6], r2
 ; CHECK-NEXT:    vmov.u8 r2, q5[15]
 ; CHECK-NEXT:    vmov.16 q6[7], r2
 ; CHECK-NEXT:    vcmp.i16 ne, q6, zr
 ; CHECK-NEXT:    vpsel q3, q1, q0
-; CHECK-NEXT:    vmov.32 q1[0], r5
+; CHECK-NEXT:    vmov.32 q1[0], r4
 ; CHECK-NEXT:    vmov.u16 r2, q3[0]
-; CHECK-NEXT:    vmov.u8 r5, q7[9]
+; CHECK-NEXT:    vmov.u8 r4, q7[9]
 ; CHECK-NEXT:    vmov.32 q0[0], r2
 ; CHECK-NEXT:    vmov.u16 r2, q3[1]
 ; CHECK-NEXT:    vmov.32 q0[1], r2
 ; CHECK-NEXT:    vmov.u16 r2, q3[2]
 ; CHECK-NEXT:    vmov.32 q0[2], r2
 ; CHECK-NEXT:    vmov.u16 r2, q3[3]
 ; CHECK-NEXT:    vmov.32 q0[3], r2
-; CHECK-NEXT:    vmov.32 q1[2], r5
+; CHECK-NEXT:    vmov.32 q1[2], r4
 ; CHECK-NEXT:    vcmp.i32 ne, q0, zr
+; CHECK-NEXT:    vand q1, q1, q2
 ; CHECK-NEXT:    vmrs r2, p0
-; CHECK-NEXT:    vldrw.u32 q6, [sp, #48] @ 16-byte Reload
-; CHECK-NEXT:    vldrw.u32 q5, [sp] @ 16-byte Reload
-; CHECK-NEXT:    vand q1, q1, q5
-; CHECK-NEXT:    vmov r5, s4
-; CHECK-NEXT:    and r6, r2, #1
-; CHECK-NEXT:    rsbs r6, r6, #0
-; CHECK-NEXT:    vmov.32 q4[0], r6
-; CHECK-NEXT:    vmov.32 q4[1], r6
-; CHECK-NEXT:    ubfx r6, r2, #4, #1
-; CHECK-NEXT:    rsbs r6, r6, #0
-; CHECK-NEXT:    vmov.32 q4[2], r6
-; CHECK-NEXT:    vmov.32 q4[3], r6
-; CHECK-NEXT:    vmov.u8 r6, q6[8]
-; CHECK-NEXT:    vmov.32 q0[0], r6
-; CHECK-NEXT:    vmov.u8 r6, q6[9]
-; CHECK-NEXT:    vmov.32 q0[2], r6
-; CHECK-NEXT:    vand q0, q0, q5
-; CHECK-NEXT:    vmov r6, s0
-; CHECK-NEXT:    umull r6, r5, r5, r6
-; CHECK-NEXT:    vmov.32 q2[0], r6
-; CHECK-NEXT:    vmov r6, s2
-; CHECK-NEXT:    vmov.32 q2[1], r5
-; CHECK-NEXT:    vmov r5, s6
-; CHECK-NEXT:    umull r6, r5, r5, r6
-; CHECK-NEXT:    vmov.32 q2[2], r6
-; CHECK-NEXT:    vmov.32 q2[3], r5
-; CHECK-NEXT:    vand q0, q2, q4
-; CHECK-NEXT:    vmov r5, s0
-; CHECK-NEXT:    vmov r6, s1
-; CHECK-NEXT:    vmov r4, s3
-; CHECK-NEXT:    adds.w r5, r5, r12
-; CHECK-NEXT:    adcs r6, r3
-; CHECK-NEXT:    vmov r3, s2
-; CHECK-NEXT:    adds r3, r3, r5
-; CHECK-NEXT:    adc.w r12, r6, r4
-; CHECK-NEXT:    ubfx r6, r2, #8, #1
-; CHECK-NEXT:    rsbs r6, r6, #0
+; CHECK-NEXT:    vldrw.u32 q5, [sp, #32] @ 16-byte Reload
+; CHECK-NEXT:    vmov r4, s6
+; CHECK-NEXT:    vmov r3, s4
+; CHECK-NEXT:    and r5, r2, #1
+; CHECK-NEXT:    rsbs r5, r5, #0
+; CHECK-NEXT:    vmov.32 q4[0], r5
+; CHECK-NEXT:    ubfx r5, r2, #4, #1
+; CHECK-NEXT:    rsbs r5, r5, #0
+; CHECK-NEXT:    vmov.32 q4[2], r5
+; CHECK-NEXT:    vmov.u8 r5, q5[8]
+; CHECK-NEXT:    vmov.32 q0[0], r5
+; CHECK-NEXT:    vmov.u8 r5, q5[9]
+; CHECK-NEXT:    vmov.32 q0[2], r5
+; CHECK-NEXT:    vand q0, q0, q2
+; CHECK-NEXT:    vmov r5, s2
+; CHECK-NEXT:    muls r5, r4, r5
+; CHECK-NEXT:    vmov r4, s0
+; CHECK-NEXT:    muls r3, r4, r3
+; CHECK-NEXT:    vmov.32 q0[0], r3
+; CHECK-NEXT:    vmov.32 q0[2], r5
+; CHECK-NEXT:    vand q0, q0, q4
+; CHECK-NEXT:    vmov r3, s0
+; CHECK-NEXT:    vmov r4, s2
+; CHECK-NEXT:    adds.w r3, r3, r12
+; CHECK-NEXT:    adc r5, lr, #0
+; CHECK-NEXT:    adds r3, r3, r4
+; CHECK-NEXT:    adc r12, r5, #0
+; CHECK-NEXT:    ubfx r5, r2, #8, #1
 ; CHECK-NEXT:    ubfx r2, r2, #12, #1
-; CHECK-NEXT:    vmov.32 q4[0], r6
+; CHECK-NEXT:    rsbs r5, r5, #0
+; CHECK-NEXT:    vmov.32 q4[0], r5
 ; CHECK-NEXT:    rsbs r2, r2, #0
-; CHECK-NEXT:    vmov.32 q4[1], r6
-; CHECK-NEXT:    vmov.u8 r6, q7[10]
 ; CHECK-NEXT:    vmov.32 q4[2], r2
-; CHECK-NEXT:    vmov.32 q1[0], r6
-; CHECK-NEXT:    vmov.32 q4[3], r2
-; CHECK-NEXT:    vmov.u8 r2, q6[10]
+; CHECK-NEXT:    vmov.u8 r2, q5[10]
+; CHECK-NEXT:    vmov.u8 r5, q7[10]
 ; CHECK-NEXT:    vmov.32 q0[0], r2
-; CHECK-NEXT:    vmov.u8 r2, q6[11]
-; CHECK-NEXT:    vmov.u8 r6, q7[11]
+; CHECK-NEXT:    vmov.u8 r2, q5[11]
+; CHECK-NEXT:    vmov.32 q1[0], r5
+; CHECK-NEXT:    vmov.u8 r5, q7[11]
 ; CHECK-NEXT:    vmov.32 q0[2], r2
-; CHECK-NEXT:    vmov.32 q1[2], r6
-; CHECK-NEXT:    vand q0, q0, q5
-; CHECK-NEXT:    vand q1, q1, q5
-; CHECK-NEXT:    vmov r2, s0
-; CHECK-NEXT:    vmov r6, s4
-; CHECK-NEXT:    umull r2, r6, r6, r2
-; CHECK-NEXT:    vmov.32 q2[0], r2
+; CHECK-NEXT:    vmov.32 q1[2], r5
+; CHECK-NEXT:    vand q0, q0, q2
+; CHECK-NEXT:    vand q1, q1, q2
 ; CHECK-NEXT:    vmov r2, s2
-; CHECK-NEXT:    vmov.32 q2[1], r6
-; CHECK-NEXT:    vmov r6, s6
-; CHECK-NEXT:    umull r2, r6, r6, r2
-; CHECK-NEXT:    vmov.32 q2[2], r2
-; CHECK-NEXT:    vmov.32 q2[3], r6
-; CHECK-NEXT:    vand q0, q2, q4
-; CHECK-NEXT:    vmov r6, s0
-; CHECK-NEXT:    vmov r2, s1
+; CHECK-NEXT:    vmov r5, s6
+; CHECK-NEXT:    vmov r4, s4
+; CHECK-NEXT:    muls r2, r5, r2
+; CHECK-NEXT:    vmov r5, s0
+; CHECK-NEXT:    muls r5, r4, r5
+; CHECK-NEXT:    vmov.u8 r4, q7[12]
+; CHECK-NEXT:    vmov.32 q1[0], r4
+; CHECK-NEXT:    vmov.u8 r4, q7[13]
+; CHECK-NEXT:    vmov.32 q0[0], r5
+; CHECK-NEXT:    vmov.32 q1[2], r4
+; CHECK-NEXT:    vmov.32 q0[2], r2
+; CHECK-NEXT:    vand q1, q1, q2
+; CHECK-NEXT:    vand q0, q0, q4
+; CHECK-NEXT:    vmov r4, s6
+; CHECK-NEXT:    vmov r2, s0
 ; CHECK-NEXT:    vmov r5, s2
-; CHECK-NEXT:    adds r3, r3, r6
-; CHECK-NEXT:    vmov r6, s3
-; CHECK-NEXT:    adc.w r2, r2, r12
-; CHECK-NEXT:    adds.w r12, r3, r5
-; CHECK-NEXT:    vmov.u8 r5, q7[12]
-; CHECK-NEXT:    vmov.32 q1[0], r5
-; CHECK-NEXT:    vmov.u8 r5, q7[13]
-; CHECK-NEXT:    vmov.32 q1[2], r5
-; CHECK-NEXT:    vand q1, q1, q5
-; CHECK-NEXT:    vmov r5, s4
-; CHECK-NEXT:    adc.w r3, r2, r6
+; CHECK-NEXT:    adds r2, r2, r3
+; CHECK-NEXT:    adc r3, r12, #0
+; CHECK-NEXT:    adds.w r12, r2, r5
 ; CHECK-NEXT:    vmov.u16 r2, q3[4]
+; CHECK-NEXT:    adc lr, r3, #0
 ; CHECK-NEXT:    vmov.32 q0[0], r2
 ; CHECK-NEXT:    vmov.u16 r2, q3[5]
 ; CHECK-NEXT:    vmov.32 q0[1], r2
 ; CHECK-NEXT:    vmov.u16 r2, q3[6]
 ; CHECK-NEXT:    vmov.32 q0[2], r2
 ; CHECK-NEXT:    vmov.u16 r2, q3[7]
 ; CHECK-NEXT:    vmov.32 q0[3], r2
+; CHECK-NEXT:    vmov r3, s4
 ; CHECK-NEXT:    vcmp.i32 ne, q0, zr
 ; CHECK-NEXT:    vmrs r2, p0
-; CHECK-NEXT:    and r6, r2, #1
-; CHECK-NEXT:    rsbs r6, r6, #0
-; CHECK-NEXT:    vmov.32 q3[0], r6
-; CHECK-NEXT:    vmov.32 q3[1], r6
-; CHECK-NEXT:    ubfx r6, r2, #4, #1
-; CHECK-NEXT:    rsbs r6, r6, #0
-; CHECK-NEXT:    vmov.32 q3[2], r6
-; CHECK-NEXT:    vmov.32 q3[3], r6
-; CHECK-NEXT:    vmov.u8 r6, q6[12]
-; CHECK-NEXT:    vmov.32 q0[0], r6
-; CHECK-NEXT:    vmov.u8 r6, q6[13]
-; CHECK-NEXT:    vmov.32 q0[2], r6
-; CHECK-NEXT:    vand q0, q0, q5
-; CHECK-NEXT:    vmov r6, s0
-; CHECK-NEXT:    umull r6, r5, r5, r6
-; CHECK-NEXT:    vmov.32 q2[0], r6
-; CHECK-NEXT:    vmov r6, s2
-; CHECK-NEXT:    vmov.32 q2[1], r5
-; CHECK-NEXT:    vmov r5, s6
-; CHECK-NEXT:    umull r6, r5, r5, r6
-; CHECK-NEXT:    vmov.32 q2[2], r6
-; CHECK-NEXT:    vmov.32 q2[3], r5
-; CHECK-NEXT:    vand q0, q2, q3
-; CHECK-NEXT:    vmov r5, s0
-; CHECK-NEXT:    vmov r6, s1
-; CHECK-NEXT:    vmov r4, s3
-; CHECK-NEXT:    adds.w r5, r5, r12
-; CHECK-NEXT:    adcs r6, r3
-; CHECK-NEXT:    vmov r3, s2
-; CHECK-NEXT:    adds r3, r3, r5
-; CHECK-NEXT:    adc.w r12, r6, r4
-; CHECK-NEXT:    ubfx r6, r2, #8, #1
-; CHECK-NEXT:    rsbs r6, r6, #0
+; CHECK-NEXT:    and r5, r2, #1
+; CHECK-NEXT:    rsbs r5, r5, #0
+; CHECK-NEXT:    vmov.32 q3[0], r5
+; CHECK-NEXT:    ubfx r5, r2, #4, #1
+; CHECK-NEXT:    rsbs r5, r5, #0
+; CHECK-NEXT:    vmov.32 q3[2], r5
+; CHECK-NEXT:    vmov.u8 r5, q5[12]
+; CHECK-NEXT:    vmov.32 q0[0], r5
+; CHECK-NEXT:    vmov.u8 r5, q5[13]
+; CHECK-NEXT:    vmov.32 q0[2], r5
+; CHECK-NEXT:    vand q0, q0, q2
+; CHECK-NEXT:    vmov r5, s2
+; CHECK-NEXT:    muls r5, r4, r5
+; CHECK-NEXT:    vmov r4, s0
+; CHECK-NEXT:    muls r3, r4, r3
+; CHECK-NEXT:    vmov.32 q0[0], r3
+; CHECK-NEXT:    vmov.32 q0[2], r5
+; CHECK-NEXT:    vand q0, q0, q3
+; CHECK-NEXT:    vmov r3, s0
+; CHECK-NEXT:    vmov r4, s2
+; CHECK-NEXT:    adds.w r3, r3, r12
+; CHECK-NEXT:    adc r5, lr, #0
+; CHECK-NEXT:    adds r3, r3, r4
+; CHECK-NEXT:    adc r12, r5, #0
+; CHECK-NEXT:    ubfx r5, r2, #8, #1
 ; CHECK-NEXT:    ubfx r2, r2, #12, #1
-; CHECK-NEXT:    vmov.32 q3[0], r6
+; CHECK-NEXT:    rsbs r5, r5, #0
+; CHECK-NEXT:    vmov.32 q3[0], r5
 ; CHECK-NEXT:    rsbs r2, r2, #0
-; CHECK-NEXT:    vmov.32 q3[1], r6
-; CHECK-NEXT:    vmov.u8 r6, q7[14]
 ; CHECK-NEXT:    vmov.32 q3[2], r2
-; CHECK-NEXT:    vmov.32 q1[0], r6
-; CHECK-NEXT:    vmov.32 q3[3], r2
-; CHECK-NEXT:    vmov.u8 r2, q6[14]
+; CHECK-NEXT:    vmov.u8 r2, q5[14]
+; CHECK-NEXT:    vmov.u8 r5, q7[14]
 ; CHECK-NEXT:    vmov.32 q0[0], r2
-; CHECK-NEXT:    vmov.u8 r2, q6[15]
-; CHECK-NEXT:    vmov.u8 r6, q7[15]
+; CHECK-NEXT:    vmov.u8 r2, q5[15]
+; CHECK-NEXT:    vmov.32 q1[0], r5
+; CHECK-NEXT:    vmov.u8 r5, q7[15]
 ; CHECK-NEXT:    vmov.32 q0[2], r2
-; CHECK-NEXT:    vmov.32 q1[2], r6
-; CHECK-NEXT:    vand q0, q0, q5
-; CHECK-NEXT:    vand q1, q1, q5
-; CHECK-NEXT:    vmov r2, s0
-; CHECK-NEXT:    vmov r6, s4
-; CHECK-NEXT:    umull r2, r6, r6, r2
-; CHECK-NEXT:    vmov.32 q2[0], r2
+; CHECK-NEXT:    vmov.32 q1[2], r5
+; CHECK-NEXT:    vand q0, q0, q2
+; CHECK-NEXT:    vand q1, q1, q2
 ; CHECK-NEXT:    vmov r2, s2
-; CHECK-NEXT:    vmov.32 q2[1], r6
-; CHECK-NEXT:    vmov r6, s6
-; CHECK-NEXT:    umull r2, r6, r6, r2
-; CHECK-NEXT:    vmov.32 q2[2], r2
-; CHECK-NEXT:    vmov.32 q2[3], r6
-; CHECK-NEXT:    vand q0, q2, q3
-; CHECK-NEXT:    vmov r6, s0
-; CHECK-NEXT:    vmov r2, s1
+; CHECK-NEXT:    vmov r5, s6
+; CHECK-NEXT:    vmov r4, s4
+; CHECK-NEXT:    muls r2, r5, r2
+; CHECK-NEXT:    vmov r5, s0
+; CHECK-NEXT:    muls r5, r4, r5
+; CHECK-NEXT:    vmov.32 q0[0], r5
+; CHECK-NEXT:    vmov.32 q0[2], r2
+; CHECK-NEXT:    vand q0, q0, q3
+; CHECK-NEXT:    vmov r2, s0
 ; CHECK-NEXT:    vmov r5, s2
-; CHECK-NEXT:    adds r3, r3, r6
-; CHECK-NEXT:    vmov r6, s3
-; CHECK-NEXT:    adc.w r2, r2, r12
-; CHECK-NEXT:    adds r3, r3, r5
-; CHECK-NEXT:    adcs r2, r6
-; CHECK-NEXT:    adds r0, r0, r3
-; CHECK-NEXT:    adcs r1, r2
-; CHECK-NEXT:    add sp, #80
+; CHECK-NEXT:    adds r2, r2, r3
+; CHECK-NEXT:    adc r3, r12, #0
+; CHECK-NEXT:    adds r2, r2, r5
+; CHECK-NEXT:    adc r3, r3, #0
+; CHECK-NEXT:    adds r0, r0, r2
+; CHECK-NEXT:    adcs r1, r3
+; CHECK-NEXT:    add sp, #48
 ; CHECK-NEXT:    vpop {d8, d9, d10, d11, d12, d13, d14, d15}
-; CHECK-NEXT:    pop {r4, r5, r6, pc}
+; CHECK-NEXT:    pop {r4, r5, r7, pc}
 entry:
   %c = icmp eq <16 x i8> %b, zeroinitializer
   %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
   %s = select <16 x i1> %c, <16 x i64> %m, <16 x i64> zeroinitializer
   %z = call i64 @llvm.experimental.vector.reduce.add.v16i64(<16 x i64> %s)
   %r = add i64 %z, %a
entry:
   %z = call i64 @llvm.experimental.vector.reduce.add.v16i64(<16 x i64> %s)
   %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, <2 x i8> %b, 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:    .vsave {d8, d9}
-; CHECK-NEXT:    vpush {d8, d9}
 ; CHECK-NEXT:    vmov.i64 q3, #0xff
 ; CHECK-NEXT:    vand q1, q1, q3
-; CHECK-NEXT:    vand q4, q0, q3
-; CHECK-NEXT:    vmov r2, s4
-; CHECK-NEXT:    vmov r3, s16
-; CHECK-NEXT:    umull r2, r3, r3, r2
-; CHECK-NEXT:    vmov.32 q0[0], r2
+; CHECK-NEXT:    vand q0, q0, q3
 ; CHECK-NEXT:    vmov r2, s6
-; CHECK-NEXT:    vmov.32 q0[1], r3
-; CHECK-NEXT:    vmov r3, s18
+; CHECK-NEXT:    vmov r3, s2
+; CHECK-NEXT:    mul r12, r3, r2
+; CHECK-NEXT:    vmov r3, s4
+; CHECK-NEXT:    vmov r2, s0
 ; CHECK-NEXT:    vand q1, q2, q3
-; CHECK-NEXT:    umull r2, r3, r3, r2
-; CHECK-NEXT:    vmov.32 q0[2], r2
+; CHECK-NEXT:    muls r2, r3, r2
+; CHECK-NEXT:    vmov.32 q0[0], r2
 ; CHECK-NEXT:    vmov r2, s4
-; CHECK-NEXT:    vmov.32 q0[3], r3
+; CHECK-NEXT:    vmov.32 q0[2], r12
+; CHECK-NEXT:    mov.w r12, #0
 ; CHECK-NEXT:    cmp r2, #0
 ; CHECK-NEXT:    cset r2, eq
 ; CHECK-NEXT:    tst.w r2, #1
 ; CHECK-NEXT:    csetm r2, ne
 ; CHECK-NEXT:    vmov.32 q2[0], r2
-; CHECK-NEXT:    vmov.32 q2[1], r2
 ; CHECK-NEXT:    vmov r2, s6
 ; CHECK-NEXT:    cmp r2, #0
 ; CHECK-NEXT:    cset r2, eq
 ; CHECK-NEXT:    tst.w r2, #1
 ; CHECK-NEXT:    csetm r2, ne
 ; CHECK-NEXT:    vmov.32 q2[2], r2
-; CHECK-NEXT:    vmov.32 q2[3], r2
 ; CHECK-NEXT:    vand q0, q0, q2
 ; 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:    vpop {d8, d9}
-; CHECK-NEXT:    pop {r7, pc}
+; CHECK-NEXT:    bx lr
 entry:
   %c = icmp eq <2 x i8> %b, zeroinitializer
   %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
   %s = select <2 x i1> %c, <2 x i64> %m, <2 x i64> zeroinitializer
   %z = call i64 @llvm.experimental.vector.reduce.add.v2i64(<2 x i64> %s)
   %r = add i64 %z, %a

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, d8
+; CHECK-NEXT:    movs r1, #0
+; CHECK-NEXT:    vmov.f32 s22, s17
+; 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 d9, r0, r1
-; CHECK-NEXT:    vmov r2, s2
-; CHECK-NEXT:    vmov r3, s3
+; CHECK-NEXT:    movs r1, #0
 ; CHECK-NEXT:    vmov r6, s0
-; CHECK-NEXT:    vmov r7, s1
 ; CHECK-NEXT:    vmov d8, r4, r5
+; CHECK-NEXT:    vmov r2, s2
 ; 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:    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() {
 ; X32-LABEL: zero_test:
 ; X32:       # %bb.0: # %entry
-; X32-NEXT:    xorps %xmm0, %xmm0
-; X32-NEXT:    movlps %xmm0, (%eax)
+; X32-NEXT:    movl $0, (%eax)
 ; X32-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) {
 ; CHECK-LABEL: store_64:
 ; CHECK:       # %bb.0: # %BB
 ; CHECK-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; CHECK-NEXT:    xorps %xmm0, %xmm0
-; CHECK-NEXT:    movlps %xmm0, (%eax)
+; CHECK-NEXT:    movl $0, 4(%eax)
+; CHECK-NEXT:    movl $0, (%eax)
 ; CHECK-NEXT:    retl
 BB:
   store <2 x i32> zeroinitializer, <2 x i32>* %ptr
   ret void
 }
 
 define <2 x i32> @load_64(<2 x i32>* %ptr) {
 ; CHECK-LABEL: load_64:
 ; CHECK:       # %bb.0: # %BB
 ; CHECK-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; CHECK-NEXT:    movsd {{.*#+}} xmm0 = mem[0],zero
 ; CHECK-NEXT:    retl
 BB:
   %t = load <2 x i32>, <2 x i32>* %ptr
   ret <2 x i32> %t
 }

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

 ; AVX512-NEXT:    movntiq %rax, 24(%rdi)
 ; AVX512-NEXT:    movntiq %rax, 16(%rdi)
 ; AVX512-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
+; SSE-LABEL: test_zero_v8f32_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_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
   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
+; SSE-LABEL: test_zero_v4i64_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_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
   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
+; SSE-LABEL: test_zero_v8i32_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_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
   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
+; SSE-LABEL: test_zero_v16i16_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_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
   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
+; SSE-LABEL: test_zero_v32i8_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_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)

 ; AVX512-NEXT:    movntiq %rax, 40(%rdi)
 ; AVX512-NEXT:    movntiq %rax, 32(%rdi)
 ; AVX512-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
+; SSE-LABEL: test_zero_v16f32_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_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)

 ; AVX512-NEXT:    movntiq %rax, 40(%rdi)
 ; AVX512-NEXT:    movntiq %rax, 32(%rdi)
 ; AVX512-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
+; SSE-LABEL: test_zero_v8i64_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_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)

 ; AVX512-NEXT:    movntiq %rax, 40(%rdi)
 ; AVX512-NEXT:    movntiq %rax, 32(%rdi)
 ; AVX512-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
+; SSE-LABEL: test_zero_v16i32_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_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)

 ; AVX512-NEXT:    movntiq %rax, 40(%rdi)
 ; AVX512-NEXT:    movntiq %rax, 32(%rdi)
 ; AVX512-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
+; SSE-LABEL: test_zero_v32i16_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_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)

 ; AVX512-NEXT:    movntiq %rax, 40(%rdi)
 ; AVX512-NEXT:    movntiq %rax, 32(%rdi)
 ; AVX512-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
+; SSE-LABEL: test_zero_v64i8_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_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)

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/vec_zero_cse.ll

 @S1 = external global <2 x i64>
 @S2 = external 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

 ; X32-SSE-LABEL: var_funnnel_v2i64:
 ; X32-SSE:       # %bb.0:
 ; X32-SSE-NEXT:    movdqa {{.*#+}} xmm3 = [63,0,63,0]
 ; X32-SSE-NEXT:    movdqa %xmm2, %xmm4
 ; X32-SSE-NEXT:    pandn %xmm3, %xmm4
 ; X32-SSE-NEXT:    psrlq $1, %xmm1
 ; X32-SSE-NEXT:    movdqa %xmm1, %xmm5
 ; X32-SSE-NEXT:    psrlq %xmm4, %xmm5
-; X32-SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[2,3,2,3]
+; X32-SSE-NEXT:    pxor %xmm6, %xmm6
+; X32-SSE-NEXT:    punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm6[2],xmm4[3],xmm6[3]
 ; X32-SSE-NEXT:    psrlq %xmm4, %xmm1
 ; X32-SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm5[0],xmm1[1]
 ; X32-SSE-NEXT:    pand %xmm3, %xmm2
 ; X32-SSE-NEXT:    movdqa %xmm0, %xmm3
 ; X32-SSE-NEXT:    psllq %xmm2, %xmm3
-; X32-SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,3,2,3]
+; X32-SSE-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm6[2],xmm2[3],xmm6[3]
 ; X32-SSE-NEXT:    psllq %xmm2, %xmm0
 ; X32-SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm3[0],xmm0[1]
 ; X32-SSE-NEXT:    orpd %xmm1, %xmm0
 ; X32-SSE-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
 }
 

 ; X32-SSE:       # %bb.0:
 ; X32-SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,0,1]
 ; X32-SSE-NEXT:    movdqa {{.*#+}} xmm3 = [63,0,63,0]
 ; X32-SSE-NEXT:    movdqa %xmm2, %xmm4
 ; X32-SSE-NEXT:    pandn %xmm3, %xmm4
 ; X32-SSE-NEXT:    psrlq $1, %xmm1
 ; X32-SSE-NEXT:    movdqa %xmm1, %xmm5
 ; X32-SSE-NEXT:    psrlq %xmm4, %xmm5
-; X32-SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[2,3,2,3]
+; X32-SSE-NEXT:    pxor %xmm6, %xmm6
+; X32-SSE-NEXT:    punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm6[2],xmm4[3],xmm6[3]
 ; X32-SSE-NEXT:    psrlq %xmm4, %xmm1
 ; X32-SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm5[0],xmm1[1]
 ; X32-SSE-NEXT:    pand %xmm3, %xmm2
 ; X32-SSE-NEXT:    movdqa %xmm0, %xmm3
 ; X32-SSE-NEXT:    psllq %xmm2, %xmm3
-; X32-SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,3,2,3]
+; X32-SSE-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm6[2],xmm2[3],xmm6[3]
 ; X32-SSE-NEXT:    psllq %xmm2, %xmm0
 ; X32-SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm3[0],xmm0[1]
 ; X32-SSE-NEXT:    orpd %xmm1, %xmm0
 ; X32-SSE-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
 ;
 ; X32-SSE-LABEL: constant_funnnel_v2i64:
 ; X32-SSE:       # %bb.0:
 ; X32-SSE-NEXT:    movdqa {{.*#+}} xmm2 = [63,0,63,0]
 ; X32-SSE-NEXT:    movdqa {{.*#+}} xmm3 = <4,u,14,u>
 ; X32-SSE-NEXT:    movdqa %xmm3, %xmm4
 ; X32-SSE-NEXT:    pandn %xmm2, %xmm4
-; X32-SSE-NEXT:    psrlq $1, %xmm1
 ; X32-SSE-NEXT:    movdqa %xmm1, %xmm5
+; X32-SSE-NEXT:    psrlq $1, %xmm5
 ; X32-SSE-NEXT:    psrlq %xmm4, %xmm5
-; X32-SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[2,3,2,3]
-; X32-SSE-NEXT:    psrlq %xmm4, %xmm1
+; X32-SSE-NEXT:    psrlq $50, %xmm1
 ; X32-SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm5[0],xmm1[1]
 ; X32-SSE-NEXT:    pand %xmm2, %xmm3
 ; X32-SSE-NEXT:    movdqa %xmm0, %xmm2
 ; X32-SSE-NEXT:    psllq %xmm3, %xmm2
-; X32-SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,3,2,3]
-; X32-SSE-NEXT:    psllq %xmm3, %xmm0
+; X32-SSE-NEXT:    psllq $14, %xmm0
 ; X32-SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm2[0],xmm0[1]
 ; X32-SSE-NEXT:    orpd %xmm1, %xmm0
 ; X32-SSE-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
 ;
 ; X32-SSE-LABEL: splatconstant_funnnel_v2i64:
 ; X32-SSE:       # %bb.0:
 ; X32-SSE-NEXT:    psrlq $50, %xmm1
-; X32-SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm1[0,1]
 ; X32-SSE-NEXT:    psllq $14, %xmm0
-; X32-SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm0[0,1]
-; X32-SSE-NEXT:    orpd %xmm1, %xmm0
+; X32-SSE-NEXT:    por %xmm1, %xmm0
 ; X32-SSE-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

 ; X32-SSE-LABEL: var_funnnel_v2i64:
 ; X32-SSE:       # %bb.0:
 ; X32-SSE-NEXT:    movdqa {{.*#+}} xmm2 = [63,0,63,0]
 ; X32-SSE-NEXT:    pxor %xmm3, %xmm3
 ; X32-SSE-NEXT:    psubq %xmm1, %xmm3
 ; X32-SSE-NEXT:    pand %xmm2, %xmm1
 ; X32-SSE-NEXT:    movdqa %xmm0, %xmm4
 ; X32-SSE-NEXT:    psllq %xmm1, %xmm4
-; X32-SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
-; X32-SSE-NEXT:    movdqa %xmm0, %xmm5
-; X32-SSE-NEXT:    psllq %xmm1, %xmm5
-; X32-SSE-NEXT:    movsd {{.*#+}} xmm5 = xmm4[0],xmm5[1]
+; X32-SSE-NEXT:    pxor %xmm5, %xmm5
+; X32-SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm5[2],xmm1[3],xmm5[3]
+; X32-SSE-NEXT:    movdqa %xmm0, %xmm6
+; X32-SSE-NEXT:    psllq %xmm1, %xmm6
+; X32-SSE-NEXT:    movsd {{.*#+}} xmm6 = xmm4[0],xmm6[1]
 ; X32-SSE-NEXT:    pand %xmm2, %xmm3
 ; X32-SSE-NEXT:    movdqa %xmm0, %xmm1
 ; X32-SSE-NEXT:    psrlq %xmm3, %xmm1
-; X32-SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[2,3,2,3]
-; X32-SSE-NEXT:    psrlq %xmm2, %xmm0
+; X32-SSE-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm5[2],xmm3[3],xmm5[3]
+; X32-SSE-NEXT:    psrlq %xmm3, %xmm0
 ; X32-SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
-; X32-SSE-NEXT:    orpd %xmm5, %xmm0
+; X32-SSE-NEXT:    orpd %xmm6, %xmm0
 ; X32-SSE-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:

 ; X32-SSE:       # %bb.0:
 ; X32-SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,0,1]
 ; X32-SSE-NEXT:    movdqa {{.*#+}} xmm2 = [63,0,63,0]
 ; X32-SSE-NEXT:    pxor %xmm3, %xmm3
 ; X32-SSE-NEXT:    psubq %xmm1, %xmm3
 ; X32-SSE-NEXT:    pand %xmm2, %xmm1
 ; X32-SSE-NEXT:    movdqa %xmm0, %xmm4
 ; X32-SSE-NEXT:    psllq %xmm1, %xmm4
-; X32-SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
-; X32-SSE-NEXT:    movdqa %xmm0, %xmm5
-; X32-SSE-NEXT:    psllq %xmm1, %xmm5
-; X32-SSE-NEXT:    movsd {{.*#+}} xmm5 = xmm4[0],xmm5[1]
+; X32-SSE-NEXT:    pxor %xmm5, %xmm5
+; X32-SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm5[2],xmm1[3],xmm5[3]
+; X32-SSE-NEXT:    movdqa %xmm0, %xmm6
+; X32-SSE-NEXT:    psllq %xmm1, %xmm6
+; X32-SSE-NEXT:    movsd {{.*#+}} xmm6 = xmm4[0],xmm6[1]
 ; X32-SSE-NEXT:    pand %xmm2, %xmm3
 ; X32-SSE-NEXT:    movdqa %xmm0, %xmm1
 ; X32-SSE-NEXT:    psrlq %xmm3, %xmm1
-; X32-SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[2,3,2,3]
-; X32-SSE-NEXT:    psrlq %xmm2, %xmm0
+; X32-SSE-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm5[2],xmm3[3],xmm5[3]
+; X32-SSE-NEXT:    psrlq %xmm3, %xmm0
 ; X32-SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
-; X32-SSE-NEXT:    orpd %xmm5, %xmm0
+; X32-SSE-NEXT:    orpd %xmm6, %xmm0
 ; X32-SSE-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:

 ; X32-SSE:       # %bb.0:
 ; X32-SSE-NEXT:    movdqa {{.*#+}} xmm1 = [63,0,63,0]
 ; X32-SSE-NEXT:    movdqa {{.*#+}} xmm2 = <4,u,14,u>
 ; X32-SSE-NEXT:    pxor %xmm3, %xmm3
 ; X32-SSE-NEXT:    psubq %xmm2, %xmm3
 ; X32-SSE-NEXT:    pand %xmm1, %xmm2
 ; X32-SSE-NEXT:    movdqa %xmm0, %xmm4
 ; X32-SSE-NEXT:    psllq %xmm2, %xmm4
-; X32-SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,3,2,3]
-; X32-SSE-NEXT:    movdqa %xmm0, %xmm5
-; X32-SSE-NEXT:    psllq %xmm2, %xmm5
-; X32-SSE-NEXT:    movsd {{.*#+}} xmm5 = xmm4[0],xmm5[1]
+; X32-SSE-NEXT:    movdqa %xmm0, %xmm2
+; X32-SSE-NEXT:    psllq $14, %xmm2
+; X32-SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm4[0],xmm2[1]
 ; X32-SSE-NEXT:    pand %xmm1, %xmm3
 ; X32-SSE-NEXT:    movdqa %xmm0, %xmm1
 ; X32-SSE-NEXT:    psrlq %xmm3, %xmm1
-; X32-SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[2,3,2,3]
-; X32-SSE-NEXT:    psrlq %xmm2, %xmm0
+; X32-SSE-NEXT:    psrlq $50, %xmm0
 ; X32-SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
-; X32-SSE-NEXT:    orpd %xmm5, %xmm0
+; X32-SSE-NEXT:    orpd %xmm2, %xmm0
 ; X32-SSE-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
 ;
 ; X32-SSE-LABEL: splatconstant_funnnel_v2i64:
 ; X32-SSE:       # %bb.0:
 ; X32-SSE-NEXT:    movdqa %xmm0, %xmm1
 ; X32-SSE-NEXT:    psrlq $50, %xmm1
-; X32-SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm1[0,1]
 ; X32-SSE-NEXT:    psllq $14, %xmm0
-; X32-SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm0[0,1]
-; X32-SSE-NEXT:    orpd %xmm1, %xmm0
+; X32-SSE-NEXT:    por %xmm1, %xmm0
 ; X32-SSE-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

 ;
 ; X32-SSE-LABEL: var_funnnel_v2i64:
 ; X32-SSE:       # %bb.0:
 ; X32-SSE-NEXT:    movdqa {{.*#+}} xmm3 = [63,0,63,0]
 ; X32-SSE-NEXT:    movdqa %xmm2, %xmm4
 ; X32-SSE-NEXT:    pand %xmm3, %xmm4
 ; X32-SSE-NEXT:    movdqa %xmm1, %xmm5
 ; X32-SSE-NEXT:    psrlq %xmm4, %xmm5
-; X32-SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[2,3,2,3]
+; X32-SSE-NEXT:    pxor %xmm6, %xmm6
+; X32-SSE-NEXT:    punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm6[2],xmm4[3],xmm6[3]
 ; X32-SSE-NEXT:    psrlq %xmm4, %xmm1
 ; X32-SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm5[0],xmm1[1]
 ; X32-SSE-NEXT:    pandn %xmm3, %xmm2
 ; X32-SSE-NEXT:    psllq $1, %xmm0
 ; X32-SSE-NEXT:    movdqa %xmm0, %xmm3
 ; X32-SSE-NEXT:    psllq %xmm2, %xmm3
-; X32-SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,3,2,3]
+; X32-SSE-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm6[2],xmm2[3],xmm6[3]
 ; X32-SSE-NEXT:    psllq %xmm2, %xmm0
 ; X32-SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm3[0],xmm0[1]
 ; X32-SSE-NEXT:    orpd %xmm1, %xmm0
 ; X32-SSE-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
 }
 

 ; X32-SSE-LABEL: splatvar_funnnel_v2i64:
 ; X32-SSE:       # %bb.0:
 ; X32-SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,0,1]
 ; X32-SSE-NEXT:    movdqa {{.*#+}} xmm3 = [63,0,63,0]
 ; X32-SSE-NEXT:    movdqa %xmm2, %xmm4
 ; X32-SSE-NEXT:    pand %xmm3, %xmm4
 ; X32-SSE-NEXT:    movdqa %xmm1, %xmm5
 ; X32-SSE-NEXT:    psrlq %xmm4, %xmm5
-; X32-SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[2,3,2,3]
+; X32-SSE-NEXT:    pxor %xmm6, %xmm6
+; X32-SSE-NEXT:    punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm6[2],xmm4[3],xmm6[3]
 ; X32-SSE-NEXT:    psrlq %xmm4, %xmm1
 ; X32-SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm5[0],xmm1[1]
 ; X32-SSE-NEXT:    pandn %xmm3, %xmm2
 ; X32-SSE-NEXT:    psllq $1, %xmm0
 ; X32-SSE-NEXT:    movdqa %xmm0, %xmm3
 ; X32-SSE-NEXT:    psllq %xmm2, %xmm3
-; X32-SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,3,2,3]
+; X32-SSE-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm6[2],xmm2[3],xmm6[3]
 ; X32-SSE-NEXT:    psllq %xmm2, %xmm0
 ; X32-SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm3[0],xmm0[1]
 ; X32-SSE-NEXT:    orpd %xmm1, %xmm0
 ; X32-SSE-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
 }

 ; X32-SSE-LABEL: constant_funnnel_v2i64:
 ; X32-SSE:       # %bb.0:
 ; X32-SSE-NEXT:    movdqa {{.*#+}} xmm2 = [63,0,63,0]
 ; X32-SSE-NEXT:    movdqa {{.*#+}} xmm3 = <4,u,14,u>
 ; X32-SSE-NEXT:    movdqa %xmm3, %xmm4
 ; X32-SSE-NEXT:    pand %xmm2, %xmm4
 ; X32-SSE-NEXT:    movdqa %xmm1, %xmm5
 ; X32-SSE-NEXT:    psrlq %xmm4, %xmm5
-; X32-SSE-NEXT:    pshufd {{.*#+}} xmm4 = xmm4[2,3,2,3]
-; X32-SSE-NEXT:    psrlq %xmm4, %xmm1
+; X32-SSE-NEXT:    psrlq $14, %xmm1
 ; X32-SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm5[0],xmm1[1]
 ; X32-SSE-NEXT:    pandn %xmm2, %xmm3
-; X32-SSE-NEXT:    psllq $1, %xmm0
 ; X32-SSE-NEXT:    movdqa %xmm0, %xmm2
+; X32-SSE-NEXT:    psllq $1, %xmm2
 ; X32-SSE-NEXT:    psllq %xmm3, %xmm2
-; X32-SSE-NEXT:    pshufd {{.*#+}} xmm3 = xmm3[2,3,2,3]
-; X32-SSE-NEXT:    psllq %xmm3, %xmm0
+; X32-SSE-NEXT:    psllq $50, %xmm0
 ; X32-SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm2[0],xmm0[1]
 ; X32-SSE-NEXT:    orpd %xmm1, %xmm0
 ; X32-SSE-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
 ;
 ; X32-SSE-LABEL: splatconstant_funnnel_v2i64:
 ; X32-SSE:       # %bb.0:
 ; X32-SSE-NEXT:    psrlq $14, %xmm1
-; X32-SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm1[0,1]
 ; X32-SSE-NEXT:    psllq $50, %xmm0
-; X32-SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm0[0,1]
-; X32-SSE-NEXT:    orpd %xmm1, %xmm0
+; X32-SSE-NEXT:    por %xmm1, %xmm0
 ; X32-SSE-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

 ; X32-SSE-LABEL: var_funnnel_v2i64:
 ; X32-SSE:       # %bb.0:
 ; X32-SSE-NEXT:    movdqa {{.*#+}} xmm2 = [63,0,63,0]
 ; X32-SSE-NEXT:    pxor %xmm3, %xmm3
 ; X32-SSE-NEXT:    psubq %xmm1, %xmm3
 ; X32-SSE-NEXT:    pand %xmm2, %xmm1
 ; X32-SSE-NEXT:    movdqa %xmm0, %xmm4
 ; X32-SSE-NEXT:    psrlq %xmm1, %xmm4
-; X32-SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
-; X32-SSE-NEXT:    movdqa %xmm0, %xmm5
-; X32-SSE-NEXT:    psrlq %xmm1, %xmm5
-; X32-SSE-NEXT:    movsd {{.*#+}} xmm5 = xmm4[0],xmm5[1]
+; X32-SSE-NEXT:    pxor %xmm5, %xmm5
+; X32-SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm5[2],xmm1[3],xmm5[3]
+; X32-SSE-NEXT:    movdqa %xmm0, %xmm6
+; X32-SSE-NEXT:    psrlq %xmm1, %xmm6
+; X32-SSE-NEXT:    movsd {{.*#+}} xmm6 = xmm4[0],xmm6[1]
 ; X32-SSE-NEXT:    pand %xmm2, %xmm3
 ; X32-SSE-NEXT:    movdqa %xmm0, %xmm1
 ; X32-SSE-NEXT:    psllq %xmm3, %xmm1
-; X32-SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[2,3,2,3]
-; X32-SSE-NEXT:    psllq %xmm2, %xmm0
+; X32-SSE-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm5[2],xmm3[3],xmm5[3]
+; X32-SSE-NEXT:    psllq %xmm3, %xmm0
 ; X32-SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
-; X32-SSE-NEXT:    orpd %xmm5, %xmm0
+; X32-SSE-NEXT:    orpd %xmm6, %xmm0
 ; X32-SSE-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:

 ; X32-SSE:       # %bb.0:
 ; X32-SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,1,0,1]
 ; X32-SSE-NEXT:    movdqa {{.*#+}} xmm2 = [63,0,63,0]
 ; X32-SSE-NEXT:    pxor %xmm3, %xmm3
 ; X32-SSE-NEXT:    psubq %xmm1, %xmm3
 ; X32-SSE-NEXT:    pand %xmm2, %xmm1
 ; X32-SSE-NEXT:    movdqa %xmm0, %xmm4
 ; X32-SSE-NEXT:    psrlq %xmm1, %xmm4
-; X32-SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
-; X32-SSE-NEXT:    movdqa %xmm0, %xmm5
-; X32-SSE-NEXT:    psrlq %xmm1, %xmm5
-; X32-SSE-NEXT:    movsd {{.*#+}} xmm5 = xmm4[0],xmm5[1]
+; X32-SSE-NEXT:    pxor %xmm5, %xmm5
+; X32-SSE-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm5[2],xmm1[3],xmm5[3]
+; X32-SSE-NEXT:    movdqa %xmm0, %xmm6
+; X32-SSE-NEXT:    psrlq %xmm1, %xmm6
+; X32-SSE-NEXT:    movsd {{.*#+}} xmm6 = xmm4[0],xmm6[1]
 ; X32-SSE-NEXT:    pand %xmm2, %xmm3
 ; X32-SSE-NEXT:    movdqa %xmm0, %xmm1
 ; X32-SSE-NEXT:    psllq %xmm3, %xmm1
-; X32-SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[2,3,2,3]
-; X32-SSE-NEXT:    psllq %xmm2, %xmm0
+; X32-SSE-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm5[2],xmm3[3],xmm5[3]
+; X32-SSE-NEXT:    psllq %xmm3, %xmm0
 ; X32-SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
-; X32-SSE-NEXT:    orpd %xmm5, %xmm0
+; X32-SSE-NEXT:    orpd %xmm6, %xmm0
 ; X32-SSE-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:

 ; X32-SSE:       # %bb.0:
 ; X32-SSE-NEXT:    movdqa {{.*#+}} xmm1 = [63,0,63,0]
 ; X32-SSE-NEXT:    movdqa {{.*#+}} xmm2 = <4,u,14,u>
 ; X32-SSE-NEXT:    pxor %xmm3, %xmm3
 ; X32-SSE-NEXT:    psubq %xmm2, %xmm3
 ; X32-SSE-NEXT:    pand %xmm1, %xmm2
 ; X32-SSE-NEXT:    movdqa %xmm0, %xmm4
 ; X32-SSE-NEXT:    psrlq %xmm2, %xmm4
-; X32-SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,3,2,3]
-; X32-SSE-NEXT:    movdqa %xmm0, %xmm5
-; X32-SSE-NEXT:    psrlq %xmm2, %xmm5
-; X32-SSE-NEXT:    movsd {{.*#+}} xmm5 = xmm4[0],xmm5[1]
+; X32-SSE-NEXT:    movdqa %xmm0, %xmm2
+; X32-SSE-NEXT:    psrlq $14, %xmm2
+; X32-SSE-NEXT:    movsd {{.*#+}} xmm2 = xmm4[0],xmm2[1]
 ; X32-SSE-NEXT:    pand %xmm1, %xmm3
 ; X32-SSE-NEXT:    movdqa %xmm0, %xmm1
 ; X32-SSE-NEXT:    psllq %xmm3, %xmm1
-; X32-SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm3[2,3,2,3]
-; X32-SSE-NEXT:    psllq %xmm2, %xmm0
+; X32-SSE-NEXT:    psllq $50, %xmm0
 ; X32-SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
-; X32-SSE-NEXT:    orpd %xmm5, %xmm0
+; X32-SSE-NEXT:    orpd %xmm2, %xmm0
 ; X32-SSE-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
 ;
 ; X32-SSE-LABEL: splatconstant_funnnel_v2i64:
 ; X32-SSE:       # %bb.0:
 ; X32-SSE-NEXT:    movdqa %xmm0, %xmm1
 ; X32-SSE-NEXT:    psllq $50, %xmm1
-; X32-SSE-NEXT:    movsd {{.*#+}} xmm1 = xmm1[0,1]
 ; X32-SSE-NEXT:    psrlq $14, %xmm0
-; X32-SSE-NEXT:    movsd {{.*#+}} xmm0 = xmm0[0,1]
-; X32-SSE-NEXT:    orpd %xmm1, %xmm0
+; X32-SSE-NEXT:    por %xmm1, %xmm0
 ; X32-SSE-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-shuffle-combining.ll

; AVX-NEXT:    retq
   %tmp10 = insertelement <8 x i16> %tmp9, i16 %cvt1, i32 5
   %tmp11 = insertelement <8 x i16> %tmp10, i16 %cvt2, i32 6
   %tmp12 = insertelement <8 x i16> %tmp11, i16 undef, i32 7
   %tmp13 = shufflevector <8 x i16> %tmp12, <8 x i16> undef, <8 x i32> <i32 0, i32 1, i32 10, i32 3, i32 4, i32 5, i32 6, i32 7>
   ret <8 x i16> %tmp13
 }
 
 define void @PR43024() {
-; SSE2-LABEL: PR43024:
-; SSE2:       # %bb.0:
-; SSE2-NEXT:    movaps {{.*#+}} xmm0 = [NaN,NaN,0.0E+0,0.0E+0]
-; SSE2-NEXT:    movaps %xmm0, (%rax)
-; SSE2-NEXT:    movaps %xmm0, %xmm1
-; SSE2-NEXT:    shufps {{.*#+}} xmm1 = xmm1[1,1],xmm0[1,1]
-; SSE2-NEXT:    addss %xmm0, %xmm1
-; SSE2-NEXT:    xorps %xmm0, %xmm0
-; SSE2-NEXT:    addss %xmm0, %xmm1
-; SSE2-NEXT:    addss %xmm0, %xmm1
-; SSE2-NEXT:    movss %xmm1, (%rax)
-; SSE2-NEXT:    retq
-;
-; SSSE3-LABEL: PR43024:
-; SSSE3:       # %bb.0:
-; SSSE3-NEXT:    movaps {{.*#+}} xmm0 = [NaN,NaN,0.0E+0,0.0E+0]
-; SSSE3-NEXT:    movaps %xmm0, (%rax)
-; SSSE3-NEXT:    movshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
-; SSSE3-NEXT:    addss %xmm0, %xmm1
-; SSSE3-NEXT:    xorps %xmm0, %xmm0
-; SSSE3-NEXT:    addss %xmm0, %xmm1
-; SSSE3-NEXT:    addss %xmm0, %xmm1
-; SSSE3-NEXT:    movss %xmm1, (%rax)
-; SSSE3-NEXT:    retq
-;
-; SSE41-LABEL: PR43024:
-; SSE41:       # %bb.0:
-; SSE41-NEXT:    movaps {{.*#+}} xmm0 = [NaN,NaN,0.0E+0,0.0E+0]
-; SSE41-NEXT:    movaps %xmm0, (%rax)
-; SSE41-NEXT:    movshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
-; SSE41-NEXT:    addss %xmm0, %xmm1
-; SSE41-NEXT:    xorps %xmm0, %xmm0
-; SSE41-NEXT:    addss %xmm0, %xmm1
-; SSE41-NEXT:    addss %xmm0, %xmm1
-; SSE41-NEXT:    movss %xmm1, (%rax)
-; SSE41-NEXT:    retq
+; 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:    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:    pmuludq {{.*}}(%rip), %xmm1
-; SSE-NEXT:    pmuludq {{.*}}(%rip), %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

 ; SSE-LABEL: trunc_mul_const_v16i16_v16i8:
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    pmullw {{.*}}(%rip), %xmm0
 ; SSE-NEXT:    pmullw {{.*}}(%rip), %xmm1
 ; SSE-NEXT:    movdqa {{.*#+}} xmm2 = [255,255,255,255,255,255,255,255]
 ; SSE-NEXT:    pand %xmm2, %xmm1
 ; SSE-NEXT:    pand %xmm2, %xmm0
 ; SSE-NEXT:    packuswb %xmm1, %xmm0
+; SSE-NEXT:    pand {{.*}}(%rip), %xmm0
 ; SSE-NEXT:    retq
 ;
 ; AVX1-LABEL: trunc_mul_const_v16i16_v16i8:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vpmullw {{.*}}(%rip), %xmm0, %xmm1
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; AVX1-NEXT:    vpmullw {{.*}}(%rip), %xmm0, %xmm0
 ; AVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = [255,255,255,255,255,255,255,255]

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
 }