xref: /external/llvm/test/CodeGen/X86/avx512bwvl-intrinsics.ll

; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=knl -mattr=+avx512bw -mattr=+avx512vl --show-mc-encoding| FileCheck %s

; 256-bit

define <8 x i32> @test_cmp_b_256(<32 x i8> %a0, <32 x i8> %a1) {
; CHECK-LABEL: test_cmp_b_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpcmpeqb %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3f,0xc1,0x00]
; CHECK-NEXT:    kmovd %k0, %r8d ## encoding: [0xc5,0x7b,0x93,0xc0]
; CHECK-NEXT:    vpcmpltb %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3f,0xc1,0x01]
; CHECK-NEXT:    kmovd %k0, %r9d ## encoding: [0xc5,0x7b,0x93,0xc8]
; CHECK-NEXT:    vpcmpleb %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3f,0xc1,0x02]
; CHECK-NEXT:    kmovd %k0, %r10d ## encoding: [0xc5,0x7b,0x93,0xd0]
; CHECK-NEXT:    vpcmpunordb %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3f,0xc1,0x03]
; CHECK-NEXT:    kmovd %k0, %esi ## encoding: [0xc5,0xfb,0x93,0xf0]
; CHECK-NEXT:    vpcmpneqb %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3f,0xc1,0x04]
; CHECK-NEXT:    kmovd %k0, %edi ## encoding: [0xc5,0xfb,0x93,0xf8]
; CHECK-NEXT:    vpcmpnltb %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3f,0xc1,0x05]
; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
; CHECK-NEXT:    vpcmpnleb %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3f,0xc1,0x06]
; CHECK-NEXT:    kmovd %k0, %ecx ## encoding: [0xc5,0xfb,0x93,0xc8]
; CHECK-NEXT:    vpcmpordb %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3f,0xc1,0x07]
; CHECK-NEXT:    kmovd %k0, %edx ## encoding: [0xc5,0xfb,0x93,0xd0]
; CHECK-NEXT:    vmovd %edi, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0x6e,0xc7]
; CHECK-NEXT:    vpinsrd $1, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc0,0x01]
; CHECK-NEXT:    vpinsrd $2, %ecx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc1,0x02]
; CHECK-NEXT:    vpinsrd $3, %edx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc2,0x03]
; CHECK-NEXT:    vmovd %r8d, %xmm1 ## encoding: [0x62,0xd1,0x7d,0x08,0x6e,0xc8]
; CHECK-NEXT:    vpinsrd $1, %r9d, %xmm1, %xmm1 ## encoding: [0xc4,0xc3,0x71,0x22,0xc9,0x01]
; CHECK-NEXT:    vpinsrd $2, %r10d, %xmm1, %xmm1 ## encoding: [0xc4,0xc3,0x71,0x22,0xca,0x02]
; CHECK-NEXT:    vpinsrd $3, %esi, %xmm1, %xmm1 ## encoding: [0xc4,0xe3,0x71,0x22,0xce,0x03]
; CHECK-NEXT:    vinserti32x4 $1, %xmm0, %ymm1, %ymm0 ## encoding: [0x62,0xf3,0x75,0x28,0x38,0xc0,0x01]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res0 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 0, i32 -1)
  %vec0 = insertelement <8 x i32> undef, i32 %res0, i32 0
  %res1 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 1, i32 -1)
  %vec1 = insertelement <8 x i32> %vec0, i32 %res1, i32 1
  %res2 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 2, i32 -1)
  %vec2 = insertelement <8 x i32> %vec1, i32 %res2, i32 2
  %res3 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 3, i32 -1)
  %vec3 = insertelement <8 x i32> %vec2, i32 %res3, i32 3
  %res4 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 4, i32 -1)
  %vec4 = insertelement <8 x i32> %vec3, i32 %res4, i32 4
  %res5 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 5, i32 -1)
  %vec5 = insertelement <8 x i32> %vec4, i32 %res5, i32 5
  %res6 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 6, i32 -1)
  %vec6 = insertelement <8 x i32> %vec5, i32 %res6, i32 6
  %res7 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 7, i32 -1)
  %vec7 = insertelement <8 x i32> %vec6, i32 %res7, i32 7
  ret <8 x i32> %vec7
}

define <8 x i32> @test_mask_cmp_b_256(<32 x i8> %a0, <32 x i8> %a1, i32 %mask) {
; CHECK-LABEL: test_mask_cmp_b_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
; CHECK-NEXT:    vpcmpeqb %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3f,0xc1,0x00]
; CHECK-NEXT:    kmovd %k0, %r8d ## encoding: [0xc5,0x7b,0x93,0xc0]
; CHECK-NEXT:    vpcmpltb %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3f,0xc1,0x01]
; CHECK-NEXT:    kmovd %k0, %r9d ## encoding: [0xc5,0x7b,0x93,0xc8]
; CHECK-NEXT:    vpcmpleb %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3f,0xc1,0x02]
; CHECK-NEXT:    kmovd %k0, %r10d ## encoding: [0xc5,0x7b,0x93,0xd0]
; CHECK-NEXT:    vpcmpunordb %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3f,0xc1,0x03]
; CHECK-NEXT:    kmovd %k0, %esi ## encoding: [0xc5,0xfb,0x93,0xf0]
; CHECK-NEXT:    vpcmpneqb %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3f,0xc1,0x04]
; CHECK-NEXT:    kmovd %k0, %edi ## encoding: [0xc5,0xfb,0x93,0xf8]
; CHECK-NEXT:    vpcmpnltb %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3f,0xc1,0x05]
; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
; CHECK-NEXT:    vpcmpnleb %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3f,0xc1,0x06]
; CHECK-NEXT:    kmovd %k0, %ecx ## encoding: [0xc5,0xfb,0x93,0xc8]
; CHECK-NEXT:    vpcmpordb %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3f,0xc1,0x07]
; CHECK-NEXT:    kmovd %k0, %edx ## encoding: [0xc5,0xfb,0x93,0xd0]
; CHECK-NEXT:    vmovd %edi, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0x6e,0xc7]
; CHECK-NEXT:    vpinsrd $1, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc0,0x01]
; CHECK-NEXT:    vpinsrd $2, %ecx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc1,0x02]
; CHECK-NEXT:    vpinsrd $3, %edx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc2,0x03]
; CHECK-NEXT:    vmovd %r8d, %xmm1 ## encoding: [0x62,0xd1,0x7d,0x08,0x6e,0xc8]
; CHECK-NEXT:    vpinsrd $1, %r9d, %xmm1, %xmm1 ## encoding: [0xc4,0xc3,0x71,0x22,0xc9,0x01]
; CHECK-NEXT:    vpinsrd $2, %r10d, %xmm1, %xmm1 ## encoding: [0xc4,0xc3,0x71,0x22,0xca,0x02]
; CHECK-NEXT:    vpinsrd $3, %esi, %xmm1, %xmm1 ## encoding: [0xc4,0xe3,0x71,0x22,0xce,0x03]
; CHECK-NEXT:    vinserti32x4 $1, %xmm0, %ymm1, %ymm0 ## encoding: [0x62,0xf3,0x75,0x28,0x38,0xc0,0x01]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res0 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 0, i32 %mask)
  %vec0 = insertelement <8 x i32> undef, i32 %res0, i32 0
  %res1 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 1, i32 %mask)
  %vec1 = insertelement <8 x i32> %vec0, i32 %res1, i32 1
  %res2 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 2, i32 %mask)
  %vec2 = insertelement <8 x i32> %vec1, i32 %res2, i32 2
  %res3 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 3, i32 %mask)
  %vec3 = insertelement <8 x i32> %vec2, i32 %res3, i32 3
  %res4 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 4, i32 %mask)
  %vec4 = insertelement <8 x i32> %vec3, i32 %res4, i32 4
  %res5 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 5, i32 %mask)
  %vec5 = insertelement <8 x i32> %vec4, i32 %res5, i32 5
  %res6 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 6, i32 %mask)
  %vec6 = insertelement <8 x i32> %vec5, i32 %res6, i32 6
  %res7 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 7, i32 %mask)
  %vec7 = insertelement <8 x i32> %vec6, i32 %res7, i32 7
  ret <8 x i32> %vec7
}

declare i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8>, <32 x i8>, i32, i32) nounwind readnone

define <8 x i32> @test_ucmp_b_256(<32 x i8> %a0, <32 x i8> %a1) {
; CHECK-LABEL: test_ucmp_b_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpcmpequb %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3e,0xc1,0x00]
; CHECK-NEXT:    kmovd %k0, %r8d ## encoding: [0xc5,0x7b,0x93,0xc0]
; CHECK-NEXT:    vpcmpltub %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3e,0xc1,0x01]
; CHECK-NEXT:    kmovd %k0, %r9d ## encoding: [0xc5,0x7b,0x93,0xc8]
; CHECK-NEXT:    vpcmpleub %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3e,0xc1,0x02]
; CHECK-NEXT:    kmovd %k0, %r10d ## encoding: [0xc5,0x7b,0x93,0xd0]
; CHECK-NEXT:    vpcmpunordub %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3e,0xc1,0x03]
; CHECK-NEXT:    kmovd %k0, %esi ## encoding: [0xc5,0xfb,0x93,0xf0]
; CHECK-NEXT:    vpcmpnequb %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3e,0xc1,0x04]
; CHECK-NEXT:    kmovd %k0, %edi ## encoding: [0xc5,0xfb,0x93,0xf8]
; CHECK-NEXT:    vpcmpnltub %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3e,0xc1,0x05]
; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
; CHECK-NEXT:    vpcmpnleub %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3e,0xc1,0x06]
; CHECK-NEXT:    kmovd %k0, %ecx ## encoding: [0xc5,0xfb,0x93,0xc8]
; CHECK-NEXT:    vpcmpordub %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x28,0x3e,0xc1,0x07]
; CHECK-NEXT:    kmovd %k0, %edx ## encoding: [0xc5,0xfb,0x93,0xd0]
; CHECK-NEXT:    vmovd %edi, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0x6e,0xc7]
; CHECK-NEXT:    vpinsrd $1, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc0,0x01]
; CHECK-NEXT:    vpinsrd $2, %ecx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc1,0x02]
; CHECK-NEXT:    vpinsrd $3, %edx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc2,0x03]
; CHECK-NEXT:    vmovd %r8d, %xmm1 ## encoding: [0x62,0xd1,0x7d,0x08,0x6e,0xc8]
; CHECK-NEXT:    vpinsrd $1, %r9d, %xmm1, %xmm1 ## encoding: [0xc4,0xc3,0x71,0x22,0xc9,0x01]
; CHECK-NEXT:    vpinsrd $2, %r10d, %xmm1, %xmm1 ## encoding: [0xc4,0xc3,0x71,0x22,0xca,0x02]
; CHECK-NEXT:    vpinsrd $3, %esi, %xmm1, %xmm1 ## encoding: [0xc4,0xe3,0x71,0x22,0xce,0x03]
; CHECK-NEXT:    vinserti32x4 $1, %xmm0, %ymm1, %ymm0 ## encoding: [0x62,0xf3,0x75,0x28,0x38,0xc0,0x01]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res0 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 0, i32 -1)
  %vec0 = insertelement <8 x i32> undef, i32 %res0, i32 0
  %res1 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 1, i32 -1)
  %vec1 = insertelement <8 x i32> %vec0, i32 %res1, i32 1
  %res2 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 2, i32 -1)
  %vec2 = insertelement <8 x i32> %vec1, i32 %res2, i32 2
  %res3 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 3, i32 -1)
  %vec3 = insertelement <8 x i32> %vec2, i32 %res3, i32 3
  %res4 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 4, i32 -1)
  %vec4 = insertelement <8 x i32> %vec3, i32 %res4, i32 4
  %res5 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 5, i32 -1)
  %vec5 = insertelement <8 x i32> %vec4, i32 %res5, i32 5
  %res6 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 6, i32 -1)
  %vec6 = insertelement <8 x i32> %vec5, i32 %res6, i32 6
  %res7 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 7, i32 -1)
  %vec7 = insertelement <8 x i32> %vec6, i32 %res7, i32 7
  ret <8 x i32> %vec7
}

define <8 x i32> @test_mask_ucmp_b_256(<32 x i8> %a0, <32 x i8> %a1, i32 %mask) {
; CHECK-LABEL: test_mask_ucmp_b_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
; CHECK-NEXT:    vpcmpequb %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x00]
; CHECK-NEXT:    kmovd %k0, %r8d ## encoding: [0xc5,0x7b,0x93,0xc0]
; CHECK-NEXT:    vpcmpltub %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x01]
; CHECK-NEXT:    kmovd %k0, %r9d ## encoding: [0xc5,0x7b,0x93,0xc8]
; CHECK-NEXT:    vpcmpleub %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x02]
; CHECK-NEXT:    kmovd %k0, %r10d ## encoding: [0xc5,0x7b,0x93,0xd0]
; CHECK-NEXT:    vpcmpunordub %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x03]
; CHECK-NEXT:    kmovd %k0, %esi ## encoding: [0xc5,0xfb,0x93,0xf0]
; CHECK-NEXT:    vpcmpnequb %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x04]
; CHECK-NEXT:    kmovd %k0, %edi ## encoding: [0xc5,0xfb,0x93,0xf8]
; CHECK-NEXT:    vpcmpnltub %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x05]
; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
; CHECK-NEXT:    vpcmpnleub %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x06]
; CHECK-NEXT:    kmovd %k0, %ecx ## encoding: [0xc5,0xfb,0x93,0xc8]
; CHECK-NEXT:    vpcmpordub %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x07]
; CHECK-NEXT:    kmovd %k0, %edx ## encoding: [0xc5,0xfb,0x93,0xd0]
; CHECK-NEXT:    vmovd %edi, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0x6e,0xc7]
; CHECK-NEXT:    vpinsrd $1, %eax, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc0,0x01]
; CHECK-NEXT:    vpinsrd $2, %ecx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc1,0x02]
; CHECK-NEXT:    vpinsrd $3, %edx, %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x22,0xc2,0x03]
; CHECK-NEXT:    vmovd %r8d, %xmm1 ## encoding: [0x62,0xd1,0x7d,0x08,0x6e,0xc8]
; CHECK-NEXT:    vpinsrd $1, %r9d, %xmm1, %xmm1 ## encoding: [0xc4,0xc3,0x71,0x22,0xc9,0x01]
; CHECK-NEXT:    vpinsrd $2, %r10d, %xmm1, %xmm1 ## encoding: [0xc4,0xc3,0x71,0x22,0xca,0x02]
; CHECK-NEXT:    vpinsrd $3, %esi, %xmm1, %xmm1 ## encoding: [0xc4,0xe3,0x71,0x22,0xce,0x03]
; CHECK-NEXT:    vinserti32x4 $1, %xmm0, %ymm1, %ymm0 ## encoding: [0x62,0xf3,0x75,0x28,0x38,0xc0,0x01]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res0 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 0, i32 %mask)
  %vec0 = insertelement <8 x i32> undef, i32 %res0, i32 0
  %res1 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 1, i32 %mask)
  %vec1 = insertelement <8 x i32> %vec0, i32 %res1, i32 1
  %res2 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 2, i32 %mask)
  %vec2 = insertelement <8 x i32> %vec1, i32 %res2, i32 2
  %res3 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 3, i32 %mask)
  %vec3 = insertelement <8 x i32> %vec2, i32 %res3, i32 3
  %res4 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 4, i32 %mask)
  %vec4 = insertelement <8 x i32> %vec3, i32 %res4, i32 4
  %res5 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 5, i32 %mask)
  %vec5 = insertelement <8 x i32> %vec4, i32 %res5, i32 5
  %res6 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 6, i32 %mask)
  %vec6 = insertelement <8 x i32> %vec5, i32 %res6, i32 6
  %res7 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 7, i32 %mask)
  %vec7 = insertelement <8 x i32> %vec6, i32 %res7, i32 7
  ret <8 x i32> %vec7
}

declare i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8>, <32 x i8>, i32, i32) nounwind readnone

define <8 x i16> @test_cmp_w_256(<16 x i16> %a0, <16 x i16> %a1) {
; CHECK-LABEL: test_cmp_w_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpcmpeqw %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0xfd,0x28,0x3f,0xc1,0x00]
; CHECK-NEXT:    vpcmpltw %ymm1, %ymm0, %k1 ## encoding: [0x62,0xf3,0xfd,0x28,0x3f,0xc9,0x01]
; CHECK-NEXT:    vpcmplew %ymm1, %ymm0, %k2 ## encoding: [0x62,0xf3,0xfd,0x28,0x3f,0xd1,0x02]
; CHECK-NEXT:    vpcmpunordw %ymm1, %ymm0, %k3 ## encoding: [0x62,0xf3,0xfd,0x28,0x3f,0xd9,0x03]
; CHECK-NEXT:    vpcmpneqw %ymm1, %ymm0, %k4 ## encoding: [0x62,0xf3,0xfd,0x28,0x3f,0xe1,0x04]
; CHECK-NEXT:    vpcmpnltw %ymm1, %ymm0, %k5 ## encoding: [0x62,0xf3,0xfd,0x28,0x3f,0xe9,0x05]
; CHECK-NEXT:    vpcmpnlew %ymm1, %ymm0, %k6 ## encoding: [0x62,0xf3,0xfd,0x28,0x3f,0xf1,0x06]
; CHECK-NEXT:    vpcmpordw %ymm1, %ymm0, %k7 ## encoding: [0x62,0xf3,0xfd,0x28,0x3f,0xf9,0x07]
; CHECK-NEXT:    kmovw %k1, %eax ## encoding: [0xc5,0xf8,0x93,0xc1]
; CHECK-NEXT:    kmovw %k0, %ecx ## encoding: [0xc5,0xf8,0x93,0xc8]
; CHECK-NEXT:    vmovd %ecx, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0x6e,0xc1]
; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xc4,0xc0,0x01]
; CHECK-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xc4,0xc0,0x02]
; CHECK-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
; CHECK-NEXT:    vpinsrw $3, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xc4,0xc0,0x03]
; CHECK-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xc4,0xc0,0x04]
; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xc4,0xc0,0x05]
; CHECK-NEXT:    kmovw %k6, %eax ## encoding: [0xc5,0xf8,0x93,0xc6]
; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xc4,0xc0,0x06]
; CHECK-NEXT:    kmovw %k7, %eax ## encoding: [0xc5,0xf8,0x93,0xc7]
; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xc4,0xc0,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res0 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 0, i16 -1)
  %vec0 = insertelement <8 x i16> undef, i16 %res0, i32 0
  %res1 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 1, i16 -1)
  %vec1 = insertelement <8 x i16> %vec0, i16 %res1, i32 1
  %res2 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 2, i16 -1)
  %vec2 = insertelement <8 x i16> %vec1, i16 %res2, i32 2
  %res3 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 3, i16 -1)
  %vec3 = insertelement <8 x i16> %vec2, i16 %res3, i32 3
  %res4 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 4, i16 -1)
  %vec4 = insertelement <8 x i16> %vec3, i16 %res4, i32 4
  %res5 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 5, i16 -1)
  %vec5 = insertelement <8 x i16> %vec4, i16 %res5, i32 5
  %res6 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 6, i16 -1)
  %vec6 = insertelement <8 x i16> %vec5, i16 %res6, i32 6
  %res7 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 7, i16 -1)
  %vec7 = insertelement <8 x i16> %vec6, i16 %res7, i32 7
  ret <8 x i16> %vec7
}

define <8 x i16> @test_mask_cmp_w_256(<16 x i16> %a0, <16 x i16> %a1, i16 %mask) {
; CHECK-LABEL: test_mask_cmp_w_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpcmpeqw %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x3f,0xc1,0x00]
; CHECK-NEXT:    vpcmpltw %ymm1, %ymm0, %k2 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x3f,0xd1,0x01]
; CHECK-NEXT:    vpcmplew %ymm1, %ymm0, %k3 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x3f,0xd9,0x02]
; CHECK-NEXT:    vpcmpunordw %ymm1, %ymm0, %k4 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x3f,0xe1,0x03]
; CHECK-NEXT:    vpcmpneqw %ymm1, %ymm0, %k5 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x3f,0xe9,0x04]
; CHECK-NEXT:    vpcmpnltw %ymm1, %ymm0, %k6 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x3f,0xf1,0x05]
; CHECK-NEXT:    vpcmpnlew %ymm1, %ymm0, %k7 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x3f,0xf9,0x06]
; CHECK-NEXT:    vpcmpordw %ymm1, %ymm0, %k1 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x3f,0xc9,0x07]
; CHECK-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
; CHECK-NEXT:    kmovw %k0, %ecx ## encoding: [0xc5,0xf8,0x93,0xc8]
; CHECK-NEXT:    vmovd %ecx, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0x6e,0xc1]
; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xc4,0xc0,0x01]
; CHECK-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xc4,0xc0,0x02]
; CHECK-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
; CHECK-NEXT:    vpinsrw $3, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xc4,0xc0,0x03]
; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xc4,0xc0,0x04]
; CHECK-NEXT:    kmovw %k6, %eax ## encoding: [0xc5,0xf8,0x93,0xc6]
; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xc4,0xc0,0x05]
; CHECK-NEXT:    kmovw %k7, %eax ## encoding: [0xc5,0xf8,0x93,0xc7]
; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xc4,0xc0,0x06]
; CHECK-NEXT:    kmovw %k1, %eax ## encoding: [0xc5,0xf8,0x93,0xc1]
; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xc4,0xc0,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res0 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 0, i16 %mask)
  %vec0 = insertelement <8 x i16> undef, i16 %res0, i32 0
  %res1 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 1, i16 %mask)
  %vec1 = insertelement <8 x i16> %vec0, i16 %res1, i32 1
  %res2 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 2, i16 %mask)
  %vec2 = insertelement <8 x i16> %vec1, i16 %res2, i32 2
  %res3 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 3, i16 %mask)
  %vec3 = insertelement <8 x i16> %vec2, i16 %res3, i32 3
  %res4 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 4, i16 %mask)
  %vec4 = insertelement <8 x i16> %vec3, i16 %res4, i32 4
  %res5 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 5, i16 %mask)
  %vec5 = insertelement <8 x i16> %vec4, i16 %res5, i32 5
  %res6 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 6, i16 %mask)
  %vec6 = insertelement <8 x i16> %vec5, i16 %res6, i32 6
  %res7 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 7, i16 %mask)
  %vec7 = insertelement <8 x i16> %vec6, i16 %res7, i32 7
  ret <8 x i16> %vec7
}

declare i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16>, <16 x i16>, i32, i16) nounwind readnone

define <8 x i16> @test_ucmp_w_256(<16 x i16> %a0, <16 x i16> %a1) {
; CHECK-LABEL: test_ucmp_w_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpcmpequw %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf3,0xfd,0x28,0x3e,0xc1,0x00]
; CHECK-NEXT:    vpcmpltuw %ymm1, %ymm0, %k1 ## encoding: [0x62,0xf3,0xfd,0x28,0x3e,0xc9,0x01]
; CHECK-NEXT:    vpcmpleuw %ymm1, %ymm0, %k2 ## encoding: [0x62,0xf3,0xfd,0x28,0x3e,0xd1,0x02]
; CHECK-NEXT:    vpcmpunorduw %ymm1, %ymm0, %k3 ## encoding: [0x62,0xf3,0xfd,0x28,0x3e,0xd9,0x03]
; CHECK-NEXT:    vpcmpnequw %ymm1, %ymm0, %k4 ## encoding: [0x62,0xf3,0xfd,0x28,0x3e,0xe1,0x04]
; CHECK-NEXT:    vpcmpnltuw %ymm1, %ymm0, %k5 ## encoding: [0x62,0xf3,0xfd,0x28,0x3e,0xe9,0x05]
; CHECK-NEXT:    vpcmpnleuw %ymm1, %ymm0, %k6 ## encoding: [0x62,0xf3,0xfd,0x28,0x3e,0xf1,0x06]
; CHECK-NEXT:    vpcmporduw %ymm1, %ymm0, %k7 ## encoding: [0x62,0xf3,0xfd,0x28,0x3e,0xf9,0x07]
; CHECK-NEXT:    kmovw %k1, %eax ## encoding: [0xc5,0xf8,0x93,0xc1]
; CHECK-NEXT:    kmovw %k0, %ecx ## encoding: [0xc5,0xf8,0x93,0xc8]
; CHECK-NEXT:    vmovd %ecx, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0x6e,0xc1]
; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xc4,0xc0,0x01]
; CHECK-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xc4,0xc0,0x02]
; CHECK-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
; CHECK-NEXT:    vpinsrw $3, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xc4,0xc0,0x03]
; CHECK-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xc4,0xc0,0x04]
; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xc4,0xc0,0x05]
; CHECK-NEXT:    kmovw %k6, %eax ## encoding: [0xc5,0xf8,0x93,0xc6]
; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xc4,0xc0,0x06]
; CHECK-NEXT:    kmovw %k7, %eax ## encoding: [0xc5,0xf8,0x93,0xc7]
; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xc4,0xc0,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res0 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 0, i16 -1)
  %vec0 = insertelement <8 x i16> undef, i16 %res0, i32 0
  %res1 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 1, i16 -1)
  %vec1 = insertelement <8 x i16> %vec0, i16 %res1, i32 1
  %res2 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 2, i16 -1)
  %vec2 = insertelement <8 x i16> %vec1, i16 %res2, i32 2
  %res3 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 3, i16 -1)
  %vec3 = insertelement <8 x i16> %vec2, i16 %res3, i32 3
  %res4 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 4, i16 -1)
  %vec4 = insertelement <8 x i16> %vec3, i16 %res4, i32 4
  %res5 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 5, i16 -1)
  %vec5 = insertelement <8 x i16> %vec4, i16 %res5, i32 5
  %res6 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 6, i16 -1)
  %vec6 = insertelement <8 x i16> %vec5, i16 %res6, i32 6
  %res7 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 7, i16 -1)
  %vec7 = insertelement <8 x i16> %vec6, i16 %res7, i32 7
  ret <8 x i16> %vec7
}

define <8 x i16> @test_mask_ucmp_w_256(<16 x i16> %a0, <16 x i16> %a1, i16 %mask) {
; CHECK-LABEL: test_mask_ucmp_w_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpcmpequw %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x3e,0xc1,0x00]
; CHECK-NEXT:    vpcmpltuw %ymm1, %ymm0, %k2 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x3e,0xd1,0x01]
; CHECK-NEXT:    vpcmpleuw %ymm1, %ymm0, %k3 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x3e,0xd9,0x02]
; CHECK-NEXT:    vpcmpunorduw %ymm1, %ymm0, %k4 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x3e,0xe1,0x03]
; CHECK-NEXT:    vpcmpnequw %ymm1, %ymm0, %k5 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x3e,0xe9,0x04]
; CHECK-NEXT:    vpcmpnltuw %ymm1, %ymm0, %k6 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x3e,0xf1,0x05]
; CHECK-NEXT:    vpcmpnleuw %ymm1, %ymm0, %k7 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x3e,0xf9,0x06]
; CHECK-NEXT:    vpcmporduw %ymm1, %ymm0, %k1 {%k1} ## encoding: [0x62,0xf3,0xfd,0x29,0x3e,0xc9,0x07]
; CHECK-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
; CHECK-NEXT:    kmovw %k0, %ecx ## encoding: [0xc5,0xf8,0x93,0xc8]
; CHECK-NEXT:    vmovd %ecx, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0x6e,0xc1]
; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xc4,0xc0,0x01]
; CHECK-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xc4,0xc0,0x02]
; CHECK-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
; CHECK-NEXT:    vpinsrw $3, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xc4,0xc0,0x03]
; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xc4,0xc0,0x04]
; CHECK-NEXT:    kmovw %k6, %eax ## encoding: [0xc5,0xf8,0x93,0xc6]
; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xc4,0xc0,0x05]
; CHECK-NEXT:    kmovw %k7, %eax ## encoding: [0xc5,0xf8,0x93,0xc7]
; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xc4,0xc0,0x06]
; CHECK-NEXT:    kmovw %k1, %eax ## encoding: [0xc5,0xf8,0x93,0xc1]
; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xc4,0xc0,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res0 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 0, i16 %mask)
  %vec0 = insertelement <8 x i16> undef, i16 %res0, i32 0
  %res1 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 1, i16 %mask)
  %vec1 = insertelement <8 x i16> %vec0, i16 %res1, i32 1
  %res2 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 2, i16 %mask)
  %vec2 = insertelement <8 x i16> %vec1, i16 %res2, i32 2
  %res3 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 3, i16 %mask)
  %vec3 = insertelement <8 x i16> %vec2, i16 %res3, i32 3
  %res4 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 4, i16 %mask)
  %vec4 = insertelement <8 x i16> %vec3, i16 %res4, i32 4
  %res5 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 5, i16 %mask)
  %vec5 = insertelement <8 x i16> %vec4, i16 %res5, i32 5
  %res6 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 6, i16 %mask)
  %vec6 = insertelement <8 x i16> %vec5, i16 %res6, i32 6
  %res7 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 7, i16 %mask)
  %vec7 = insertelement <8 x i16> %vec6, i16 %res7, i32 7
  ret <8 x i16> %vec7
}

declare i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16>, <16 x i16>, i32, i16) nounwind readnone

; 128-bit

define i16 @test_pcmpeq_b_128(<16 x i8> %a, <16 x i8> %b) {
; CHECK-LABEL: test_pcmpeq_b_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpcmpeqb %xmm1, %xmm0, %k0 ## encoding: [0x62,0xf1,0x7d,0x08,0x74,0xc1]
; CHECK-NEXT:    kmovw %k0, %eax ## encoding: [0xc5,0xf8,0x93,0xc0]
; CHECK-NEXT:    ## kill: %AX<def> %AX<kill> %EAX<kill>
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call i16 @llvm.x86.avx512.mask.pcmpeq.b.128(<16 x i8> %a, <16 x i8> %b, i16 -1)
  ret i16 %res
}

define i16 @test_mask_pcmpeq_b_128(<16 x i8> %a, <16 x i8> %b, i16 %mask) {
; CHECK-LABEL: test_mask_pcmpeq_b_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpcmpeqb %xmm1, %xmm0, %k0 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0x74,0xc1]
; CHECK-NEXT:    kmovw %k0, %eax ## encoding: [0xc5,0xf8,0x93,0xc0]
; CHECK-NEXT:    ## kill: %AX<def> %AX<kill> %EAX<kill>
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call i16 @llvm.x86.avx512.mask.pcmpeq.b.128(<16 x i8> %a, <16 x i8> %b, i16 %mask)
  ret i16 %res
}

declare i16 @llvm.x86.avx512.mask.pcmpeq.b.128(<16 x i8>, <16 x i8>, i16)

define i8 @test_pcmpeq_w_128(<8 x i16> %a, <8 x i16> %b) {
; CHECK-LABEL: test_pcmpeq_w_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpcmpeqw %xmm1, %xmm0, %k0 ## encoding: [0x62,0xf1,0x7d,0x08,0x75,0xc1]
; CHECK-NEXT:    kmovw %k0, %eax ## encoding: [0xc5,0xf8,0x93,0xc0]
; CHECK-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call i8 @llvm.x86.avx512.mask.pcmpeq.w.128(<8 x i16> %a, <8 x i16> %b, i8 -1)
  ret i8 %res
}

define i8 @test_mask_pcmpeq_w_128(<8 x i16> %a, <8 x i16> %b, i8 %mask) {
; CHECK-LABEL: test_mask_pcmpeq_w_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpcmpeqw %xmm1, %xmm0, %k0 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0x75,0xc1]
; CHECK-NEXT:    kmovw %k0, %eax ## encoding: [0xc5,0xf8,0x93,0xc0]
; CHECK-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call i8 @llvm.x86.avx512.mask.pcmpeq.w.128(<8 x i16> %a, <8 x i16> %b, i8 %mask)
  ret i8 %res
}

declare i8 @llvm.x86.avx512.mask.pcmpeq.w.128(<8 x i16>, <8 x i16>, i8)

define i16 @test_pcmpgt_b_128(<16 x i8> %a, <16 x i8> %b) {
; CHECK-LABEL: test_pcmpgt_b_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpcmpgtb %xmm1, %xmm0, %k0 ## encoding: [0x62,0xf1,0x7d,0x08,0x64,0xc1]
; CHECK-NEXT:    kmovw %k0, %eax ## encoding: [0xc5,0xf8,0x93,0xc0]
; CHECK-NEXT:    ## kill: %AX<def> %AX<kill> %EAX<kill>
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call i16 @llvm.x86.avx512.mask.pcmpgt.b.128(<16 x i8> %a, <16 x i8> %b, i16 -1)
  ret i16 %res
}

define i16 @test_mask_pcmpgt_b_128(<16 x i8> %a, <16 x i8> %b, i16 %mask) {
; CHECK-LABEL: test_mask_pcmpgt_b_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpcmpgtb %xmm1, %xmm0, %k0 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0x64,0xc1]
; CHECK-NEXT:    kmovw %k0, %eax ## encoding: [0xc5,0xf8,0x93,0xc0]
; CHECK-NEXT:    ## kill: %AX<def> %AX<kill> %EAX<kill>
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call i16 @llvm.x86.avx512.mask.pcmpgt.b.128(<16 x i8> %a, <16 x i8> %b, i16 %mask)
  ret i16 %res
}

declare i16 @llvm.x86.avx512.mask.pcmpgt.b.128(<16 x i8>, <16 x i8>, i16)

define i8 @test_pcmpgt_w_128(<8 x i16> %a, <8 x i16> %b) {
; CHECK-LABEL: test_pcmpgt_w_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpcmpgtw %xmm1, %xmm0, %k0 ## encoding: [0x62,0xf1,0x7d,0x08,0x65,0xc1]
; CHECK-NEXT:    kmovw %k0, %eax ## encoding: [0xc5,0xf8,0x93,0xc0]
; CHECK-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call i8 @llvm.x86.avx512.mask.pcmpgt.w.128(<8 x i16> %a, <8 x i16> %b, i8 -1)
  ret i8 %res
}

define i8 @test_mask_pcmpgt_w_128(<8 x i16> %a, <8 x i16> %b, i8 %mask) {
; CHECK-LABEL: test_mask_pcmpgt_w_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpcmpgtw %xmm1, %xmm0, %k0 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0x65,0xc1]
; CHECK-NEXT:    kmovw %k0, %eax ## encoding: [0xc5,0xf8,0x93,0xc0]
; CHECK-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call i8 @llvm.x86.avx512.mask.pcmpgt.w.128(<8 x i16> %a, <8 x i16> %b, i8 %mask)
  ret i8 %res
}

declare i8 @llvm.x86.avx512.mask.pcmpgt.w.128(<8 x i16>, <8 x i16>, i8)

define <8 x i16> @test_cmp_b_128(<16 x i8> %a0, <16 x i8> %a1) {
; CHECK-LABEL: test_cmp_b_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpcmpeqb %xmm1, %xmm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x08,0x3f,0xc1,0x00]
; CHECK-NEXT:    vpcmpltb %xmm1, %xmm0, %k1 ## encoding: [0x62,0xf3,0x7d,0x08,0x3f,0xc9,0x01]
; CHECK-NEXT:    vpcmpleb %xmm1, %xmm0, %k2 ## encoding: [0x62,0xf3,0x7d,0x08,0x3f,0xd1,0x02]
; CHECK-NEXT:    vpcmpunordb %xmm1, %xmm0, %k3 ## encoding: [0x62,0xf3,0x7d,0x08,0x3f,0xd9,0x03]
; CHECK-NEXT:    vpcmpneqb %xmm1, %xmm0, %k4 ## encoding: [0x62,0xf3,0x7d,0x08,0x3f,0xe1,0x04]
; CHECK-NEXT:    vpcmpnltb %xmm1, %xmm0, %k5 ## encoding: [0x62,0xf3,0x7d,0x08,0x3f,0xe9,0x05]
; CHECK-NEXT:    vpcmpnleb %xmm1, %xmm0, %k6 ## encoding: [0x62,0xf3,0x7d,0x08,0x3f,0xf1,0x06]
; CHECK-NEXT:    vpcmpordb %xmm1, %xmm0, %k7 ## encoding: [0x62,0xf3,0x7d,0x08,0x3f,0xf9,0x07]
; CHECK-NEXT:    kmovw %k1, %eax ## encoding: [0xc5,0xf8,0x93,0xc1]
; CHECK-NEXT:    kmovw %k0, %ecx ## encoding: [0xc5,0xf8,0x93,0xc8]
; CHECK-NEXT:    vmovd %ecx, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0x6e,0xc1]
; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xc4,0xc0,0x01]
; CHECK-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xc4,0xc0,0x02]
; CHECK-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
; CHECK-NEXT:    vpinsrw $3, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xc4,0xc0,0x03]
; CHECK-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xc4,0xc0,0x04]
; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xc4,0xc0,0x05]
; CHECK-NEXT:    kmovw %k6, %eax ## encoding: [0xc5,0xf8,0x93,0xc6]
; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xc4,0xc0,0x06]
; CHECK-NEXT:    kmovw %k7, %eax ## encoding: [0xc5,0xf8,0x93,0xc7]
; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xc4,0xc0,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res0 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 0, i16 -1)
  %vec0 = insertelement <8 x i16> undef, i16 %res0, i32 0
  %res1 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 1, i16 -1)
  %vec1 = insertelement <8 x i16> %vec0, i16 %res1, i32 1
  %res2 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 2, i16 -1)
  %vec2 = insertelement <8 x i16> %vec1, i16 %res2, i32 2
  %res3 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 3, i16 -1)
  %vec3 = insertelement <8 x i16> %vec2, i16 %res3, i32 3
  %res4 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 4, i16 -1)
  %vec4 = insertelement <8 x i16> %vec3, i16 %res4, i32 4
  %res5 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 5, i16 -1)
  %vec5 = insertelement <8 x i16> %vec4, i16 %res5, i32 5
  %res6 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 6, i16 -1)
  %vec6 = insertelement <8 x i16> %vec5, i16 %res6, i32 6
  %res7 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 7, i16 -1)
  %vec7 = insertelement <8 x i16> %vec6, i16 %res7, i32 7
  ret <8 x i16> %vec7
}

define <8 x i16> @test_mask_cmp_b_128(<16 x i8> %a0, <16 x i8> %a1, i16 %mask) {
; CHECK-LABEL: test_mask_cmp_b_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpcmpeqb %xmm1, %xmm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x3f,0xc1,0x00]
; CHECK-NEXT:    vpcmpltb %xmm1, %xmm0, %k2 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x3f,0xd1,0x01]
; CHECK-NEXT:    vpcmpleb %xmm1, %xmm0, %k3 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x3f,0xd9,0x02]
; CHECK-NEXT:    vpcmpunordb %xmm1, %xmm0, %k4 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x3f,0xe1,0x03]
; CHECK-NEXT:    vpcmpneqb %xmm1, %xmm0, %k5 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x3f,0xe9,0x04]
; CHECK-NEXT:    vpcmpnltb %xmm1, %xmm0, %k6 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x3f,0xf1,0x05]
; CHECK-NEXT:    vpcmpnleb %xmm1, %xmm0, %k7 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x3f,0xf9,0x06]
; CHECK-NEXT:    vpcmpordb %xmm1, %xmm0, %k1 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x3f,0xc9,0x07]
; CHECK-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
; CHECK-NEXT:    kmovw %k0, %ecx ## encoding: [0xc5,0xf8,0x93,0xc8]
; CHECK-NEXT:    vmovd %ecx, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0x6e,0xc1]
; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xc4,0xc0,0x01]
; CHECK-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xc4,0xc0,0x02]
; CHECK-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
; CHECK-NEXT:    vpinsrw $3, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xc4,0xc0,0x03]
; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xc4,0xc0,0x04]
; CHECK-NEXT:    kmovw %k6, %eax ## encoding: [0xc5,0xf8,0x93,0xc6]
; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xc4,0xc0,0x05]
; CHECK-NEXT:    kmovw %k7, %eax ## encoding: [0xc5,0xf8,0x93,0xc7]
; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xc4,0xc0,0x06]
; CHECK-NEXT:    kmovw %k1, %eax ## encoding: [0xc5,0xf8,0x93,0xc1]
; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xc4,0xc0,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res0 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 0, i16 %mask)
  %vec0 = insertelement <8 x i16> undef, i16 %res0, i32 0
  %res1 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 1, i16 %mask)
  %vec1 = insertelement <8 x i16> %vec0, i16 %res1, i32 1
  %res2 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 2, i16 %mask)
  %vec2 = insertelement <8 x i16> %vec1, i16 %res2, i32 2
  %res3 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 3, i16 %mask)
  %vec3 = insertelement <8 x i16> %vec2, i16 %res3, i32 3
  %res4 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 4, i16 %mask)
  %vec4 = insertelement <8 x i16> %vec3, i16 %res4, i32 4
  %res5 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 5, i16 %mask)
  %vec5 = insertelement <8 x i16> %vec4, i16 %res5, i32 5
  %res6 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 6, i16 %mask)
  %vec6 = insertelement <8 x i16> %vec5, i16 %res6, i32 6
  %res7 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 7, i16 %mask)
  %vec7 = insertelement <8 x i16> %vec6, i16 %res7, i32 7
  ret <8 x i16> %vec7
}

declare i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8>, <16 x i8>, i32, i16) nounwind readnone

define <8 x i16> @test_ucmp_b_128(<16 x i8> %a0, <16 x i8> %a1) {
; CHECK-LABEL: test_ucmp_b_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpcmpequb %xmm1, %xmm0, %k0 ## encoding: [0x62,0xf3,0x7d,0x08,0x3e,0xc1,0x00]
; CHECK-NEXT:    vpcmpltub %xmm1, %xmm0, %k1 ## encoding: [0x62,0xf3,0x7d,0x08,0x3e,0xc9,0x01]
; CHECK-NEXT:    vpcmpleub %xmm1, %xmm0, %k2 ## encoding: [0x62,0xf3,0x7d,0x08,0x3e,0xd1,0x02]
; CHECK-NEXT:    vpcmpunordub %xmm1, %xmm0, %k3 ## encoding: [0x62,0xf3,0x7d,0x08,0x3e,0xd9,0x03]
; CHECK-NEXT:    vpcmpnequb %xmm1, %xmm0, %k4 ## encoding: [0x62,0xf3,0x7d,0x08,0x3e,0xe1,0x04]
; CHECK-NEXT:    vpcmpnltub %xmm1, %xmm0, %k5 ## encoding: [0x62,0xf3,0x7d,0x08,0x3e,0xe9,0x05]
; CHECK-NEXT:    vpcmpnleub %xmm1, %xmm0, %k6 ## encoding: [0x62,0xf3,0x7d,0x08,0x3e,0xf1,0x06]
; CHECK-NEXT:    vpcmpordub %xmm1, %xmm0, %k7 ## encoding: [0x62,0xf3,0x7d,0x08,0x3e,0xf9,0x07]
; CHECK-NEXT:    kmovw %k1, %eax ## encoding: [0xc5,0xf8,0x93,0xc1]
; CHECK-NEXT:    kmovw %k0, %ecx ## encoding: [0xc5,0xf8,0x93,0xc8]
; CHECK-NEXT:    vmovd %ecx, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0x6e,0xc1]
; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xc4,0xc0,0x01]
; CHECK-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xc4,0xc0,0x02]
; CHECK-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
; CHECK-NEXT:    vpinsrw $3, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xc4,0xc0,0x03]
; CHECK-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xc4,0xc0,0x04]
; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xc4,0xc0,0x05]
; CHECK-NEXT:    kmovw %k6, %eax ## encoding: [0xc5,0xf8,0x93,0xc6]
; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xc4,0xc0,0x06]
; CHECK-NEXT:    kmovw %k7, %eax ## encoding: [0xc5,0xf8,0x93,0xc7]
; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xc4,0xc0,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res0 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 0, i16 -1)
  %vec0 = insertelement <8 x i16> undef, i16 %res0, i32 0
  %res1 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 1, i16 -1)
  %vec1 = insertelement <8 x i16> %vec0, i16 %res1, i32 1
  %res2 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 2, i16 -1)
  %vec2 = insertelement <8 x i16> %vec1, i16 %res2, i32 2
  %res3 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 3, i16 -1)
  %vec3 = insertelement <8 x i16> %vec2, i16 %res3, i32 3
  %res4 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 4, i16 -1)
  %vec4 = insertelement <8 x i16> %vec3, i16 %res4, i32 4
  %res5 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 5, i16 -1)
  %vec5 = insertelement <8 x i16> %vec4, i16 %res5, i32 5
  %res6 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 6, i16 -1)
  %vec6 = insertelement <8 x i16> %vec5, i16 %res6, i32 6
  %res7 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 7, i16 -1)
  %vec7 = insertelement <8 x i16> %vec6, i16 %res7, i32 7
  ret <8 x i16> %vec7
}

define <8 x i16> @test_mask_ucmp_b_128(<16 x i8> %a0, <16 x i8> %a1, i16 %mask) {
; CHECK-LABEL: test_mask_ucmp_b_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpcmpequb %xmm1, %xmm0, %k0 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x3e,0xc1,0x00]
; CHECK-NEXT:    vpcmpltub %xmm1, %xmm0, %k2 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x3e,0xd1,0x01]
; CHECK-NEXT:    vpcmpleub %xmm1, %xmm0, %k3 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x3e,0xd9,0x02]
; CHECK-NEXT:    vpcmpunordub %xmm1, %xmm0, %k4 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x3e,0xe1,0x03]
; CHECK-NEXT:    vpcmpnequb %xmm1, %xmm0, %k5 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x3e,0xe9,0x04]
; CHECK-NEXT:    vpcmpnltub %xmm1, %xmm0, %k6 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x3e,0xf1,0x05]
; CHECK-NEXT:    vpcmpnleub %xmm1, %xmm0, %k7 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x3e,0xf9,0x06]
; CHECK-NEXT:    vpcmpordub %xmm1, %xmm0, %k1 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x3e,0xc9,0x07]
; CHECK-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
; CHECK-NEXT:    kmovw %k0, %ecx ## encoding: [0xc5,0xf8,0x93,0xc8]
; CHECK-NEXT:    vmovd %ecx, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0x6e,0xc1]
; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xc4,0xc0,0x01]
; CHECK-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xc4,0xc0,0x02]
; CHECK-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
; CHECK-NEXT:    vpinsrw $3, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xc4,0xc0,0x03]
; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xc4,0xc0,0x04]
; CHECK-NEXT:    kmovw %k6, %eax ## encoding: [0xc5,0xf8,0x93,0xc6]
; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xc4,0xc0,0x05]
; CHECK-NEXT:    kmovw %k7, %eax ## encoding: [0xc5,0xf8,0x93,0xc7]
; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xc4,0xc0,0x06]
; CHECK-NEXT:    kmovw %k1, %eax ## encoding: [0xc5,0xf8,0x93,0xc1]
; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xc4,0xc0,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res0 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 0, i16 %mask)
  %vec0 = insertelement <8 x i16> undef, i16 %res0, i32 0
  %res1 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 1, i16 %mask)
  %vec1 = insertelement <8 x i16> %vec0, i16 %res1, i32 1
  %res2 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 2, i16 %mask)
  %vec2 = insertelement <8 x i16> %vec1, i16 %res2, i32 2
  %res3 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 3, i16 %mask)
  %vec3 = insertelement <8 x i16> %vec2, i16 %res3, i32 3
  %res4 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 4, i16 %mask)
  %vec4 = insertelement <8 x i16> %vec3, i16 %res4, i32 4
  %res5 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 5, i16 %mask)
  %vec5 = insertelement <8 x i16> %vec4, i16 %res5, i32 5
  %res6 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 6, i16 %mask)
  %vec6 = insertelement <8 x i16> %vec5, i16 %res6, i32 6
  %res7 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 7, i16 %mask)
  %vec7 = insertelement <8 x i16> %vec6, i16 %res7, i32 7
  ret <8 x i16> %vec7
}

declare i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8>, <16 x i8>, i32, i16) nounwind readnone

define <8 x i8> @test_cmp_w_128(<8 x i16> %a0, <8 x i16> %a1) {
; CHECK-LABEL: test_cmp_w_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpcmpeqw %xmm1, %xmm0, %k0 ## encoding: [0x62,0xf3,0xfd,0x08,0x3f,0xc1,0x00]
; CHECK-NEXT:    vpcmpltw %xmm1, %xmm0, %k1 ## encoding: [0x62,0xf3,0xfd,0x08,0x3f,0xc9,0x01]
; CHECK-NEXT:    vpcmplew %xmm1, %xmm0, %k2 ## encoding: [0x62,0xf3,0xfd,0x08,0x3f,0xd1,0x02]
; CHECK-NEXT:    vpcmpunordw %xmm1, %xmm0, %k3 ## encoding: [0x62,0xf3,0xfd,0x08,0x3f,0xd9,0x03]
; CHECK-NEXT:    vpcmpneqw %xmm1, %xmm0, %k4 ## encoding: [0x62,0xf3,0xfd,0x08,0x3f,0xe1,0x04]
; CHECK-NEXT:    vpcmpnltw %xmm1, %xmm0, %k5 ## encoding: [0x62,0xf3,0xfd,0x08,0x3f,0xe9,0x05]
; CHECK-NEXT:    vpcmpnlew %xmm1, %xmm0, %k6 ## encoding: [0x62,0xf3,0xfd,0x08,0x3f,0xf1,0x06]
; CHECK-NEXT:    vpcmpordw %xmm1, %xmm0, %k7 ## encoding: [0x62,0xf3,0xfd,0x08,0x3f,0xf9,0x07]
; CHECK-NEXT:    kmovw %k0, %eax ## encoding: [0xc5,0xf8,0x93,0xc0]
; CHECK-NEXT:    vpinsrb $0, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf3,0x7d,0x08,0x20,0xc0,0x00]
; CHECK-NEXT:    kmovw %k1, %eax ## encoding: [0xc5,0xf8,0x93,0xc1]
; CHECK-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf3,0x7d,0x08,0x20,0xc0,0x02]
; CHECK-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
; CHECK-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf3,0x7d,0x08,0x20,0xc0,0x04]
; CHECK-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
; CHECK-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf3,0x7d,0x08,0x20,0xc0,0x06]
; CHECK-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
; CHECK-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf3,0x7d,0x08,0x20,0xc0,0x08]
; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
; CHECK-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf3,0x7d,0x08,0x20,0xc0,0x0a]
; CHECK-NEXT:    kmovw %k6, %eax ## encoding: [0xc5,0xf8,0x93,0xc6]
; CHECK-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf3,0x7d,0x08,0x20,0xc0,0x0c]
; CHECK-NEXT:    kmovw %k7, %eax ## encoding: [0xc5,0xf8,0x93,0xc7]
; CHECK-NEXT:    vpinsrb $14, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf3,0x7d,0x08,0x20,0xc0,0x0e]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res0 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 0, i8 -1)
  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
  %res1 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 1, i8 -1)
  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
  %res2 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 2, i8 -1)
  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
  %res3 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 3, i8 -1)
  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
  %res4 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 4, i8 -1)
  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
  %res5 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 5, i8 -1)
  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
  %res6 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 6, i8 -1)
  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
  %res7 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 7, i8 -1)
  %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
  ret <8 x i8> %vec7
}

define <8 x i8> @test_mask_cmp_w_128(<8 x i16> %a0, <8 x i16> %a1, i8 %mask) {
; CHECK-LABEL: test_mask_cmp_w_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpcmpeqw %xmm1, %xmm0, %k0 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x3f,0xc1,0x00]
; CHECK-NEXT:    vpcmpltw %xmm1, %xmm0, %k2 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x3f,0xd1,0x01]
; CHECK-NEXT:    vpcmplew %xmm1, %xmm0, %k3 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x3f,0xd9,0x02]
; CHECK-NEXT:    vpcmpunordw %xmm1, %xmm0, %k4 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x3f,0xe1,0x03]
; CHECK-NEXT:    vpcmpneqw %xmm1, %xmm0, %k5 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x3f,0xe9,0x04]
; CHECK-NEXT:    vpcmpnltw %xmm1, %xmm0, %k6 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x3f,0xf1,0x05]
; CHECK-NEXT:    vpcmpnlew %xmm1, %xmm0, %k7 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x3f,0xf9,0x06]
; CHECK-NEXT:    vpcmpordw %xmm1, %xmm0, %k1 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x3f,0xc9,0x07]
; CHECK-NEXT:    kmovw %k0, %eax ## encoding: [0xc5,0xf8,0x93,0xc0]
; CHECK-NEXT:    vpinsrb $0, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf3,0x7d,0x08,0x20,0xc0,0x00]
; CHECK-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
; CHECK-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf3,0x7d,0x08,0x20,0xc0,0x02]
; CHECK-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
; CHECK-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf3,0x7d,0x08,0x20,0xc0,0x04]
; CHECK-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
; CHECK-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf3,0x7d,0x08,0x20,0xc0,0x06]
; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
; CHECK-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf3,0x7d,0x08,0x20,0xc0,0x08]
; CHECK-NEXT:    kmovw %k6, %eax ## encoding: [0xc5,0xf8,0x93,0xc6]
; CHECK-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf3,0x7d,0x08,0x20,0xc0,0x0a]
; CHECK-NEXT:    kmovw %k7, %eax ## encoding: [0xc5,0xf8,0x93,0xc7]
; CHECK-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf3,0x7d,0x08,0x20,0xc0,0x0c]
; CHECK-NEXT:    kmovw %k1, %eax ## encoding: [0xc5,0xf8,0x93,0xc1]
; CHECK-NEXT:    vpinsrb $14, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf3,0x7d,0x08,0x20,0xc0,0x0e]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res0 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 0, i8 %mask)
  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
  %res1 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 1, i8 %mask)
  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
  %res2 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 2, i8 %mask)
  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
  %res3 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 3, i8 %mask)
  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
  %res4 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 4, i8 %mask)
  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
  %res5 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 5, i8 %mask)
  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
  %res6 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 6, i8 %mask)
  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
  %res7 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 7, i8 %mask)
  %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
  ret <8 x i8> %vec7
}

declare i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16>, <8 x i16>, i32, i8) nounwind readnone

define <8 x i8> @test_ucmp_w_128(<8 x i16> %a0, <8 x i16> %a1) {
; CHECK-LABEL: test_ucmp_w_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpcmpequw %xmm1, %xmm0, %k0 ## encoding: [0x62,0xf3,0xfd,0x08,0x3e,0xc1,0x00]
; CHECK-NEXT:    vpcmpltuw %xmm1, %xmm0, %k1 ## encoding: [0x62,0xf3,0xfd,0x08,0x3e,0xc9,0x01]
; CHECK-NEXT:    vpcmpleuw %xmm1, %xmm0, %k2 ## encoding: [0x62,0xf3,0xfd,0x08,0x3e,0xd1,0x02]
; CHECK-NEXT:    vpcmpunorduw %xmm1, %xmm0, %k3 ## encoding: [0x62,0xf3,0xfd,0x08,0x3e,0xd9,0x03]
; CHECK-NEXT:    vpcmpnequw %xmm1, %xmm0, %k4 ## encoding: [0x62,0xf3,0xfd,0x08,0x3e,0xe1,0x04]
; CHECK-NEXT:    vpcmpnltuw %xmm1, %xmm0, %k5 ## encoding: [0x62,0xf3,0xfd,0x08,0x3e,0xe9,0x05]
; CHECK-NEXT:    vpcmpnleuw %xmm1, %xmm0, %k6 ## encoding: [0x62,0xf3,0xfd,0x08,0x3e,0xf1,0x06]
; CHECK-NEXT:    vpcmporduw %xmm1, %xmm0, %k7 ## encoding: [0x62,0xf3,0xfd,0x08,0x3e,0xf9,0x07]
; CHECK-NEXT:    kmovw %k0, %eax ## encoding: [0xc5,0xf8,0x93,0xc0]
; CHECK-NEXT:    vpinsrb $0, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf3,0x7d,0x08,0x20,0xc0,0x00]
; CHECK-NEXT:    kmovw %k1, %eax ## encoding: [0xc5,0xf8,0x93,0xc1]
; CHECK-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf3,0x7d,0x08,0x20,0xc0,0x02]
; CHECK-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
; CHECK-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf3,0x7d,0x08,0x20,0xc0,0x04]
; CHECK-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
; CHECK-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf3,0x7d,0x08,0x20,0xc0,0x06]
; CHECK-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
; CHECK-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf3,0x7d,0x08,0x20,0xc0,0x08]
; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
; CHECK-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf3,0x7d,0x08,0x20,0xc0,0x0a]
; CHECK-NEXT:    kmovw %k6, %eax ## encoding: [0xc5,0xf8,0x93,0xc6]
; CHECK-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf3,0x7d,0x08,0x20,0xc0,0x0c]
; CHECK-NEXT:    kmovw %k7, %eax ## encoding: [0xc5,0xf8,0x93,0xc7]
; CHECK-NEXT:    vpinsrb $14, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf3,0x7d,0x08,0x20,0xc0,0x0e]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res0 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 0, i8 -1)
  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
  %res1 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 1, i8 -1)
  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
  %res2 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 2, i8 -1)
  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
  %res3 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 3, i8 -1)
  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
  %res4 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 4, i8 -1)
  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
  %res5 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 5, i8 -1)
  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
  %res6 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 6, i8 -1)
  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
  %res7 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 7, i8 -1)
  %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
  ret <8 x i8> %vec7
}

define <8 x i8> @test_mask_ucmp_w_128(<8 x i16> %a0, <8 x i16> %a1, i8 %mask) {
; CHECK-LABEL: test_mask_ucmp_w_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpcmpequw %xmm1, %xmm0, %k0 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x3e,0xc1,0x00]
; CHECK-NEXT:    vpcmpltuw %xmm1, %xmm0, %k2 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x3e,0xd1,0x01]
; CHECK-NEXT:    vpcmpleuw %xmm1, %xmm0, %k3 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x3e,0xd9,0x02]
; CHECK-NEXT:    vpcmpunorduw %xmm1, %xmm0, %k4 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x3e,0xe1,0x03]
; CHECK-NEXT:    vpcmpnequw %xmm1, %xmm0, %k5 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x3e,0xe9,0x04]
; CHECK-NEXT:    vpcmpnltuw %xmm1, %xmm0, %k6 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x3e,0xf1,0x05]
; CHECK-NEXT:    vpcmpnleuw %xmm1, %xmm0, %k7 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x3e,0xf9,0x06]
; CHECK-NEXT:    vpcmporduw %xmm1, %xmm0, %k1 {%k1} ## encoding: [0x62,0xf3,0xfd,0x09,0x3e,0xc9,0x07]
; CHECK-NEXT:    kmovw %k0, %eax ## encoding: [0xc5,0xf8,0x93,0xc0]
; CHECK-NEXT:    vpinsrb $0, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf3,0x7d,0x08,0x20,0xc0,0x00]
; CHECK-NEXT:    kmovw %k2, %eax ## encoding: [0xc5,0xf8,0x93,0xc2]
; CHECK-NEXT:    vpinsrb $2, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf3,0x7d,0x08,0x20,0xc0,0x02]
; CHECK-NEXT:    kmovw %k3, %eax ## encoding: [0xc5,0xf8,0x93,0xc3]
; CHECK-NEXT:    vpinsrb $4, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf3,0x7d,0x08,0x20,0xc0,0x04]
; CHECK-NEXT:    kmovw %k4, %eax ## encoding: [0xc5,0xf8,0x93,0xc4]
; CHECK-NEXT:    vpinsrb $6, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf3,0x7d,0x08,0x20,0xc0,0x06]
; CHECK-NEXT:    kmovw %k5, %eax ## encoding: [0xc5,0xf8,0x93,0xc5]
; CHECK-NEXT:    vpinsrb $8, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf3,0x7d,0x08,0x20,0xc0,0x08]
; CHECK-NEXT:    kmovw %k6, %eax ## encoding: [0xc5,0xf8,0x93,0xc6]
; CHECK-NEXT:    vpinsrb $10, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf3,0x7d,0x08,0x20,0xc0,0x0a]
; CHECK-NEXT:    kmovw %k7, %eax ## encoding: [0xc5,0xf8,0x93,0xc7]
; CHECK-NEXT:    vpinsrb $12, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf3,0x7d,0x08,0x20,0xc0,0x0c]
; CHECK-NEXT:    kmovw %k1, %eax ## encoding: [0xc5,0xf8,0x93,0xc1]
; CHECK-NEXT:    vpinsrb $14, %eax, %xmm0, %xmm0 ## encoding: [0x62,0xf3,0x7d,0x08,0x20,0xc0,0x0e]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res0 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 0, i8 %mask)
  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
  %res1 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 1, i8 %mask)
  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
  %res2 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 2, i8 %mask)
  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
  %res3 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 3, i8 %mask)
  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
  %res4 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 4, i8 %mask)
  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
  %res5 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 5, i8 %mask)
  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
  %res6 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 6, i8 %mask)
  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
  %res7 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 7, i8 %mask)
  %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
  ret <8 x i8> %vec7
}

declare i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16>, <8 x i16>, i32, i8) nounwind readnone

declare <8 x float> @llvm.x86.avx512.mask.vfmadd.ps.256(<8 x float>, <8 x float>, <8 x float>, i8) nounwind readnone

define <8 x float> @test_mask_vfmadd256_ps(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2, i8 %mask) {
; CHECK-LABEL: test_mask_vfmadd256_ps:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vfmadd213ps %ymm2, %ymm1, %ymm0 {%k1} ## encoding: [0x62,0xf2,0x75,0x29,0xa8,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask.vfmadd.ps.256(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2, i8 %mask) nounwind
  ret <8 x float> %res
}

declare <4 x float> @llvm.x86.avx512.mask.vfmadd.ps.128(<4 x float>, <4 x float>, <4 x float>, i8) nounwind readnone

define <4 x float> @test_mask_vfmadd128_ps(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2, i8 %mask) {
; CHECK-LABEL: test_mask_vfmadd128_ps:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vfmadd213ps %xmm2, %xmm1, %xmm0 {%k1} ## encoding: [0x62,0xf2,0x75,0x09,0xa8,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask.vfmadd.ps.128(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2, i8 %mask) nounwind
  ret <4 x float> %res
}

declare <4 x double> @llvm.x86.avx512.mask.vfmadd.pd.256(<4 x double>, <4 x double>, <4 x double>, i8)

define <4 x double> @test_mask_fmadd256_pd(<4 x double> %a, <4 x double> %b, <4 x double> %c, i8 %mask) {
; CHECK-LABEL: test_mask_fmadd256_pd:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vfmadd213pd %ymm2, %ymm1, %ymm0 {%k1} ## encoding: [0x62,0xf2,0xf5,0x29,0xa8,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <4 x double> @llvm.x86.avx512.mask.vfmadd.pd.256(<4 x double> %a, <4 x double> %b, <4 x double> %c, i8 %mask)
  ret <4 x double> %res
}

declare <2 x double> @llvm.x86.avx512.mask.vfmadd.pd.128(<2 x double>, <2 x double>, <2 x double>, i8)

define <2 x double> @test_mask_fmadd128_pd(<2 x double> %a, <2 x double> %b, <2 x double> %c, i8 %mask) {
; CHECK-LABEL: test_mask_fmadd128_pd:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vfmadd213pd %xmm2, %xmm1, %xmm0 {%k1} ## encoding: [0x62,0xf2,0xf5,0x09,0xa8,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <2 x double> @llvm.x86.avx512.mask.vfmadd.pd.128(<2 x double> %a, <2 x double> %b, <2 x double> %c, i8 %mask)
  ret <2 x double> %res
}

define <2 x double>@test_int_x86_avx512_mask_vfmadd_pd_128(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_vfmadd_pd_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vmovaps %zmm0, %zmm3 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xd8]
; CHECK-NEXT:    vfmadd213pd %xmm2, %xmm1, %xmm3 {%k1} ## encoding: [0x62,0xf2,0xf5,0x09,0xa8,0xda]
; CHECK-NEXT:    vfmadd213pd %xmm2, %xmm1, %xmm0 ## encoding: [0x62,0xf2,0xf5,0x08,0xa8,0xc2]
; CHECK-NEXT:    vaddpd %xmm0, %xmm3, %xmm0 ## encoding: [0x62,0xf1,0xe5,0x08,0x58,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <2 x double> @llvm.x86.avx512.mask.vfmadd.pd.128(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 %x3)
  %res1 = call <2 x double> @llvm.x86.avx512.mask.vfmadd.pd.128(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 -1)
  %res2 = fadd <2 x double> %res, %res1
  ret <2 x double> %res2
}

declare <2 x double> @llvm.x86.avx512.mask3.vfmadd.pd.128(<2 x double>, <2 x double>, <2 x double>, i8)

define <2 x double>@test_int_x86_avx512_mask3_vfmadd_pd_128(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask3_vfmadd_pd_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vmovaps %zmm2, %zmm3 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xda]
; CHECK-NEXT:    vfmadd231pd %xmm1, %xmm0, %xmm3 {%k1} ## encoding: [0x62,0xf2,0xfd,0x09,0xb8,0xd9]
; CHECK-NEXT:    vfmadd213pd %xmm2, %xmm1, %xmm0 ## encoding: [0x62,0xf2,0xf5,0x08,0xa8,0xc2]
; CHECK-NEXT:    vaddpd %xmm0, %xmm3, %xmm0 ## encoding: [0x62,0xf1,0xe5,0x08,0x58,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <2 x double> @llvm.x86.avx512.mask3.vfmadd.pd.128(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 %x3)
  %res1 = call <2 x double> @llvm.x86.avx512.mask3.vfmadd.pd.128(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 -1)
  %res2 = fadd <2 x double> %res, %res1
  ret <2 x double> %res2
}

declare <2 x double> @llvm.x86.avx512.maskz.vfmadd.pd.128(<2 x double>, <2 x double>, <2 x double>, i8)

define <2 x double>@test_int_x86_avx512_maskz_vfmadd_pd_128(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_maskz_vfmadd_pd_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vmovaps %zmm0, %zmm3 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xd8]
; CHECK-NEXT:    vfmadd213pd %xmm2, %xmm1, %xmm3 {%k1} {z} ## encoding: [0x62,0xf2,0xf5,0x89,0xa8,0xda]
; CHECK-NEXT:    vfmadd213pd %xmm2, %xmm1, %xmm0 ## encoding: [0x62,0xf2,0xf5,0x08,0xa8,0xc2]
; CHECK-NEXT:    vaddpd %xmm0, %xmm3, %xmm0 ## encoding: [0x62,0xf1,0xe5,0x08,0x58,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <2 x double> @llvm.x86.avx512.maskz.vfmadd.pd.128(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 %x3)
  %res1 = call <2 x double> @llvm.x86.avx512.maskz.vfmadd.pd.128(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 -1)
  %res2 = fadd <2 x double> %res, %res1
  ret <2 x double> %res2
}

define <4 x double>@test_int_x86_avx512_mask_vfmadd_pd_256(<4 x double> %x0, <4 x double> %x1, <4 x double> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_vfmadd_pd_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vmovaps %zmm0, %zmm3 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xd8]
; CHECK-NEXT:    vfmadd213pd %ymm2, %ymm1, %ymm3 {%k1} ## encoding: [0x62,0xf2,0xf5,0x29,0xa8,0xda]
; CHECK-NEXT:    vfmadd213pd %ymm2, %ymm1, %ymm0 ## encoding: [0x62,0xf2,0xf5,0x28,0xa8,0xc2]
; CHECK-NEXT:    vaddpd %ymm0, %ymm3, %ymm0 ## encoding: [0x62,0xf1,0xe5,0x28,0x58,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <4 x double> @llvm.x86.avx512.mask.vfmadd.pd.256(<4 x double> %x0, <4 x double> %x1, <4 x double> %x2, i8 %x3)
  %res1 = call <4 x double> @llvm.x86.avx512.mask.vfmadd.pd.256(<4 x double> %x0, <4 x double> %x1, <4 x double> %x2, i8 -1)
  %res2 = fadd <4 x double> %res, %res1
  ret <4 x double> %res2
}

declare <4 x double> @llvm.x86.avx512.mask3.vfmadd.pd.256(<4 x double>, <4 x double>, <4 x double>, i8)

define <4 x double>@test_int_x86_avx512_mask3_vfmadd_pd_256(<4 x double> %x0, <4 x double> %x1, <4 x double> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask3_vfmadd_pd_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vmovaps %zmm2, %zmm3 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xda]
; CHECK-NEXT:    vfmadd231pd %ymm1, %ymm0, %ymm3 {%k1} ## encoding: [0x62,0xf2,0xfd,0x29,0xb8,0xd9]
; CHECK-NEXT:    vfmadd213pd %ymm2, %ymm1, %ymm0 ## encoding: [0x62,0xf2,0xf5,0x28,0xa8,0xc2]
; CHECK-NEXT:    vaddpd %ymm0, %ymm3, %ymm0 ## encoding: [0x62,0xf1,0xe5,0x28,0x58,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <4 x double> @llvm.x86.avx512.mask3.vfmadd.pd.256(<4 x double> %x0, <4 x double> %x1, <4 x double> %x2, i8 %x3)
  %res1 = call <4 x double> @llvm.x86.avx512.mask3.vfmadd.pd.256(<4 x double> %x0, <4 x double> %x1, <4 x double> %x2, i8 -1)
  %res2 = fadd <4 x double> %res, %res1
  ret <4 x double> %res2
}

declare <4 x double> @llvm.x86.avx512.maskz.vfmadd.pd.256(<4 x double>, <4 x double>, <4 x double>, i8)

define <4 x double>@test_int_x86_avx512_maskz_vfmadd_pd_256(<4 x double> %x0, <4 x double> %x1, <4 x double> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_maskz_vfmadd_pd_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vmovaps %zmm0, %zmm3 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xd8]
; CHECK-NEXT:    vfmadd213pd %ymm2, %ymm1, %ymm3 {%k1} {z} ## encoding: [0x62,0xf2,0xf5,0xa9,0xa8,0xda]
; CHECK-NEXT:    vfmadd213pd %ymm2, %ymm1, %ymm0 ## encoding: [0x62,0xf2,0xf5,0x28,0xa8,0xc2]
; CHECK-NEXT:    vaddpd %ymm0, %ymm3, %ymm0 ## encoding: [0x62,0xf1,0xe5,0x28,0x58,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <4 x double> @llvm.x86.avx512.maskz.vfmadd.pd.256(<4 x double> %x0, <4 x double> %x1, <4 x double> %x2, i8 %x3)
  %res1 = call <4 x double> @llvm.x86.avx512.maskz.vfmadd.pd.256(<4 x double> %x0, <4 x double> %x1, <4 x double> %x2, i8 -1)
  %res2 = fadd <4 x double> %res, %res1
  ret <4 x double> %res2
}

define <4 x float>@test_int_x86_avx512_mask_vfmadd_ps_128(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_vfmadd_ps_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vmovaps %zmm0, %zmm3 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xd8]
; CHECK-NEXT:    vfmadd213ps %xmm2, %xmm1, %xmm3 {%k1} ## encoding: [0x62,0xf2,0x75,0x09,0xa8,0xda]
; CHECK-NEXT:    vfmadd213ps %xmm2, %xmm1, %xmm0 ## encoding: [0x62,0xf2,0x75,0x08,0xa8,0xc2]
; CHECK-NEXT:    vaddps %xmm0, %xmm3, %xmm0 ## encoding: [0x62,0xf1,0x64,0x08,0x58,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask.vfmadd.ps.128(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 %x3)
  %res1 = call <4 x float> @llvm.x86.avx512.mask.vfmadd.ps.128(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 -1)
  %res2 = fadd <4 x float> %res, %res1
  ret <4 x float> %res2
}

declare <4 x float> @llvm.x86.avx512.mask3.vfmadd.ps.128(<4 x float>, <4 x float>, <4 x float>, i8)

define <4 x float>@test_int_x86_avx512_mask3_vfmadd_ps_128(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask3_vfmadd_ps_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vmovaps %zmm2, %zmm3 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xda]
; CHECK-NEXT:    vfmadd231ps %xmm1, %xmm0, %xmm3 {%k1} ## encoding: [0x62,0xf2,0x7d,0x09,0xb8,0xd9]
; CHECK-NEXT:    vfmadd213ps %xmm2, %xmm1, %xmm0 ## encoding: [0x62,0xf2,0x75,0x08,0xa8,0xc2]
; CHECK-NEXT:    vaddps %xmm0, %xmm3, %xmm0 ## encoding: [0x62,0xf1,0x64,0x08,0x58,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask3.vfmadd.ps.128(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 %x3)
  %res1 = call <4 x float> @llvm.x86.avx512.mask3.vfmadd.ps.128(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 -1)
  %res2 = fadd <4 x float> %res, %res1
  ret <4 x float> %res2
}

declare <4 x float> @llvm.x86.avx512.maskz.vfmadd.ps.128(<4 x float>, <4 x float>, <4 x float>, i8)

define <4 x float>@test_int_x86_avx512_maskz_vfmadd_ps_128(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_maskz_vfmadd_ps_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vmovaps %zmm0, %zmm3 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xd8]
; CHECK-NEXT:    vfmadd213ps %xmm2, %xmm1, %xmm3 {%k1} {z} ## encoding: [0x62,0xf2,0x75,0x89,0xa8,0xda]
; CHECK-NEXT:    vfmadd213ps %xmm2, %xmm1, %xmm0 ## encoding: [0x62,0xf2,0x75,0x08,0xa8,0xc2]
; CHECK-NEXT:    vaddps %xmm0, %xmm3, %xmm0 ## encoding: [0x62,0xf1,0x64,0x08,0x58,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.maskz.vfmadd.ps.128(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 %x3)
  %res1 = call <4 x float> @llvm.x86.avx512.maskz.vfmadd.ps.128(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 -1)
  %res2 = fadd <4 x float> %res, %res1
  ret <4 x float> %res2
}

define <8 x float>@test_int_x86_avx512_mask_vfmadd_ps_256(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_vfmadd_ps_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vmovaps %zmm0, %zmm3 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xd8]
; CHECK-NEXT:    vfmadd213ps %ymm2, %ymm1, %ymm3 {%k1} ## encoding: [0x62,0xf2,0x75,0x29,0xa8,0xda]
; CHECK-NEXT:    vfmadd213ps %ymm2, %ymm1, %ymm0 ## encoding: [0x62,0xf2,0x75,0x28,0xa8,0xc2]
; CHECK-NEXT:    vaddps %ymm0, %ymm3, %ymm0 ## encoding: [0x62,0xf1,0x64,0x28,0x58,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask.vfmadd.ps.256(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2, i8 %x3)
  %res1 = call <8 x float> @llvm.x86.avx512.mask.vfmadd.ps.256(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2, i8 -1)
  %res2 = fadd <8 x float> %res, %res1
  ret <8 x float> %res2
}

declare <8 x float> @llvm.x86.avx512.mask3.vfmadd.ps.256(<8 x float>, <8 x float>, <8 x float>, i8)

define <8 x float>@test_int_x86_avx512_mask3_vfmadd_ps_256(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask3_vfmadd_ps_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vmovaps %zmm2, %zmm3 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xda]
; CHECK-NEXT:    vfmadd231ps %ymm1, %ymm0, %ymm3 {%k1} ## encoding: [0x62,0xf2,0x7d,0x29,0xb8,0xd9]
; CHECK-NEXT:    vfmadd213ps %ymm2, %ymm1, %ymm0 ## encoding: [0x62,0xf2,0x75,0x28,0xa8,0xc2]
; CHECK-NEXT:    vaddps %ymm0, %ymm3, %ymm0 ## encoding: [0x62,0xf1,0x64,0x28,0x58,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask3.vfmadd.ps.256(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2, i8 %x3)
  %res1 = call <8 x float> @llvm.x86.avx512.mask3.vfmadd.ps.256(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2, i8 -1)
  %res2 = fadd <8 x float> %res, %res1
  ret <8 x float> %res2
}

declare <8 x float> @llvm.x86.avx512.maskz.vfmadd.ps.256(<8 x float>, <8 x float>, <8 x float>, i8)

define <8 x float>@test_int_x86_avx512_maskz_vfmadd_ps_256(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_maskz_vfmadd_ps_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vmovaps %zmm0, %zmm3 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xd8]
; CHECK-NEXT:    vfmadd213ps %ymm2, %ymm1, %ymm3 {%k1} {z} ## encoding: [0x62,0xf2,0x75,0xa9,0xa8,0xda]
; CHECK-NEXT:    vfmadd213ps %ymm2, %ymm1, %ymm0 ## encoding: [0x62,0xf2,0x75,0x28,0xa8,0xc2]
; CHECK-NEXT:    vaddps %ymm0, %ymm3, %ymm0 ## encoding: [0x62,0xf1,0x64,0x28,0x58,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.maskz.vfmadd.ps.256(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2, i8 %x3)
  %res1 = call <8 x float> @llvm.x86.avx512.maskz.vfmadd.ps.256(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2, i8 -1)
  %res2 = fadd <8 x float> %res, %res1
  ret <8 x float> %res2
}


declare <2 x double> @llvm.x86.avx512.mask3.vfmsub.pd.128(<2 x double>, <2 x double>, <2 x double>, i8)

define <2 x double>@test_int_x86_avx512_mask3_vfmsub_pd_128(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask3_vfmsub_pd_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vmovaps %zmm2, %zmm3 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xda]
; CHECK-NEXT:    vfmsub231pd %xmm1, %xmm0, %xmm3 {%k1} ## encoding: [0x62,0xf2,0xfd,0x09,0xba,0xd9]
; CHECK-NEXT:    vfmsub213pd %xmm2, %xmm1, %xmm0 ## encoding: [0x62,0xf2,0xf5,0x08,0xaa,0xc2]
; CHECK-NEXT:    vaddpd %xmm0, %xmm3, %xmm0 ## encoding: [0x62,0xf1,0xe5,0x08,0x58,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <2 x double> @llvm.x86.avx512.mask3.vfmsub.pd.128(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 %x3)
  %res1 = call <2 x double> @llvm.x86.avx512.mask3.vfmsub.pd.128(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 -1)
  %res2 = fadd <2 x double> %res, %res1
  ret <2 x double> %res2
}


declare <4 x double> @llvm.x86.avx512.mask3.vfmsub.pd.256(<4 x double>, <4 x double>, <4 x double>, i8)

define <4 x double>@test_int_x86_avx512_mask3_vfmsub_pd_256(<4 x double> %x0, <4 x double> %x1, <4 x double> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask3_vfmsub_pd_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vmovaps %zmm2, %zmm3 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xda]
; CHECK-NEXT:    vfmsub231pd %ymm1, %ymm0, %ymm3 {%k1} ## encoding: [0x62,0xf2,0xfd,0x29,0xba,0xd9]
; CHECK-NEXT:    vfmsub213pd %ymm2, %ymm1, %ymm0 ## encoding: [0x62,0xf2,0xf5,0x28,0xaa,0xc2]
; CHECK-NEXT:    vaddpd %ymm0, %ymm3, %ymm0 ## encoding: [0x62,0xf1,0xe5,0x28,0x58,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <4 x double> @llvm.x86.avx512.mask3.vfmsub.pd.256(<4 x double> %x0, <4 x double> %x1, <4 x double> %x2, i8 %x3)
  %res1 = call <4 x double> @llvm.x86.avx512.mask3.vfmsub.pd.256(<4 x double> %x0, <4 x double> %x1, <4 x double> %x2, i8 -1)
  %res2 = fadd <4 x double> %res, %res1
  ret <4 x double> %res2
}

declare <4 x float> @llvm.x86.avx512.mask3.vfmsub.ps.128(<4 x float>, <4 x float>, <4 x float>, i8)

define <4 x float>@test_int_x86_avx512_mask3_vfmsub_ps_128(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask3_vfmsub_ps_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vmovaps %zmm2, %zmm3 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xda]
; CHECK-NEXT:    vfmsub231ps %xmm1, %xmm0, %xmm3 {%k1} ## encoding: [0x62,0xf2,0x7d,0x09,0xba,0xd9]
; CHECK-NEXT:    vfmsub213ps %xmm2, %xmm1, %xmm0 ## encoding: [0x62,0xf2,0x75,0x08,0xaa,0xc2]
; CHECK-NEXT:    vaddps %xmm0, %xmm3, %xmm0 ## encoding: [0x62,0xf1,0x64,0x08,0x58,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask3.vfmsub.ps.128(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 %x3)
  %res1 = call <4 x float> @llvm.x86.avx512.mask3.vfmsub.ps.128(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 -1)
  %res2 = fadd <4 x float> %res, %res1
  ret <4 x float> %res2
}

declare <8 x float> @llvm.x86.avx512.mask3.vfmsub.ps.256(<8 x float>, <8 x float>, <8 x float>, i8)

define <8 x float>@test_int_x86_avx512_mask3_vfmsub_ps_256(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask3_vfmsub_ps_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vmovaps %zmm2, %zmm3 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xda]
; CHECK-NEXT:    vfmsub231ps %ymm1, %ymm0, %ymm3 {%k1} ## encoding: [0x62,0xf2,0x7d,0x29,0xba,0xd9]
; CHECK-NEXT:    vfmsub213ps %ymm2, %ymm1, %ymm0 ## encoding: [0x62,0xf2,0x75,0x28,0xaa,0xc2]
; CHECK-NEXT:    vaddps %ymm0, %ymm3, %ymm0 ## encoding: [0x62,0xf1,0x64,0x28,0x58,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask3.vfmsub.ps.256(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2, i8 %x3)
  %res1 = call <8 x float> @llvm.x86.avx512.mask3.vfmsub.ps.256(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2, i8 -1)
  %res2 = fadd <8 x float> %res, %res1
  ret <8 x float> %res2
}

declare <8 x float> @llvm.x86.avx512.mask.vfnmadd.ps.256(<8 x float>, <8 x float>, <8 x float>, i8) nounwind readnone

define <8 x float> @test_mask_vfnmadd256_ps(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2, i8 %mask) {
; CHECK-LABEL: test_mask_vfnmadd256_ps:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vfnmadd213ps %ymm2, %ymm1, %ymm0 {%k1} ## encoding: [0x62,0xf2,0x75,0x29,0xac,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask.vfnmadd.ps.256(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2, i8 %mask) nounwind
  ret <8 x float> %res
}

declare <4 x float> @llvm.x86.avx512.mask.vfnmadd.ps.128(<4 x float>, <4 x float>, <4 x float>, i8) nounwind readnone

define <4 x float> @test_mask_vfnmadd128_ps(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2, i8 %mask) {
; CHECK-LABEL: test_mask_vfnmadd128_ps:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vfnmadd213ps %xmm2, %xmm1, %xmm0 {%k1} ## encoding: [0x62,0xf2,0x75,0x09,0xac,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask.vfnmadd.ps.128(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2, i8 %mask) nounwind
  ret <4 x float> %res
}

declare <4 x double> @llvm.x86.avx512.mask.vfnmadd.pd.256(<4 x double>, <4 x double>, <4 x double>, i8) nounwind readnone

define <4 x double> @test_mask_vfnmadd256_pd(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2, i8 %mask) {
; CHECK-LABEL: test_mask_vfnmadd256_pd:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vfnmadd213pd %ymm2, %ymm1, %ymm0 {%k1} ## encoding: [0x62,0xf2,0xf5,0x29,0xac,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <4 x double> @llvm.x86.avx512.mask.vfnmadd.pd.256(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2, i8 %mask) nounwind
  ret <4 x double> %res
}

declare <2 x double> @llvm.x86.avx512.mask.vfnmadd.pd.128(<2 x double>, <2 x double>, <2 x double>, i8) nounwind readnone

define <2 x double> @test_mask_vfnmadd128_pd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2, i8 %mask) {
; CHECK-LABEL: test_mask_vfnmadd128_pd:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vfnmadd213pd %xmm2, %xmm1, %xmm0 {%k1} ## encoding: [0x62,0xf2,0xf5,0x09,0xac,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <2 x double> @llvm.x86.avx512.mask.vfnmadd.pd.128(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2, i8 %mask) nounwind
  ret <2 x double> %res
}

declare <8 x float> @llvm.x86.avx512.mask.vfnmsub.ps.256(<8 x float>, <8 x float>, <8 x float>, i8) nounwind readnone

define <8 x float> @test_mask_vfnmsub256_ps(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2, i8 %mask) {
; CHECK-LABEL: test_mask_vfnmsub256_ps:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vfnmsub213ps %ymm2, %ymm1, %ymm0 {%k1} ## encoding: [0x62,0xf2,0x75,0x29,0xae,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask.vfnmsub.ps.256(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2, i8 %mask) nounwind
  ret <8 x float> %res
}

declare <4 x float> @llvm.x86.avx512.mask.vfnmsub.ps.128(<4 x float>, <4 x float>, <4 x float>, i8) nounwind readnone

define <4 x float> @test_mask_vfnmsub128_ps(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2, i8 %mask) {
; CHECK-LABEL: test_mask_vfnmsub128_ps:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vfnmsub213ps %xmm2, %xmm1, %xmm0 {%k1} ## encoding: [0x62,0xf2,0x75,0x09,0xae,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask.vfnmsub.ps.128(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2, i8 %mask) nounwind
  ret <4 x float> %res
}

declare <4 x double> @llvm.x86.avx512.mask.vfnmsub.pd.256(<4 x double>, <4 x double>, <4 x double>, i8) nounwind readnone

define <4 x double> @test_mask_vfnmsub256_pd(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2, i8 %mask) {
; CHECK-LABEL: test_mask_vfnmsub256_pd:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vfnmsub213pd %ymm2, %ymm1, %ymm0 {%k1} ## encoding: [0x62,0xf2,0xf5,0x29,0xae,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <4 x double> @llvm.x86.avx512.mask.vfnmsub.pd.256(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2, i8 %mask) nounwind
  ret <4 x double> %res
}

declare <2 x double> @llvm.x86.avx512.mask.vfnmsub.pd.128(<2 x double>, <2 x double>, <2 x double>, i8) nounwind readnone

define <2 x double> @test_mask_vfnmsub128_pd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2, i8 %mask) {
; CHECK-LABEL: test_mask_vfnmsub128_pd:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vfnmsub213pd %xmm2, %xmm1, %xmm0 {%k1} ## encoding: [0x62,0xf2,0xf5,0x09,0xae,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <2 x double> @llvm.x86.avx512.mask.vfnmsub.pd.128(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2, i8 %mask) nounwind
  ret <2 x double> %res
}


define <2 x double>@test_int_x86_avx512_mask_vfnmsub_pd_128(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_vfnmsub_pd_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vmovaps %zmm0, %zmm3 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xd8]
; CHECK-NEXT:    vfnmsub213pd %xmm2, %xmm1, %xmm3 {%k1} ## encoding: [0x62,0xf2,0xf5,0x09,0xae,0xda]
; CHECK-NEXT:    vfnmsub213pd %xmm2, %xmm1, %xmm0 ## encoding: [0x62,0xf2,0xf5,0x08,0xae,0xc2]
; CHECK-NEXT:    vaddpd %xmm0, %xmm3, %xmm0 ## encoding: [0x62,0xf1,0xe5,0x08,0x58,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <2 x double> @llvm.x86.avx512.mask.vfnmsub.pd.128(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 %x3)
  %res1 = call <2 x double> @llvm.x86.avx512.mask.vfnmsub.pd.128(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 -1)
  %res2 = fadd <2 x double> %res, %res1
  ret <2 x double> %res2
}

declare <2 x double> @llvm.x86.avx512.mask3.vfnmsub.pd.128(<2 x double>, <2 x double>, <2 x double>, i8)

define <2 x double>@test_int_x86_avx512_mask3_vfnmsub_pd_128(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask3_vfnmsub_pd_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vmovaps %zmm2, %zmm3 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xda]
; CHECK-NEXT:    vfnmsub231pd %xmm1, %xmm0, %xmm3 {%k1} ## encoding: [0x62,0xf2,0xfd,0x09,0xbe,0xd9]
; CHECK-NEXT:    vfnmsub213pd %xmm2, %xmm1, %xmm0 ## encoding: [0x62,0xf2,0xf5,0x08,0xae,0xc2]
; CHECK-NEXT:    vaddpd %xmm0, %xmm3, %xmm0 ## encoding: [0x62,0xf1,0xe5,0x08,0x58,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <2 x double> @llvm.x86.avx512.mask3.vfnmsub.pd.128(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 %x3)
  %res1 = call <2 x double> @llvm.x86.avx512.mask3.vfnmsub.pd.128(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 -1)
  %res2 = fadd <2 x double> %res, %res1
  ret <2 x double> %res2
}

define <4 x double>@test_int_x86_avx512_mask_vfnmsub_pd_256(<4 x double> %x0, <4 x double> %x1, <4 x double> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_vfnmsub_pd_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vmovaps %zmm0, %zmm3 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xd8]
; CHECK-NEXT:    vfnmsub213pd %ymm2, %ymm1, %ymm3 {%k1} ## encoding: [0x62,0xf2,0xf5,0x29,0xae,0xda]
; CHECK-NEXT:    vfnmsub213pd %ymm2, %ymm1, %ymm0 ## encoding: [0x62,0xf2,0xf5,0x28,0xae,0xc2]
; CHECK-NEXT:    vaddpd %ymm0, %ymm3, %ymm0 ## encoding: [0x62,0xf1,0xe5,0x28,0x58,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <4 x double> @llvm.x86.avx512.mask.vfnmsub.pd.256(<4 x double> %x0, <4 x double> %x1, <4 x double> %x2, i8 %x3)
  %res1 = call <4 x double> @llvm.x86.avx512.mask.vfnmsub.pd.256(<4 x double> %x0, <4 x double> %x1, <4 x double> %x2, i8 -1)
  %res2 = fadd <4 x double> %res, %res1
  ret <4 x double> %res2
}

declare <4 x double> @llvm.x86.avx512.mask3.vfnmsub.pd.256(<4 x double>, <4 x double>, <4 x double>, i8)

define <4 x double>@test_int_x86_avx512_mask3_vfnmsub_pd_256(<4 x double> %x0, <4 x double> %x1, <4 x double> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask3_vfnmsub_pd_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vmovaps %zmm2, %zmm3 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xda]
; CHECK-NEXT:    vfnmsub231pd %ymm1, %ymm0, %ymm3 {%k1} ## encoding: [0x62,0xf2,0xfd,0x29,0xbe,0xd9]
; CHECK-NEXT:    vfnmsub213pd %ymm2, %ymm1, %ymm0 ## encoding: [0x62,0xf2,0xf5,0x28,0xae,0xc2]
; CHECK-NEXT:    vaddpd %ymm0, %ymm3, %ymm0 ## encoding: [0x62,0xf1,0xe5,0x28,0x58,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <4 x double> @llvm.x86.avx512.mask3.vfnmsub.pd.256(<4 x double> %x0, <4 x double> %x1, <4 x double> %x2, i8 %x3)
  %res1 = call <4 x double> @llvm.x86.avx512.mask3.vfnmsub.pd.256(<4 x double> %x0, <4 x double> %x1, <4 x double> %x2, i8 -1)
  %res2 = fadd <4 x double> %res, %res1
  ret <4 x double> %res2
}

define <4 x float>@test_int_x86_avx512_mask_vfnmsub_ps_128(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_vfnmsub_ps_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vmovaps %zmm0, %zmm3 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xd8]
; CHECK-NEXT:    vfnmsub213ps %xmm2, %xmm1, %xmm3 {%k1} ## encoding: [0x62,0xf2,0x75,0x09,0xae,0xda]
; CHECK-NEXT:    vfnmsub213ps %xmm2, %xmm1, %xmm0 ## encoding: [0x62,0xf2,0x75,0x08,0xae,0xc2]
; CHECK-NEXT:    vaddps %xmm0, %xmm3, %xmm0 ## encoding: [0x62,0xf1,0x64,0x08,0x58,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask.vfnmsub.ps.128(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 %x3)
  %res1 = call <4 x float> @llvm.x86.avx512.mask.vfnmsub.ps.128(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 -1)
  %res2 = fadd <4 x float> %res, %res1
  ret <4 x float> %res2
}

declare <4 x float> @llvm.x86.avx512.mask3.vfnmsub.ps.128(<4 x float>, <4 x float>, <4 x float>, i8)

define <4 x float>@test_int_x86_avx512_mask3_vfnmsub_ps_128(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask3_vfnmsub_ps_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vmovaps %zmm2, %zmm3 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xda]
; CHECK-NEXT:    vfnmsub231ps %xmm1, %xmm0, %xmm3 {%k1} ## encoding: [0x62,0xf2,0x7d,0x09,0xbe,0xd9]
; CHECK-NEXT:    vfnmsub213ps %xmm2, %xmm1, %xmm0 ## encoding: [0x62,0xf2,0x75,0x08,0xae,0xc2]
; CHECK-NEXT:    vaddps %xmm0, %xmm3, %xmm0 ## encoding: [0x62,0xf1,0x64,0x08,0x58,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask3.vfnmsub.ps.128(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 %x3)
  %res1 = call <4 x float> @llvm.x86.avx512.mask3.vfnmsub.ps.128(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 -1)
  %res2 = fadd <4 x float> %res, %res1
  ret <4 x float> %res2
}

define <8 x float>@test_int_x86_avx512_mask_vfnmsub_ps_256(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_vfnmsub_ps_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vmovaps %zmm0, %zmm3 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xd8]
; CHECK-NEXT:    vfnmsub213ps %ymm2, %ymm1, %ymm3 {%k1} ## encoding: [0x62,0xf2,0x75,0x29,0xae,0xda]
; CHECK-NEXT:    vfnmsub213ps %ymm2, %ymm1, %ymm0 ## encoding: [0x62,0xf2,0x75,0x28,0xae,0xc2]
; CHECK-NEXT:    vaddps %ymm0, %ymm3, %ymm0 ## encoding: [0x62,0xf1,0x64,0x28,0x58,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask.vfnmsub.ps.256(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2, i8 %x3)
  %res1 = call <8 x float> @llvm.x86.avx512.mask.vfnmsub.ps.256(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2, i8 -1)
  %res2 = fadd <8 x float> %res, %res1
  ret <8 x float> %res2
}

declare <8 x float> @llvm.x86.avx512.mask3.vfnmsub.ps.256(<8 x float>, <8 x float>, <8 x float>, i8)

define <8 x float>@test_int_x86_avx512_mask3_vfnmsub_ps_256(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask3_vfnmsub_ps_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vmovaps %zmm2, %zmm3 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xda]
; CHECK-NEXT:    vfnmsub231ps %ymm1, %ymm0, %ymm3 {%k1} ## encoding: [0x62,0xf2,0x7d,0x29,0xbe,0xd9]
; CHECK-NEXT:    vfnmsub213ps %ymm2, %ymm1, %ymm0 ## encoding: [0x62,0xf2,0x75,0x28,0xae,0xc2]
; CHECK-NEXT:    vaddps %ymm0, %ymm3, %ymm0 ## encoding: [0x62,0xf1,0x64,0x28,0x58,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask3.vfnmsub.ps.256(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2, i8 %x3)
  %res1 = call <8 x float> @llvm.x86.avx512.mask3.vfnmsub.ps.256(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2, i8 -1)
  %res2 = fadd <8 x float> %res, %res1
  ret <8 x float> %res2
}

define <2 x double>@test_int_x86_avx512_mask_vfnmadd_pd_128(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_vfnmadd_pd_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vmovaps %zmm0, %zmm3 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xd8]
; CHECK-NEXT:    vfnmadd213pd %xmm2, %xmm1, %xmm3 {%k1} ## encoding: [0x62,0xf2,0xf5,0x09,0xac,0xda]
; CHECK-NEXT:    vfnmadd213pd %xmm2, %xmm1, %xmm0 ## encoding: [0x62,0xf2,0xf5,0x08,0xac,0xc2]
; CHECK-NEXT:    vaddpd %xmm0, %xmm3, %xmm0 ## encoding: [0x62,0xf1,0xe5,0x08,0x58,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <2 x double> @llvm.x86.avx512.mask.vfnmadd.pd.128(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 %x3)
  %res1 = call <2 x double> @llvm.x86.avx512.mask.vfnmadd.pd.128(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 -1)
  %res2 = fadd <2 x double> %res, %res1
  ret <2 x double> %res2
}

define <4 x double>@test_int_x86_avx512_mask_vfnmadd_pd_256(<4 x double> %x0, <4 x double> %x1, <4 x double> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_vfnmadd_pd_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vmovaps %zmm0, %zmm3 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xd8]
; CHECK-NEXT:    vfnmadd213pd %ymm2, %ymm1, %ymm3 {%k1} ## encoding: [0x62,0xf2,0xf5,0x29,0xac,0xda]
; CHECK-NEXT:    vfnmadd213pd %ymm2, %ymm1, %ymm0 ## encoding: [0x62,0xf2,0xf5,0x28,0xac,0xc2]
; CHECK-NEXT:    vaddpd %ymm0, %ymm3, %ymm0 ## encoding: [0x62,0xf1,0xe5,0x28,0x58,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <4 x double> @llvm.x86.avx512.mask.vfnmadd.pd.256(<4 x double> %x0, <4 x double> %x1, <4 x double> %x2, i8 %x3)
  %res1 = call <4 x double> @llvm.x86.avx512.mask.vfnmadd.pd.256(<4 x double> %x0, <4 x double> %x1, <4 x double> %x2, i8 -1)
  %res2 = fadd <4 x double> %res, %res1
  ret <4 x double> %res2
}

define <4 x float>@test_int_x86_avx512_mask_vfnmadd_ps_128(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_vfnmadd_ps_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vmovaps %zmm0, %zmm3 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xd8]
; CHECK-NEXT:    vfnmadd213ps %xmm2, %xmm1, %xmm3 {%k1} ## encoding: [0x62,0xf2,0x75,0x09,0xac,0xda]
; CHECK-NEXT:    vfnmadd213ps %xmm2, %xmm1, %xmm0 ## encoding: [0x62,0xf2,0x75,0x08,0xac,0xc2]
; CHECK-NEXT:    vaddps %xmm0, %xmm3, %xmm0 ## encoding: [0x62,0xf1,0x64,0x08,0x58,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask.vfnmadd.ps.128(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 %x3)
  %res1 = call <4 x float> @llvm.x86.avx512.mask.vfnmadd.ps.128(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 -1)
  %res2 = fadd <4 x float> %res, %res1
  ret <4 x float> %res2
}

define <8 x float>@test_int_x86_avx512_mask_vfnmadd_ps_256(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_vfnmadd_ps_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vmovaps %zmm0, %zmm3 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xd8]
; CHECK-NEXT:    vfnmadd213ps %ymm2, %ymm1, %ymm3 {%k1} ## encoding: [0x62,0xf2,0x75,0x29,0xac,0xda]
; CHECK-NEXT:    vfnmadd213ps %ymm2, %ymm1, %ymm0 ## encoding: [0x62,0xf2,0x75,0x28,0xac,0xc2]
; CHECK-NEXT:    vaddps %ymm0, %ymm3, %ymm0 ## encoding: [0x62,0xf1,0x64,0x28,0x58,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask.vfnmadd.ps.256(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2, i8 %x3)
  %res1 = call <8 x float> @llvm.x86.avx512.mask.vfnmadd.ps.256(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2, i8 -1)
  %res2 = fadd <8 x float> %res, %res1
  ret <8 x float> %res2
}

declare <8 x float> @llvm.x86.avx512.mask.vfmaddsub.ps.256(<8 x float>, <8 x float>, <8 x float>, i8) nounwind readnone

define <8 x float> @test_mask_fmaddsub256_ps(<8 x float> %a, <8 x float> %b, <8 x float> %c, i8 %mask) {
; CHECK-LABEL: test_mask_fmaddsub256_ps:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vfmaddsub213ps %ymm2, %ymm1, %ymm0 {%k1} ## encoding: [0x62,0xf2,0x75,0x29,0xa6,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask.vfmaddsub.ps.256(<8 x float> %a, <8 x float> %b, <8 x float> %c, i8 %mask)
  ret <8 x float> %res
}

declare <4 x float> @llvm.x86.avx512.mask.vfmaddsub.ps.128(<4 x float>, <4 x float>, <4 x float>, i8) nounwind readnone

define <4 x float> @test_mask_fmaddsub128_ps(<4 x float> %a, <4 x float> %b, <4 x float> %c, i8 %mask) {
; CHECK-LABEL: test_mask_fmaddsub128_ps:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vfmaddsub213ps %xmm2, %xmm1, %xmm0 {%k1} ## encoding: [0x62,0xf2,0x75,0x09,0xa6,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask.vfmaddsub.ps.128(<4 x float> %a, <4 x float> %b, <4 x float> %c, i8 %mask)
  ret <4 x float> %res
}

declare <4 x double> @llvm.x86.avx512.mask.vfmaddsub.pd.256(<4 x double>, <4 x double>, <4 x double>, i8) nounwind readnone

define <4 x double> @test_mask_vfmaddsub256_pd(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2, i8 %mask) {
; CHECK-LABEL: test_mask_vfmaddsub256_pd:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vfmaddsub213pd %ymm2, %ymm1, %ymm0 {%k1} ## encoding: [0x62,0xf2,0xf5,0x29,0xa6,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <4 x double> @llvm.x86.avx512.mask.vfmaddsub.pd.256(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2, i8 %mask) nounwind
  ret <4 x double> %res
}

declare <2 x double> @llvm.x86.avx512.mask.vfmaddsub.pd.128(<2 x double>, <2 x double>, <2 x double>, i8) nounwind readnone

define <2 x double> @test_mask_vfmaddsub128_pd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2, i8 %mask) {
; CHECK-LABEL: test_mask_vfmaddsub128_pd:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vfmaddsub213pd %xmm2, %xmm1, %xmm0 {%k1} ## encoding: [0x62,0xf2,0xf5,0x09,0xa6,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <2 x double> @llvm.x86.avx512.mask.vfmaddsub.pd.128(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2, i8 %mask) nounwind
  ret <2 x double> %res
}

define <2 x double>@test_int_x86_avx512_mask_vfmaddsub_pd_128(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_vfmaddsub_pd_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vmovaps %zmm0, %zmm3 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xd8]
; CHECK-NEXT:    vfmaddsub213pd %xmm2, %xmm1, %xmm3 {%k1} ## encoding: [0x62,0xf2,0xf5,0x09,0xa6,0xda]
; CHECK-NEXT:    vfmaddsub213pd %xmm2, %xmm1, %xmm0 ## encoding: [0x62,0xf2,0xf5,0x08,0xa6,0xc2]
; CHECK-NEXT:    vaddpd %xmm0, %xmm3, %xmm0 ## encoding: [0x62,0xf1,0xe5,0x08,0x58,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <2 x double> @llvm.x86.avx512.mask.vfmaddsub.pd.128(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 %x3)
  %res1 = call <2 x double> @llvm.x86.avx512.mask.vfmaddsub.pd.128(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 -1)
  %res2 = fadd <2 x double> %res, %res1
  ret <2 x double> %res2
}

declare <2 x double> @llvm.x86.avx512.mask3.vfmaddsub.pd.128(<2 x double>, <2 x double>, <2 x double>, i8)

define <2 x double>@test_int_x86_avx512_mask3_vfmaddsub_pd_128(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask3_vfmaddsub_pd_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vmovaps %zmm2, %zmm3 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xda]
; CHECK-NEXT:    vfmaddsub231pd %xmm1, %xmm0, %xmm3 {%k1} ## encoding: [0x62,0xf2,0xfd,0x09,0xb6,0xd9]
; CHECK-NEXT:    vfmaddsub213pd %xmm2, %xmm1, %xmm0 ## encoding: [0x62,0xf2,0xf5,0x08,0xa6,0xc2]
; CHECK-NEXT:    vaddpd %xmm0, %xmm3, %xmm0 ## encoding: [0x62,0xf1,0xe5,0x08,0x58,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <2 x double> @llvm.x86.avx512.mask3.vfmaddsub.pd.128(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 %x3)
  %res1 = call <2 x double> @llvm.x86.avx512.mask3.vfmaddsub.pd.128(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 -1)
  %res2 = fadd <2 x double> %res, %res1
  ret <2 x double> %res2
}

declare <2 x double> @llvm.x86.avx512.maskz.vfmaddsub.pd.128(<2 x double>, <2 x double>, <2 x double>, i8)

define <2 x double>@test_int_x86_avx512_maskz_vfmaddsub_pd_128(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_maskz_vfmaddsub_pd_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vmovaps %zmm0, %zmm3 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xd8]
; CHECK-NEXT:    vfmaddsub213pd %xmm2, %xmm1, %xmm3 {%k1} {z} ## encoding: [0x62,0xf2,0xf5,0x89,0xa6,0xda]
; CHECK-NEXT:    vfmaddsub213pd %xmm2, %xmm1, %xmm0 ## encoding: [0x62,0xf2,0xf5,0x08,0xa6,0xc2]
; CHECK-NEXT:    vaddpd %xmm0, %xmm3, %xmm0 ## encoding: [0x62,0xf1,0xe5,0x08,0x58,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <2 x double> @llvm.x86.avx512.maskz.vfmaddsub.pd.128(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 %x3)
  %res1 = call <2 x double> @llvm.x86.avx512.maskz.vfmaddsub.pd.128(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 -1)
  %res2 = fadd <2 x double> %res, %res1
  ret <2 x double> %res2
}

define <4 x double>@test_int_x86_avx512_mask_vfmaddsub_pd_256(<4 x double> %x0, <4 x double> %x1, <4 x double> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_vfmaddsub_pd_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vmovaps %zmm0, %zmm3 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xd8]
; CHECK-NEXT:    vfmaddsub213pd %ymm2, %ymm1, %ymm3 {%k1} ## encoding: [0x62,0xf2,0xf5,0x29,0xa6,0xda]
; CHECK-NEXT:    vfmaddsub213pd %ymm2, %ymm1, %ymm0 ## encoding: [0x62,0xf2,0xf5,0x28,0xa6,0xc2]
; CHECK-NEXT:    vaddpd %ymm0, %ymm3, %ymm0 ## encoding: [0x62,0xf1,0xe5,0x28,0x58,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <4 x double> @llvm.x86.avx512.mask.vfmaddsub.pd.256(<4 x double> %x0, <4 x double> %x1, <4 x double> %x2, i8 %x3)
  %res1 = call <4 x double> @llvm.x86.avx512.mask.vfmaddsub.pd.256(<4 x double> %x0, <4 x double> %x1, <4 x double> %x2, i8 -1)
  %res2 = fadd <4 x double> %res, %res1
  ret <4 x double> %res2
}

declare <4 x double> @llvm.x86.avx512.mask3.vfmaddsub.pd.256(<4 x double>, <4 x double>, <4 x double>, i8)

define <4 x double>@test_int_x86_avx512_mask3_vfmaddsub_pd_256(<4 x double> %x0, <4 x double> %x1, <4 x double> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask3_vfmaddsub_pd_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vmovaps %zmm2, %zmm3 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xda]
; CHECK-NEXT:    vfmaddsub231pd %ymm1, %ymm0, %ymm3 {%k1} ## encoding: [0x62,0xf2,0xfd,0x29,0xb6,0xd9]
; CHECK-NEXT:    vfmaddsub213pd %ymm2, %ymm1, %ymm0 ## encoding: [0x62,0xf2,0xf5,0x28,0xa6,0xc2]
; CHECK-NEXT:    vaddpd %ymm0, %ymm3, %ymm0 ## encoding: [0x62,0xf1,0xe5,0x28,0x58,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <4 x double> @llvm.x86.avx512.mask3.vfmaddsub.pd.256(<4 x double> %x0, <4 x double> %x1, <4 x double> %x2, i8 %x3)
  %res1 = call <4 x double> @llvm.x86.avx512.mask3.vfmaddsub.pd.256(<4 x double> %x0, <4 x double> %x1, <4 x double> %x2, i8 -1)
  %res2 = fadd <4 x double> %res, %res1
  ret <4 x double> %res2
}

declare <4 x double> @llvm.x86.avx512.maskz.vfmaddsub.pd.256(<4 x double>, <4 x double>, <4 x double>, i8)

define <4 x double>@test_int_x86_avx512_maskz_vfmaddsub_pd_256(<4 x double> %x0, <4 x double> %x1, <4 x double> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_maskz_vfmaddsub_pd_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vmovaps %zmm0, %zmm3 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xd8]
; CHECK-NEXT:    vfmaddsub213pd %ymm2, %ymm1, %ymm3 {%k1} {z} ## encoding: [0x62,0xf2,0xf5,0xa9,0xa6,0xda]
; CHECK-NEXT:    vfmaddsub213pd %ymm2, %ymm1, %ymm0 ## encoding: [0x62,0xf2,0xf5,0x28,0xa6,0xc2]
; CHECK-NEXT:    vaddpd %ymm0, %ymm3, %ymm0 ## encoding: [0x62,0xf1,0xe5,0x28,0x58,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <4 x double> @llvm.x86.avx512.maskz.vfmaddsub.pd.256(<4 x double> %x0, <4 x double> %x1, <4 x double> %x2, i8 %x3)
  %res1 = call <4 x double> @llvm.x86.avx512.maskz.vfmaddsub.pd.256(<4 x double> %x0, <4 x double> %x1, <4 x double> %x2, i8 -1)
  %res2 = fadd <4 x double> %res, %res1
  ret <4 x double> %res2
}

define <4 x float>@test_int_x86_avx512_mask_vfmaddsub_ps_128(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_vfmaddsub_ps_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vmovaps %zmm0, %zmm3 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xd8]
; CHECK-NEXT:    vfmaddsub213ps %xmm2, %xmm1, %xmm3 {%k1} ## encoding: [0x62,0xf2,0x75,0x09,0xa6,0xda]
; CHECK-NEXT:    vfmaddsub213ps %xmm2, %xmm1, %xmm0 ## encoding: [0x62,0xf2,0x75,0x08,0xa6,0xc2]
; CHECK-NEXT:    vaddps %xmm0, %xmm3, %xmm0 ## encoding: [0x62,0xf1,0x64,0x08,0x58,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask.vfmaddsub.ps.128(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 %x3)
  %res1 = call <4 x float> @llvm.x86.avx512.mask.vfmaddsub.ps.128(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 -1)
  %res2 = fadd <4 x float> %res, %res1
  ret <4 x float> %res2
}

declare <4 x float> @llvm.x86.avx512.mask3.vfmaddsub.ps.128(<4 x float>, <4 x float>, <4 x float>, i8)

define <4 x float>@test_int_x86_avx512_mask3_vfmaddsub_ps_128(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask3_vfmaddsub_ps_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vmovaps %zmm2, %zmm3 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xda]
; CHECK-NEXT:    vfmaddsub231ps %xmm1, %xmm0, %xmm3 {%k1} ## encoding: [0x62,0xf2,0x7d,0x09,0xb6,0xd9]
; CHECK-NEXT:    vfmaddsub213ps %xmm2, %xmm1, %xmm0 ## encoding: [0x62,0xf2,0x75,0x08,0xa6,0xc2]
; CHECK-NEXT:    vaddps %xmm0, %xmm3, %xmm0 ## encoding: [0x62,0xf1,0x64,0x08,0x58,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask3.vfmaddsub.ps.128(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 %x3)
  %res1 = call <4 x float> @llvm.x86.avx512.mask3.vfmaddsub.ps.128(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 -1)
  %res2 = fadd <4 x float> %res, %res1
  ret <4 x float> %res2
}

declare <4 x float> @llvm.x86.avx512.maskz.vfmaddsub.ps.128(<4 x float>, <4 x float>, <4 x float>, i8)

define <4 x float>@test_int_x86_avx512_maskz_vfmaddsub_ps_128(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_maskz_vfmaddsub_ps_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vmovaps %zmm0, %zmm3 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xd8]
; CHECK-NEXT:    vfmaddsub213ps %xmm2, %xmm1, %xmm3 {%k1} {z} ## encoding: [0x62,0xf2,0x75,0x89,0xa6,0xda]
; CHECK-NEXT:    vfmaddsub213ps %xmm2, %xmm1, %xmm0 ## encoding: [0x62,0xf2,0x75,0x08,0xa6,0xc2]
; CHECK-NEXT:    vaddps %xmm0, %xmm3, %xmm0 ## encoding: [0x62,0xf1,0x64,0x08,0x58,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.maskz.vfmaddsub.ps.128(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 %x3)
  %res1 = call <4 x float> @llvm.x86.avx512.maskz.vfmaddsub.ps.128(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 -1)
  %res2 = fadd <4 x float> %res, %res1
  ret <4 x float> %res2
}

define <8 x float>@test_int_x86_avx512_mask_vfmaddsub_ps_256(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_vfmaddsub_ps_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vmovaps %zmm0, %zmm3 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xd8]
; CHECK-NEXT:    vfmaddsub213ps %ymm2, %ymm1, %ymm3 {%k1} ## encoding: [0x62,0xf2,0x75,0x29,0xa6,0xda]
; CHECK-NEXT:    vfmaddsub213ps %ymm2, %ymm1, %ymm0 ## encoding: [0x62,0xf2,0x75,0x28,0xa6,0xc2]
; CHECK-NEXT:    vaddps %ymm0, %ymm3, %ymm0 ## encoding: [0x62,0xf1,0x64,0x28,0x58,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask.vfmaddsub.ps.256(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2, i8 %x3)
  %res1 = call <8 x float> @llvm.x86.avx512.mask.vfmaddsub.ps.256(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2, i8 -1)
  %res2 = fadd <8 x float> %res, %res1
  ret <8 x float> %res2
}

declare <8 x float> @llvm.x86.avx512.mask3.vfmaddsub.ps.256(<8 x float>, <8 x float>, <8 x float>, i8)

define <8 x float>@test_int_x86_avx512_mask3_vfmaddsub_ps_256(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask3_vfmaddsub_ps_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vmovaps %zmm2, %zmm3 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xda]
; CHECK-NEXT:    vfmaddsub231ps %ymm1, %ymm0, %ymm3 {%k1} ## encoding: [0x62,0xf2,0x7d,0x29,0xb6,0xd9]
; CHECK-NEXT:    vfmaddsub213ps %ymm2, %ymm1, %ymm0 ## encoding: [0x62,0xf2,0x75,0x28,0xa6,0xc2]
; CHECK-NEXT:    vaddps %ymm0, %ymm3, %ymm0 ## encoding: [0x62,0xf1,0x64,0x28,0x58,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask3.vfmaddsub.ps.256(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2, i8 %x3)
  %res1 = call <8 x float> @llvm.x86.avx512.mask3.vfmaddsub.ps.256(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2, i8 -1)
  %res2 = fadd <8 x float> %res, %res1
  ret <8 x float> %res2
}

declare <8 x float> @llvm.x86.avx512.maskz.vfmaddsub.ps.256(<8 x float>, <8 x float>, <8 x float>, i8)

define <8 x float>@test_int_x86_avx512_maskz_vfmaddsub_ps_256(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_maskz_vfmaddsub_ps_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vmovaps %zmm0, %zmm3 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xd8]
; CHECK-NEXT:    vfmaddsub213ps %ymm2, %ymm1, %ymm3 {%k1} {z} ## encoding: [0x62,0xf2,0x75,0xa9,0xa6,0xda]
; CHECK-NEXT:    vfmaddsub213ps %ymm2, %ymm1, %ymm0 ## encoding: [0x62,0xf2,0x75,0x28,0xa6,0xc2]
; CHECK-NEXT:    vaddps %ymm0, %ymm3, %ymm0 ## encoding: [0x62,0xf1,0x64,0x28,0x58,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.maskz.vfmaddsub.ps.256(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2, i8 %x3)
  %res1 = call <8 x float> @llvm.x86.avx512.maskz.vfmaddsub.ps.256(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2, i8 -1)
  %res2 = fadd <8 x float> %res, %res1
  ret <8 x float> %res2
}

declare <2 x double> @llvm.x86.avx512.mask3.vfmsubadd.pd.128(<2 x double>, <2 x double>, <2 x double>, i8)

define <2 x double>@test_int_x86_avx512_mask3_vfmsubadd_pd_128(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask3_vfmsubadd_pd_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vmovaps %zmm2, %zmm3 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xda]
; CHECK-NEXT:    vfmsubadd231pd %xmm1, %xmm0, %xmm3 {%k1} ## encoding: [0x62,0xf2,0xfd,0x09,0xb7,0xd9]
; CHECK-NEXT:    vfmsubadd213pd %xmm2, %xmm1, %xmm0 ## encoding: [0x62,0xf2,0xf5,0x08,0xa7,0xc2]
; CHECK-NEXT:    vaddpd %xmm0, %xmm3, %xmm0 ## encoding: [0x62,0xf1,0xe5,0x08,0x58,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <2 x double> @llvm.x86.avx512.mask3.vfmsubadd.pd.128(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 %x3)
  %res1 = call <2 x double> @llvm.x86.avx512.mask3.vfmsubadd.pd.128(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 -1)
  %res2=fadd <2 x double> %res, %res1
  ret <2 x double> %res2
}

declare <4 x double> @llvm.x86.avx512.mask3.vfmsubadd.pd.256(<4 x double>, <4 x double>, <4 x double>, i8)

define <4 x double>@test_int_x86_avx512_mask3_vfmsubadd_pd_256(<4 x double> %x0, <4 x double> %x1, <4 x double> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask3_vfmsubadd_pd_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vmovaps %zmm2, %zmm3 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xda]
; CHECK-NEXT:    vfmsubadd231pd %ymm1, %ymm0, %ymm3 {%k1} ## encoding: [0x62,0xf2,0xfd,0x29,0xb7,0xd9]
; CHECK-NEXT:    vfmsubadd213pd %ymm2, %ymm1, %ymm0 ## encoding: [0x62,0xf2,0xf5,0x28,0xa7,0xc2]
; CHECK-NEXT:    vaddpd %ymm0, %ymm3, %ymm0 ## encoding: [0x62,0xf1,0xe5,0x28,0x58,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <4 x double> @llvm.x86.avx512.mask3.vfmsubadd.pd.256(<4 x double> %x0, <4 x double> %x1, <4 x double> %x2, i8 %x3)
  %res1 = call <4 x double> @llvm.x86.avx512.mask3.vfmsubadd.pd.256(<4 x double> %x0, <4 x double> %x1, <4 x double> %x2, i8 -1)
  %res2=fadd <4 x double> %res, %res1
  ret <4 x double> %res2
}

declare <4 x float> @llvm.x86.avx512.mask3.vfmsubadd.ps.128(<4 x float>, <4 x float>, <4 x float>, i8)

define <4 x float>@test_int_x86_avx512_mask3_vfmsubadd_ps_128(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask3_vfmsubadd_ps_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vmovaps %zmm2, %zmm3 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xda]
; CHECK-NEXT:    vfmsubadd231ps %xmm1, %xmm0, %xmm3 {%k1} ## encoding: [0x62,0xf2,0x7d,0x09,0xb7,0xd9]
; CHECK-NEXT:    vfmsubadd213ps %xmm2, %xmm1, %xmm0 ## encoding: [0x62,0xf2,0x75,0x08,0xa7,0xc2]
; CHECK-NEXT:    vaddps %xmm0, %xmm3, %xmm0 ## encoding: [0x62,0xf1,0x64,0x08,0x58,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask3.vfmsubadd.ps.128(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 %x3)
  %res1 = call <4 x float> @llvm.x86.avx512.mask3.vfmsubadd.ps.128(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 -1)
  %res2=fadd <4 x float> %res, %res1
  ret <4 x float> %res2
}

declare <8 x float> @llvm.x86.avx512.mask3.vfmsubadd.ps.256(<8 x float>, <8 x float>, <8 x float>, i8)

define <8 x float>@test_int_x86_avx512_mask3_vfmsubadd_ps_256(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask3_vfmsubadd_ps_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vmovaps %zmm2, %zmm3 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xda]
; CHECK-NEXT:    vfmsubadd231ps %ymm1, %ymm0, %ymm3 {%k1} ## encoding: [0x62,0xf2,0x7d,0x29,0xb7,0xd9]
; CHECK-NEXT:    vfmsubadd213ps %ymm2, %ymm1, %ymm0 ## encoding: [0x62,0xf2,0x75,0x28,0xa7,0xc2]
; CHECK-NEXT:    vaddps %ymm0, %ymm3, %ymm0 ## encoding: [0x62,0xf1,0x64,0x28,0x58,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask3.vfmsubadd.ps.256(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2, i8 %x3)
  %res1 = call <8 x float> @llvm.x86.avx512.mask3.vfmsubadd.ps.256(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2, i8 -1)
  %res2=fadd <8 x float> %res, %res1
  ret <8 x float> %res2
}


define <4 x float> @test_mask_vfmadd128_ps_r(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2, i8 %mask) {
; CHECK-LABEL: test_mask_vfmadd128_ps_r:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vfmadd213ps %xmm2, %xmm1, %xmm0 {%k1} ## encoding: [0x62,0xf2,0x75,0x09,0xa8,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask.vfmadd.ps.128(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2, i8 %mask) nounwind
  ret <4 x float> %res
}

define <4 x float> @test_mask_vfmadd128_ps_rz(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2) {
; CHECK-LABEL: test_mask_vfmadd128_ps_rz:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vfmadd213ps %xmm2, %xmm1, %xmm0 ## encoding: [0x62,0xf2,0x75,0x08,0xa8,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask.vfmadd.ps.128(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2, i8 -1) nounwind
  ret <4 x float> %res
}

define <4 x float> @test_mask_vfmadd128_ps_rmk(<4 x float> %a0, <4 x float> %a1, <4 x float>* %ptr_a2, i8 %mask) {
; CHECK-LABEL: test_mask_vfmadd128_ps_rmk:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vfmadd213ps (%rdi), %xmm1, %xmm0 {%k1} ## encoding: [0x62,0xf2,0x75,0x09,0xa8,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %a2 = load <4 x float>, <4 x float>* %ptr_a2
  %res = call <4 x float> @llvm.x86.avx512.mask.vfmadd.ps.128(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2, i8 %mask) nounwind
  ret <4 x float> %res
}

define <4 x float> @test_mask_vfmadd128_ps_rmka(<4 x float> %a0, <4 x float> %a1, <4 x float>* %ptr_a2, i8 %mask) {
; CHECK-LABEL: test_mask_vfmadd128_ps_rmka:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vfmadd213ps (%rdi), %xmm1, %xmm0 {%k1} ## encoding: [0x62,0xf2,0x75,0x09,0xa8,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %a2 = load <4 x float>, <4 x float>* %ptr_a2, align 8
  %res = call <4 x float> @llvm.x86.avx512.mask.vfmadd.ps.128(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2, i8 %mask) nounwind
  ret <4 x float> %res
}

define <4 x float> @test_mask_vfmadd128_ps_rmkz(<4 x float> %a0, <4 x float> %a1, <4 x float>* %ptr_a2) {
; CHECK-LABEL: test_mask_vfmadd128_ps_rmkz:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vfmadd213ps (%rdi), %xmm1, %xmm0 ## encoding: [0x62,0xf2,0x75,0x08,0xa8,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %a2 = load <4 x float>, <4 x float>* %ptr_a2
  %res = call <4 x float> @llvm.x86.avx512.mask.vfmadd.ps.128(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2, i8 -1) nounwind
  ret <4 x float> %res
}

define <4 x float> @test_mask_vfmadd128_ps_rmkza(<4 x float> %a0, <4 x float> %a1, <4 x float>* %ptr_a2) {
; CHECK-LABEL: test_mask_vfmadd128_ps_rmkza:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vfmadd213ps (%rdi), %xmm1, %xmm0 ## encoding: [0x62,0xf2,0x75,0x08,0xa8,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %a2 = load <4 x float>, <4 x float>* %ptr_a2, align 4
  %res = call <4 x float> @llvm.x86.avx512.mask.vfmadd.ps.128(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2, i8 -1) nounwind
  ret <4 x float> %res
}

define <4 x float> @test_mask_vfmadd128_ps_rmb(<4 x float> %a0, <4 x float> %a1, float* %ptr_a2, i8 %mask) {
; CHECK-LABEL: test_mask_vfmadd128_ps_rmb:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vfmadd213ps (%rdi){1to4}, %xmm1, %xmm0 {%k1} ## encoding: [0x62,0xf2,0x75,0x19,0xa8,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %q = load float, float* %ptr_a2
  %vecinit.i = insertelement <4 x float> undef, float %q, i32 0
  %vecinit2.i = insertelement <4 x float> %vecinit.i, float %q, i32 1
  %vecinit4.i = insertelement <4 x float> %vecinit2.i, float %q, i32 2
  %vecinit6.i = insertelement <4 x float> %vecinit4.i, float %q, i32 3
  %res = call <4 x float> @llvm.x86.avx512.mask.vfmadd.ps.128(<4 x float> %a0, <4 x float> %a1, <4 x float> %vecinit6.i, i8 %mask) nounwind
  ret <4 x float> %res
}

define <4 x float> @test_mask_vfmadd128_ps_rmba(<4 x float> %a0, <4 x float> %a1, float* %ptr_a2, i8 %mask) {
; CHECK-LABEL: test_mask_vfmadd128_ps_rmba:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vfmadd213ps (%rdi){1to4}, %xmm1, %xmm0 {%k1} ## encoding: [0x62,0xf2,0x75,0x19,0xa8,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %q = load float, float* %ptr_a2, align 4
  %vecinit.i = insertelement <4 x float> undef, float %q, i32 0
  %vecinit2.i = insertelement <4 x float> %vecinit.i, float %q, i32 1
  %vecinit4.i = insertelement <4 x float> %vecinit2.i, float %q, i32 2
  %vecinit6.i = insertelement <4 x float> %vecinit4.i, float %q, i32 3
  %res = call <4 x float> @llvm.x86.avx512.mask.vfmadd.ps.128(<4 x float> %a0, <4 x float> %a1, <4 x float> %vecinit6.i, i8 %mask) nounwind
  ret <4 x float> %res
}

define <4 x float> @test_mask_vfmadd128_ps_rmbz(<4 x float> %a0, <4 x float> %a1, float* %ptr_a2) {
; CHECK-LABEL: test_mask_vfmadd128_ps_rmbz:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vfmadd213ps (%rdi){1to4}, %xmm1, %xmm0 ## encoding: [0x62,0xf2,0x75,0x18,0xa8,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %q = load float, float* %ptr_a2
  %vecinit.i = insertelement <4 x float> undef, float %q, i32 0
  %vecinit2.i = insertelement <4 x float> %vecinit.i, float %q, i32 1
  %vecinit4.i = insertelement <4 x float> %vecinit2.i, float %q, i32 2
  %vecinit6.i = insertelement <4 x float> %vecinit4.i, float %q, i32 3
  %res = call <4 x float> @llvm.x86.avx512.mask.vfmadd.ps.128(<4 x float> %a0, <4 x float> %a1, <4 x float> %vecinit6.i, i8 -1) nounwind
  ret <4 x float> %res
}

define <4 x float> @test_mask_vfmadd128_ps_rmbza(<4 x float> %a0, <4 x float> %a1, float* %ptr_a2) {
; CHECK-LABEL: test_mask_vfmadd128_ps_rmbza:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vfmadd213ps (%rdi){1to4}, %xmm1, %xmm0 ## encoding: [0x62,0xf2,0x75,0x18,0xa8,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %q = load float, float* %ptr_a2, align 4
  %vecinit.i = insertelement <4 x float> undef, float %q, i32 0
  %vecinit2.i = insertelement <4 x float> %vecinit.i, float %q, i32 1
  %vecinit4.i = insertelement <4 x float> %vecinit2.i, float %q, i32 2
  %vecinit6.i = insertelement <4 x float> %vecinit4.i, float %q, i32 3
  %res = call <4 x float> @llvm.x86.avx512.mask.vfmadd.ps.128(<4 x float> %a0, <4 x float> %a1, <4 x float> %vecinit6.i, i8 -1) nounwind
  ret <4 x float> %res
}

define <2 x double> @test_mask_vfmadd128_pd_r(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2, i8 %mask) {
; CHECK-LABEL: test_mask_vfmadd128_pd_r:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vfmadd213pd %xmm2, %xmm1, %xmm0 {%k1} ## encoding: [0x62,0xf2,0xf5,0x09,0xa8,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <2 x double> @llvm.x86.avx512.mask.vfmadd.pd.128(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2, i8 %mask) nounwind
  ret <2 x double> %res
}

define <2 x double> @test_mask_vfmadd128_pd_rz(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2) {
; CHECK-LABEL: test_mask_vfmadd128_pd_rz:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vfmadd213pd %xmm2, %xmm1, %xmm0 ## encoding: [0x62,0xf2,0xf5,0x08,0xa8,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <2 x double> @llvm.x86.avx512.mask.vfmadd.pd.128(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2, i8 -1) nounwind
  ret <2 x double> %res
}

define <2 x double> @test_mask_vfmadd128_pd_rmk(<2 x double> %a0, <2 x double> %a1, <2 x double>* %ptr_a2, i8 %mask) {
; CHECK-LABEL: test_mask_vfmadd128_pd_rmk:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vfmadd213pd (%rdi), %xmm1, %xmm0 {%k1} ## encoding: [0x62,0xf2,0xf5,0x09,0xa8,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %a2 = load <2 x double>, <2 x double>* %ptr_a2
  %res = call <2 x double> @llvm.x86.avx512.mask.vfmadd.pd.128(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2, i8 %mask) nounwind
  ret <2 x double> %res
}

define <2 x double> @test_mask_vfmadd128_pd_rmkz(<2 x double> %a0, <2 x double> %a1, <2 x double>* %ptr_a2) {
; CHECK-LABEL: test_mask_vfmadd128_pd_rmkz:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vfmadd213pd (%rdi), %xmm1, %xmm0 ## encoding: [0x62,0xf2,0xf5,0x08,0xa8,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %a2 = load <2 x double>, <2 x double>* %ptr_a2
  %res = call <2 x double> @llvm.x86.avx512.mask.vfmadd.pd.128(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2, i8 -1) nounwind
  ret <2 x double> %res
}

define <4 x double> @test_mask_vfmadd256_pd_r(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2, i8 %mask) {
; CHECK-LABEL: test_mask_vfmadd256_pd_r:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vfmadd213pd %ymm2, %ymm1, %ymm0 {%k1} ## encoding: [0x62,0xf2,0xf5,0x29,0xa8,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <4 x double> @llvm.x86.avx512.mask.vfmadd.pd.256(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2, i8 %mask) nounwind
  ret <4 x double> %res
}

define <4 x double> @test_mask_vfmadd256_pd_rz(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2) {
; CHECK-LABEL: test_mask_vfmadd256_pd_rz:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vfmadd213pd %ymm2, %ymm1, %ymm0 ## encoding: [0x62,0xf2,0xf5,0x28,0xa8,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <4 x double> @llvm.x86.avx512.mask.vfmadd.pd.256(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2, i8 -1) nounwind
  ret <4 x double> %res
}

define <4 x double> @test_mask_vfmadd256_pd_rmk(<4 x double> %a0, <4 x double> %a1, <4 x double>* %ptr_a2, i8 %mask) {
; CHECK-LABEL: test_mask_vfmadd256_pd_rmk:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vfmadd213pd (%rdi), %ymm1, %ymm0 {%k1} ## encoding: [0x62,0xf2,0xf5,0x29,0xa8,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %a2 = load <4 x double>, <4 x double>* %ptr_a2
  %res = call <4 x double> @llvm.x86.avx512.mask.vfmadd.pd.256(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2, i8 %mask) nounwind
  ret <4 x double> %res
}

define <4 x double> @test_mask_vfmadd256_pd_rmkz(<4 x double> %a0, <4 x double> %a1, <4 x double>* %ptr_a2) {
; CHECK-LABEL: test_mask_vfmadd256_pd_rmkz:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vfmadd213pd (%rdi), %ymm1, %ymm0 ## encoding: [0x62,0xf2,0xf5,0x28,0xa8,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %a2 = load <4 x double>, <4 x double>* %ptr_a2
  %res = call <4 x double> @llvm.x86.avx512.mask.vfmadd.pd.256(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2, i8 -1) nounwind
  ret <4 x double> %res
}
define <8 x i16> @test_mask_add_epi16_rr_128(<8 x i16> %a, <8 x i16> %b) {
; CHECK-LABEL: test_mask_add_epi16_rr_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpaddw %xmm1, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xfd,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.mask.padd.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> zeroinitializer, i8 -1)
  ret <8 x i16> %res
}

define <8 x i16> @test_mask_add_epi16_rrk_128(<8 x i16> %a, <8 x i16> %b, <8 x i16> %passThru, i8 %mask) {
; CHECK-LABEL: test_mask_add_epi16_rrk_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpaddw %xmm1, %xmm0, %xmm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0xfd,0xd1]
; CHECK-NEXT:    vmovaps %zmm2, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.mask.padd.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> %passThru, i8 %mask)
  ret <8 x i16> %res
}

define <8 x i16> @test_mask_add_epi16_rrkz_128(<8 x i16> %a, <8 x i16> %b, i8 %mask) {
; CHECK-LABEL: test_mask_add_epi16_rrkz_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpaddw %xmm1, %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0x89,0xfd,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.mask.padd.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> zeroinitializer, i8 %mask)
  ret <8 x i16> %res
}

define <8 x i16> @test_mask_add_epi16_rm_128(<8 x i16> %a, <8 x i16>* %ptr_b) {
; CHECK-LABEL: test_mask_add_epi16_rm_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpaddw (%rdi), %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xfd,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <8 x i16>, <8 x i16>* %ptr_b
  %res = call <8 x i16> @llvm.x86.avx512.mask.padd.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> zeroinitializer, i8 -1)
  ret <8 x i16> %res
}

define <8 x i16> @test_mask_add_epi16_rmk_128(<8 x i16> %a, <8 x i16>* %ptr_b, <8 x i16> %passThru, i8 %mask) {
; CHECK-LABEL: test_mask_add_epi16_rmk_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpaddw (%rdi), %xmm0, %xmm1 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0xfd,0x0f]
; CHECK-NEXT:    vmovaps %zmm1, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <8 x i16>, <8 x i16>* %ptr_b
  %res = call <8 x i16> @llvm.x86.avx512.mask.padd.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> %passThru, i8 %mask)
  ret <8 x i16> %res
}

define <8 x i16> @test_mask_add_epi16_rmkz_128(<8 x i16> %a, <8 x i16>* %ptr_b, i8 %mask) {
; CHECK-LABEL: test_mask_add_epi16_rmkz_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpaddw (%rdi), %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0x89,0xfd,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <8 x i16>, <8 x i16>* %ptr_b
  %res = call <8 x i16> @llvm.x86.avx512.mask.padd.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> zeroinitializer, i8 %mask)
  ret <8 x i16> %res
}

declare <8 x i16> @llvm.x86.avx512.mask.padd.w.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)

define <16 x i16> @test_mask_add_epi16_rr_256(<16 x i16> %a, <16 x i16> %b) {
; CHECK-LABEL: test_mask_add_epi16_rr_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpaddw %ymm1, %ymm0, %ymm0 ## encoding: [0x62,0xf1,0x7d,0x28,0xfd,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.mask.padd.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> zeroinitializer, i16 -1)
  ret <16 x i16> %res
}

define <16 x i16> @test_mask_add_epi16_rrk_256(<16 x i16> %a, <16 x i16> %b, <16 x i16> %passThru, i16 %mask) {
; CHECK-LABEL: test_mask_add_epi16_rrk_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpaddw %ymm1, %ymm0, %ymm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0xfd,0xd1]
; CHECK-NEXT:    vmovaps %zmm2, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.mask.padd.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> %passThru, i16 %mask)
  ret <16 x i16> %res
}

define <16 x i16> @test_mask_add_epi16_rrkz_256(<16 x i16> %a, <16 x i16> %b, i16 %mask) {
; CHECK-LABEL: test_mask_add_epi16_rrkz_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpaddw %ymm1, %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xa9,0xfd,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.mask.padd.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> zeroinitializer, i16 %mask)
  ret <16 x i16> %res
}

define <16 x i16> @test_mask_add_epi16_rm_256(<16 x i16> %a, <16 x i16>* %ptr_b) {
; CHECK-LABEL: test_mask_add_epi16_rm_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpaddw (%rdi), %ymm0, %ymm0 ## encoding: [0x62,0xf1,0x7d,0x28,0xfd,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <16 x i16>, <16 x i16>* %ptr_b
  %res = call <16 x i16> @llvm.x86.avx512.mask.padd.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> zeroinitializer, i16 -1)
  ret <16 x i16> %res
}

define <16 x i16> @test_mask_add_epi16_rmk_256(<16 x i16> %a, <16 x i16>* %ptr_b, <16 x i16> %passThru, i16 %mask) {
; CHECK-LABEL: test_mask_add_epi16_rmk_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpaddw (%rdi), %ymm0, %ymm1 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0xfd,0x0f]
; CHECK-NEXT:    vmovaps %zmm1, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <16 x i16>, <16 x i16>* %ptr_b
  %res = call <16 x i16> @llvm.x86.avx512.mask.padd.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> %passThru, i16 %mask)
  ret <16 x i16> %res
}

define <16 x i16> @test_mask_add_epi16_rmkz_256(<16 x i16> %a, <16 x i16>* %ptr_b, i16 %mask) {
; CHECK-LABEL: test_mask_add_epi16_rmkz_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpaddw (%rdi), %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xa9,0xfd,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <16 x i16>, <16 x i16>* %ptr_b
  %res = call <16 x i16> @llvm.x86.avx512.mask.padd.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> zeroinitializer, i16 %mask)
  ret <16 x i16> %res
}

declare <16 x i16> @llvm.x86.avx512.mask.padd.w.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)

define <8 x i16> @test_mask_sub_epi16_rr_128(<8 x i16> %a, <8 x i16> %b) {
; CHECK-LABEL: test_mask_sub_epi16_rr_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpsubw %xmm1, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xf9,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.mask.psub.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> zeroinitializer, i8 -1)
  ret <8 x i16> %res
}

define <8 x i16> @test_mask_sub_epi16_rrk_128(<8 x i16> %a, <8 x i16> %b, <8 x i16> %passThru, i8 %mask) {
; CHECK-LABEL: test_mask_sub_epi16_rrk_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpsubw %xmm1, %xmm0, %xmm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0xf9,0xd1]
; CHECK-NEXT:    vmovaps %zmm2, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.mask.psub.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> %passThru, i8 %mask)
  ret <8 x i16> %res
}

define <8 x i16> @test_mask_sub_epi16_rrkz_128(<8 x i16> %a, <8 x i16> %b, i8 %mask) {
; CHECK-LABEL: test_mask_sub_epi16_rrkz_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpsubw %xmm1, %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0x89,0xf9,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.mask.psub.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> zeroinitializer, i8 %mask)
  ret <8 x i16> %res
}

define <8 x i16> @test_mask_sub_epi16_rm_128(<8 x i16> %a, <8 x i16>* %ptr_b) {
; CHECK-LABEL: test_mask_sub_epi16_rm_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpsubw (%rdi), %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xf9,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <8 x i16>, <8 x i16>* %ptr_b
  %res = call <8 x i16> @llvm.x86.avx512.mask.psub.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> zeroinitializer, i8 -1)
  ret <8 x i16> %res
}

define <8 x i16> @test_mask_sub_epi16_rmk_128(<8 x i16> %a, <8 x i16>* %ptr_b, <8 x i16> %passThru, i8 %mask) {
; CHECK-LABEL: test_mask_sub_epi16_rmk_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpsubw (%rdi), %xmm0, %xmm1 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0xf9,0x0f]
; CHECK-NEXT:    vmovaps %zmm1, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <8 x i16>, <8 x i16>* %ptr_b
  %res = call <8 x i16> @llvm.x86.avx512.mask.psub.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> %passThru, i8 %mask)
  ret <8 x i16> %res
}

define <8 x i16> @test_mask_sub_epi16_rmkz_128(<8 x i16> %a, <8 x i16>* %ptr_b, i8 %mask) {
; CHECK-LABEL: test_mask_sub_epi16_rmkz_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpsubw (%rdi), %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0x89,0xf9,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <8 x i16>, <8 x i16>* %ptr_b
  %res = call <8 x i16> @llvm.x86.avx512.mask.psub.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> zeroinitializer, i8 %mask)
  ret <8 x i16> %res
}

declare <8 x i16> @llvm.x86.avx512.mask.psub.w.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)

define <16 x i16> @test_mask_sub_epi16_rr_256(<16 x i16> %a, <16 x i16> %b) {
; CHECK-LABEL: test_mask_sub_epi16_rr_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpsubw %ymm1, %ymm0, %ymm0 ## encoding: [0x62,0xf1,0x7d,0x28,0xf9,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.mask.psub.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> zeroinitializer, i16 -1)
  ret <16 x i16> %res
}

define <16 x i16> @test_mask_sub_epi16_rrk_256(<16 x i16> %a, <16 x i16> %b, <16 x i16> %passThru, i16 %mask) {
; CHECK-LABEL: test_mask_sub_epi16_rrk_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpsubw %ymm1, %ymm0, %ymm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0xf9,0xd1]
; CHECK-NEXT:    vmovaps %zmm2, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.mask.psub.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> %passThru, i16 %mask)
  ret <16 x i16> %res
}

define <16 x i16> @test_mask_sub_epi16_rrkz_256(<16 x i16> %a, <16 x i16> %b, i16 %mask) {
; CHECK-LABEL: test_mask_sub_epi16_rrkz_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpsubw %ymm1, %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xa9,0xf9,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.mask.psub.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> zeroinitializer, i16 %mask)
  ret <16 x i16> %res
}

define <16 x i16> @test_mask_sub_epi16_rm_256(<16 x i16> %a, <16 x i16>* %ptr_b) {
; CHECK-LABEL: test_mask_sub_epi16_rm_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpsubw (%rdi), %ymm0, %ymm0 ## encoding: [0x62,0xf1,0x7d,0x28,0xf9,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <16 x i16>, <16 x i16>* %ptr_b
  %res = call <16 x i16> @llvm.x86.avx512.mask.psub.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> zeroinitializer, i16 -1)
  ret <16 x i16> %res
}

define <16 x i16> @test_mask_sub_epi16_rmk_256(<16 x i16> %a, <16 x i16>* %ptr_b, <16 x i16> %passThru, i16 %mask) {
; CHECK-LABEL: test_mask_sub_epi16_rmk_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpsubw (%rdi), %ymm0, %ymm1 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0xf9,0x0f]
; CHECK-NEXT:    vmovaps %zmm1, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <16 x i16>, <16 x i16>* %ptr_b
  %res = call <16 x i16> @llvm.x86.avx512.mask.psub.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> %passThru, i16 %mask)
  ret <16 x i16> %res
}

define <16 x i16> @test_mask_sub_epi16_rmkz_256(<16 x i16> %a, <16 x i16>* %ptr_b, i16 %mask) {
; CHECK-LABEL: test_mask_sub_epi16_rmkz_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpsubw (%rdi), %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xa9,0xf9,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <16 x i16>, <16 x i16>* %ptr_b
  %res = call <16 x i16> @llvm.x86.avx512.mask.psub.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> zeroinitializer, i16 %mask)
  ret <16 x i16> %res
}

declare <16 x i16> @llvm.x86.avx512.mask.psub.w.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)

define <32 x i16> @test_mask_add_epi16_rr_512(<32 x i16> %a, <32 x i16> %b) {
; CHECK-LABEL: test_mask_add_epi16_rr_512:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpaddw %zmm1, %zmm0, %zmm0 ## encoding: [0x62,0xf1,0x7d,0x48,0xfd,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <32 x i16> @llvm.x86.avx512.mask.padd.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> zeroinitializer, i32 -1)
  ret <32 x i16> %res
}

define <32 x i16> @test_mask_add_epi16_rrk_512(<32 x i16> %a, <32 x i16> %b, <32 x i16> %passThru, i32 %mask) {
; CHECK-LABEL: test_mask_add_epi16_rrk_512:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
; CHECK-NEXT:    vpaddw %zmm1, %zmm0, %zmm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x49,0xfd,0xd1]
; CHECK-NEXT:    vmovaps %zmm2, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <32 x i16> @llvm.x86.avx512.mask.padd.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> %passThru, i32 %mask)
  ret <32 x i16> %res
}

define <32 x i16> @test_mask_add_epi16_rrkz_512(<32 x i16> %a, <32 x i16> %b, i32 %mask) {
; CHECK-LABEL: test_mask_add_epi16_rrkz_512:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
; CHECK-NEXT:    vpaddw %zmm1, %zmm0, %zmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xc9,0xfd,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <32 x i16> @llvm.x86.avx512.mask.padd.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> zeroinitializer, i32 %mask)
  ret <32 x i16> %res
}

define <32 x i16> @test_mask_add_epi16_rm_512(<32 x i16> %a, <32 x i16>* %ptr_b) {
; CHECK-LABEL: test_mask_add_epi16_rm_512:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpaddw (%rdi), %zmm0, %zmm0 ## encoding: [0x62,0xf1,0x7d,0x48,0xfd,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <32 x i16>, <32 x i16>* %ptr_b
  %res = call <32 x i16> @llvm.x86.avx512.mask.padd.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> zeroinitializer, i32 -1)
  ret <32 x i16> %res
}

define <32 x i16> @test_mask_add_epi16_rmk_512(<32 x i16> %a, <32 x i16>* %ptr_b, <32 x i16> %passThru, i32 %mask) {
; CHECK-LABEL: test_mask_add_epi16_rmk_512:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovd %esi, %k1 ## encoding: [0xc5,0xfb,0x92,0xce]
; CHECK-NEXT:    vpaddw (%rdi), %zmm0, %zmm1 {%k1} ## encoding: [0x62,0xf1,0x7d,0x49,0xfd,0x0f]
; CHECK-NEXT:    vmovaps %zmm1, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <32 x i16>, <32 x i16>* %ptr_b
  %res = call <32 x i16> @llvm.x86.avx512.mask.padd.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> %passThru, i32 %mask)
  ret <32 x i16> %res
}

define <32 x i16> @test_mask_add_epi16_rmkz_512(<32 x i16> %a, <32 x i16>* %ptr_b, i32 %mask) {
; CHECK-LABEL: test_mask_add_epi16_rmkz_512:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovd %esi, %k1 ## encoding: [0xc5,0xfb,0x92,0xce]
; CHECK-NEXT:    vpaddw (%rdi), %zmm0, %zmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xc9,0xfd,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <32 x i16>, <32 x i16>* %ptr_b
  %res = call <32 x i16> @llvm.x86.avx512.mask.padd.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> zeroinitializer, i32 %mask)
  ret <32 x i16> %res
}

declare <32 x i16> @llvm.x86.avx512.mask.padd.w.512(<32 x i16>, <32 x i16>, <32 x i16>, i32)

define <32 x i16> @test_mask_sub_epi16_rr_512(<32 x i16> %a, <32 x i16> %b) {
; CHECK-LABEL: test_mask_sub_epi16_rr_512:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpsubw %zmm1, %zmm0, %zmm0 ## encoding: [0x62,0xf1,0x7d,0x48,0xf9,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <32 x i16> @llvm.x86.avx512.mask.psub.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> zeroinitializer, i32 -1)
  ret <32 x i16> %res
}

define <32 x i16> @test_mask_sub_epi16_rrk_512(<32 x i16> %a, <32 x i16> %b, <32 x i16> %passThru, i32 %mask) {
; CHECK-LABEL: test_mask_sub_epi16_rrk_512:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
; CHECK-NEXT:    vpsubw %zmm1, %zmm0, %zmm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x49,0xf9,0xd1]
; CHECK-NEXT:    vmovaps %zmm2, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <32 x i16> @llvm.x86.avx512.mask.psub.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> %passThru, i32 %mask)
  ret <32 x i16> %res
}

define <32 x i16> @test_mask_sub_epi16_rrkz_512(<32 x i16> %a, <32 x i16> %b, i32 %mask) {
; CHECK-LABEL: test_mask_sub_epi16_rrkz_512:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
; CHECK-NEXT:    vpsubw %zmm1, %zmm0, %zmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xc9,0xf9,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <32 x i16> @llvm.x86.avx512.mask.psub.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> zeroinitializer, i32 %mask)
  ret <32 x i16> %res
}

define <32 x i16> @test_mask_sub_epi16_rm_512(<32 x i16> %a, <32 x i16>* %ptr_b) {
; CHECK-LABEL: test_mask_sub_epi16_rm_512:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpsubw (%rdi), %zmm0, %zmm0 ## encoding: [0x62,0xf1,0x7d,0x48,0xf9,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <32 x i16>, <32 x i16>* %ptr_b
  %res = call <32 x i16> @llvm.x86.avx512.mask.psub.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> zeroinitializer, i32 -1)
  ret <32 x i16> %res
}

define <32 x i16> @test_mask_sub_epi16_rmk_512(<32 x i16> %a, <32 x i16>* %ptr_b, <32 x i16> %passThru, i32 %mask) {
; CHECK-LABEL: test_mask_sub_epi16_rmk_512:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovd %esi, %k1 ## encoding: [0xc5,0xfb,0x92,0xce]
; CHECK-NEXT:    vpsubw (%rdi), %zmm0, %zmm1 {%k1} ## encoding: [0x62,0xf1,0x7d,0x49,0xf9,0x0f]
; CHECK-NEXT:    vmovaps %zmm1, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <32 x i16>, <32 x i16>* %ptr_b
  %res = call <32 x i16> @llvm.x86.avx512.mask.psub.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> %passThru, i32 %mask)
  ret <32 x i16> %res
}

define <32 x i16> @test_mask_sub_epi16_rmkz_512(<32 x i16> %a, <32 x i16>* %ptr_b, i32 %mask) {
; CHECK-LABEL: test_mask_sub_epi16_rmkz_512:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovd %esi, %k1 ## encoding: [0xc5,0xfb,0x92,0xce]
; CHECK-NEXT:    vpsubw (%rdi), %zmm0, %zmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xc9,0xf9,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <32 x i16>, <32 x i16>* %ptr_b
  %res = call <32 x i16> @llvm.x86.avx512.mask.psub.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> zeroinitializer, i32 %mask)
  ret <32 x i16> %res
}

declare <32 x i16> @llvm.x86.avx512.mask.psub.w.512(<32 x i16>, <32 x i16>, <32 x i16>, i32)

define <32 x i16> @test_mask_mullo_epi16_rr_512(<32 x i16> %a, <32 x i16> %b) {
; CHECK-LABEL: test_mask_mullo_epi16_rr_512:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpmullw %zmm1, %zmm0, %zmm0 ## encoding: [0x62,0xf1,0x7d,0x48,0xd5,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <32 x i16> @llvm.x86.avx512.mask.pmull.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> zeroinitializer, i32 -1)
  ret <32 x i16> %res
}

define <32 x i16> @test_mask_mullo_epi16_rrk_512(<32 x i16> %a, <32 x i16> %b, <32 x i16> %passThru, i32 %mask) {
; CHECK-LABEL: test_mask_mullo_epi16_rrk_512:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
; CHECK-NEXT:    vpmullw %zmm1, %zmm0, %zmm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x49,0xd5,0xd1]
; CHECK-NEXT:    vmovaps %zmm2, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <32 x i16> @llvm.x86.avx512.mask.pmull.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> %passThru, i32 %mask)
  ret <32 x i16> %res
}

define <32 x i16> @test_mask_mullo_epi16_rrkz_512(<32 x i16> %a, <32 x i16> %b, i32 %mask) {
; CHECK-LABEL: test_mask_mullo_epi16_rrkz_512:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
; CHECK-NEXT:    vpmullw %zmm1, %zmm0, %zmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xc9,0xd5,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <32 x i16> @llvm.x86.avx512.mask.pmull.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> zeroinitializer, i32 %mask)
  ret <32 x i16> %res
}

define <32 x i16> @test_mask_mullo_epi16_rm_512(<32 x i16> %a, <32 x i16>* %ptr_b) {
; CHECK-LABEL: test_mask_mullo_epi16_rm_512:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpmullw (%rdi), %zmm0, %zmm0 ## encoding: [0x62,0xf1,0x7d,0x48,0xd5,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <32 x i16>, <32 x i16>* %ptr_b
  %res = call <32 x i16> @llvm.x86.avx512.mask.pmull.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> zeroinitializer, i32 -1)
  ret <32 x i16> %res
}

define <32 x i16> @test_mask_mullo_epi16_rmk_512(<32 x i16> %a, <32 x i16>* %ptr_b, <32 x i16> %passThru, i32 %mask) {
; CHECK-LABEL: test_mask_mullo_epi16_rmk_512:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovd %esi, %k1 ## encoding: [0xc5,0xfb,0x92,0xce]
; CHECK-NEXT:    vpmullw (%rdi), %zmm0, %zmm1 {%k1} ## encoding: [0x62,0xf1,0x7d,0x49,0xd5,0x0f]
; CHECK-NEXT:    vmovaps %zmm1, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <32 x i16>, <32 x i16>* %ptr_b
  %res = call <32 x i16> @llvm.x86.avx512.mask.pmull.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> %passThru, i32 %mask)
  ret <32 x i16> %res
}

define <32 x i16> @test_mask_mullo_epi16_rmkz_512(<32 x i16> %a, <32 x i16>* %ptr_b, i32 %mask) {
; CHECK-LABEL: test_mask_mullo_epi16_rmkz_512:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovd %esi, %k1 ## encoding: [0xc5,0xfb,0x92,0xce]
; CHECK-NEXT:    vpmullw (%rdi), %zmm0, %zmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xc9,0xd5,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <32 x i16>, <32 x i16>* %ptr_b
  %res = call <32 x i16> @llvm.x86.avx512.mask.pmull.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> zeroinitializer, i32 %mask)
  ret <32 x i16> %res
}

declare <32 x i16> @llvm.x86.avx512.mask.pmull.w.512(<32 x i16>, <32 x i16>, <32 x i16>, i32)

define <8 x i16> @test_mask_mullo_epi16_rr_128(<8 x i16> %a, <8 x i16> %b) {
; CHECK-LABEL: test_mask_mullo_epi16_rr_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpmullw %xmm1, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xd5,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.mask.pmull.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> zeroinitializer, i8 -1)
  ret <8 x i16> %res
}

define <8 x i16> @test_mask_mullo_epi16_rrk_128(<8 x i16> %a, <8 x i16> %b, <8 x i16> %passThru, i8 %mask) {
; CHECK-LABEL: test_mask_mullo_epi16_rrk_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpmullw %xmm1, %xmm0, %xmm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0xd5,0xd1]
; CHECK-NEXT:    vmovaps %zmm2, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.mask.pmull.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> %passThru, i8 %mask)
  ret <8 x i16> %res
}

define <8 x i16> @test_mask_mullo_epi16_rrkz_128(<8 x i16> %a, <8 x i16> %b, i8 %mask) {
; CHECK-LABEL: test_mask_mullo_epi16_rrkz_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpmullw %xmm1, %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0x89,0xd5,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.mask.pmull.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> zeroinitializer, i8 %mask)
  ret <8 x i16> %res
}

define <8 x i16> @test_mask_mullo_epi16_rm_128(<8 x i16> %a, <8 x i16>* %ptr_b) {
; CHECK-LABEL: test_mask_mullo_epi16_rm_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpmullw (%rdi), %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xd5,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <8 x i16>, <8 x i16>* %ptr_b
  %res = call <8 x i16> @llvm.x86.avx512.mask.pmull.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> zeroinitializer, i8 -1)
  ret <8 x i16> %res
}

define <8 x i16> @test_mask_mullo_epi16_rmk_128(<8 x i16> %a, <8 x i16>* %ptr_b, <8 x i16> %passThru, i8 %mask) {
; CHECK-LABEL: test_mask_mullo_epi16_rmk_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpmullw (%rdi), %xmm0, %xmm1 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0xd5,0x0f]
; CHECK-NEXT:    vmovaps %zmm1, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <8 x i16>, <8 x i16>* %ptr_b
  %res = call <8 x i16> @llvm.x86.avx512.mask.pmull.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> %passThru, i8 %mask)
  ret <8 x i16> %res
}

define <8 x i16> @test_mask_mullo_epi16_rmkz_128(<8 x i16> %a, <8 x i16>* %ptr_b, i8 %mask) {
; CHECK-LABEL: test_mask_mullo_epi16_rmkz_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpmullw (%rdi), %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0x89,0xd5,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <8 x i16>, <8 x i16>* %ptr_b
  %res = call <8 x i16> @llvm.x86.avx512.mask.pmull.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> zeroinitializer, i8 %mask)
  ret <8 x i16> %res
}

declare <8 x i16> @llvm.x86.avx512.mask.pmull.w.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)

define <16 x i16> @test_mask_mullo_epi16_rr_256(<16 x i16> %a, <16 x i16> %b) {
; CHECK-LABEL: test_mask_mullo_epi16_rr_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpmullw %ymm1, %ymm0, %ymm0 ## encoding: [0x62,0xf1,0x7d,0x28,0xd5,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.mask.pmull.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> zeroinitializer, i16 -1)
  ret <16 x i16> %res
}

define <16 x i16> @test_mask_mullo_epi16_rrk_256(<16 x i16> %a, <16 x i16> %b, <16 x i16> %passThru, i16 %mask) {
; CHECK-LABEL: test_mask_mullo_epi16_rrk_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpmullw %ymm1, %ymm0, %ymm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0xd5,0xd1]
; CHECK-NEXT:    vmovaps %zmm2, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.mask.pmull.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> %passThru, i16 %mask)
  ret <16 x i16> %res
}

define <16 x i16> @test_mask_mullo_epi16_rrkz_256(<16 x i16> %a, <16 x i16> %b, i16 %mask) {
; CHECK-LABEL: test_mask_mullo_epi16_rrkz_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpmullw %ymm1, %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xa9,0xd5,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.mask.pmull.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> zeroinitializer, i16 %mask)
  ret <16 x i16> %res
}

define <16 x i16> @test_mask_mullo_epi16_rm_256(<16 x i16> %a, <16 x i16>* %ptr_b) {
; CHECK-LABEL: test_mask_mullo_epi16_rm_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpmullw (%rdi), %ymm0, %ymm0 ## encoding: [0x62,0xf1,0x7d,0x28,0xd5,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <16 x i16>, <16 x i16>* %ptr_b
  %res = call <16 x i16> @llvm.x86.avx512.mask.pmull.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> zeroinitializer, i16 -1)
  ret <16 x i16> %res
}

define <16 x i16> @test_mask_mullo_epi16_rmk_256(<16 x i16> %a, <16 x i16>* %ptr_b, <16 x i16> %passThru, i16 %mask) {
; CHECK-LABEL: test_mask_mullo_epi16_rmk_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpmullw (%rdi), %ymm0, %ymm1 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0xd5,0x0f]
; CHECK-NEXT:    vmovaps %zmm1, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <16 x i16>, <16 x i16>* %ptr_b
  %res = call <16 x i16> @llvm.x86.avx512.mask.pmull.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> %passThru, i16 %mask)
  ret <16 x i16> %res
}

define <16 x i16> @test_mask_mullo_epi16_rmkz_256(<16 x i16> %a, <16 x i16>* %ptr_b, i16 %mask) {
; CHECK-LABEL: test_mask_mullo_epi16_rmkz_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpmullw (%rdi), %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xa9,0xd5,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <16 x i16>, <16 x i16>* %ptr_b
  %res = call <16 x i16> @llvm.x86.avx512.mask.pmull.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> zeroinitializer, i16 %mask)
  ret <16 x i16> %res
}

declare <16 x i16> @llvm.x86.avx512.mask.pmull.w.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)


define <8 x i16> @test_mask_packs_epi32_rr_128(<4 x i32> %a, <4 x i32> %b) {
; CHECK-LABEL: test_mask_packs_epi32_rr_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpackssdw %xmm1, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0x6b,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.mask.packssdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> zeroinitializer, i8 -1)
  ret <8 x i16> %res
}

define <8 x i16> @test_mask_packs_epi32_rrk_128(<4 x i32> %a, <4 x i32> %b, <8 x i16> %passThru, i8 %mask) {
; CHECK-LABEL: test_mask_packs_epi32_rrk_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpackssdw %xmm1, %xmm0, %xmm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0x6b,0xd1]
; CHECK-NEXT:    vmovaps %zmm2, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.mask.packssdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> %passThru, i8 %mask)
  ret <8 x i16> %res
}

define <8 x i16> @test_mask_packs_epi32_rrkz_128(<4 x i32> %a, <4 x i32> %b, i8 %mask) {
; CHECK-LABEL: test_mask_packs_epi32_rrkz_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpackssdw %xmm1, %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0x89,0x6b,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.mask.packssdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> zeroinitializer, i8 %mask)
  ret <8 x i16> %res
}

define <8 x i16> @test_mask_packs_epi32_rm_128(<4 x i32> %a, <4 x i32>* %ptr_b) {
; CHECK-LABEL: test_mask_packs_epi32_rm_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpackssdw (%rdi), %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0x6b,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <4 x i32>, <4 x i32>* %ptr_b
  %res = call <8 x i16> @llvm.x86.avx512.mask.packssdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> zeroinitializer, i8 -1)
  ret <8 x i16> %res
}

define <8 x i16> @test_mask_packs_epi32_rmk_128(<4 x i32> %a, <4 x i32>* %ptr_b, <8 x i16> %passThru, i8 %mask) {
; CHECK-LABEL: test_mask_packs_epi32_rmk_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpackssdw (%rdi), %xmm0, %xmm1 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0x6b,0x0f]
; CHECK-NEXT:    vmovaps %zmm1, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <4 x i32>, <4 x i32>* %ptr_b
  %res = call <8 x i16> @llvm.x86.avx512.mask.packssdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> %passThru, i8 %mask)
  ret <8 x i16> %res
}

define <8 x i16> @test_mask_packs_epi32_rmkz_128(<4 x i32> %a, <4 x i32>* %ptr_b, i8 %mask) {
; CHECK-LABEL: test_mask_packs_epi32_rmkz_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpackssdw (%rdi), %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0x89,0x6b,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <4 x i32>, <4 x i32>* %ptr_b
  %res = call <8 x i16> @llvm.x86.avx512.mask.packssdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> zeroinitializer, i8 %mask)
  ret <8 x i16> %res
}

define <8 x i16> @test_mask_packs_epi32_rmb_128(<4 x i32> %a, i32* %ptr_b) {
; CHECK-LABEL: test_mask_packs_epi32_rmb_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpackssdw (%rdi){1to4}, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x18,0x6b,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %q = load i32, i32* %ptr_b
  %vecinit.i = insertelement <4 x i32> undef, i32 %q, i32 0
  %b = shufflevector <4 x i32> %vecinit.i, <4 x i32> undef, <4 x i32> zeroinitializer
  %res = call <8 x i16> @llvm.x86.avx512.mask.packssdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> zeroinitializer, i8 -1)
  ret <8 x i16> %res
}

define <8 x i16> @test_mask_packs_epi32_rmbk_128(<4 x i32> %a, i32* %ptr_b, <8 x i16> %passThru, i8 %mask) {
; CHECK-LABEL: test_mask_packs_epi32_rmbk_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpackssdw (%rdi){1to4}, %xmm0, %xmm1 {%k1} ## encoding: [0x62,0xf1,0x7d,0x19,0x6b,0x0f]
; CHECK-NEXT:    vmovaps %zmm1, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %q = load i32, i32* %ptr_b
  %vecinit.i = insertelement <4 x i32> undef, i32 %q, i32 0
  %b = shufflevector <4 x i32> %vecinit.i, <4 x i32> undef, <4 x i32> zeroinitializer
  %res = call <8 x i16> @llvm.x86.avx512.mask.packssdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> %passThru, i8 %mask)
  ret <8 x i16> %res
}

define <8 x i16> @test_mask_packs_epi32_rmbkz_128(<4 x i32> %a, i32* %ptr_b, i8 %mask) {
; CHECK-LABEL: test_mask_packs_epi32_rmbkz_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpackssdw (%rdi){1to4}, %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0x99,0x6b,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %q = load i32, i32* %ptr_b
  %vecinit.i = insertelement <4 x i32> undef, i32 %q, i32 0
  %b = shufflevector <4 x i32> %vecinit.i, <4 x i32> undef, <4 x i32> zeroinitializer
  %res = call <8 x i16> @llvm.x86.avx512.mask.packssdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> zeroinitializer, i8 %mask)
  ret <8 x i16> %res
}

declare <8 x i16> @llvm.x86.avx512.mask.packssdw.128(<4 x i32>, <4 x i32>, <8 x i16>, i8)

define <16 x i16> @test_mask_packs_epi32_rr_256(<8 x i32> %a, <8 x i32> %b) {
; CHECK-LABEL: test_mask_packs_epi32_rr_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpackssdw %ymm1, %ymm0, %ymm0 ## encoding: [0x62,0xf1,0x7d,0x28,0x6b,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.mask.packssdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> zeroinitializer, i16 -1)
  ret <16 x i16> %res
}

define <16 x i16> @test_mask_packs_epi32_rrk_256(<8 x i32> %a, <8 x i32> %b, <16 x i16> %passThru, i16 %mask) {
; CHECK-LABEL: test_mask_packs_epi32_rrk_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpackssdw %ymm1, %ymm0, %ymm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0x6b,0xd1]
; CHECK-NEXT:    vmovaps %zmm2, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.mask.packssdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> %passThru, i16 %mask)
  ret <16 x i16> %res
}

define <16 x i16> @test_mask_packs_epi32_rrkz_256(<8 x i32> %a, <8 x i32> %b, i16 %mask) {
; CHECK-LABEL: test_mask_packs_epi32_rrkz_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpackssdw %ymm1, %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xa9,0x6b,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.mask.packssdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> zeroinitializer, i16 %mask)
  ret <16 x i16> %res
}

define <16 x i16> @test_mask_packs_epi32_rm_256(<8 x i32> %a, <8 x i32>* %ptr_b) {
; CHECK-LABEL: test_mask_packs_epi32_rm_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpackssdw (%rdi), %ymm0, %ymm0 ## encoding: [0x62,0xf1,0x7d,0x28,0x6b,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <8 x i32>, <8 x i32>* %ptr_b
  %res = call <16 x i16> @llvm.x86.avx512.mask.packssdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> zeroinitializer, i16 -1)
  ret <16 x i16> %res
}

define <16 x i16> @test_mask_packs_epi32_rmk_256(<8 x i32> %a, <8 x i32>* %ptr_b, <16 x i16> %passThru, i16 %mask) {
; CHECK-LABEL: test_mask_packs_epi32_rmk_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpackssdw (%rdi), %ymm0, %ymm1 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0x6b,0x0f]
; CHECK-NEXT:    vmovaps %zmm1, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <8 x i32>, <8 x i32>* %ptr_b
  %res = call <16 x i16> @llvm.x86.avx512.mask.packssdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> %passThru, i16 %mask)
  ret <16 x i16> %res
}

define <16 x i16> @test_mask_packs_epi32_rmkz_256(<8 x i32> %a, <8 x i32>* %ptr_b, i16 %mask) {
; CHECK-LABEL: test_mask_packs_epi32_rmkz_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpackssdw (%rdi), %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xa9,0x6b,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <8 x i32>, <8 x i32>* %ptr_b
  %res = call <16 x i16> @llvm.x86.avx512.mask.packssdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> zeroinitializer, i16 %mask)
  ret <16 x i16> %res
}

define <16 x i16> @test_mask_packs_epi32_rmb_256(<8 x i32> %a, i32* %ptr_b) {
; CHECK-LABEL: test_mask_packs_epi32_rmb_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpackssdw (%rdi){1to8}, %ymm0, %ymm0 ## encoding: [0x62,0xf1,0x7d,0x38,0x6b,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %q = load i32, i32* %ptr_b
  %vecinit.i = insertelement <8 x i32> undef, i32 %q, i32 0
  %b = shufflevector <8 x i32> %vecinit.i, <8 x i32> undef, <8 x i32> zeroinitializer
  %res = call <16 x i16> @llvm.x86.avx512.mask.packssdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> zeroinitializer, i16 -1)
  ret <16 x i16> %res
}

define <16 x i16> @test_mask_packs_epi32_rmbk_256(<8 x i32> %a, i32* %ptr_b, <16 x i16> %passThru, i16 %mask) {
; CHECK-LABEL: test_mask_packs_epi32_rmbk_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpackssdw (%rdi){1to8}, %ymm0, %ymm1 {%k1} ## encoding: [0x62,0xf1,0x7d,0x39,0x6b,0x0f]
; CHECK-NEXT:    vmovaps %zmm1, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %q = load i32, i32* %ptr_b
  %vecinit.i = insertelement <8 x i32> undef, i32 %q, i32 0
  %b = shufflevector <8 x i32> %vecinit.i, <8 x i32> undef, <8 x i32> zeroinitializer
  %res = call <16 x i16> @llvm.x86.avx512.mask.packssdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> %passThru, i16 %mask)
  ret <16 x i16> %res
}

define <16 x i16> @test_mask_packs_epi32_rmbkz_256(<8 x i32> %a, i32* %ptr_b, i16 %mask) {
; CHECK-LABEL: test_mask_packs_epi32_rmbkz_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpackssdw (%rdi){1to8}, %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xb9,0x6b,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %q = load i32, i32* %ptr_b
  %vecinit.i = insertelement <8 x i32> undef, i32 %q, i32 0
  %b = shufflevector <8 x i32> %vecinit.i, <8 x i32> undef, <8 x i32> zeroinitializer
  %res = call <16 x i16> @llvm.x86.avx512.mask.packssdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> zeroinitializer, i16 %mask)
  ret <16 x i16> %res
}

declare <16 x i16> @llvm.x86.avx512.mask.packssdw.256(<8 x i32>, <8 x i32>, <16 x i16>, i16)

define <16 x i8> @test_mask_packs_epi16_rr_128(<8 x i16> %a, <8 x i16> %b) {
; CHECK-LABEL: test_mask_packs_epi16_rr_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0x63,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i8> @llvm.x86.avx512.mask.packsswb.128(<8 x i16> %a, <8 x i16> %b, <16 x i8> zeroinitializer, i16 -1)
  ret <16 x i8> %res
}

define <16 x i8> @test_mask_packs_epi16_rrk_128(<8 x i16> %a, <8 x i16> %b, <16 x i8> %passThru, i16 %mask) {
; CHECK-LABEL: test_mask_packs_epi16_rrk_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpacksswb %xmm1, %xmm0, %xmm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0x63,0xd1]
; CHECK-NEXT:    vmovaps %zmm2, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i8> @llvm.x86.avx512.mask.packsswb.128(<8 x i16> %a, <8 x i16> %b, <16 x i8> %passThru, i16 %mask)
  ret <16 x i8> %res
}

define <16 x i8> @test_mask_packs_epi16_rrkz_128(<8 x i16> %a, <8 x i16> %b, i16 %mask) {
; CHECK-LABEL: test_mask_packs_epi16_rrkz_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0x89,0x63,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i8> @llvm.x86.avx512.mask.packsswb.128(<8 x i16> %a, <8 x i16> %b, <16 x i8> zeroinitializer, i16 %mask)
  ret <16 x i8> %res
}

define <16 x i8> @test_mask_packs_epi16_rm_128(<8 x i16> %a, <8 x i16>* %ptr_b) {
; CHECK-LABEL: test_mask_packs_epi16_rm_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpacksswb (%rdi), %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0x63,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <8 x i16>, <8 x i16>* %ptr_b
  %res = call <16 x i8> @llvm.x86.avx512.mask.packsswb.128(<8 x i16> %a, <8 x i16> %b, <16 x i8> zeroinitializer, i16 -1)
  ret <16 x i8> %res
}

define <16 x i8> @test_mask_packs_epi16_rmk_128(<8 x i16> %a, <8 x i16>* %ptr_b, <16 x i8> %passThru, i16 %mask) {
; CHECK-LABEL: test_mask_packs_epi16_rmk_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpacksswb (%rdi), %xmm0, %xmm1 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0x63,0x0f]
; CHECK-NEXT:    vmovaps %zmm1, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <8 x i16>, <8 x i16>* %ptr_b
  %res = call <16 x i8> @llvm.x86.avx512.mask.packsswb.128(<8 x i16> %a, <8 x i16> %b, <16 x i8> %passThru, i16 %mask)
  ret <16 x i8> %res
}

define <16 x i8> @test_mask_packs_epi16_rmkz_128(<8 x i16> %a, <8 x i16>* %ptr_b, i16 %mask) {
; CHECK-LABEL: test_mask_packs_epi16_rmkz_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpacksswb (%rdi), %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0x89,0x63,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <8 x i16>, <8 x i16>* %ptr_b
  %res = call <16 x i8> @llvm.x86.avx512.mask.packsswb.128(<8 x i16> %a, <8 x i16> %b, <16 x i8> zeroinitializer, i16 %mask)
  ret <16 x i8> %res
}

declare <16 x i8> @llvm.x86.avx512.mask.packsswb.128(<8 x i16>, <8 x i16>, <16 x i8>, i16)

define <32 x i8> @test_mask_packs_epi16_rr_256(<16 x i16> %a, <16 x i16> %b) {
; CHECK-LABEL: test_mask_packs_epi16_rr_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpacksswb %ymm1, %ymm0, %ymm0 ## encoding: [0x62,0xf1,0x7d,0x28,0x63,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <32 x i8> @llvm.x86.avx512.mask.packsswb.256(<16 x i16> %a, <16 x i16> %b, <32 x i8> zeroinitializer, i32 -1)
  ret <32 x i8> %res
}

define <32 x i8> @test_mask_packs_epi16_rrk_256(<16 x i16> %a, <16 x i16> %b, <32 x i8> %passThru, i32 %mask) {
; CHECK-LABEL: test_mask_packs_epi16_rrk_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
; CHECK-NEXT:    vpacksswb %ymm1, %ymm0, %ymm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0x63,0xd1]
; CHECK-NEXT:    vmovaps %zmm2, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <32 x i8> @llvm.x86.avx512.mask.packsswb.256(<16 x i16> %a, <16 x i16> %b, <32 x i8> %passThru, i32 %mask)
  ret <32 x i8> %res
}

define <32 x i8> @test_mask_packs_epi16_rrkz_256(<16 x i16> %a, <16 x i16> %b, i32 %mask) {
; CHECK-LABEL: test_mask_packs_epi16_rrkz_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
; CHECK-NEXT:    vpacksswb %ymm1, %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xa9,0x63,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <32 x i8> @llvm.x86.avx512.mask.packsswb.256(<16 x i16> %a, <16 x i16> %b, <32 x i8> zeroinitializer, i32 %mask)
  ret <32 x i8> %res
}

define <32 x i8> @test_mask_packs_epi16_rm_256(<16 x i16> %a, <16 x i16>* %ptr_b) {
; CHECK-LABEL: test_mask_packs_epi16_rm_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpacksswb (%rdi), %ymm0, %ymm0 ## encoding: [0x62,0xf1,0x7d,0x28,0x63,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <16 x i16>, <16 x i16>* %ptr_b
  %res = call <32 x i8> @llvm.x86.avx512.mask.packsswb.256(<16 x i16> %a, <16 x i16> %b, <32 x i8> zeroinitializer, i32 -1)
  ret <32 x i8> %res
}

define <32 x i8> @test_mask_packs_epi16_rmk_256(<16 x i16> %a, <16 x i16>* %ptr_b, <32 x i8> %passThru, i32 %mask) {
; CHECK-LABEL: test_mask_packs_epi16_rmk_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovd %esi, %k1 ## encoding: [0xc5,0xfb,0x92,0xce]
; CHECK-NEXT:    vpacksswb (%rdi), %ymm0, %ymm1 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0x63,0x0f]
; CHECK-NEXT:    vmovaps %zmm1, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <16 x i16>, <16 x i16>* %ptr_b
  %res = call <32 x i8> @llvm.x86.avx512.mask.packsswb.256(<16 x i16> %a, <16 x i16> %b, <32 x i8> %passThru, i32 %mask)
  ret <32 x i8> %res
}

define <32 x i8> @test_mask_packs_epi16_rmkz_256(<16 x i16> %a, <16 x i16>* %ptr_b, i32 %mask) {
; CHECK-LABEL: test_mask_packs_epi16_rmkz_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovd %esi, %k1 ## encoding: [0xc5,0xfb,0x92,0xce]
; CHECK-NEXT:    vpacksswb (%rdi), %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xa9,0x63,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <16 x i16>, <16 x i16>* %ptr_b
  %res = call <32 x i8> @llvm.x86.avx512.mask.packsswb.256(<16 x i16> %a, <16 x i16> %b, <32 x i8> zeroinitializer, i32 %mask)
  ret <32 x i8> %res
}

declare <32 x i8> @llvm.x86.avx512.mask.packsswb.256(<16 x i16>, <16 x i16>, <32 x i8>, i32)


define <8 x i16> @test_mask_packus_epi32_rr_128(<4 x i32> %a, <4 x i32> %b) {
; CHECK-LABEL: test_mask_packus_epi32_rr_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0 ## encoding: [0x62,0xf2,0x7d,0x08,0x2b,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.mask.packusdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> zeroinitializer, i8 -1)
  ret <8 x i16> %res
}

define <8 x i16> @test_mask_packus_epi32_rrk_128(<4 x i32> %a, <4 x i32> %b, <8 x i16> %passThru, i8 %mask) {
; CHECK-LABEL: test_mask_packus_epi32_rrk_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpackusdw %xmm1, %xmm0, %xmm2 {%k1} ## encoding: [0x62,0xf2,0x7d,0x09,0x2b,0xd1]
; CHECK-NEXT:    vmovaps %zmm2, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.mask.packusdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> %passThru, i8 %mask)
  ret <8 x i16> %res
}

define <8 x i16> @test_mask_packus_epi32_rrkz_128(<4 x i32> %a, <4 x i32> %b, i8 %mask) {
; CHECK-LABEL: test_mask_packus_epi32_rrkz_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf2,0x7d,0x89,0x2b,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.mask.packusdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> zeroinitializer, i8 %mask)
  ret <8 x i16> %res
}

define <8 x i16> @test_mask_packus_epi32_rm_128(<4 x i32> %a, <4 x i32>* %ptr_b) {
; CHECK-LABEL: test_mask_packus_epi32_rm_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpackusdw (%rdi), %xmm0, %xmm0 ## encoding: [0x62,0xf2,0x7d,0x08,0x2b,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <4 x i32>, <4 x i32>* %ptr_b
  %res = call <8 x i16> @llvm.x86.avx512.mask.packusdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> zeroinitializer, i8 -1)
  ret <8 x i16> %res
}

define <8 x i16> @test_mask_packus_epi32_rmk_128(<4 x i32> %a, <4 x i32>* %ptr_b, <8 x i16> %passThru, i8 %mask) {
; CHECK-LABEL: test_mask_packus_epi32_rmk_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpackusdw (%rdi), %xmm0, %xmm1 {%k1} ## encoding: [0x62,0xf2,0x7d,0x09,0x2b,0x0f]
; CHECK-NEXT:    vmovaps %zmm1, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <4 x i32>, <4 x i32>* %ptr_b
  %res = call <8 x i16> @llvm.x86.avx512.mask.packusdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> %passThru, i8 %mask)
  ret <8 x i16> %res
}

define <8 x i16> @test_mask_packus_epi32_rmkz_128(<4 x i32> %a, <4 x i32>* %ptr_b, i8 %mask) {
; CHECK-LABEL: test_mask_packus_epi32_rmkz_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpackusdw (%rdi), %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf2,0x7d,0x89,0x2b,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <4 x i32>, <4 x i32>* %ptr_b
  %res = call <8 x i16> @llvm.x86.avx512.mask.packusdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> zeroinitializer, i8 %mask)
  ret <8 x i16> %res
}

define <8 x i16> @test_mask_packus_epi32_rmb_128(<4 x i32> %a, i32* %ptr_b) {
; CHECK-LABEL: test_mask_packus_epi32_rmb_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpackusdw (%rdi){1to4}, %xmm0, %xmm0 ## encoding: [0x62,0xf2,0x7d,0x18,0x2b,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %q = load i32, i32* %ptr_b
  %vecinit.i = insertelement <4 x i32> undef, i32 %q, i32 0
  %b = shufflevector <4 x i32> %vecinit.i, <4 x i32> undef, <4 x i32> zeroinitializer
  %res = call <8 x i16> @llvm.x86.avx512.mask.packusdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> zeroinitializer, i8 -1)
  ret <8 x i16> %res
}

define <8 x i16> @test_mask_packus_epi32_rmbk_128(<4 x i32> %a, i32* %ptr_b, <8 x i16> %passThru, i8 %mask) {
; CHECK-LABEL: test_mask_packus_epi32_rmbk_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpackusdw (%rdi){1to4}, %xmm0, %xmm1 {%k1} ## encoding: [0x62,0xf2,0x7d,0x19,0x2b,0x0f]
; CHECK-NEXT:    vmovaps %zmm1, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %q = load i32, i32* %ptr_b
  %vecinit.i = insertelement <4 x i32> undef, i32 %q, i32 0
  %b = shufflevector <4 x i32> %vecinit.i, <4 x i32> undef, <4 x i32> zeroinitializer
  %res = call <8 x i16> @llvm.x86.avx512.mask.packusdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> %passThru, i8 %mask)
  ret <8 x i16> %res
}

define <8 x i16> @test_mask_packus_epi32_rmbkz_128(<4 x i32> %a, i32* %ptr_b, i8 %mask) {
; CHECK-LABEL: test_mask_packus_epi32_rmbkz_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpackusdw (%rdi){1to4}, %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf2,0x7d,0x99,0x2b,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %q = load i32, i32* %ptr_b
  %vecinit.i = insertelement <4 x i32> undef, i32 %q, i32 0
  %b = shufflevector <4 x i32> %vecinit.i, <4 x i32> undef, <4 x i32> zeroinitializer
  %res = call <8 x i16> @llvm.x86.avx512.mask.packusdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> zeroinitializer, i8 %mask)
  ret <8 x i16> %res
}

declare <8 x i16> @llvm.x86.avx512.mask.packusdw.128(<4 x i32>, <4 x i32>, <8 x i16>, i8)

define <16 x i16> @test_mask_packus_epi32_rr_256(<8 x i32> %a, <8 x i32> %b) {
; CHECK-LABEL: test_mask_packus_epi32_rr_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpackusdw %ymm1, %ymm0, %ymm0 ## encoding: [0x62,0xf2,0x7d,0x28,0x2b,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.mask.packusdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> zeroinitializer, i16 -1)
  ret <16 x i16> %res
}

define <16 x i16> @test_mask_packus_epi32_rrk_256(<8 x i32> %a, <8 x i32> %b, <16 x i16> %passThru, i16 %mask) {
; CHECK-LABEL: test_mask_packus_epi32_rrk_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpackusdw %ymm1, %ymm0, %ymm2 {%k1} ## encoding: [0x62,0xf2,0x7d,0x29,0x2b,0xd1]
; CHECK-NEXT:    vmovaps %zmm2, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.mask.packusdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> %passThru, i16 %mask)
  ret <16 x i16> %res
}

define <16 x i16> @test_mask_packus_epi32_rrkz_256(<8 x i32> %a, <8 x i32> %b, i16 %mask) {
; CHECK-LABEL: test_mask_packus_epi32_rrkz_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpackusdw %ymm1, %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf2,0x7d,0xa9,0x2b,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.mask.packusdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> zeroinitializer, i16 %mask)
  ret <16 x i16> %res
}

define <16 x i16> @test_mask_packus_epi32_rm_256(<8 x i32> %a, <8 x i32>* %ptr_b) {
; CHECK-LABEL: test_mask_packus_epi32_rm_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpackusdw (%rdi), %ymm0, %ymm0 ## encoding: [0x62,0xf2,0x7d,0x28,0x2b,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <8 x i32>, <8 x i32>* %ptr_b
  %res = call <16 x i16> @llvm.x86.avx512.mask.packusdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> zeroinitializer, i16 -1)
  ret <16 x i16> %res
}

define <16 x i16> @test_mask_packus_epi32_rmk_256(<8 x i32> %a, <8 x i32>* %ptr_b, <16 x i16> %passThru, i16 %mask) {
; CHECK-LABEL: test_mask_packus_epi32_rmk_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpackusdw (%rdi), %ymm0, %ymm1 {%k1} ## encoding: [0x62,0xf2,0x7d,0x29,0x2b,0x0f]
; CHECK-NEXT:    vmovaps %zmm1, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <8 x i32>, <8 x i32>* %ptr_b
  %res = call <16 x i16> @llvm.x86.avx512.mask.packusdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> %passThru, i16 %mask)
  ret <16 x i16> %res
}

define <16 x i16> @test_mask_packus_epi32_rmkz_256(<8 x i32> %a, <8 x i32>* %ptr_b, i16 %mask) {
; CHECK-LABEL: test_mask_packus_epi32_rmkz_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpackusdw (%rdi), %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf2,0x7d,0xa9,0x2b,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <8 x i32>, <8 x i32>* %ptr_b
  %res = call <16 x i16> @llvm.x86.avx512.mask.packusdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> zeroinitializer, i16 %mask)
  ret <16 x i16> %res
}

define <16 x i16> @test_mask_packus_epi32_rmb_256(<8 x i32> %a, i32* %ptr_b) {
; CHECK-LABEL: test_mask_packus_epi32_rmb_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpackusdw (%rdi){1to8}, %ymm0, %ymm0 ## encoding: [0x62,0xf2,0x7d,0x38,0x2b,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %q = load i32, i32* %ptr_b
  %vecinit.i = insertelement <8 x i32> undef, i32 %q, i32 0
  %b = shufflevector <8 x i32> %vecinit.i, <8 x i32> undef, <8 x i32> zeroinitializer
  %res = call <16 x i16> @llvm.x86.avx512.mask.packusdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> zeroinitializer, i16 -1)
  ret <16 x i16> %res
}

define <16 x i16> @test_mask_packus_epi32_rmbk_256(<8 x i32> %a, i32* %ptr_b, <16 x i16> %passThru, i16 %mask) {
; CHECK-LABEL: test_mask_packus_epi32_rmbk_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpackusdw (%rdi){1to8}, %ymm0, %ymm1 {%k1} ## encoding: [0x62,0xf2,0x7d,0x39,0x2b,0x0f]
; CHECK-NEXT:    vmovaps %zmm1, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %q = load i32, i32* %ptr_b
  %vecinit.i = insertelement <8 x i32> undef, i32 %q, i32 0
  %b = shufflevector <8 x i32> %vecinit.i, <8 x i32> undef, <8 x i32> zeroinitializer
  %res = call <16 x i16> @llvm.x86.avx512.mask.packusdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> %passThru, i16 %mask)
  ret <16 x i16> %res
}

define <16 x i16> @test_mask_packus_epi32_rmbkz_256(<8 x i32> %a, i32* %ptr_b, i16 %mask) {
; CHECK-LABEL: test_mask_packus_epi32_rmbkz_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpackusdw (%rdi){1to8}, %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf2,0x7d,0xb9,0x2b,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %q = load i32, i32* %ptr_b
  %vecinit.i = insertelement <8 x i32> undef, i32 %q, i32 0
  %b = shufflevector <8 x i32> %vecinit.i, <8 x i32> undef, <8 x i32> zeroinitializer
  %res = call <16 x i16> @llvm.x86.avx512.mask.packusdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> zeroinitializer, i16 %mask)
  ret <16 x i16> %res
}

declare <16 x i16> @llvm.x86.avx512.mask.packusdw.256(<8 x i32>, <8 x i32>, <16 x i16>, i16)

define <16 x i8> @test_mask_packus_epi16_rr_128(<8 x i16> %a, <8 x i16> %b) {
; CHECK-LABEL: test_mask_packus_epi16_rr_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0x67,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i8> @llvm.x86.avx512.mask.packuswb.128(<8 x i16> %a, <8 x i16> %b, <16 x i8> zeroinitializer, i16 -1)
  ret <16 x i8> %res
}

define <16 x i8> @test_mask_packus_epi16_rrk_128(<8 x i16> %a, <8 x i16> %b, <16 x i8> %passThru, i16 %mask) {
; CHECK-LABEL: test_mask_packus_epi16_rrk_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpackuswb %xmm1, %xmm0, %xmm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0x67,0xd1]
; CHECK-NEXT:    vmovaps %zmm2, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i8> @llvm.x86.avx512.mask.packuswb.128(<8 x i16> %a, <8 x i16> %b, <16 x i8> %passThru, i16 %mask)
  ret <16 x i8> %res
}

define <16 x i8> @test_mask_packus_epi16_rrkz_128(<8 x i16> %a, <8 x i16> %b, i16 %mask) {
; CHECK-LABEL: test_mask_packus_epi16_rrkz_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0x89,0x67,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i8> @llvm.x86.avx512.mask.packuswb.128(<8 x i16> %a, <8 x i16> %b, <16 x i8> zeroinitializer, i16 %mask)
  ret <16 x i8> %res
}

define <16 x i8> @test_mask_packus_epi16_rm_128(<8 x i16> %a, <8 x i16>* %ptr_b) {
; CHECK-LABEL: test_mask_packus_epi16_rm_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpackuswb (%rdi), %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0x67,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <8 x i16>, <8 x i16>* %ptr_b
  %res = call <16 x i8> @llvm.x86.avx512.mask.packuswb.128(<8 x i16> %a, <8 x i16> %b, <16 x i8> zeroinitializer, i16 -1)
  ret <16 x i8> %res
}

define <16 x i8> @test_mask_packus_epi16_rmk_128(<8 x i16> %a, <8 x i16>* %ptr_b, <16 x i8> %passThru, i16 %mask) {
; CHECK-LABEL: test_mask_packus_epi16_rmk_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpackuswb (%rdi), %xmm0, %xmm1 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0x67,0x0f]
; CHECK-NEXT:    vmovaps %zmm1, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <8 x i16>, <8 x i16>* %ptr_b
  %res = call <16 x i8> @llvm.x86.avx512.mask.packuswb.128(<8 x i16> %a, <8 x i16> %b, <16 x i8> %passThru, i16 %mask)
  ret <16 x i8> %res
}

define <16 x i8> @test_mask_packus_epi16_rmkz_128(<8 x i16> %a, <8 x i16>* %ptr_b, i16 %mask) {
; CHECK-LABEL: test_mask_packus_epi16_rmkz_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpackuswb (%rdi), %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0x89,0x67,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <8 x i16>, <8 x i16>* %ptr_b
  %res = call <16 x i8> @llvm.x86.avx512.mask.packuswb.128(<8 x i16> %a, <8 x i16> %b, <16 x i8> zeroinitializer, i16 %mask)
  ret <16 x i8> %res
}

declare <16 x i8> @llvm.x86.avx512.mask.packuswb.128(<8 x i16>, <8 x i16>, <16 x i8>, i16)

define <32 x i8> @test_mask_packus_epi16_rr_256(<16 x i16> %a, <16 x i16> %b) {
; CHECK-LABEL: test_mask_packus_epi16_rr_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpackuswb %ymm1, %ymm0, %ymm0 ## encoding: [0x62,0xf1,0x7d,0x28,0x67,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <32 x i8> @llvm.x86.avx512.mask.packuswb.256(<16 x i16> %a, <16 x i16> %b, <32 x i8> zeroinitializer, i32 -1)
  ret <32 x i8> %res
}

define <32 x i8> @test_mask_packus_epi16_rrk_256(<16 x i16> %a, <16 x i16> %b, <32 x i8> %passThru, i32 %mask) {
; CHECK-LABEL: test_mask_packus_epi16_rrk_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
; CHECK-NEXT:    vpackuswb %ymm1, %ymm0, %ymm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0x67,0xd1]
; CHECK-NEXT:    vmovaps %zmm2, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <32 x i8> @llvm.x86.avx512.mask.packuswb.256(<16 x i16> %a, <16 x i16> %b, <32 x i8> %passThru, i32 %mask)
  ret <32 x i8> %res
}

define <32 x i8> @test_mask_packus_epi16_rrkz_256(<16 x i16> %a, <16 x i16> %b, i32 %mask) {
; CHECK-LABEL: test_mask_packus_epi16_rrkz_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
; CHECK-NEXT:    vpackuswb %ymm1, %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xa9,0x67,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <32 x i8> @llvm.x86.avx512.mask.packuswb.256(<16 x i16> %a, <16 x i16> %b, <32 x i8> zeroinitializer, i32 %mask)
  ret <32 x i8> %res
}

define <32 x i8> @test_mask_packus_epi16_rm_256(<16 x i16> %a, <16 x i16>* %ptr_b) {
; CHECK-LABEL: test_mask_packus_epi16_rm_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpackuswb (%rdi), %ymm0, %ymm0 ## encoding: [0x62,0xf1,0x7d,0x28,0x67,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <16 x i16>, <16 x i16>* %ptr_b
  %res = call <32 x i8> @llvm.x86.avx512.mask.packuswb.256(<16 x i16> %a, <16 x i16> %b, <32 x i8> zeroinitializer, i32 -1)
  ret <32 x i8> %res
}

define <32 x i8> @test_mask_packus_epi16_rmk_256(<16 x i16> %a, <16 x i16>* %ptr_b, <32 x i8> %passThru, i32 %mask) {
; CHECK-LABEL: test_mask_packus_epi16_rmk_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovd %esi, %k1 ## encoding: [0xc5,0xfb,0x92,0xce]
; CHECK-NEXT:    vpackuswb (%rdi), %ymm0, %ymm1 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0x67,0x0f]
; CHECK-NEXT:    vmovaps %zmm1, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <16 x i16>, <16 x i16>* %ptr_b
  %res = call <32 x i8> @llvm.x86.avx512.mask.packuswb.256(<16 x i16> %a, <16 x i16> %b, <32 x i8> %passThru, i32 %mask)
  ret <32 x i8> %res
}

define <32 x i8> @test_mask_packus_epi16_rmkz_256(<16 x i16> %a, <16 x i16>* %ptr_b, i32 %mask) {
; CHECK-LABEL: test_mask_packus_epi16_rmkz_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovd %esi, %k1 ## encoding: [0xc5,0xfb,0x92,0xce]
; CHECK-NEXT:    vpackuswb (%rdi), %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xa9,0x67,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <16 x i16>, <16 x i16>* %ptr_b
  %res = call <32 x i8> @llvm.x86.avx512.mask.packuswb.256(<16 x i16> %a, <16 x i16> %b, <32 x i8> zeroinitializer, i32 %mask)
  ret <32 x i8> %res
}

declare <32 x i8> @llvm.x86.avx512.mask.packuswb.256(<16 x i16>, <16 x i16>, <32 x i8>, i32)

define <8 x i16> @test_mask_adds_epi16_rr_128(<8 x i16> %a, <8 x i16> %b) {
; CHECK-LABEL: test_mask_adds_epi16_rr_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpaddsw %xmm1, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xed,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.mask.padds.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> zeroinitializer, i8 -1)
  ret <8 x i16> %res
}

define <8 x i16> @test_mask_adds_epi16_rrk_128(<8 x i16> %a, <8 x i16> %b, <8 x i16> %passThru, i8 %mask) {
; CHECK-LABEL: test_mask_adds_epi16_rrk_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpaddsw %xmm1, %xmm0, %xmm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0xed,0xd1]
; CHECK-NEXT:    vmovaps %zmm2, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.mask.padds.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> %passThru, i8 %mask)
  ret <8 x i16> %res
}

define <8 x i16> @test_mask_adds_epi16_rrkz_128(<8 x i16> %a, <8 x i16> %b, i8 %mask) {
; CHECK-LABEL: test_mask_adds_epi16_rrkz_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpaddsw %xmm1, %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0x89,0xed,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.mask.padds.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> zeroinitializer, i8 %mask)
  ret <8 x i16> %res
}

define <8 x i16> @test_mask_adds_epi16_rm_128(<8 x i16> %a, <8 x i16>* %ptr_b) {
; CHECK-LABEL: test_mask_adds_epi16_rm_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpaddsw (%rdi), %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xed,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <8 x i16>, <8 x i16>* %ptr_b
  %res = call <8 x i16> @llvm.x86.avx512.mask.padds.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> zeroinitializer, i8 -1)
  ret <8 x i16> %res
}

define <8 x i16> @test_mask_adds_epi16_rmk_128(<8 x i16> %a, <8 x i16>* %ptr_b, <8 x i16> %passThru, i8 %mask) {
; CHECK-LABEL: test_mask_adds_epi16_rmk_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpaddsw (%rdi), %xmm0, %xmm1 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0xed,0x0f]
; CHECK-NEXT:    vmovaps %zmm1, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <8 x i16>, <8 x i16>* %ptr_b
  %res = call <8 x i16> @llvm.x86.avx512.mask.padds.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> %passThru, i8 %mask)
  ret <8 x i16> %res
}

define <8 x i16> @test_mask_adds_epi16_rmkz_128(<8 x i16> %a, <8 x i16>* %ptr_b, i8 %mask) {
; CHECK-LABEL: test_mask_adds_epi16_rmkz_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpaddsw (%rdi), %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0x89,0xed,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <8 x i16>, <8 x i16>* %ptr_b
  %res = call <8 x i16> @llvm.x86.avx512.mask.padds.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> zeroinitializer, i8 %mask)
  ret <8 x i16> %res
}

declare <8 x i16> @llvm.x86.avx512.mask.padds.w.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)

define <16 x i16> @test_mask_adds_epi16_rr_256(<16 x i16> %a, <16 x i16> %b) {
; CHECK-LABEL: test_mask_adds_epi16_rr_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpaddsw %ymm1, %ymm0, %ymm0 ## encoding: [0x62,0xf1,0x7d,0x28,0xed,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.mask.padds.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> zeroinitializer, i16 -1)
  ret <16 x i16> %res
}

define <16 x i16> @test_mask_adds_epi16_rrk_256(<16 x i16> %a, <16 x i16> %b, <16 x i16> %passThru, i16 %mask) {
; CHECK-LABEL: test_mask_adds_epi16_rrk_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpaddsw %ymm1, %ymm0, %ymm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0xed,0xd1]
; CHECK-NEXT:    vmovaps %zmm2, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.mask.padds.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> %passThru, i16 %mask)
  ret <16 x i16> %res
}

define <16 x i16> @test_mask_adds_epi16_rrkz_256(<16 x i16> %a, <16 x i16> %b, i16 %mask) {
; CHECK-LABEL: test_mask_adds_epi16_rrkz_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpaddsw %ymm1, %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xa9,0xed,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.mask.padds.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> zeroinitializer, i16 %mask)
  ret <16 x i16> %res
}

define <16 x i16> @test_mask_adds_epi16_rm_256(<16 x i16> %a, <16 x i16>* %ptr_b) {
; CHECK-LABEL: test_mask_adds_epi16_rm_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpaddsw (%rdi), %ymm0, %ymm0 ## encoding: [0x62,0xf1,0x7d,0x28,0xed,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <16 x i16>, <16 x i16>* %ptr_b
  %res = call <16 x i16> @llvm.x86.avx512.mask.padds.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> zeroinitializer, i16 -1)
  ret <16 x i16> %res
}

define <16 x i16> @test_mask_adds_epi16_rmk_256(<16 x i16> %a, <16 x i16>* %ptr_b, <16 x i16> %passThru, i16 %mask) {
; CHECK-LABEL: test_mask_adds_epi16_rmk_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpaddsw (%rdi), %ymm0, %ymm1 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0xed,0x0f]
; CHECK-NEXT:    vmovaps %zmm1, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <16 x i16>, <16 x i16>* %ptr_b
  %res = call <16 x i16> @llvm.x86.avx512.mask.padds.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> %passThru, i16 %mask)
  ret <16 x i16> %res
}

define <16 x i16> @test_mask_adds_epi16_rmkz_256(<16 x i16> %a, <16 x i16>* %ptr_b, i16 %mask) {
; CHECK-LABEL: test_mask_adds_epi16_rmkz_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpaddsw (%rdi), %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xa9,0xed,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <16 x i16>, <16 x i16>* %ptr_b
  %res = call <16 x i16> @llvm.x86.avx512.mask.padds.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> zeroinitializer, i16 %mask)
  ret <16 x i16> %res
}

declare <16 x i16> @llvm.x86.avx512.mask.padds.w.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)

define <8 x i16> @test_mask_subs_epi16_rr_128(<8 x i16> %a, <8 x i16> %b) {
; CHECK-LABEL: test_mask_subs_epi16_rr_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpsubsw %xmm1, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xe9,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.mask.psubs.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> zeroinitializer, i8 -1)
  ret <8 x i16> %res
}

define <8 x i16> @test_mask_subs_epi16_rrk_128(<8 x i16> %a, <8 x i16> %b, <8 x i16> %passThru, i8 %mask) {
; CHECK-LABEL: test_mask_subs_epi16_rrk_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpsubsw %xmm1, %xmm0, %xmm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0xe9,0xd1]
; CHECK-NEXT:    vmovaps %zmm2, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.mask.psubs.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> %passThru, i8 %mask)
  ret <8 x i16> %res
}

define <8 x i16> @test_mask_subs_epi16_rrkz_128(<8 x i16> %a, <8 x i16> %b, i8 %mask) {
; CHECK-LABEL: test_mask_subs_epi16_rrkz_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpsubsw %xmm1, %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0x89,0xe9,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.mask.psubs.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> zeroinitializer, i8 %mask)
  ret <8 x i16> %res
}

define <8 x i16> @test_mask_subs_epi16_rm_128(<8 x i16> %a, <8 x i16>* %ptr_b) {
; CHECK-LABEL: test_mask_subs_epi16_rm_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpsubsw (%rdi), %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xe9,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <8 x i16>, <8 x i16>* %ptr_b
  %res = call <8 x i16> @llvm.x86.avx512.mask.psubs.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> zeroinitializer, i8 -1)
  ret <8 x i16> %res
}

define <8 x i16> @test_mask_subs_epi16_rmk_128(<8 x i16> %a, <8 x i16>* %ptr_b, <8 x i16> %passThru, i8 %mask) {
; CHECK-LABEL: test_mask_subs_epi16_rmk_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpsubsw (%rdi), %xmm0, %xmm1 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0xe9,0x0f]
; CHECK-NEXT:    vmovaps %zmm1, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <8 x i16>, <8 x i16>* %ptr_b
  %res = call <8 x i16> @llvm.x86.avx512.mask.psubs.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> %passThru, i8 %mask)
  ret <8 x i16> %res
}

define <8 x i16> @test_mask_subs_epi16_rmkz_128(<8 x i16> %a, <8 x i16>* %ptr_b, i8 %mask) {
; CHECK-LABEL: test_mask_subs_epi16_rmkz_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpsubsw (%rdi), %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0x89,0xe9,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <8 x i16>, <8 x i16>* %ptr_b
  %res = call <8 x i16> @llvm.x86.avx512.mask.psubs.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> zeroinitializer, i8 %mask)
  ret <8 x i16> %res
}

declare <8 x i16> @llvm.x86.avx512.mask.psubs.w.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)

define <16 x i16> @test_mask_subs_epi16_rr_256(<16 x i16> %a, <16 x i16> %b) {
; CHECK-LABEL: test_mask_subs_epi16_rr_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpsubsw %ymm1, %ymm0, %ymm0 ## encoding: [0x62,0xf1,0x7d,0x28,0xe9,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.mask.psubs.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> zeroinitializer, i16 -1)
  ret <16 x i16> %res
}

define <16 x i16> @test_mask_subs_epi16_rrk_256(<16 x i16> %a, <16 x i16> %b, <16 x i16> %passThru, i16 %mask) {
; CHECK-LABEL: test_mask_subs_epi16_rrk_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpsubsw %ymm1, %ymm0, %ymm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0xe9,0xd1]
; CHECK-NEXT:    vmovaps %zmm2, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.mask.psubs.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> %passThru, i16 %mask)
  ret <16 x i16> %res
}

define <16 x i16> @test_mask_subs_epi16_rrkz_256(<16 x i16> %a, <16 x i16> %b, i16 %mask) {
; CHECK-LABEL: test_mask_subs_epi16_rrkz_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpsubsw %ymm1, %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xa9,0xe9,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.mask.psubs.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> zeroinitializer, i16 %mask)
  ret <16 x i16> %res
}

define <16 x i16> @test_mask_subs_epi16_rm_256(<16 x i16> %a, <16 x i16>* %ptr_b) {
; CHECK-LABEL: test_mask_subs_epi16_rm_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpsubsw (%rdi), %ymm0, %ymm0 ## encoding: [0x62,0xf1,0x7d,0x28,0xe9,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <16 x i16>, <16 x i16>* %ptr_b
  %res = call <16 x i16> @llvm.x86.avx512.mask.psubs.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> zeroinitializer, i16 -1)
  ret <16 x i16> %res
}

define <16 x i16> @test_mask_subs_epi16_rmk_256(<16 x i16> %a, <16 x i16>* %ptr_b, <16 x i16> %passThru, i16 %mask) {
; CHECK-LABEL: test_mask_subs_epi16_rmk_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpsubsw (%rdi), %ymm0, %ymm1 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0xe9,0x0f]
; CHECK-NEXT:    vmovaps %zmm1, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <16 x i16>, <16 x i16>* %ptr_b
  %res = call <16 x i16> @llvm.x86.avx512.mask.psubs.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> %passThru, i16 %mask)
  ret <16 x i16> %res
}

define <16 x i16> @test_mask_subs_epi16_rmkz_256(<16 x i16> %a, <16 x i16>* %ptr_b, i16 %mask) {
; CHECK-LABEL: test_mask_subs_epi16_rmkz_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpsubsw (%rdi), %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xa9,0xe9,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <16 x i16>, <16 x i16>* %ptr_b
  %res = call <16 x i16> @llvm.x86.avx512.mask.psubs.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> zeroinitializer, i16 %mask)
  ret <16 x i16> %res
}

declare <16 x i16> @llvm.x86.avx512.mask.psubs.w.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)

define <8 x i16> @test_mask_adds_epu16_rr_128(<8 x i16> %a, <8 x i16> %b) {
; CHECK-LABEL: test_mask_adds_epu16_rr_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpaddusw %xmm1, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xdd,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.mask.paddus.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> zeroinitializer, i8 -1)
  ret <8 x i16> %res
}

define <8 x i16> @test_mask_adds_epu16_rrk_128(<8 x i16> %a, <8 x i16> %b, <8 x i16> %passThru, i8 %mask) {
; CHECK-LABEL: test_mask_adds_epu16_rrk_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpaddusw %xmm1, %xmm0, %xmm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0xdd,0xd1]
; CHECK-NEXT:    vmovaps %zmm2, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.mask.paddus.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> %passThru, i8 %mask)
  ret <8 x i16> %res
}

define <8 x i16> @test_mask_adds_epu16_rrkz_128(<8 x i16> %a, <8 x i16> %b, i8 %mask) {
; CHECK-LABEL: test_mask_adds_epu16_rrkz_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpaddusw %xmm1, %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0x89,0xdd,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.mask.paddus.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> zeroinitializer, i8 %mask)
  ret <8 x i16> %res
}

define <8 x i16> @test_mask_adds_epu16_rm_128(<8 x i16> %a, <8 x i16>* %ptr_b) {
; CHECK-LABEL: test_mask_adds_epu16_rm_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpaddusw (%rdi), %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xdd,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <8 x i16>, <8 x i16>* %ptr_b
  %res = call <8 x i16> @llvm.x86.avx512.mask.paddus.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> zeroinitializer, i8 -1)
  ret <8 x i16> %res
}

define <8 x i16> @test_mask_adds_epu16_rmk_128(<8 x i16> %a, <8 x i16>* %ptr_b, <8 x i16> %passThru, i8 %mask) {
; CHECK-LABEL: test_mask_adds_epu16_rmk_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpaddusw (%rdi), %xmm0, %xmm1 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0xdd,0x0f]
; CHECK-NEXT:    vmovaps %zmm1, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <8 x i16>, <8 x i16>* %ptr_b
  %res = call <8 x i16> @llvm.x86.avx512.mask.paddus.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> %passThru, i8 %mask)
  ret <8 x i16> %res
}

define <8 x i16> @test_mask_adds_epu16_rmkz_128(<8 x i16> %a, <8 x i16>* %ptr_b, i8 %mask) {
; CHECK-LABEL: test_mask_adds_epu16_rmkz_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpaddusw (%rdi), %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0x89,0xdd,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <8 x i16>, <8 x i16>* %ptr_b
  %res = call <8 x i16> @llvm.x86.avx512.mask.paddus.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> zeroinitializer, i8 %mask)
  ret <8 x i16> %res
}

declare <8 x i16> @llvm.x86.avx512.mask.paddus.w.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)

define <16 x i16> @test_mask_adds_epu16_rr_256(<16 x i16> %a, <16 x i16> %b) {
; CHECK-LABEL: test_mask_adds_epu16_rr_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpaddusw %ymm1, %ymm0, %ymm0 ## encoding: [0x62,0xf1,0x7d,0x28,0xdd,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.mask.paddus.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> zeroinitializer, i16 -1)
  ret <16 x i16> %res
}

define <16 x i16> @test_mask_adds_epu16_rrk_256(<16 x i16> %a, <16 x i16> %b, <16 x i16> %passThru, i16 %mask) {
; CHECK-LABEL: test_mask_adds_epu16_rrk_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpaddusw %ymm1, %ymm0, %ymm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0xdd,0xd1]
; CHECK-NEXT:    vmovaps %zmm2, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.mask.paddus.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> %passThru, i16 %mask)
  ret <16 x i16> %res
}

define <16 x i16> @test_mask_adds_epu16_rrkz_256(<16 x i16> %a, <16 x i16> %b, i16 %mask) {
; CHECK-LABEL: test_mask_adds_epu16_rrkz_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpaddusw %ymm1, %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xa9,0xdd,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.mask.paddus.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> zeroinitializer, i16 %mask)
  ret <16 x i16> %res
}

define <16 x i16> @test_mask_adds_epu16_rm_256(<16 x i16> %a, <16 x i16>* %ptr_b) {
; CHECK-LABEL: test_mask_adds_epu16_rm_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpaddusw (%rdi), %ymm0, %ymm0 ## encoding: [0x62,0xf1,0x7d,0x28,0xdd,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <16 x i16>, <16 x i16>* %ptr_b
  %res = call <16 x i16> @llvm.x86.avx512.mask.paddus.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> zeroinitializer, i16 -1)
  ret <16 x i16> %res
}

define <16 x i16> @test_mask_adds_epu16_rmk_256(<16 x i16> %a, <16 x i16>* %ptr_b, <16 x i16> %passThru, i16 %mask) {
; CHECK-LABEL: test_mask_adds_epu16_rmk_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpaddusw (%rdi), %ymm0, %ymm1 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0xdd,0x0f]
; CHECK-NEXT:    vmovaps %zmm1, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <16 x i16>, <16 x i16>* %ptr_b
  %res = call <16 x i16> @llvm.x86.avx512.mask.paddus.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> %passThru, i16 %mask)
  ret <16 x i16> %res
}

define <16 x i16> @test_mask_adds_epu16_rmkz_256(<16 x i16> %a, <16 x i16>* %ptr_b, i16 %mask) {
; CHECK-LABEL: test_mask_adds_epu16_rmkz_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpaddusw (%rdi), %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xa9,0xdd,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <16 x i16>, <16 x i16>* %ptr_b
  %res = call <16 x i16> @llvm.x86.avx512.mask.paddus.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> zeroinitializer, i16 %mask)
  ret <16 x i16> %res
}

declare <16 x i16> @llvm.x86.avx512.mask.paddus.w.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)

define <8 x i16> @test_mask_subs_epu16_rr_128(<8 x i16> %a, <8 x i16> %b) {
; CHECK-LABEL: test_mask_subs_epu16_rr_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpsubusw %xmm1, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xd9,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.mask.psubus.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> zeroinitializer, i8 -1)
  ret <8 x i16> %res
}

define <8 x i16> @test_mask_subs_epu16_rrk_128(<8 x i16> %a, <8 x i16> %b, <8 x i16> %passThru, i8 %mask) {
; CHECK-LABEL: test_mask_subs_epu16_rrk_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpsubusw %xmm1, %xmm0, %xmm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0xd9,0xd1]
; CHECK-NEXT:    vmovaps %zmm2, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.mask.psubus.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> %passThru, i8 %mask)
  ret <8 x i16> %res
}

define <8 x i16> @test_mask_subs_epu16_rrkz_128(<8 x i16> %a, <8 x i16> %b, i8 %mask) {
; CHECK-LABEL: test_mask_subs_epu16_rrkz_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpsubusw %xmm1, %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0x89,0xd9,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.mask.psubus.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> zeroinitializer, i8 %mask)
  ret <8 x i16> %res
}

define <8 x i16> @test_mask_subs_epu16_rm_128(<8 x i16> %a, <8 x i16>* %ptr_b) {
; CHECK-LABEL: test_mask_subs_epu16_rm_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpsubusw (%rdi), %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xd9,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <8 x i16>, <8 x i16>* %ptr_b
  %res = call <8 x i16> @llvm.x86.avx512.mask.psubus.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> zeroinitializer, i8 -1)
  ret <8 x i16> %res
}

define <8 x i16> @test_mask_subs_epu16_rmk_128(<8 x i16> %a, <8 x i16>* %ptr_b, <8 x i16> %passThru, i8 %mask) {
; CHECK-LABEL: test_mask_subs_epu16_rmk_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpsubusw (%rdi), %xmm0, %xmm1 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0xd9,0x0f]
; CHECK-NEXT:    vmovaps %zmm1, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <8 x i16>, <8 x i16>* %ptr_b
  %res = call <8 x i16> @llvm.x86.avx512.mask.psubus.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> %passThru, i8 %mask)
  ret <8 x i16> %res
}

define <8 x i16> @test_mask_subs_epu16_rmkz_128(<8 x i16> %a, <8 x i16>* %ptr_b, i8 %mask) {
; CHECK-LABEL: test_mask_subs_epu16_rmkz_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpsubusw (%rdi), %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0x89,0xd9,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <8 x i16>, <8 x i16>* %ptr_b
  %res = call <8 x i16> @llvm.x86.avx512.mask.psubus.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> zeroinitializer, i8 %mask)
  ret <8 x i16> %res
}

declare <8 x i16> @llvm.x86.avx512.mask.psubus.w.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)

define <16 x i16> @test_mask_subs_epu16_rr_256(<16 x i16> %a, <16 x i16> %b) {
; CHECK-LABEL: test_mask_subs_epu16_rr_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpsubusw %ymm1, %ymm0, %ymm0 ## encoding: [0x62,0xf1,0x7d,0x28,0xd9,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.mask.psubus.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> zeroinitializer, i16 -1)
  ret <16 x i16> %res
}

define <16 x i16> @test_mask_subs_epu16_rrk_256(<16 x i16> %a, <16 x i16> %b, <16 x i16> %passThru, i16 %mask) {
; CHECK-LABEL: test_mask_subs_epu16_rrk_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpsubusw %ymm1, %ymm0, %ymm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0xd9,0xd1]
; CHECK-NEXT:    vmovaps %zmm2, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.mask.psubus.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> %passThru, i16 %mask)
  ret <16 x i16> %res
}

define <16 x i16> @test_mask_subs_epu16_rrkz_256(<16 x i16> %a, <16 x i16> %b, i16 %mask) {
; CHECK-LABEL: test_mask_subs_epu16_rrkz_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpsubusw %ymm1, %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xa9,0xd9,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.mask.psubus.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> zeroinitializer, i16 %mask)
  ret <16 x i16> %res
}

define <16 x i16> @test_mask_subs_epu16_rm_256(<16 x i16> %a, <16 x i16>* %ptr_b) {
; CHECK-LABEL: test_mask_subs_epu16_rm_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpsubusw (%rdi), %ymm0, %ymm0 ## encoding: [0x62,0xf1,0x7d,0x28,0xd9,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <16 x i16>, <16 x i16>* %ptr_b
  %res = call <16 x i16> @llvm.x86.avx512.mask.psubus.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> zeroinitializer, i16 -1)
  ret <16 x i16> %res
}

define <16 x i16> @test_mask_subs_epu16_rmk_256(<16 x i16> %a, <16 x i16>* %ptr_b, <16 x i16> %passThru, i16 %mask) {
; CHECK-LABEL: test_mask_subs_epu16_rmk_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpsubusw (%rdi), %ymm0, %ymm1 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0xd9,0x0f]
; CHECK-NEXT:    vmovaps %zmm1, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <16 x i16>, <16 x i16>* %ptr_b
  %res = call <16 x i16> @llvm.x86.avx512.mask.psubus.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> %passThru, i16 %mask)
  ret <16 x i16> %res
}

define <16 x i16> @test_mask_subs_epu16_rmkz_256(<16 x i16> %a, <16 x i16>* %ptr_b, i16 %mask) {
; CHECK-LABEL: test_mask_subs_epu16_rmkz_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpsubusw (%rdi), %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xa9,0xd9,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <16 x i16>, <16 x i16>* %ptr_b
  %res = call <16 x i16> @llvm.x86.avx512.mask.psubus.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> zeroinitializer, i16 %mask)
  ret <16 x i16> %res
}

declare <16 x i16> @llvm.x86.avx512.mask.psubus.w.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)

define <16 x i8> @test_mask_adds_epi8_rr_128(<16 x i8> %a, <16 x i8> %b) {
; CHECK-LABEL: test_mask_adds_epi8_rr_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpaddsb %xmm1, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xec,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i8> @llvm.x86.avx512.mask.padds.b.128(<16 x i8> %a, <16 x i8> %b, <16 x i8> zeroinitializer, i16 -1)
  ret <16 x i8> %res
}

define <16 x i8> @test_mask_adds_epi8_rrk_128(<16 x i8> %a, <16 x i8> %b, <16 x i8> %passThru, i16 %mask) {
; CHECK-LABEL: test_mask_adds_epi8_rrk_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpaddsb %xmm1, %xmm0, %xmm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0xec,0xd1]
; CHECK-NEXT:    vmovaps %zmm2, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i8> @llvm.x86.avx512.mask.padds.b.128(<16 x i8> %a, <16 x i8> %b, <16 x i8> %passThru, i16 %mask)
  ret <16 x i8> %res
}

define <16 x i8> @test_mask_adds_epi8_rrkz_128(<16 x i8> %a, <16 x i8> %b, i16 %mask) {
; CHECK-LABEL: test_mask_adds_epi8_rrkz_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpaddsb %xmm1, %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0x89,0xec,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i8> @llvm.x86.avx512.mask.padds.b.128(<16 x i8> %a, <16 x i8> %b, <16 x i8> zeroinitializer, i16 %mask)
  ret <16 x i8> %res
}

define <16 x i8> @test_mask_adds_epi8_rm_128(<16 x i8> %a, <16 x i8>* %ptr_b) {
; CHECK-LABEL: test_mask_adds_epi8_rm_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpaddsb (%rdi), %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xec,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <16 x i8>, <16 x i8>* %ptr_b
  %res = call <16 x i8> @llvm.x86.avx512.mask.padds.b.128(<16 x i8> %a, <16 x i8> %b, <16 x i8> zeroinitializer, i16 -1)
  ret <16 x i8> %res
}

define <16 x i8> @test_mask_adds_epi8_rmk_128(<16 x i8> %a, <16 x i8>* %ptr_b, <16 x i8> %passThru, i16 %mask) {
; CHECK-LABEL: test_mask_adds_epi8_rmk_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpaddsb (%rdi), %xmm0, %xmm1 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0xec,0x0f]
; CHECK-NEXT:    vmovaps %zmm1, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <16 x i8>, <16 x i8>* %ptr_b
  %res = call <16 x i8> @llvm.x86.avx512.mask.padds.b.128(<16 x i8> %a, <16 x i8> %b, <16 x i8> %passThru, i16 %mask)
  ret <16 x i8> %res
}

define <16 x i8> @test_mask_adds_epi8_rmkz_128(<16 x i8> %a, <16 x i8>* %ptr_b, i16 %mask) {
; CHECK-LABEL: test_mask_adds_epi8_rmkz_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpaddsb (%rdi), %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0x89,0xec,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <16 x i8>, <16 x i8>* %ptr_b
  %res = call <16 x i8> @llvm.x86.avx512.mask.padds.b.128(<16 x i8> %a, <16 x i8> %b, <16 x i8> zeroinitializer, i16 %mask)
  ret <16 x i8> %res
}

declare <16 x i8> @llvm.x86.avx512.mask.padds.b.128(<16 x i8>, <16 x i8>, <16 x i8>, i16)

define <32 x i8> @test_mask_adds_epi8_rr_256(<32 x i8> %a, <32 x i8> %b) {
; CHECK-LABEL: test_mask_adds_epi8_rr_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpaddsb %ymm1, %ymm0, %ymm0 ## encoding: [0x62,0xf1,0x7d,0x28,0xec,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <32 x i8> @llvm.x86.avx512.mask.padds.b.256(<32 x i8> %a, <32 x i8> %b, <32 x i8> zeroinitializer, i32 -1)
  ret <32 x i8> %res
}

define <32 x i8> @test_mask_adds_epi8_rrk_256(<32 x i8> %a, <32 x i8> %b, <32 x i8> %passThru, i32 %mask) {
; CHECK-LABEL: test_mask_adds_epi8_rrk_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
; CHECK-NEXT:    vpaddsb %ymm1, %ymm0, %ymm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0xec,0xd1]
; CHECK-NEXT:    vmovaps %zmm2, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <32 x i8> @llvm.x86.avx512.mask.padds.b.256(<32 x i8> %a, <32 x i8> %b, <32 x i8> %passThru, i32 %mask)
  ret <32 x i8> %res
}

define <32 x i8> @test_mask_adds_epi8_rrkz_256(<32 x i8> %a, <32 x i8> %b, i32 %mask) {
; CHECK-LABEL: test_mask_adds_epi8_rrkz_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
; CHECK-NEXT:    vpaddsb %ymm1, %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xa9,0xec,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <32 x i8> @llvm.x86.avx512.mask.padds.b.256(<32 x i8> %a, <32 x i8> %b, <32 x i8> zeroinitializer, i32 %mask)
  ret <32 x i8> %res
}

define <32 x i8> @test_mask_adds_epi8_rm_256(<32 x i8> %a, <32 x i8>* %ptr_b) {
; CHECK-LABEL: test_mask_adds_epi8_rm_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpaddsb (%rdi), %ymm0, %ymm0 ## encoding: [0x62,0xf1,0x7d,0x28,0xec,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <32 x i8>, <32 x i8>* %ptr_b
  %res = call <32 x i8> @llvm.x86.avx512.mask.padds.b.256(<32 x i8> %a, <32 x i8> %b, <32 x i8> zeroinitializer, i32 -1)
  ret <32 x i8> %res
}

define <32 x i8> @test_mask_adds_epi8_rmk_256(<32 x i8> %a, <32 x i8>* %ptr_b, <32 x i8> %passThru, i32 %mask) {
; CHECK-LABEL: test_mask_adds_epi8_rmk_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovd %esi, %k1 ## encoding: [0xc5,0xfb,0x92,0xce]
; CHECK-NEXT:    vpaddsb (%rdi), %ymm0, %ymm1 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0xec,0x0f]
; CHECK-NEXT:    vmovaps %zmm1, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <32 x i8>, <32 x i8>* %ptr_b
  %res = call <32 x i8> @llvm.x86.avx512.mask.padds.b.256(<32 x i8> %a, <32 x i8> %b, <32 x i8> %passThru, i32 %mask)
  ret <32 x i8> %res
}

define <32 x i8> @test_mask_adds_epi8_rmkz_256(<32 x i8> %a, <32 x i8>* %ptr_b, i32 %mask) {
; CHECK-LABEL: test_mask_adds_epi8_rmkz_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovd %esi, %k1 ## encoding: [0xc5,0xfb,0x92,0xce]
; CHECK-NEXT:    vpaddsb (%rdi), %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xa9,0xec,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <32 x i8>, <32 x i8>* %ptr_b
  %res = call <32 x i8> @llvm.x86.avx512.mask.padds.b.256(<32 x i8> %a, <32 x i8> %b, <32 x i8> zeroinitializer, i32 %mask)
  ret <32 x i8> %res
}

declare <32 x i8> @llvm.x86.avx512.mask.padds.b.256(<32 x i8>, <32 x i8>, <32 x i8>, i32)

define <16 x i8> @test_mask_subs_epi8_rr_128(<16 x i8> %a, <16 x i8> %b) {
; CHECK-LABEL: test_mask_subs_epi8_rr_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpsubsb %xmm1, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xe8,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i8> @llvm.x86.avx512.mask.psubs.b.128(<16 x i8> %a, <16 x i8> %b, <16 x i8> zeroinitializer, i16 -1)
  ret <16 x i8> %res
}

define <16 x i8> @test_mask_subs_epi8_rrk_128(<16 x i8> %a, <16 x i8> %b, <16 x i8> %passThru, i16 %mask) {
; CHECK-LABEL: test_mask_subs_epi8_rrk_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpsubsb %xmm1, %xmm0, %xmm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0xe8,0xd1]
; CHECK-NEXT:    vmovaps %zmm2, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i8> @llvm.x86.avx512.mask.psubs.b.128(<16 x i8> %a, <16 x i8> %b, <16 x i8> %passThru, i16 %mask)
  ret <16 x i8> %res
}

define <16 x i8> @test_mask_subs_epi8_rrkz_128(<16 x i8> %a, <16 x i8> %b, i16 %mask) {
; CHECK-LABEL: test_mask_subs_epi8_rrkz_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpsubsb %xmm1, %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0x89,0xe8,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i8> @llvm.x86.avx512.mask.psubs.b.128(<16 x i8> %a, <16 x i8> %b, <16 x i8> zeroinitializer, i16 %mask)
  ret <16 x i8> %res
}

define <16 x i8> @test_mask_subs_epi8_rm_128(<16 x i8> %a, <16 x i8>* %ptr_b) {
; CHECK-LABEL: test_mask_subs_epi8_rm_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpsubsb (%rdi), %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xe8,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <16 x i8>, <16 x i8>* %ptr_b
  %res = call <16 x i8> @llvm.x86.avx512.mask.psubs.b.128(<16 x i8> %a, <16 x i8> %b, <16 x i8> zeroinitializer, i16 -1)
  ret <16 x i8> %res
}

define <16 x i8> @test_mask_subs_epi8_rmk_128(<16 x i8> %a, <16 x i8>* %ptr_b, <16 x i8> %passThru, i16 %mask) {
; CHECK-LABEL: test_mask_subs_epi8_rmk_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpsubsb (%rdi), %xmm0, %xmm1 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0xe8,0x0f]
; CHECK-NEXT:    vmovaps %zmm1, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <16 x i8>, <16 x i8>* %ptr_b
  %res = call <16 x i8> @llvm.x86.avx512.mask.psubs.b.128(<16 x i8> %a, <16 x i8> %b, <16 x i8> %passThru, i16 %mask)
  ret <16 x i8> %res
}

define <16 x i8> @test_mask_subs_epi8_rmkz_128(<16 x i8> %a, <16 x i8>* %ptr_b, i16 %mask) {
; CHECK-LABEL: test_mask_subs_epi8_rmkz_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpsubsb (%rdi), %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0x89,0xe8,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <16 x i8>, <16 x i8>* %ptr_b
  %res = call <16 x i8> @llvm.x86.avx512.mask.psubs.b.128(<16 x i8> %a, <16 x i8> %b, <16 x i8> zeroinitializer, i16 %mask)
  ret <16 x i8> %res
}

declare <16 x i8> @llvm.x86.avx512.mask.psubs.b.128(<16 x i8>, <16 x i8>, <16 x i8>, i16)

define <32 x i8> @test_mask_subs_epi8_rr_256(<32 x i8> %a, <32 x i8> %b) {
; CHECK-LABEL: test_mask_subs_epi8_rr_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpsubsb %ymm1, %ymm0, %ymm0 ## encoding: [0x62,0xf1,0x7d,0x28,0xe8,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <32 x i8> @llvm.x86.avx512.mask.psubs.b.256(<32 x i8> %a, <32 x i8> %b, <32 x i8> zeroinitializer, i32 -1)
  ret <32 x i8> %res
}

define <32 x i8> @test_mask_subs_epi8_rrk_256(<32 x i8> %a, <32 x i8> %b, <32 x i8> %passThru, i32 %mask) {
; CHECK-LABEL: test_mask_subs_epi8_rrk_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
; CHECK-NEXT:    vpsubsb %ymm1, %ymm0, %ymm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0xe8,0xd1]
; CHECK-NEXT:    vmovaps %zmm2, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <32 x i8> @llvm.x86.avx512.mask.psubs.b.256(<32 x i8> %a, <32 x i8> %b, <32 x i8> %passThru, i32 %mask)
  ret <32 x i8> %res
}

define <32 x i8> @test_mask_subs_epi8_rrkz_256(<32 x i8> %a, <32 x i8> %b, i32 %mask) {
; CHECK-LABEL: test_mask_subs_epi8_rrkz_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
; CHECK-NEXT:    vpsubsb %ymm1, %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xa9,0xe8,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <32 x i8> @llvm.x86.avx512.mask.psubs.b.256(<32 x i8> %a, <32 x i8> %b, <32 x i8> zeroinitializer, i32 %mask)
  ret <32 x i8> %res
}

define <32 x i8> @test_mask_subs_epi8_rm_256(<32 x i8> %a, <32 x i8>* %ptr_b) {
; CHECK-LABEL: test_mask_subs_epi8_rm_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpsubsb (%rdi), %ymm0, %ymm0 ## encoding: [0x62,0xf1,0x7d,0x28,0xe8,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <32 x i8>, <32 x i8>* %ptr_b
  %res = call <32 x i8> @llvm.x86.avx512.mask.psubs.b.256(<32 x i8> %a, <32 x i8> %b, <32 x i8> zeroinitializer, i32 -1)
  ret <32 x i8> %res
}

define <32 x i8> @test_mask_subs_epi8_rmk_256(<32 x i8> %a, <32 x i8>* %ptr_b, <32 x i8> %passThru, i32 %mask) {
; CHECK-LABEL: test_mask_subs_epi8_rmk_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovd %esi, %k1 ## encoding: [0xc5,0xfb,0x92,0xce]
; CHECK-NEXT:    vpsubsb (%rdi), %ymm0, %ymm1 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0xe8,0x0f]
; CHECK-NEXT:    vmovaps %zmm1, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <32 x i8>, <32 x i8>* %ptr_b
  %res = call <32 x i8> @llvm.x86.avx512.mask.psubs.b.256(<32 x i8> %a, <32 x i8> %b, <32 x i8> %passThru, i32 %mask)
  ret <32 x i8> %res
}

define <32 x i8> @test_mask_subs_epi8_rmkz_256(<32 x i8> %a, <32 x i8>* %ptr_b, i32 %mask) {
; CHECK-LABEL: test_mask_subs_epi8_rmkz_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovd %esi, %k1 ## encoding: [0xc5,0xfb,0x92,0xce]
; CHECK-NEXT:    vpsubsb (%rdi), %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xa9,0xe8,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <32 x i8>, <32 x i8>* %ptr_b
  %res = call <32 x i8> @llvm.x86.avx512.mask.psubs.b.256(<32 x i8> %a, <32 x i8> %b, <32 x i8> zeroinitializer, i32 %mask)
  ret <32 x i8> %res
}

declare <32 x i8> @llvm.x86.avx512.mask.psubs.b.256(<32 x i8>, <32 x i8>, <32 x i8>, i32)

define <16 x i8> @test_mask_adds_epu8_rr_128(<16 x i8> %a, <16 x i8> %b) {
; CHECK-LABEL: test_mask_adds_epu8_rr_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpaddusb %xmm1, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xdc,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i8> @llvm.x86.avx512.mask.paddus.b.128(<16 x i8> %a, <16 x i8> %b, <16 x i8> zeroinitializer, i16 -1)
  ret <16 x i8> %res
}

define <16 x i8> @test_mask_adds_epu8_rrk_128(<16 x i8> %a, <16 x i8> %b, <16 x i8> %passThru, i16 %mask) {
; CHECK-LABEL: test_mask_adds_epu8_rrk_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpaddusb %xmm1, %xmm0, %xmm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0xdc,0xd1]
; CHECK-NEXT:    vmovaps %zmm2, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i8> @llvm.x86.avx512.mask.paddus.b.128(<16 x i8> %a, <16 x i8> %b, <16 x i8> %passThru, i16 %mask)
  ret <16 x i8> %res
}

define <16 x i8> @test_mask_adds_epu8_rrkz_128(<16 x i8> %a, <16 x i8> %b, i16 %mask) {
; CHECK-LABEL: test_mask_adds_epu8_rrkz_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpaddusb %xmm1, %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0x89,0xdc,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i8> @llvm.x86.avx512.mask.paddus.b.128(<16 x i8> %a, <16 x i8> %b, <16 x i8> zeroinitializer, i16 %mask)
  ret <16 x i8> %res
}

define <16 x i8> @test_mask_adds_epu8_rm_128(<16 x i8> %a, <16 x i8>* %ptr_b) {
; CHECK-LABEL: test_mask_adds_epu8_rm_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpaddusb (%rdi), %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xdc,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <16 x i8>, <16 x i8>* %ptr_b
  %res = call <16 x i8> @llvm.x86.avx512.mask.paddus.b.128(<16 x i8> %a, <16 x i8> %b, <16 x i8> zeroinitializer, i16 -1)
  ret <16 x i8> %res
}

define <16 x i8> @test_mask_adds_epu8_rmk_128(<16 x i8> %a, <16 x i8>* %ptr_b, <16 x i8> %passThru, i16 %mask) {
; CHECK-LABEL: test_mask_adds_epu8_rmk_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpaddusb (%rdi), %xmm0, %xmm1 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0xdc,0x0f]
; CHECK-NEXT:    vmovaps %zmm1, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <16 x i8>, <16 x i8>* %ptr_b
  %res = call <16 x i8> @llvm.x86.avx512.mask.paddus.b.128(<16 x i8> %a, <16 x i8> %b, <16 x i8> %passThru, i16 %mask)
  ret <16 x i8> %res
}

define <16 x i8> @test_mask_adds_epu8_rmkz_128(<16 x i8> %a, <16 x i8>* %ptr_b, i16 %mask) {
; CHECK-LABEL: test_mask_adds_epu8_rmkz_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpaddusb (%rdi), %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0x89,0xdc,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <16 x i8>, <16 x i8>* %ptr_b
  %res = call <16 x i8> @llvm.x86.avx512.mask.paddus.b.128(<16 x i8> %a, <16 x i8> %b, <16 x i8> zeroinitializer, i16 %mask)
  ret <16 x i8> %res
}

declare <16 x i8> @llvm.x86.avx512.mask.paddus.b.128(<16 x i8>, <16 x i8>, <16 x i8>, i16)

define <32 x i8> @test_mask_adds_epu8_rr_256(<32 x i8> %a, <32 x i8> %b) {
; CHECK-LABEL: test_mask_adds_epu8_rr_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpaddusb %ymm1, %ymm0, %ymm0 ## encoding: [0x62,0xf1,0x7d,0x28,0xdc,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <32 x i8> @llvm.x86.avx512.mask.paddus.b.256(<32 x i8> %a, <32 x i8> %b, <32 x i8> zeroinitializer, i32 -1)
  ret <32 x i8> %res
}

define <32 x i8> @test_mask_adds_epu8_rrk_256(<32 x i8> %a, <32 x i8> %b, <32 x i8> %passThru, i32 %mask) {
; CHECK-LABEL: test_mask_adds_epu8_rrk_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
; CHECK-NEXT:    vpaddusb %ymm1, %ymm0, %ymm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0xdc,0xd1]
; CHECK-NEXT:    vmovaps %zmm2, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <32 x i8> @llvm.x86.avx512.mask.paddus.b.256(<32 x i8> %a, <32 x i8> %b, <32 x i8> %passThru, i32 %mask)
  ret <32 x i8> %res
}

define <32 x i8> @test_mask_adds_epu8_rrkz_256(<32 x i8> %a, <32 x i8> %b, i32 %mask) {
; CHECK-LABEL: test_mask_adds_epu8_rrkz_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
; CHECK-NEXT:    vpaddusb %ymm1, %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xa9,0xdc,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <32 x i8> @llvm.x86.avx512.mask.paddus.b.256(<32 x i8> %a, <32 x i8> %b, <32 x i8> zeroinitializer, i32 %mask)
  ret <32 x i8> %res
}

define <32 x i8> @test_mask_adds_epu8_rm_256(<32 x i8> %a, <32 x i8>* %ptr_b) {
; CHECK-LABEL: test_mask_adds_epu8_rm_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpaddusb (%rdi), %ymm0, %ymm0 ## encoding: [0x62,0xf1,0x7d,0x28,0xdc,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <32 x i8>, <32 x i8>* %ptr_b
  %res = call <32 x i8> @llvm.x86.avx512.mask.paddus.b.256(<32 x i8> %a, <32 x i8> %b, <32 x i8> zeroinitializer, i32 -1)
  ret <32 x i8> %res
}

define <32 x i8> @test_mask_adds_epu8_rmk_256(<32 x i8> %a, <32 x i8>* %ptr_b, <32 x i8> %passThru, i32 %mask) {
; CHECK-LABEL: test_mask_adds_epu8_rmk_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovd %esi, %k1 ## encoding: [0xc5,0xfb,0x92,0xce]
; CHECK-NEXT:    vpaddusb (%rdi), %ymm0, %ymm1 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0xdc,0x0f]
; CHECK-NEXT:    vmovaps %zmm1, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <32 x i8>, <32 x i8>* %ptr_b
  %res = call <32 x i8> @llvm.x86.avx512.mask.paddus.b.256(<32 x i8> %a, <32 x i8> %b, <32 x i8> %passThru, i32 %mask)
  ret <32 x i8> %res
}

define <32 x i8> @test_mask_adds_epu8_rmkz_256(<32 x i8> %a, <32 x i8>* %ptr_b, i32 %mask) {
; CHECK-LABEL: test_mask_adds_epu8_rmkz_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovd %esi, %k1 ## encoding: [0xc5,0xfb,0x92,0xce]
; CHECK-NEXT:    vpaddusb (%rdi), %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xa9,0xdc,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <32 x i8>, <32 x i8>* %ptr_b
  %res = call <32 x i8> @llvm.x86.avx512.mask.paddus.b.256(<32 x i8> %a, <32 x i8> %b, <32 x i8> zeroinitializer, i32 %mask)
  ret <32 x i8> %res
}

declare <32 x i8> @llvm.x86.avx512.mask.paddus.b.256(<32 x i8>, <32 x i8>, <32 x i8>, i32)

define <16 x i8> @test_mask_subs_epu8_rr_128(<16 x i8> %a, <16 x i8> %b) {
; CHECK-LABEL: test_mask_subs_epu8_rr_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpsubusb %xmm1, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xd8,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i8> @llvm.x86.avx512.mask.psubus.b.128(<16 x i8> %a, <16 x i8> %b, <16 x i8> zeroinitializer, i16 -1)
  ret <16 x i8> %res
}

define <16 x i8> @test_mask_subs_epu8_rrk_128(<16 x i8> %a, <16 x i8> %b, <16 x i8> %passThru, i16 %mask) {
; CHECK-LABEL: test_mask_subs_epu8_rrk_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpsubusb %xmm1, %xmm0, %xmm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0xd8,0xd1]
; CHECK-NEXT:    vmovaps %zmm2, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i8> @llvm.x86.avx512.mask.psubus.b.128(<16 x i8> %a, <16 x i8> %b, <16 x i8> %passThru, i16 %mask)
  ret <16 x i8> %res
}

define <16 x i8> @test_mask_subs_epu8_rrkz_128(<16 x i8> %a, <16 x i8> %b, i16 %mask) {
; CHECK-LABEL: test_mask_subs_epu8_rrkz_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpsubusb %xmm1, %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0x89,0xd8,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i8> @llvm.x86.avx512.mask.psubus.b.128(<16 x i8> %a, <16 x i8> %b, <16 x i8> zeroinitializer, i16 %mask)
  ret <16 x i8> %res
}

define <16 x i8> @test_mask_subs_epu8_rm_128(<16 x i8> %a, <16 x i8>* %ptr_b) {
; CHECK-LABEL: test_mask_subs_epu8_rm_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpsubusb (%rdi), %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xd8,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <16 x i8>, <16 x i8>* %ptr_b
  %res = call <16 x i8> @llvm.x86.avx512.mask.psubus.b.128(<16 x i8> %a, <16 x i8> %b, <16 x i8> zeroinitializer, i16 -1)
  ret <16 x i8> %res
}

define <16 x i8> @test_mask_subs_epu8_rmk_128(<16 x i8> %a, <16 x i8>* %ptr_b, <16 x i8> %passThru, i16 %mask) {
; CHECK-LABEL: test_mask_subs_epu8_rmk_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpsubusb (%rdi), %xmm0, %xmm1 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0xd8,0x0f]
; CHECK-NEXT:    vmovaps %zmm1, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <16 x i8>, <16 x i8>* %ptr_b
  %res = call <16 x i8> @llvm.x86.avx512.mask.psubus.b.128(<16 x i8> %a, <16 x i8> %b, <16 x i8> %passThru, i16 %mask)
  ret <16 x i8> %res
}

define <16 x i8> @test_mask_subs_epu8_rmkz_128(<16 x i8> %a, <16 x i8>* %ptr_b, i16 %mask) {
; CHECK-LABEL: test_mask_subs_epu8_rmkz_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpsubusb (%rdi), %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0x89,0xd8,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <16 x i8>, <16 x i8>* %ptr_b
  %res = call <16 x i8> @llvm.x86.avx512.mask.psubus.b.128(<16 x i8> %a, <16 x i8> %b, <16 x i8> zeroinitializer, i16 %mask)
  ret <16 x i8> %res
}

declare <16 x i8> @llvm.x86.avx512.mask.psubus.b.128(<16 x i8>, <16 x i8>, <16 x i8>, i16)

define <32 x i8> @test_mask_subs_epu8_rr_256(<32 x i8> %a, <32 x i8> %b) {
; CHECK-LABEL: test_mask_subs_epu8_rr_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpsubusb %ymm1, %ymm0, %ymm0 ## encoding: [0x62,0xf1,0x7d,0x28,0xd8,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <32 x i8> @llvm.x86.avx512.mask.psubus.b.256(<32 x i8> %a, <32 x i8> %b, <32 x i8> zeroinitializer, i32 -1)
  ret <32 x i8> %res
}

define <32 x i8> @test_mask_subs_epu8_rrk_256(<32 x i8> %a, <32 x i8> %b, <32 x i8> %passThru, i32 %mask) {
; CHECK-LABEL: test_mask_subs_epu8_rrk_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
; CHECK-NEXT:    vpsubusb %ymm1, %ymm0, %ymm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0xd8,0xd1]
; CHECK-NEXT:    vmovaps %zmm2, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <32 x i8> @llvm.x86.avx512.mask.psubus.b.256(<32 x i8> %a, <32 x i8> %b, <32 x i8> %passThru, i32 %mask)
  ret <32 x i8> %res
}

define <32 x i8> @test_mask_subs_epu8_rrkz_256(<32 x i8> %a, <32 x i8> %b, i32 %mask) {
; CHECK-LABEL: test_mask_subs_epu8_rrkz_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
; CHECK-NEXT:    vpsubusb %ymm1, %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xa9,0xd8,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <32 x i8> @llvm.x86.avx512.mask.psubus.b.256(<32 x i8> %a, <32 x i8> %b, <32 x i8> zeroinitializer, i32 %mask)
  ret <32 x i8> %res
}

define <32 x i8> @test_mask_subs_epu8_rm_256(<32 x i8> %a, <32 x i8>* %ptr_b) {
; CHECK-LABEL: test_mask_subs_epu8_rm_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpsubusb (%rdi), %ymm0, %ymm0 ## encoding: [0x62,0xf1,0x7d,0x28,0xd8,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <32 x i8>, <32 x i8>* %ptr_b
  %res = call <32 x i8> @llvm.x86.avx512.mask.psubus.b.256(<32 x i8> %a, <32 x i8> %b, <32 x i8> zeroinitializer, i32 -1)
  ret <32 x i8> %res
}

define <32 x i8> @test_mask_subs_epu8_rmk_256(<32 x i8> %a, <32 x i8>* %ptr_b, <32 x i8> %passThru, i32 %mask) {
; CHECK-LABEL: test_mask_subs_epu8_rmk_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovd %esi, %k1 ## encoding: [0xc5,0xfb,0x92,0xce]
; CHECK-NEXT:    vpsubusb (%rdi), %ymm0, %ymm1 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0xd8,0x0f]
; CHECK-NEXT:    vmovaps %zmm1, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <32 x i8>, <32 x i8>* %ptr_b
  %res = call <32 x i8> @llvm.x86.avx512.mask.psubus.b.256(<32 x i8> %a, <32 x i8> %b, <32 x i8> %passThru, i32 %mask)
  ret <32 x i8> %res
}

define <32 x i8> @test_mask_subs_epu8_rmkz_256(<32 x i8> %a, <32 x i8>* %ptr_b, i32 %mask) {
; CHECK-LABEL: test_mask_subs_epu8_rmkz_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovd %esi, %k1 ## encoding: [0xc5,0xfb,0x92,0xce]
; CHECK-NEXT:    vpsubusb (%rdi), %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xa9,0xd8,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %b = load <32 x i8>, <32 x i8>* %ptr_b
  %res = call <32 x i8> @llvm.x86.avx512.mask.psubus.b.256(<32 x i8> %a, <32 x i8> %b, <32 x i8> zeroinitializer, i32 %mask)
  ret <32 x i8> %res
}

declare <32 x i8> @llvm.x86.avx512.mask.psubus.b.256(<32 x i8>, <32 x i8>, <32 x i8>, i32)

declare <16 x i8> @llvm.x86.avx512.mask.pmaxs.b.128(<16 x i8>, <16 x i8>, <16 x i8>, i16)

define <16 x i8>@test_int_x86_avx512_mask_pmaxs_b_128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %x2, i16 %mask) {
; CHECK-LABEL: test_int_x86_avx512_mask_pmaxs_b_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpmaxsb %xmm1, %xmm0, %xmm2 {%k1} ## encoding: [0x62,0xf2,0x7d,0x09,0x3c,0xd1]
; CHECK-NEXT:    vpmaxsb %xmm1, %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf2,0x7d,0x89,0x3c,0xc1]
; CHECK-NEXT:    vpaddb %xmm0, %xmm2, %xmm0 ## encoding: [0x62,0xf1,0x6d,0x08,0xfc,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i8> @llvm.x86.avx512.mask.pmaxs.b.128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %x2 ,i16 %mask)
  %res1 = call <16 x i8> @llvm.x86.avx512.mask.pmaxs.b.128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> zeroinitializer, i16 %mask)
  %res2 = add <16 x i8> %res, %res1
  ret <16 x i8> %res2
}

declare <32 x i8> @llvm.x86.avx512.mask.pmaxs.b.256(<32 x i8>, <32 x i8>, <32 x i8>, i32)

define <32 x i8>@test_int_x86_avx512_mask_pmaxs_b_256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_pmaxs_b_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
; CHECK-NEXT:    vpmaxsb %ymm1, %ymm0, %ymm2 {%k1} ## encoding: [0x62,0xf2,0x7d,0x29,0x3c,0xd1]
; CHECK-NEXT:    vpmaxsb %ymm1, %ymm0, %ymm0 ## encoding: [0x62,0xf2,0x7d,0x28,0x3c,0xc1]
; CHECK-NEXT:    vpaddb %ymm0, %ymm2, %ymm0 ## encoding: [0x62,0xf1,0x6d,0x28,0xfc,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <32 x i8> @llvm.x86.avx512.mask.pmaxs.b.256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 %x3)
  %res1 = call <32 x i8> @llvm.x86.avx512.mask.pmaxs.b.256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 -1)
  %res2 = add <32 x i8> %res, %res1
  ret <32 x i8> %res2
}

declare <8 x i16> @llvm.x86.avx512.mask.pmaxs.w.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)

define <8 x i16>@test_int_x86_avx512_mask_pmaxs_w_128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_pmaxs_w_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpmaxsw %xmm1, %xmm0, %xmm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0xee,0xd1]
; CHECK-NEXT:    vpmaxsw %xmm1, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xee,0xc1]
; CHECK-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 ## encoding: [0x62,0xf1,0x6d,0x08,0xfd,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.mask.pmaxs.w.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3)
  %res1 = call <8 x i16> @llvm.x86.avx512.mask.pmaxs.w.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 -1)
  %res2 = add <8 x i16> %res, %res1
  ret <8 x i16> %res2
}

declare <16 x i16> @llvm.x86.avx512.mask.pmaxs.w.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)

define <16 x i16>@test_int_x86_avx512_mask_pmaxs_w_256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %mask) {
; CHECK-LABEL: test_int_x86_avx512_mask_pmaxs_w_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpmaxsw %ymm1, %ymm0, %ymm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0xee,0xd1]
; CHECK-NEXT:    vpmaxsw %ymm1, %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xa9,0xee,0xc1]
; CHECK-NEXT:    vpaddw %ymm0, %ymm2, %ymm0 ## encoding: [0x62,0xf1,0x6d,0x28,0xfd,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.mask.pmaxs.w.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %mask)
  %res1 = call <16 x i16> @llvm.x86.avx512.mask.pmaxs.w.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> zeroinitializer, i16 %mask)
  %res2 = add <16 x i16> %res, %res1
  ret <16 x i16> %res2
}

declare <16 x i8> @llvm.x86.avx512.mask.pmaxu.b.128(<16 x i8>, <16 x i8>, <16 x i8>, i16)

define <16 x i8>@test_int_x86_avx512_mask_pmaxu_b_128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %x2,i16 %mask) {
; CHECK-LABEL: test_int_x86_avx512_mask_pmaxu_b_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpmaxub %xmm1, %xmm0, %xmm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0xde,0xd1]
; CHECK-NEXT:    vpmaxub %xmm1, %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0x89,0xde,0xc1]
; CHECK-NEXT:    vpaddb %xmm0, %xmm2, %xmm0 ## encoding: [0x62,0xf1,0x6d,0x08,0xfc,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i8> @llvm.x86.avx512.mask.pmaxu.b.128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %x2, i16 %mask)
  %res1 = call <16 x i8> @llvm.x86.avx512.mask.pmaxu.b.128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> zeroinitializer, i16 %mask)
  %res2 = add <16 x i8> %res, %res1
  ret <16 x i8> %res2
}

declare <32 x i8> @llvm.x86.avx512.mask.pmaxu.b.256(<32 x i8>, <32 x i8>, <32 x i8>, i32)

define <32 x i8>@test_int_x86_avx512_mask_pmaxu_b_256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_pmaxu_b_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
; CHECK-NEXT:    vpmaxub %ymm1, %ymm0, %ymm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0xde,0xd1]
; CHECK-NEXT:    vpmaxub %ymm1, %ymm0, %ymm0 ## encoding: [0x62,0xf1,0x7d,0x28,0xde,0xc1]
; CHECK-NEXT:    vpaddb %ymm0, %ymm2, %ymm0 ## encoding: [0x62,0xf1,0x6d,0x28,0xfc,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <32 x i8> @llvm.x86.avx512.mask.pmaxu.b.256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 %x3)
  %res1 = call <32 x i8> @llvm.x86.avx512.mask.pmaxu.b.256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 -1)
  %res2 = add <32 x i8> %res, %res1
  ret <32 x i8> %res2
}

declare <8 x i16> @llvm.x86.avx512.mask.pmaxu.w.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)

define <8 x i16>@test_int_x86_avx512_mask_pmaxu_w_128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_pmaxu_w_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpmaxuw %xmm1, %xmm0, %xmm2 {%k1} ## encoding: [0x62,0xf2,0x7d,0x09,0x3e,0xd1]
; CHECK-NEXT:    vpmaxuw %xmm1, %xmm0, %xmm0 ## encoding: [0x62,0xf2,0x7d,0x08,0x3e,0xc1]
; CHECK-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 ## encoding: [0x62,0xf1,0x6d,0x08,0xfd,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.mask.pmaxu.w.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3)
  %res1 = call <8 x i16> @llvm.x86.avx512.mask.pmaxu.w.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 -1)
  %res2 = add <8 x i16> %res, %res1
  ret <8 x i16> %res2
}

declare <16 x i16> @llvm.x86.avx512.mask.pmaxu.w.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)

define <16 x i16>@test_int_x86_avx512_mask_pmaxu_w_256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %mask) {
; CHECK-LABEL: test_int_x86_avx512_mask_pmaxu_w_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpmaxuw %ymm1, %ymm0, %ymm2 {%k1} ## encoding: [0x62,0xf2,0x7d,0x29,0x3e,0xd1]
; CHECK-NEXT:    vpmaxuw %ymm1, %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf2,0x7d,0xa9,0x3e,0xc1]
; CHECK-NEXT:    vpaddw %ymm0, %ymm2, %ymm0 ## encoding: [0x62,0xf1,0x6d,0x28,0xfd,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.mask.pmaxu.w.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %mask)
  %res1 = call <16 x i16> @llvm.x86.avx512.mask.pmaxu.w.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> zeroinitializer, i16 %mask)
  %res2 = add <16 x i16> %res, %res1
  ret <16 x i16> %res2
}

declare <16 x i8> @llvm.x86.avx512.mask.pmins.b.128(<16 x i8>, <16 x i8>, <16 x i8>, i16)

define <16 x i8>@test_int_x86_avx512_mask_pmins_b_128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %x2, i16 %mask) {
; CHECK-LABEL: test_int_x86_avx512_mask_pmins_b_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpminsb %xmm1, %xmm0, %xmm2 {%k1} ## encoding: [0x62,0xf2,0x7d,0x09,0x38,0xd1]
; CHECK-NEXT:    vpminsb %xmm1, %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf2,0x7d,0x89,0x38,0xc1]
; CHECK-NEXT:    vpaddb %xmm0, %xmm2, %xmm0 ## encoding: [0x62,0xf1,0x6d,0x08,0xfc,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i8> @llvm.x86.avx512.mask.pmins.b.128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %x2, i16 %mask)
  %res1 = call <16 x i8> @llvm.x86.avx512.mask.pmins.b.128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> zeroinitializer, i16 %mask)
  %res2 = add <16 x i8> %res, %res1
  ret <16 x i8> %res2
}

declare <32 x i8> @llvm.x86.avx512.mask.pmins.b.256(<32 x i8>, <32 x i8>, <32 x i8>, i32)

define <32 x i8>@test_int_x86_avx512_mask_pmins_b_256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_pmins_b_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
; CHECK-NEXT:    vpminsb %ymm1, %ymm0, %ymm2 {%k1} ## encoding: [0x62,0xf2,0x7d,0x29,0x38,0xd1]
; CHECK-NEXT:    vpminsb %ymm1, %ymm0, %ymm0 ## encoding: [0x62,0xf2,0x7d,0x28,0x38,0xc1]
; CHECK-NEXT:    vpaddb %ymm0, %ymm2, %ymm0 ## encoding: [0x62,0xf1,0x6d,0x28,0xfc,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <32 x i8> @llvm.x86.avx512.mask.pmins.b.256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 %x3)
  %res1 = call <32 x i8> @llvm.x86.avx512.mask.pmins.b.256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 -1)
  %res2 = add <32 x i8> %res, %res1
  ret <32 x i8> %res2
}

declare <8 x i16> @llvm.x86.avx512.mask.pmins.w.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)

define <8 x i16>@test_int_x86_avx512_mask_pmins_w_128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_pmins_w_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpminsw %xmm1, %xmm0, %xmm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0xea,0xd1]
; CHECK-NEXT:    vpminsw %xmm1, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xea,0xc1]
; CHECK-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 ## encoding: [0x62,0xf1,0x6d,0x08,0xfd,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.mask.pmins.w.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3)
  %res1 = call <8 x i16> @llvm.x86.avx512.mask.pmins.w.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 -1)
  %res2 = add <8 x i16> %res, %res1
  ret <8 x i16> %res2
}

declare <16 x i16> @llvm.x86.avx512.mask.pmins.w.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)

define <16 x i16>@test_int_x86_avx512_mask_pmins_w_256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %mask) {
; CHECK-LABEL: test_int_x86_avx512_mask_pmins_w_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpminsw %ymm1, %ymm0, %ymm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0xea,0xd1]
; CHECK-NEXT:    vpminsw %ymm1, %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xa9,0xea,0xc1]
; CHECK-NEXT:    vpaddw %ymm0, %ymm2, %ymm0 ## encoding: [0x62,0xf1,0x6d,0x28,0xfd,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.mask.pmins.w.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %mask)
  %res1 = call <16 x i16> @llvm.x86.avx512.mask.pmins.w.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> zeroinitializer, i16 %mask)
  %res2 = add <16 x i16> %res, %res1
  ret <16 x i16> %res2
}

declare <16 x i8> @llvm.x86.avx512.mask.pminu.b.128(<16 x i8>, <16 x i8>, <16 x i8>, i16)

define <16 x i8>@test_int_x86_avx512_mask_pminu_b_128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %x2, i16 %mask) {
; CHECK-LABEL: test_int_x86_avx512_mask_pminu_b_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpminub %xmm1, %xmm0, %xmm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0xda,0xd1]
; CHECK-NEXT:    vpminub %xmm1, %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0x89,0xda,0xc1]
; CHECK-NEXT:    vpaddb %xmm0, %xmm2, %xmm0 ## encoding: [0x62,0xf1,0x6d,0x08,0xfc,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i8> @llvm.x86.avx512.mask.pminu.b.128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %x2, i16 %mask)
  %res1 = call <16 x i8> @llvm.x86.avx512.mask.pminu.b.128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> zeroinitializer, i16 %mask)
  %res2 = add <16 x i8> %res, %res1
  ret <16 x i8> %res2
}

declare <32 x i8> @llvm.x86.avx512.mask.pminu.b.256(<32 x i8>, <32 x i8>, <32 x i8>, i32)

define <32 x i8>@test_int_x86_avx512_mask_pminu_b_256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_pminu_b_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
; CHECK-NEXT:    vpminub %ymm1, %ymm0, %ymm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0xda,0xd1]
; CHECK-NEXT:    vpminub %ymm1, %ymm0, %ymm0 ## encoding: [0x62,0xf1,0x7d,0x28,0xda,0xc1]
; CHECK-NEXT:    vpaddb %ymm0, %ymm2, %ymm0 ## encoding: [0x62,0xf1,0x6d,0x28,0xfc,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <32 x i8> @llvm.x86.avx512.mask.pminu.b.256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 %x3)
  %res1 = call <32 x i8> @llvm.x86.avx512.mask.pminu.b.256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 -1)
  %res2 = add <32 x i8> %res, %res1
  ret <32 x i8> %res2
}

declare <8 x i16> @llvm.x86.avx512.mask.pminu.w.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)

define <8 x i16>@test_int_x86_avx512_mask_pminu_w_128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_pminu_w_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpminuw %xmm1, %xmm0, %xmm2 {%k1} ## encoding: [0x62,0xf2,0x7d,0x09,0x3a,0xd1]
; CHECK-NEXT:    vpminuw %xmm1, %xmm0, %xmm0 ## encoding: [0x62,0xf2,0x7d,0x08,0x3a,0xc1]
; CHECK-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 ## encoding: [0x62,0xf1,0x6d,0x08,0xfd,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.mask.pminu.w.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3)
  %res1 = call <8 x i16> @llvm.x86.avx512.mask.pminu.w.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 -1)
  %res2 = add <8 x i16> %res, %res1
  ret <8 x i16> %res2
}

declare <16 x i16> @llvm.x86.avx512.mask.pminu.w.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)

define <16 x i16>@test_int_x86_avx512_mask_pminu_w_256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %mask) {
; CHECK-LABEL: test_int_x86_avx512_mask_pminu_w_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpminuw %ymm1, %ymm0, %ymm2 {%k1} ## encoding: [0x62,0xf2,0x7d,0x29,0x3a,0xd1]
; CHECK-NEXT:    vpminuw %ymm1, %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf2,0x7d,0xa9,0x3a,0xc1]
; CHECK-NEXT:    vpaddw %ymm0, %ymm2, %ymm0 ## encoding: [0x62,0xf1,0x6d,0x28,0xfd,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.mask.pminu.w.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %mask)
  %res1 = call <16 x i16> @llvm.x86.avx512.mask.pminu.w.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> zeroinitializer, i16 %mask)
  %res2 = add <16 x i16> %res, %res1
  ret <16 x i16> %res2
}

declare <8 x i16> @llvm.x86.avx512.mask.vpermt2var.hi.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)

define <8 x i16>@test_int_x86_avx512_mask_vpermt2var_hi_128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_vpermt2var_hi_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vmovaps %zmm1, %zmm3 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xd9]
; CHECK-NEXT:    vpermt2w %xmm2, %xmm0, %xmm3 {%k1} ## encoding: [0x62,0xf2,0xfd,0x09,0x7d,0xda]
; CHECK-NEXT:    vpermt2w %xmm2, %xmm0, %xmm1 ## encoding: [0x62,0xf2,0xfd,0x08,0x7d,0xca]
; CHECK-NEXT:    vpaddw %xmm1, %xmm3, %xmm0 ## encoding: [0x62,0xf1,0x65,0x08,0xfd,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.mask.vpermt2var.hi.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3)
  %res1 = call <8 x i16> @llvm.x86.avx512.mask.vpermt2var.hi.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 -1)
  %res2 = add <8 x i16> %res, %res1
  ret <8 x i16> %res2
}

declare <8 x i16> @llvm.x86.avx512.maskz.vpermt2var.hi.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)

define <8 x i16>@test_int_x86_avx512_maskz_vpermt2var_hi_128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_maskz_vpermt2var_hi_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vmovaps %zmm1, %zmm3 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xd9]
; CHECK-NEXT:    vpermt2w %xmm2, %xmm0, %xmm3 {%k1} {z} ## encoding: [0x62,0xf2,0xfd,0x89,0x7d,0xda]
; CHECK-NEXT:    vpermt2w %xmm2, %xmm0, %xmm1 ## encoding: [0x62,0xf2,0xfd,0x08,0x7d,0xca]
; CHECK-NEXT:    vpaddw %xmm1, %xmm3, %xmm0 ## encoding: [0x62,0xf1,0x65,0x08,0xfd,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.maskz.vpermt2var.hi.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3)
  %res1 = call <8 x i16> @llvm.x86.avx512.maskz.vpermt2var.hi.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 -1)
  %res2 = add <8 x i16> %res, %res1
  ret <8 x i16> %res2
}

declare <16 x i16> @llvm.x86.avx512.mask.vpermt2var.hi.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)

define <16 x i16>@test_int_x86_avx512_mask_vpermt2var_hi_256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_vpermt2var_hi_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vmovaps %zmm1, %zmm3 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xd9]
; CHECK-NEXT:    vpermt2w %ymm2, %ymm0, %ymm3 {%k1} ## encoding: [0x62,0xf2,0xfd,0x29,0x7d,0xda]
; CHECK-NEXT:    vpermt2w %ymm2, %ymm0, %ymm1 ## encoding: [0x62,0xf2,0xfd,0x28,0x7d,0xca]
; CHECK-NEXT:    vpaddw %ymm1, %ymm3, %ymm0 ## encoding: [0x62,0xf1,0x65,0x28,0xfd,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.mask.vpermt2var.hi.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %x3)
  %res1 = call <16 x i16> @llvm.x86.avx512.mask.vpermt2var.hi.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 -1)
  %res2 = add <16 x i16> %res, %res1
  ret <16 x i16> %res2
}

declare <16 x i16> @llvm.x86.avx512.maskz.vpermt2var.hi.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)

define <16 x i16>@test_int_x86_avx512_maskz_vpermt2var_hi_256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %x3) {
; CHECK-LABEL: test_int_x86_avx512_maskz_vpermt2var_hi_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vmovaps %zmm1, %zmm3 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xd9]
; CHECK-NEXT:    vpermt2w %ymm2, %ymm0, %ymm3 {%k1} {z} ## encoding: [0x62,0xf2,0xfd,0xa9,0x7d,0xda]
; CHECK-NEXT:    vpermt2w %ymm2, %ymm0, %ymm1 ## encoding: [0x62,0xf2,0xfd,0x28,0x7d,0xca]
; CHECK-NEXT:    vpaddw %ymm1, %ymm3, %ymm0 ## encoding: [0x62,0xf1,0x65,0x28,0xfd,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.maskz.vpermt2var.hi.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %x3)
  %res1 = call <16 x i16> @llvm.x86.avx512.maskz.vpermt2var.hi.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 -1)
  %res2 = add <16 x i16> %res, %res1
  ret <16 x i16> %res2
}

declare <8 x i16> @llvm.x86.avx512.mask.vpermi2var.hi.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)

define <8 x i16>@test_int_x86_avx512_mask_vpermi2var_hi_128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_vpermi2var_hi_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vmovaps %zmm1, %zmm3 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xd9]
; CHECK-NEXT:    vpermi2w %xmm2, %xmm0, %xmm3 {%k1} ## encoding: [0x62,0xf2,0xfd,0x09,0x75,0xda]
; CHECK-NEXT:    vpermi2w %xmm2, %xmm0, %xmm1 ## encoding: [0x62,0xf2,0xfd,0x08,0x75,0xca]
; CHECK-NEXT:    vpaddw %xmm1, %xmm3, %xmm0 ## encoding: [0x62,0xf1,0x65,0x08,0xfd,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.mask.vpermi2var.hi.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3)
  %res1 = call <8 x i16> @llvm.x86.avx512.mask.vpermi2var.hi.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 -1)
  %res2 = add <8 x i16> %res, %res1
  ret <8 x i16> %res2
}

declare <16 x i16> @llvm.x86.avx512.mask.vpermi2var.hi.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)

define <16 x i16>@test_int_x86_avx512_mask_vpermi2var_hi_256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_vpermi2var_hi_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vmovaps %zmm1, %zmm3 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xd9]
; CHECK-NEXT:    vpermi2w %ymm2, %ymm0, %ymm3 {%k1} ## encoding: [0x62,0xf2,0xfd,0x29,0x75,0xda]
; CHECK-NEXT:    vpermi2w %ymm2, %ymm0, %ymm1 ## encoding: [0x62,0xf2,0xfd,0x28,0x75,0xca]
; CHECK-NEXT:    vpaddw %ymm1, %ymm3, %ymm0 ## encoding: [0x62,0xf1,0x65,0x28,0xfd,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.mask.vpermi2var.hi.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %x3)
  %res1 = call <16 x i16> @llvm.x86.avx512.mask.vpermi2var.hi.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 -1)
  %res2 = add <16 x i16> %res, %res1
  ret <16 x i16> %res2
}

declare <16 x i8> @llvm.x86.avx512.mask.pavg.b.128(<16 x i8>, <16 x i8>, <16 x i8>, i16)

define <16 x i8>@test_int_x86_avx512_mask_pavg_b_128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %x2, i16 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_pavg_b_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpavgb %xmm1, %xmm0, %xmm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0xe0,0xd1]
; CHECK-NEXT:    vpavgb %xmm1, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xe0,0xc1]
; CHECK-NEXT:    vpaddb %xmm0, %xmm2, %xmm0 ## encoding: [0x62,0xf1,0x6d,0x08,0xfc,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i8> @llvm.x86.avx512.mask.pavg.b.128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %x2, i16 %x3)
  %res1 = call <16 x i8> @llvm.x86.avx512.mask.pavg.b.128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %x2, i16 -1)
  %res2 = add <16 x i8> %res, %res1
  ret <16 x i8> %res2
}

declare <32 x i8> @llvm.x86.avx512.mask.pavg.b.256(<32 x i8>, <32 x i8>, <32 x i8>, i32)

define <32 x i8>@test_int_x86_avx512_mask_pavg_b_256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_pavg_b_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
; CHECK-NEXT:    vpavgb %ymm1, %ymm0, %ymm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0xe0,0xd1]
; CHECK-NEXT:    vpavgb %ymm1, %ymm0, %ymm0 ## encoding: [0x62,0xf1,0x7d,0x28,0xe0,0xc1]
; CHECK-NEXT:    vpaddb %ymm0, %ymm2, %ymm0 ## encoding: [0x62,0xf1,0x6d,0x28,0xfc,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <32 x i8> @llvm.x86.avx512.mask.pavg.b.256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 %x3)
  %res1 = call <32 x i8> @llvm.x86.avx512.mask.pavg.b.256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 -1)
  %res2 = add <32 x i8> %res, %res1
  ret <32 x i8> %res2
}

declare <8 x i16> @llvm.x86.avx512.mask.pavg.w.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)

define <8 x i16>@test_int_x86_avx512_mask_pavg_w_128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_pavg_w_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpavgw %xmm1, %xmm0, %xmm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0xe3,0xd1]
; CHECK-NEXT:    vpavgw %xmm1, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xe3,0xc1]
; CHECK-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 ## encoding: [0x62,0xf1,0x6d,0x08,0xfd,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.mask.pavg.w.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3)
  %res1 = call <8 x i16> @llvm.x86.avx512.mask.pavg.w.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 -1)
  %res2 = add <8 x i16> %res, %res1
  ret <8 x i16> %res2
}

declare <16 x i16> @llvm.x86.avx512.mask.pavg.w.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)

define <16 x i16>@test_int_x86_avx512_mask_pavg_w_256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_pavg_w_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpavgw %ymm1, %ymm0, %ymm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0xe3,0xd1]
; CHECK-NEXT:    vpavgw %ymm1, %ymm0, %ymm0 ## encoding: [0x62,0xf1,0x7d,0x28,0xe3,0xc1]
; CHECK-NEXT:    vpaddw %ymm0, %ymm2, %ymm0 ## encoding: [0x62,0xf1,0x6d,0x28,0xfd,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.mask.pavg.w.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %x3)
  %res1 = call <16 x i16> @llvm.x86.avx512.mask.pavg.w.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 -1)
  %res2 = add <16 x i16> %res, %res1
  ret <16 x i16> %res2
}

declare <16 x i8> @llvm.x86.avx512.mask.pshuf.b.128(<16 x i8>, <16 x i8>, <16 x i8>, i16)

define <16 x i8>@test_int_x86_avx512_mask_pshuf_b_128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %x2, i16 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_pshuf_b_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpshufb %xmm1, %xmm0, %xmm2 {%k1} ## encoding: [0x62,0xf2,0x7d,0x09,0x00,0xd1]
; CHECK-NEXT:    vpshufb %xmm1, %xmm0, %xmm0 ## encoding: [0xc4,0xe2,0x79,0x00,0xc1]
; CHECK-NEXT:    vpaddb %xmm0, %xmm2, %xmm0 ## encoding: [0x62,0xf1,0x6d,0x08,0xfc,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i8> @llvm.x86.avx512.mask.pshuf.b.128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %x2, i16 %x3)
  %res1 = call <16 x i8> @llvm.x86.avx512.mask.pshuf.b.128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %x2, i16 -1)
  %res2 = add <16 x i8> %res, %res1
  ret <16 x i8> %res2
}

declare <32 x i8> @llvm.x86.avx512.mask.pshuf.b.256(<32 x i8>, <32 x i8>, <32 x i8>, i32)

define <32 x i8>@test_int_x86_avx512_mask_pshuf_b_256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_pshuf_b_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
; CHECK-NEXT:    vpshufb %ymm1, %ymm0, %ymm2 {%k1} ## encoding: [0x62,0xf2,0x7d,0x29,0x00,0xd1]
; CHECK-NEXT:    vpshufb %ymm1, %ymm0, %ymm0 ## encoding: [0xc4,0xe2,0x7d,0x00,0xc1]
; CHECK-NEXT:    vpaddb %ymm0, %ymm2, %ymm0 ## encoding: [0x62,0xf1,0x6d,0x28,0xfc,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <32 x i8> @llvm.x86.avx512.mask.pshuf.b.256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 %x3)
  %res1 = call <32 x i8> @llvm.x86.avx512.mask.pshuf.b.256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 -1)
  %res2 = add <32 x i8> %res, %res1
  ret <32 x i8> %res2
}

declare <16 x i8> @llvm.x86.avx512.mask.pabs.b.128(<16 x i8>, <16 x i8>, i16)

define <16 x i8>@test_int_x86_avx512_mask_pabs_b_128(<16 x i8> %x0, <16 x i8> %x1, i16 %x2) {
; CHECK-LABEL: test_int_x86_avx512_mask_pabs_b_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpabsb %xmm0, %xmm1 {%k1} ## encoding: [0x62,0xf2,0x7d,0x09,0x1c,0xc8]
; CHECK-NEXT:    vpabsb %xmm0, %xmm0 ## encoding: [0x62,0xf2,0x7d,0x08,0x1c,0xc0]
; CHECK-NEXT:    vpaddb %xmm0, %xmm1, %xmm0 ## encoding: [0x62,0xf1,0x75,0x08,0xfc,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i8> @llvm.x86.avx512.mask.pabs.b.128(<16 x i8> %x0, <16 x i8> %x1, i16 %x2)
  %res1 = call <16 x i8> @llvm.x86.avx512.mask.pabs.b.128(<16 x i8> %x0, <16 x i8> %x1, i16 -1)
  %res2 = add <16 x i8> %res, %res1
  ret <16 x i8> %res2
}

declare <32 x i8> @llvm.x86.avx512.mask.pabs.b.256(<32 x i8>, <32 x i8>, i32)

define <32 x i8>@test_int_x86_avx512_mask_pabs_b_256(<32 x i8> %x0, <32 x i8> %x1, i32 %x2) {
; CHECK-LABEL: test_int_x86_avx512_mask_pabs_b_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
; CHECK-NEXT:    vpabsb %ymm0, %ymm1 {%k1} ## encoding: [0x62,0xf2,0x7d,0x29,0x1c,0xc8]
; CHECK-NEXT:    vpabsb %ymm0, %ymm0 ## encoding: [0x62,0xf2,0x7d,0x28,0x1c,0xc0]
; CHECK-NEXT:    vpaddb %ymm0, %ymm1, %ymm0 ## encoding: [0x62,0xf1,0x75,0x28,0xfc,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <32 x i8> @llvm.x86.avx512.mask.pabs.b.256(<32 x i8> %x0, <32 x i8> %x1, i32 %x2)
  %res1 = call <32 x i8> @llvm.x86.avx512.mask.pabs.b.256(<32 x i8> %x0, <32 x i8> %x1, i32 -1)
  %res2 = add <32 x i8> %res, %res1
  ret <32 x i8> %res2
}

declare <8 x i16> @llvm.x86.avx512.mask.pabs.w.128(<8 x i16>, <8 x i16>, i8)

define <8 x i16>@test_int_x86_avx512_mask_pabs_w_128(<8 x i16> %x0, <8 x i16> %x1, i8 %x2) {
; CHECK-LABEL: test_int_x86_avx512_mask_pabs_w_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpabsw %xmm0, %xmm1 {%k1} ## encoding: [0x62,0xf2,0x7d,0x09,0x1d,0xc8]
; CHECK-NEXT:    vpabsw %xmm0, %xmm0 ## encoding: [0x62,0xf2,0x7d,0x08,0x1d,0xc0]
; CHECK-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 ## encoding: [0x62,0xf1,0x75,0x08,0xfd,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.mask.pabs.w.128(<8 x i16> %x0, <8 x i16> %x1, i8 %x2)
  %res1 = call <8 x i16> @llvm.x86.avx512.mask.pabs.w.128(<8 x i16> %x0, <8 x i16> %x1, i8 -1)
  %res2 = add <8 x i16> %res, %res1
  ret <8 x i16> %res2
}

declare <16 x i16> @llvm.x86.avx512.mask.pabs.w.256(<16 x i16>, <16 x i16>, i16)

define <16 x i16>@test_int_x86_avx512_mask_pabs_w_256(<16 x i16> %x0, <16 x i16> %x1, i16 %x2) {
; CHECK-LABEL: test_int_x86_avx512_mask_pabs_w_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpabsw %ymm0, %ymm1 {%k1} ## encoding: [0x62,0xf2,0x7d,0x29,0x1d,0xc8]
; CHECK-NEXT:    vpabsw %ymm0, %ymm0 ## encoding: [0x62,0xf2,0x7d,0x28,0x1d,0xc0]
; CHECK-NEXT:    vpaddw %ymm0, %ymm1, %ymm0 ## encoding: [0x62,0xf1,0x75,0x28,0xfd,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.mask.pabs.w.256(<16 x i16> %x0, <16 x i16> %x1, i16 %x2)
  %res1 = call <16 x i16> @llvm.x86.avx512.mask.pabs.w.256(<16 x i16> %x0, <16 x i16> %x1, i16 -1)
  %res2 = add <16 x i16> %res, %res1
  ret <16 x i16> %res2
}

declare <8 x i16> @llvm.x86.avx512.mask.pmulhu.w.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)

define <8 x i16>@test_int_x86_avx512_mask_pmulhu_w_128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_pmulhu_w_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpmulhuw %xmm1, %xmm0, %xmm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0xe4,0xd1]
; CHECK-NEXT:    vpmulhuw %xmm1, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xe4,0xc1]
; CHECK-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 ## encoding: [0x62,0xf1,0x6d,0x08,0xfd,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.mask.pmulhu.w.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3)
  %res1 = call <8 x i16> @llvm.x86.avx512.mask.pmulhu.w.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 -1)
  %res2 = add <8 x i16> %res, %res1
  ret <8 x i16> %res2
}

declare <16 x i16> @llvm.x86.avx512.mask.pmulhu.w.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)

define <16 x i16>@test_int_x86_avx512_mask_pmulhu_w_256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_pmulhu_w_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpmulhuw %ymm1, %ymm0, %ymm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0xe4,0xd1]
; CHECK-NEXT:    vpmulhuw %ymm1, %ymm0, %ymm0 ## encoding: [0x62,0xf1,0x7d,0x28,0xe4,0xc1]
; CHECK-NEXT:    vpaddw %ymm0, %ymm2, %ymm0 ## encoding: [0x62,0xf1,0x6d,0x28,0xfd,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.mask.pmulhu.w.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %x3)
  %res1 = call <16 x i16> @llvm.x86.avx512.mask.pmulhu.w.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 -1)
  %res2 = add <16 x i16> %res, %res1
  ret <16 x i16> %res2
}

declare <8 x i16> @llvm.x86.avx512.mask.pmulh.w.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)

define <8 x i16>@test_int_x86_avx512_mask_pmulh_w_128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_pmulh_w_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpmulhw %xmm1, %xmm0, %xmm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0xe5,0xd1]
; CHECK-NEXT:    vpmulhw %xmm1, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xe5,0xc1]
; CHECK-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 ## encoding: [0x62,0xf1,0x6d,0x08,0xfd,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.mask.pmulh.w.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3)
  %res1 = call <8 x i16> @llvm.x86.avx512.mask.pmulh.w.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 -1)
  %res2 = add <8 x i16> %res, %res1
  ret <8 x i16> %res2
}

declare <16 x i16> @llvm.x86.avx512.mask.pmulh.w.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)

define <16 x i16>@test_int_x86_avx512_mask_pmulh_w_256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_pmulh_w_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpmulhw %ymm1, %ymm0, %ymm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0xe5,0xd1]
; CHECK-NEXT:    vpmulhw %ymm1, %ymm0, %ymm0 ## encoding: [0x62,0xf1,0x7d,0x28,0xe5,0xc1]
; CHECK-NEXT:    vpaddw %ymm0, %ymm2, %ymm0 ## encoding: [0x62,0xf1,0x6d,0x28,0xfd,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.mask.pmulh.w.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %x3)
  %res1 = call <16 x i16> @llvm.x86.avx512.mask.pmulh.w.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 -1)
  %res2 = add <16 x i16> %res, %res1
  ret <16 x i16> %res2
}

declare <8 x i16> @llvm.x86.avx512.mask.pmul.hr.sw.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)

define <8 x i16>@test_int_x86_avx512_mask_pmulhr_sw_128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_pmulhr_sw_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpmulhrsw %xmm1, %xmm0, %xmm2 {%k1} ## encoding: [0x62,0xf2,0x7d,0x09,0x0b,0xd1]
; CHECK-NEXT:    vpmulhrsw %xmm1, %xmm0, %xmm0 ## encoding: [0x62,0xf2,0x7d,0x08,0x0b,0xc1]
; CHECK-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 ## encoding: [0x62,0xf1,0x6d,0x08,0xfd,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.mask.pmul.hr.sw.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3)
  %res1 = call <8 x i16> @llvm.x86.avx512.mask.pmul.hr.sw.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 -1)
  %res2 = add <8 x i16> %res, %res1
  ret <8 x i16> %res2
}

declare <16 x i16> @llvm.x86.avx512.mask.pmul.hr.sw.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)

define <16 x i16>@test_int_x86_avx512_mask_pmulhr_sw_256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_pmulhr_sw_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpmulhrsw %ymm1, %ymm0, %ymm2 {%k1} ## encoding: [0x62,0xf2,0x7d,0x29,0x0b,0xd1]
; CHECK-NEXT:    vpmulhrsw %ymm1, %ymm0, %ymm0 ## encoding: [0x62,0xf2,0x7d,0x28,0x0b,0xc1]
; CHECK-NEXT:    vpaddw %ymm0, %ymm2, %ymm0 ## encoding: [0x62,0xf1,0x6d,0x28,0xfd,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.mask.pmul.hr.sw.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %x3)
  %res1 = call <16 x i16> @llvm.x86.avx512.mask.pmul.hr.sw.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 -1)
  %res2 = add <16 x i16> %res, %res1
  ret <16 x i16> %res2
}

declare <16 x i8> @llvm.x86.avx512.mask.pmov.wb.128(<8 x i16>, <16 x i8>, i8)

define <16 x i8>@test_int_x86_avx512_mask_pmov_wb_128(<8 x i16> %x0, <16 x i8> %x1, i8 %x2) {
; CHECK-LABEL: test_int_x86_avx512_mask_pmov_wb_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpmovwb %xmm0, %xmm1 {%k1} ## encoding: [0x62,0xf2,0x7e,0x09,0x30,0xc1]
; CHECK-NEXT:    vpmovwb %xmm0, %xmm2 {%k1} {z} ## encoding: [0x62,0xf2,0x7e,0x89,0x30,0xc2]
; CHECK-NEXT:    vpmovwb %xmm0, %xmm0 ## encoding: [0x62,0xf2,0x7e,0x08,0x30,0xc0]
; CHECK-NEXT:    vpaddb %xmm1, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xfc,0xc1]
; CHECK-NEXT:    vpaddb %xmm2, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xfc,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
    %res0 = call <16 x i8> @llvm.x86.avx512.mask.pmov.wb.128(<8 x i16> %x0, <16 x i8> %x1, i8 -1)
    %res1 = call <16 x i8> @llvm.x86.avx512.mask.pmov.wb.128(<8 x i16> %x0, <16 x i8> %x1, i8 %x2)
    %res2 = call <16 x i8> @llvm.x86.avx512.mask.pmov.wb.128(<8 x i16> %x0, <16 x i8> zeroinitializer, i8 %x2)
    %res3 = add <16 x i8> %res0, %res1
    %res4 = add <16 x i8> %res3, %res2
    ret <16 x i8> %res4
}

declare void @llvm.x86.avx512.mask.pmov.wb.mem.128(i8* %ptr, <8 x i16>, i8)

define void @test_int_x86_avx512_mask_pmov_wb_mem_128(i8* %ptr, <8 x i16> %x1, i8 %x2) {
; CHECK-LABEL: test_int_x86_avx512_mask_pmov_wb_mem_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpmovwb %xmm0, (%rdi) ## encoding: [0x62,0xf2,0x7e,0x08,0x30,0x07]
; CHECK-NEXT:    vpmovwb %xmm0, (%rdi) {%k1} ## encoding: [0x62,0xf2,0x7e,0x09,0x30,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
    call void @llvm.x86.avx512.mask.pmov.wb.mem.128(i8* %ptr, <8 x i16> %x1, i8 -1)
    call void @llvm.x86.avx512.mask.pmov.wb.mem.128(i8* %ptr, <8 x i16> %x1, i8 %x2)
    ret void
}

declare <16 x i8> @llvm.x86.avx512.mask.pmovs.wb.128(<8 x i16>, <16 x i8>, i8)

define <16 x i8>@test_int_x86_avx512_mask_pmovs_wb_128(<8 x i16> %x0, <16 x i8> %x1, i8 %x2) {
; CHECK-LABEL: test_int_x86_avx512_mask_pmovs_wb_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpmovswb %xmm0, %xmm1 {%k1} ## encoding: [0x62,0xf2,0x7e,0x09,0x20,0xc1]
; CHECK-NEXT:    vpmovswb %xmm0, %xmm2 {%k1} {z} ## encoding: [0x62,0xf2,0x7e,0x89,0x20,0xc2]
; CHECK-NEXT:    vpmovswb %xmm0, %xmm0 ## encoding: [0x62,0xf2,0x7e,0x08,0x20,0xc0]
; CHECK-NEXT:    vpaddb %xmm1, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xfc,0xc1]
; CHECK-NEXT:    vpaddb %xmm2, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xfc,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
    %res0 = call <16 x i8> @llvm.x86.avx512.mask.pmovs.wb.128(<8 x i16> %x0, <16 x i8> %x1, i8 -1)
    %res1 = call <16 x i8> @llvm.x86.avx512.mask.pmovs.wb.128(<8 x i16> %x0, <16 x i8> %x1, i8 %x2)
    %res2 = call <16 x i8> @llvm.x86.avx512.mask.pmovs.wb.128(<8 x i16> %x0, <16 x i8> zeroinitializer, i8 %x2)
    %res3 = add <16 x i8> %res0, %res1
    %res4 = add <16 x i8> %res3, %res2
    ret <16 x i8> %res4
}

declare void @llvm.x86.avx512.mask.pmovs.wb.mem.128(i8* %ptr, <8 x i16>, i8)

define void @test_int_x86_avx512_mask_pmovs_wb_mem_128(i8* %ptr, <8 x i16> %x1, i8 %x2) {
; CHECK-LABEL: test_int_x86_avx512_mask_pmovs_wb_mem_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpmovswb %xmm0, (%rdi) ## encoding: [0x62,0xf2,0x7e,0x08,0x20,0x07]
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpmovswb %xmm0, (%rdi) {%k1} ## encoding: [0x62,0xf2,0x7e,0x09,0x20,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
    call void @llvm.x86.avx512.mask.pmovs.wb.mem.128(i8* %ptr, <8 x i16> %x1, i8 -1)
    call void @llvm.x86.avx512.mask.pmovs.wb.mem.128(i8* %ptr, <8 x i16> %x1, i8 %x2)
    ret void
}

declare <16 x i8> @llvm.x86.avx512.mask.pmovus.wb.128(<8 x i16>, <16 x i8>, i8)

define <16 x i8>@test_int_x86_avx512_mask_pmovus_wb_128(<8 x i16> %x0, <16 x i8> %x1, i8 %x2) {
; CHECK-LABEL: test_int_x86_avx512_mask_pmovus_wb_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpmovuswb %xmm0, %xmm1 {%k1} ## encoding: [0x62,0xf2,0x7e,0x09,0x10,0xc1]
; CHECK-NEXT:    vpmovuswb %xmm0, %xmm2 {%k1} {z} ## encoding: [0x62,0xf2,0x7e,0x89,0x10,0xc2]
; CHECK-NEXT:    vpmovuswb %xmm0, %xmm0 ## encoding: [0x62,0xf2,0x7e,0x08,0x10,0xc0]
; CHECK-NEXT:    vpaddb %xmm1, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xfc,0xc1]
; CHECK-NEXT:    vpaddb %xmm2, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xfc,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
    %res0 = call <16 x i8> @llvm.x86.avx512.mask.pmovus.wb.128(<8 x i16> %x0, <16 x i8> %x1, i8 -1)
    %res1 = call <16 x i8> @llvm.x86.avx512.mask.pmovus.wb.128(<8 x i16> %x0, <16 x i8> %x1, i8 %x2)
    %res2 = call <16 x i8> @llvm.x86.avx512.mask.pmovus.wb.128(<8 x i16> %x0, <16 x i8> zeroinitializer, i8 %x2)
    %res3 = add <16 x i8> %res0, %res1
    %res4 = add <16 x i8> %res3, %res2
    ret <16 x i8> %res4
}

declare void @llvm.x86.avx512.mask.pmovus.wb.mem.128(i8* %ptr, <8 x i16>, i8)

define void @test_int_x86_avx512_mask_pmovus_wb_mem_128(i8* %ptr, <8 x i16> %x1, i8 %x2) {
; CHECK-LABEL: test_int_x86_avx512_mask_pmovus_wb_mem_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpmovuswb %xmm0, (%rdi) ## encoding: [0x62,0xf2,0x7e,0x08,0x10,0x07]
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpmovuswb %xmm0, (%rdi) {%k1} ## encoding: [0x62,0xf2,0x7e,0x09,0x10,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
    call void @llvm.x86.avx512.mask.pmovus.wb.mem.128(i8* %ptr, <8 x i16> %x1, i8 -1)
    call void @llvm.x86.avx512.mask.pmovus.wb.mem.128(i8* %ptr, <8 x i16> %x1, i8 %x2)
    ret void
}

declare <16 x i8> @llvm.x86.avx512.mask.pmov.wb.256(<16 x i16>, <16 x i8>, i16)

define <16 x i8>@test_int_x86_avx512_mask_pmov_wb_256(<16 x i16> %x0, <16 x i8> %x1, i16 %x2) {
; CHECK-LABEL: test_int_x86_avx512_mask_pmov_wb_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpmovwb %ymm0, %xmm1 {%k1} ## encoding: [0x62,0xf2,0x7e,0x29,0x30,0xc1]
; CHECK-NEXT:    vpmovwb %ymm0, %xmm2 {%k1} {z} ## encoding: [0x62,0xf2,0x7e,0xa9,0x30,0xc2]
; CHECK-NEXT:    vpmovwb %ymm0, %xmm0 ## encoding: [0x62,0xf2,0x7e,0x28,0x30,0xc0]
; CHECK-NEXT:    vpaddb %xmm1, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xfc,0xc1]
; CHECK-NEXT:    vpaddb %xmm2, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xfc,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
    %res0 = call <16 x i8> @llvm.x86.avx512.mask.pmov.wb.256(<16 x i16> %x0, <16 x i8> %x1, i16 -1)
    %res1 = call <16 x i8> @llvm.x86.avx512.mask.pmov.wb.256(<16 x i16> %x0, <16 x i8> %x1, i16 %x2)
    %res2 = call <16 x i8> @llvm.x86.avx512.mask.pmov.wb.256(<16 x i16> %x0, <16 x i8> zeroinitializer, i16 %x2)
    %res3 = add <16 x i8> %res0, %res1
    %res4 = add <16 x i8> %res3, %res2
    ret <16 x i8> %res4
}

declare void @llvm.x86.avx512.mask.pmov.wb.mem.256(i8* %ptr, <16 x i16>, i16)

define void @test_int_x86_avx512_mask_pmov_wb_mem_256(i8* %ptr, <16 x i16> %x1, i16 %x2) {
; CHECK-LABEL: test_int_x86_avx512_mask_pmov_wb_mem_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpmovwb %ymm0, (%rdi) ## encoding: [0x62,0xf2,0x7e,0x28,0x30,0x07]
; CHECK-NEXT:    vpmovwb %ymm0, (%rdi) {%k1} ## encoding: [0x62,0xf2,0x7e,0x29,0x30,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
    call void @llvm.x86.avx512.mask.pmov.wb.mem.256(i8* %ptr, <16 x i16> %x1, i16 -1)
    call void @llvm.x86.avx512.mask.pmov.wb.mem.256(i8* %ptr, <16 x i16> %x1, i16 %x2)
    ret void
}

declare <16 x i8> @llvm.x86.avx512.mask.pmovs.wb.256(<16 x i16>, <16 x i8>, i16)

define <16 x i8>@test_int_x86_avx512_mask_pmovs_wb_256(<16 x i16> %x0, <16 x i8> %x1, i16 %x2) {
; CHECK-LABEL: test_int_x86_avx512_mask_pmovs_wb_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpmovswb %ymm0, %xmm1 {%k1} ## encoding: [0x62,0xf2,0x7e,0x29,0x20,0xc1]
; CHECK-NEXT:    vpmovswb %ymm0, %xmm2 {%k1} {z} ## encoding: [0x62,0xf2,0x7e,0xa9,0x20,0xc2]
; CHECK-NEXT:    vpmovswb %ymm0, %xmm0 ## encoding: [0x62,0xf2,0x7e,0x28,0x20,0xc0]
; CHECK-NEXT:    vpaddb %xmm1, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xfc,0xc1]
; CHECK-NEXT:    vpaddb %xmm2, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xfc,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
    %res0 = call <16 x i8> @llvm.x86.avx512.mask.pmovs.wb.256(<16 x i16> %x0, <16 x i8> %x1, i16 -1)
    %res1 = call <16 x i8> @llvm.x86.avx512.mask.pmovs.wb.256(<16 x i16> %x0, <16 x i8> %x1, i16 %x2)
    %res2 = call <16 x i8> @llvm.x86.avx512.mask.pmovs.wb.256(<16 x i16> %x0, <16 x i8> zeroinitializer, i16 %x2)
    %res3 = add <16 x i8> %res0, %res1
    %res4 = add <16 x i8> %res3, %res2
    ret <16 x i8> %res4
}

declare void @llvm.x86.avx512.mask.pmovs.wb.mem.256(i8* %ptr, <16 x i16>, i16)

define void @test_int_x86_avx512_mask_pmovs_wb_mem_256(i8* %ptr, <16 x i16> %x1, i16 %x2) {
; CHECK-LABEL: test_int_x86_avx512_mask_pmovs_wb_mem_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpmovswb %ymm0, (%rdi) ## encoding: [0x62,0xf2,0x7e,0x28,0x20,0x07]
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpmovswb %ymm0, (%rdi) {%k1} ## encoding: [0x62,0xf2,0x7e,0x29,0x20,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
    call void @llvm.x86.avx512.mask.pmovs.wb.mem.256(i8* %ptr, <16 x i16> %x1, i16 -1)
    call void @llvm.x86.avx512.mask.pmovs.wb.mem.256(i8* %ptr, <16 x i16> %x1, i16 %x2)
    ret void
}

declare <16 x i8> @llvm.x86.avx512.mask.pmovus.wb.256(<16 x i16>, <16 x i8>, i16)

define <16 x i8>@test_int_x86_avx512_mask_pmovus_wb_256(<16 x i16> %x0, <16 x i8> %x1, i16 %x2) {
; CHECK-LABEL: test_int_x86_avx512_mask_pmovus_wb_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpmovuswb %ymm0, %xmm1 {%k1} ## encoding: [0x62,0xf2,0x7e,0x29,0x10,0xc1]
; CHECK-NEXT:    vpmovuswb %ymm0, %xmm2 {%k1} {z} ## encoding: [0x62,0xf2,0x7e,0xa9,0x10,0xc2]
; CHECK-NEXT:    vpmovuswb %ymm0, %xmm0 ## encoding: [0x62,0xf2,0x7e,0x28,0x10,0xc0]
; CHECK-NEXT:    vpaddb %xmm1, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xfc,0xc1]
; CHECK-NEXT:    vpaddb %xmm2, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xfc,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
    %res0 = call <16 x i8> @llvm.x86.avx512.mask.pmovus.wb.256(<16 x i16> %x0, <16 x i8> %x1, i16 -1)
    %res1 = call <16 x i8> @llvm.x86.avx512.mask.pmovus.wb.256(<16 x i16> %x0, <16 x i8> %x1, i16 %x2)
    %res2 = call <16 x i8> @llvm.x86.avx512.mask.pmovus.wb.256(<16 x i16> %x0, <16 x i8> zeroinitializer, i16 %x2)
    %res3 = add <16 x i8> %res0, %res1
    %res4 = add <16 x i8> %res3, %res2
    ret <16 x i8> %res4
}

declare void @llvm.x86.avx512.mask.pmovus.wb.mem.256(i8* %ptr, <16 x i16>, i16)

define void @test_int_x86_avx512_mask_pmovus_wb_mem_256(i8* %ptr, <16 x i16> %x1, i16 %x2) {
; CHECK-LABEL: test_int_x86_avx512_mask_pmovus_wb_mem_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpmovuswb %ymm0, (%rdi) ## encoding: [0x62,0xf2,0x7e,0x28,0x10,0x07]
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpmovuswb %ymm0, (%rdi) {%k1} ## encoding: [0x62,0xf2,0x7e,0x29,0x10,0x07]
; CHECK-NEXT:    retq ## encoding: [0xc3]
    call void @llvm.x86.avx512.mask.pmovus.wb.mem.256(i8* %ptr, <16 x i16> %x1, i16 -1)
    call void @llvm.x86.avx512.mask.pmovus.wb.mem.256(i8* %ptr, <16 x i16> %x1, i16 %x2)
    ret void
}

declare <4 x i32> @llvm.x86.avx512.mask.pmaddw.d.128(<8 x i16>, <8 x i16>, <4 x i32>, i8)

define <4 x i32>@test_int_x86_avx512_mask_pmaddw_d_128(<8 x i16> %x0, <8 x i16> %x1, <4 x i32> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_pmaddw_d_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpmaddwd %xmm1, %xmm0, %xmm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0xf5,0xd1]
; CHECK-NEXT:    vpmaddwd %xmm1, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xf5,0xc1]
; CHECK-NEXT:    vpaddd %xmm0, %xmm2, %xmm0 ## encoding: [0x62,0xf1,0x6d,0x08,0xfe,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <4 x i32> @llvm.x86.avx512.mask.pmaddw.d.128(<8 x i16> %x0, <8 x i16> %x1, <4 x i32> %x2, i8 %x3)
  %res1 = call <4 x i32> @llvm.x86.avx512.mask.pmaddw.d.128(<8 x i16> %x0, <8 x i16> %x1, <4 x i32> %x2, i8 -1)
  %res2 = add <4 x i32> %res, %res1
  ret <4 x i32> %res2
}

declare <8 x i32> @llvm.x86.avx512.mask.pmaddw.d.256(<16 x i16>, <16 x i16>, <8 x i32>, i8)

define <8 x i32>@test_int_x86_avx512_mask_pmaddw_d_256(<16 x i16> %x0, <16 x i16> %x1, <8 x i32> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_pmaddw_d_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpmaddwd %ymm1, %ymm0, %ymm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0xf5,0xd1]
; CHECK-NEXT:    vpmaddwd %ymm1, %ymm0, %ymm0 ## encoding: [0x62,0xf1,0x7d,0x28,0xf5,0xc1]
; CHECK-NEXT:    vpaddd %ymm0, %ymm2, %ymm0 ## encoding: [0x62,0xf1,0x6d,0x28,0xfe,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x i32> @llvm.x86.avx512.mask.pmaddw.d.256(<16 x i16> %x0, <16 x i16> %x1, <8 x i32> %x2, i8 %x3)
  %res1 = call <8 x i32> @llvm.x86.avx512.mask.pmaddw.d.256(<16 x i16> %x0, <16 x i16> %x1, <8 x i32> %x2, i8 -1)
  %res2 = add <8 x i32> %res, %res1
  ret <8 x i32> %res2
}

declare <8 x i16> @llvm.x86.avx512.mask.pmaddubs.w.128(<16 x i8>, <16 x i8>, <8 x i16>, i8)

define <8 x i16>@test_int_x86_avx512_mask_pmaddubs_w_128(<16 x i8> %x0, <16 x i8> %x1, <8 x i16> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_pmaddubs_w_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpmaddubsw %xmm1, %xmm0, %xmm2 {%k1} ## encoding: [0x62,0xf2,0x7d,0x09,0x04,0xd1]
; CHECK-NEXT:    vpmaddubsw %xmm1, %xmm0, %xmm0 ## encoding: [0x62,0xf2,0x7d,0x08,0x04,0xc1]
; CHECK-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 ## encoding: [0x62,0xf1,0x6d,0x08,0xfd,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.mask.pmaddubs.w.128(<16 x i8> %x0, <16 x i8> %x1, <8 x i16> %x2, i8 %x3)
  %res1 = call <8 x i16> @llvm.x86.avx512.mask.pmaddubs.w.128(<16 x i8> %x0, <16 x i8> %x1, <8 x i16> %x2, i8 -1)
  %res2 = add <8 x i16> %res, %res1
  ret <8 x i16> %res2
}

declare <16 x i16> @llvm.x86.avx512.mask.pmaddubs.w.256(<32 x i8>, <32 x i8>, <16 x i16>, i16)

define <16 x i16>@test_int_x86_avx512_mask_pmaddubs_w_256(<32 x i8> %x0, <32 x i8> %x1, <16 x i16> %x2, i16 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_pmaddubs_w_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpmaddubsw %ymm1, %ymm0, %ymm2 {%k1} ## encoding: [0x62,0xf2,0x7d,0x29,0x04,0xd1]
; CHECK-NEXT:    vpmaddubsw %ymm1, %ymm0, %ymm0 ## encoding: [0x62,0xf2,0x7d,0x28,0x04,0xc1]
; CHECK-NEXT:    vpaddw %ymm0, %ymm2, %ymm0 ## encoding: [0x62,0xf1,0x6d,0x28,0xfd,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.mask.pmaddubs.w.256(<32 x i8> %x0, <32 x i8> %x1, <16 x i16> %x2, i16 %x3)
  %res1 = call <16 x i16> @llvm.x86.avx512.mask.pmaddubs.w.256(<32 x i8> %x0, <32 x i8> %x1, <16 x i16> %x2, i16 -1)
  %res2 = add <16 x i16> %res, %res1
  ret <16 x i16> %res2
}

declare <8 x i16> @llvm.x86.avx512.mask.dbpsadbw.128(<16 x i8>, <16 x i8>, i32, <8 x i16>, i8)

define <8 x i16>@test_int_x86_avx512_mask_dbpsadbw_128(<16 x i8> %x0, <16 x i8> %x1, <8 x i16> %x3, i8 %x4) {
; CHECK-LABEL: test_int_x86_avx512_mask_dbpsadbw_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vdbpsadbw $2, %xmm1, %xmm0, %xmm2 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x42,0xd1,0x02]
; CHECK-NEXT:    vdbpsadbw $2, %xmm1, %xmm0, %xmm3 {%k1} {z} ## encoding: [0x62,0xf3,0x7d,0x89,0x42,0xd9,0x02]
; CHECK-NEXT:    vdbpsadbw $2, %xmm1, %xmm0, %xmm0 ## encoding: [0x62,0xf3,0x7d,0x08,0x42,0xc1,0x02]
; CHECK-NEXT:    vpaddw %xmm3, %xmm2, %xmm1 ## encoding: [0x62,0xf1,0x6d,0x08,0xfd,0xcb]
; CHECK-NEXT:    vpaddw %xmm1, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xfd,0xc1]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.mask.dbpsadbw.128(<16 x i8> %x0, <16 x i8> %x1, i32 2, <8 x i16> %x3, i8 %x4)
  %res1 = call <8 x i16> @llvm.x86.avx512.mask.dbpsadbw.128(<16 x i8> %x0, <16 x i8> %x1, i32 2, <8 x i16> zeroinitializer, i8 %x4)
  %res2 = call <8 x i16> @llvm.x86.avx512.mask.dbpsadbw.128(<16 x i8> %x0, <16 x i8> %x1, i32 2, <8 x i16> %x3, i8 -1)
  %res3 = add <8 x i16> %res, %res1
  %res4 = add <8 x i16> %res2, %res3
  ret <8 x i16> %res4
}

declare <16 x i16> @llvm.x86.avx512.mask.dbpsadbw.256(<32 x i8>, <32 x i8>, i32, <16 x i16>, i16)

define <16 x i16>@test_int_x86_avx512_mask_dbpsadbw_256(<32 x i8> %x0, <32 x i8> %x1, <16 x i16> %x3, i16 %x4) {
; CHECK-LABEL: test_int_x86_avx512_mask_dbpsadbw_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vdbpsadbw $2, %ymm1, %ymm0, %ymm2 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x42,0xd1,0x02]
; CHECK-NEXT:    vdbpsadbw $2, %ymm1, %ymm0, %ymm3 {%k1} {z} ## encoding: [0x62,0xf3,0x7d,0xa9,0x42,0xd9,0x02]
; CHECK-NEXT:    vdbpsadbw $2, %ymm1, %ymm0, %ymm0 ## encoding: [0x62,0xf3,0x7d,0x28,0x42,0xc1,0x02]
; CHECK-NEXT:    vpaddw %ymm3, %ymm2, %ymm1 ## encoding: [0x62,0xf1,0x6d,0x28,0xfd,0xcb]
; CHECK-NEXT:    vpaddw %ymm0, %ymm1, %ymm0 ## encoding: [0x62,0xf1,0x75,0x28,0xfd,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.mask.dbpsadbw.256(<32 x i8> %x0, <32 x i8> %x1, i32 2, <16 x i16> %x3, i16 %x4)
  %res1 = call <16 x i16> @llvm.x86.avx512.mask.dbpsadbw.256(<32 x i8> %x0, <32 x i8> %x1, i32 2, <16 x i16> zeroinitializer, i16 %x4)
  %res2 = call <16 x i16> @llvm.x86.avx512.mask.dbpsadbw.256(<32 x i8> %x0, <32 x i8> %x1, i32 2, <16 x i16> %x3, i16 -1)
  %res3 = add <16 x i16> %res, %res1
  %res4 = add <16 x i16> %res3, %res2
  ret <16 x i16> %res4
}

declare i16 @llvm.x86.avx512.cvtb2mask.128(<16 x i8>)

define i16@test_int_x86_avx512_cvtb2mask_128(<16 x i8> %x0) {
; CHECK-LABEL: test_int_x86_avx512_cvtb2mask_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpmovb2m %xmm0, %k0 ## encoding: [0x62,0xf2,0x7e,0x08,0x29,0xc0]
; CHECK-NEXT:    kmovw %k0, %eax ## encoding: [0xc5,0xf8,0x93,0xc0]
; CHECK-NEXT:    ## kill: %AX<def> %AX<kill> %EAX<kill>
; CHECK-NEXT:    retq ## encoding: [0xc3]
    %res = call i16 @llvm.x86.avx512.cvtb2mask.128(<16 x i8> %x0)
    ret i16 %res
}

declare i32 @llvm.x86.avx512.cvtb2mask.256(<32 x i8>)

define i32@test_int_x86_avx512_cvtb2mask_256(<32 x i8> %x0) {
; CHECK-LABEL: test_int_x86_avx512_cvtb2mask_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpmovb2m %ymm0, %k0 ## encoding: [0x62,0xf2,0x7e,0x28,0x29,0xc0]
; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
    %res = call i32 @llvm.x86.avx512.cvtb2mask.256(<32 x i8> %x0)
    ret i32 %res
}

declare i8 @llvm.x86.avx512.cvtw2mask.128(<8 x i16>)

define i8@test_int_x86_avx512_cvtw2mask_128(<8 x i16> %x0) {
; CHECK-LABEL: test_int_x86_avx512_cvtw2mask_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpmovw2m %xmm0, %k0 ## encoding: [0x62,0xf2,0xfe,0x08,0x29,0xc0]
; CHECK-NEXT:    kmovw %k0, %eax ## encoding: [0xc5,0xf8,0x93,0xc0]
; CHECK-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
; CHECK-NEXT:    retq ## encoding: [0xc3]
    %res = call i8 @llvm.x86.avx512.cvtw2mask.128(<8 x i16> %x0)
    ret i8 %res
}

declare i16 @llvm.x86.avx512.cvtw2mask.256(<16 x i16>)

define i16@test_int_x86_avx512_cvtw2mask_256(<16 x i16> %x0) {
; CHECK-LABEL: test_int_x86_avx512_cvtw2mask_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    vpmovw2m %ymm0, %k0 ## encoding: [0x62,0xf2,0xfe,0x28,0x29,0xc0]
; CHECK-NEXT:    kmovw %k0, %eax ## encoding: [0xc5,0xf8,0x93,0xc0]
; CHECK-NEXT:    ## kill: %AX<def> %AX<kill> %EAX<kill>
; CHECK-NEXT:    retq ## encoding: [0xc3]
    %res = call i16 @llvm.x86.avx512.cvtw2mask.256(<16 x i16> %x0)
    ret i16 %res
}

declare <16 x i8> @llvm.x86.avx512.cvtmask2b.128(i16)

define <16 x i8>@test_int_x86_avx512_cvtmask2b_128(i16 %x0) {
; CHECK-LABEL: test_int_x86_avx512_cvtmask2b_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k0 ## encoding: [0xc5,0xf8,0x92,0xc7]
; CHECK-NEXT:    vpmovm2b %k0, %xmm0 ## encoding: [0x62,0xf2,0x7e,0x08,0x28,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i8> @llvm.x86.avx512.cvtmask2b.128(i16 %x0)
  ret <16 x i8> %res
}

declare <32 x i8> @llvm.x86.avx512.cvtmask2b.256(i32)

define <32 x i8>@test_int_x86_avx512_cvtmask2b_256(i32 %x0) {
; CHECK-LABEL: test_int_x86_avx512_cvtmask2b_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovd %edi, %k0 ## encoding: [0xc5,0xfb,0x92,0xc7]
; CHECK-NEXT:    vpmovm2b %k0, %ymm0 ## encoding: [0x62,0xf2,0x7e,0x28,0x28,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <32 x i8> @llvm.x86.avx512.cvtmask2b.256(i32 %x0)
  ret <32 x i8> %res
}

declare <8 x i16> @llvm.x86.avx512.cvtmask2w.128(i8)

define <8 x i16>@test_int_x86_avx512_cvtmask2w_128(i8 %x0) {
; CHECK-LABEL: test_int_x86_avx512_cvtmask2w_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k0 ## encoding: [0xc5,0xf8,0x92,0xc7]
; CHECK-NEXT:    vpmovm2w %k0, %xmm0 ## encoding: [0x62,0xf2,0xfe,0x08,0x28,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.cvtmask2w.128(i8 %x0)
  ret <8 x i16> %res
}

declare <16 x i16> @llvm.x86.avx512.cvtmask2w.256(i16)

define <16 x i16>@test_int_x86_avx512_cvtmask2w_256(i16 %x0) {
; CHECK-LABEL: test_int_x86_avx512_cvtmask2w_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k0 ## encoding: [0xc5,0xf8,0x92,0xc7]
; CHECK-NEXT:    vpmovm2w %k0, %ymm0 ## encoding: [0x62,0xf2,0xfe,0x28,0x28,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.cvtmask2w.256(i16 %x0)
  ret <16 x i16> %res
}

declare <8 x i16> @llvm.x86.avx512.mask.psrl.w.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)

define <8 x i16>@test_int_x86_avx512_mask_psrl_w_128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_psrl_w_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpsrlw %xmm1, %xmm0, %xmm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0xd1,0xd1]
; CHECK-NEXT:    vpsrlw %xmm1, %xmm0, %xmm3 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0x89,0xd1,0xd9]
; CHECK-NEXT:    vpsrlw %xmm1, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xd1,0xc1]
; CHECK-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 ## encoding: [0x62,0xf1,0x6d,0x08,0xfd,0xc0]
; CHECK-NEXT:    vpaddw %xmm0, %xmm3, %xmm0 ## encoding: [0x62,0xf1,0x65,0x08,0xfd,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.mask.psrl.w.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3)
  %res1 = call <8 x i16> @llvm.x86.avx512.mask.psrl.w.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 -1)
  %res2 = call <8 x i16> @llvm.x86.avx512.mask.psrl.w.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> zeroinitializer, i8 %x3)
  %res3 = add <8 x i16> %res, %res1
  %res4 = add <8 x i16> %res2, %res3
  ret <8 x i16> %res4
}

declare <16 x i16> @llvm.x86.avx512.mask.psrl.w.256(<16 x i16>, <8 x i16>, <16 x i16>, i16)

define <16 x i16>@test_int_x86_avx512_mask_psrl_w_256(<16 x i16> %x0, <8 x i16> %x1, <16 x i16> %x2, i16 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_psrl_w_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpsrlw %xmm1, %ymm0, %ymm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0xd1,0xd1]
; CHECK-NEXT:    vpsrlw %xmm1, %ymm0, %ymm3 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xa9,0xd1,0xd9]
; CHECK-NEXT:    vpsrlw %xmm1, %ymm0, %ymm0 ## encoding: [0x62,0xf1,0x7d,0x28,0xd1,0xc1]
; CHECK-NEXT:    vpaddw %ymm0, %ymm2, %ymm0 ## encoding: [0x62,0xf1,0x6d,0x28,0xfd,0xc0]
; CHECK-NEXT:    vpaddw %ymm3, %ymm0, %ymm0 ## encoding: [0x62,0xf1,0x7d,0x28,0xfd,0xc3]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.mask.psrl.w.256(<16 x i16> %x0, <8 x i16> %x1, <16 x i16> %x2, i16 %x3)
  %res1 = call <16 x i16> @llvm.x86.avx512.mask.psrl.w.256(<16 x i16> %x0, <8 x i16> %x1, <16 x i16> %x2, i16 -1)
  %res2 = call <16 x i16> @llvm.x86.avx512.mask.psrl.w.256(<16 x i16> %x0, <8 x i16> %x1, <16 x i16> zeroinitializer, i16 %x3)
  %res3 = add <16 x i16> %res, %res1
  %res4 = add <16 x i16> %res3, %res2
  ret <16 x i16> %res4
}

declare <8 x i16> @llvm.x86.avx512.mask.psrl.wi.128(<8 x i16>, i32, <8 x i16>, i8)

define <8 x i16>@test_int_x86_avx512_mask_psrl_wi_128(<8 x i16> %x0, i32 %x1, <8 x i16> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_psrl_wi_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpsrlw $3, %xmm0, %xmm1 {%k1} ## encoding: [0x62,0xf1,0x75,0x09,0x71,0xd0,0x03]
; CHECK-NEXT:    vpsrlw $3, %xmm0, %xmm2 {%k1} {z} ## encoding: [0x62,0xf1,0x6d,0x89,0x71,0xd0,0x03]
; CHECK-NEXT:    vpsrlw $3, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0x71,0xd0,0x03]
; CHECK-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 ## encoding: [0x62,0xf1,0x75,0x08,0xfd,0xc0]
; CHECK-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 ## encoding: [0x62,0xf1,0x6d,0x08,0xfd,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.mask.psrl.wi.128(<8 x i16> %x0, i32 3, <8 x i16> %x2, i8 %x3)
  %res1 = call <8 x i16> @llvm.x86.avx512.mask.psrl.wi.128(<8 x i16> %x0, i32 3, <8 x i16> %x2, i8 -1)
  %res2 = call <8 x i16> @llvm.x86.avx512.mask.psrl.wi.128(<8 x i16> %x0, i32 3, <8 x i16> zeroinitializer, i8 %x3)
  %res3 = add <8 x i16> %res, %res1
  %res4 = add <8 x i16> %res2, %res3
  ret <8 x i16> %res4
}

declare <16 x i16> @llvm.x86.avx512.mask.psrl.wi.256(<16 x i16>, i32, <16 x i16>, i16)

define <16 x i16>@test_int_x86_avx512_mask_psrl_wi_256(<16 x i16> %x0, i32 %x1, <16 x i16> %x2, i16 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_psrl_wi_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpsrlw $3, %ymm0, %ymm1 {%k1} ## encoding: [0x62,0xf1,0x75,0x29,0x71,0xd0,0x03]
; CHECK-NEXT:    vpsrlw $3, %ymm0, %ymm2 {%k1} {z} ## encoding: [0x62,0xf1,0x6d,0xa9,0x71,0xd0,0x03]
; CHECK-NEXT:    vpsrlw $3, %ymm0, %ymm0 ## encoding: [0x62,0xf1,0x7d,0x28,0x71,0xd0,0x03]
; CHECK-NEXT:    vpaddw %ymm0, %ymm1, %ymm0 ## encoding: [0x62,0xf1,0x75,0x28,0xfd,0xc0]
; CHECK-NEXT:    vpaddw %ymm2, %ymm0, %ymm0 ## encoding: [0x62,0xf1,0x7d,0x28,0xfd,0xc2]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.mask.psrl.wi.256(<16 x i16> %x0, i32 3, <16 x i16> %x2, i16 %x3)
  %res1 = call <16 x i16> @llvm.x86.avx512.mask.psrl.wi.256(<16 x i16> %x0, i32 3, <16 x i16> %x2, i16 -1)
  %res2 = call <16 x i16> @llvm.x86.avx512.mask.psrl.wi.256(<16 x i16> %x0, i32 3, <16 x i16> zeroinitializer, i16 %x3)
  %res3 = add <16 x i16> %res, %res1
  %res4 = add <16 x i16> %res3, %res2
  ret <16 x i16> %res4
}

declare <16 x i16> @llvm.x86.avx512.mask.psrlv16.hi(<16 x i16>, <16 x i16>, <16 x i16>, i16)

define <16 x i16>@test_int_x86_avx512_mask_psrlv16_hi(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_psrlv16_hi:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpsrlvw %ymm1, %ymm0, %ymm2 {%k1} ## encoding: [0x62,0xf2,0xfd,0x29,0x10,0xd1]
; CHECK-NEXT:    vpsrlvw %ymm1, %ymm0, %ymm3 {%k1} {z} ## encoding: [0x62,0xf2,0xfd,0xa9,0x10,0xd9]
; CHECK-NEXT:    vpsrlvw %ymm1, %ymm0, %ymm0 ## encoding: [0x62,0xf2,0xfd,0x28,0x10,0xc1]
; CHECK-NEXT:    vpaddw %ymm3, %ymm2, %ymm1 ## encoding: [0x62,0xf1,0x6d,0x28,0xfd,0xcb]
; CHECK-NEXT:    vpaddw %ymm0, %ymm1, %ymm0 ## encoding: [0x62,0xf1,0x75,0x28,0xfd,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.mask.psrlv16.hi(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %x3)
  %res1 = call <16 x i16> @llvm.x86.avx512.mask.psrlv16.hi(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> zeroinitializer, i16 %x3)
  %res2 = call <16 x i16> @llvm.x86.avx512.mask.psrlv16.hi(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 -1)
  %res3 = add <16 x i16> %res, %res1
  %res4 = add <16 x i16> %res3, %res2
  ret <16 x i16> %res4
}

declare <8 x i16> @llvm.x86.avx512.mask.psrlv8.hi(<8 x i16>, <8 x i16>, <8 x i16>, i8)

define <8 x i16>@test_int_x86_avx512_mask_psrlv8_hi(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_psrlv8_hi:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpsrlvw %xmm1, %xmm0, %xmm2 {%k1} ## encoding: [0x62,0xf2,0xfd,0x09,0x10,0xd1]
; CHECK-NEXT:    vpsrlvw %xmm1, %xmm0, %xmm3 {%k1} {z} ## encoding: [0x62,0xf2,0xfd,0x89,0x10,0xd9]
; CHECK-NEXT:    vpsrlvw %xmm1, %xmm0, %xmm0 ## encoding: [0x62,0xf2,0xfd,0x08,0x10,0xc1]
; CHECK-NEXT:    vpaddw %xmm3, %xmm2, %xmm1 ## encoding: [0x62,0xf1,0x6d,0x08,0xfd,0xcb]
; CHECK-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 ## encoding: [0x62,0xf1,0x75,0x08,0xfd,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.mask.psrlv8.hi(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3)
  %res1 = call <8 x i16> @llvm.x86.avx512.mask.psrlv8.hi(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> zeroinitializer, i8 %x3)
  %res2 = call <8 x i16> @llvm.x86.avx512.mask.psrlv8.hi(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 -1)
  %res3 = add <8 x i16> %res, %res1
  %res4 = add <8 x i16> %res3, %res2
  ret <8 x i16> %res4
}

declare <8 x i16> @llvm.x86.avx512.mask.psra.w.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)

define <8 x i16>@test_int_x86_avx512_mask_psra_w_128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_psra_w_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpsraw %xmm1, %xmm0, %xmm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0xe1,0xd1]
; CHECK-NEXT:    vpsraw %xmm1, %xmm0, %xmm3 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0x89,0xe1,0xd9]
; CHECK-NEXT:    vpsraw %xmm1, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xe1,0xc1]
; CHECK-NEXT:    vpaddw %xmm3, %xmm2, %xmm1 ## encoding: [0x62,0xf1,0x6d,0x08,0xfd,0xcb]
; CHECK-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 ## encoding: [0x62,0xf1,0x75,0x08,0xfd,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.mask.psra.w.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3)
  %res1 = call <8 x i16> @llvm.x86.avx512.mask.psra.w.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> zeroinitializer, i8 %x3)
  %res2 = call <8 x i16> @llvm.x86.avx512.mask.psra.w.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 -1)
  %res3 = add <8 x i16> %res, %res1
  %res4 = add <8 x i16> %res3, %res2
  ret <8 x i16> %res4
}

declare <8 x i16> @llvm.x86.avx512.mask.psra.wi.128(<8 x i16>, i32, <8 x i16>, i8)

define <8 x i16>@test_int_x86_avx512_mask_psra_wi_128(<8 x i16> %x0, i32 %x1, <8 x i16> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_psra_wi_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpsraw $3, %xmm0, %xmm1 {%k1} ## encoding: [0x62,0xf1,0x75,0x09,0x71,0xe0,0x03]
; CHECK-NEXT:    vpsraw $3, %xmm0, %xmm2 {%k1} {z} ## encoding: [0x62,0xf1,0x6d,0x89,0x71,0xe0,0x03]
; CHECK-NEXT:    vpsraw $3, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0x71,0xe0,0x03]
; CHECK-NEXT:    vpaddw %xmm2, %xmm1, %xmm1 ## encoding: [0x62,0xf1,0x75,0x08,0xfd,0xca]
; CHECK-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 ## encoding: [0x62,0xf1,0x75,0x08,0xfd,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.mask.psra.wi.128(<8 x i16> %x0, i32 3, <8 x i16> %x2, i8 %x3)
  %res1 = call <8 x i16> @llvm.x86.avx512.mask.psra.wi.128(<8 x i16> %x0, i32 3, <8 x i16> zeroinitializer, i8 %x3)
  %res2 = call <8 x i16> @llvm.x86.avx512.mask.psra.wi.128(<8 x i16> %x0, i32 3, <8 x i16> %x2, i8 -1)
  %res3 = add <8 x i16> %res, %res1
  %res4 = add <8 x i16> %res3, %res2
  ret <8 x i16> %res4
}

declare <16 x i16> @llvm.x86.avx512.mask.psra.w.256(<16 x i16>, <8 x i16>, <16 x i16>, i16)

define <16 x i16>@test_int_x86_avx512_mask_psra_w_256(<16 x i16> %x0, <8 x i16> %x1, <16 x i16> %x2, i16 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_psra_w_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpsraw %xmm1, %ymm0, %ymm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0xe1,0xd1]
; CHECK-NEXT:    vpsraw %xmm1, %ymm0, %ymm3 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xa9,0xe1,0xd9]
; CHECK-NEXT:    vpsraw %xmm1, %ymm0, %ymm0 ## encoding: [0x62,0xf1,0x7d,0x28,0xe1,0xc1]
; CHECK-NEXT:    vpaddw %ymm3, %ymm2, %ymm1 ## encoding: [0x62,0xf1,0x6d,0x28,0xfd,0xcb]
; CHECK-NEXT:    vpaddw %ymm0, %ymm1, %ymm0 ## encoding: [0x62,0xf1,0x75,0x28,0xfd,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.mask.psra.w.256(<16 x i16> %x0, <8 x i16> %x1, <16 x i16> %x2, i16 %x3)
  %res1 = call <16 x i16> @llvm.x86.avx512.mask.psra.w.256(<16 x i16> %x0, <8 x i16> %x1, <16 x i16> zeroinitializer, i16 %x3)
  %res2 = call <16 x i16> @llvm.x86.avx512.mask.psra.w.256(<16 x i16> %x0, <8 x i16> %x1, <16 x i16> %x2, i16 -1)
  %res3 = add <16 x i16> %res, %res1
  %res4 = add <16 x i16> %res3, %res2
  ret <16 x i16> %res4
}

declare <16 x i16> @llvm.x86.avx512.mask.psra.wi.256(<16 x i16>, i32, <16 x i16>, i16)

define <16 x i16>@test_int_x86_avx512_mask_psra_wi_256(<16 x i16> %x0, i32 %x1, <16 x i16> %x2, i16 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_psra_wi_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpsraw $3, %ymm0, %ymm1 {%k1} ## encoding: [0x62,0xf1,0x75,0x29,0x71,0xe0,0x03]
; CHECK-NEXT:    vpsraw $3, %ymm0, %ymm2 {%k1} {z} ## encoding: [0x62,0xf1,0x6d,0xa9,0x71,0xe0,0x03]
; CHECK-NEXT:    vpsraw $3, %ymm0, %ymm0 ## encoding: [0x62,0xf1,0x7d,0x28,0x71,0xe0,0x03]
; CHECK-NEXT:    vpaddw %ymm2, %ymm1, %ymm1 ## encoding: [0x62,0xf1,0x75,0x28,0xfd,0xca]
; CHECK-NEXT:    vpaddw %ymm0, %ymm1, %ymm0 ## encoding: [0x62,0xf1,0x75,0x28,0xfd,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.mask.psra.wi.256(<16 x i16> %x0, i32 3, <16 x i16> %x2, i16 %x3)
  %res1 = call <16 x i16> @llvm.x86.avx512.mask.psra.wi.256(<16 x i16> %x0, i32 3, <16 x i16> zeroinitializer, i16 %x3)
  %res2 = call <16 x i16> @llvm.x86.avx512.mask.psra.wi.256(<16 x i16> %x0, i32 3, <16 x i16> %x2, i16 -1)
  %res3 = add <16 x i16> %res, %res1
  %res4 = add <16 x i16> %res3, %res2
  ret <16 x i16> %res4
}

declare <16 x i16> @llvm.x86.avx512.mask.psrav16.hi(<16 x i16>, <16 x i16>, <16 x i16>, i16)

define <16 x i16>@test_int_x86_avx512_mask_psrav16_hi(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_psrav16_hi:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpsravw %ymm1, %ymm0, %ymm2 {%k1} ## encoding: [0x62,0xf2,0xfd,0x29,0x11,0xd1]
; CHECK-NEXT:    vpsravw %ymm1, %ymm0, %ymm3 {%k1} {z} ## encoding: [0x62,0xf2,0xfd,0xa9,0x11,0xd9]
; CHECK-NEXT:    vpsravw %ymm1, %ymm0, %ymm0 ## encoding: [0x62,0xf2,0xfd,0x28,0x11,0xc1]
; CHECK-NEXT:    vpaddw %ymm3, %ymm2, %ymm1 ## encoding: [0x62,0xf1,0x6d,0x28,0xfd,0xcb]
; CHECK-NEXT:    vpaddw %ymm0, %ymm1, %ymm0 ## encoding: [0x62,0xf1,0x75,0x28,0xfd,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.mask.psrav16.hi(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %x3)
  %res1 = call <16 x i16> @llvm.x86.avx512.mask.psrav16.hi(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> zeroinitializer, i16 %x3)
  %res2 = call <16 x i16> @llvm.x86.avx512.mask.psrav16.hi(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 -1)
  %res3 = add <16 x i16> %res, %res1
  %res4 = add <16 x i16> %res3, %res2
  ret <16 x i16> %res4
}

declare <8 x i16> @llvm.x86.avx512.mask.psrav8.hi(<8 x i16>, <8 x i16>, <8 x i16>, i8)

define <8 x i16>@test_int_x86_avx512_mask_psrav8_hi(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_psrav8_hi:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpsravw %xmm1, %xmm0, %xmm2 {%k1} ## encoding: [0x62,0xf2,0xfd,0x09,0x11,0xd1]
; CHECK-NEXT:    vpsravw %xmm1, %xmm0, %xmm3 {%k1} {z} ## encoding: [0x62,0xf2,0xfd,0x89,0x11,0xd9]
; CHECK-NEXT:    vpsravw %xmm1, %xmm0, %xmm0 ## encoding: [0x62,0xf2,0xfd,0x08,0x11,0xc1]
; CHECK-NEXT:    vpaddw %xmm3, %xmm2, %xmm1 ## encoding: [0x62,0xf1,0x6d,0x08,0xfd,0xcb]
; CHECK-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 ## encoding: [0x62,0xf1,0x75,0x08,0xfd,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.mask.psrav8.hi(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3)
  %res1 = call <8 x i16> @llvm.x86.avx512.mask.psrav8.hi(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> zeroinitializer, i8 %x3)
  %res2 = call <8 x i16> @llvm.x86.avx512.mask.psrav8.hi(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 -1)
  %res3 = add <8 x i16> %res, %res1
  %res4 = add <8 x i16> %res3, %res2
  ret <8 x i16> %res4
}


declare <8 x i16> @llvm.x86.avx512.mask.psll.w.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)

define <8 x i16>@test_int_x86_avx512_mask_psll_w_128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_psll_w_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpsllw %xmm1, %xmm0, %xmm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x09,0xf1,0xd1]
; CHECK-NEXT:    vpsllw %xmm1, %xmm0, %xmm3 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0x89,0xf1,0xd9]
; CHECK-NEXT:    vpsllw %xmm1, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xf1,0xc1]
; CHECK-NEXT:    vpaddw %xmm3, %xmm2, %xmm1 ## encoding: [0x62,0xf1,0x6d,0x08,0xfd,0xcb]
; CHECK-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 ## encoding: [0x62,0xf1,0x75,0x08,0xfd,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.mask.psll.w.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3)
  %res1 = call <8 x i16> @llvm.x86.avx512.mask.psll.w.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> zeroinitializer, i8 %x3)
  %res2 = call <8 x i16> @llvm.x86.avx512.mask.psll.w.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 -1)
  %res3 = add <8 x i16> %res, %res1
  %res4 = add <8 x i16> %res3, %res2
  ret <8 x i16> %res4
}

declare <16 x i16> @llvm.x86.avx512.mask.psll.w.256(<16 x i16>, <8 x i16>, <16 x i16>, i16)

define <16 x i16>@test_int_x86_avx512_mask_psll_w_256(<16 x i16> %x0, <8 x i16> %x1, <16 x i16> %x2, i16 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_psll_w_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpsllw %xmm1, %ymm0, %ymm2 {%k1} ## encoding: [0x62,0xf1,0x7d,0x29,0xf1,0xd1]
; CHECK-NEXT:    vpsllw %xmm1, %ymm0, %ymm3 {%k1} {z} ## encoding: [0x62,0xf1,0x7d,0xa9,0xf1,0xd9]
; CHECK-NEXT:    vpsllw %xmm1, %ymm0, %ymm0 ## encoding: [0x62,0xf1,0x7d,0x28,0xf1,0xc1]
; CHECK-NEXT:    vpaddw %ymm3, %ymm2, %ymm1 ## encoding: [0x62,0xf1,0x6d,0x28,0xfd,0xcb]
; CHECK-NEXT:    vpaddw %ymm0, %ymm1, %ymm0 ## encoding: [0x62,0xf1,0x75,0x28,0xfd,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.mask.psll.w.256(<16 x i16> %x0, <8 x i16> %x1, <16 x i16> %x2, i16 %x3)
  %res1 = call <16 x i16> @llvm.x86.avx512.mask.psll.w.256(<16 x i16> %x0, <8 x i16> %x1, <16 x i16> zeroinitializer, i16 %x3)
  %res2 = call <16 x i16> @llvm.x86.avx512.mask.psll.w.256(<16 x i16> %x0, <8 x i16> %x1, <16 x i16> %x2, i16 -1)
  %res3 = add <16 x i16> %res, %res1
  %res4 = add <16 x i16> %res3, %res2
  ret <16 x i16> %res4
}

declare <8 x i16> @llvm.x86.avx512.mask.psll.wi.128(<8 x i16>, i32, <8 x i16>, i8)

define <8 x i16>@test_int_x86_avx512_mask_psll_wi_128(<8 x i16> %x0, i32 %x1, <8 x i16> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_psll_wi_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpsllw $3, %xmm0, %xmm1 {%k1} ## encoding: [0x62,0xf1,0x75,0x09,0x71,0xf0,0x03]
; CHECK-NEXT:    vpsllw $3, %xmm0, %xmm2 {%k1} {z} ## encoding: [0x62,0xf1,0x6d,0x89,0x71,0xf0,0x03]
; CHECK-NEXT:    vpsllw $3, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0x71,0xf0,0x03]
; CHECK-NEXT:    vpaddw %xmm2, %xmm1, %xmm1 ## encoding: [0x62,0xf1,0x75,0x08,0xfd,0xca]
; CHECK-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 ## encoding: [0x62,0xf1,0x75,0x08,0xfd,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.mask.psll.wi.128(<8 x i16> %x0, i32 3, <8 x i16> %x2, i8 %x3)
  %res1 = call <8 x i16> @llvm.x86.avx512.mask.psll.wi.128(<8 x i16> %x0, i32 3, <8 x i16> zeroinitializer, i8 %x3)
  %res2 = call <8 x i16> @llvm.x86.avx512.mask.psll.wi.128(<8 x i16> %x0, i32 3, <8 x i16> %x2, i8 -1)
  %res3 = add <8 x i16> %res, %res1
  %res4 = add <8 x i16> %res3, %res2
  ret <8 x i16> %res4
}

declare <16 x i16> @llvm.x86.avx512.mask.psll.wi.256(<16 x i16>, i32, <16 x i16>, i16)

define <16 x i16>@test_int_x86_avx512_mask_psll_wi_256(<16 x i16> %x0, i32 %x1, <16 x i16> %x2, i16 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_psll_wi_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpsllw $3, %ymm0, %ymm1 {%k1} ## encoding: [0x62,0xf1,0x75,0x29,0x71,0xf0,0x03]
; CHECK-NEXT:    vpsllw $3, %ymm0, %ymm2 {%k1} {z} ## encoding: [0x62,0xf1,0x6d,0xa9,0x71,0xf0,0x03]
; CHECK-NEXT:    vpsllw $3, %ymm0, %ymm0 ## encoding: [0x62,0xf1,0x7d,0x28,0x71,0xf0,0x03]
; CHECK-NEXT:    vpaddw %ymm2, %ymm1, %ymm1 ## encoding: [0x62,0xf1,0x75,0x28,0xfd,0xca]
; CHECK-NEXT:    vpaddw %ymm0, %ymm1, %ymm0 ## encoding: [0x62,0xf1,0x75,0x28,0xfd,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.mask.psll.wi.256(<16 x i16> %x0, i32 3, <16 x i16> %x2, i16 %x3)
  %res1 = call <16 x i16> @llvm.x86.avx512.mask.psll.wi.256(<16 x i16> %x0, i32 3, <16 x i16> zeroinitializer, i16 %x3)
  %res2 = call <16 x i16> @llvm.x86.avx512.mask.psll.wi.256(<16 x i16> %x0, i32 3, <16 x i16> %x2, i16 -1)
  %res3 = add <16 x i16> %res, %res1
  %res4 = add <16 x i16> %res3, %res2
  ret <16 x i16> %res4
}

declare <16 x i16> @llvm.x86.avx512.mask.psllv16.hi(<16 x i16>, <16 x i16>, <16 x i16>, i16)

define <16 x i16>@test_int_x86_avx512_mask_psllv16_hi(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_psllv16_hi:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpsllvw %ymm1, %ymm0, %ymm2 {%k1} ## encoding: [0x62,0xf2,0xfd,0x29,0x12,0xd1]
; CHECK-NEXT:    vpsllvw %ymm1, %ymm0, %ymm3 {%k1} {z} ## encoding: [0x62,0xf2,0xfd,0xa9,0x12,0xd9]
; CHECK-NEXT:    vpsllvw %ymm1, %ymm0, %ymm0 ## encoding: [0x62,0xf2,0xfd,0x28,0x12,0xc1]
; CHECK-NEXT:    vpaddw %ymm3, %ymm2, %ymm1 ## encoding: [0x62,0xf1,0x6d,0x28,0xfd,0xcb]
; CHECK-NEXT:    vpaddw %ymm0, %ymm1, %ymm0 ## encoding: [0x62,0xf1,0x75,0x28,0xfd,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.mask.psllv16.hi(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %x3)
  %res1 = call <16 x i16> @llvm.x86.avx512.mask.psllv16.hi(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> zeroinitializer, i16 %x3)
  %res2 = call <16 x i16> @llvm.x86.avx512.mask.psllv16.hi(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 -1)
  %res3 = add <16 x i16> %res, %res1
  %res4 = add <16 x i16> %res3, %res2
  ret <16 x i16> %res4
}

declare <8 x i16> @llvm.x86.avx512.mask.psllv8.hi(<8 x i16>, <8 x i16>, <8 x i16>, i8)

define <8 x i16>@test_int_x86_avx512_mask_psllv8_hi(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_psllv8_hi:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpsllvw %xmm1, %xmm0, %xmm2 {%k1} ## encoding: [0x62,0xf2,0xfd,0x09,0x12,0xd1]
; CHECK-NEXT:    vpsllvw %xmm1, %xmm0, %xmm3 {%k1} {z} ## encoding: [0x62,0xf2,0xfd,0x89,0x12,0xd9]
; CHECK-NEXT:    vpsllvw %xmm1, %xmm0, %xmm0 ## encoding: [0x62,0xf2,0xfd,0x08,0x12,0xc1]
; CHECK-NEXT:    vpaddw %xmm3, %xmm2, %xmm1 ## encoding: [0x62,0xf1,0x6d,0x08,0xfd,0xcb]
; CHECK-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 ## encoding: [0x62,0xf1,0x75,0x08,0xfd,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.mask.psllv8.hi(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3)
  %res1 = call <8 x i16> @llvm.x86.avx512.mask.psllv8.hi(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> zeroinitializer, i8 %x3)
  %res2 = call <8 x i16> @llvm.x86.avx512.mask.psllv8.hi(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 -1)
  %res3 = add <8 x i16> %res, %res1
  %res4 = add <8 x i16> %res3, %res2
  ret <8 x i16> %res4
}

declare <8 x i16> @llvm.x86.avx512.mask.pmovzxb.w.128(<16 x i8>, <8 x i16>, i8)

define <8 x i16>@test_int_x86_avx512_mask_pmovzxb_w_128(<16 x i8> %x0, <8 x i16> %x1, i8 %x2) {
; CHECK-LABEL: test_int_x86_avx512_mask_pmovzxb_w_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpmovzxbw %xmm0, %xmm1 {%k1} ## encoding: [0x62,0xf2,0x7d,0x09,0x30,0xc8]
; CHECK-NEXT:    ## xmm1 {%k1} = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
; CHECK-NEXT:    vpmovzxbw %xmm0, %xmm2 {%k1} {z} ## encoding: [0x62,0xf2,0x7d,0x89,0x30,0xd0]
; CHECK-NEXT:    ## xmm2 {%k1} {z} = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
; CHECK-NEXT:    vpmovzxbw %xmm0, %xmm0 ## encoding: [0x62,0xf2,0x7d,0x08,0x30,0xc0]
; CHECK-NEXT:    ## xmm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
; CHECK-NEXT:    vpaddw %xmm2, %xmm1, %xmm1 ## encoding: [0x62,0xf1,0x75,0x08,0xfd,0xca]
; CHECK-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 ## encoding: [0x62,0xf1,0x75,0x08,0xfd,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.mask.pmovzxb.w.128(<16 x i8> %x0, <8 x i16> %x1, i8 %x2)
  %res1 = call <8 x i16> @llvm.x86.avx512.mask.pmovzxb.w.128(<16 x i8> %x0, <8 x i16> zeroinitializer, i8 %x2)
  %res2 = call <8 x i16> @llvm.x86.avx512.mask.pmovzxb.w.128(<16 x i8> %x0, <8 x i16> %x1, i8 -1)
  %res3 = add <8 x i16> %res, %res1
  %res4 = add <8 x i16> %res3, %res2
  ret <8 x i16> %res4
}

declare <16 x i16> @llvm.x86.avx512.mask.pmovzxb.w.256(<16 x i8>, <16 x i16>, i16)

define <16 x i16>@test_int_x86_avx512_mask_pmovzxb_w_256(<16 x i8> %x0, <16 x i16> %x1, i16 %x2) {
; CHECK-LABEL: test_int_x86_avx512_mask_pmovzxb_w_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpmovzxbw %xmm0, %ymm1 {%k1} ## encoding: [0x62,0xf2,0x7d,0x29,0x30,0xc8]
; CHECK-NEXT:    ## ymm1 {%k1} = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
; CHECK-NEXT:    vpmovzxbw %xmm0, %ymm2 {%k1} {z} ## encoding: [0x62,0xf2,0x7d,0xa9,0x30,0xd0]
; CHECK-NEXT:    ## ymm2 {%k1} {z} = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
; CHECK-NEXT:    vpmovzxbw %xmm0, %ymm0 ## encoding: [0x62,0xf2,0x7d,0x28,0x30,0xc0]
; CHECK-NEXT:    ## ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
; CHECK-NEXT:    vpaddw %ymm2, %ymm1, %ymm1 ## encoding: [0x62,0xf1,0x75,0x28,0xfd,0xca]
; CHECK-NEXT:    vpaddw %ymm0, %ymm1, %ymm0 ## encoding: [0x62,0xf1,0x75,0x28,0xfd,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.mask.pmovzxb.w.256(<16 x i8> %x0, <16 x i16> %x1, i16 %x2)
  %res1 = call <16 x i16> @llvm.x86.avx512.mask.pmovzxb.w.256(<16 x i8> %x0, <16 x i16> zeroinitializer, i16 %x2)
  %res2 = call <16 x i16> @llvm.x86.avx512.mask.pmovzxb.w.256(<16 x i8> %x0, <16 x i16> %x1, i16 -1)
  %res3 = add <16 x i16> %res, %res1
  %res4 = add <16 x i16> %res3, %res2
  ret <16 x i16> %res4
}


declare <8 x i16> @llvm.x86.avx512.mask.pmovsxb.w.128(<16 x i8>, <8 x i16>, i8)

define <8 x i16>@test_int_x86_avx512_mask_pmovsxb_w_128(<16 x i8> %x0, <8 x i16> %x1, i8 %x2) {
; CHECK-LABEL: test_int_x86_avx512_mask_pmovsxb_w_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpmovsxbw %xmm0, %xmm1 {%k1} ## encoding: [0x62,0xf2,0x7d,0x09,0x20,0xc8]
; CHECK-NEXT:    vpmovsxbw %xmm0, %xmm2 {%k1} {z} ## encoding: [0x62,0xf2,0x7d,0x89,0x20,0xd0]
; CHECK-NEXT:    vpmovsxbw %xmm0, %xmm0 ## encoding: [0x62,0xf2,0x7d,0x08,0x20,0xc0]
; CHECK-NEXT:    vpaddw %xmm2, %xmm1, %xmm1 ## encoding: [0x62,0xf1,0x75,0x08,0xfd,0xca]
; CHECK-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 ## encoding: [0x62,0xf1,0x75,0x08,0xfd,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.mask.pmovsxb.w.128(<16 x i8> %x0, <8 x i16> %x1, i8 %x2)
  %res1 = call <8 x i16> @llvm.x86.avx512.mask.pmovsxb.w.128(<16 x i8> %x0, <8 x i16> zeroinitializer, i8 %x2)
  %res2 = call <8 x i16> @llvm.x86.avx512.mask.pmovsxb.w.128(<16 x i8> %x0, <8 x i16> %x1, i8 -1)
  %res3 = add <8 x i16> %res, %res1
  %res4 = add <8 x i16> %res3, %res2
  ret <8 x i16> %res4
}

declare <16 x i16> @llvm.x86.avx512.mask.pmovsxb.w.256(<16 x i8>, <16 x i16>, i16)

define <16 x i16>@test_int_x86_avx512_mask_pmovsxb_w_256(<16 x i8> %x0, <16 x i16> %x1, i16 %x2) {
; CHECK-LABEL: test_int_x86_avx512_mask_pmovsxb_w_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpmovsxbw %xmm0, %ymm1 {%k1} ## encoding: [0x62,0xf2,0x7d,0x29,0x20,0xc8]
; CHECK-NEXT:    vpmovsxbw %xmm0, %ymm2 {%k1} {z} ## encoding: [0x62,0xf2,0x7d,0xa9,0x20,0xd0]
; CHECK-NEXT:    vpmovsxbw %xmm0, %ymm0 ## encoding: [0x62,0xf2,0x7d,0x28,0x20,0xc0]
; CHECK-NEXT:    vpaddw %ymm2, %ymm1, %ymm1 ## encoding: [0x62,0xf1,0x75,0x28,0xfd,0xca]
; CHECK-NEXT:    vpaddw %ymm0, %ymm1, %ymm0 ## encoding: [0x62,0xf1,0x75,0x28,0xfd,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.mask.pmovsxb.w.256(<16 x i8> %x0, <16 x i16> %x1, i16 %x2)
  %res1 = call <16 x i16> @llvm.x86.avx512.mask.pmovsxb.w.256(<16 x i8> %x0, <16 x i16> zeroinitializer, i16 %x2)
  %res2 = call <16 x i16> @llvm.x86.avx512.mask.pmovsxb.w.256(<16 x i8> %x0, <16 x i16> %x1, i16 -1)
  %res3 = add <16 x i16> %res, %res1
  %res4 = add <16 x i16> %res3, %res2
  ret <16 x i16> %res4
}

declare <2 x i64> @llvm.x86.avx512.mask.pmovsxd.q.128(<4 x i32>, <2 x i64>, i8)

define <2 x i64>@test_int_x86_avx512_mask_pmovsxd_q_128(<4 x i32> %x0, <2 x i64> %x1, i8 %x2) {
; CHECK-LABEL: test_int_x86_avx512_mask_pmovsxd_q_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpmovsxdq %xmm0, %xmm1 {%k1} ## encoding: [0x62,0xf2,0x7d,0x09,0x25,0xc8]
; CHECK-NEXT:    vpmovsxdq %xmm0, %xmm2 {%k1} {z} ## encoding: [0x62,0xf2,0x7d,0x89,0x25,0xd0]
; CHECK-NEXT:    vpmovsxdq %xmm0, %xmm0 ## encoding: [0x62,0xf2,0x7d,0x08,0x25,0xc0]
; CHECK-NEXT:    vpaddq %xmm2, %xmm1, %xmm1 ## encoding: [0x62,0xf1,0xf5,0x08,0xd4,0xca]
; CHECK-NEXT:    vpaddq %xmm0, %xmm1, %xmm0 ## encoding: [0x62,0xf1,0xf5,0x08,0xd4,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <2 x i64> @llvm.x86.avx512.mask.pmovsxd.q.128(<4 x i32> %x0, <2 x i64> %x1, i8 %x2)
  %res1 = call <2 x i64> @llvm.x86.avx512.mask.pmovsxd.q.128(<4 x i32> %x0, <2 x i64> zeroinitializer, i8 %x2)
  %res2 = call <2 x i64> @llvm.x86.avx512.mask.pmovsxd.q.128(<4 x i32> %x0, <2 x i64> %x1, i8 -1)
  %res3 = add <2 x i64> %res, %res1
  %res4 = add <2 x i64> %res3, %res2
  ret <2 x i64> %res4
}

declare <4 x i64> @llvm.x86.avx512.mask.pmovsxd.q.256(<4 x i32>, <4 x i64>, i8)

define <4 x i64>@test_int_x86_avx512_mask_pmovsxd_q_256(<4 x i32> %x0, <4 x i64> %x1, i8 %x2) {
; CHECK-LABEL: test_int_x86_avx512_mask_pmovsxd_q_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpmovsxdq %xmm0, %ymm1 {%k1} ## encoding: [0x62,0xf2,0x7d,0x29,0x25,0xc8]
; CHECK-NEXT:    vpmovsxdq %xmm0, %ymm2 {%k1} {z} ## encoding: [0x62,0xf2,0x7d,0xa9,0x25,0xd0]
; CHECK-NEXT:    vpmovsxdq %xmm0, %ymm0 ## encoding: [0x62,0xf2,0x7d,0x28,0x25,0xc0]
; CHECK-NEXT:    vpaddq %ymm2, %ymm1, %ymm1 ## encoding: [0x62,0xf1,0xf5,0x28,0xd4,0xca]
; CHECK-NEXT:    vpaddq %ymm0, %ymm1, %ymm0 ## encoding: [0x62,0xf1,0xf5,0x28,0xd4,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <4 x i64> @llvm.x86.avx512.mask.pmovsxd.q.256(<4 x i32> %x0, <4 x i64> %x1, i8 %x2)
  %res1 = call <4 x i64> @llvm.x86.avx512.mask.pmovsxd.q.256(<4 x i32> %x0, <4 x i64> zeroinitializer, i8 %x2)
  %res2 = call <4 x i64> @llvm.x86.avx512.mask.pmovsxd.q.256(<4 x i32> %x0, <4 x i64> %x1, i8 -1)
  %res3 = add <4 x i64> %res, %res1
  %res4 = add <4 x i64> %res3, %res2
  ret <4 x i64> %res4
}

declare <8 x i16> @llvm.x86.avx512.mask.permvar.hi.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)

define <8 x i16>@test_int_x86_avx512_mask_permvar_hi_128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_permvar_hi_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpermw %xmm0, %xmm1, %xmm2 {%k1} ## encoding: [0x62,0xf2,0xf5,0x09,0x8d,0xd0]
; CHECK-NEXT:    vpermw %xmm0, %xmm1, %xmm3 {%k1} {z} ## encoding: [0x62,0xf2,0xf5,0x89,0x8d,0xd8]
; CHECK-NEXT:    vpermw %xmm0, %xmm1, %xmm0 ## encoding: [0x62,0xf2,0xf5,0x08,0x8d,0xc0]
; CHECK-NEXT:    vpaddw %xmm3, %xmm2, %xmm1 ## encoding: [0x62,0xf1,0x6d,0x08,0xfd,0xcb]
; CHECK-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 ## encoding: [0x62,0xf1,0x75,0x08,0xfd,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.mask.permvar.hi.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3)
  %res1 = call <8 x i16> @llvm.x86.avx512.mask.permvar.hi.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> zeroinitializer, i8 %x3)
  %res2 = call <8 x i16> @llvm.x86.avx512.mask.permvar.hi.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 -1)
  %res3 = add <8 x i16> %res, %res1
  %res4 = add <8 x i16> %res3, %res2
  ret <8 x i16> %res4
}

declare <16 x i16> @llvm.x86.avx512.mask.permvar.hi.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)

define <16 x i16>@test_int_x86_avx512_mask_permvar_hi_256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_permvar_hi_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vpermw %ymm0, %ymm1, %ymm2 {%k1} ## encoding: [0x62,0xf2,0xf5,0x29,0x8d,0xd0]
; CHECK-NEXT:    vpermw %ymm0, %ymm1, %ymm3 {%k1} {z} ## encoding: [0x62,0xf2,0xf5,0xa9,0x8d,0xd8]
; CHECK-NEXT:    vpermw %ymm0, %ymm1, %ymm0 ## encoding: [0x62,0xf2,0xf5,0x28,0x8d,0xc0]
; CHECK-NEXT:    vpaddw %ymm3, %ymm2, %ymm1 ## encoding: [0x62,0xf1,0x6d,0x28,0xfd,0xcb]
; CHECK-NEXT:    vpaddw %ymm0, %ymm1, %ymm0 ## encoding: [0x62,0xf1,0x75,0x28,0xfd,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.mask.permvar.hi.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %x3)
  %res1 = call <16 x i16> @llvm.x86.avx512.mask.permvar.hi.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> zeroinitializer, i16 %x3)
  %res2 = call <16 x i16> @llvm.x86.avx512.mask.permvar.hi.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 -1)
  %res3 = add <16 x i16> %res, %res1
  %res4 = add <16 x i16> %res3, %res2
  ret <16 x i16> %res4
}

declare i16 @llvm.x86.avx512.ptestm.b.128(<16 x i8>, <16 x i8>, i16)

define i16@test_int_x86_avx512_ptestm_b_128(<16 x i8> %x0, <16 x i8> %x1, i16 %x2) {
; CHECK-LABEL: test_int_x86_avx512_ptestm_b_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vptestmb %xmm1, %xmm0, %k0 {%k1} ## encoding: [0x62,0xf2,0x7d,0x09,0x26,0xc1]
; CHECK-NEXT:    kmovw %k0, %ecx ## encoding: [0xc5,0xf8,0x93,0xc8]
; CHECK-NEXT:    vptestmb %xmm1, %xmm0, %k0 ## encoding: [0x62,0xf2,0x7d,0x08,0x26,0xc1]
; CHECK-NEXT:    kmovw %k0, %eax ## encoding: [0xc5,0xf8,0x93,0xc0]
; CHECK-NEXT:    addl %ecx, %eax ## encoding: [0x01,0xc8]
; CHECK-NEXT:    ## kill: %AX<def> %AX<kill> %EAX<kill>
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call i16 @llvm.x86.avx512.ptestm.b.128(<16 x i8> %x0, <16 x i8> %x1, i16 %x2)
  %res1 = call i16 @llvm.x86.avx512.ptestm.b.128(<16 x i8> %x0, <16 x i8> %x1, i16-1)
  %res2 = add i16 %res, %res1
  ret i16 %res2
}

declare i32 @llvm.x86.avx512.ptestm.b.256(<32 x i8>, <32 x i8>, i32)

define i32@test_int_x86_avx512_ptestm_b_256(<32 x i8> %x0, <32 x i8> %x1, i32 %x2) {
; CHECK-LABEL: test_int_x86_avx512_ptestm_b_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
; CHECK-NEXT:    vptestmb %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf2,0x7d,0x29,0x26,0xc1]
; CHECK-NEXT:    kmovd %k0, %ecx ## encoding: [0xc5,0xfb,0x93,0xc8]
; CHECK-NEXT:    vptestmb %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf2,0x7d,0x28,0x26,0xc1]
; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
; CHECK-NEXT:    addl %ecx, %eax ## encoding: [0x01,0xc8]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call i32 @llvm.x86.avx512.ptestm.b.256(<32 x i8> %x0, <32 x i8> %x1, i32 %x2)
  %res1 = call i32 @llvm.x86.avx512.ptestm.b.256(<32 x i8> %x0, <32 x i8> %x1, i32-1)
  %res2 = add i32 %res, %res1
  ret i32 %res2
}

declare i8 @llvm.x86.avx512.ptestm.w.128(<8 x i16>, <8 x i16>, i8)

define i8@test_int_x86_avx512_ptestm_w_128(<8 x i16> %x0, <8 x i16> %x1, i8 %x2) {
; CHECK-LABEL: test_int_x86_avx512_ptestm_w_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vptestmw %xmm1, %xmm0, %k0 {%k1} ## encoding: [0x62,0xf2,0xfd,0x09,0x26,0xc1]
; CHECK-NEXT:    kmovw %k0, %ecx ## encoding: [0xc5,0xf8,0x93,0xc8]
; CHECK-NEXT:    vptestmw %xmm1, %xmm0, %k0 ## encoding: [0x62,0xf2,0xfd,0x08,0x26,0xc1]
; CHECK-NEXT:    kmovw %k0, %eax ## encoding: [0xc5,0xf8,0x93,0xc0]
; CHECK-NEXT:    addb %cl, %al ## encoding: [0x00,0xc8]
; CHECK-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call i8 @llvm.x86.avx512.ptestm.w.128(<8 x i16> %x0, <8 x i16> %x1, i8 %x2)
  %res1 = call i8 @llvm.x86.avx512.ptestm.w.128(<8 x i16> %x0, <8 x i16> %x1, i8-1)
  %res2 = add i8 %res, %res1
  ret i8 %res2
}

declare i16 @llvm.x86.avx512.ptestm.w.256(<16 x i16>, <16 x i16>, i16)

define i16@test_int_x86_avx512_ptestm_w_256(<16 x i16> %x0, <16 x i16> %x1, i16 %x2) {
; CHECK-LABEL: test_int_x86_avx512_ptestm_w_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vptestmw %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf2,0xfd,0x29,0x26,0xc1]
; CHECK-NEXT:    kmovw %k0, %ecx ## encoding: [0xc5,0xf8,0x93,0xc8]
; CHECK-NEXT:    vptestmw %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf2,0xfd,0x28,0x26,0xc1]
; CHECK-NEXT:    kmovw %k0, %eax ## encoding: [0xc5,0xf8,0x93,0xc0]
; CHECK-NEXT:    addl %ecx, %eax ## encoding: [0x01,0xc8]
; CHECK-NEXT:    ## kill: %AX<def> %AX<kill> %EAX<kill>
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call i16 @llvm.x86.avx512.ptestm.w.256(<16 x i16> %x0, <16 x i16> %x1, i16 %x2)
  %res1 = call i16 @llvm.x86.avx512.ptestm.w.256(<16 x i16> %x0, <16 x i16> %x1, i16-1)
  %res2 = add i16 %res, %res1
  ret i16 %res2
}

declare i16 @llvm.x86.avx512.ptestnm.b.128(<16 x i8>, <16 x i8>, i16)

define i16@test_int_x86_avx512_ptestnm_b_128(<16 x i8> %x0, <16 x i8> %x1, i16 %x2) {
; CHECK-LABEL: test_int_x86_avx512_ptestnm_b_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vptestnmb %xmm1, %xmm0, %k0 {%k1} ## encoding: [0x62,0xf2,0x7e,0x09,0x26,0xc1]
; CHECK-NEXT:    kmovw %k0, %ecx ## encoding: [0xc5,0xf8,0x93,0xc8]
; CHECK-NEXT:    vptestnmb %xmm1, %xmm0, %k0 ## encoding: [0x62,0xf2,0x7e,0x08,0x26,0xc1]
; CHECK-NEXT:    kmovw %k0, %eax ## encoding: [0xc5,0xf8,0x93,0xc0]
; CHECK-NEXT:    addl %ecx, %eax ## encoding: [0x01,0xc8]
; CHECK-NEXT:    ## kill: %AX<def> %AX<kill> %EAX<kill>
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call i16 @llvm.x86.avx512.ptestnm.b.128(<16 x i8> %x0, <16 x i8> %x1, i16 %x2)
  %res1 = call i16 @llvm.x86.avx512.ptestnm.b.128(<16 x i8> %x0, <16 x i8> %x1, i16-1)
  %res2 = add i16 %res, %res1
  ret i16 %res2
}

declare i32 @llvm.x86.avx512.ptestnm.b.256(<32 x i8>, <32 x i8>, i32)

define i32@test_int_x86_avx512_ptestnm_b_256(<32 x i8> %x0, <32 x i8> %x1, i32 %x2) {
; CHECK-LABEL: test_int_x86_avx512_ptestnm_b_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
; CHECK-NEXT:    vptestnmb %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf2,0x7e,0x29,0x26,0xc1]
; CHECK-NEXT:    kmovd %k0, %ecx ## encoding: [0xc5,0xfb,0x93,0xc8]
; CHECK-NEXT:    vptestnmb %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf2,0x7e,0x28,0x26,0xc1]
; CHECK-NEXT:    kmovd %k0, %eax ## encoding: [0xc5,0xfb,0x93,0xc0]
; CHECK-NEXT:    addl %ecx, %eax ## encoding: [0x01,0xc8]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call i32 @llvm.x86.avx512.ptestnm.b.256(<32 x i8> %x0, <32 x i8> %x1, i32 %x2)
  %res1 = call i32 @llvm.x86.avx512.ptestnm.b.256(<32 x i8> %x0, <32 x i8> %x1, i32-1)
  %res2 = add i32 %res, %res1
  ret i32 %res2
}

declare i8 @llvm.x86.avx512.ptestnm.w.128(<8 x i16>, <8 x i16>, i8 %x2)

define i8@test_int_x86_avx512_ptestnm_w_128(<8 x i16> %x0, <8 x i16> %x1, i8 %x2) {
; CHECK-LABEL: test_int_x86_avx512_ptestnm_w_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vptestnmw %xmm1, %xmm0, %k0 {%k1} ## encoding: [0x62,0xf2,0xfe,0x09,0x26,0xc1]
; CHECK-NEXT:    kmovw %k0, %ecx ## encoding: [0xc5,0xf8,0x93,0xc8]
; CHECK-NEXT:    vptestnmw %xmm1, %xmm0, %k0 ## encoding: [0x62,0xf2,0xfe,0x08,0x26,0xc1]
; CHECK-NEXT:    kmovw %k0, %eax ## encoding: [0xc5,0xf8,0x93,0xc0]
; CHECK-NEXT:    addb %cl, %al ## encoding: [0x00,0xc8]
; CHECK-NEXT:    ## kill: %AL<def> %AL<kill> %EAX<kill>
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call i8 @llvm.x86.avx512.ptestnm.w.128(<8 x i16> %x0, <8 x i16> %x1, i8 %x2)
  %res1 = call i8 @llvm.x86.avx512.ptestnm.w.128(<8 x i16> %x0, <8 x i16> %x1, i8-1)
  %res2 = add i8 %res, %res1
  ret i8 %res2
}

declare i16 @llvm.x86.avx512.ptestnm.w.256(<16 x i16>, <16 x i16>, i16 %x2)

define i16@test_int_x86_avx512_ptestnm_w_256(<16 x i16> %x0, <16 x i16> %x1, i16 %x2) {
; CHECK-LABEL: test_int_x86_avx512_ptestnm_w_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT:    vptestnmw %ymm1, %ymm0, %k0 {%k1} ## encoding: [0x62,0xf2,0xfe,0x29,0x26,0xc1]
; CHECK-NEXT:    kmovw %k0, %ecx ## encoding: [0xc5,0xf8,0x93,0xc8]
; CHECK-NEXT:    vptestnmw %ymm1, %ymm0, %k0 ## encoding: [0x62,0xf2,0xfe,0x28,0x26,0xc1]
; CHECK-NEXT:    kmovw %k0, %eax ## encoding: [0xc5,0xf8,0x93,0xc0]
; CHECK-NEXT:    addl %ecx, %eax ## encoding: [0x01,0xc8]
; CHECK-NEXT:    ## kill: %AX<def> %AX<kill> %EAX<kill>
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call i16 @llvm.x86.avx512.ptestnm.w.256(<16 x i16> %x0, <16 x i16> %x1, i16 %x2)
  %res1 = call i16 @llvm.x86.avx512.ptestnm.w.256(<16 x i16> %x0, <16 x i16> %x1, i16-1)
  %res2 = add i16 %res, %res1
  ret i16 %res2
}

declare <32 x i8> @llvm.x86.avx512.mask.pbroadcast.b.gpr.256(i8, <32 x i8>, i32)

define <32 x i8>@test_int_x86_avx512_mask_pbroadcast_b_gpr_256(i8 %x0, <32 x i8> %x1, i32 %mask) {
; CHECK-LABEL: test_int_x86_avx512_mask_pbroadcast_b_gpr_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovd %esi, %k1 ## encoding: [0xc5,0xfb,0x92,0xce]
; CHECK-NEXT:    vpbroadcastb %dil, %ymm0 {%k1} ## encoding: [0x62,0xf2,0x7d,0x29,0x7a,0xc7]
; CHECK-NEXT:    vpbroadcastb %dil, %ymm1 {%k1} {z} ## encoding: [0x62,0xf2,0x7d,0xa9,0x7a,0xcf]
; CHECK-NEXT:    vpbroadcastb %dil, %ymm2 ## encoding: [0x62,0xf2,0x7d,0x28,0x7a,0xd7]
; CHECK-NEXT:    vpaddb %ymm0, %ymm2, %ymm0 ## encoding: [0x62,0xf1,0x6d,0x28,0xfc,0xc0]
; CHECK-NEXT:    vpaddb %ymm0, %ymm1, %ymm0 ## encoding: [0x62,0xf1,0x75,0x28,0xfc,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <32 x i8> @llvm.x86.avx512.mask.pbroadcast.b.gpr.256(i8 %x0, <32 x i8> %x1, i32 -1)
  %res1 = call <32 x i8> @llvm.x86.avx512.mask.pbroadcast.b.gpr.256(i8 %x0, <32 x i8> %x1, i32 %mask)
  %res2 = call <32 x i8> @llvm.x86.avx512.mask.pbroadcast.b.gpr.256(i8 %x0, <32 x i8> zeroinitializer, i32 %mask)
  %res3 = add <32 x i8> %res, %res1
  %res4 = add <32 x i8> %res2, %res3
  ret <32 x i8> %res4
}

declare <16 x i8> @llvm.x86.avx512.mask.pbroadcast.b.gpr.128(i8, <16 x i8>, i16)

define <16 x i8>@test_int_x86_avx512_mask_pbroadcast_b_gpr_128(i8 %x0, <16 x i8> %x1, i16 %mask) {
; CHECK-LABEL: test_int_x86_avx512_mask_pbroadcast_b_gpr_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpbroadcastb %dil, %xmm0 {%k1} ## encoding: [0x62,0xf2,0x7d,0x09,0x7a,0xc7]
; CHECK-NEXT:    vpbroadcastb %dil, %xmm1 {%k1} {z} ## encoding: [0x62,0xf2,0x7d,0x89,0x7a,0xcf]
; CHECK-NEXT:    vpbroadcastb %dil, %xmm2 ## encoding: [0x62,0xf2,0x7d,0x08,0x7a,0xd7]
; CHECK-NEXT:    vpaddb %xmm0, %xmm2, %xmm0 ## encoding: [0x62,0xf1,0x6d,0x08,0xfc,0xc0]
; CHECK-NEXT:    vpaddb %xmm0, %xmm1, %xmm0 ## encoding: [0x62,0xf1,0x75,0x08,0xfc,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i8> @llvm.x86.avx512.mask.pbroadcast.b.gpr.128(i8 %x0, <16 x i8> %x1, i16 -1)
  %res1 = call <16 x i8> @llvm.x86.avx512.mask.pbroadcast.b.gpr.128(i8 %x0, <16 x i8> %x1, i16 %mask)
  %res2 = call <16 x i8> @llvm.x86.avx512.mask.pbroadcast.b.gpr.128(i8 %x0, <16 x i8> zeroinitializer, i16 %mask)
  %res3 = add <16 x i8> %res, %res1
  %res4 = add <16 x i8> %res2, %res3
  ret <16 x i8> %res4
}

declare <16 x i16> @llvm.x86.avx512.mask.pbroadcast.w.gpr.256(i16, <16 x i16>, i16)

define <16 x i16>@test_int_x86_avx512_mask_pbroadcast_w_gpr_256(i16 %x0, <16 x i16> %x1, i16 %mask) {
; CHECK-LABEL: test_int_x86_avx512_mask_pbroadcast_w_gpr_256:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpbroadcastw %di, %ymm0 {%k1} ## encoding: [0x62,0xf2,0x7d,0x29,0x7b,0xc7]
; CHECK-NEXT:    vpbroadcastw %di, %ymm1 {%k1} {z} ## encoding: [0x62,0xf2,0x7d,0xa9,0x7b,0xcf]
; CHECK-NEXT:    vpbroadcastw %di, %ymm2 ## encoding: [0x62,0xf2,0x7d,0x28,0x7b,0xd7]
; CHECK-NEXT:    vpaddw %ymm0, %ymm2, %ymm0 ## encoding: [0x62,0xf1,0x6d,0x28,0xfd,0xc0]
; CHECK-NEXT:    vpaddw %ymm0, %ymm1, %ymm0 ## encoding: [0x62,0xf1,0x75,0x28,0xfd,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.mask.pbroadcast.w.gpr.256(i16 %x0, <16 x i16> %x1, i16 -1)
  %res1 = call <16 x i16> @llvm.x86.avx512.mask.pbroadcast.w.gpr.256(i16 %x0, <16 x i16> %x1, i16 %mask)
  %res2 = call <16 x i16> @llvm.x86.avx512.mask.pbroadcast.w.gpr.256(i16 %x0, <16 x i16> zeroinitializer, i16 %mask)
  %res3 = add <16 x i16> %res, %res1
  %res4 = add <16 x i16> %res2, %res3
  ret <16 x i16> %res4
}

declare <8 x i16> @llvm.x86.avx512.mask.pbroadcast.w.gpr.128(i16, <8 x i16>, i8)

define <8 x i16>@test_int_x86_avx512_mask_pbroadcast_w_gpr_128(i16 %x0, <8 x i16> %x1, i8 %mask) {
; CHECK-LABEL: test_int_x86_avx512_mask_pbroadcast_w_gpr_128:
; CHECK:       ## BB#0:
; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT:    vpbroadcastw %di, %xmm0 {%k1} ## encoding: [0x62,0xf2,0x7d,0x09,0x7b,0xc7]
; CHECK-NEXT:    vpbroadcastw %di, %xmm1 {%k1} {z} ## encoding: [0x62,0xf2,0x7d,0x89,0x7b,0xcf]
; CHECK-NEXT:    vpbroadcastw %di, %xmm2 ## encoding: [0x62,0xf2,0x7d,0x08,0x7b,0xd7]
; CHECK-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 ## encoding: [0x62,0xf1,0x6d,0x08,0xfd,0xc0]
; CHECK-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 ## encoding: [0x62,0xf1,0x75,0x08,0xfd,0xc0]
; CHECK-NEXT:    retq ## encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.mask.pbroadcast.w.gpr.128(i16 %x0, <8 x i16> %x1, i8 -1)
  %res1 = call <8 x i16> @llvm.x86.avx512.mask.pbroadcast.w.gpr.128(i16 %x0, <8 x i16> %x1, i8 %mask)
  %res2 = call <8 x i16> @llvm.x86.avx512.mask.pbroadcast.w.gpr.128(i16 %x0, <8 x i16> zeroinitializer, i8 %mask)
  %res3 = add <8 x i16> %res, %res1
  %res4 = add <8 x i16> %res2, %res3
  ret <8 x i16> %res4
}