test/CodeGen/X86/stack-folding-int-avx512.ll - llvm-project/llvm - Git at Google

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -O3 -disable-peephole -mtriple=x86_64-unknown-unknown -mattr=+avx512f,+avx512bw,+avx512dq,+avx512vbmi,+avx512cd,+avx512vpopcntdq,+avx512vnni < %s | FileCheck %s

 target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
 target triple = "x86_64-unknown-unknown"

 ; Stack reload folding tests.
 ;
 ; By including a nop call with sideeffects we can force a partial register spill of the
 ; relevant registers and check that the reload is correctly folded into the instruction.

 define <16 x i32> @stack_fold_valignd(<16 x i32> %a, <16 x i32> %b) {
 ; CHECK-LABEL: stack_fold_valignd:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    subq $56, %rsp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 64
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
 ; CHECK-NEXT:    valignd $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    # zmm0 = mem[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15],zmm0[0]
 ; CHECK-NEXT:    addq $56, %rsp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 8
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = shufflevector <16 x i32> %a, <16 x i32> %b, <16 x i32><i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16>
   ret <16 x i32> %2
 }

 define <16 x i32> @stack_fold_valignd_mask(<16 x i32> %a, <16 x i32> %b, <16 x i32>* %passthru, i16 %mask) {
 ; CHECK-LABEL: stack_fold_valignd_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    subq $56, %rsp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 64
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %esi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rdi), %zmm1
 ; CHECK-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
 ; CHECK-NEXT:    valignd $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm1 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    # zmm1 {%k1} = mem[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15],zmm0[0]
 ; CHECK-NEXT:    vmovdqa64 %zmm1, %zmm0
 ; CHECK-NEXT:    addq $56, %rsp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 8
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = shufflevector <16 x i32> %a, <16 x i32> %b, <16 x i32> <i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16>
   %3 = bitcast i16 %mask to <16 x i1>
   %4 = load <16 x i32>, <16 x i32>* %passthru
   %5 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> %4
   ret <16 x i32> %5
 }

 define <16 x i32> @stack_fold_valignd_maskz(<16 x i32> %a, <16 x i32> %b, i16 %mask) {
 ; CHECK-LABEL: stack_fold_valignd_maskz:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    subq $56, %rsp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 64
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
 ; CHECK-NEXT:    valignd $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    # zmm0 {%k1} {z} = mem[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15],zmm0[0]
 ; CHECK-NEXT:    addq $56, %rsp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 8
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = shufflevector <16 x i32> %a, <16 x i32> %b, <16 x i32> <i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16>
   %3 = bitcast i16 %mask to <16 x i1>
   %4 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> zeroinitializer
   ret <16 x i32> %4
 }

 define <8 x i64> @stack_fold_valignq(<8 x i64> %a, <8 x i64> %b) {
 ; CHECK-LABEL: stack_fold_valignq:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    subq $56, %rsp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 64
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
 ; CHECK-NEXT:    valignq $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    # zmm0 = mem[1,2,3,4,5,6,7],zmm0[0]
 ; CHECK-NEXT:    addq $56, %rsp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 8
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8>
   ret <8 x i64> %2
 }

 define <8 x i64> @stack_fold_valignq_mask(<8 x i64> %a, <8 x i64> %b, <8 x i64>* %passthru, i8 %mask) {
 ; CHECK-LABEL: stack_fold_valignq_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    subq $56, %rsp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 64
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %esi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rdi), %zmm1
 ; CHECK-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
 ; CHECK-NEXT:    valignq $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm1 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    # zmm1 {%k1} = mem[1,2,3,4,5,6,7],zmm0[0]
 ; CHECK-NEXT:    vmovdqa64 %zmm1, %zmm0
 ; CHECK-NEXT:    addq $56, %rsp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 8
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8>
   %3 = bitcast i8 %mask to <8 x i1>
   %4 = load <8 x i64>, <8 x i64>* %passthru
   %5 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> %4
   ret <8 x i64> %5
 }

 define <8 x i64> @stack_fold_valignq_maskz(<8 x i64> %a, <8 x i64> %b, i8 %mask) {
 ; CHECK-LABEL: stack_fold_valignq_maskz:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    subq $56, %rsp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 64
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
 ; CHECK-NEXT:    valignq $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    # zmm0 {%k1} {z} = mem[1,2,3,4,5,6,7],zmm0[0]
 ; CHECK-NEXT:    addq $56, %rsp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 8
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8>
   %3 = bitcast i8 %mask to <8 x i1>
   %4 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> zeroinitializer
   ret <8 x i64> %4
 }

 define <64 x i8> @stack_fold_pavgb(<64 x i8> %a0, <64 x i8> %a1) {
 ; CHECK-LABEL: stack_fold_pavgb:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpavgb {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <64 x i8> @llvm.x86.avx512.pavg.b.512(<64 x i8> %a0, <64 x i8> %a1)
   ret <64 x i8> %2
 }
 declare <64 x i8> @llvm.x86.avx512.pavg.b.512(<64 x i8>, <64 x i8>)

 define <64 x i8> @stack_fold_pavgb_commuted(<64 x i8> %a0, <64 x i8> %a1) {
 ; CHECK-LABEL: stack_fold_pavgb_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpavgb {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <64 x i8> @llvm.x86.avx512.pavg.b.512(<64 x i8> %a1, <64 x i8> %a0)
   ret <64 x i8> %2
 }

 define <64 x i8> @stack_fold_pavgb_mask(<64 x i8> %a0, <64 x i8> %a1, <64 x i8>* %a2, i64 %mask) {
 ; CHECK-LABEL: stack_fold_pavgb_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovdqa64 %zmm0, %zmm1
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovq %rsi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rdi), %zmm0
 ; CHECK-NEXT:    vpavgb {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <64 x i8> @llvm.x86.avx512.pavg.b.512(<64 x i8> %a0, <64 x i8> %a1)
   %3 = bitcast i64 %mask to <64 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %4 = load <64 x i8>, <64 x i8>* %a2
   %5 = select <64 x i1> %3, <64 x i8> %2, <64 x i8> %4
   ret <64 x i8> %5
 }

 define <64 x i8> @stack_fold_pavgb_mask_commuted(<64 x i8> %a0, <64 x i8> %a1, <64 x i8>* %a2, i64 %mask) {
 ; CHECK-LABEL: stack_fold_pavgb_mask_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovdqa64 %zmm0, %zmm1
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovq %rsi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rdi), %zmm0
 ; CHECK-NEXT:    vpavgb {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <64 x i8> @llvm.x86.avx512.pavg.b.512(<64 x i8> %a1, <64 x i8> %a0)
   %3 = bitcast i64 %mask to <64 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %4 = load <64 x i8>, <64 x i8>* %a2
   %5 = select <64 x i1> %3, <64 x i8> %2, <64 x i8> %4
   ret <64 x i8> %5
 }

 define <64 x i8> @stack_fold_pavgb_maskz(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) {
 ; CHECK-LABEL: stack_fold_pavgb_maskz:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovq %rdi, %k1
 ; CHECK-NEXT:    vpavgb {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <64 x i8> @llvm.x86.avx512.pavg.b.512(<64 x i8> %a0, <64 x i8> %a1)
   %3 = bitcast i64 %mask to <64 x i1>
   %4 = select <64 x i1> %3, <64 x i8> %2, <64 x i8> zeroinitializer
   ret <64 x i8> %4
 }

 define <64 x i8> @stack_fold_pavgb_maskz_commuted(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) {
 ; CHECK-LABEL: stack_fold_pavgb_maskz_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovq %rdi, %k1
 ; CHECK-NEXT:    vpavgb {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <64 x i8> @llvm.x86.avx512.pavg.b.512(<64 x i8> %a1, <64 x i8> %a0)
   %3 = bitcast i64 %mask to <64 x i1>
   %4 = select <64 x i1> %3, <64 x i8> %2, <64 x i8> zeroinitializer
   ret <64 x i8> %4
 }

 define <32 x i16> @stack_fold_pavgw(<32 x i16> %a0, <32 x i16> %a1) {
 ; CHECK-LABEL: stack_fold_pavgw:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpavgw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <32 x i16> @llvm.x86.avx512.pavg.w.512(<32 x i16> %a0, <32 x i16> %a1)
   ret <32 x i16> %2
 }
 declare <32 x i16> @llvm.x86.avx512.pavg.w.512(<32 x i16>, <32 x i16>)

 define <32 x i16> @stack_fold_pavgw_commuted(<32 x i16> %a0, <32 x i16> %a1) {
 ; CHECK-LABEL: stack_fold_pavgw_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpavgw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <32 x i16> @llvm.x86.avx512.pavg.w.512(<32 x i16> %a1, <32 x i16> %a0)
   ret <32 x i16> %2
 }

 define <32 x i16> @stack_fold_pavgw_mask(<32 x i16> %a0, <32 x i16> %a1, <32 x i16>* %a2, i32 %mask) {
 ; CHECK-LABEL: stack_fold_pavgw_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovdqa64 %zmm0, %zmm1
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %esi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rdi), %zmm0
 ; CHECK-NEXT:    vpavgw {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <32 x i16> @llvm.x86.avx512.pavg.w.512(<32 x i16> %a0, <32 x i16> %a1)
   %3 = bitcast i32 %mask to <32 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %4 = load <32 x i16>, <32 x i16>* %a2
   %5 = select <32 x i1> %3, <32 x i16> %2, <32 x i16> %4
   ret <32 x i16> %5
 }

 define <32 x i16> @stack_fold_pavgw_mask_commuted(<32 x i16> %a0, <32 x i16> %a1, <32 x i16>* %a2, i32 %mask) {
 ; CHECK-LABEL: stack_fold_pavgw_mask_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovdqa64 %zmm0, %zmm1
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %esi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rdi), %zmm0
 ; CHECK-NEXT:    vpavgw {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <32 x i16> @llvm.x86.avx512.pavg.w.512(<32 x i16> %a1, <32 x i16> %a0)
   %3 = bitcast i32 %mask to <32 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %4 = load <32 x i16>, <32 x i16>* %a2
   %5 = select <32 x i1> %3, <32 x i16> %2, <32 x i16> %4
   ret <32 x i16> %5
 }

 define <32 x i16> @stack_fold_pavgw_maskz(<32 x i16> %a0, <32 x i16> %a1, i32 %mask) {
 ; CHECK-LABEL: stack_fold_pavgw_maskz:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpavgw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <32 x i16> @llvm.x86.avx512.pavg.w.512(<32 x i16> %a0, <32 x i16> %a1)
   %3 = bitcast i32 %mask to <32 x i1>
   %4 = select <32 x i1> %3, <32 x i16> %2, <32 x i16> zeroinitializer
   ret <32 x i16> %4
 }

 define <32 x i16> @stack_fold_pavgw_maskz_commuted(<32 x i16> %a0, <32 x i16> %a1, i32 %mask) {
 ; CHECK-LABEL: stack_fold_pavgw_maskz_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpavgw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <32 x i16> @llvm.x86.avx512.pavg.w.512(<32 x i16> %a1, <32 x i16> %a0)
   %3 = bitcast i32 %mask to <32 x i1>
   %4 = select <32 x i1> %3, <32 x i16> %2, <32 x i16> zeroinitializer
   ret <32 x i16> %4
 }

 define <4 x i32> @stack_fold_extracti32x4(<16 x i16> %a0, <16 x i32> %a1) {
 ; CHECK-LABEL: stack_fold_extracti32x4:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
 ; CHECK-NEXT:    vextracti32x4 $3, %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Folded Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; CHECK-NEXT:    vzeroupper
 ; CHECK-NEXT:    retq
   ; zext forces execution domain
   %1 = zext <16 x i16> %a0 to <16 x i32>
   %2 = shufflevector <16 x i32> %1, <16 x i32> %a1, <4 x i32> <i32 12, i32 13, i32 14, i32 15>
   %3 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   ret <4 x i32> %2
 }

 define <2 x i64> @stack_fold_extracti64x2(<8 x i32> %a0, <8 x i64> %a1) {
 ; CHECK-LABEL: stack_fold_extracti64x2:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vpmovzxdq {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero
 ; CHECK-NEXT:    vextracti32x4 $3, %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Folded Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; CHECK-NEXT:    vzeroupper
 ; CHECK-NEXT:    retq
   ; zext forces execution domain
   %1 = zext <8 x i32> %a0 to <8 x i64>
   %2 = shufflevector <8 x i64> %1, <8 x i64> %a1, <2 x i32> <i32 6, i32 7>
   %3 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   ret <2 x i64> %2
 }

 define <8 x i32> @stack_fold_extracti32x8(<16 x i16> %a0, <16 x i32> %a1) {
 ; CHECK-LABEL: stack_fold_extracti32x8:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
 ; CHECK-NEXT:    vextracti64x4 $1, %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Folded Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; CHECK-NEXT:    retq
   ; zext forces execution domain
   %1 = zext <16 x i16> %a0 to <16 x i32>
   %2 = shufflevector <16 x i32> %1, <16 x i32> %a1, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
   %3 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   ret <8 x i32> %2
 }

 define <4 x i64> @stack_fold_extracti64x4(<8 x i32> %a0, <8 x i64> %a1) {
 ; CHECK-LABEL: stack_fold_extracti64x4:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vpmovzxdq {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero
 ; CHECK-NEXT:    vextracti64x4 $1, %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Folded Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; CHECK-NEXT:    retq
   ; zext forces execution domain
   %1 = zext <8 x i32> %a0 to <8 x i64>
   %2 = shufflevector <8 x i64> %1, <8 x i64> %a1, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
   %3 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   ret <4 x i64> %2
 }

 define <16 x i32> @stack_fold_inserti32x8(<8 x i32> %a0, <8 x i32> %a1) {
 ; CHECK-LABEL: stack_fold_inserti32x8:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; CHECK-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 32-byte Folded Reload
 ; CHECK-NEXT:    vpternlogd $255, %zmm1, %zmm1, %zmm1
 ; CHECK-NEXT:    vpsubd %zmm1, %zmm0, %zmm0
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = shufflevector <8 x i32> %a0, <8 x i32> %a1, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
   ; add forces execution domain
   %3 = add <16 x i32> %2, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
   ret <16 x i32> %3
 }

 define <8 x i64> @stack_fold_inserti64x4(<4 x i64> %a0, <4 x i64> %a1) {
 ; CHECK-LABEL: stack_fold_inserti64x4:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; CHECK-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vinserti64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 32-byte Folded Reload
 ; CHECK-NEXT:    vpternlogd $255, %zmm1, %zmm1, %zmm1
 ; CHECK-NEXT:    vpsubq %zmm1, %zmm0, %zmm0
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = shufflevector <4 x i64> %a0, <4 x i64> %a1, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
   ; add forces execution domain
   %3 = add <8 x i64> %2, <i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1>
   ret <8 x i64> %3
 }

 define <64 x i8> @stack_fold_pabsb(<64 x i8> %a0) {
 ; CHECK-LABEL: stack_fold_pabsb:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpabsb {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp sgt <64 x i8> %a0, zeroinitializer
   %3 = sub <64 x i8> zeroinitializer, %a0
   %4 = select <64 x i1> %2, <64 x i8> %a0, <64 x i8> %3
   ret <64 x i8> %4
 }

 define <64 x i8> @stack_fold_pabsb_mask(<64 x i8> %passthru, <64 x i8> %a0, i64 %mask) {
 ; CHECK-LABEL: stack_fold_pabsb_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    subq $56, %rsp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 64
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovq %rdi, %k1
 ; CHECK-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
 ; CHECK-NEXT:    vpabsb {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    addq $56, %rsp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 8
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp sgt <64 x i8> %a0, zeroinitializer
   %3 = sub <64 x i8> zeroinitializer, %a0
   %4 = select <64 x i1> %2, <64 x i8> %a0, <64 x i8> %3
   %5 = bitcast i64 %mask to <64 x i1>
   %6 = select <64 x i1> %5, <64 x i8> %4, <64 x i8> %passthru
   ret <64 x i8> %6
 }

 define <64 x i8> @stack_fold_pabsb_maskz(<64 x i8> %a0, i64 %mask) {
 ; CHECK-LABEL: stack_fold_pabsb_maskz:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovq %rdi, %k1
 ; CHECK-NEXT:    vpabsb {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp sgt <64 x i8> %a0, zeroinitializer
   %3 = sub <64 x i8> zeroinitializer, %a0
   %4 = select <64 x i1> %2, <64 x i8> %a0, <64 x i8> %3
   %5 = bitcast i64 %mask to <64 x i1>
   %6 = select <64 x i1> %5, <64 x i8> %4, <64 x i8> zeroinitializer
   ret <64 x i8> %6
 }

 define <16 x i32> @stack_fold_pabsd(<16 x i32> %a0) {
 ; CHECK-LABEL: stack_fold_pabsd:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpabsd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp sgt <16 x i32> %a0, zeroinitializer
   %3 = sub <16 x i32> zeroinitializer, %a0
   %4 = select <16 x i1> %2, <16 x i32> %a0, <16 x i32> %3
   ret <16 x i32> %4
 }

 define <16 x i32> @stack_fold_pabsd_mask(<16 x i32> %passthru, <16 x i32> %a0, i16 %mask) {
 ; CHECK-LABEL: stack_fold_pabsd_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    subq $56, %rsp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 64
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
 ; CHECK-NEXT:    vpabsd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    addq $56, %rsp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 8
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp sgt <16 x i32> %a0, zeroinitializer
   %3 = sub <16 x i32> zeroinitializer, %a0
   %4 = select <16 x i1> %2, <16 x i32> %a0, <16 x i32> %3
   %5 = bitcast i16 %mask to <16 x i1>
   %6 = select <16 x i1> %5, <16 x i32> %4, <16 x i32> %passthru
   ret <16 x i32> %6
 }

 define <16 x i32> @stack_fold_pabsd_maskz(<16 x i32> %a0, i16 %mask) {
 ; CHECK-LABEL: stack_fold_pabsd_maskz:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpabsd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp sgt <16 x i32> %a0, zeroinitializer
   %3 = sub <16 x i32> zeroinitializer, %a0
   %4 = select <16 x i1> %2, <16 x i32> %a0, <16 x i32> %3
   %5 = bitcast i16 %mask to <16 x i1>
   %6 = select <16 x i1> %5, <16 x i32> %4, <16 x i32> zeroinitializer
   ret <16 x i32> %6
 }

 define <8 x i64> @stack_fold_pabsq(<8 x i64> %a0) {
 ; CHECK-LABEL: stack_fold_pabsq:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpabsq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp sgt <8 x i64> %a0, zeroinitializer
   %3 = sub <8 x i64> zeroinitializer, %a0
   %4 = select <8 x i1> %2, <8 x i64> %a0, <8 x i64> %3
   ret <8 x i64> %4
 }

 define <8 x i64> @stack_fold_pabsq_mask(<8 x i64> %passthru, <8 x i64> %a0, i8 %mask) {
 ; CHECK-LABEL: stack_fold_pabsq_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    subq $56, %rsp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 64
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
 ; CHECK-NEXT:    vpabsq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    addq $56, %rsp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 8
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp sgt <8 x i64> %a0, zeroinitializer
   %3 = sub <8 x i64> zeroinitializer, %a0
   %4 = select <8 x i1> %2, <8 x i64> %a0, <8 x i64> %3
   %5 = bitcast i8 %mask to <8 x i1>
   %6 = select <8 x i1> %5, <8 x i64> %4, <8 x i64> %passthru
   ret <8 x i64> %6
 }

 define <8 x i64> @stack_fold_pabsq_maskz(<8 x i64> %a0, i8 %mask) {
 ; CHECK-LABEL: stack_fold_pabsq_maskz:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpabsq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp sgt <8 x i64> %a0, zeroinitializer
   %3 = sub <8 x i64> zeroinitializer, %a0
   %4 = select <8 x i1> %2, <8 x i64> %a0, <8 x i64> %3
   %5 = bitcast i8 %mask to <8 x i1>
   %6 = select <8 x i1> %5, <8 x i64> %4, <8 x i64> zeroinitializer
   ret <8 x i64> %6
 }

 define <32 x i16> @stack_fold_pabsw(<32 x i16> %a0) {
 ; CHECK-LABEL: stack_fold_pabsw:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpabsw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp sgt <32 x i16> %a0, zeroinitializer
   %3 = sub <32 x i16> zeroinitializer, %a0
   %4 = select <32 x i1> %2, <32 x i16> %a0, <32 x i16> %3
   ret <32 x i16> %4
 }

 define <32 x i16> @stack_fold_pabsw_mask(<32 x i16> %passthru, <32 x i16> %a0, i32 %mask) {
 ; CHECK-LABEL: stack_fold_pabsw_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    subq $56, %rsp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 64
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
 ; CHECK-NEXT:    vpabsw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    addq $56, %rsp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 8
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp sgt <32 x i16> %a0, zeroinitializer
   %3 = sub <32 x i16> zeroinitializer, %a0
   %4 = select <32 x i1> %2, <32 x i16> %a0, <32 x i16> %3
   %5 = bitcast i32 %mask to <32 x i1>
   %6 = select <32 x i1> %5, <32 x i16> %4, <32 x i16> %passthru
   ret <32 x i16> %6
 }

 define <32 x i16> @stack_fold_pabsw_maskz(<32 x i16> %a0, i32 %mask) {
 ; CHECK-LABEL: stack_fold_pabsw_maskz:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpabsw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp sgt <32 x i16> %a0, zeroinitializer
   %3 = sub <32 x i16> zeroinitializer, %a0
   %4 = select <32 x i1> %2, <32 x i16> %a0, <32 x i16> %3
   %5 = bitcast i32 %mask to <32 x i1>
   %6 = select <32 x i1> %5, <32 x i16> %4, <32 x i16> zeroinitializer
   ret <32 x i16> %6
 }

 define <32 x i16> @stack_fold_packssdw(<16 x i32> %a0, <16 x i32> %a1) {
 ; CHECK-LABEL: stack_fold_packssdw:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpackssdw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <32 x i16> @llvm.x86.avx512.packssdw.512(<16 x i32> %a0, <16 x i32> %a1)
   ret <32 x i16> %2
 }
 declare <32 x i16> @llvm.x86.avx512.packssdw.512(<16 x i32>, <16 x i32>) nounwind readnone

 define <64 x i8> @stack_fold_packsswb(<32 x i16> %a0, <32 x i16> %a1) {
 ; CHECK-LABEL: stack_fold_packsswb:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpacksswb {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <64 x i8> @llvm.x86.avx512.packsswb.512(<32 x i16> %a0, <32 x i16> %a1)
   ret <64 x i8> %2
 }
 declare <64 x i8> @llvm.x86.avx512.packsswb.512(<32 x i16>, <32 x i16>) nounwind readnone

 define <32 x i16> @stack_fold_packusdw(<16 x i32> %a0, <16 x i32> %a1) {
 ; CHECK-LABEL: stack_fold_packusdw:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpackusdw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <32 x i16> @llvm.x86.avx512.packusdw.512(<16 x i32> %a0, <16 x i32> %a1)
   ret <32 x i16> %2
 }
 declare <32 x i16> @llvm.x86.avx512.packusdw.512(<16 x i32>, <16 x i32>) nounwind readnone

 define <32 x i16> @stack_fold_packusdw_mask(<32 x i16>* %passthru, <16 x i32> %a0, <16 x i32> %a1, i32 %mask) {
 ; CHECK-LABEL: stack_fold_packusdw_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vmovdqa64 (%rdi), %zmm2
 ; CHECK-NEXT:    kmovd %esi, %k1
 ; CHECK-NEXT:    vpackusdw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    vmovdqa64 %zmm2, %zmm0
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = load <32 x i16>, <32 x i16>* %passthru
   %3 = call <32 x i16> @llvm.x86.avx512.packusdw.512(<16 x i32> %a0, <16 x i32> %a1)
   %4 = bitcast i32 %mask to <32 x i1>
   %5 = select <32 x i1> %4, <32 x i16> %3, <32 x i16> %2
   ret <32 x i16> %5
 }

 define <32 x i16> @stack_fold_packusdw_maskz(<16 x i32> %a0, <16 x i32> %a1, i32 %mask) {
 ; CHECK-LABEL: stack_fold_packusdw_maskz:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpackusdw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <32 x i16> @llvm.x86.avx512.packusdw.512(<16 x i32> %a0, <16 x i32> %a1)
   %3 = bitcast i32 %mask to <32 x i1>
   %4 = select <32 x i1> %3, <32 x i16> %2, <32 x i16> zeroinitializer
   ret <32 x i16> %4
 }

 define <64 x i8> @stack_fold_packuswb(<32 x i16> %a0, <32 x i16> %a1) {
 ; CHECK-LABEL: stack_fold_packuswb:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpackuswb {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <64 x i8> @llvm.x86.avx512.packuswb.512(<32 x i16> %a0, <32 x i16> %a1)
   ret <64 x i8> %2
 }
 declare <64 x i8> @llvm.x86.avx512.packuswb.512(<32 x i16>, <32 x i16>) nounwind readnone

 define <64 x i8> @stack_fold_paddb(<64 x i8> %a0, <64 x i8> %a1) {
 ; CHECK-LABEL: stack_fold_paddb:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpaddb {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = add <64 x i8> %a0, %a1
   ret <64 x i8> %2
 }

 define <64 x i8> @stack_fold_paddb_commuted(<64 x i8> %a0, <64 x i8> %a1) {
 ; CHECK-LABEL: stack_fold_paddb_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpaddb {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = add <64 x i8> %a1, %a0
   ret <64 x i8> %2
 }

 define <64 x i8> @stack_fold_paddb_mask(<64 x i8> %a0, <64 x i8> %a1, <64 x i8>* %a2, i64 %mask) {
 ; CHECK-LABEL: stack_fold_paddb_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovdqa64 %zmm0, %zmm1
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovq %rsi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rdi), %zmm0
 ; CHECK-NEXT:    vpaddb {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = add <64 x i8> %a0, %a1
   %3 = bitcast i64 %mask to <64 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %4 = load <64 x i8>, <64 x i8>* %a2
   %5 = select <64 x i1> %3, <64 x i8> %2, <64 x i8> %4
   ret <64 x i8> %5
 }

 define <64 x i8> @stack_fold_paddb_mask_commuted(<64 x i8> %a0, <64 x i8> %a1, <64 x i8>* %a2, i64 %mask) {
 ; CHECK-LABEL: stack_fold_paddb_mask_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovdqa64 %zmm0, %zmm1
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovq %rsi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rdi), %zmm0
 ; CHECK-NEXT:    vpaddb {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = add <64 x i8> %a1, %a0
   %3 = bitcast i64 %mask to <64 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %4 = load <64 x i8>, <64 x i8>* %a2
   %5 = select <64 x i1> %3, <64 x i8> %2, <64 x i8> %4
   ret <64 x i8> %5
 }

 define <64 x i8> @stack_fold_paddb_maskz(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) {
 ; CHECK-LABEL: stack_fold_paddb_maskz:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovq %rdi, %k1
 ; CHECK-NEXT:    vpaddb {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = add <64 x i8> %a0, %a1
   %3 = bitcast i64 %mask to <64 x i1>
   %4 = select <64 x i1> %3, <64 x i8> %2, <64 x i8> zeroinitializer
   ret <64 x i8> %4
 }

 define <64 x i8> @stack_fold_paddb_maskz_commuted(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) {
 ; CHECK-LABEL: stack_fold_paddb_maskz_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovq %rdi, %k1
 ; CHECK-NEXT:    vpaddb {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = add <64 x i8> %a1, %a0
   %3 = bitcast i64 %mask to <64 x i1>
   %4 = select <64 x i1> %3, <64 x i8> %2, <64 x i8> zeroinitializer
   ret <64 x i8> %4
 }

 define <16 x i32> @stack_fold_paddd(<16 x i32> %a0, <16 x i32> %a1) {
 ; CHECK-LABEL: stack_fold_paddd:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpaddd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = add <16 x i32> %a0, %a1
   ret <16 x i32> %2
 }

 define <16 x i32> @stack_fold_paddd_commuted(<16 x i32> %a0, <16 x i32> %a1) {
 ; CHECK-LABEL: stack_fold_paddd_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpaddd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = add <16 x i32> %a1, %a0
   ret <16 x i32> %2
 }

 define <16 x i32> @stack_fold_paddd_mask(<16 x i32> %a0, <16 x i32> %a1, <16 x i32>* %a2, i16 %mask) {
 ; CHECK-LABEL: stack_fold_paddd_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovdqa64 %zmm0, %zmm1
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %esi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rdi), %zmm0
 ; CHECK-NEXT:    vpaddd {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = add <16 x i32> %a0, %a1
   %3 = bitcast i16 %mask to <16 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %4 = load <16 x i32>, <16 x i32>* %a2
   %5 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> %4
   ret <16 x i32> %5
 }

 define <16 x i32> @stack_fold_paddd_mask_commuted(<16 x i32> %a0, <16 x i32> %a1, <16 x i32>* %a2, i16 %mask) {
 ; CHECK-LABEL: stack_fold_paddd_mask_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovdqa64 %zmm0, %zmm1
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %esi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rdi), %zmm0
 ; CHECK-NEXT:    vpaddd {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = add <16 x i32> %a1, %a0
   %3 = bitcast i16 %mask to <16 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %4 = load <16 x i32>, <16 x i32>* %a2
   %5 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> %4
   ret <16 x i32> %5
 }

 define <16 x i32> @stack_fold_paddd_maskz(<16 x i32> %a0, <16 x i32> %a1, i16 %mask) {
 ; CHECK-LABEL: stack_fold_paddd_maskz:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpaddd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = add <16 x i32> %a0, %a1
   %3 = bitcast i16 %mask to <16 x i1>
   %4 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> zeroinitializer
   ret <16 x i32> %4
 }

 define <16 x i32> @stack_fold_paddd_maskz_commuted(<16 x i32> %a0, <16 x i32> %a1, i16 %mask) {
 ; CHECK-LABEL: stack_fold_paddd_maskz_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpaddd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = add <16 x i32> %a1, %a0
   %3 = bitcast i16 %mask to <16 x i1>
   %4 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> zeroinitializer
   ret <16 x i32> %4
 }

 define <8 x i64> @stack_fold_paddq(<8 x i64> %a0, <8 x i64> %a1) {
 ; CHECK-LABEL: stack_fold_paddq:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpaddq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = add <8 x i64> %a0, %a1
   ret <8 x i64> %2
 }

 define <8 x i64> @stack_fold_paddq_commuted(<8 x i64> %a0, <8 x i64> %a1) {
 ; CHECK-LABEL: stack_fold_paddq_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpaddq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = add <8 x i64> %a1, %a0
   ret <8 x i64> %2
 }

 define <8 x i64> @stack_fold_paddq_mask(<8 x i64> %a0, <8 x i64> %a1, <8 x i64>* %a2, i8 %mask) {
 ; CHECK-LABEL: stack_fold_paddq_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovdqa64 %zmm0, %zmm1
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %esi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rdi), %zmm0
 ; CHECK-NEXT:    vpaddq {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = add <8 x i64> %a0, %a1
   %3 = bitcast i8 %mask to <8 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %4 = load <8 x i64>, <8 x i64>* %a2
   %5 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> %4
   ret <8 x i64> %5
 }

 define <8 x i64> @stack_fold_paddq_mask_commuted(<8 x i64> %a0, <8 x i64> %a1, <8 x i64>* %a2, i8 %mask) {
 ; CHECK-LABEL: stack_fold_paddq_mask_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovdqa64 %zmm0, %zmm1
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %esi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rdi), %zmm0
 ; CHECK-NEXT:    vpaddq {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = add <8 x i64> %a1, %a0
   %3 = bitcast i8 %mask to <8 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %4 = load <8 x i64>, <8 x i64>* %a2
   %5 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> %4
   ret <8 x i64> %5
 }

 define <8 x i64> @stack_fold_paddq_maskz(<8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
 ; CHECK-LABEL: stack_fold_paddq_maskz:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpaddq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = add <8 x i64> %a0, %a1
   %3 = bitcast i8 %mask to <8 x i1>
   %4 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> zeroinitializer
   ret <8 x i64> %4
 }

 define <8 x i64> @stack_fold_paddq_maskz_commuted(<8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
 ; CHECK-LABEL: stack_fold_paddq_maskz_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpaddq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = add <8 x i64> %a1, %a0
   %3 = bitcast i8 %mask to <8 x i1>
   %4 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> zeroinitializer
   ret <8 x i64> %4
 }

 define <64 x i8> @stack_fold_paddsb(<64 x i8> %a0, <64 x i8> %a1) {
 ; CHECK-LABEL: stack_fold_paddsb:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpaddsb {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <64 x i8> @llvm.sadd.sat.v64i8(<64 x i8> %a0, <64 x i8> %a1)
   ret <64 x i8> %2
 }

 define <64 x i8> @stack_fold_paddsb_commuted(<64 x i8> %a0, <64 x i8> %a1) {
 ; CHECK-LABEL: stack_fold_paddsb_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpaddsb {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <64 x i8> @llvm.sadd.sat.v64i8(<64 x i8> %a1, <64 x i8> %a0)
   ret <64 x i8> %2
 }

 define <64 x i8> @stack_fold_paddsb_mask(<64 x i8> %a0, <64 x i8> %a1, <64 x i8>* %a2, i64 %mask) {
 ; CHECK-LABEL: stack_fold_paddsb_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovdqa64 %zmm0, %zmm1
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovq %rsi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rdi), %zmm0
 ; CHECK-NEXT:    vpaddsb {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <64 x i8> @llvm.sadd.sat.v64i8(<64 x i8> %a0, <64 x i8> %a1)
   %3 = bitcast i64 %mask to <64 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %4 = load <64 x i8>, <64 x i8>* %a2
   %5 = select <64 x i1> %3, <64 x i8> %2, <64 x i8> %4
   ret <64 x i8> %5
 }

 define <64 x i8> @stack_fold_paddsb_mask_commuted(<64 x i8> %a0, <64 x i8> %a1, <64 x i8>* %a2, i64 %mask) {
 ; CHECK-LABEL: stack_fold_paddsb_mask_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovdqa64 %zmm0, %zmm1
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovq %rsi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rdi), %zmm0
 ; CHECK-NEXT:    vpaddsb {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <64 x i8> @llvm.sadd.sat.v64i8(<64 x i8> %a1, <64 x i8> %a0)
   %3 = bitcast i64 %mask to <64 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %4 = load <64 x i8>, <64 x i8>* %a2
   %5 = select <64 x i1> %3, <64 x i8> %2, <64 x i8> %4
   ret <64 x i8> %5
 }

 define <64 x i8> @stack_fold_paddsb_maskz(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) {
 ; CHECK-LABEL: stack_fold_paddsb_maskz:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovq %rdi, %k1
 ; CHECK-NEXT:    vpaddsb {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <64 x i8> @llvm.sadd.sat.v64i8(<64 x i8> %a0, <64 x i8> %a1)
   %3 = bitcast i64 %mask to <64 x i1>
   %4 = select <64 x i1> %3, <64 x i8> %2, <64 x i8> zeroinitializer
   ret <64 x i8> %4
 }

 define <64 x i8> @stack_fold_paddsb_maskz_commuted(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) {
 ; CHECK-LABEL: stack_fold_paddsb_maskz_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovq %rdi, %k1
 ; CHECK-NEXT:    vpaddsb {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <64 x i8> @llvm.sadd.sat.v64i8(<64 x i8> %a1, <64 x i8> %a0)
   %3 = bitcast i64 %mask to <64 x i1>
   %4 = select <64 x i1> %3, <64 x i8> %2, <64 x i8> zeroinitializer
   ret <64 x i8> %4
 }

 define <32 x i16> @stack_fold_paddsw(<32 x i16> %a0, <32 x i16> %a1) {
 ; CHECK-LABEL: stack_fold_paddsw:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpaddsw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <32 x i16> @llvm.sadd.sat.v32i16(<32 x i16> %a0, <32 x i16> %a1)
   ret <32 x i16> %2
 }

 define <32 x i16> @stack_fold_paddsw_commuted(<32 x i16> %a0, <32 x i16> %a1) {
 ; CHECK-LABEL: stack_fold_paddsw_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpaddsw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <32 x i16> @llvm.sadd.sat.v32i16(<32 x i16> %a1, <32 x i16> %a0)
   ret <32 x i16> %2
 }

 define <32 x i16> @stack_fold_paddsw_mask(<32 x i16> %a0, <32 x i16> %a1, <32 x i16>* %a2, i32 %mask) {
 ; CHECK-LABEL: stack_fold_paddsw_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovdqa64 %zmm0, %zmm1
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %esi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rdi), %zmm0
 ; CHECK-NEXT:    vpaddsw {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <32 x i16> @llvm.sadd.sat.v32i16(<32 x i16> %a0, <32 x i16> %a1)
   %3 = bitcast i32 %mask to <32 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %4 = load <32 x i16>, <32 x i16>* %a2
   %5 = select <32 x i1> %3, <32 x i16> %2, <32 x i16> %4
   ret <32 x i16> %5
 }

 define <32 x i16> @stack_fold_paddsw_mask_commuted(<32 x i16> %a0, <32 x i16> %a1, <32 x i16>* %a2, i32 %mask) {
 ; CHECK-LABEL: stack_fold_paddsw_mask_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovdqa64 %zmm0, %zmm1
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %esi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rdi), %zmm0
 ; CHECK-NEXT:    vpaddsw {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <32 x i16> @llvm.sadd.sat.v32i16(<32 x i16> %a1, <32 x i16> %a0)
   %3 = bitcast i32 %mask to <32 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %4 = load <32 x i16>, <32 x i16>* %a2
   %5 = select <32 x i1> %3, <32 x i16> %2, <32 x i16> %4
   ret <32 x i16> %5
 }

 define <32 x i16> @stack_fold_paddsw_maskz(<32 x i16> %a0, <32 x i16> %a1, i32 %mask) {
 ; CHECK-LABEL: stack_fold_paddsw_maskz:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpaddsw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <32 x i16> @llvm.sadd.sat.v32i16(<32 x i16> %a0, <32 x i16> %a1)
   %3 = bitcast i32 %mask to <32 x i1>
   %4 = select <32 x i1> %3, <32 x i16> %2, <32 x i16> zeroinitializer
   ret <32 x i16> %4
 }

 define <32 x i16> @stack_fold_paddsw_maskz_commuted(<32 x i16> %a0, <32 x i16> %a1, i32 %mask) {
 ; CHECK-LABEL: stack_fold_paddsw_maskz_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpaddsw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <32 x i16> @llvm.sadd.sat.v32i16(<32 x i16> %a1, <32 x i16> %a0)
   %3 = bitcast i32 %mask to <32 x i1>
   %4 = select <32 x i1> %3, <32 x i16> %2, <32 x i16> zeroinitializer
   ret <32 x i16> %4
 }

 define <64 x i8> @stack_fold_paddusb(<64 x i8> %a0, <64 x i8> %a1) {
 ; CHECK-LABEL: stack_fold_paddusb:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpaddusb {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <64 x i8> @llvm.uadd.sat.v64i8(<64 x i8> %a0, <64 x i8> %a1)
   ret <64 x i8> %2
 }

 define <64 x i8> @stack_fold_paddusb_commuted(<64 x i8> %a0, <64 x i8> %a1) {
 ; CHECK-LABEL: stack_fold_paddusb_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpaddusb {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <64 x i8> @llvm.uadd.sat.v64i8(<64 x i8> %a1, <64 x i8> %a0)
   ret <64 x i8> %2
 }

 define <64 x i8> @stack_fold_paddusb_mask(<64 x i8> %a0, <64 x i8> %a1, <64 x i8>* %a2, i64 %mask) {
 ; CHECK-LABEL: stack_fold_paddusb_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovdqa64 %zmm0, %zmm1
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovq %rsi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rdi), %zmm0
 ; CHECK-NEXT:    vpaddusb {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <64 x i8> @llvm.uadd.sat.v64i8(<64 x i8> %a0, <64 x i8> %a1)
   %3 = bitcast i64 %mask to <64 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %4 = load <64 x i8>, <64 x i8>* %a2
   %5 = select <64 x i1> %3, <64 x i8> %2, <64 x i8> %4
   ret <64 x i8> %5
 }

 define <64 x i8> @stack_fold_paddusb_mask_commuted(<64 x i8> %a0, <64 x i8> %a1, <64 x i8>* %a2, i64 %mask) {
 ; CHECK-LABEL: stack_fold_paddusb_mask_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovdqa64 %zmm0, %zmm1
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovq %rsi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rdi), %zmm0
 ; CHECK-NEXT:    vpaddusb {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <64 x i8> @llvm.uadd.sat.v64i8(<64 x i8> %a1, <64 x i8> %a0)
   %3 = bitcast i64 %mask to <64 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %4 = load <64 x i8>, <64 x i8>* %a2
   %5 = select <64 x i1> %3, <64 x i8> %2, <64 x i8> %4
   ret <64 x i8> %5
 }

 define <64 x i8> @stack_fold_paddusb_maskz(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) {
 ; CHECK-LABEL: stack_fold_paddusb_maskz:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovq %rdi, %k1
 ; CHECK-NEXT:    vpaddusb {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <64 x i8> @llvm.uadd.sat.v64i8(<64 x i8> %a0, <64 x i8> %a1)
   %3 = bitcast i64 %mask to <64 x i1>
   %4 = select <64 x i1> %3, <64 x i8> %2, <64 x i8> zeroinitializer
   ret <64 x i8> %4
 }

 define <64 x i8> @stack_fold_paddusb_maskz_commuted(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) {
 ; CHECK-LABEL: stack_fold_paddusb_maskz_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovq %rdi, %k1
 ; CHECK-NEXT:    vpaddusb {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <64 x i8> @llvm.uadd.sat.v64i8(<64 x i8> %a1, <64 x i8> %a0)
   %3 = bitcast i64 %mask to <64 x i1>
   %4 = select <64 x i1> %3, <64 x i8> %2, <64 x i8> zeroinitializer
   ret <64 x i8> %4
 }

 define <32 x i16> @stack_fold_paddusw(<32 x i16> %a0, <32 x i16> %a1) {
 ; CHECK-LABEL: stack_fold_paddusw:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpaddusw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <32 x i16> @llvm.uadd.sat.v32i16(<32 x i16> %a0, <32 x i16> %a1)
   ret <32 x i16> %2
 }

 define <32 x i16> @stack_fold_paddusw_commuted(<32 x i16> %a0, <32 x i16> %a1) {
 ; CHECK-LABEL: stack_fold_paddusw_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpaddusw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <32 x i16> @llvm.uadd.sat.v32i16(<32 x i16> %a1, <32 x i16> %a0)
   ret <32 x i16> %2
 }

 define <32 x i16> @stack_fold_paddusw_mask(<32 x i16> %a0, <32 x i16> %a1, <32 x i16>* %a2, i32 %mask) {
 ; CHECK-LABEL: stack_fold_paddusw_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovdqa64 %zmm0, %zmm1
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %esi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rdi), %zmm0
 ; CHECK-NEXT:    vpaddusw {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <32 x i16> @llvm.uadd.sat.v32i16(<32 x i16> %a0, <32 x i16> %a1)
   %3 = bitcast i32 %mask to <32 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %4 = load <32 x i16>, <32 x i16>* %a2
   %5 = select <32 x i1> %3, <32 x i16> %2, <32 x i16> %4
   ret <32 x i16> %5
 }

 define <32 x i16> @stack_fold_paddusw_mask_commuted(<32 x i16> %a0, <32 x i16> %a1, <32 x i16>* %a2, i32 %mask) {
 ; CHECK-LABEL: stack_fold_paddusw_mask_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovdqa64 %zmm0, %zmm1
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %esi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rdi), %zmm0
 ; CHECK-NEXT:    vpaddusw {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <32 x i16> @llvm.uadd.sat.v32i16(<32 x i16> %a1, <32 x i16> %a0)
   %3 = bitcast i32 %mask to <32 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %4 = load <32 x i16>, <32 x i16>* %a2
   %5 = select <32 x i1> %3, <32 x i16> %2, <32 x i16> %4
   ret <32 x i16> %5
 }

 define <32 x i16> @stack_fold_paddusw_maskz(<32 x i16> %a0, <32 x i16> %a1, i32 %mask) {
 ; CHECK-LABEL: stack_fold_paddusw_maskz:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpaddusw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <32 x i16> @llvm.uadd.sat.v32i16(<32 x i16> %a0, <32 x i16> %a1)
   %3 = bitcast i32 %mask to <32 x i1>
   %4 = select <32 x i1> %3, <32 x i16> %2, <32 x i16> zeroinitializer
   ret <32 x i16> %4
 }

 define <32 x i16> @stack_fold_paddusw_maskz_commuted(<32 x i16> %a0, <32 x i16> %a1, i32 %mask) {
 ; CHECK-LABEL: stack_fold_paddusw_maskz_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpaddusw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <32 x i16> @llvm.uadd.sat.v32i16(<32 x i16> %a1, <32 x i16> %a0)
   %3 = bitcast i32 %mask to <32 x i1>
   %4 = select <32 x i1> %3, <32 x i16> %2, <32 x i16> zeroinitializer
   ret <32 x i16> %4
 }

 define <32 x i16> @stack_fold_paddw(<32 x i16> %a0, <32 x i16> %a1) {
 ; CHECK-LABEL: stack_fold_paddw:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpaddw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = add <32 x i16> %a0, %a1
   ret <32 x i16> %2
 }

 define <32 x i16> @stack_fold_paddw_commuted(<32 x i16> %a0, <32 x i16> %a1) {
 ; CHECK-LABEL: stack_fold_paddw_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpaddw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = add <32 x i16> %a1, %a0
   ret <32 x i16> %2
 }

 define <32 x i16> @stack_fold_paddw_mask(<32 x i16> %a0, <32 x i16> %a1, <32 x i16>* %a2, i32 %mask) {
 ; CHECK-LABEL: stack_fold_paddw_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovdqa64 %zmm0, %zmm1
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %esi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rdi), %zmm0
 ; CHECK-NEXT:    vpaddw {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = add <32 x i16> %a0, %a1
   %3 = bitcast i32 %mask to <32 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %4 = load <32 x i16>, <32 x i16>* %a2
   %5 = select <32 x i1> %3, <32 x i16> %2, <32 x i16> %4
   ret <32 x i16> %5
 }

 define <32 x i16> @stack_fold_paddw_mask_commuted(<32 x i16> %a0, <32 x i16> %a1, <32 x i16>* %a2, i32 %mask) {
 ; CHECK-LABEL: stack_fold_paddw_mask_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovdqa64 %zmm0, %zmm1
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %esi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rdi), %zmm0
 ; CHECK-NEXT:    vpaddw {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = add <32 x i16> %a1, %a0
   %3 = bitcast i32 %mask to <32 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %4 = load <32 x i16>, <32 x i16>* %a2
   %5 = select <32 x i1> %3, <32 x i16> %2, <32 x i16> %4
   ret <32 x i16> %5
 }

 define <32 x i16> @stack_fold_paddw_maskz(<32 x i16> %a0, <32 x i16> %a1, i32 %mask) {
 ; CHECK-LABEL: stack_fold_paddw_maskz:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpaddw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = add <32 x i16> %a0, %a1
   %3 = bitcast i32 %mask to <32 x i1>
   %4 = select <32 x i1> %3, <32 x i16> %2, <32 x i16> zeroinitializer
   ret <32 x i16> %4
 }

 define <32 x i16> @stack_fold_paddw_maskz_commuted(<32 x i16> %a0, <32 x i16> %a1, i32 %mask) {
 ; CHECK-LABEL: stack_fold_paddw_maskz_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpaddw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = add <32 x i16> %a1, %a0
   %3 = bitcast i32 %mask to <32 x i1>
   %4 = select <32 x i1> %3, <32 x i16> %2, <32 x i16> zeroinitializer
   ret <32 x i16> %4
 }

 define <64 x i8> @stack_fold_palignr(<64 x i8> %a0, <64 x i8> %a1) {
 ; CHECK-LABEL: stack_fold_palignr:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    subq $56, %rsp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 64
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
 ; CHECK-NEXT:    vpalignr $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    # zmm0 = mem[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15],zmm0[0],mem[17,18,19,20,21,22,23,24,25,26,27,28,29,30,31],zmm0[16],mem[33,34,35,36,37,38,39,40,41,42,43,44,45,46,47],zmm0[32],mem[49,50,51,52,53,54,55,56,57,58,59,60,61,62,63],zmm0[48]
 ; CHECK-NEXT:    addq $56, %rsp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 8
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = shufflevector <64 x i8> %a1, <64 x i8> %a0, <64 x i32> <i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 64, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 80, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 96, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63, i32 112>
   ret <64 x i8> %2
 }

 define <64 x i8> @stack_fold_palignr_mask(<64 x i8> %a0, <64 x i8> %a1, <64 x i8>* %passthru, i64 %mask) {
 ; CHECK-LABEL: stack_fold_palignr_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    subq $56, %rsp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 64
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovq %rsi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rdi), %zmm1
 ; CHECK-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
 ; CHECK-NEXT:    vpalignr $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm1 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    # zmm1 {%k1} = mem[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15],zmm0[0],mem[17,18,19,20,21,22,23,24,25,26,27,28,29,30,31],zmm0[16],mem[33,34,35,36,37,38,39,40,41,42,43,44,45,46,47],zmm0[32],mem[49,50,51,52,53,54,55,56,57,58,59,60,61,62,63],zmm0[48]
 ; CHECK-NEXT:    vmovdqa64 %zmm1, %zmm0
 ; CHECK-NEXT:    addq $56, %rsp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 8
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = shufflevector <64 x i8> %a1, <64 x i8> %a0, <64 x i32> <i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 64, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 80, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 96, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63, i32 112>
   %3 = bitcast i64 %mask to <64 x i1>
   %4 = load <64 x i8>, <64 x i8>* %passthru
   %5 = select <64 x i1> %3, <64 x i8> %2, <64 x i8> %4
   ret <64 x i8> %5
 }

 define <64 x i8> @stack_fold_palignr_maskz(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) {
 ; CHECK-LABEL: stack_fold_palignr_maskz:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    subq $56, %rsp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 64
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovq %rdi, %k1
 ; CHECK-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
 ; CHECK-NEXT:    vpalignr $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    # zmm0 {%k1} {z} = mem[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15],zmm0[0],mem[17,18,19,20,21,22,23,24,25,26,27,28,29,30,31],zmm0[16],mem[33,34,35,36,37,38,39,40,41,42,43,44,45,46,47],zmm0[32],mem[49,50,51,52,53,54,55,56,57,58,59,60,61,62,63],zmm0[48]
 ; CHECK-NEXT:    addq $56, %rsp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 8
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = shufflevector <64 x i8> %a1, <64 x i8> %a0, <64 x i32> <i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 64, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 80, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 96, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63, i32 112>
   %3 = bitcast i64 %mask to <64 x i1>
   %4 = select <64 x i1> %3, <64 x i8> %2, <64 x i8> zeroinitializer
   ret <64 x i8> %4
 }

 define <16 x i32> @stack_fold_pandd(<16 x i32> %a0, <16 x i32> %a1) {
 ; CHECK-LABEL: stack_fold_pandd:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vandps {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = and <16 x i32> %a0, %a1
   ret <16 x i32> %2
 }

 define <16 x i32> @stack_fold_pandd_commuted(<16 x i32> %a0, <16 x i32> %a1) {
 ; CHECK-LABEL: stack_fold_pandd_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vandps {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = and <16 x i32> %a1, %a0
   ret <16 x i32> %2
 }

 define <16 x i32> @stack_fold_pandd_mask(<16 x i32> %a0, <16 x i32> %a1, <16 x i32>* %a2, i16 %mask) {
 ; CHECK-LABEL: stack_fold_pandd_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovaps %zmm0, %zmm1
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %esi, %k1
 ; CHECK-NEXT:    vmovaps (%rdi), %zmm0
 ; CHECK-NEXT:    vandps {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = and <16 x i32> %a0, %a1
   %3 = bitcast i16 %mask to <16 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %4 = load <16 x i32>, <16 x i32>* %a2
   %5 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> %4
   ret <16 x i32> %5
 }

 define <16 x i32> @stack_fold_pandd_mask_commuted(<16 x i32> %a0, <16 x i32> %a1, <16 x i32>* %a2, i16 %mask) {
 ; CHECK-LABEL: stack_fold_pandd_mask_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovaps %zmm0, %zmm1
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %esi, %k1
 ; CHECK-NEXT:    vmovaps (%rdi), %zmm0
 ; CHECK-NEXT:    vandps {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = and <16 x i32> %a1, %a0
   %3 = bitcast i16 %mask to <16 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %4 = load <16 x i32>, <16 x i32>* %a2
   %5 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> %4
   ret <16 x i32> %5
 }

 define <16 x i32> @stack_fold_pandd_maskz(<16 x i32> %a0, <16 x i32> %a1, i16 %mask) {
 ; CHECK-LABEL: stack_fold_pandd_maskz:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vandps {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = and <16 x i32> %a0, %a1
   %3 = bitcast i16 %mask to <16 x i1>
   %4 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> zeroinitializer
   ret <16 x i32> %4
 }

 define <16 x i32> @stack_fold_pandd_maskz_commuted(<16 x i32> %a0, <16 x i32> %a1, i16 %mask) {
 ; CHECK-LABEL: stack_fold_pandd_maskz_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vandps {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = and <16 x i32> %a1, %a0
   %3 = bitcast i16 %mask to <16 x i1>
   %4 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> zeroinitializer
   ret <16 x i32> %4
 }

 define <8 x i64> @stack_fold_pandq(<8 x i64> %a0, <8 x i64> %a1) {
 ; CHECK-LABEL: stack_fold_pandq:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vandps {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = and <8 x i64> %a0, %a1
   ret <8 x i64> %2
 }

 define <8 x i64> @stack_fold_pandq_commuted(<8 x i64> %a0, <8 x i64> %a1) {
 ; CHECK-LABEL: stack_fold_pandq_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vandps {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = and <8 x i64> %a1, %a0
   ret <8 x i64> %2
 }

 define <8 x i64> @stack_fold_pandq_mask(<8 x i64> %a0, <8 x i64> %a1, <8 x i64>* %a2, i8 %mask) {
 ; CHECK-LABEL: stack_fold_pandq_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovapd %zmm0, %zmm1
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %esi, %k1
 ; CHECK-NEXT:    vmovapd (%rdi), %zmm0
 ; CHECK-NEXT:    vandpd {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = and <8 x i64> %a0, %a1
   %3 = bitcast i8 %mask to <8 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %4 = load <8 x i64>, <8 x i64>* %a2
   %5 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> %4
   ret <8 x i64> %5
 }

 define <8 x i64> @stack_fold_pandq_mask_commuted(<8 x i64> %a0, <8 x i64> %a1, <8 x i64>* %a2, i8 %mask) {
 ; CHECK-LABEL: stack_fold_pandq_mask_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovapd %zmm0, %zmm1
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %esi, %k1
 ; CHECK-NEXT:    vmovapd (%rdi), %zmm0
 ; CHECK-NEXT:    vandpd {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = and <8 x i64> %a1, %a0
   %3 = bitcast i8 %mask to <8 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %4 = load <8 x i64>, <8 x i64>* %a2
   %5 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> %4
   ret <8 x i64> %5
 }

 define <8 x i64> @stack_fold_pandq_maskz(<8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
 ; CHECK-LABEL: stack_fold_pandq_maskz:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vandpd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = and <8 x i64> %a0, %a1
   %3 = bitcast i8 %mask to <8 x i1>
   %4 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> zeroinitializer
   ret <8 x i64> %4
 }

 define <8 x i64> @stack_fold_pandq_maskz_commuted(<8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
 ; CHECK-LABEL: stack_fold_pandq_maskz_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vandpd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = and <8 x i64> %a1, %a0
   %3 = bitcast i8 %mask to <8 x i1>
   %4 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> zeroinitializer
   ret <8 x i64> %4
 }

 define <16 x i32> @stack_fold_vpconflictd(<16 x i32> %a0) {
 ; CHECK-LABEL: stack_fold_vpconflictd:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpconflictd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <16 x i32> @llvm.x86.avx512.conflict.d.512(<16 x i32> %a0)
   ret <16 x i32> %2
 }
 declare <16 x i32> @llvm.x86.avx512.mask.conflict.d.512(<16 x i32>, <16 x i32>, i16) nounwind readonly

 define <8 x i64> @stack_fold_vpconflictq(<8 x i64> %a0) {
 ; CHECK-LABEL: stack_fold_vpconflictq:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpconflictq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <8 x i64> @llvm.x86.avx512.conflict.q.512(<8 x i64> %a0)
   ret <8 x i64> %2
 }
 declare <8 x i64> @llvm.x86.avx512.mask.conflict.q.512(<8 x i64>, <8 x i64>, i8) nounwind readnone

 define i64 @stack_fold_pcmpeqb(<64 x i8> %a0, <64 x i8> %a1) {
 ; CHECK-LABEL: stack_fold_pcmpeqb:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpcmpeqb {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %k0 # 64-byte Folded Reload
 ; CHECK-NEXT:    kmovq %k0, %rax
 ; CHECK-NEXT:    vzeroupper
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp eq <64 x i8> %a0, %a1
   %3 = bitcast <64 x i1> %2 to i64
   ret i64 %3
 }

 define i16 @stack_fold_pcmpeqd(<16 x i32> %a0, <16 x i32> %a1) {
 ; CHECK-LABEL: stack_fold_pcmpeqd:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpcmpeqd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %k0 # 64-byte Folded Reload
 ; CHECK-NEXT:    kmovd %k0, %eax
 ; CHECK-NEXT:    # kill: def $ax killed $ax killed $eax
 ; CHECK-NEXT:    vzeroupper
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp eq <16 x i32> %a0, %a1
   %3 = bitcast <16 x i1> %2 to i16
   ret i16 %3
 }

 define i8 @stack_fold_pcmpeqq(<8 x i64> %a0, <8 x i64> %a1) {
 ; CHECK-LABEL: stack_fold_pcmpeqq:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpcmpeqq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %k0 # 64-byte Folded Reload
 ; CHECK-NEXT:    kmovd %k0, %eax
 ; CHECK-NEXT:    # kill: def $al killed $al killed $eax
 ; CHECK-NEXT:    vzeroupper
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp eq <8 x i64> %a0, %a1
   %3 = bitcast <8 x i1> %2 to i8
   ret i8 %3
 }

 define i32 @stack_fold_pcmpeqw(<32 x i16> %a0, <32 x i16> %a1) {
 ; CHECK-LABEL: stack_fold_pcmpeqw:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpcmpeqw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %k0 # 64-byte Folded Reload
 ; CHECK-NEXT:    kmovd %k0, %eax
 ; CHECK-NEXT:    vzeroupper
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp eq <32 x i16> %a0, %a1
   %3 = bitcast <32 x i1> %2 to i32
   ret i32 %3
 }

 define <16 x i32> @stack_fold_pcmpeqd_mask(<16 x i32> %a0, <16 x i32> %a1, <16 x i32>* %a2, i16 %mask, <16 x i32> %b0, <16 x i32> %b1) {
 ; CHECK-LABEL: stack_fold_pcmpeqd_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    subq $184, %rsp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 192
 ; CHECK-NEXT:    vmovups %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovups %zmm2, (%rsp) # 64-byte Spill
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
 ; CHECK-NEXT:    vpaddd (%rdi), %zmm0, %zmm0
 ; CHECK-NEXT:    kmovd %esi, %k1
 ; CHECK-NEXT:    vpcmpeqd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %k1 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
 ; CHECK-NEXT:    vpblendmd (%rsp), %zmm0, %zmm0 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    addq $184, %rsp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 8
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   ; load and add are here to keep the operations below the side effecting block and to avoid folding the wrong load
   %2 = load <16 x i32>, <16 x i32>* %a2
   %3 = add <16 x i32> %a1, %2
   %4 = bitcast i16 %mask to <16 x i1>
   %5 = icmp eq <16 x i32> %3, %a0
   %6 = and <16 x i1> %4, %5
   %7 = select <16 x i1> %6, <16 x i32> %b0, <16 x i32> %b1
   ret <16 x i32> %7
 }

 define <16 x i32> @stack_fold_pcmpeqd_mask_commuted(<16 x i32> %a0, <16 x i32> %a1, <16 x i32>* %a2, i16 %mask, <16 x i32> %b0, <16 x i32> %b1) {
 ; CHECK-LABEL: stack_fold_pcmpeqd_mask_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    subq $184, %rsp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 192
 ; CHECK-NEXT:    vmovups %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovups %zmm2, (%rsp) # 64-byte Spill
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
 ; CHECK-NEXT:    vpaddd (%rdi), %zmm0, %zmm0
 ; CHECK-NEXT:    kmovd %esi, %k1
 ; CHECK-NEXT:    vpcmpeqd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %k1 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
 ; CHECK-NEXT:    vpblendmd (%rsp), %zmm0, %zmm0 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    addq $184, %rsp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 8
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   ; load and add are here to keep the operations below the side effecting block and to avoid folding the wrong load
   %2 = load <16 x i32>, <16 x i32>* %a2
   %3 = add <16 x i32> %a1, %2
   %4 = bitcast i16 %mask to <16 x i1>
   %5 = icmp eq <16 x i32> %a0, %3
   %6 = and <16 x i1> %4, %5
   %7 = select <16 x i1> %6, <16 x i32> %b0, <16 x i32> %b1
   ret <16 x i32> %7
 }

 define <16 x i32> @stack_fold_pcmpled_mask(<16 x i32> %a0, <16 x i32> %a1, <16 x i32>* %a2, i16 %mask, <16 x i32> %b0, <16 x i32> %b1) {
 ; CHECK-LABEL: stack_fold_pcmpled_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    subq $184, %rsp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 192
 ; CHECK-NEXT:    vmovups %zmm3, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovups %zmm2, (%rsp) # 64-byte Spill
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
 ; CHECK-NEXT:    vpaddd (%rdi), %zmm0, %zmm0
 ; CHECK-NEXT:    kmovd %esi, %k1
 ; CHECK-NEXT:    vpcmpled {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %k1 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
 ; CHECK-NEXT:    vpblendmd (%rsp), %zmm0, %zmm0 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    addq $184, %rsp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 8
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   ; load and add are here to keep the operations below the side effecting block and to avoid folding the wrong load
   %2 = load <16 x i32>, <16 x i32>* %a2
   %3 = add <16 x i32> %a1, %2
   %4 = bitcast i16 %mask to <16 x i1>
   %5 = icmp sge <16 x i32> %a0, %3
   %6 = and <16 x i1> %4, %5
   %7 = select <16 x i1> %6, <16 x i32> %b0, <16 x i32> %b1
   ret <16 x i32> %7
 }

 define i16 @stack_fold_pcmpleud(<16 x i32> %a0, <16 x i32> %a1, <16 x i32>* %a2, i16 %mask) {
 ; CHECK-LABEL: stack_fold_pcmpleud:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    subq $56, %rsp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 64
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
 ; CHECK-NEXT:    vpaddd (%rdi), %zmm0, %zmm0
 ; CHECK-NEXT:    vpcmpleud {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %k0 # 64-byte Folded Reload
 ; CHECK-NEXT:    kmovd %k0, %eax
 ; CHECK-NEXT:    andl %esi, %eax
 ; CHECK-NEXT:    # kill: def $ax killed $ax killed $eax
 ; CHECK-NEXT:    addq $56, %rsp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 8
 ; CHECK-NEXT:    vzeroupper
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = load <16 x i32>, <16 x i32>* %a2
   %3 = add <16 x i32> %a1, %2
   %4 = bitcast i16 %mask to <16 x i1>
   %5 = icmp uge <16 x i32> %a0, %3
   %6 = and <16 x i1> %5, %4
   %7 = bitcast <16 x i1> %6 to i16
   ret i16 %7
 }

 define <64 x i8> @stack_fold_permbvar(<64 x i8> %a0, <64 x i8> %a1) {
 ; CHECK-LABEL: stack_fold_permbvar:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpermb {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <64 x i8> @llvm.x86.avx512.permvar.qi.512(<64 x i8> %a1, <64 x i8> %a0)
   ret <64 x i8> %2
 }
 declare <64 x i8> @llvm.x86.avx512.permvar.qi.512(<64 x i8>, <64 x i8>) nounwind readonly

 define <64 x i8> @stack_fold_permbvar_mask(<64 x i8>* %passthru, <64 x i8> %a0, <64 x i8> %a1, i64 %mask) {
 ; CHECK-LABEL: stack_fold_permbvar_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovq %rsi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rdi), %zmm2
 ; CHECK-NEXT:    vpermb {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    vmovdqa64 %zmm2, %zmm0
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <64 x i8> @llvm.x86.avx512.permvar.qi.512(<64 x i8> %a1, <64 x i8> %a0)
   %3 = bitcast i64 %mask to <64 x i1>
   ; load needed to keep the operation from being scheduled above the asm block
   %4 = load <64 x i8>, <64 x i8>* %passthru
   %5 = select <64 x i1> %3, <64 x i8> %2, <64 x i8> %4
   ret <64 x i8> %5
 }

 define <64 x i8> @stack_fold_permbvar_maskz(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) {
 ; CHECK-LABEL: stack_fold_permbvar_maskz:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovq %rdi, %k1
 ; CHECK-NEXT:    vpermb {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <64 x i8> @llvm.x86.avx512.permvar.qi.512(<64 x i8> %a1, <64 x i8> %a0)
   %3 = bitcast i64 %mask to <64 x i1>
   %4 = select <64 x i1> %3, <64 x i8> %2, <64 x i8> zeroinitializer
   ret <64 x i8> %4
 }

 define <16 x i32> @stack_fold_permd(<16 x i32> %a0, <16 x i32> %a1) {
 ; CHECK-LABEL: stack_fold_permd:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpermd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    vpternlogd $255, %zmm1, %zmm1, %zmm1
 ; CHECK-NEXT:    vpsubd %zmm1, %zmm0, %zmm0
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <16 x i32> @llvm.x86.avx512.permvar.si.512(<16 x i32> %a1, <16 x i32> %a0)
   ; add forces execution domain
   %3 = add <16 x i32> %2, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
   ret <16 x i32> %3
 }
 declare <16 x i32> @llvm.x86.avx512.permvar.si.512(<16 x i32>, <16 x i32>) nounwind readonly

 define <64 x i8> @stack_fold_vpermi2b(<64 x i8> %x0, <64 x i8> %x1, <64 x i8> %x2) {
 ; CHECK-LABEL: stack_fold_vpermi2b:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpermi2b {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <64 x i8> @llvm.x86.avx512.vpermi2var.qi.512(<64 x i8> %x1, <64 x i8> %x0, <64 x i8> %x2)
   ret <64 x i8> %2
 }

 define <16 x i32> @stack_fold_vpermi2d(<16 x i32> %x0, <16 x i32> %x1, <16 x i32> %x2) {
 ; CHECK-LABEL: stack_fold_vpermi2d:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpermi2d {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <16 x i32> @llvm.x86.avx512.vpermi2var.d.512(<16 x i32> %x1, <16 x i32> %x0, <16 x i32> %x2)
   ret <16 x i32> %2
 }

 define <8 x i64> @stack_fold_vpermi2q(<8 x i64> %x0, <8 x i64> %x1, <8 x i64> %x2) {
 ; CHECK-LABEL: stack_fold_vpermi2q:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpermi2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <8 x i64> @llvm.x86.avx512.vpermi2var.q.512(<8 x i64> %x1, <8 x i64> %x0, <8 x i64> %x2)
   ret <8 x i64> %2
 }

 define <32 x i16> @stack_fold_vpermi2w(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x2) {
 ; CHECK-LABEL: stack_fold_vpermi2w:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpermi2w {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <32 x i16> @llvm.x86.avx512.vpermi2var.hi.512(<32 x i16> %x1, <32 x i16> %x0, <32 x i16> %x2)
   ret <32 x i16> %2
 }

 define <8 x i64> @stack_fold_permq(<8 x i64> %a0) {
 ; CHECK-LABEL: stack_fold_permq:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpermq $235, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    # zmm0 = mem[3,2,2,3,7,6,6,7]
 ; CHECK-NEXT:    vpternlogd $255, %zmm1, %zmm1, %zmm1
 ; CHECK-NEXT:    vpsubq %zmm1, %zmm0, %zmm0
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = shufflevector <8 x i64> %a0, <8 x i64> undef, <8 x i32> <i32 3, i32 2, i32 2, i32 3, i32 7, i32 6, i32 6, i32 7>
   ; add forces execution domain
   %3 = add <8 x i64> %2, <i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1>
   ret <8 x i64> %3
 }

 define <8 x i64> @stack_fold_permq_mask(<8 x i64>* %passthru, <8 x i64> %a0, i8 %mask) {
 ; CHECK-LABEL: stack_fold_permq_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %esi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rdi), %zmm0
 ; CHECK-NEXT:    vpermq $235, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    # zmm0 {%k1} = mem[3,2,2,3,7,6,6,7]
 ; CHECK-NEXT:    vpternlogd $255, %zmm1, %zmm1, %zmm1
 ; CHECK-NEXT:    vpsubq %zmm1, %zmm0, %zmm0
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = shufflevector <8 x i64> %a0, <8 x i64> undef, <8 x i32> <i32 3, i32 2, i32 2, i32 3, i32 7, i32 6, i32 6, i32 7>
   %3 = bitcast i8 %mask to <8 x i1>
   ; load needed to keep the operation from being scheduled above the asm block
   %4 = load <8 x i64>, <8 x i64>* %passthru
   %5 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> %4
   ; add forces execution domain
   %6 = add <8 x i64> %5, <i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1>
   ret <8 x i64> %6
 }

 define <8 x i64> @stack_fold_permq_maskz(<8 x i64>* %passthru, <8 x i64> %a0, i8 %mask) {
 ; CHECK-LABEL: stack_fold_permq_maskz:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %esi, %k1
 ; CHECK-NEXT:    vpermq $235, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    # zmm0 {%k1} {z} = mem[3,2,2,3,7,6,6,7]
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = shufflevector <8 x i64> %a0, <8 x i64> undef, <8 x i32> <i32 3, i32 2, i32 2, i32 3, i32 7, i32 6, i32 6, i32 7>
   %3 = bitcast i8 %mask to <8 x i1>
   %4 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> zeroinitializer
   ret <8 x i64> %4
 }

 define <8 x i64> @stack_fold_permqvar(<8 x i64> %a0, <8 x i64> %a1) {
 ; CHECK-LABEL: stack_fold_permqvar:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpermq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    vpternlogd $255, %zmm1, %zmm1, %zmm1
 ; CHECK-NEXT:    vpsubq %zmm1, %zmm0, %zmm0
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <8 x i64> @llvm.x86.avx512.permvar.di.512(<8 x i64> %a1, <8 x i64> %a0)
   ; add forces execution domain
   %3 = add <8 x i64> %2, <i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1>
   ret <8 x i64> %3
 }
 declare <8 x i64> @llvm.x86.avx512.permvar.di.512(<8 x i64>, <8 x i64>) nounwind readonly

 define <8 x i64> @stack_fold_permqvar_mask(<8 x i64>* %passthru, <8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
 ; CHECK-LABEL: stack_fold_permqvar_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %esi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rdi), %zmm1
 ; CHECK-NEXT:    vpermq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm1 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0
 ; CHECK-NEXT:    vpsubq %zmm0, %zmm1, %zmm0
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <8 x i64> @llvm.x86.avx512.permvar.di.512(<8 x i64> %a1, <8 x i64> %a0)
   %3 = bitcast i8 %mask to <8 x i1>
   ; load needed to keep the operation from being scheduled above the asm block
   %4 = load <8 x i64>, <8 x i64>* %passthru
   %5 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> %4
   ; add forces execution domain
   %6 = add <8 x i64> %5, <i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1>
   ret <8 x i64> %6
 }

 define <64 x i8> @stack_fold_vpermt2b(<64 x i8> %x0, <64 x i8> %x1, <64 x i8> %x2) {
 ; CHECK-LABEL: stack_fold_vpermt2b:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpermt2b {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <64 x i8> @llvm.x86.avx512.vpermi2var.qi.512(<64 x i8> %x0, <64 x i8> %x1, <64 x i8> %x2)
   ret <64 x i8> %2
 }
 declare <64 x i8> @llvm.x86.avx512.vpermi2var.qi.512(<64 x i8>, <64 x i8>, <64 x i8>)

 define <16 x i32> @stack_fold_vpermt2d(<16 x i32> %x0, <16 x i32> %x1, <16 x i32> %x2) {
 ; CHECK-LABEL: stack_fold_vpermt2d:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpermt2d {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <16 x i32> @llvm.x86.avx512.vpermi2var.d.512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32> %x2)
   ret <16 x i32> %2
 }
 declare <16 x i32> @llvm.x86.avx512.vpermi2var.d.512(<16 x i32>, <16 x i32>, <16 x i32>)

 define <8 x i64> @stack_fold_vpermt2q(<8 x i64> %x0, <8 x i64> %x1, <8 x i64> %x2) {
 ; CHECK-LABEL: stack_fold_vpermt2q:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpermt2q {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <8 x i64> @llvm.x86.avx512.vpermi2var.q.512(<8 x i64> %x0, <8 x i64> %x1, <8 x i64> %x2)
   ret <8 x i64> %2
 }
 declare <8 x i64> @llvm.x86.avx512.vpermi2var.q.512(<8 x i64>, <8 x i64>, <8 x i64>)

 define <32 x i16> @stack_fold_vpermt2w(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x2) {
 ; CHECK-LABEL: stack_fold_vpermt2w:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpermt2w {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <32 x i16> @llvm.x86.avx512.vpermi2var.hi.512(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x2)
   ret <32 x i16> %2
 }
 declare <32 x i16> @llvm.x86.avx512.vpermi2var.hi.512(<32 x i16>, <32 x i16>, <32 x i16>)

 define <32 x i16> @stack_fold_permwvar(<32 x i16> %a0, <32 x i16> %a1) {
 ; CHECK-LABEL: stack_fold_permwvar:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpermw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <32 x i16> @llvm.x86.avx512.permvar.hi.512(<32 x i16> %a1, <32 x i16> %a0)
   ret <32 x i16> %2
 }
 declare <32 x i16> @llvm.x86.avx512.permvar.hi.512(<32 x i16>, <32 x i16>) nounwind readonly

 define <32 x i16> @stack_fold_permwvar_mask(<32 x i16>* %passthru, <32 x i16> %a0, <32 x i16> %a1, i32 %mask) {
 ; CHECK-LABEL: stack_fold_permwvar_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %esi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rdi), %zmm2
 ; CHECK-NEXT:    vpermw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    vmovdqa64 %zmm2, %zmm0
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <32 x i16> @llvm.x86.avx512.permvar.hi.512(<32 x i16> %a1, <32 x i16> %a0)
   %3 = bitcast i32 %mask to <32 x i1>
   ; load needed to keep the operation from being scheduled above the asm block
   %4 = load <32 x i16>, <32 x i16>* %passthru
   %5 = select <32 x i1> %3, <32 x i16> %2, <32 x i16> %4
   ret <32 x i16> %5
 }

 define <32 x i16> @stack_fold_permwvar_maskz(<32 x i16> %a0, <32 x i16> %a1, i32 %mask) {
 ; CHECK-LABEL: stack_fold_permwvar_maskz:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpermw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <32 x i16> @llvm.x86.avx512.permvar.hi.512(<32 x i16> %a1, <32 x i16> %a0)
   %3 = bitcast i32 %mask to <32 x i1>
   %4 = select <32 x i1> %3, <32 x i16> %2, <32 x i16> zeroinitializer
   ret <32 x i16> %4
 }

 define i32 @stack_fold_pextrd(<4 x i32> %a0, <4 x i32> %a1) {
 ; CHECK-LABEL: stack_fold_pextrd:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    pushq %rbp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 16
 ; CHECK-NEXT:    pushq %r15
 ; CHECK-NEXT:    .cfi_def_cfa_offset 24
 ; CHECK-NEXT:    pushq %r14
 ; CHECK-NEXT:    .cfi_def_cfa_offset 32
 ; CHECK-NEXT:    pushq %r13
 ; CHECK-NEXT:    .cfi_def_cfa_offset 40
 ; CHECK-NEXT:    pushq %r12
 ; CHECK-NEXT:    .cfi_def_cfa_offset 48
 ; CHECK-NEXT:    pushq %rbx
 ; CHECK-NEXT:    .cfi_def_cfa_offset 56
 ; CHECK-NEXT:    .cfi_offset %rbx, -56
 ; CHECK-NEXT:    .cfi_offset %r12, -48
 ; CHECK-NEXT:    .cfi_offset %r13, -40
 ; CHECK-NEXT:    .cfi_offset %r14, -32
 ; CHECK-NEXT:    .cfi_offset %r15, -24
 ; CHECK-NEXT:    .cfi_offset %rbp, -16
 ; CHECK-NEXT:    vpaddd %xmm1, %xmm0, %xmm0
 ; CHECK-NEXT:    vpextrd $1, %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Folded Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    movl {{[-0-9]+}}(%r{{[sb]}}p), %eax # 4-byte Reload
 ; CHECK-NEXT:    popq %rbx
 ; CHECK-NEXT:    .cfi_def_cfa_offset 48
 ; CHECK-NEXT:    popq %r12
 ; CHECK-NEXT:    .cfi_def_cfa_offset 40
 ; CHECK-NEXT:    popq %r13
 ; CHECK-NEXT:    .cfi_def_cfa_offset 32
 ; CHECK-NEXT:    popq %r14
 ; CHECK-NEXT:    .cfi_def_cfa_offset 24
 ; CHECK-NEXT:    popq %r15
 ; CHECK-NEXT:    .cfi_def_cfa_offset 16
 ; CHECK-NEXT:    popq %rbp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 8
 ; CHECK-NEXT:    retq
   ; add forces execution domain
   %1 = add <4 x i32> %a0, %a1
   %2 = extractelement <4 x i32> %1, i32 1
   %3 = tail call <2 x i64> asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
   ret i32 %2
 }

 define i64 @stack_fold_pextrq(<2 x i64> %a0) {
 ; CHECK-LABEL: stack_fold_pextrq:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    pushq %rbp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 16
 ; CHECK-NEXT:    pushq %r15
 ; CHECK-NEXT:    .cfi_def_cfa_offset 24
 ; CHECK-NEXT:    pushq %r14
 ; CHECK-NEXT:    .cfi_def_cfa_offset 32
 ; CHECK-NEXT:    pushq %r13
 ; CHECK-NEXT:    .cfi_def_cfa_offset 40
 ; CHECK-NEXT:    pushq %r12
 ; CHECK-NEXT:    .cfi_def_cfa_offset 48
 ; CHECK-NEXT:    pushq %rbx
 ; CHECK-NEXT:    .cfi_def_cfa_offset 56
 ; CHECK-NEXT:    .cfi_offset %rbx, -56
 ; CHECK-NEXT:    .cfi_offset %r12, -48
 ; CHECK-NEXT:    .cfi_offset %r13, -40
 ; CHECK-NEXT:    .cfi_offset %r14, -32
 ; CHECK-NEXT:    .cfi_offset %r15, -24
 ; CHECK-NEXT:    .cfi_offset %rbp, -16
 ; CHECK-NEXT:    vpextrq $1, %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Folded Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
 ; CHECK-NEXT:    popq %rbx
 ; CHECK-NEXT:    .cfi_def_cfa_offset 48
 ; CHECK-NEXT:    popq %r12
 ; CHECK-NEXT:    .cfi_def_cfa_offset 40
 ; CHECK-NEXT:    popq %r13
 ; CHECK-NEXT:    .cfi_def_cfa_offset 32
 ; CHECK-NEXT:    popq %r14
 ; CHECK-NEXT:    .cfi_def_cfa_offset 24
 ; CHECK-NEXT:    popq %r15
 ; CHECK-NEXT:    .cfi_def_cfa_offset 16
 ; CHECK-NEXT:    popq %rbp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 8
 ; CHECK-NEXT:    retq
   %1 = extractelement <2 x i64> %a0, i32 1
   %2 = tail call <2 x i64> asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
   ret i64 %1
 }

 define <16 x i8> @stack_fold_pinsrb(<16 x i8> %a0, i8 %a1) {
 ; CHECK-LABEL: stack_fold_pinsrb:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    pushq %rbp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 16
 ; CHECK-NEXT:    pushq %r15
 ; CHECK-NEXT:    .cfi_def_cfa_offset 24
 ; CHECK-NEXT:    pushq %r14
 ; CHECK-NEXT:    .cfi_def_cfa_offset 32
 ; CHECK-NEXT:    pushq %r13
 ; CHECK-NEXT:    .cfi_def_cfa_offset 40
 ; CHECK-NEXT:    pushq %r12
 ; CHECK-NEXT:    .cfi_def_cfa_offset 48
 ; CHECK-NEXT:    pushq %rbx
 ; CHECK-NEXT:    .cfi_def_cfa_offset 56
 ; CHECK-NEXT:    .cfi_offset %rbx, -56
 ; CHECK-NEXT:    .cfi_offset %r12, -48
 ; CHECK-NEXT:    .cfi_offset %r13, -40
 ; CHECK-NEXT:    .cfi_offset %r14, -32
 ; CHECK-NEXT:    .cfi_offset %r15, -24
 ; CHECK-NEXT:    .cfi_offset %rbp, -16
 ; CHECK-NEXT:    movl %edi, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpinsrb $1, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 4-byte Folded Reload
 ; CHECK-NEXT:    popq %rbx
 ; CHECK-NEXT:    .cfi_def_cfa_offset 48
 ; CHECK-NEXT:    popq %r12
 ; CHECK-NEXT:    .cfi_def_cfa_offset 40
 ; CHECK-NEXT:    popq %r13
 ; CHECK-NEXT:    .cfi_def_cfa_offset 32
 ; CHECK-NEXT:    popq %r14
 ; CHECK-NEXT:    .cfi_def_cfa_offset 24
 ; CHECK-NEXT:    popq %r15
 ; CHECK-NEXT:    .cfi_def_cfa_offset 16
 ; CHECK-NEXT:    popq %rbp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 8
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
   %2 = insertelement <16 x i8> %a0, i8 %a1, i32 1
   ret <16 x i8> %2
 }

 define <4 x i32> @stack_fold_pinsrd(<4 x i32> %a0, i32 %a1) {
 ; CHECK-LABEL: stack_fold_pinsrd:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    pushq %rbp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 16
 ; CHECK-NEXT:    pushq %r15
 ; CHECK-NEXT:    .cfi_def_cfa_offset 24
 ; CHECK-NEXT:    pushq %r14
 ; CHECK-NEXT:    .cfi_def_cfa_offset 32
 ; CHECK-NEXT:    pushq %r13
 ; CHECK-NEXT:    .cfi_def_cfa_offset 40
 ; CHECK-NEXT:    pushq %r12
 ; CHECK-NEXT:    .cfi_def_cfa_offset 48
 ; CHECK-NEXT:    pushq %rbx
 ; CHECK-NEXT:    .cfi_def_cfa_offset 56
 ; CHECK-NEXT:    .cfi_offset %rbx, -56
 ; CHECK-NEXT:    .cfi_offset %r12, -48
 ; CHECK-NEXT:    .cfi_offset %r13, -40
 ; CHECK-NEXT:    .cfi_offset %r14, -32
 ; CHECK-NEXT:    .cfi_offset %r15, -24
 ; CHECK-NEXT:    .cfi_offset %rbp, -16
 ; CHECK-NEXT:    movl %edi, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpinsrd $1, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 4-byte Folded Reload
 ; CHECK-NEXT:    popq %rbx
 ; CHECK-NEXT:    .cfi_def_cfa_offset 48
 ; CHECK-NEXT:    popq %r12
 ; CHECK-NEXT:    .cfi_def_cfa_offset 40
 ; CHECK-NEXT:    popq %r13
 ; CHECK-NEXT:    .cfi_def_cfa_offset 32
 ; CHECK-NEXT:    popq %r14
 ; CHECK-NEXT:    .cfi_def_cfa_offset 24
 ; CHECK-NEXT:    popq %r15
 ; CHECK-NEXT:    .cfi_def_cfa_offset 16
 ; CHECK-NEXT:    popq %rbp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 8
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
   %2 = insertelement <4 x i32> %a0, i32 %a1, i32 1
   ret <4 x i32> %2
 }

 define <2 x i64> @stack_fold_pinsrq(<2 x i64> %a0, i64 %a1) {
 ; CHECK-LABEL: stack_fold_pinsrq:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    pushq %rbp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 16
 ; CHECK-NEXT:    pushq %r15
 ; CHECK-NEXT:    .cfi_def_cfa_offset 24
 ; CHECK-NEXT:    pushq %r14
 ; CHECK-NEXT:    .cfi_def_cfa_offset 32
 ; CHECK-NEXT:    pushq %r13
 ; CHECK-NEXT:    .cfi_def_cfa_offset 40
 ; CHECK-NEXT:    pushq %r12
 ; CHECK-NEXT:    .cfi_def_cfa_offset 48
 ; CHECK-NEXT:    pushq %rbx
 ; CHECK-NEXT:    .cfi_def_cfa_offset 56
 ; CHECK-NEXT:    .cfi_offset %rbx, -56
 ; CHECK-NEXT:    .cfi_offset %r12, -48
 ; CHECK-NEXT:    .cfi_offset %r13, -40
 ; CHECK-NEXT:    .cfi_offset %r14, -32
 ; CHECK-NEXT:    .cfi_offset %r15, -24
 ; CHECK-NEXT:    .cfi_offset %rbp, -16
 ; CHECK-NEXT:    movq %rdi, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpinsrq $1, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 8-byte Folded Reload
 ; CHECK-NEXT:    popq %rbx
 ; CHECK-NEXT:    .cfi_def_cfa_offset 48
 ; CHECK-NEXT:    popq %r12
 ; CHECK-NEXT:    .cfi_def_cfa_offset 40
 ; CHECK-NEXT:    popq %r13
 ; CHECK-NEXT:    .cfi_def_cfa_offset 32
 ; CHECK-NEXT:    popq %r14
 ; CHECK-NEXT:    .cfi_def_cfa_offset 24
 ; CHECK-NEXT:    popq %r15
 ; CHECK-NEXT:    .cfi_def_cfa_offset 16
 ; CHECK-NEXT:    popq %rbp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 8
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
   %2 = insertelement <2 x i64> %a0, i64 %a1, i32 1
   ret <2 x i64> %2
 }

 define <8 x i16> @stack_fold_pinsrw(<8 x i16> %a0, i16 %a1) {
 ; CHECK-LABEL: stack_fold_pinsrw:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    pushq %rbp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 16
 ; CHECK-NEXT:    pushq %r15
 ; CHECK-NEXT:    .cfi_def_cfa_offset 24
 ; CHECK-NEXT:    pushq %r14
 ; CHECK-NEXT:    .cfi_def_cfa_offset 32
 ; CHECK-NEXT:    pushq %r13
 ; CHECK-NEXT:    .cfi_def_cfa_offset 40
 ; CHECK-NEXT:    pushq %r12
 ; CHECK-NEXT:    .cfi_def_cfa_offset 48
 ; CHECK-NEXT:    pushq %rbx
 ; CHECK-NEXT:    .cfi_def_cfa_offset 56
 ; CHECK-NEXT:    .cfi_offset %rbx, -56
 ; CHECK-NEXT:    .cfi_offset %r12, -48
 ; CHECK-NEXT:    .cfi_offset %r13, -40
 ; CHECK-NEXT:    .cfi_offset %r14, -32
 ; CHECK-NEXT:    .cfi_offset %r15, -24
 ; CHECK-NEXT:    .cfi_offset %rbp, -16
 ; CHECK-NEXT:    movl %edi, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpinsrw $1, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 4-byte Folded Reload
 ; CHECK-NEXT:    popq %rbx
 ; CHECK-NEXT:    .cfi_def_cfa_offset 48
 ; CHECK-NEXT:    popq %r12
 ; CHECK-NEXT:    .cfi_def_cfa_offset 40
 ; CHECK-NEXT:    popq %r13
 ; CHECK-NEXT:    .cfi_def_cfa_offset 32
 ; CHECK-NEXT:    popq %r14
 ; CHECK-NEXT:    .cfi_def_cfa_offset 24
 ; CHECK-NEXT:    popq %r15
 ; CHECK-NEXT:    .cfi_def_cfa_offset 16
 ; CHECK-NEXT:    popq %rbp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 8
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{rax},~{rbx},~{rcx},~{rdx},~{rsi},~{rdi},~{rbp},~{r8},~{r9},~{r10},~{r11},~{r12},~{r13},~{r14},~{r15}"()
   %2 = insertelement <8 x i16> %a0, i16 %a1, i32 1
   ret <8 x i16> %2
 }

 define <16 x i32> @stack_fold_vplzcntd(<16 x i32> %a0) {
 ; CHECK-LABEL: stack_fold_vplzcntd:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vplzcntd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <16 x i32> @llvm.ctlz.v16i32(<16 x i32> %a0, i1 false)
   ret <16 x i32> %2
 }

 define <8 x i64> @stack_fold_vplzcntq(<8 x i64> %a0) {
 ; CHECK-LABEL: stack_fold_vplzcntq:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vplzcntq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <8 x i64> @llvm.ctlz.v8i64(<8 x i64> %a0, i1 false)
   ret <8 x i64> %2
 }

 define <32 x i16> @stack_fold_pmaddubsw_zmm(<64 x i8> %a0, <64 x i8> %a1) {
 ; CHECK-LABEL: stack_fold_pmaddubsw_zmm:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpmaddubsw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <32 x i16> @llvm.x86.avx512.pmaddubs.w.512(<64 x i8> %a0, <64 x i8> %a1)
   ret <32 x i16> %2
 }
 declare <32 x i16> @llvm.x86.avx512.pmaddubs.w.512(<64 x i8>, <64 x i8>) nounwind readnone

 define <32 x i16> @stack_fold_pmaddubsw_zmm_mask(<32 x i16>* %passthru, <64 x i8> %a0, <64 x i8> %a1, i32 %mask) {
 ; CHECK-LABEL: stack_fold_pmaddubsw_zmm_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %esi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rdi), %zmm2
 ; CHECK-NEXT:    vpmaddubsw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    vmovdqa64 %zmm2, %zmm0
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <32 x i16> @llvm.x86.avx512.pmaddubs.w.512(<64 x i8> %a0, <64 x i8> %a1)
   %3 = bitcast i32 %mask to <32 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %4 = load <32 x i16>, <32 x i16>* %passthru
   %5 = select <32 x i1> %3, <32 x i16> %2, <32 x i16> %4
   ret <32 x i16> %5
 }

 define <32 x i16> @stack_fold_pmaddubsw_zmm_maskz(<64 x i8> %a0, <64 x i8> %a1, i32 %mask) {
 ; CHECK-LABEL: stack_fold_pmaddubsw_zmm_maskz:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpmaddubsw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <32 x i16> @llvm.x86.avx512.pmaddubs.w.512(<64 x i8> %a0, <64 x i8> %a1)
   %3 = bitcast i32 %mask to <32 x i1>
   %4 = select <32 x i1> %3, <32 x i16> %2, <32 x i16> zeroinitializer
   ret <32 x i16> %4
 }

 define <16 x i32> @stack_fold_pmaddwd_zmm(<32 x i16> %a0, <32 x i16> %a1) {
 ; CHECK-LABEL: stack_fold_pmaddwd_zmm:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpmaddwd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <16 x i32> @llvm.x86.avx512.pmaddw.d.512(<32 x i16> %a0, <32 x i16> %a1)
   ret <16 x i32> %2
 }
 declare <16 x i32> @llvm.x86.avx512.pmaddw.d.512(<32 x i16>, <32 x i16>) nounwind readnone

 define <16 x i32> @stack_fold_pmaddwd_zmm_commuted(<32 x i16> %a0, <32 x i16> %a1) {
 ; CHECK-LABEL: stack_fold_pmaddwd_zmm_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpmaddwd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <16 x i32> @llvm.x86.avx512.pmaddw.d.512(<32 x i16> %a1, <32 x i16> %a0)
   ret <16 x i32> %2
 }

 define <16 x i32> @stack_fold_pmaddwd_zmm_mask(<16 x i32>* %passthru, <32 x i16> %a0, <32 x i16> %a1, i16 %mask) {
 ; CHECK-LABEL: stack_fold_pmaddwd_zmm_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %esi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rdi), %zmm2
 ; CHECK-NEXT:    vpmaddwd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    vmovdqa64 %zmm2, %zmm0
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <16 x i32> @llvm.x86.avx512.pmaddw.d.512(<32 x i16> %a0, <32 x i16> %a1)
   %3 = bitcast i16 %mask to <16 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %4 = load <16 x i32>, <16 x i32>* %passthru
   %5 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> %4
   ret <16 x i32> %5
 }

 define <16 x i32> @stack_fold_pmaddwd_zmm_mask_commuted(<16 x i32>* %passthru, <32 x i16> %a0, <32 x i16> %a1, i16 %mask) {
 ; CHECK-LABEL: stack_fold_pmaddwd_zmm_mask_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %esi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rdi), %zmm2
 ; CHECK-NEXT:    vpmaddwd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    vmovdqa64 %zmm2, %zmm0
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <16 x i32> @llvm.x86.avx512.pmaddw.d.512(<32 x i16> %a1, <32 x i16> %a0)
   %3 = bitcast i16 %mask to <16 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %4 = load <16 x i32>, <16 x i32>* %passthru
   %5 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> %4
   ret <16 x i32> %5
 }

 define <16 x i32> @stack_fold_pmaddwd_zmm_maskz(<16 x i32>* %passthru, <32 x i16> %a0, <32 x i16> %a1, i16 %mask) {
 ; CHECK-LABEL: stack_fold_pmaddwd_zmm_maskz:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %esi, %k1
 ; CHECK-NEXT:    vpmaddwd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <16 x i32> @llvm.x86.avx512.pmaddw.d.512(<32 x i16> %a0, <32 x i16> %a1)
   %3 = bitcast i16 %mask to <16 x i1>
   %4 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> zeroinitializer
   ret <16 x i32> %4
 }

 define <16 x i32> @stack_fold_pmaddwd_zmm_maskz_commuted(<16 x i32>* %passthru, <32 x i16> %a0, <32 x i16> %a1, i16 %mask) {
 ; CHECK-LABEL: stack_fold_pmaddwd_zmm_maskz_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %esi, %k1
 ; CHECK-NEXT:    vpmaddwd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <16 x i32> @llvm.x86.avx512.pmaddw.d.512(<32 x i16> %a1, <32 x i16> %a0)
   %3 = bitcast i16 %mask to <16 x i1>
   %4 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> zeroinitializer
   ret <16 x i32> %4
 }

 define <64 x i8> @stack_fold_pmaxsb(<64 x i8> %a0, <64 x i8> %a1) {
 ; CHECK-LABEL: stack_fold_pmaxsb:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpmaxsb {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp sgt <64 x i8> %a0, %a1
   %3 = select <64 x i1> %2, <64 x i8> %a0, <64 x i8> %a1
   ret <64 x i8> %3
 }

 define <64 x i8> @stack_fold_pmaxsb_commuted(<64 x i8> %a0, <64 x i8> %a1) {
 ; CHECK-LABEL: stack_fold_pmaxsb_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpmaxsb {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp sgt <64 x i8> %a1, %a0
   %3 = select <64 x i1> %2, <64 x i8> %a1, <64 x i8> %a0
   ret <64 x i8> %3
 }

 define <64 x i8> @stack_fold_pmaxsb_mask(<64 x i8> %a0, <64 x i8> %a1, i64 %mask, <64 x i8>* %passthru) {
 ; CHECK-LABEL: stack_fold_pmaxsb_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovq %rdi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rsi), %zmm2
 ; CHECK-NEXT:    vpmaxsb {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    vmovdqa64 %zmm2, %zmm0
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp sgt <64 x i8> %a0, %a1
   %3 = select <64 x i1> %2, <64 x i8> %a0, <64 x i8> %a1
   %4 = bitcast i64 %mask to <64 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %5 = load <64 x i8>, <64 x i8>* %passthru
   %6 = select <64 x i1> %4, <64 x i8> %3, <64 x i8> %5
   ret <64 x i8> %6
 }

 define <64 x i8> @stack_fold_pmaxsb_mask_commuted(<64 x i8> %a0, <64 x i8> %a1, i64 %mask, <64 x i8>* %passthru) {
 ; CHECK-LABEL: stack_fold_pmaxsb_mask_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovq %rdi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rsi), %zmm2
 ; CHECK-NEXT:    vpmaxsb {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    vmovdqa64 %zmm2, %zmm0
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp sgt <64 x i8> %a1, %a0
   %3 = select <64 x i1> %2, <64 x i8> %a1, <64 x i8> %a0
   %4 = bitcast i64 %mask to <64 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %5 = load <64 x i8>, <64 x i8>* %passthru
   %6 = select <64 x i1> %4, <64 x i8> %3, <64 x i8> %5
   ret <64 x i8> %6
 }

 define <64 x i8> @stack_fold_pmaxsb_maskz(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) {
 ; CHECK-LABEL: stack_fold_pmaxsb_maskz:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovq %rdi, %k1
 ; CHECK-NEXT:    vpmaxsb {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp sgt <64 x i8> %a0, %a1
   %3 = select <64 x i1> %2, <64 x i8> %a0, <64 x i8> %a1
   %4 = bitcast i64 %mask to <64 x i1>
   %5 = select <64 x i1> %4, <64 x i8> %3, <64 x i8> zeroinitializer
   ret <64 x i8> %5
 }

 define <64 x i8> @stack_fold_pmaxsb_maskz_commuted(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) {
 ; CHECK-LABEL: stack_fold_pmaxsb_maskz_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovq %rdi, %k1
 ; CHECK-NEXT:    vpmaxsb {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp sgt <64 x i8> %a1, %a0
   %3 = select <64 x i1> %2, <64 x i8> %a1, <64 x i8> %a0
   %4 = bitcast i64 %mask to <64 x i1>
   %5 = select <64 x i1> %4, <64 x i8> %3, <64 x i8> zeroinitializer
   ret <64 x i8> %5
 }

 define <16 x i32> @stack_fold_pmaxsd(<16 x i32> %a0, <16 x i32> %a1) {
 ; CHECK-LABEL: stack_fold_pmaxsd:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpmaxsd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp sgt <16 x i32> %a0, %a1
   %3 = select <16 x i1> %2, <16 x i32> %a0, <16 x i32> %a1
   ret <16 x i32> %3
 }

 define <16 x i32> @stack_fold_pmaxsd_commuted(<16 x i32> %a0, <16 x i32> %a1) {
 ; CHECK-LABEL: stack_fold_pmaxsd_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpmaxsd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp sgt <16 x i32> %a1, %a0
   %3 = select <16 x i1> %2, <16 x i32> %a1, <16 x i32> %a0
   ret <16 x i32> %3
 }

 define <16 x i32> @stack_fold_pmaxsd_mask(<16 x i32> %a0, <16 x i32> %a1, i16 %mask, <16 x i32>* %passthru) {
 ; CHECK-LABEL: stack_fold_pmaxsd_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rsi), %zmm2
 ; CHECK-NEXT:    vpmaxsd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    vmovdqa64 %zmm2, %zmm0
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp sgt <16 x i32> %a0, %a1
   %3 = select <16 x i1> %2, <16 x i32> %a0, <16 x i32> %a1
   %4 = bitcast i16 %mask to <16 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %5 = load <16 x i32>, <16 x i32>* %passthru
   %6 = select <16 x i1> %4, <16 x i32> %3, <16 x i32> %5
   ret <16 x i32> %6
 }

 define <16 x i32> @stack_fold_pmaxsd_mask_commuted(<16 x i32> %a0, <16 x i32> %a1, i16 %mask, <16 x i32>* %passthru) {
 ; CHECK-LABEL: stack_fold_pmaxsd_mask_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rsi), %zmm2
 ; CHECK-NEXT:    vpmaxsd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    vmovdqa64 %zmm2, %zmm0
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp sgt <16 x i32> %a1, %a0
   %3 = select <16 x i1> %2, <16 x i32> %a1, <16 x i32> %a0
   %4 = bitcast i16 %mask to <16 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %5 = load <16 x i32>, <16 x i32>* %passthru
   %6 = select <16 x i1> %4, <16 x i32> %3, <16 x i32> %5
   ret <16 x i32> %6
 }

 define <16 x i32> @stack_fold_pmaxsd_maskz(<16 x i32> %a0, <16 x i32> %a1, i16 %mask) {
 ; CHECK-LABEL: stack_fold_pmaxsd_maskz:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpmaxsd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp sgt <16 x i32> %a0, %a1
   %3 = select <16 x i1> %2, <16 x i32> %a0, <16 x i32> %a1
   %4 = bitcast i16 %mask to <16 x i1>
   %5 = select <16 x i1> %4, <16 x i32> %3, <16 x i32> zeroinitializer
   ret <16 x i32> %5
 }

 define <16 x i32> @stack_fold_pmaxsd_maskz_commuted(<16 x i32> %a0, <16 x i32> %a1, i16 %mask) {
 ; CHECK-LABEL: stack_fold_pmaxsd_maskz_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpmaxsd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp sgt <16 x i32> %a1, %a0
   %3 = select <16 x i1> %2, <16 x i32> %a1, <16 x i32> %a0
   %4 = bitcast i16 %mask to <16 x i1>
   %5 = select <16 x i1> %4, <16 x i32> %3, <16 x i32> zeroinitializer
   ret <16 x i32> %5
 }

 define <8 x i64> @stack_fold_pmaxsq(<8 x i64> %a0, <8 x i64> %a1) {
 ; CHECK-LABEL: stack_fold_pmaxsq:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpmaxsq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp sgt <8 x i64> %a0, %a1
   %3 = select <8 x i1> %2, <8 x i64> %a0, <8 x i64> %a1
   ret <8 x i64> %3
 }

 define <8 x i64> @stack_fold_pmaxsq_commuted(<8 x i64> %a0, <8 x i64> %a1) {
 ; CHECK-LABEL: stack_fold_pmaxsq_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpmaxsq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp sgt <8 x i64> %a1, %a0
   %3 = select <8 x i1> %2, <8 x i64> %a1, <8 x i64> %a0
   ret <8 x i64> %3
 }

 define <8 x i64> @stack_fold_pmaxsq_mask(<8 x i64> %a0, <8 x i64> %a1, i8 %mask, <8 x i64>* %passthru) {
 ; CHECK-LABEL: stack_fold_pmaxsq_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rsi), %zmm2
 ; CHECK-NEXT:    vpmaxsq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    vmovdqa64 %zmm2, %zmm0
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp sgt <8 x i64> %a0, %a1
   %3 = select <8 x i1> %2, <8 x i64> %a0, <8 x i64> %a1
   %4 = bitcast i8 %mask to <8 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %5 = load <8 x i64>, <8 x i64>* %passthru
   %6 = select <8 x i1> %4, <8 x i64> %3, <8 x i64> %5
   ret <8 x i64> %6
 }

 define <8 x i64> @stack_fold_pmaxsq_mask_commuted(<8 x i64> %a0, <8 x i64> %a1, i8 %mask, <8 x i64>* %passthru) {
 ; CHECK-LABEL: stack_fold_pmaxsq_mask_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rsi), %zmm2
 ; CHECK-NEXT:    vpmaxsq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    vmovdqa64 %zmm2, %zmm0
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp sgt <8 x i64> %a1, %a0
   %3 = select <8 x i1> %2, <8 x i64> %a1, <8 x i64> %a0
   %4 = bitcast i8 %mask to <8 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %5 = load <8 x i64>, <8 x i64>* %passthru
   %6 = select <8 x i1> %4, <8 x i64> %3, <8 x i64> %5
   ret <8 x i64> %6
 }

 define <8 x i64> @stack_fold_pmaxsq_maskz(<8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
 ; CHECK-LABEL: stack_fold_pmaxsq_maskz:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpmaxsq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp sgt <8 x i64> %a0, %a1
   %3 = select <8 x i1> %2, <8 x i64> %a0, <8 x i64> %a1
   %4 = bitcast i8 %mask to <8 x i1>
   %5 = select <8 x i1> %4, <8 x i64> %3, <8 x i64> zeroinitializer
   ret <8 x i64> %5
 }

 define <8 x i64> @stack_fold_pmaxsq_maskz_commuted(<8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
 ; CHECK-LABEL: stack_fold_pmaxsq_maskz_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpmaxsq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp sgt <8 x i64> %a1, %a0
   %3 = select <8 x i1> %2, <8 x i64> %a1, <8 x i64> %a0
   %4 = bitcast i8 %mask to <8 x i1>
   %5 = select <8 x i1> %4, <8 x i64> %3, <8 x i64> zeroinitializer
   ret <8 x i64> %5
 }

 define <32 x i16> @stack_fold_pmaxsw(<32 x i16> %a0, <32 x i16> %a1) {
 ; CHECK-LABEL: stack_fold_pmaxsw:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpmaxsw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp sgt <32 x i16> %a0, %a1
   %3 = select <32 x i1> %2, <32 x i16> %a0, <32 x i16> %a1
   ret <32 x i16> %3
 }

 define <32 x i16> @stack_fold_pmaxsw_commuted(<32 x i16> %a0, <32 x i16> %a1) {
 ; CHECK-LABEL: stack_fold_pmaxsw_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpmaxsw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp sgt <32 x i16> %a1, %a0
   %3 = select <32 x i1> %2, <32 x i16> %a1, <32 x i16> %a0
   ret <32 x i16> %3
 }

 define <32 x i16> @stack_fold_pmaxsw_mask(<32 x i16> %a0, <32 x i16> %a1, i32 %mask, <32 x i16>* %passthru) {
 ; CHECK-LABEL: stack_fold_pmaxsw_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rsi), %zmm2
 ; CHECK-NEXT:    vpmaxsw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    vmovdqa64 %zmm2, %zmm0
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp sgt <32 x i16> %a0, %a1
   %3 = select <32 x i1> %2, <32 x i16> %a0, <32 x i16> %a1
   %4 = bitcast i32 %mask to <32 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %5 = load <32 x i16>, <32 x i16>* %passthru
   %6 = select <32 x i1> %4, <32 x i16> %3, <32 x i16> %5
   ret <32 x i16> %6
 }

 define <32 x i16> @stack_fold_pmaxsw_mask_commuted(<32 x i16> %a0, <32 x i16> %a1, i32 %mask, <32 x i16>* %passthru) {
 ; CHECK-LABEL: stack_fold_pmaxsw_mask_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rsi), %zmm2
 ; CHECK-NEXT:    vpmaxsw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    vmovdqa64 %zmm2, %zmm0
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp sgt <32 x i16> %a1, %a0
   %3 = select <32 x i1> %2, <32 x i16> %a1, <32 x i16> %a0
   %4 = bitcast i32 %mask to <32 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %5 = load <32 x i16>, <32 x i16>* %passthru
   %6 = select <32 x i1> %4, <32 x i16> %3, <32 x i16> %5
   ret <32 x i16> %6
 }

 define <32 x i16> @stack_fold_pmaxsw_maskz(<32 x i16> %a0, <32 x i16> %a1, i32 %mask) {
 ; CHECK-LABEL: stack_fold_pmaxsw_maskz:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpmaxsw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp sgt <32 x i16> %a0, %a1
   %3 = select <32 x i1> %2, <32 x i16> %a0, <32 x i16> %a1
   %4 = bitcast i32 %mask to <32 x i1>
   %5 = select <32 x i1> %4, <32 x i16> %3, <32 x i16> zeroinitializer
   ret <32 x i16> %5
 }

 define <32 x i16> @stack_fold_pmaxsw_maskz_commuted(<32 x i16> %a0, <32 x i16> %a1, i32 %mask) {
 ; CHECK-LABEL: stack_fold_pmaxsw_maskz_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpmaxsw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp sgt <32 x i16> %a1, %a0
   %3 = select <32 x i1> %2, <32 x i16> %a1, <32 x i16> %a0
   %4 = bitcast i32 %mask to <32 x i1>
   %5 = select <32 x i1> %4, <32 x i16> %3, <32 x i16> zeroinitializer
   ret <32 x i16> %5
 }

 define <64 x i8> @stack_fold_pmaxub(<64 x i8> %a0, <64 x i8> %a1) {
 ; CHECK-LABEL: stack_fold_pmaxub:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpmaxub {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp ugt <64 x i8> %a0, %a1
   %3 = select <64 x i1> %2, <64 x i8> %a0, <64 x i8> %a1
   ret <64 x i8> %3
 }

 define <64 x i8> @stack_fold_pmaxub_commuted(<64 x i8> %a0, <64 x i8> %a1) {
 ; CHECK-LABEL: stack_fold_pmaxub_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpmaxub {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp ugt <64 x i8> %a1, %a0
   %3 = select <64 x i1> %2, <64 x i8> %a1, <64 x i8> %a0
   ret <64 x i8> %3
 }

 define <64 x i8> @stack_fold_pmaxub_mask(<64 x i8> %a0, <64 x i8> %a1, i64 %mask, <64 x i8>* %passthru) {
 ; CHECK-LABEL: stack_fold_pmaxub_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovq %rdi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rsi), %zmm2
 ; CHECK-NEXT:    vpmaxub {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    vmovdqa64 %zmm2, %zmm0
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp ugt <64 x i8> %a0, %a1
   %3 = select <64 x i1> %2, <64 x i8> %a0, <64 x i8> %a1
   %4 = bitcast i64 %mask to <64 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %5 = load <64 x i8>, <64 x i8>* %passthru
   %6 = select <64 x i1> %4, <64 x i8> %3, <64 x i8> %5
   ret <64 x i8> %6
 }

 define <64 x i8> @stack_fold_pmaxub_mask_commuted(<64 x i8> %a0, <64 x i8> %a1, i64 %mask, <64 x i8>* %passthru) {
 ; CHECK-LABEL: stack_fold_pmaxub_mask_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovq %rdi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rsi), %zmm2
 ; CHECK-NEXT:    vpmaxub {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    vmovdqa64 %zmm2, %zmm0
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp ugt <64 x i8> %a1, %a0
   %3 = select <64 x i1> %2, <64 x i8> %a1, <64 x i8> %a0
   %4 = bitcast i64 %mask to <64 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %5 = load <64 x i8>, <64 x i8>* %passthru
   %6 = select <64 x i1> %4, <64 x i8> %3, <64 x i8> %5
   ret <64 x i8> %6
 }

 define <64 x i8> @stack_fold_pmaxub_maskz(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) {
 ; CHECK-LABEL: stack_fold_pmaxub_maskz:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovq %rdi, %k1
 ; CHECK-NEXT:    vpmaxub {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp ugt <64 x i8> %a0, %a1
   %3 = select <64 x i1> %2, <64 x i8> %a0, <64 x i8> %a1
   %4 = bitcast i64 %mask to <64 x i1>
   %5 = select <64 x i1> %4, <64 x i8> %3, <64 x i8> zeroinitializer
   ret <64 x i8> %5
 }

 define <64 x i8> @stack_fold_pmaxub_maskz_commuted(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) {
 ; CHECK-LABEL: stack_fold_pmaxub_maskz_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovq %rdi, %k1
 ; CHECK-NEXT:    vpmaxub {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp ugt <64 x i8> %a1, %a0
   %3 = select <64 x i1> %2, <64 x i8> %a1, <64 x i8> %a0
   %4 = bitcast i64 %mask to <64 x i1>
   %5 = select <64 x i1> %4, <64 x i8> %3, <64 x i8> zeroinitializer
   ret <64 x i8> %5
 }

 define <16 x i32> @stack_fold_pmaxud(<16 x i32> %a0, <16 x i32> %a1) {
 ; CHECK-LABEL: stack_fold_pmaxud:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpmaxud {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp ugt <16 x i32> %a0, %a1
   %3 = select <16 x i1> %2, <16 x i32> %a0, <16 x i32> %a1
   ret <16 x i32> %3
 }

 define <16 x i32> @stack_fold_pmaxud_commuted(<16 x i32> %a0, <16 x i32> %a1) {
 ; CHECK-LABEL: stack_fold_pmaxud_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpmaxud {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp ugt <16 x i32> %a1, %a0
   %3 = select <16 x i1> %2, <16 x i32> %a1, <16 x i32> %a0
   ret <16 x i32> %3
 }

 define <16 x i32> @stack_fold_pmaxud_mask(<16 x i32> %a0, <16 x i32> %a1, i16 %mask, <16 x i32>* %passthru) {
 ; CHECK-LABEL: stack_fold_pmaxud_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rsi), %zmm2
 ; CHECK-NEXT:    vpmaxud {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    vmovdqa64 %zmm2, %zmm0
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp ugt <16 x i32> %a0, %a1
   %3 = select <16 x i1> %2, <16 x i32> %a0, <16 x i32> %a1
   %4 = bitcast i16 %mask to <16 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %5 = load <16 x i32>, <16 x i32>* %passthru
   %6 = select <16 x i1> %4, <16 x i32> %3, <16 x i32> %5
   ret <16 x i32> %6
 }

 define <16 x i32> @stack_fold_pmaxud_mask_commuted(<16 x i32> %a0, <16 x i32> %a1, i16 %mask, <16 x i32>* %passthru) {
 ; CHECK-LABEL: stack_fold_pmaxud_mask_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rsi), %zmm2
 ; CHECK-NEXT:    vpmaxud {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    vmovdqa64 %zmm2, %zmm0
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp ugt <16 x i32> %a1, %a0
   %3 = select <16 x i1> %2, <16 x i32> %a1, <16 x i32> %a0
   %4 = bitcast i16 %mask to <16 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %5 = load <16 x i32>, <16 x i32>* %passthru
   %6 = select <16 x i1> %4, <16 x i32> %3, <16 x i32> %5
   ret <16 x i32> %6
 }

 define <16 x i32> @stack_fold_pmaxud_maskz(<16 x i32> %a0, <16 x i32> %a1, i16 %mask) {
 ; CHECK-LABEL: stack_fold_pmaxud_maskz:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpmaxud {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp ugt <16 x i32> %a0, %a1
   %3 = select <16 x i1> %2, <16 x i32> %a0, <16 x i32> %a1
   %4 = bitcast i16 %mask to <16 x i1>
   %5 = select <16 x i1> %4, <16 x i32> %3, <16 x i32> zeroinitializer
   ret <16 x i32> %5
 }

 define <16 x i32> @stack_fold_pmaxud_maskz_commuted(<16 x i32> %a0, <16 x i32> %a1, i16 %mask) {
 ; CHECK-LABEL: stack_fold_pmaxud_maskz_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpmaxud {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp ugt <16 x i32> %a1, %a0
   %3 = select <16 x i1> %2, <16 x i32> %a1, <16 x i32> %a0
   %4 = bitcast i16 %mask to <16 x i1>
   %5 = select <16 x i1> %4, <16 x i32> %3, <16 x i32> zeroinitializer
   ret <16 x i32> %5
 }

 define <8 x i64> @stack_fold_pmaxuq(<8 x i64> %a0, <8 x i64> %a1) {
 ; CHECK-LABEL: stack_fold_pmaxuq:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpmaxuq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp ugt <8 x i64> %a0, %a1
   %3 = select <8 x i1> %2, <8 x i64> %a0, <8 x i64> %a1
   ret <8 x i64> %3
 }

 define <8 x i64> @stack_fold_pmaxuq_commuted(<8 x i64> %a0, <8 x i64> %a1) {
 ; CHECK-LABEL: stack_fold_pmaxuq_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpmaxuq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp ugt <8 x i64> %a1, %a0
   %3 = select <8 x i1> %2, <8 x i64> %a1, <8 x i64> %a0
   ret <8 x i64> %3
 }

 define <8 x i64> @stack_fold_pmaxuq_mask(<8 x i64> %a0, <8 x i64> %a1, i8 %mask, <8 x i64>* %passthru) {
 ; CHECK-LABEL: stack_fold_pmaxuq_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rsi), %zmm2
 ; CHECK-NEXT:    vpmaxuq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    vmovdqa64 %zmm2, %zmm0
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp ugt <8 x i64> %a0, %a1
   %3 = select <8 x i1> %2, <8 x i64> %a0, <8 x i64> %a1
   %4 = bitcast i8 %mask to <8 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %5 = load <8 x i64>, <8 x i64>* %passthru
   %6 = select <8 x i1> %4, <8 x i64> %3, <8 x i64> %5
   ret <8 x i64> %6
 }

 define <8 x i64> @stack_fold_pmaxuq_mask_commuted(<8 x i64> %a0, <8 x i64> %a1, i8 %mask, <8 x i64>* %passthru) {
 ; CHECK-LABEL: stack_fold_pmaxuq_mask_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rsi), %zmm2
 ; CHECK-NEXT:    vpmaxuq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    vmovdqa64 %zmm2, %zmm0
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp ugt <8 x i64> %a1, %a0
   %3 = select <8 x i1> %2, <8 x i64> %a1, <8 x i64> %a0
   %4 = bitcast i8 %mask to <8 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %5 = load <8 x i64>, <8 x i64>* %passthru
   %6 = select <8 x i1> %4, <8 x i64> %3, <8 x i64> %5
   ret <8 x i64> %6
 }

 define <8 x i64> @stack_fold_pmaxuq_maskz(<8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
 ; CHECK-LABEL: stack_fold_pmaxuq_maskz:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpmaxuq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp ugt <8 x i64> %a0, %a1
   %3 = select <8 x i1> %2, <8 x i64> %a0, <8 x i64> %a1
   %4 = bitcast i8 %mask to <8 x i1>
   %5 = select <8 x i1> %4, <8 x i64> %3, <8 x i64> zeroinitializer
   ret <8 x i64> %5
 }

 define <8 x i64> @stack_fold_pmaxuq_maskz_commuted(<8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
 ; CHECK-LABEL: stack_fold_pmaxuq_maskz_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpmaxuq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp ugt <8 x i64> %a1, %a0
   %3 = select <8 x i1> %2, <8 x i64> %a1, <8 x i64> %a0
   %4 = bitcast i8 %mask to <8 x i1>
   %5 = select <8 x i1> %4, <8 x i64> %3, <8 x i64> zeroinitializer
   ret <8 x i64> %5
 }

 define <32 x i16> @stack_fold_pmaxuw(<32 x i16> %a0, <32 x i16> %a1) {
 ; CHECK-LABEL: stack_fold_pmaxuw:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpmaxuw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp ugt <32 x i16> %a0, %a1
   %3 = select <32 x i1> %2, <32 x i16> %a0, <32 x i16> %a1
   ret <32 x i16> %3
 }

 define <32 x i16> @stack_fold_pmaxuw_commuted(<32 x i16> %a0, <32 x i16> %a1) {
 ; CHECK-LABEL: stack_fold_pmaxuw_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpmaxuw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp ugt <32 x i16> %a1, %a0
   %3 = select <32 x i1> %2, <32 x i16> %a1, <32 x i16> %a0
   ret <32 x i16> %3
 }

 define <32 x i16> @stack_fold_pmaxuw_mask(<32 x i16> %a0, <32 x i16> %a1, i32 %mask, <32 x i16>* %passthru) {
 ; CHECK-LABEL: stack_fold_pmaxuw_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rsi), %zmm2
 ; CHECK-NEXT:    vpmaxuw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    vmovdqa64 %zmm2, %zmm0
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp ugt <32 x i16> %a0, %a1
   %3 = select <32 x i1> %2, <32 x i16> %a0, <32 x i16> %a1
   %4 = bitcast i32 %mask to <32 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %5 = load <32 x i16>, <32 x i16>* %passthru
   %6 = select <32 x i1> %4, <32 x i16> %3, <32 x i16> %5
   ret <32 x i16> %6
 }

 define <32 x i16> @stack_fold_pmaxuw_mask_commuted(<32 x i16> %a0, <32 x i16> %a1, i32 %mask, <32 x i16>* %passthru) {
 ; CHECK-LABEL: stack_fold_pmaxuw_mask_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rsi), %zmm2
 ; CHECK-NEXT:    vpmaxuw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    vmovdqa64 %zmm2, %zmm0
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp ugt <32 x i16> %a1, %a0
   %3 = select <32 x i1> %2, <32 x i16> %a1, <32 x i16> %a0
   %4 = bitcast i32 %mask to <32 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %5 = load <32 x i16>, <32 x i16>* %passthru
   %6 = select <32 x i1> %4, <32 x i16> %3, <32 x i16> %5
   ret <32 x i16> %6
 }

 define <32 x i16> @stack_fold_pmaxuw_maskz(<32 x i16> %a0, <32 x i16> %a1, i32 %mask) {
 ; CHECK-LABEL: stack_fold_pmaxuw_maskz:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpmaxuw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp ugt <32 x i16> %a0, %a1
   %3 = select <32 x i1> %2, <32 x i16> %a0, <32 x i16> %a1
   %4 = bitcast i32 %mask to <32 x i1>
   %5 = select <32 x i1> %4, <32 x i16> %3, <32 x i16> zeroinitializer
   ret <32 x i16> %5
 }

 define <32 x i16> @stack_fold_pmaxuw_maskz_commuted(<32 x i16> %a0, <32 x i16> %a1, i32 %mask) {
 ; CHECK-LABEL: stack_fold_pmaxuw_maskz_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpmaxuw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp ugt <32 x i16> %a1, %a0
   %3 = select <32 x i1> %2, <32 x i16> %a1, <32 x i16> %a0
   %4 = bitcast i32 %mask to <32 x i1>
   %5 = select <32 x i1> %4, <32 x i16> %3, <32 x i16> zeroinitializer
   ret <32 x i16> %5
 }

 define <64 x i8> @stack_fold_pminsb(<64 x i8> %a0, <64 x i8> %a1) {
 ; CHECK-LABEL: stack_fold_pminsb:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpminsb {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp slt <64 x i8> %a0, %a1
   %3 = select <64 x i1> %2, <64 x i8> %a0, <64 x i8> %a1
   ret <64 x i8> %3
 }

 define <64 x i8> @stack_fold_pminsb_commuted(<64 x i8> %a0, <64 x i8> %a1) {
 ; CHECK-LABEL: stack_fold_pminsb_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpminsb {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp slt <64 x i8> %a1, %a0
   %3 = select <64 x i1> %2, <64 x i8> %a1, <64 x i8> %a0
   ret <64 x i8> %3
 }

 define <64 x i8> @stack_fold_pminsb_mask(<64 x i8> %a0, <64 x i8> %a1, i64 %mask, <64 x i8>* %passthru) {
 ; CHECK-LABEL: stack_fold_pminsb_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovq %rdi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rsi), %zmm2
 ; CHECK-NEXT:    vpminsb {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    vmovdqa64 %zmm2, %zmm0
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp slt <64 x i8> %a0, %a1
   %3 = select <64 x i1> %2, <64 x i8> %a0, <64 x i8> %a1
   %4 = bitcast i64 %mask to <64 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %5 = load <64 x i8>, <64 x i8>* %passthru
   %6 = select <64 x i1> %4, <64 x i8> %3, <64 x i8> %5
   ret <64 x i8> %6
 }

 define <64 x i8> @stack_fold_pminsb_mask_commuted(<64 x i8> %a0, <64 x i8> %a1, i64 %mask, <64 x i8>* %passthru) {
 ; CHECK-LABEL: stack_fold_pminsb_mask_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovq %rdi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rsi), %zmm2
 ; CHECK-NEXT:    vpminsb {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    vmovdqa64 %zmm2, %zmm0
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp slt <64 x i8> %a1, %a0
   %3 = select <64 x i1> %2, <64 x i8> %a1, <64 x i8> %a0
   %4 = bitcast i64 %mask to <64 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %5 = load <64 x i8>, <64 x i8>* %passthru
   %6 = select <64 x i1> %4, <64 x i8> %3, <64 x i8> %5
   ret <64 x i8> %6
 }

 define <64 x i8> @stack_fold_pminsb_maskz(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) {
 ; CHECK-LABEL: stack_fold_pminsb_maskz:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovq %rdi, %k1
 ; CHECK-NEXT:    vpminsb {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp slt <64 x i8> %a0, %a1
   %3 = select <64 x i1> %2, <64 x i8> %a0, <64 x i8> %a1
   %4 = bitcast i64 %mask to <64 x i1>
   %5 = select <64 x i1> %4, <64 x i8> %3, <64 x i8> zeroinitializer
   ret <64 x i8> %5
 }

 define <64 x i8> @stack_fold_pminsb_maskz_commuted(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) {
 ; CHECK-LABEL: stack_fold_pminsb_maskz_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovq %rdi, %k1
 ; CHECK-NEXT:    vpminsb {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp slt <64 x i8> %a1, %a0
   %3 = select <64 x i1> %2, <64 x i8> %a1, <64 x i8> %a0
   %4 = bitcast i64 %mask to <64 x i1>
   %5 = select <64 x i1> %4, <64 x i8> %3, <64 x i8> zeroinitializer
   ret <64 x i8> %5
 }

 define <16 x i32> @stack_fold_pminsd(<16 x i32> %a0, <16 x i32> %a1) {
 ; CHECK-LABEL: stack_fold_pminsd:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpminsd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp slt <16 x i32> %a0, %a1
   %3 = select <16 x i1> %2, <16 x i32> %a0, <16 x i32> %a1
   ret <16 x i32> %3
 }

 define <16 x i32> @stack_fold_pminsd_commuted(<16 x i32> %a0, <16 x i32> %a1) {
 ; CHECK-LABEL: stack_fold_pminsd_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpminsd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp slt <16 x i32> %a1, %a0
   %3 = select <16 x i1> %2, <16 x i32> %a1, <16 x i32> %a0
   ret <16 x i32> %3
 }

 define <16 x i32> @stack_fold_pminsd_mask(<16 x i32> %a0, <16 x i32> %a1, i16 %mask, <16 x i32>* %passthru) {
 ; CHECK-LABEL: stack_fold_pminsd_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rsi), %zmm2
 ; CHECK-NEXT:    vpminsd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    vmovdqa64 %zmm2, %zmm0
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp slt <16 x i32> %a0, %a1
   %3 = select <16 x i1> %2, <16 x i32> %a0, <16 x i32> %a1
   %4 = bitcast i16 %mask to <16 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %5 = load <16 x i32>, <16 x i32>* %passthru
   %6 = select <16 x i1> %4, <16 x i32> %3, <16 x i32> %5
   ret <16 x i32> %6
 }

 define <16 x i32> @stack_fold_pminsd_mask_commuted(<16 x i32> %a0, <16 x i32> %a1, i16 %mask, <16 x i32>* %passthru) {
 ; CHECK-LABEL: stack_fold_pminsd_mask_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rsi), %zmm2
 ; CHECK-NEXT:    vpminsd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    vmovdqa64 %zmm2, %zmm0
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp slt <16 x i32> %a1, %a0
   %3 = select <16 x i1> %2, <16 x i32> %a1, <16 x i32> %a0
   %4 = bitcast i16 %mask to <16 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %5 = load <16 x i32>, <16 x i32>* %passthru
   %6 = select <16 x i1> %4, <16 x i32> %3, <16 x i32> %5
   ret <16 x i32> %6
 }

 define <16 x i32> @stack_fold_pminsd_maskz(<16 x i32> %a0, <16 x i32> %a1, i16 %mask) {
 ; CHECK-LABEL: stack_fold_pminsd_maskz:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpminsd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp slt <16 x i32> %a0, %a1
   %3 = select <16 x i1> %2, <16 x i32> %a0, <16 x i32> %a1
   %4 = bitcast i16 %mask to <16 x i1>
   %5 = select <16 x i1> %4, <16 x i32> %3, <16 x i32> zeroinitializer
   ret <16 x i32> %5
 }

 define <16 x i32> @stack_fold_pminsd_maskz_commuted(<16 x i32> %a0, <16 x i32> %a1, i16 %mask) {
 ; CHECK-LABEL: stack_fold_pminsd_maskz_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpminsd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp slt <16 x i32> %a1, %a0
   %3 = select <16 x i1> %2, <16 x i32> %a1, <16 x i32> %a0
   %4 = bitcast i16 %mask to <16 x i1>
   %5 = select <16 x i1> %4, <16 x i32> %3, <16 x i32> zeroinitializer
   ret <16 x i32> %5
 }

 define <8 x i64> @stack_fold_pminsq(<8 x i64> %a0, <8 x i64> %a1) {
 ; CHECK-LABEL: stack_fold_pminsq:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpminsq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp slt <8 x i64> %a0, %a1
   %3 = select <8 x i1> %2, <8 x i64> %a0, <8 x i64> %a1
   ret <8 x i64> %3
 }

 define <8 x i64> @stack_fold_pminsq_commuted(<8 x i64> %a0, <8 x i64> %a1) {
 ; CHECK-LABEL: stack_fold_pminsq_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpminsq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp slt <8 x i64> %a1, %a0
   %3 = select <8 x i1> %2, <8 x i64> %a1, <8 x i64> %a0
   ret <8 x i64> %3
 }

 define <8 x i64> @stack_fold_pminsq_mask(<8 x i64> %a0, <8 x i64> %a1, i8 %mask, <8 x i64>* %passthru) {
 ; CHECK-LABEL: stack_fold_pminsq_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rsi), %zmm2
 ; CHECK-NEXT:    vpminsq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    vmovdqa64 %zmm2, %zmm0
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp slt <8 x i64> %a0, %a1
   %3 = select <8 x i1> %2, <8 x i64> %a0, <8 x i64> %a1
   %4 = bitcast i8 %mask to <8 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %5 = load <8 x i64>, <8 x i64>* %passthru
   %6 = select <8 x i1> %4, <8 x i64> %3, <8 x i64> %5
   ret <8 x i64> %6
 }

 define <8 x i64> @stack_fold_pminsq_mask_commuted(<8 x i64> %a0, <8 x i64> %a1, i8 %mask, <8 x i64>* %passthru) {
 ; CHECK-LABEL: stack_fold_pminsq_mask_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rsi), %zmm2
 ; CHECK-NEXT:    vpminsq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    vmovdqa64 %zmm2, %zmm0
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp slt <8 x i64> %a1, %a0
   %3 = select <8 x i1> %2, <8 x i64> %a1, <8 x i64> %a0
   %4 = bitcast i8 %mask to <8 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %5 = load <8 x i64>, <8 x i64>* %passthru
   %6 = select <8 x i1> %4, <8 x i64> %3, <8 x i64> %5
   ret <8 x i64> %6
 }

 define <8 x i64> @stack_fold_pminsq_maskz(<8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
 ; CHECK-LABEL: stack_fold_pminsq_maskz:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpminsq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp slt <8 x i64> %a0, %a1
   %3 = select <8 x i1> %2, <8 x i64> %a0, <8 x i64> %a1
   %4 = bitcast i8 %mask to <8 x i1>
   %5 = select <8 x i1> %4, <8 x i64> %3, <8 x i64> zeroinitializer
   ret <8 x i64> %5
 }

 define <8 x i64> @stack_fold_pminsq_maskz_commuted(<8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
 ; CHECK-LABEL: stack_fold_pminsq_maskz_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpminsq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp slt <8 x i64> %a1, %a0
   %3 = select <8 x i1> %2, <8 x i64> %a1, <8 x i64> %a0
   %4 = bitcast i8 %mask to <8 x i1>
   %5 = select <8 x i1> %4, <8 x i64> %3, <8 x i64> zeroinitializer
   ret <8 x i64> %5
 }

 define <32 x i16> @stack_fold_pminsw(<32 x i16> %a0, <32 x i16> %a1) {
 ; CHECK-LABEL: stack_fold_pminsw:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpminsw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp slt <32 x i16> %a0, %a1
   %3 = select <32 x i1> %2, <32 x i16> %a0, <32 x i16> %a1
   ret <32 x i16> %3
 }

 define <32 x i16> @stack_fold_pminsw_commuted(<32 x i16> %a0, <32 x i16> %a1) {
 ; CHECK-LABEL: stack_fold_pminsw_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpminsw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp slt <32 x i16> %a1, %a0
   %3 = select <32 x i1> %2, <32 x i16> %a1, <32 x i16> %a0
   ret <32 x i16> %3
 }

 define <32 x i16> @stack_fold_pminsw_mask(<32 x i16> %a0, <32 x i16> %a1, i32 %mask, <32 x i16>* %passthru) {
 ; CHECK-LABEL: stack_fold_pminsw_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rsi), %zmm2
 ; CHECK-NEXT:    vpminsw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    vmovdqa64 %zmm2, %zmm0
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp slt <32 x i16> %a0, %a1
   %3 = select <32 x i1> %2, <32 x i16> %a0, <32 x i16> %a1
   %4 = bitcast i32 %mask to <32 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %5 = load <32 x i16>, <32 x i16>* %passthru
   %6 = select <32 x i1> %4, <32 x i16> %3, <32 x i16> %5
   ret <32 x i16> %6
 }

 define <32 x i16> @stack_fold_pminsw_mask_commuted(<32 x i16> %a0, <32 x i16> %a1, i32 %mask, <32 x i16>* %passthru) {
 ; CHECK-LABEL: stack_fold_pminsw_mask_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rsi), %zmm2
 ; CHECK-NEXT:    vpminsw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    vmovdqa64 %zmm2, %zmm0
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp slt <32 x i16> %a1, %a0
   %3 = select <32 x i1> %2, <32 x i16> %a1, <32 x i16> %a0
   %4 = bitcast i32 %mask to <32 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %5 = load <32 x i16>, <32 x i16>* %passthru
   %6 = select <32 x i1> %4, <32 x i16> %3, <32 x i16> %5
   ret <32 x i16> %6
 }

 define <32 x i16> @stack_fold_pminsw_maskz(<32 x i16> %a0, <32 x i16> %a1, i32 %mask) {
 ; CHECK-LABEL: stack_fold_pminsw_maskz:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpminsw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp slt <32 x i16> %a0, %a1
   %3 = select <32 x i1> %2, <32 x i16> %a0, <32 x i16> %a1
   %4 = bitcast i32 %mask to <32 x i1>
   %5 = select <32 x i1> %4, <32 x i16> %3, <32 x i16> zeroinitializer
   ret <32 x i16> %5
 }

 define <32 x i16> @stack_fold_pminsw_maskz_commuted(<32 x i16> %a0, <32 x i16> %a1, i32 %mask) {
 ; CHECK-LABEL: stack_fold_pminsw_maskz_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpminsw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp slt <32 x i16> %a1, %a0
   %3 = select <32 x i1> %2, <32 x i16> %a1, <32 x i16> %a0
   %4 = bitcast i32 %mask to <32 x i1>
   %5 = select <32 x i1> %4, <32 x i16> %3, <32 x i16> zeroinitializer
   ret <32 x i16> %5
 }

 define <64 x i8> @stack_fold_pminub(<64 x i8> %a0, <64 x i8> %a1) {
 ; CHECK-LABEL: stack_fold_pminub:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpminub {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp ult <64 x i8> %a0, %a1
   %3 = select <64 x i1> %2, <64 x i8> %a0, <64 x i8> %a1
   ret <64 x i8> %3
 }

 define <64 x i8> @stack_fold_pminub_commuted(<64 x i8> %a0, <64 x i8> %a1) {
 ; CHECK-LABEL: stack_fold_pminub_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpminub {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp ult <64 x i8> %a1, %a0
   %3 = select <64 x i1> %2, <64 x i8> %a1, <64 x i8> %a0
   ret <64 x i8> %3
 }

 define <64 x i8> @stack_fold_pminub_mask(<64 x i8> %a0, <64 x i8> %a1, i64 %mask, <64 x i8>* %passthru) {
 ; CHECK-LABEL: stack_fold_pminub_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovq %rdi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rsi), %zmm2
 ; CHECK-NEXT:    vpminub {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    vmovdqa64 %zmm2, %zmm0
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp ult <64 x i8> %a0, %a1
   %3 = select <64 x i1> %2, <64 x i8> %a0, <64 x i8> %a1
   %4 = bitcast i64 %mask to <64 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %5 = load <64 x i8>, <64 x i8>* %passthru
   %6 = select <64 x i1> %4, <64 x i8> %3, <64 x i8> %5
   ret <64 x i8> %6
 }

 define <64 x i8> @stack_fold_pminub_mask_commuted(<64 x i8> %a0, <64 x i8> %a1, i64 %mask, <64 x i8>* %passthru) {
 ; CHECK-LABEL: stack_fold_pminub_mask_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovq %rdi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rsi), %zmm2
 ; CHECK-NEXT:    vpminub {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    vmovdqa64 %zmm2, %zmm0
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp ult <64 x i8> %a1, %a0
   %3 = select <64 x i1> %2, <64 x i8> %a1, <64 x i8> %a0
   %4 = bitcast i64 %mask to <64 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %5 = load <64 x i8>, <64 x i8>* %passthru
   %6 = select <64 x i1> %4, <64 x i8> %3, <64 x i8> %5
   ret <64 x i8> %6
 }

 define <64 x i8> @stack_fold_pminub_maskz(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) {
 ; CHECK-LABEL: stack_fold_pminub_maskz:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovq %rdi, %k1
 ; CHECK-NEXT:    vpminub {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp ult <64 x i8> %a0, %a1
   %3 = select <64 x i1> %2, <64 x i8> %a0, <64 x i8> %a1
   %4 = bitcast i64 %mask to <64 x i1>
   %5 = select <64 x i1> %4, <64 x i8> %3, <64 x i8> zeroinitializer
   ret <64 x i8> %5
 }

 define <64 x i8> @stack_fold_pminub_maskz_commuted(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) {
 ; CHECK-LABEL: stack_fold_pminub_maskz_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovq %rdi, %k1
 ; CHECK-NEXT:    vpminub {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp ult <64 x i8> %a1, %a0
   %3 = select <64 x i1> %2, <64 x i8> %a1, <64 x i8> %a0
   %4 = bitcast i64 %mask to <64 x i1>
   %5 = select <64 x i1> %4, <64 x i8> %3, <64 x i8> zeroinitializer
   ret <64 x i8> %5
 }

 define <16 x i32> @stack_fold_pminud(<16 x i32> %a0, <16 x i32> %a1) {
 ; CHECK-LABEL: stack_fold_pminud:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpminud {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp ult <16 x i32> %a0, %a1
   %3 = select <16 x i1> %2, <16 x i32> %a0, <16 x i32> %a1
   ret <16 x i32> %3
 }

 define <16 x i32> @stack_fold_pminud_commuted(<16 x i32> %a0, <16 x i32> %a1) {
 ; CHECK-LABEL: stack_fold_pminud_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpminud {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp ult <16 x i32> %a1, %a0
   %3 = select <16 x i1> %2, <16 x i32> %a1, <16 x i32> %a0
   ret <16 x i32> %3
 }

 define <16 x i32> @stack_fold_pminud_mask(<16 x i32> %a0, <16 x i32> %a1, i16 %mask, <16 x i32>* %passthru) {
 ; CHECK-LABEL: stack_fold_pminud_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rsi), %zmm2
 ; CHECK-NEXT:    vpminud {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    vmovdqa64 %zmm2, %zmm0
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp ult <16 x i32> %a0, %a1
   %3 = select <16 x i1> %2, <16 x i32> %a0, <16 x i32> %a1
   %4 = bitcast i16 %mask to <16 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %5 = load <16 x i32>, <16 x i32>* %passthru
   %6 = select <16 x i1> %4, <16 x i32> %3, <16 x i32> %5
   ret <16 x i32> %6
 }

 define <16 x i32> @stack_fold_pminud_mask_commuted(<16 x i32> %a0, <16 x i32> %a1, i16 %mask, <16 x i32>* %passthru) {
 ; CHECK-LABEL: stack_fold_pminud_mask_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rsi), %zmm2
 ; CHECK-NEXT:    vpminud {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    vmovdqa64 %zmm2, %zmm0
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp ult <16 x i32> %a1, %a0
   %3 = select <16 x i1> %2, <16 x i32> %a1, <16 x i32> %a0
   %4 = bitcast i16 %mask to <16 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %5 = load <16 x i32>, <16 x i32>* %passthru
   %6 = select <16 x i1> %4, <16 x i32> %3, <16 x i32> %5
   ret <16 x i32> %6
 }

 define <16 x i32> @stack_fold_pminud_maskz(<16 x i32> %a0, <16 x i32> %a1, i16 %mask) {
 ; CHECK-LABEL: stack_fold_pminud_maskz:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpminud {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp ult <16 x i32> %a0, %a1
   %3 = select <16 x i1> %2, <16 x i32> %a0, <16 x i32> %a1
   %4 = bitcast i16 %mask to <16 x i1>
   %5 = select <16 x i1> %4, <16 x i32> %3, <16 x i32> zeroinitializer
   ret <16 x i32> %5
 }

 define <16 x i32> @stack_fold_pminud_maskz_commuted(<16 x i32> %a0, <16 x i32> %a1, i16 %mask) {
 ; CHECK-LABEL: stack_fold_pminud_maskz_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpminud {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp ult <16 x i32> %a1, %a0
   %3 = select <16 x i1> %2, <16 x i32> %a1, <16 x i32> %a0
   %4 = bitcast i16 %mask to <16 x i1>
   %5 = select <16 x i1> %4, <16 x i32> %3, <16 x i32> zeroinitializer
   ret <16 x i32> %5
 }

 define <8 x i64> @stack_fold_pminuq(<8 x i64> %a0, <8 x i64> %a1) {
 ; CHECK-LABEL: stack_fold_pminuq:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpminuq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp ult <8 x i64> %a0, %a1
   %3 = select <8 x i1> %2, <8 x i64> %a0, <8 x i64> %a1
   ret <8 x i64> %3
 }

 define <8 x i64> @stack_fold_pminuq_commuted(<8 x i64> %a0, <8 x i64> %a1) {
 ; CHECK-LABEL: stack_fold_pminuq_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpminuq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp ult <8 x i64> %a1, %a0
   %3 = select <8 x i1> %2, <8 x i64> %a1, <8 x i64> %a0
   ret <8 x i64> %3
 }

 define <8 x i64> @stack_fold_pminuq_mask(<8 x i64> %a0, <8 x i64> %a1, i8 %mask, <8 x i64>* %passthru) {
 ; CHECK-LABEL: stack_fold_pminuq_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rsi), %zmm2
 ; CHECK-NEXT:    vpminuq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    vmovdqa64 %zmm2, %zmm0
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp ult <8 x i64> %a0, %a1
   %3 = select <8 x i1> %2, <8 x i64> %a0, <8 x i64> %a1
   %4 = bitcast i8 %mask to <8 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %5 = load <8 x i64>, <8 x i64>* %passthru
   %6 = select <8 x i1> %4, <8 x i64> %3, <8 x i64> %5
   ret <8 x i64> %6
 }

 define <8 x i64> @stack_fold_pminuq_mask_commuted(<8 x i64> %a0, <8 x i64> %a1, i8 %mask, <8 x i64>* %passthru) {
 ; CHECK-LABEL: stack_fold_pminuq_mask_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rsi), %zmm2
 ; CHECK-NEXT:    vpminuq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    vmovdqa64 %zmm2, %zmm0
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp ult <8 x i64> %a1, %a0
   %3 = select <8 x i1> %2, <8 x i64> %a1, <8 x i64> %a0
   %4 = bitcast i8 %mask to <8 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %5 = load <8 x i64>, <8 x i64>* %passthru
   %6 = select <8 x i1> %4, <8 x i64> %3, <8 x i64> %5
   ret <8 x i64> %6
 }

 define <8 x i64> @stack_fold_pminuq_maskz(<8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
 ; CHECK-LABEL: stack_fold_pminuq_maskz:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpminuq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp ult <8 x i64> %a0, %a1
   %3 = select <8 x i1> %2, <8 x i64> %a0, <8 x i64> %a1
   %4 = bitcast i8 %mask to <8 x i1>
   %5 = select <8 x i1> %4, <8 x i64> %3, <8 x i64> zeroinitializer
   ret <8 x i64> %5
 }

 define <8 x i64> @stack_fold_pminuq_maskz_commuted(<8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
 ; CHECK-LABEL: stack_fold_pminuq_maskz_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpminuq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp ult <8 x i64> %a1, %a0
   %3 = select <8 x i1> %2, <8 x i64> %a1, <8 x i64> %a0
   %4 = bitcast i8 %mask to <8 x i1>
   %5 = select <8 x i1> %4, <8 x i64> %3, <8 x i64> zeroinitializer
   ret <8 x i64> %5
 }

 define <32 x i16> @stack_fold_pminuw(<32 x i16> %a0, <32 x i16> %a1) {
 ; CHECK-LABEL: stack_fold_pminuw:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpminuw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp ult <32 x i16> %a0, %a1
   %3 = select <32 x i1> %2, <32 x i16> %a0, <32 x i16> %a1
   ret <32 x i16> %3
 }

 define <32 x i16> @stack_fold_pminuw_commuted(<32 x i16> %a0, <32 x i16> %a1) {
 ; CHECK-LABEL: stack_fold_pminuw_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpminuw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp ult <32 x i16> %a1, %a0
   %3 = select <32 x i1> %2, <32 x i16> %a1, <32 x i16> %a0
   ret <32 x i16> %3
 }

 define <32 x i16> @stack_fold_pminuw_mask(<32 x i16> %a0, <32 x i16> %a1, i32 %mask, <32 x i16>* %passthru) {
 ; CHECK-LABEL: stack_fold_pminuw_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rsi), %zmm2
 ; CHECK-NEXT:    vpminuw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    vmovdqa64 %zmm2, %zmm0
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp ult <32 x i16> %a0, %a1
   %3 = select <32 x i1> %2, <32 x i16> %a0, <32 x i16> %a1
   %4 = bitcast i32 %mask to <32 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %5 = load <32 x i16>, <32 x i16>* %passthru
   %6 = select <32 x i1> %4, <32 x i16> %3, <32 x i16> %5
   ret <32 x i16> %6
 }

 define <32 x i16> @stack_fold_pminuw_mask_commuted(<32 x i16> %a0, <32 x i16> %a1, i32 %mask, <32 x i16>* %passthru) {
 ; CHECK-LABEL: stack_fold_pminuw_mask_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rsi), %zmm2
 ; CHECK-NEXT:    vpminuw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    vmovdqa64 %zmm2, %zmm0
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp ult <32 x i16> %a1, %a0
   %3 = select <32 x i1> %2, <32 x i16> %a1, <32 x i16> %a0
   %4 = bitcast i32 %mask to <32 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %5 = load <32 x i16>, <32 x i16>* %passthru
   %6 = select <32 x i1> %4, <32 x i16> %3, <32 x i16> %5
   ret <32 x i16> %6
 }

 define <32 x i16> @stack_fold_pminuw_maskz(<32 x i16> %a0, <32 x i16> %a1, i32 %mask) {
 ; CHECK-LABEL: stack_fold_pminuw_maskz:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpminuw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp ult <32 x i16> %a0, %a1
   %3 = select <32 x i1> %2, <32 x i16> %a0, <32 x i16> %a1
   %4 = bitcast i32 %mask to <32 x i1>
   %5 = select <32 x i1> %4, <32 x i16> %3, <32 x i16> zeroinitializer
   ret <32 x i16> %5
 }

 define <32 x i16> @stack_fold_pminuw_maskz_commuted(<32 x i16> %a0, <32 x i16> %a1, i32 %mask) {
 ; CHECK-LABEL: stack_fold_pminuw_maskz_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpminuw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = icmp ult <32 x i16> %a1, %a0
   %3 = select <32 x i1> %2, <32 x i16> %a1, <32 x i16> %a0
   %4 = bitcast i32 %mask to <32 x i1>
   %5 = select <32 x i1> %4, <32 x i16> %3, <32 x i16> zeroinitializer
   ret <32 x i16> %5
 }

 define <16 x i8> @stack_fold_vpmovdb(<16 x i32> %a0) {
 ; CHECK-LABEL: stack_fold_vpmovdb:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vpmovdb %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Folded Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; CHECK-NEXT:    vzeroupper
 ; CHECK-NEXT:    retq
   %1 = call <16 x i8> @llvm.x86.avx512.mask.pmov.db.512(<16 x i32> %a0, <16 x i8> undef, i16 -1)
   %2 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   ret <16 x i8> %1
 }
 declare <16 x i8> @llvm.x86.avx512.mask.pmov.db.512(<16 x i32>, <16 x i8>, i16)

 define <16 x i16> @stack_fold_vpmovdw(<16 x i32> %a0) {
 ; CHECK-LABEL: stack_fold_vpmovdw:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vpmovdw %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Folded Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; CHECK-NEXT:    retq
   %1 = call <16 x i16> @llvm.x86.avx512.mask.pmov.dw.512(<16 x i32> %a0, <16 x i16> undef, i16 -1)
   %2 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   ret <16 x i16> %1
 }
 declare <16 x i16> @llvm.x86.avx512.mask.pmov.dw.512(<16 x i32>, <16 x i16>, i16)

 define <2 x i64> @stack_fold_movq_load(<2 x i64> %a0) {
 ; CHECK-LABEL: stack_fold_movq_load:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vmovq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
 ; CHECK-NEXT:    # xmm0 = mem[0],zero
 ; CHECK-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
 ; CHECK-NEXT:    vpsubq %xmm1, %xmm0, %xmm0
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = shufflevector <2 x i64> %a0, <2 x i64> zeroinitializer, <2 x i32> <i32 0, i32 2>
   ; add forces execution domain
   %3 = add <2 x i64> %2, <i64 1, i64 1>
   ret <2 x i64> %3
 }

 define <8 x i32> @stack_fold_vpmovqd(<8 x i64> %a0) {
 ; CHECK-LABEL: stack_fold_vpmovqd:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vpmovqd %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Folded Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; CHECK-NEXT:    retq
   %1 = trunc <8 x i64> %a0 to <8 x i32>
   %2 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   ret <8 x i32> %1
 }
 declare <8 x i32> @llvm.x86.avx512.mask.pmov.qd.512(<8 x i64>, <8 x i32>, i8)

 define <8 x i16> @stack_fold_vpmovqw(<8 x i64> %a0) {
 ; CHECK-LABEL: stack_fold_vpmovqw:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vpmovqw %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Folded Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; CHECK-NEXT:    vzeroupper
 ; CHECK-NEXT:    retq
   %1 = call <8 x i16> @llvm.x86.avx512.mask.pmov.qw.512(<8 x i64> %a0, <8 x i16> undef, i8 -1)
   %2 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   ret <8 x i16> %1
 }
 declare <8 x i16> @llvm.x86.avx512.mask.pmov.qw.512(<8 x i64>, <8 x i16>, i8)

 define <32 x i8> @stack_fold_vpmovwb(<32 x i16> %a0) {
 ; CHECK-LABEL: stack_fold_vpmovwb:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vpmovwb %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Folded Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; CHECK-NEXT:    retq
   %1 = trunc <32 x i16> %a0 to <32 x i8>
   %2 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   ret <32 x i8> %1
 }
 declare <32 x i8> @llvm.x86.avx512.mask.pmov.wb.512(<32 x i16>, <32 x i8>, i32)

 define <16 x i8> @stack_fold_vpmovsdb(<16 x i32> %a0) {
 ; CHECK-LABEL: stack_fold_vpmovsdb:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vpmovsdb %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Folded Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; CHECK-NEXT:    vzeroupper
 ; CHECK-NEXT:    retq
   %1 = call <16 x i8> @llvm.x86.avx512.mask.pmovs.db.512(<16 x i32> %a0, <16 x i8> undef, i16 -1)
   %2 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   ret <16 x i8> %1
 }
 declare <16 x i8> @llvm.x86.avx512.mask.pmovs.db.512(<16 x i32>, <16 x i8>, i16)

 define <16 x i16> @stack_fold_vpmovsdw(<16 x i32> %a0) {
 ; CHECK-LABEL: stack_fold_vpmovsdw:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vpmovsdw %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Folded Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; CHECK-NEXT:    retq
   %1 = call <16 x i16> @llvm.x86.avx512.mask.pmovs.dw.512(<16 x i32> %a0, <16 x i16> undef, i16 -1)
   %2 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   ret <16 x i16> %1
 }
 declare <16 x i16> @llvm.x86.avx512.mask.pmovs.dw.512(<16 x i32>, <16 x i16>, i16)

 define <8 x i32> @stack_fold_vpmovsqd(<8 x i64> %a0) {
 ; CHECK-LABEL: stack_fold_vpmovsqd:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vpmovsqd %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Folded Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; CHECK-NEXT:    retq
   %1 = call <8 x i32> @llvm.x86.avx512.mask.pmovs.qd.512(<8 x i64> %a0, <8 x i32> undef, i8 -1)
   %2 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   ret <8 x i32> %1
 }
 declare <8 x i32> @llvm.x86.avx512.mask.pmovs.qd.512(<8 x i64>, <8 x i32>, i8)

 define <8 x i16> @stack_fold_vpmovsqw(<8 x i64> %a0) {
 ; CHECK-LABEL: stack_fold_vpmovsqw:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vpmovsqw %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Folded Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; CHECK-NEXT:    vzeroupper
 ; CHECK-NEXT:    retq
   %1 = call <8 x i16> @llvm.x86.avx512.mask.pmovs.qw.512(<8 x i64> %a0, <8 x i16> undef, i8 -1)
   %2 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   ret <8 x i16> %1
 }
 declare <8 x i16> @llvm.x86.avx512.mask.pmovs.qw.512(<8 x i64>, <8 x i16>, i8)

 define <32 x i8> @stack_fold_vpmovswb(<32 x i16> %a0) {
 ; CHECK-LABEL: stack_fold_vpmovswb:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vpmovswb %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Folded Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; CHECK-NEXT:    retq
   %1 = call <32 x i8> @llvm.x86.avx512.mask.pmovs.wb.512(<32 x i16> %a0, <32 x i8> undef, i32 -1)
   %2 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   ret <32 x i8> %1
 }
 declare <32 x i8> @llvm.x86.avx512.mask.pmovs.wb.512(<32 x i16>, <32 x i8>, i32)

 define <16 x i32> @stack_fold_pmovsxbd_zmm(<16 x i8> %a0) {
 ; CHECK-LABEL: stack_fold_pmovsxbd_zmm:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpmovsxbd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 16-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = sext <16 x i8> %a0 to <16 x i32>
   ret <16 x i32> %2
 }

 define <8 x i64> @stack_fold_pmovsxbq_zmm(<16 x i8> %a0) {
 ; CHECK-LABEL: stack_fold_pmovsxbq_zmm:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpmovsxbq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 16-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = shufflevector <16 x i8> %a0, <16 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
   %3 = sext <8 x i8> %2 to <8 x i64>
   ret <8 x i64> %3
 }

 define <32 x i16> @stack_fold_pmovsxbw_zmm(<32 x i8> %a0) {
 ; CHECK-LABEL: stack_fold_pmovsxbw_zmm:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpmovsxbw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 32-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = sext <32 x i8> %a0 to <32 x i16>
   ret <32 x i16> %2
 }

 define <8 x i64> @stack_fold_pmovsxdq_zmm(<8 x i32> %a0) {
 ; CHECK-LABEL: stack_fold_pmovsxdq_zmm:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpmovsxdq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 32-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = sext <8 x i32> %a0 to <8 x i64>
   ret <8 x i64> %2
 }

 define <16 x i32> @stack_fold_pmovsxwd_zmm(<16 x i16> %a0) {
 ; CHECK-LABEL: stack_fold_pmovsxwd_zmm:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpmovsxwd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 32-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = sext <16 x i16> %a0 to <16 x i32>
   ret <16 x i32> %2
 }

 define <8 x i64> @stack_fold_pmovsxwq_zmm(<8 x i16> %a0) {
 ; CHECK-LABEL: stack_fold_pmovsxwq_zmm:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpmovsxwq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 16-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = sext <8 x i16> %a0 to <8 x i64>
   ret <8 x i64> %2
 }

 define <8 x i64> @stack_fold_pmovsxwq_mask_zmm(<8 x i64> %passthru, <8 x i16> %a0, i8 %mask) {
 ; CHECK-LABEL: stack_fold_pmovsxwq_mask_zmm:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpmovsxwq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 {%k1} # 16-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = sext <8 x i16> %a0 to <8 x i64>
   %3 = bitcast i8 %mask to <8 x i1>
   %4 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> %passthru
   ret <8 x i64> %4
 }

 define <8 x i64> @stack_fold_pmovsxwq_maskz_zmm(<8 x i16> %a0, i8 %mask) {
 ; CHECK-LABEL: stack_fold_pmovsxwq_maskz_zmm:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpmovsxwq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 {%k1} {z} # 16-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = sext <8 x i16> %a0 to <8 x i64>
   %3 = bitcast i8 %mask to <8 x i1>
   %4 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> zeroinitializer
   ret <8 x i64> %4
 }

 define <16 x i8> @stack_fold_vpmovusdb(<16 x i32> %a0) {
 ; CHECK-LABEL: stack_fold_vpmovusdb:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vpmovusdb %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Folded Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; CHECK-NEXT:    vzeroupper
 ; CHECK-NEXT:    retq
   %1 = call <16 x i8> @llvm.x86.avx512.mask.pmovus.db.512(<16 x i32> %a0, <16 x i8> undef, i16 -1)
   %2 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   ret <16 x i8> %1
 }
 declare <16 x i8> @llvm.x86.avx512.mask.pmovus.db.512(<16 x i32>, <16 x i8>, i16)

 define <16 x i16> @stack_fold_vpmovusdw(<16 x i32> %a0) {
 ; CHECK-LABEL: stack_fold_vpmovusdw:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vpmovusdw %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Folded Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; CHECK-NEXT:    retq
   %1 = call <16 x i16> @llvm.x86.avx512.mask.pmovus.dw.512(<16 x i32> %a0, <16 x i16> undef, i16 -1)
   %2 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   ret <16 x i16> %1
 }
 declare <16 x i16> @llvm.x86.avx512.mask.pmovus.dw.512(<16 x i32>, <16 x i16>, i16)

 define <8 x i32> @stack_fold_vpmovusqd(<8 x i64> %a0) {
 ; CHECK-LABEL: stack_fold_vpmovusqd:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vpmovusqd %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Folded Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; CHECK-NEXT:    retq
   %1 = call <8 x i32> @llvm.x86.avx512.mask.pmovus.qd.512(<8 x i64> %a0, <8 x i32> undef, i8 -1)
   %2 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   ret <8 x i32> %1
 }
 declare <8 x i32> @llvm.x86.avx512.mask.pmovus.qd.512(<8 x i64>, <8 x i32>, i8)

 define <8 x i16> @stack_fold_vpmovusqw(<8 x i64> %a0) {
 ; CHECK-LABEL: stack_fold_vpmovusqw:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vpmovusqw %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Folded Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
 ; CHECK-NEXT:    vzeroupper
 ; CHECK-NEXT:    retq
   %1 = call <8 x i16> @llvm.x86.avx512.mask.pmovus.qw.512(<8 x i64> %a0, <8 x i16> undef, i8 -1)
   %2 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   ret <8 x i16> %1
 }
 declare <8 x i16> @llvm.x86.avx512.mask.pmovus.qw.512(<8 x i64>, <8 x i16>, i8)

 define <32 x i8> @stack_fold_vpmovuswb(<32 x i16> %a0) {
 ; CHECK-LABEL: stack_fold_vpmovuswb:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vpmovuswb %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Folded Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
 ; CHECK-NEXT:    retq
   %1 = call <32 x i8> @llvm.x86.avx512.mask.pmovus.wb.512(<32 x i16> %a0, <32 x i8> undef, i32 -1)
   %2 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   ret <32 x i8> %1
 }
 declare <32 x i8> @llvm.x86.avx512.mask.pmovus.wb.512(<32 x i16>, <32 x i8>, i32)

 define <16 x i32> @stack_fold_pmovzxbd_zmm(<16 x i8> %a0) {
 ; CHECK-LABEL: stack_fold_pmovzxbd_zmm:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpmovzxbd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 16-byte Folded Reload
 ; CHECK-NEXT:    # zmm0 = mem[0],zero,zero,zero,mem[1],zero,zero,zero,mem[2],zero,zero,zero,mem[3],zero,zero,zero,mem[4],zero,zero,zero,mem[5],zero,zero,zero,mem[6],zero,zero,zero,mem[7],zero,zero,zero,mem[8],zero,zero,zero,mem[9],zero,zero,zero,mem[10],zero,zero,zero,mem[11],zero,zero,zero,mem[12],zero,zero,zero,mem[13],zero,zero,zero,mem[14],zero,zero,zero,mem[15],zero,zero,zero
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = zext <16 x i8> %a0 to <16 x i32>
   ret <16 x i32> %2
 }

 define <8 x i64> @stack_fold_pmovzxbq_zmm(<16 x i8> %a0) {
 ; CHECK-LABEL: stack_fold_pmovzxbq_zmm:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpmovzxbq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 16-byte Folded Reload
 ; CHECK-NEXT:    # zmm0 = mem[0],zero,zero,zero,zero,zero,zero,zero,mem[1],zero,zero,zero,zero,zero,zero,zero,mem[2],zero,zero,zero,zero,zero,zero,zero,mem[3],zero,zero,zero,zero,zero,zero,zero,mem[4],zero,zero,zero,zero,zero,zero,zero,mem[5],zero,zero,zero,zero,zero,zero,zero,mem[6],zero,zero,zero,zero,zero,zero,zero,mem[7],zero,zero,zero,zero,zero,zero,zero
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = shufflevector <16 x i8> %a0, <16 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
   %3 = zext <8 x i8> %2 to <8 x i64>
   ret <8 x i64> %3
 }

 define <32 x i16> @stack_fold_pmovzxbw_zmm(<32 x i8> %a0) {
 ; CHECK-LABEL: stack_fold_pmovzxbw_zmm:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpmovzxbw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 32-byte Folded Reload
 ; CHECK-NEXT:    # zmm0 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero,mem[8],zero,mem[9],zero,mem[10],zero,mem[11],zero,mem[12],zero,mem[13],zero,mem[14],zero,mem[15],zero,mem[16],zero,mem[17],zero,mem[18],zero,mem[19],zero,mem[20],zero,mem[21],zero,mem[22],zero,mem[23],zero,mem[24],zero,mem[25],zero,mem[26],zero,mem[27],zero,mem[28],zero,mem[29],zero,mem[30],zero,mem[31],zero
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = zext <32 x i8> %a0 to <32 x i16>
   ret <32 x i16> %2
 }

 define <8 x i64> @stack_fold_pmovzxdq_zmm(<8 x i32> %a0) {
 ; CHECK-LABEL: stack_fold_pmovzxdq_zmm:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpmovzxdq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 32-byte Folded Reload
 ; CHECK-NEXT:    # zmm0 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = zext <8 x i32> %a0 to <8 x i64>
   ret <8 x i64> %2
 }

 define <16 x i32> @stack_fold_pmovzxwd_zmm(<16 x i16> %a0) {
 ; CHECK-LABEL: stack_fold_pmovzxwd_zmm:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpmovzxwd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 32-byte Folded Reload
 ; CHECK-NEXT:    # zmm0 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero,mem[8],zero,mem[9],zero,mem[10],zero,mem[11],zero,mem[12],zero,mem[13],zero,mem[14],zero,mem[15],zero
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = zext <16 x i16> %a0 to <16 x i32>
   ret <16 x i32> %2
 }

 define <8 x i64> @stack_fold_pmovzxwq_zmm(<8 x i16> %a0) {
 ; CHECK-LABEL: stack_fold_pmovzxwq_zmm:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpmovzxwq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 16-byte Folded Reload
 ; CHECK-NEXT:    # zmm0 = mem[0],zero,zero,zero,mem[1],zero,zero,zero,mem[2],zero,zero,zero,mem[3],zero,zero,zero,mem[4],zero,zero,zero,mem[5],zero,zero,zero,mem[6],zero,zero,zero,mem[7],zero,zero,zero
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = zext <8 x i16> %a0 to <8 x i64>
   ret <8 x i64> %2
 }

 define <8 x i64> @stack_fold_pmovzxwq_mask_zmm(<8 x i64> %passthru, <8 x i16> %a0, i8 %mask) {
 ; CHECK-LABEL: stack_fold_pmovzxwq_mask_zmm:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpmovzxwq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 {%k1} # 16-byte Folded Reload
 ; CHECK-NEXT:    # zmm0 {%k1} = mem[0],zero,zero,zero,mem[1],zero,zero,zero,mem[2],zero,zero,zero,mem[3],zero,zero,zero,mem[4],zero,zero,zero,mem[5],zero,zero,zero,mem[6],zero,zero,zero,mem[7],zero,zero,zero
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = zext <8 x i16> %a0 to <8 x i64>
   %3 = bitcast i8 %mask to <8 x i1>
   %4 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> %passthru
   ret <8 x i64> %4
 }

 define <8 x i64> @stack_fold_pmovzxwq_maskz_zmm(<8 x i16> %a0, i8 %mask) {
 ; CHECK-LABEL: stack_fold_pmovzxwq_maskz_zmm:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpmovzxwq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 {%k1} {z} # 16-byte Folded Reload
 ; CHECK-NEXT:    # zmm0 {%k1} {z} = mem[0],zero,zero,zero,mem[1],zero,zero,zero,mem[2],zero,zero,zero,mem[3],zero,zero,zero,mem[4],zero,zero,zero,mem[5],zero,zero,zero,mem[6],zero,zero,zero,mem[7],zero,zero,zero
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = zext <8 x i16> %a0 to <8 x i64>
   %3 = bitcast i8 %mask to <8 x i1>
   %4 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> zeroinitializer
   ret <8 x i64> %4
 }

 define <16 x i32> @stack_fold_pmulld(<16 x i32> %a0, <16 x i32> %a1) {
 ; CHECK-LABEL: stack_fold_pmulld:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpmulld {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = mul <16 x i32> %a0, %a1
   ret <16 x i32> %2
 }

 define <16 x i32> @stack_fold_pmulld_commuted(<16 x i32> %a0, <16 x i32> %a1) {
 ; CHECK-LABEL: stack_fold_pmulld_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpmulld {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = mul <16 x i32> %a1, %a0
   ret <16 x i32> %2
 }

 define <16 x i32> @stack_fold_pmulld_mask(<16 x i32> %a0, <16 x i32> %a1, <16 x i32>* %a2, i16 %mask) {
 ; CHECK-LABEL: stack_fold_pmulld_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovdqa64 %zmm0, %zmm1
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %esi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rdi), %zmm0
 ; CHECK-NEXT:    vpmulld {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = mul <16 x i32> %a0, %a1
   %3 = bitcast i16 %mask to <16 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %4 = load <16 x i32>, <16 x i32>* %a2
   %5 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> %4
   ret <16 x i32> %5
 }

 define <16 x i32> @stack_fold_pmulld_mask_commuted(<16 x i32> %a0, <16 x i32> %a1, <16 x i32>* %a2, i16 %mask) {
 ; CHECK-LABEL: stack_fold_pmulld_mask_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovdqa64 %zmm0, %zmm1
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %esi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rdi), %zmm0
 ; CHECK-NEXT:    vpmulld {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = mul <16 x i32> %a1, %a0
   %3 = bitcast i16 %mask to <16 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %4 = load <16 x i32>, <16 x i32>* %a2
   %5 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> %4
   ret <16 x i32> %5
 }

 define <16 x i32> @stack_fold_pmulld_maskz(<16 x i32> %a0, <16 x i32> %a1, i16 %mask) {
 ; CHECK-LABEL: stack_fold_pmulld_maskz:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpmulld {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = mul <16 x i32> %a0, %a1
   %3 = bitcast i16 %mask to <16 x i1>
   %4 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> zeroinitializer
   ret <16 x i32> %4
 }

 define <16 x i32> @stack_fold_pmulld_maskz_commuted(<16 x i32> %a0, <16 x i32> %a1, i16 %mask) {
 ; CHECK-LABEL: stack_fold_pmulld_maskz_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpmulld {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = mul <16 x i32> %a1, %a0
   %3 = bitcast i16 %mask to <16 x i1>
   %4 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> zeroinitializer
   ret <16 x i32> %4
 }

 define <8 x i64> @stack_fold_pmullq(<8 x i64> %a0, <8 x i64> %a1) {
 ; CHECK-LABEL: stack_fold_pmullq:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpmullq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = mul <8 x i64> %a0, %a1
   ret <8 x i64> %2
 }

 define <8 x i64> @stack_fold_pmullq_commuted(<8 x i64> %a0, <8 x i64> %a1) {
 ; CHECK-LABEL: stack_fold_pmullq_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpmullq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = mul <8 x i64> %a1, %a0
   ret <8 x i64> %2
 }

 define <8 x i64> @stack_fold_pmullq_mask(<8 x i64> %a0, <8 x i64> %a1, <8 x i64>* %a2, i8 %mask) {
 ; CHECK-LABEL: stack_fold_pmullq_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovdqa64 %zmm0, %zmm1
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %esi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rdi), %zmm0
 ; CHECK-NEXT:    vpmullq {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = mul <8 x i64> %a0, %a1
   %3 = bitcast i8 %mask to <8 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %4 = load <8 x i64>, <8 x i64>* %a2
   %5 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> %4
   ret <8 x i64> %5
 }

 define <8 x i64> @stack_fold_pmullq_mask_commuted(<8 x i64> %a0, <8 x i64> %a1, <8 x i64>* %a2, i8 %mask) {
 ; CHECK-LABEL: stack_fold_pmullq_mask_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovdqa64 %zmm0, %zmm1
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %esi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rdi), %zmm0
 ; CHECK-NEXT:    vpmullq {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = mul <8 x i64> %a1, %a0
   %3 = bitcast i8 %mask to <8 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %4 = load <8 x i64>, <8 x i64>* %a2
   %5 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> %4
   ret <8 x i64> %5
 }

 define <8 x i64> @stack_fold_pmullq_maskz(<8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
 ; CHECK-LABEL: stack_fold_pmullq_maskz:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpmullq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = mul <8 x i64> %a0, %a1
   %3 = bitcast i8 %mask to <8 x i1>
   %4 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> zeroinitializer
   ret <8 x i64> %4
 }

 define <8 x i64> @stack_fold_pmullq_maskz_commuted(<8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
 ; CHECK-LABEL: stack_fold_pmullq_maskz_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpmullq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = mul <8 x i64> %a1, %a0
   %3 = bitcast i8 %mask to <8 x i1>
   %4 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> zeroinitializer
   ret <8 x i64> %4
 }

 define <32 x i16> @stack_fold_pmullw(<32 x i16> %a0, <32 x i16> %a1) {
 ; CHECK-LABEL: stack_fold_pmullw:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpmullw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = mul <32 x i16> %a0, %a1
   ret <32 x i16> %2
 }

 define <32 x i16> @stack_fold_pmullw_commuted(<32 x i16> %a0, <32 x i16> %a1) {
 ; CHECK-LABEL: stack_fold_pmullw_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpmullw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = mul <32 x i16> %a1, %a0
   ret <32 x i16> %2
 }

 define <32 x i16> @stack_fold_pmullw_mask(<32 x i16> %a0, <32 x i16> %a1, <32 x i16>* %a2, i32 %mask) {
 ; CHECK-LABEL: stack_fold_pmullw_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovdqa64 %zmm0, %zmm1
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %esi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rdi), %zmm0
 ; CHECK-NEXT:    vpmullw {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = mul <32 x i16> %a0, %a1
   %3 = bitcast i32 %mask to <32 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %4 = load <32 x i16>, <32 x i16>* %a2
   %5 = select <32 x i1> %3, <32 x i16> %2, <32 x i16> %4
   ret <32 x i16> %5
 }

 define <32 x i16> @stack_fold_pmullw_mask_commuted(<32 x i16> %a0, <32 x i16> %a1, <32 x i16>* %a2, i32 %mask) {
 ; CHECK-LABEL: stack_fold_pmullw_mask_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovdqa64 %zmm0, %zmm1
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %esi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rdi), %zmm0
 ; CHECK-NEXT:    vpmullw {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = mul <32 x i16> %a1, %a0
   %3 = bitcast i32 %mask to <32 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %4 = load <32 x i16>, <32 x i16>* %a2
   %5 = select <32 x i1> %3, <32 x i16> %2, <32 x i16> %4
   ret <32 x i16> %5
 }

 define <32 x i16> @stack_fold_pmullw_maskz(<32 x i16> %a0, <32 x i16> %a1, i32 %mask) {
 ; CHECK-LABEL: stack_fold_pmullw_maskz:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpmullw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = mul <32 x i16> %a0, %a1
   %3 = bitcast i32 %mask to <32 x i1>
   %4 = select <32 x i1> %3, <32 x i16> %2, <32 x i16> zeroinitializer
   ret <32 x i16> %4
 }

 define <32 x i16> @stack_fold_pmullw_maskz_commuted(<32 x i16> %a0, <32 x i16> %a1, i32 %mask) {
 ; CHECK-LABEL: stack_fold_pmullw_maskz_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpmullw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = mul <32 x i16> %a1, %a0
   %3 = bitcast i32 %mask to <32 x i1>
   %4 = select <32 x i1> %3, <32 x i16> %2, <32 x i16> zeroinitializer
   ret <32 x i16> %4
 }

 define <8 x i64> @stack_fold_pmuldq(<8 x i64> %a0, <8 x i64> %a1) {
 ; CHECK-LABEL: stack_fold_pmuldq:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpmuldq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = shl <8 x i64> %a0, <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>
   %3 = ashr <8 x i64> %2, <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>
   %4 = shl <8 x i64> %a1, <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>
   %5 = ashr <8 x i64> %4, <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>
   %6 = mul <8 x i64> %3, %5
   ret <8 x i64> %6
 }

 define <8 x i64> @stack_fold_pmuldq_commuted(<8 x i64> %a0, <8 x i64> %a1) {
 ; CHECK-LABEL: stack_fold_pmuldq_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpmuldq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = shl <8 x i64> %a0, <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>
   %3 = ashr <8 x i64> %2, <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>
   %4 = shl <8 x i64> %a1, <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>
   %5 = ashr <8 x i64> %4, <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>
   %6 = mul <8 x i64> %5, %3
   ret <8 x i64> %6
 }

 define <8 x i64> @stack_fold_pmuldq_mask(<8 x i64> %a0, <8 x i64> %a1, <8 x i64>* %a2, i8 %mask) {
 ; CHECK-LABEL: stack_fold_pmuldq_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovdqa64 %zmm0, %zmm1
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %esi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rdi), %zmm0
 ; CHECK-NEXT:    vpmuldq {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = shl <8 x i64> %a0, <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>
   %3 = ashr <8 x i64> %2, <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>
   %4 = shl <8 x i64> %a1, <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>
   %5 = ashr <8 x i64> %4, <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>
   %6 = mul <8 x i64> %3, %5
   %7 = bitcast i8 %mask to <8 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %8 = load <8 x i64>, <8 x i64>* %a2
   %9 = select <8 x i1> %7, <8 x i64> %6, <8 x i64> %8
   ret <8 x i64> %9
 }

 define <8 x i64> @stack_fold_pmuldq_mask_commuted(<8 x i64> %a0, <8 x i64> %a1, <8 x i64>* %a2, i8 %mask) {
 ; CHECK-LABEL: stack_fold_pmuldq_mask_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovdqa64 %zmm0, %zmm1
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %esi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rdi), %zmm0
 ; CHECK-NEXT:    vpmuldq {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = shl <8 x i64> %a0, <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>
   %3 = ashr <8 x i64> %2, <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>
   %4 = shl <8 x i64> %a1, <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>
   %5 = ashr <8 x i64> %4, <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>
   %6 = mul <8 x i64> %5, %3
   %7 = bitcast i8 %mask to <8 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %8 = load <8 x i64>, <8 x i64>* %a2
   %9 = select <8 x i1> %7, <8 x i64> %6, <8 x i64> %8
   ret <8 x i64> %9
 }

 define <8 x i64> @stack_fold_pmuldq_maskz(<8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
 ; CHECK-LABEL: stack_fold_pmuldq_maskz:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpmuldq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = shl <8 x i64> %a0, <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>
   %3 = ashr <8 x i64> %2, <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>
   %4 = shl <8 x i64> %a1, <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>
   %5 = ashr <8 x i64> %4, <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>
   %6 = mul <8 x i64> %3, %5
   %7 = bitcast i8 %mask to <8 x i1>
   %8 = select <8 x i1> %7, <8 x i64> %6, <8 x i64> zeroinitializer
   ret <8 x i64> %8
 }

 define <8 x i64> @stack_fold_pmuldq_maskz_commuted(<8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
 ; CHECK-LABEL: stack_fold_pmuldq_maskz_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpmuldq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = shl <8 x i64> %a0, <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>
   %3 = ashr <8 x i64> %2, <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>
   %4 = shl <8 x i64> %a1, <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>
   %5 = ashr <8 x i64> %4, <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>
   %6 = mul <8 x i64> %5, %3
   %7 = bitcast i8 %mask to <8 x i1>
   %8 = select <8 x i1> %7, <8 x i64> %6, <8 x i64> zeroinitializer
   ret <8 x i64> %8
 }


 define <8 x i64> @stack_fold_pmuludq(<8 x i64> %a0, <8 x i64> %a1) {
 ; CHECK-LABEL: stack_fold_pmuludq:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpmuludq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = and <8 x i64> %a0, <i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295>
   %3 = and <8 x i64> %a1, <i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295>
   %4 = mul <8 x i64> %2, %3
   ret <8 x i64> %4
 }

 define <8 x i64> @stack_fold_pmuludq_commuted(<8 x i64> %a0, <8 x i64> %a1) {
 ; CHECK-LABEL: stack_fold_pmuludq_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpmuludq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = and <8 x i64> %a0, <i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295>
   %3 = and <8 x i64> %a1, <i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295>
   %4 = mul <8 x i64> %3, %2
   ret <8 x i64> %4
 }

 define <8 x i64> @stack_fold_pmuludq_mask(<8 x i64> %a0, <8 x i64> %a1, <8 x i64>* %a2, i8 %mask) {
 ; CHECK-LABEL: stack_fold_pmuludq_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovdqa64 %zmm0, %zmm1
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %esi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rdi), %zmm0
 ; CHECK-NEXT:    vpmuludq {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = and <8 x i64> %a0, <i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295>
   %3 = and <8 x i64> %a1, <i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295>
   %4 = mul <8 x i64> %2, %3
   %5 = bitcast i8 %mask to <8 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %6 = load <8 x i64>, <8 x i64>* %a2
   %7 = select <8 x i1> %5, <8 x i64> %4, <8 x i64> %6
   ret <8 x i64> %7
 }

 define <8 x i64> @stack_fold_pmuludq_mask_commuted(<8 x i64> %a0, <8 x i64> %a1, <8 x i64>* %a2, i8 %mask) {
 ; CHECK-LABEL: stack_fold_pmuludq_mask_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovdqa64 %zmm0, %zmm1
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %esi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rdi), %zmm0
 ; CHECK-NEXT:    vpmuludq {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = and <8 x i64> %a0, <i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295>
   %3 = and <8 x i64> %a1, <i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295>
   %4 = mul <8 x i64> %3, %2
   %5 = bitcast i8 %mask to <8 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %6 = load <8 x i64>, <8 x i64>* %a2
   %7 = select <8 x i1> %5, <8 x i64> %4, <8 x i64> %6
   ret <8 x i64> %7
 }

 define <8 x i64> @stack_fold_pmuludq_maskz(<8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
 ; CHECK-LABEL: stack_fold_pmuludq_maskz:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpmuludq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = and <8 x i64> %a0, <i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295>
   %3 = and <8 x i64> %a1, <i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295>
   %4 = mul <8 x i64> %2, %3
   %5 = bitcast i8 %mask to <8 x i1>
   %6 = select <8 x i1> %5, <8 x i64> %4, <8 x i64> zeroinitializer
   ret <8 x i64> %6
 }

 define <8 x i64> @stack_fold_pmuludq_maskz_commuted(<8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
 ; CHECK-LABEL: stack_fold_pmuludq_maskz_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpmuludq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = and <8 x i64> %a0, <i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295>
   %3 = and <8 x i64> %a1, <i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295, i64 4294967295>
   %4 = mul <8 x i64> %3, %2
   %5 = bitcast i8 %mask to <8 x i1>
   %6 = select <8 x i1> %5, <8 x i64> %4, <8 x i64> zeroinitializer
   ret <8 x i64> %6
 }

 define <16 x i32> @stack_fold_vpopcntd(<16 x i32> %a0) {
 ; CHECK-LABEL: stack_fold_vpopcntd:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpopcntd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %a0)
   ret <16 x i32> %2
 }
 declare <16 x i32> @llvm.ctpop.v16i32(<16 x i32>) nounwind readonly

 define <8 x i64> @stack_fold_vpopcntq(<8 x i64> %a0) {
 ; CHECK-LABEL: stack_fold_vpopcntq:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpopcntq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %a0)
   ret <8 x i64> %2
 }
 declare <8 x i64> @llvm.ctpop.v8i64(<8 x i64>) nounwind readnone

 define <16 x i32> @stack_fold_pord(<16 x i32> %a0, <16 x i32> %a1) {
 ; CHECK-LABEL: stack_fold_pord:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vorps {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = or <16 x i32> %a0, %a1
   ret <16 x i32> %2
 }

 define <16 x i32> @stack_fold_pord_commuted(<16 x i32> %a0, <16 x i32> %a1) {
 ; CHECK-LABEL: stack_fold_pord_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vorps {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = or <16 x i32> %a1, %a0
   ret <16 x i32> %2
 }

 define <16 x i32> @stack_fold_pord_mask(<16 x i32> %a0, <16 x i32> %a1, <16 x i32>* %a2, i16 %mask) {
 ; CHECK-LABEL: stack_fold_pord_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovaps %zmm0, %zmm1
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %esi, %k1
 ; CHECK-NEXT:    vmovaps (%rdi), %zmm0
 ; CHECK-NEXT:    vorps {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = or <16 x i32> %a0, %a1
   %3 = bitcast i16 %mask to <16 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %4 = load <16 x i32>, <16 x i32>* %a2
   %5 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> %4
   ret <16 x i32> %5
 }

 define <16 x i32> @stack_fold_pord_mask_commuted(<16 x i32> %a0, <16 x i32> %a1, <16 x i32>* %a2, i16 %mask) {
 ; CHECK-LABEL: stack_fold_pord_mask_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovaps %zmm0, %zmm1
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %esi, %k1
 ; CHECK-NEXT:    vmovaps (%rdi), %zmm0
 ; CHECK-NEXT:    vorps {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = or <16 x i32> %a1, %a0
   %3 = bitcast i16 %mask to <16 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %4 = load <16 x i32>, <16 x i32>* %a2
   %5 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> %4
   ret <16 x i32> %5
 }

 define <16 x i32> @stack_fold_pord_maskz(<16 x i32> %a0, <16 x i32> %a1, i16 %mask) {
 ; CHECK-LABEL: stack_fold_pord_maskz:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vorps {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = or <16 x i32> %a0, %a1
   %3 = bitcast i16 %mask to <16 x i1>
   %4 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> zeroinitializer
   ret <16 x i32> %4
 }

 define <16 x i32> @stack_fold_pord_maskz_commuted(<16 x i32> %a0, <16 x i32> %a1, i16 %mask) {
 ; CHECK-LABEL: stack_fold_pord_maskz_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vorps {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = or <16 x i32> %a1, %a0
   %3 = bitcast i16 %mask to <16 x i1>
   %4 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> zeroinitializer
   ret <16 x i32> %4
 }

 define <8 x i64> @stack_fold_porq(<8 x i64> %a0, <8 x i64> %a1) {
 ; CHECK-LABEL: stack_fold_porq:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vorps {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = or <8 x i64> %a0, %a1
   ret <8 x i64> %2
 }

 define <8 x i64> @stack_fold_porq_commuted(<8 x i64> %a0, <8 x i64> %a1) {
 ; CHECK-LABEL: stack_fold_porq_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vorps {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = or <8 x i64> %a1, %a0
   ret <8 x i64> %2
 }

 define <8 x i64> @stack_fold_porq_mask(<8 x i64> %a0, <8 x i64> %a1, <8 x i64>* %a2, i8 %mask) {
 ; CHECK-LABEL: stack_fold_porq_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovapd %zmm0, %zmm1
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %esi, %k1
 ; CHECK-NEXT:    vmovapd (%rdi), %zmm0
 ; CHECK-NEXT:    vorpd {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = or <8 x i64> %a0, %a1
   %3 = bitcast i8 %mask to <8 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %4 = load <8 x i64>, <8 x i64>* %a2
   %5 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> %4
   ret <8 x i64> %5
 }

 define <8 x i64> @stack_fold_porq_mask_commuted(<8 x i64> %a0, <8 x i64> %a1, <8 x i64>* %a2, i8 %mask) {
 ; CHECK-LABEL: stack_fold_porq_mask_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovapd %zmm0, %zmm1
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %esi, %k1
 ; CHECK-NEXT:    vmovapd (%rdi), %zmm0
 ; CHECK-NEXT:    vorpd {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = or <8 x i64> %a1, %a0
   %3 = bitcast i8 %mask to <8 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %4 = load <8 x i64>, <8 x i64>* %a2
   %5 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> %4
   ret <8 x i64> %5
 }

 define <8 x i64> @stack_fold_porq_maskz(<8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
 ; CHECK-LABEL: stack_fold_porq_maskz:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vorpd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = or <8 x i64> %a0, %a1
   %3 = bitcast i8 %mask to <8 x i1>
   %4 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> zeroinitializer
   ret <8 x i64> %4
 }

 define <8 x i64> @stack_fold_porq_maskz_commuted(<8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
 ; CHECK-LABEL: stack_fold_porq_maskz_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vorpd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = or <8 x i64> %a1, %a0
   %3 = bitcast i8 %mask to <8 x i1>
   %4 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> zeroinitializer
   ret <8 x i64> %4
 }

 define <8 x i64> @stack_fold_psadbw(<64 x i8> %a0, <64 x i8> %a1) {
 ; CHECK-LABEL: stack_fold_psadbw:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpsadbw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <8 x i64> @llvm.x86.avx512.psad.bw.512(<64 x i8> %a0, <64 x i8> %a1)
   ret <8 x i64> %2
 }
 declare <8 x i64> @llvm.x86.avx512.psad.bw.512(<64 x i8>, <64 x i8>) nounwind readnone

 define <8 x i64> @stack_fold_psadbw_commute(<64 x i8> %a0, <64 x i8> %a1) {
 ; CHECK-LABEL: stack_fold_psadbw_commute:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpsadbw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <8 x i64> @llvm.x86.avx512.psad.bw.512(<64 x i8> %a1, <64 x i8> %a0)
   ret <8 x i64> %2
 }

 define <64 x i8> @stack_fold_pshufb_zmm(<64 x i8> %a0, <64 x i8> %a1) {
 ; CHECK-LABEL: stack_fold_pshufb_zmm:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpshufb {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <64 x i8> @llvm.x86.avx512.pshuf.b.512(<64 x i8> %a0, <64 x i8> %a1)
   ret <64 x i8> %2
 }
 declare <64 x i8> @llvm.x86.avx512.pshuf.b.512(<64 x i8>, <64 x i8>)

 define <64 x i8> @stack_fold_pshufb_zmm_mask(<64 x i8>* %passthru, <64 x i8> %a0, <64 x i8> %a1, i64 %mask) {
 ; CHECK-LABEL: stack_fold_pshufb_zmm_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vmovdqa64 (%rdi), %zmm2
 ; CHECK-NEXT:    kmovq %rsi, %k1
 ; CHECK-NEXT:    vpshufb {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    vmovdqa64 %zmm2, %zmm0
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = load <64 x i8>, <64 x i8>* %passthru
   %3 = call <64 x i8> @llvm.x86.avx512.pshuf.b.512(<64 x i8> %a0, <64 x i8> %a1)
   %4 = bitcast i64 %mask to <64 x i1>
   %5 = select <64 x i1> %4, <64 x i8> %3, <64 x i8> %2
   ret <64 x i8> %5
 }

 define <64 x i8> @stack_fold_pshufb_zmm_maskz(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) {
 ; CHECK-LABEL: stack_fold_pshufb_zmm_maskz:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovq %rdi, %k1
 ; CHECK-NEXT:    vpshufb {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <64 x i8> @llvm.x86.avx512.pshuf.b.512(<64 x i8> %a0, <64 x i8> %a1)
   %3 = bitcast i64 %mask to <64 x i1>
   %4 = select <64 x i1> %3, <64 x i8> %2, <64 x i8> zeroinitializer
   ret <64 x i8> %4
 }

 define <16 x i32> @stack_fold_pshufd_zmm(<16 x i32> %a0) {
 ; CHECK-LABEL: stack_fold_pshufd_zmm:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpshufd $27, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    # zmm0 = mem[3,2,1,0,7,6,5,4,11,10,9,8,15,14,13,12]
 ; CHECK-NEXT:    vpternlogd $255, %zmm1, %zmm1, %zmm1
 ; CHECK-NEXT:    vpsubd %zmm1, %zmm0, %zmm0
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = shufflevector <16 x i32> %a0, <16 x i32> undef, <16 x i32> <i32 3, i32 2, i32 1, i32 0, i32 7, i32 6, i32 5, i32 4, i32 11, i32 10, i32 9, i32 8, i32 15, i32 14, i32 13, i32 12>
   %3 = add <16 x i32> %2, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
   ret <16 x i32> %3
 }

 define <16 x i32> @stack_fold_pshufd_zmm_mask(<16 x i32> %passthru, <16 x i32> %a0, i16 %mask) {
 ; CHECK-LABEL: stack_fold_pshufd_zmm_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    subq $56, %rsp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 64
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
 ; CHECK-NEXT:    vpshufd $27, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    # zmm0 {%k1} = mem[3,2,1,0,7,6,5,4,11,10,9,8,15,14,13,12]
 ; CHECK-NEXT:    addq $56, %rsp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 8
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = shufflevector <16 x i32> %a0, <16 x i32> undef, <16 x i32> <i32 3, i32 2, i32 1, i32 0, i32 7, i32 6, i32 5, i32 4, i32 11, i32 10, i32 9, i32 8, i32 15, i32 14, i32 13, i32 12>
   %3 = bitcast i16 %mask to <16 x i1>
   %4 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> %passthru
   ret <16 x i32> %4
 }

 define <16 x i32> @stack_fold_pshufd_zmm_maskz(<16 x i32> %a0, i16 %mask) {
 ; CHECK-LABEL: stack_fold_pshufd_zmm_maskz:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpshufd $27, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    # zmm0 {%k1} {z} = mem[3,2,1,0,7,6,5,4,11,10,9,8,15,14,13,12]
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = shufflevector <16 x i32> %a0, <16 x i32> undef, <16 x i32> <i32 3, i32 2, i32 1, i32 0, i32 7, i32 6, i32 5, i32 4, i32 11, i32 10, i32 9, i32 8, i32 15, i32 14, i32 13, i32 12>
   %3 = bitcast i16 %mask to <16 x i1>
   %4 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> zeroinitializer
   ret <16 x i32> %4
 }

 define <32 x i16> @stack_fold_pshufhw_zmm(<32 x i16> %a0) {
 ; CHECK-LABEL: stack_fold_pshufhw_zmm:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpshufhw $27, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    # zmm0 = mem[0,1,2,3,7,6,5,4,8,9,10,11,15,14,13,12,16,17,18,19,23,22,21,20,24,25,26,27,31,30,29,28]
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = shufflevector <32 x i16> %a0, <32 x i16> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 7, i32 6, i32 5, i32 4, i32 8, i32 9, i32 10, i32 11, i32 15, i32 14, i32 13, i32 12, i32 16, i32 17, i32 18, i32 19, i32 23, i32 22, i32 21, i32 20, i32 24, i32 25, i32 26, i32 27, i32 31, i32 30, i32 29, i32 28>
   ret <32 x i16> %2
 }

 define <32 x i16> @stack_fold_pshufhw_zmm_mask(<32 x i16> %passthru, <32 x i16> %a0, i32 %mask) {
 ; CHECK-LABEL: stack_fold_pshufhw_zmm_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    subq $56, %rsp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 64
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
 ; CHECK-NEXT:    vpshufhw $27, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    # zmm0 {%k1} = mem[0,1,2,3,7,6,5,4,8,9,10,11,15,14,13,12,16,17,18,19,23,22,21,20,24,25,26,27,31,30,29,28]
 ; CHECK-NEXT:    addq $56, %rsp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 8
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = shufflevector <32 x i16> %a0, <32 x i16> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 7, i32 6, i32 5, i32 4, i32 8, i32 9, i32 10, i32 11, i32 15, i32 14, i32 13, i32 12, i32 16, i32 17, i32 18, i32 19, i32 23, i32 22, i32 21, i32 20, i32 24, i32 25, i32 26, i32 27, i32 31, i32 30, i32 29, i32 28>
   %3 = bitcast i32 %mask to <32 x i1>
   %4 = select <32 x i1> %3, <32 x i16> %2, <32 x i16> %passthru
   ret <32 x i16> %4
 }

 define <32 x i16> @stack_fold_pshufhw_zmm_maskz(<32 x i16> %a0, i32 %mask) {
 ; CHECK-LABEL: stack_fold_pshufhw_zmm_maskz:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpshufhw $27, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    # zmm0 {%k1} {z} = mem[0,1,2,3,7,6,5,4,8,9,10,11,15,14,13,12,16,17,18,19,23,22,21,20,24,25,26,27,31,30,29,28]
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = shufflevector <32 x i16> %a0, <32 x i16> undef, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 7, i32 6, i32 5, i32 4, i32 8, i32 9, i32 10, i32 11, i32 15, i32 14, i32 13, i32 12, i32 16, i32 17, i32 18, i32 19, i32 23, i32 22, i32 21, i32 20, i32 24, i32 25, i32 26, i32 27, i32 31, i32 30, i32 29, i32 28>
   %3 = bitcast i32 %mask to <32 x i1>
   %4 = select <32 x i1> %3, <32 x i16> %2, <32 x i16> zeroinitializer
   ret <32 x i16> %4
 }

 define <32 x i16> @stack_fold_pshuflw_zmm(<32 x i16> %a0) {
 ; CHECK-LABEL: stack_fold_pshuflw_zmm:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpshuflw $27, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    # zmm0 = mem[3,2,1,0,4,5,6,7,11,10,9,8,12,13,14,15,19,18,17,16,20,21,22,23,27,26,25,24,28,29,30,31]
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = shufflevector <32 x i16> %a0, <32 x i16> undef, <32 x i32> <i32 3, i32 2, i32 1, i32 0, i32 4, i32 5, i32 6, i32 7, i32 11, i32 10, i32 9, i32 8, i32 12, i32 13, i32 14, i32 15, i32 19, i32 18, i32 17, i32 16, i32 20, i32 21, i32 22, i32 23, i32 27, i32 26, i32 25, i32 24, i32 28, i32 29, i32 30, i32 31>
   ret <32 x i16> %2
 }

 define <32 x i16> @stack_fold_pshuflw_zmm_mask(<32 x i16> %passthru, <32 x i16> %a0, i32 %mask) {
 ; CHECK-LABEL: stack_fold_pshuflw_zmm_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    subq $56, %rsp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 64
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
 ; CHECK-NEXT:    vpshuflw $27, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    # zmm0 {%k1} = mem[3,2,1,0,4,5,6,7,11,10,9,8,12,13,14,15,19,18,17,16,20,21,22,23,27,26,25,24,28,29,30,31]
 ; CHECK-NEXT:    addq $56, %rsp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 8
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = shufflevector <32 x i16> %a0, <32 x i16> undef, <32 x i32> <i32 3, i32 2, i32 1, i32 0, i32 4, i32 5, i32 6, i32 7, i32 11, i32 10, i32 9, i32 8, i32 12, i32 13, i32 14, i32 15, i32 19, i32 18, i32 17, i32 16, i32 20, i32 21, i32 22, i32 23, i32 27, i32 26, i32 25, i32 24, i32 28, i32 29, i32 30, i32 31>
   %3 = bitcast i32 %mask to <32 x i1>
   %4 = select <32 x i1> %3, <32 x i16> %2, <32 x i16> %passthru
   ret <32 x i16> %4
 }

 define <32 x i16> @stack_fold_pshuflw_zmm_maskz(<32 x i16> %a0, i32 %mask) {
 ; CHECK-LABEL: stack_fold_pshuflw_zmm_maskz:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpshuflw $27, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    # zmm0 {%k1} {z} = mem[3,2,1,0,4,5,6,7,11,10,9,8,12,13,14,15,19,18,17,16,20,21,22,23,27,26,25,24,28,29,30,31]
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = shufflevector <32 x i16> %a0, <32 x i16> undef, <32 x i32> <i32 3, i32 2, i32 1, i32 0, i32 4, i32 5, i32 6, i32 7, i32 11, i32 10, i32 9, i32 8, i32 12, i32 13, i32 14, i32 15, i32 19, i32 18, i32 17, i32 16, i32 20, i32 21, i32 22, i32 23, i32 27, i32 26, i32 25, i32 24, i32 28, i32 29, i32 30, i32 31>
   %3 = bitcast i32 %mask to <32 x i1>
   %4 = select <32 x i1> %3, <32 x i16> %2, <32 x i16> zeroinitializer
   ret <32 x i16> %4
 }

 define <16 x i32> @stack_fold_pslld(<16 x i32> %a0, <4 x i32> %a1) {
 ; CHECK-LABEL: stack_fold_pslld:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpslld {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 16-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <16 x i32> @llvm.x86.avx512.psll.d.512(<16 x i32> %a0, <4 x i32> %a1)
   ret <16 x i32> %2
 }
 declare <16 x i32> @llvm.x86.avx512.psll.d.512(<16 x i32>, <4 x i32>) nounwind readnone

 define <16 x i32> @stack_fold_pslld_mask(<16 x i32>* %passthru, <16 x i32> %a0, <4 x i32> %a1, i16 %mask) {
 ; CHECK-LABEL: stack_fold_pslld_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %esi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rdi), %zmm2
 ; CHECK-NEXT:    vpslld {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 {%k1} # 16-byte Folded Reload
 ; CHECK-NEXT:    vmovdqa64 %zmm2, %zmm0
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <16 x i32> @llvm.x86.avx512.psll.d.512(<16 x i32> %a0, <4 x i32> %a1)
   %3 = bitcast i16 %mask to <16 x i1>
   %4 = load <16 x i32>, <16 x i32>* %passthru
   %5 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> %4
   ret <16 x i32> %5
 }

 define <16 x i32> @stack_fold_pslld_maskz(<16 x i32> %a0, <4 x i32> %a1, i16 %mask) {
 ; CHECK-LABEL: stack_fold_pslld_maskz:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpslld {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 16-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <16 x i32> @llvm.x86.avx512.psll.d.512(<16 x i32> %a0, <4 x i32> %a1)
   %3 = bitcast i16 %mask to <16 x i1>
   %4 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> zeroinitializer
   ret <16 x i32> %4
 }

 define <16 x i32> @stack_fold_pslldi(<16 x i32> %a0) {
 ; CHECK-LABEL: stack_fold_pslldi:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpslld $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <16 x i32> @llvm.x86.avx512.pslli.d.512(<16 x i32> %a0, i32 1)
   ret <16 x i32> %2
 }
 declare <16 x i32> @llvm.x86.avx512.pslli.d.512(<16 x i32>, i32) nounwind readnone

 define <16 x i32> @stack_fold_pslldi_mask(<16 x i32>* %passthru, <16 x i32> %a0, i16 %mask) {
 ; CHECK-LABEL: stack_fold_pslldi_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %esi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rdi), %zmm1
 ; CHECK-NEXT:    vpslld $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm1 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    vmovdqa64 %zmm1, %zmm0
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <16 x i32> @llvm.x86.avx512.pslli.d.512(<16 x i32> %a0, i32 1)
   %3 = bitcast i16 %mask to <16 x i1>
   %4 = load <16 x i32>, <16 x i32>* %passthru
   %5 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> %4
   ret <16 x i32> %5
 }

 define <16 x i32> @stack_fold_pslldi_maskz(<16 x i32> %a0, i16 %mask) {
 ; CHECK-LABEL: stack_fold_pslldi_maskz:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpslld $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <16 x i32> @llvm.x86.avx512.pslli.d.512(<16 x i32> %a0, i32 1)
   %3 = bitcast i16 %mask to <16 x i1>
   %4 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> zeroinitializer
   ret <16 x i32> %4
 }

 define <64 x i8> @stack_fold_pslldq(<64 x i8> %a, <64 x i8> %b) {
 ; CHECK-LABEL: stack_fold_pslldq:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpslldq $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    # zmm0 = zero,mem[0,1,2,3,4,5,6,7,8,9,10,11,12,13,14],zero,mem[16,17,18,19,20,21,22,23,24,25,26,27,28,29,30],zero,mem[32,33,34,35,36,37,38,39,40,41,42,43,44,45,46],zero,mem[48,49,50,51,52,53,54,55,56,57,58,59,60,61,62]
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = shufflevector <64 x i8> %a, <64 x i8> zeroinitializer, <64 x i32> <i32 79, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 95, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 111, i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 127, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62>
   ret <64 x i8> %2
 }

 define <8 x i64> @stack_fold_psllq(<8 x i64> %a0, <2 x i64> %a1) {
 ; CHECK-LABEL: stack_fold_psllq:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpsllq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 16-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <8 x i64> @llvm.x86.avx512.psll.q.512(<8 x i64> %a0, <2 x i64> %a1)
   ret <8 x i64> %2
 }
 declare <8 x i64> @llvm.x86.avx512.psll.q.512(<8 x i64>, <2 x i64>) nounwind readnone

 define <8 x i64> @stack_fold_psllqi(<8 x i64> %a0) {
 ; CHECK-LABEL: stack_fold_psllqi:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpsllq $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <8 x i64> @llvm.x86.avx512.pslli.q.512(<8 x i64> %a0, i32 1)
   ret <8 x i64> %2
 }
 declare <8 x i64> @llvm.x86.avx512.pslli.q.512(<8 x i64>, i32) nounwind readnone

 define <16 x i32> @stack_fold_psllvd(<16 x i32> %a0, <16 x i32> %a1) {
 ; CHECK-LABEL: stack_fold_psllvd:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpsllvd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <16 x i32> @llvm.x86.avx512.psllv.d.512(<16 x i32> %a0, <16 x i32> %a1)
   ret <16 x i32> %2
 }
 declare <16 x i32> @llvm.x86.avx512.psllv.d.512(<16 x i32>, <16 x i32>) nounwind readnone

 define <16 x i32> @stack_fold_psllvd_mask(<16 x i32>* %passthru, <16 x i32> %a0, <16 x i32> %a1, i16 %mask) {
 ; CHECK-LABEL: stack_fold_psllvd_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %esi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rdi), %zmm2
 ; CHECK-NEXT:    vpsllvd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    vmovdqa64 %zmm2, %zmm0
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <16 x i32> @llvm.x86.avx512.psllv.d.512(<16 x i32> %a0, <16 x i32> %a1)
   %3 = bitcast i16 %mask to <16 x i1>
   %4 = load <16 x i32>, <16 x i32>* %passthru
   %5 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> %4
   ret <16 x i32> %5
 }

 define <16 x i32> @stack_fold_psllvd_maskz(<16 x i32> %a0, <16 x i32> %a1, i16 %mask) {
 ; CHECK-LABEL: stack_fold_psllvd_maskz:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vpsllvd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <16 x i32> @llvm.x86.avx512.psllv.d.512(<16 x i32> %a0, <16 x i32> %a1)
   %3 = bitcast i16 %mask to <16 x i1>
   %4 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> zeroinitializer
   ret <16 x i32> %4
 }

 define <8 x i64> @stack_fold_psllvq(<8 x i64> %a0, <8 x i64> %a1) {
 ; CHECK-LABEL: stack_fold_psllvq:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpsllvq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <8 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <8 x i64> @llvm.x86.avx512.psllv.q.512(<8 x i64> %a0, <8 x i64> %a1)
   ret <8 x i64> %2
 }
 declare <8 x i64> @llvm.x86.avx512.psllv.q.512(<8 x i64>, <8 x i64>) nounwind readnone

 define <32 x i16> @stack_fold_psllvw(<32 x i16> %a0, <32 x i16> %a1) {
 ; CHECK-LABEL: stack_fold_psllvw:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpsllvw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <32 x i16> @llvm.x86.avx512.psllv.w.512(<32 x i16> %a0, <32 x i16> %a1)
   ret <32 x i16> %2
 }
 declare <32 x i16> @llvm.x86.avx512.psllv.w.512(<32 x i16>, <32 x i16>) nounwind readnone

 define <32 x i16> @stack_fold_psllw(<32 x i16> %a0, <8 x i16> %a1) {
 ; CHECK-LABEL: stack_fold_psllw:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpsllw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 16-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <32 x i16> @llvm.x86.avx512.psll.w.512(<32 x i16> %a0, <8 x i16> %a1)
   ret <32 x i16> %2
 }
 declare <32 x i16> @llvm.x86.avx512.psll.w.512(<32 x i16>, <8 x i16>) nounwind readnone

 define <32 x i16> @stack_fold_psllwi(<32 x i16> %a0) {
 ; CHECK-LABEL: stack_fold_psllwi:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpsllw $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <32 x i16> @llvm.x86.avx512.pslli.w.512(<32 x i16> %a0, i32 1)
   ret <32 x i16> %2
 }
 declare <32 x i16> @llvm.x86.avx512.pslli.w.512(<32 x i16>, i32) nounwind readnone

 define <16 x i32> @stack_fold_psrad(<16 x i32> %a0, <4 x i32> %a1) {
 ; CHECK-LABEL: stack_fold_psrad:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpsrad {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 16-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <16 x i32> @llvm.x86.avx512.psra.d.512(<16 x i32> %a0, <4 x i32> %a1)
   ret <16 x i32> %2
 }
 declare <16 x i32> @llvm.x86.avx512.psra.d.512(<16 x i32>, <4 x i32>) nounwind readnone

 define <16 x i32> @stack_fold_psradi(<16 x i32> %a0) {
 ; CHECK-LABEL: stack_fold_psradi:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpsrad $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <16 x i32> @llvm.x86.avx512.psrai.d.512(<16 x i32> %a0, i32 1)
   ret <16 x i32> %2
 }
 declare <16 x i32> @llvm.x86.avx512.psrai.d.512(<16 x i32>, i32) nounwind readnone

 define <8 x i64> @stack_fold_psraq(<8 x i64> %a0, <2 x i64> %a1) {
 ; CHECK-LABEL: stack_fold_psraq:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpsraq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 16-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <8 x i64> @llvm.x86.avx512.psra.q.512(<8 x i64> %a0, <2 x i64> %a1)
   ret <8 x i64> %2
 }
 declare <8 x i64> @llvm.x86.avx512.psra.q.512(<8 x i64>, <2 x i64>) nounwind readnone

 define <8 x i64> @stack_fold_psraqi(<8 x i64> %a0) {
 ; CHECK-LABEL: stack_fold_psraqi:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpsraq $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <8 x i64> @llvm.x86.avx512.psrai.q.512(<8 x i64> %a0, i32 1)
   ret <8 x i64> %2
 }
 declare <8 x i64> @llvm.x86.avx512.psrai.q.512(<8 x i64>, i32) nounwind readnone

 define <16 x i32> @stack_fold_psravd(<16 x i32> %a0, <16 x i32> %a1) {
 ; CHECK-LABEL: stack_fold_psravd:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpsravd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <16 x i32> @llvm.x86.avx512.psrav.d.512(<16 x i32> %a0, <16 x i32> %a1)
   ret <16 x i32> %2
 }
 declare <16 x i32> @llvm.x86.avx512.psrav.d.512(<16 x i32>, <16 x i32>) nounwind readnone

 define <8 x i64> @stack_fold_psravq(<8 x i64> %a0, <8 x i64> %a1) {
 ; CHECK-LABEL: stack_fold_psravq:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpsravq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <8 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <8 x i64> @llvm.x86.avx512.psrav.q.512(<8 x i64> %a0, <8 x i64> %a1)
   ret <8 x i64> %2
 }
 declare <8 x i64> @llvm.x86.avx512.psrav.q.512(<8 x i64>, <8 x i64>) nounwind readnone

 define <32 x i16> @stack_fold_psravw(<32 x i16> %a0, <32 x i16> %a1) {
 ; CHECK-LABEL: stack_fold_psravw:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpsravw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <32 x i16> @llvm.x86.avx512.psrav.w.512(<32 x i16> %a0, <32 x i16> %a1)
   ret <32 x i16> %2
 }
 declare <32 x i16> @llvm.x86.avx512.psrav.w.512(<32 x i16>, <32 x i16>) nounwind readnone

 define <32 x i16> @stack_fold_psraw(<32 x i16> %a0, <8 x i16> %a1) {
 ; CHECK-LABEL: stack_fold_psraw:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpsraw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 16-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <32 x i16> @llvm.x86.avx512.psra.w.512(<32 x i16> %a0, <8 x i16> %a1)
   ret <32 x i16> %2
 }
 declare <32 x i16> @llvm.x86.avx512.psra.w.512(<32 x i16>, <8 x i16>) nounwind readnone

 define <32 x i16> @stack_fold_psrawi(<32 x i16> %a0) {
 ; CHECK-LABEL: stack_fold_psrawi:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpsraw $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <32 x i16> @llvm.x86.avx512.psrai.w.512(<32 x i16> %a0, i32 1)
   ret <32 x i16> %2
 }
 declare <32 x i16> @llvm.x86.avx512.psrai.w.512(<32 x i16>, i32) nounwind readnone

 define <16 x i32> @stack_fold_psrld(<16 x i32> %a0, <4 x i32> %a1) {
 ; CHECK-LABEL: stack_fold_psrld:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpsrld {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 16-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <16 x i32> @llvm.x86.avx512.psrl.d.512(<16 x i32> %a0, <4 x i32> %a1)
   ret <16 x i32> %2
 }
 declare <16 x i32> @llvm.x86.avx512.psrl.d.512(<16 x i32>, <4 x i32>) nounwind readnone

 define <16 x i32> @stack_fold_psrldi(<16 x i32> %a0) {
 ; CHECK-LABEL: stack_fold_psrldi:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpsrld $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <16 x i32> @llvm.x86.avx512.psrli.d.512(<16 x i32> %a0, i32 1)
   ret <16 x i32> %2
 }
 declare <16 x i32> @llvm.x86.avx512.psrli.d.512(<16 x i32>, i32) nounwind readnone

 define <64 x i8> @stack_fold_psrldq(<64 x i8> %a, <64 x i8> %b) {
 ; CHECK-LABEL: stack_fold_psrldq:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpsrldq $2, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    # zmm0 = mem[2,3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero,mem[18,19,20,21,22,23,24,25,26,27,28,29,30,31],zero,zero,mem[34,35,36,37,38,39,40,41,42,43,44,45,46,47],zero,zero,mem[50,51,52,53,54,55,56,57,58,59,60,61,62,63],zero,zero
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = shufflevector <64 x i8> %a, <64 x i8> zeroinitializer, <64 x i32> <i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 64, i32 64, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 64, i32 64, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 64, i32 64, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63, i32 64, i32 64>
   ret <64 x i8> %2
 }

 define <8 x i64> @stack_fold_psrlq(<8 x i64> %a0, <2 x i64> %a1) {
 ; CHECK-LABEL: stack_fold_psrlq:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpsrlq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 16-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <8 x i64> @llvm.x86.avx512.psrl.q.512(<8 x i64> %a0, <2 x i64> %a1)
   ret <8 x i64> %2
 }
 declare <8 x i64> @llvm.x86.avx512.psrl.q.512(<8 x i64>, <2 x i64>) nounwind readnone

 define <8 x i64> @stack_fold_psrlqi(<8 x i64> %a0) {
 ; CHECK-LABEL: stack_fold_psrlqi:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpsrlq $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <8 x i64> @llvm.x86.avx512.psrli.q.512(<8 x i64> %a0, i32 1)
   ret <8 x i64> %2
 }
 declare <8 x i64> @llvm.x86.avx512.psrli.q.512(<8 x i64>, i32) nounwind readnone

 define <16 x i32> @stack_fold_psrlvd(<16 x i32> %a0, <16 x i32> %a1) {
 ; CHECK-LABEL: stack_fold_psrlvd:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpsrlvd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <16 x i32> @llvm.x86.avx512.psrlv.d.512(<16 x i32> %a0, <16 x i32> %a1)
   ret <16 x i32> %2
 }
 declare <16 x i32> @llvm.x86.avx512.psrlv.d.512(<16 x i32>, <16 x i32>) nounwind readnone

 define <8 x i64> @stack_fold_psrlvq(<8 x i64> %a0, <8 x i64> %a1) {
 ; CHECK-LABEL: stack_fold_psrlvq:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpsrlvq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <8 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <8 x i64> @llvm.x86.avx512.psrlv.q.512(<8 x i64> %a0, <8 x i64> %a1)
   ret <8 x i64> %2
 }
 declare <8 x i64> @llvm.x86.avx512.psrlv.q.512(<8 x i64>, <8 x i64>) nounwind readnone

 define <32 x i16> @stack_fold_psrlvw(<32 x i16> %a0, <32 x i16> %a1) {
 ; CHECK-LABEL: stack_fold_psrlvw:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpsrlvw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <32 x i16> @llvm.x86.avx512.psrlv.w.512(<32 x i16> %a0, <32 x i16> %a1)
   ret <32 x i16> %2
 }
 declare <32 x i16> @llvm.x86.avx512.psrlv.w.512(<32 x i16>, <32 x i16>) nounwind readnone

 define <32 x i16> @stack_fold_psrlw(<32 x i16> %a0, <8 x i16> %a1) {
 ; CHECK-LABEL: stack_fold_psrlw:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpsrlw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 16-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <32 x i16> @llvm.x86.avx512.psrl.w.512(<32 x i16> %a0, <8 x i16> %a1)
   ret <32 x i16> %2
 }
 declare <32 x i16> @llvm.x86.avx512.psrl.w.512(<32 x i16>, <8 x i16>) nounwind readnone

 define <32 x i16> @stack_fold_psrlwi(<32 x i16> %a0) {
 ; CHECK-LABEL: stack_fold_psrlwi:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpsrlw $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <32 x i16> @llvm.x86.avx512.psrli.w.512(<32 x i16> %a0, i32 1)
   ret <32 x i16> %2
 }
 declare <32 x i16> @llvm.x86.avx512.psrli.w.512(<32 x i16>, i32) nounwind readnone

 define <64 x i8> @stack_fold_psubb(<64 x i8> %a0, <64 x i8> %a1) {
 ; CHECK-LABEL: stack_fold_psubb:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpsubb {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = sub <64 x i8> %a0, %a1
   ret <64 x i8> %2
 }

 define <16 x i32> @stack_fold_psubd(<16 x i32> %a0, <16 x i32> %a1) {
 ; CHECK-LABEL: stack_fold_psubd:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpsubd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = sub <16 x i32> %a0, %a1
   ret <16 x i32> %2
 }

 define <8 x i64> @stack_fold_psubq(<8 x i64> %a0, <8 x i64> %a1) {
 ; CHECK-LABEL: stack_fold_psubq:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpsubq {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = sub <8 x i64> %a0, %a1
   ret <8 x i64> %2
 }

 define <64 x i8> @stack_fold_psubsb(<64 x i8> %a0, <64 x i8> %a1) {
 ; CHECK-LABEL: stack_fold_psubsb:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpsubsb {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <64 x i8> @llvm.ssub.sat.v64i8(<64 x i8> %a0, <64 x i8> %a1)
   ret <64 x i8> %2
 }

 define <32 x i16> @stack_fold_psubsw(<32 x i16> %a0, <32 x i16> %a1) {
 ; CHECK-LABEL: stack_fold_psubsw:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpsubsw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <32 x i16> @llvm.ssub.sat.v32i16(<32 x i16> %a0, <32 x i16> %a1)
   ret <32 x i16> %2
 }

 define <64 x i8> @stack_fold_psubusb(<64 x i8> %a0, <64 x i8> %a1) {
 ; CHECK-LABEL: stack_fold_psubusb:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpsubusb {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <64 x i8> @llvm.usub.sat.v64i8(<64 x i8> %a0, <64 x i8> %a1)
   ret <64 x i8> %2
 }

 define <32 x i16> @stack_fold_psubusw(<32 x i16> %a0, <32 x i16> %a1) {
 ; CHECK-LABEL: stack_fold_psubusw:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpsubusw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <32 x i16> @llvm.usub.sat.v32i16(<32 x i16> %a0, <32 x i16> %a1)
   ret <32 x i16> %2
 }

 define <32 x i16> @stack_fold_psubw(<32 x i16> %a0, <32 x i16> %a1) {
 ; CHECK-LABEL: stack_fold_psubw:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpsubw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = sub <32 x i16> %a0, %a1
   ret <32 x i16> %2
 }

 define <8 x i64> @stack_fold_shufi64x2(<8 x i64> %a, <8 x i64> %b) {
 ; CHECK-LABEL: stack_fold_shufi64x2:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    subq $56, %rsp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 64
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
 ; CHECK-NEXT:    vshufi64x2 $24, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    # zmm0 = zmm0[0,1,4,5],mem[2,3,0,1]
 ; CHECK-NEXT:    addq $56, %rsp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 8
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 0, i32 1, i32 4, i32 5, i32 10, i32 11, i32 8, i32 9>
   ret <8 x i64> %2
 }

 define <8 x i64> @stack_fold_shufi64x2_mask(<8 x i64> %a, <8 x i64> %b, i8 %mask, <8 x i64>* %passthru) {
 ; CHECK-LABEL: stack_fold_shufi64x2_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    subq $56, %rsp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 64
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rsi), %zmm1
 ; CHECK-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
 ; CHECK-NEXT:    vshufi64x2 $24, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm1 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    # zmm1 {%k1} = zmm0[0,1,4,5],mem[2,3,0,1]
 ; CHECK-NEXT:    vmovdqa64 %zmm1, %zmm0
 ; CHECK-NEXT:    addq $56, %rsp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 8
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 0, i32 1, i32 4, i32 5, i32 10, i32 11, i32 8, i32 9>
   %3 = bitcast i8 %mask to <8 x i1>
   ; load needed to keep the operation from being scheduled above the asm block
   %4 = load <8 x i64>, <8 x i64>* %passthru
   %5 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> %4
   ret <8 x i64> %5
 }

 define <8 x i64> @stack_fold_shufi64x2_maskz(<8 x i64> %a, <8 x i64> %b, i8 %mask, <8 x i64>* %passthru) {
 ; CHECK-LABEL: stack_fold_shufi64x2_maskz:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    subq $56, %rsp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 64
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
 ; CHECK-NEXT:    vshufi64x2 $24, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    # zmm0 {%k1} {z} = zmm0[0,1,4,5],mem[2,3,0,1]
 ; CHECK-NEXT:    addq $56, %rsp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 8
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = shufflevector <8 x i64> %a, <8 x i64> %b, <8 x i32> <i32 0, i32 1, i32 4, i32 5, i32 10, i32 11, i32 8, i32 9>
   %3 = bitcast i8 %mask to <8 x i1>
   %4 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> zeroinitializer
   ret <8 x i64> %4
 }

 define <16 x i32> @stack_fold_shufi32x4_mask(<16 x i32> %a, <16 x i32> %b, i16 %mask, <16 x i32>* %passthru) {
 ; CHECK-LABEL: stack_fold_shufi32x4_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    subq $56, %rsp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 64
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rsi), %zmm1
 ; CHECK-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
 ; CHECK-NEXT:    vshufi32x4 $20, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm1 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    # zmm1 {%k1} = zmm0[0,1,2,3,4,5,6,7],mem[4,5,6,7,0,1,2,3]
 ; CHECK-NEXT:    vmovdqa64 %zmm1, %zmm0
 ; CHECK-NEXT:    addq $56, %rsp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 8
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = shufflevector <16 x i32> %a, <16 x i32> %b, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 20, i32 21, i32 22, i32 23, i32 16, i32 17, i32 18, i32 19>
   %3 = bitcast i16 %mask to <16 x i1>
   ; load needed to keep the operation from being scheduled above the asm block
   %4 = load <16 x i32>, <16 x i32>* %passthru
   %5 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> %4
   ret <16 x i32> %5
 }

 define <16 x i32> @stack_fold_shufi32x4_maskz(<16 x i32> %a, <16 x i32> %b, i16 %mask) {
 ; CHECK-LABEL: stack_fold_shufi32x4_maskz:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    subq $56, %rsp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 64
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload
 ; CHECK-NEXT:    vshufi32x4 $20, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    # zmm0 {%k1} {z} = zmm0[0,1,2,3,4,5,6,7],mem[4,5,6,7,0,1,2,3]
 ; CHECK-NEXT:    addq $56, %rsp
 ; CHECK-NEXT:    .cfi_def_cfa_offset 8
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm1},~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = shufflevector <16 x i32> %a, <16 x i32> %b, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 20, i32 21, i32 22, i32 23, i32 16, i32 17, i32 18, i32 19>
   %3 = bitcast i16 %mask to <16 x i1>
   %4 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> zeroinitializer
   ret <16 x i32> %4
 }

 define <16 x i32> @stack_fold_ternlogd(<16 x i32> %x0, <16 x i32> %x1, <16 x i32> %x2) {
 ; CHECK-LABEL: stack_fold_ternlogd:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpternlogd $33, {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <16 x i32> @llvm.x86.avx512.pternlog.d.512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32> %x2, i32 33)
   ret <16 x i32> %2
 }
 declare <16 x i32> @llvm.x86.avx512.pternlog.d.512(<16 x i32>, <16 x i32>, <16 x i32>, i32)

 define <8 x i64> @stack_fold_ternlogq(<8 x i64> %x0, <8 x i64> %x1, <8 x i64> %x2) {
 ; CHECK-LABEL: stack_fold_ternlogq:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm2, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpternlogq $33, {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = call <8 x i64> @llvm.x86.avx512.pternlog.q.512(<8 x i64> %x0, <8 x i64> %x1, <8 x i64> %x2, i32 33)
   ret <8 x i64> %2
 }

 declare <8 x i64> @llvm.x86.avx512.pternlog.q.512(<8 x i64>, <8 x i64>, <8 x i64>, i32)

 define <64 x i8> @stack_fold_punpckhbw_zmm(<64 x i8> %a0, <64 x i8> %a1) {
 ; CHECK-LABEL: stack_fold_punpckhbw_zmm:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vpunpckhbw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    # zmm0 = zmm0[8],mem[8],zmm0[9],mem[9],zmm0[10],mem[10],zmm0[11],mem[11],zmm0[12],mem[12],zmm0[13],mem[13],zmm0[14],mem[14],zmm0[15],mem[15],zmm0[24],mem[24],zmm0[25],mem[25],zmm0[26],mem[26],zmm0[27],mem[27],zmm0[28],mem[28],zmm0[29],mem[29],zmm0[30],mem[30],zmm0[31],mem[31],zmm0[40],mem[40],zmm0[41],mem[41],zmm0[42],mem[42],zmm0[43],mem[43],zmm0[44],mem[44],zmm0[45],mem[45],zmm0[46],mem[46],zmm0[47],mem[47],zmm0[56],mem[56],zmm0[57],mem[57],zmm0[58],mem[58],zmm0[59],mem[59],zmm0[60],mem[60],zmm0[61],mem[61],zmm0[62],mem[62],zmm0[63],mem[63]
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = shufflevector <64 x i8> %a0, <64 x i8> %a1, <64 x i32> <i32 8, i32 72, i32 9, i32 73, i32 10, i32 74, i32 11, i32 75, i32 12, i32 76, i32 13, i32 77, i32 14, i32 78, i32 15, i32 79, i32 24, i32 88, i32 25, i32 89, i32 26, i32 90, i32 27, i32 91, i32 28, i32 92, i32 29, i32 93, i32 30, i32 94, i32 31, i32 95, i32 40, i32 104, i32 41, i32 105, i32 42, i32 106, i32 43, i32 107, i32 44, i32 108, i32 45, i32 109, i32 46, i32 110, i32 47, i32 111, i32 56, i32 120, i32 57, i32 121, i32 58, i32 122, i32 59, i32 123, i32 60, i32 124, i32 61, i32 125, i32 62, i32 126, i32 63, i32 127>
   ret <64 x i8> %2
 }

 define <64 x i8> @stack_fold_punpckhbw_mask_zmm(<64 x i8>* %passthru, <64 x i8> %a0, <64 x i8> %a1, i64 %mask) {
 ; CHECK-LABEL: stack_fold_punpckhbw_mask_zmm:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovq %rsi, %k1
 ; CHECK-NEXT:    vmovdqa64 (%rdi), %zmm2
 ; CHECK-NEXT:    vpunpckhbw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm2 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    # zmm2 {%k1} = zmm0[8],mem[8],zmm0[9],mem[9],zmm0[10],mem[10],zmm0[11],mem[11],zmm0[12],mem[12],zmm0[13],mem[13],zmm0[14],mem[14],zmm0[15],mem[15],zmm0[24],mem[24],zmm0[25],mem[25],zmm0[26],mem[26],zmm0[27],mem[27],zmm0[28],mem[28],zmm0[29],mem[29],zmm0[30],mem[30],zmm0[31],mem[31],zmm0[40],mem[40],zmm0[41],mem[41],zmm0[42],mem[42],zmm0[43],mem[43],zmm0[44],mem[44],zmm0[45],mem[45],zmm0[46],mem[46],zmm0[47],mem[47],zmm0[56],mem[56],zmm0[57],mem[57],zmm0[58],mem[58],zmm0[59],mem[59],zmm0[60],mem[60],zmm0[61],mem[61],zmm0[62],mem[62],zmm0[63],mem[63]
 ; CHECK-NEXT:    vmovdqa64 %zmm2, %zmm0
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = shufflevector <64 x i8> %a0, <64 x i8> %a1, <64 x i32> <i32 8, i32 72, i32 9, i32 73, i32 10, i32 74, i32 11, i32 75, i32 12, i32 76, i32 13, i32 77, i32 14, i32 78, i32 15, i32 79, i32 24, i32 88, i32 25, i32 89, i32 26, i32 90, i32 27, i32 91, i32 28, i32 92, i32 29, i32 93, i32 30, i32 94, i32 31, i32 95, i32 40, i32 104, i32 41, i32 105, i32 42, i32 106, i32 43, i32 107, i32 44, i32 108, i32 45, i32 109, i32 46, i32 110, i32 47, i32 111, i32 56, i32 120, i32 57, i32 121, i32 58, i32 122, i32 59, i32 123, i32 60, i32 124, i32 61, i32 125, i32 62, i32 126, i32 63, i32 127>
   %3 = bitcast i64 %mask to <64 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %4 = load <64 x i8>, <64 x i8>* %passthru
   %5 = select <64 x i1> %3, <64 x i8> %2, <64 x i8> %4
   ret <64 x i8> %5
 }

 define <64 x i8> @stack_fold_punpckhbw_maskz_zmm(<64 x i8> %a0, <64 x i8> %a1, i64 %mask) {
 ; CHECK-LABEL: stack_fold_punpckhbw_maskz_zmm:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovq %rdi, %k1
 ; CHECK-NEXT:    vpunpckhbw {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    # zmm0 {%k1} {z} = zmm0[8],mem[8],zmm0[9],mem[9],zmm0[10],mem[10],zmm0[11],mem[11],zmm0[12],mem[12],zmm0[13],mem[13],zmm0[14],mem[14],zmm0[15],mem[15],zmm0[24],mem[24],zmm0[25],mem[25],zmm0[26],mem[26],zmm0[27],mem[27],zmm0[28],mem[28],zmm0[29],mem[29],zmm0[30],mem[30],zmm0[31],mem[31],zmm0[40],mem[40],zmm0[41],mem[41],zmm0[42],mem[42],zmm0[43],mem[43],zmm0[44],mem[44],zmm0[45],mem[45],zmm0[46],mem[46],zmm0[47],mem[47],zmm0[56],mem[56],zmm0[57],mem[57],zmm0[58],mem[58],zmm0[59],mem[59],zmm0[60],mem[60],zmm0[61],mem[61],zmm0[62],mem[62],zmm0[63],mem[63]
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = shufflevector <64 x i8> %a0, <64 x i8> %a1, <64 x i32> <i32 8, i32 72, i32 9, i32 73, i32 10, i32 74, i32 11, i32 75, i32 12, i32 76, i32 13, i32 77, i32 14, i32 78, i32 15, i32 79, i32 24, i32 88, i32 25, i32 89, i32 26, i32 90, i32 27, i32 91, i32 28, i32 92, i32 29, i32 93, i32 30, i32 94, i32 31, i32 95, i32 40, i32 104, i32 41, i32 105, i32 42, i32 106, i32 43, i32 107, i32 44, i32 108, i32 45, i32 109, i32 46, i32 110, i32 47, i32 111, i32 56, i32 120, i32 57, i32 121, i32 58, i32 122, i32 59, i32 123, i32 60, i32 124, i32 61, i32 125, i32 62, i32 126, i32 63, i32 127>
   %3 = bitcast i64 %mask to <64 x i1>
   %4 = select <64 x i1> %3, <64 x i8> %2, <64 x i8> zeroinitializer
   ret <64 x i8> %4
 }

 define <16 x i32> @stack_fold_pxord(<16 x i32> %a0, <16 x i32> %a1) {
 ; CHECK-LABEL: stack_fold_pxord:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vxorps {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = xor <16 x i32> %a0, %a1
   ret <16 x i32> %2
 }

 define <16 x i32> @stack_fold_pxord_commuted(<16 x i32> %a0, <16 x i32> %a1) {
 ; CHECK-LABEL: stack_fold_pxord_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vxorps {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = xor <16 x i32> %a1, %a0
   ret <16 x i32> %2
 }

 define <16 x i32> @stack_fold_pxord_mask(<16 x i32> %a0, <16 x i32> %a1, <16 x i32>* %a2, i16 %mask) {
 ; CHECK-LABEL: stack_fold_pxord_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovaps %zmm0, %zmm1
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %esi, %k1
 ; CHECK-NEXT:    vmovaps (%rdi), %zmm0
 ; CHECK-NEXT:    vxorps {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = xor <16 x i32> %a0, %a1
   %3 = bitcast i16 %mask to <16 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %4 = load <16 x i32>, <16 x i32>* %a2
   %5 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> %4
   ret <16 x i32> %5
 }

 define <16 x i32> @stack_fold_pxord_mask_commuted(<16 x i32> %a0, <16 x i32> %a1, <16 x i32>* %a2, i16 %mask) {
 ; CHECK-LABEL: stack_fold_pxord_mask_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovaps %zmm0, %zmm1
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %esi, %k1
 ; CHECK-NEXT:    vmovaps (%rdi), %zmm0
 ; CHECK-NEXT:    vxorps {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = xor <16 x i32> %a1, %a0
   %3 = bitcast i16 %mask to <16 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %4 = load <16 x i32>, <16 x i32>* %a2
   %5 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> %4
   ret <16 x i32> %5
 }

 define <16 x i32> @stack_fold_pxord_maskz(<16 x i32> %a0, <16 x i32> %a1, i16 %mask) {
 ; CHECK-LABEL: stack_fold_pxord_maskz:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vxorps {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = xor <16 x i32> %a0, %a1
   %3 = bitcast i16 %mask to <16 x i1>
   %4 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> zeroinitializer
   ret <16 x i32> %4
 }

 define <16 x i32> @stack_fold_pxord_maskz_commuted(<16 x i32> %a0, <16 x i32> %a1, i16 %mask) {
 ; CHECK-LABEL: stack_fold_pxord_maskz_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vxorps {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = xor <16 x i32> %a1, %a0
   %3 = bitcast i16 %mask to <16 x i1>
   %4 = select <16 x i1> %3, <16 x i32> %2, <16 x i32> zeroinitializer
   ret <16 x i32> %4
 }

 define <8 x i64> @stack_fold_pxorq(<8 x i64> %a0, <8 x i64> %a1) {
 ; CHECK-LABEL: stack_fold_pxorq:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vxorps {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = xor <8 x i64> %a0, %a1
   ret <8 x i64> %2
 }

 define <8 x i64> @stack_fold_pxorq_commuted(<8 x i64> %a0, <8 x i64> %a1) {
 ; CHECK-LABEL: stack_fold_pxorq_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    vxorps {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = xor <8 x i64> %a1, %a0
   ret <8 x i64> %2
 }

 define <8 x i64> @stack_fold_pxorq_mask(<8 x i64> %a0, <8 x i64> %a1, <8 x i64>* %a2, i8 %mask) {
 ; CHECK-LABEL: stack_fold_pxorq_mask:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovapd %zmm0, %zmm1
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %esi, %k1
 ; CHECK-NEXT:    vmovapd (%rdi), %zmm0
 ; CHECK-NEXT:    vxorpd {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = xor <8 x i64> %a0, %a1
   %3 = bitcast i8 %mask to <8 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %4 = load <8 x i64>, <8 x i64>* %a2
   %5 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> %4
   ret <8 x i64> %5
 }

 define <8 x i64> @stack_fold_pxorq_mask_commuted(<8 x i64> %a0, <8 x i64> %a1, <8 x i64>* %a2, i8 %mask) {
 ; CHECK-LABEL: stack_fold_pxorq_mask_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    vmovapd %zmm0, %zmm1
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %esi, %k1
 ; CHECK-NEXT:    vmovapd (%rdi), %zmm0
 ; CHECK-NEXT:    vxorpd {{[-0-9]+}}(%r{{[sb]}}p), %zmm1, %zmm0 {%k1} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm0},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = xor <8 x i64> %a1, %a0
   %3 = bitcast i8 %mask to <8 x i1>
   ; load needed to keep the operation from being scheduled about the asm block
   %4 = load <8 x i64>, <8 x i64>* %a2
   %5 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> %4
   ret <8 x i64> %5
 }

 define <8 x i64> @stack_fold_pxorq_maskz(<8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
 ; CHECK-LABEL: stack_fold_pxorq_maskz:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vxorpd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = xor <8 x i64> %a0, %a1
   %3 = bitcast i8 %mask to <8 x i1>
   %4 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> zeroinitializer
   ret <8 x i64> %4
 }

 define <8 x i64> @stack_fold_pxorq_maskz_commuted(<8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
 ; CHECK-LABEL: stack_fold_pxorq_maskz_commuted:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vmovups %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill
 ; CHECK-NEXT:    #APP
 ; CHECK-NEXT:    nop
 ; CHECK-NEXT:    #NO_APP
 ; CHECK-NEXT:    kmovd %edi, %k1
 ; CHECK-NEXT:    vxorpd {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 {%k1} {z} # 64-byte Folded Reload
 ; CHECK-NEXT:    retq
   %1 = tail call <2 x i64> asm sideeffect "nop", "=x,~{xmm2},~{xmm3},~{xmm4},~{xmm5},~{xmm6},~{xmm7},~{xmm8},~{xmm9},~{xmm10},~{xmm11},~{xmm12},~{xmm13},~{xmm14},~{xmm15},~{xmm16},~{xmm17},~{xmm18},~{xmm19},~{xmm20},~{xmm21},~{xmm22},~{xmm23},~{xmm24},~{xmm25},~{xmm26},~{xmm27},~{xmm28},~{xmm29},~{xmm30},~{xmm31},~{flags}"()
   %2 = xor <8 x i64> %a1, %a0
   %3 = bitcast i8 %mask to <8 x i1>
   %4 = select <8 x i1> %3, <8 x i64> %2, <8 x i64> zeroinitializer
   ret <8 x i64> %4
 }

 declare <64 x i8> @llvm.sadd.sat.v64i8(<64 x i8>, <64 x i8>)
 declare <32 x i16> @llvm.sadd.sat.v32i16(<32 x i16>, <32 x i16>)
 declare <64 x i8> @llvm.uadd.sat.v64i8(<64 x i8>, <64 x i8>)
 declare <32 x i16> @llvm.uadd.sat.v32i16(<32 x i16>, <32 x i16>)
 declare <16 x i32> @llvm.x86.avx512.conflict.d.512(<16 x i32>)
 declare <8 x i64> @llvm.x86.avx512.conflict.q.512(<8 x i64>)
 declare <16 x i32> @llvm.ctlz.v16i32(<16 x i32>, i1)
 declare <8 x i64> @llvm.ctlz.v8i64(<8 x i64>, i1)
 declare <64 x i8> @llvm.ssub.sat.v64i8(<64 x i8>, <64 x i8>)
 declare <32 x i16> @llvm.ssub.sat.v32i16(<32 x i16>, <32 x i16>)
 declare <64 x i8> @llvm.usub.sat.v64i8(<64 x i8>, <64 x i8>)
 declare <32 x i16> @llvm.usub.sat.v32i16(<32 x i16>, <32 x i16>)