; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+avx512bw,+avx512vl --show-mc-encoding | FileCheck %s --check-prefixes=CHECK,X86
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512bw,+avx512vl --show-mc-encoding | FileCheck %s --check-prefixes=CHECK,X64

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) {
; X86-LABEL: test_int_x86_avx512_mask_pbroadcast_b_gpr_128:
; X86:       # %bb.0:
; X86-NEXT:    movb {{[0-9]+}}(%esp), %al # encoding: [0x8a,0x44,0x24,0x04]
; X86-NEXT:    vpbroadcastb %eax, %xmm1 # encoding: [0x62,0xf2,0x7d,0x08,0x7a,0xc8]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpbroadcastb %eax, %xmm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x7a,0xc0]
; X86-NEXT:    vpbroadcastb %eax, %xmm2 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x7a,0xd0]
; X86-NEXT:    vpaddb %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfc,0xc2]
; X86-NEXT:    vpaddb %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfc,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pbroadcast_b_gpr_128:
; X64:       # %bb.0:
; X64-NEXT:    vpbroadcastb %edi, %xmm1 # encoding: [0x62,0xf2,0x7d,0x08,0x7a,0xcf]
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpbroadcastb %edi, %xmm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x7a,0xc7]
; X64-NEXT:    vpbroadcastb %edi, %xmm2 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x7a,0xd7]
; X64-NEXT:    vpaddb %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfc,0xc2]
; X64-NEXT:    vpaddb %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfc,0xc0]
; X64-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 <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) {
; X86-LABEL: test_int_x86_avx512_mask_pbroadcast_w_gpr_128:
; X86:       # %bb.0:
; X86-NEXT:    movzwl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb7,0x44,0x24,0x04]
; X86-NEXT:    vpbroadcastw %eax, %xmm1 # encoding: [0x62,0xf2,0x7d,0x08,0x7b,0xc8]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
; X86-NEXT:    kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
; X86-NEXT:    vpbroadcastw %eax, %xmm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x7b,0xc0]
; X86-NEXT:    vpbroadcastw %eax, %xmm2 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x7b,0xd0]
; X86-NEXT:    vpaddw %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc2]
; X86-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pbroadcast_w_gpr_128:
; X64:       # %bb.0:
; X64-NEXT:    vpbroadcastw %edi, %xmm1 # encoding: [0x62,0xf2,0x7d,0x08,0x7b,0xcf]
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpbroadcastw %edi, %xmm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x7b,0xc7]
; X64-NEXT:    vpbroadcastw %edi, %xmm2 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x7b,0xd7]
; X64-NEXT:    vpaddw %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc2]
; X64-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
; X64-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
}


 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) {
; X86-LABEL: test_int_x86_avx512_mask_pbroadcast_b_gpr_256:
; X86:       # %bb.0:
; X86-NEXT:    movb {{[0-9]+}}(%esp), %al # encoding: [0x8a,0x44,0x24,0x04]
; X86-NEXT:    vpbroadcastb %eax, %ymm1 # encoding: [0x62,0xf2,0x7d,0x28,0x7a,0xc8]
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpbroadcastb %eax, %ymm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x7a,0xc0]
; X86-NEXT:    vpbroadcastb %eax, %ymm2 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x7a,0xd0]
; X86-NEXT:    vpaddb %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfc,0xc2]
; X86-NEXT:    vpaddb %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfc,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pbroadcast_b_gpr_256:
; X64:       # %bb.0:
; X64-NEXT:    vpbroadcastb %edi, %ymm1 # encoding: [0x62,0xf2,0x7d,0x28,0x7a,0xcf]
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpbroadcastb %edi, %ymm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x7a,0xc7]
; X64-NEXT:    vpbroadcastb %edi, %ymm2 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x7a,0xd7]
; X64-NEXT:    vpaddb %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfc,0xc2]
; X64-NEXT:    vpaddb %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfc,0xc0]
; X64-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 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) {
; X86-LABEL: test_int_x86_avx512_mask_pbroadcast_w_gpr_256:
; X86:       # %bb.0:
; X86-NEXT:    movzwl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb7,0x44,0x24,0x04]
; X86-NEXT:    vpbroadcastw %eax, %ymm1 # encoding: [0x62,0xf2,0x7d,0x28,0x7b,0xc8]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpbroadcastw %eax, %ymm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x7b,0xc0]
; X86-NEXT:    vpbroadcastw %eax, %ymm2 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x7b,0xd0]
; X86-NEXT:    vpaddw %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc2]
; X86-NEXT:    vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pbroadcast_w_gpr_256:
; X64:       # %bb.0:
; X64-NEXT:    vpbroadcastw %edi, %ymm1 # encoding: [0x62,0xf2,0x7d,0x28,0x7b,0xcf]
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpbroadcastw %edi, %ymm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x7b,0xc7]
; X64-NEXT:    vpbroadcastw %edi, %ymm2 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x7b,0xd7]
; X64-NEXT:    vpaddw %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc2]
; X64-NEXT:    vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
; X64-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 <32 x i8> @llvm.x86.avx512.pbroadcastb.256(<16 x i8>, <32 x i8>, i32)

define <32 x i8>@test_int_x86_avx512_pbroadcastb_256(<16 x i8> %x0, <32 x i8> %x1, i32 %mask) {
; X86-LABEL: test_int_x86_avx512_pbroadcastb_256:
; X86:       # %bb.0:
; X86-NEXT:    vpbroadcastb %xmm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x78,0xd0]
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpbroadcastb %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x78,0xc8]
; X86-NEXT:    vpbroadcastb %xmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x78,0xc0]
; X86-NEXT:    vpaddb %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfc,0xc0]
; X86-NEXT:    vpaddb %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_pbroadcastb_256:
; X64:       # %bb.0:
; X64-NEXT:    vpbroadcastb %xmm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x78,0xd0]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpbroadcastb %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x78,0xc8]
; X64-NEXT:    vpbroadcastb %xmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x78,0xc0]
; X64-NEXT:    vpaddb %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfc,0xc0]
; X64-NEXT:    vpaddb %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <32 x i8> @llvm.x86.avx512.pbroadcastb.256(<16 x i8> %x0, <32 x i8> %x1, i32 -1)
  %res1 = call <32 x i8> @llvm.x86.avx512.pbroadcastb.256(<16 x i8> %x0, <32 x i8> %x1, i32 %mask)
  %res2 = call <32 x i8> @llvm.x86.avx512.pbroadcastb.256(<16 x 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.pbroadcastb.128(<16 x i8>, <16 x i8>, i16)

define <16 x i8>@test_int_x86_avx512_pbroadcastb_128(<16 x i8> %x0, <16 x i8> %x1, i16 %mask) {
; X86-LABEL: test_int_x86_avx512_pbroadcastb_128:
; X86:       # %bb.0:
; X86-NEXT:    vpbroadcastb %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x78,0xd0]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpbroadcastb %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x78,0xc8]
; X86-NEXT:    vpbroadcastb %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x78,0xc0]
; X86-NEXT:    vpaddb %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfc,0xc0]
; X86-NEXT:    vpaddb %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_pbroadcastb_128:
; X64:       # %bb.0:
; X64-NEXT:    vpbroadcastb %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x78,0xd0]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpbroadcastb %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x78,0xc8]
; X64-NEXT:    vpbroadcastb %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x78,0xc0]
; X64-NEXT:    vpaddb %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfc,0xc0]
; X64-NEXT:    vpaddb %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <16 x i8> @llvm.x86.avx512.pbroadcastb.128(<16 x i8> %x0, <16 x i8> %x1, i16 -1)
  %res1 = call <16 x i8> @llvm.x86.avx512.pbroadcastb.128(<16 x i8> %x0, <16 x i8> %x1, i16 %mask)
  %res2 = call <16 x i8> @llvm.x86.avx512.pbroadcastb.128(<16 x 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.pbroadcastw.256(<8 x i16>, <16 x i16>, i16)

define <16 x i16>@test_int_x86_avx512_pbroadcastw_256(<8 x i16> %x0, <16 x i16> %x1, i16 %mask) {
; X86-LABEL: test_int_x86_avx512_pbroadcastw_256:
; X86:       # %bb.0:
; X86-NEXT:    vpbroadcastw %xmm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x79,0xd0]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpbroadcastw %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x79,0xc8]
; X86-NEXT:    vpbroadcastw %xmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x79,0xc0]
; X86-NEXT:    vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
; X86-NEXT:    vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_pbroadcastw_256:
; X64:       # %bb.0:
; X64-NEXT:    vpbroadcastw %xmm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x79,0xd0]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpbroadcastw %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x79,0xc8]
; X64-NEXT:    vpbroadcastw %xmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x79,0xc0]
; X64-NEXT:    vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
; X64-NEXT:    vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.pbroadcastw.256(<8 x i16> %x0, <16 x i16> %x1, i16 -1)
  %res1 = call <16 x i16> @llvm.x86.avx512.pbroadcastw.256(<8 x i16> %x0, <16 x i16> %x1, i16 %mask)
  %res2 = call <16 x i16> @llvm.x86.avx512.pbroadcastw.256(<8 x 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.pbroadcastw.128(<8 x i16>, <8 x i16>, i8)

define <8 x i16>@test_int_x86_avx512_pbroadcastw_128(<8 x i16> %x0, <8 x i16> %x1, i8 %mask) {
; X86-LABEL: test_int_x86_avx512_pbroadcastw_128:
; X86:       # %bb.0:
; X86-NEXT:    vpbroadcastw %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x79,0xd0]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpbroadcastw %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x79,0xc8]
; X86-NEXT:    vpbroadcastw %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x79,0xc0]
; X86-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
; X86-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_pbroadcastw_128:
; X64:       # %bb.0:
; X64-NEXT:    vpbroadcastw %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x79,0xd0]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpbroadcastw %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x79,0xc8]
; X64-NEXT:    vpbroadcastw %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x79,0xc0]
; X64-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
; X64-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.pbroadcastw.128(<8 x i16> %x0, <8 x i16> %x1, i8 -1)
  %res1 = call <8 x i16> @llvm.x86.avx512.pbroadcastw.128(<8 x i16> %x0, <8 x i16> %x1, i8 %mask)
  %res2 = call <8 x i16> @llvm.x86.avx512.pbroadcastw.128(<8 x 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
}

declare <64 x i8> @llvm.x86.avx512.pbroadcastb.512(<16 x i8>, <64 x i8>, i64)

define <64 x i8>@test_int_x86_avx512_pbroadcastb_512(<16 x i8> %x0, <64 x i8> %x1, i64 %mask) {
; X86-LABEL: test_int_x86_avx512_pbroadcastb_512:
; X86:       # %bb.0:
; X86-NEXT:    vpbroadcastb %xmm0, %zmm2 # encoding: [0x62,0xf2,0x7d,0x48,0x78,0xd0]
; X86-NEXT:    kmovq {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpbroadcastb %xmm0, %zmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x49,0x78,0xc8]
; X86-NEXT:    vpbroadcastb %xmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xc9,0x78,0xc0]
; X86-NEXT:    vpaddb %zmm0, %zmm1, %zmm0 # encoding: [0x62,0xf1,0x75,0x48,0xfc,0xc0]
; X86-NEXT:    vpaddb %zmm0, %zmm2, %zmm0 # encoding: [0x62,0xf1,0x6d,0x48,0xfc,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_pbroadcastb_512:
; X64:       # %bb.0:
; X64-NEXT:    vpbroadcastb %xmm0, %zmm2 # encoding: [0x62,0xf2,0x7d,0x48,0x78,0xd0]
; X64-NEXT:    kmovq %rdi, %k1 # encoding: [0xc4,0xe1,0xfb,0x92,0xcf]
; X64-NEXT:    vpbroadcastb %xmm0, %zmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x49,0x78,0xc8]
; X64-NEXT:    vpbroadcastb %xmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xc9,0x78,0xc0]
; X64-NEXT:    vpaddb %zmm0, %zmm1, %zmm0 # encoding: [0x62,0xf1,0x75,0x48,0xfc,0xc0]
; X64-NEXT:    vpaddb %zmm0, %zmm2, %zmm0 # encoding: [0x62,0xf1,0x6d,0x48,0xfc,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <64 x i8> @llvm.x86.avx512.pbroadcastb.512(<16 x i8> %x0, <64 x i8> %x1, i64 -1)
  %res1 = call <64 x i8> @llvm.x86.avx512.pbroadcastb.512(<16 x i8> %x0, <64 x i8> %x1, i64 %mask)
  %res2 = call <64 x i8> @llvm.x86.avx512.pbroadcastb.512(<16 x i8> %x0, <64 x i8> zeroinitializer, i64 %mask)
  %res3 = add <64 x i8> %res, %res1
  %res4 = add <64 x i8> %res2, %res3
  ret <64 x i8> %res4
}

declare <32 x i16> @llvm.x86.avx512.pbroadcastw.512(<8 x i16>, <32 x i16>, i32)

define <32 x i16>@test_int_x86_avx512_pbroadcastw_512(<8 x i16> %x0, <32 x i16> %x1, i32 %mask) {
; X86-LABEL: test_int_x86_avx512_pbroadcastw_512:
; X86:       # %bb.0:
; X86-NEXT:    vpbroadcastw %xmm0, %zmm2 # encoding: [0x62,0xf2,0x7d,0x48,0x79,0xd0]
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpbroadcastw %xmm0, %zmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x49,0x79,0xc8]
; X86-NEXT:    vpbroadcastw %xmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xc9,0x79,0xc0]
; X86-NEXT:    vpaddw %zmm0, %zmm1, %zmm0 # encoding: [0x62,0xf1,0x75,0x48,0xfd,0xc0]
; X86-NEXT:    vpaddw %zmm0, %zmm2, %zmm0 # encoding: [0x62,0xf1,0x6d,0x48,0xfd,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_pbroadcastw_512:
; X64:       # %bb.0:
; X64-NEXT:    vpbroadcastw %xmm0, %zmm2 # encoding: [0x62,0xf2,0x7d,0x48,0x79,0xd0]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpbroadcastw %xmm0, %zmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x49,0x79,0xc8]
; X64-NEXT:    vpbroadcastw %xmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xc9,0x79,0xc0]
; X64-NEXT:    vpaddw %zmm0, %zmm1, %zmm0 # encoding: [0x62,0xf1,0x75,0x48,0xfd,0xc0]
; X64-NEXT:    vpaddw %zmm0, %zmm2, %zmm0 # encoding: [0x62,0xf1,0x6d,0x48,0xfd,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <32 x i16> @llvm.x86.avx512.pbroadcastw.512(<8 x i16> %x0, <32 x i16> %x1, i32 -1)
  %res1 = call <32 x i16> @llvm.x86.avx512.pbroadcastw.512(<8 x i16> %x0, <32 x i16> %x1, i32 %mask)
  %res2 = call <32 x i16> @llvm.x86.avx512.pbroadcastw.512(<8 x i16> %x0, <32 x i16> zeroinitializer, i32 %mask)
  %res3 = add <32 x i16> %res, %res1
  %res4 = add <32 x i16> %res2, %res3
  ret <32 x i16> %res4
}

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

define void@test_int_x86_avx512_mask_storeu_b_128(i8* %ptr1, i8* %ptr2, <16 x i8> %x1, i16 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_storeu_b_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x08]
; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx # encoding: [0x8b,0x4c,0x24,0x04]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x0c]
; X86-NEXT:    vmovdqu8 %xmm0, (%ecx) {%k1} # encoding: [0x62,0xf1,0x7f,0x09,0x7f,0x01]
; X86-NEXT:    vmovdqu %xmm0, (%eax) # EVEX TO VEX Compression encoding: [0xc5,0xfa,0x7f,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_storeu_b_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edx, %k1 # encoding: [0xc5,0xfb,0x92,0xca]
; X64-NEXT:    vmovdqu8 %xmm0, (%rdi) {%k1} # encoding: [0x62,0xf1,0x7f,0x09,0x7f,0x07]
; X64-NEXT:    vmovdqu %xmm0, (%rsi) # EVEX TO VEX Compression encoding: [0xc5,0xfa,0x7f,0x06]
; X64-NEXT:    retq # encoding: [0xc3]
  call void @llvm.x86.avx512.mask.storeu.b.128(i8* %ptr1, <16 x i8> %x1, i16 %x2)
  call void @llvm.x86.avx512.mask.storeu.b.128(i8* %ptr2, <16 x i8> %x1, i16 -1)
  ret void
}

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

define void@test_int_x86_avx512_mask_storeu_b_256(i8* %ptr1, i8* %ptr2, <32 x i8> %x1, i32 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_storeu_b_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x08]
; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx # encoding: [0x8b,0x4c,0x24,0x04]
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x0c]
; X86-NEXT:    vmovdqu8 %ymm0, (%ecx) {%k1} # encoding: [0x62,0xf1,0x7f,0x29,0x7f,0x01]
; X86-NEXT:    vmovdqu %ymm0, (%eax) # EVEX TO VEX Compression encoding: [0xc5,0xfe,0x7f,0x00]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_storeu_b_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edx, %k1 # encoding: [0xc5,0xfb,0x92,0xca]
; X64-NEXT:    vmovdqu8 %ymm0, (%rdi) {%k1} # encoding: [0x62,0xf1,0x7f,0x29,0x7f,0x07]
; X64-NEXT:    vmovdqu %ymm0, (%rsi) # EVEX TO VEX Compression encoding: [0xc5,0xfe,0x7f,0x06]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
  call void @llvm.x86.avx512.mask.storeu.b.256(i8* %ptr1, <32 x i8> %x1, i32 %x2)
  call void @llvm.x86.avx512.mask.storeu.b.256(i8* %ptr2, <32 x i8> %x1, i32 -1)
  ret void
}

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

define void@test_int_x86_avx512_mask_storeu_w_128(i8* %ptr1, i8* %ptr2, <8 x i16> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_storeu_w_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x08]
; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx # encoding: [0x8b,0x4c,0x24,0x04]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %edx # encoding: [0x0f,0xb6,0x54,0x24,0x0c]
; X86-NEXT:    kmovd %edx, %k1 # encoding: [0xc5,0xfb,0x92,0xca]
; X86-NEXT:    vmovdqu16 %xmm0, (%ecx) {%k1} # encoding: [0x62,0xf1,0xff,0x09,0x7f,0x01]
; X86-NEXT:    vmovdqu %xmm0, (%eax) # EVEX TO VEX Compression encoding: [0xc5,0xfa,0x7f,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_storeu_w_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edx, %k1 # encoding: [0xc5,0xfb,0x92,0xca]
; X64-NEXT:    vmovdqu16 %xmm0, (%rdi) {%k1} # encoding: [0x62,0xf1,0xff,0x09,0x7f,0x07]
; X64-NEXT:    vmovdqu %xmm0, (%rsi) # EVEX TO VEX Compression encoding: [0xc5,0xfa,0x7f,0x06]
; X64-NEXT:    retq # encoding: [0xc3]
  call void @llvm.x86.avx512.mask.storeu.w.128(i8* %ptr1, <8 x i16> %x1, i8 %x2)
  call void @llvm.x86.avx512.mask.storeu.w.128(i8* %ptr2, <8 x i16> %x1, i8 -1)
  ret void
}

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

define void@test_int_x86_avx512_mask_storeu_w_256(i8* %ptr1, i8* %ptr2, <16 x i16> %x1, i16 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_storeu_w_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x08]
; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx # encoding: [0x8b,0x4c,0x24,0x04]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x0c]
; X86-NEXT:    vmovdqu16 %ymm0, (%ecx) {%k1} # encoding: [0x62,0xf1,0xff,0x29,0x7f,0x01]
; X86-NEXT:    vmovdqu %ymm0, (%eax) # EVEX TO VEX Compression encoding: [0xc5,0xfe,0x7f,0x00]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_storeu_w_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edx, %k1 # encoding: [0xc5,0xfb,0x92,0xca]
; X64-NEXT:    vmovdqu16 %ymm0, (%rdi) {%k1} # encoding: [0x62,0xf1,0xff,0x29,0x7f,0x07]
; X64-NEXT:    vmovdqu %ymm0, (%rsi) # EVEX TO VEX Compression encoding: [0xc5,0xfe,0x7f,0x06]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
  call void @llvm.x86.avx512.mask.storeu.w.256(i8* %ptr1, <16 x i16> %x1, i16 %x2)
  call void @llvm.x86.avx512.mask.storeu.w.256(i8* %ptr2, <16 x i16> %x1, i16 -1)
  ret void
}

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

define <8 x i16>@test_int_x86_avx512_mask_loadu_w_128(i8* %ptr, i8* %ptr2, <8 x i16> %x1, i8 %mask) {
; X86-LABEL: test_int_x86_avx512_mask_loadu_w_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x08]
; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx # encoding: [0x8b,0x4c,0x24,0x04]
; X86-NEXT:    vmovdqu (%ecx), %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xfa,0x6f,0x01]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %edx # encoding: [0x0f,0xb6,0x54,0x24,0x0c]
; X86-NEXT:    kmovd %edx, %k1 # encoding: [0xc5,0xfb,0x92,0xca]
; X86-NEXT:    vmovdqu16 (%eax), %xmm0 {%k1} # encoding: [0x62,0xf1,0xff,0x09,0x6f,0x00]
; X86-NEXT:    vmovdqu16 (%ecx), %xmm1 {%k1} {z} # encoding: [0x62,0xf1,0xff,0x89,0x6f,0x09]
; X86-NEXT:    vpaddw %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_loadu_w_128:
; X64:       # %bb.0:
; X64-NEXT:    vmovdqu (%rdi), %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xfa,0x6f,0x07]
; X64-NEXT:    kmovd %edx, %k1 # encoding: [0xc5,0xfb,0x92,0xca]
; X64-NEXT:    vmovdqu16 (%rsi), %xmm0 {%k1} # encoding: [0x62,0xf1,0xff,0x09,0x6f,0x06]
; X64-NEXT:    vmovdqu16 (%rdi), %xmm1 {%k1} {z} # encoding: [0x62,0xf1,0xff,0x89,0x6f,0x0f]
; X64-NEXT:    vpaddw %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
    %res0 = call <8 x i16> @llvm.x86.avx512.mask.loadu.w.128(i8* %ptr, <8 x i16> %x1, i8 -1)
    %res = call <8 x i16> @llvm.x86.avx512.mask.loadu.w.128(i8* %ptr2, <8 x i16> %res0, i8 %mask)
    %res1 = call <8 x i16> @llvm.x86.avx512.mask.loadu.w.128(i8* %ptr, <8 x i16> zeroinitializer, i8 %mask)
    %res2 = add <8 x i16> %res, %res1
    ret <8 x i16> %res2
}

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

define <16 x i16>@test_int_x86_avx512_mask_loadu_w_256(i8* %ptr, i8* %ptr2, <16 x i16> %x1, i16 %mask) {
; X86-LABEL: test_int_x86_avx512_mask_loadu_w_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x08]
; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx # encoding: [0x8b,0x4c,0x24,0x04]
; X86-NEXT:    vmovdqu (%ecx), %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfe,0x6f,0x01]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x0c]
; X86-NEXT:    vmovdqu16 (%eax), %ymm0 {%k1} # encoding: [0x62,0xf1,0xff,0x29,0x6f,0x00]
; X86-NEXT:    vmovdqu16 (%ecx), %ymm1 {%k1} {z} # encoding: [0x62,0xf1,0xff,0xa9,0x6f,0x09]
; X86-NEXT:    vpaddw %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_loadu_w_256:
; X64:       # %bb.0:
; X64-NEXT:    vmovdqu (%rdi), %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfe,0x6f,0x07]
; X64-NEXT:    kmovd %edx, %k1 # encoding: [0xc5,0xfb,0x92,0xca]
; X64-NEXT:    vmovdqu16 (%rsi), %ymm0 {%k1} # encoding: [0x62,0xf1,0xff,0x29,0x6f,0x06]
; X64-NEXT:    vmovdqu16 (%rdi), %ymm1 {%k1} {z} # encoding: [0x62,0xf1,0xff,0xa9,0x6f,0x0f]
; X64-NEXT:    vpaddw %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
    %res0 = call <16 x i16> @llvm.x86.avx512.mask.loadu.w.256(i8* %ptr, <16 x i16> %x1, i16 -1)
    %res = call <16 x i16> @llvm.x86.avx512.mask.loadu.w.256(i8* %ptr2, <16 x i16> %res0, i16 %mask)
    %res1 = call <16 x i16> @llvm.x86.avx512.mask.loadu.w.256(i8* %ptr, <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.loadu.b.128(i8*, <16 x i8>, i16)

define <16 x i8>@test_int_x86_avx512_mask_loadu_b_128(i8* %ptr, i8* %ptr2, <16 x i8> %x1, i16 %mask) {
; X86-LABEL: test_int_x86_avx512_mask_loadu_b_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x08]
; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx # encoding: [0x8b,0x4c,0x24,0x04]
; X86-NEXT:    vmovdqu (%ecx), %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xfa,0x6f,0x01]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x0c]
; X86-NEXT:    vmovdqu8 (%eax), %xmm0 {%k1} # encoding: [0x62,0xf1,0x7f,0x09,0x6f,0x00]
; X86-NEXT:    vmovdqu8 (%ecx), %xmm1 {%k1} {z} # encoding: [0x62,0xf1,0x7f,0x89,0x6f,0x09]
; X86-NEXT:    vpaddb %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfc,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_loadu_b_128:
; X64:       # %bb.0:
; X64-NEXT:    vmovdqu (%rdi), %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xfa,0x6f,0x07]
; X64-NEXT:    kmovd %edx, %k1 # encoding: [0xc5,0xfb,0x92,0xca]
; X64-NEXT:    vmovdqu8 (%rsi), %xmm0 {%k1} # encoding: [0x62,0xf1,0x7f,0x09,0x6f,0x06]
; X64-NEXT:    vmovdqu8 (%rdi), %xmm1 {%k1} {z} # encoding: [0x62,0xf1,0x7f,0x89,0x6f,0x0f]
; X64-NEXT:    vpaddb %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfc,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
    %res0 = call <16 x i8> @llvm.x86.avx512.mask.loadu.b.128(i8* %ptr, <16 x i8> %x1, i16 -1)
    %res = call <16 x i8> @llvm.x86.avx512.mask.loadu.b.128(i8* %ptr2, <16 x i8> %res0, i16 %mask)
    %res1 = call <16 x i8> @llvm.x86.avx512.mask.loadu.b.128(i8* %ptr, <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.loadu.b.256(i8*, <32 x i8>, i32)

define <32 x i8>@test_int_x86_avx512_mask_loadu_b_256(i8* %ptr, i8* %ptr2, <32 x i8> %x1, i32 %mask) {
; X86-LABEL: test_int_x86_avx512_mask_loadu_b_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x08]
; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx # encoding: [0x8b,0x4c,0x24,0x04]
; X86-NEXT:    vmovdqu (%ecx), %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfe,0x6f,0x01]
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x0c]
; X86-NEXT:    vmovdqu8 (%eax), %ymm0 {%k1} # encoding: [0x62,0xf1,0x7f,0x29,0x6f,0x00]
; X86-NEXT:    vmovdqu8 (%ecx), %ymm1 {%k1} {z} # encoding: [0x62,0xf1,0x7f,0xa9,0x6f,0x09]
; X86-NEXT:    vpaddb %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfc,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_loadu_b_256:
; X64:       # %bb.0:
; X64-NEXT:    vmovdqu (%rdi), %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfe,0x6f,0x07]
; X64-NEXT:    kmovd %edx, %k1 # encoding: [0xc5,0xfb,0x92,0xca]
; X64-NEXT:    vmovdqu8 (%rsi), %ymm0 {%k1} # encoding: [0x62,0xf1,0x7f,0x29,0x6f,0x06]
; X64-NEXT:    vmovdqu8 (%rdi), %ymm1 {%k1} {z} # encoding: [0x62,0xf1,0x7f,0xa9,0x6f,0x0f]
; X64-NEXT:    vpaddb %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfc,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
    %res0 = call <32 x i8> @llvm.x86.avx512.mask.loadu.b.256(i8* %ptr, <32 x i8> %x1, i32 -1)
    %res = call <32 x i8> @llvm.x86.avx512.mask.loadu.b.256(i8* %ptr2, <32 x i8> %res0, i32 %mask)
    %res1 = call <32 x i8> @llvm.x86.avx512.mask.loadu.b.256(i8* %ptr, <32 x i8> zeroinitializer, i32 %mask)
    %res2 = add <32 x i8> %res, %res1
    ret <32 x i8> %res2
}

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

define <16 x i8>@test_int_x86_avx512_mask_palignr_128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %x3, i16 %x4) {
; X86-LABEL: test_int_x86_avx512_mask_palignr_128:
; X86:       # %bb.0:
; X86-NEXT:    vpalignr $2, %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x0f,0xd9,0x02]
; X86-NEXT:    # xmm3 = xmm1[2,3,4,5,6,7,8,9,10,11,12,13,14,15],xmm0[0,1]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpalignr $2, %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x0f,0xd1,0x02]
; X86-NEXT:    # xmm2 {%k1} = xmm1[2,3,4,5,6,7,8,9,10,11,12,13,14,15],xmm0[0,1]
; X86-NEXT:    vpalignr $2, %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0x89,0x0f,0xc1,0x02]
; X86-NEXT:    # xmm0 {%k1} {z} = xmm1[2,3,4,5,6,7,8,9,10,11,12,13,14,15],xmm0[0,1]
; X86-NEXT:    vpaddb %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfc,0xc3]
; X86-NEXT:    vpaddb %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_palignr_128:
; X64:       # %bb.0:
; X64-NEXT:    vpalignr $2, %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x0f,0xd9,0x02]
; X64-NEXT:    # xmm3 = xmm1[2,3,4,5,6,7,8,9,10,11,12,13,14,15],xmm0[0,1]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpalignr $2, %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x0f,0xd1,0x02]
; X64-NEXT:    # xmm2 {%k1} = xmm1[2,3,4,5,6,7,8,9,10,11,12,13,14,15],xmm0[0,1]
; X64-NEXT:    vpalignr $2, %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0x89,0x0f,0xc1,0x02]
; X64-NEXT:    # xmm0 {%k1} {z} = xmm1[2,3,4,5,6,7,8,9,10,11,12,13,14,15],xmm0[0,1]
; X64-NEXT:    vpaddb %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfc,0xc3]
; X64-NEXT:    vpaddb %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <16 x i8> @llvm.x86.avx512.mask.palignr.128(<16 x i8> %x0, <16 x i8> %x1, i32 2, <16 x i8> %x3, i16 %x4)
  %res1 = call <16 x i8> @llvm.x86.avx512.mask.palignr.128(<16 x i8> %x0, <16 x i8> %x1, i32 2, <16 x i8> zeroinitializer, i16 %x4)
  %res2 = call <16 x i8> @llvm.x86.avx512.mask.palignr.128(<16 x i8> %x0, <16 x i8> %x1, i32 2, <16 x i8> %x3, i16 -1)
  %res3 = add <16 x i8> %res, %res1
  %res4 = add <16 x i8> %res3, %res2
  ret <16 x i8> %res4
}

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

define <32 x i8>@test_int_x86_avx512_mask_palignr_256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x3, i32 %x4) {
; X86-LABEL: test_int_x86_avx512_mask_palignr_256:
; X86:       # %bb.0:
; X86-NEXT:    vpalignr $2, %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x7d,0x0f,0xd9,0x02]
; X86-NEXT:    # ymm3 = ymm1[2,3,4,5,6,7,8,9,10,11,12,13,14,15],ymm0[0,1],ymm1[18,19,20,21,22,23,24,25,26,27,28,29,30,31],ymm0[16,17]
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpalignr $2, %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x0f,0xd1,0x02]
; X86-NEXT:    # ymm2 {%k1} = ymm1[2,3,4,5,6,7,8,9,10,11,12,13,14,15],ymm0[0,1],ymm1[18,19,20,21,22,23,24,25,26,27,28,29,30,31],ymm0[16,17]
; X86-NEXT:    vpalignr $2, %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0xa9,0x0f,0xc1,0x02]
; X86-NEXT:    # ymm0 {%k1} {z} = ymm1[2,3,4,5,6,7,8,9,10,11,12,13,14,15],ymm0[0,1],ymm1[18,19,20,21,22,23,24,25,26,27,28,29,30,31],ymm0[16,17]
; X86-NEXT:    vpaddb %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfc,0xc3]
; X86-NEXT:    vpaddb %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_palignr_256:
; X64:       # %bb.0:
; X64-NEXT:    vpalignr $2, %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x7d,0x0f,0xd9,0x02]
; X64-NEXT:    # ymm3 = ymm1[2,3,4,5,6,7,8,9,10,11,12,13,14,15],ymm0[0,1],ymm1[18,19,20,21,22,23,24,25,26,27,28,29,30,31],ymm0[16,17]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpalignr $2, %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x0f,0xd1,0x02]
; X64-NEXT:    # ymm2 {%k1} = ymm1[2,3,4,5,6,7,8,9,10,11,12,13,14,15],ymm0[0,1],ymm1[18,19,20,21,22,23,24,25,26,27,28,29,30,31],ymm0[16,17]
; X64-NEXT:    vpalignr $2, %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0xa9,0x0f,0xc1,0x02]
; X64-NEXT:    # ymm0 {%k1} {z} = ymm1[2,3,4,5,6,7,8,9,10,11,12,13,14,15],ymm0[0,1],ymm1[18,19,20,21,22,23,24,25,26,27,28,29,30,31],ymm0[16,17]
; X64-NEXT:    vpaddb %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfc,0xc3]
; X64-NEXT:    vpaddb %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <32 x i8> @llvm.x86.avx512.mask.palignr.256(<32 x i8> %x0, <32 x i8> %x1, i32 2, <32 x i8> %x3, i32 %x4)
  %res1 = call <32 x i8> @llvm.x86.avx512.mask.palignr.256(<32 x i8> %x0, <32 x i8> %x1, i32 2, <32 x i8> zeroinitializer, i32 %x4)
  %res2 = call <32 x i8> @llvm.x86.avx512.mask.palignr.256(<32 x i8> %x0, <32 x i8> %x1, i32 2, <32 x i8> %x3, i32 -1)
  %res3 = add <32 x i8> %res, %res1
  %res4 = add <32 x i8> %res3, %res2
  ret <32 x i8> %res4
}

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

define <8 x i16>@test_int_x86_avx512_mask_pshufh_w_128(<8 x i16> %x0, i32 %x1, <8 x i16> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_pshufh_w_128:
; X86:       # %bb.0:
; X86-NEXT:    vpshufhw $3, %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xfa,0x70,0xd0,0x03]
; X86-NEXT:    # xmm2 = xmm0[0,1,2,3,7,4,4,4]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpshufhw $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7e,0x09,0x70,0xc8,0x03]
; X86-NEXT:    # xmm1 {%k1} = xmm0[0,1,2,3,7,4,4,4]
; X86-NEXT:    vpshufhw $3, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7e,0x89,0x70,0xc0,0x03]
; X86-NEXT:    # xmm0 {%k1} {z} = xmm0[0,1,2,3,7,4,4,4]
; X86-NEXT:    vpaddw %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc2]
; X86-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pshufh_w_128:
; X64:       # %bb.0:
; X64-NEXT:    vpshufhw $3, %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xfa,0x70,0xd0,0x03]
; X64-NEXT:    # xmm2 = xmm0[0,1,2,3,7,4,4,4]
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpshufhw $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7e,0x09,0x70,0xc8,0x03]
; X64-NEXT:    # xmm1 {%k1} = xmm0[0,1,2,3,7,4,4,4]
; X64-NEXT:    vpshufhw $3, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7e,0x89,0x70,0xc0,0x03]
; X64-NEXT:    # xmm0 {%k1} {z} = xmm0[0,1,2,3,7,4,4,4]
; X64-NEXT:    vpaddw %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc2]
; X64-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.mask.pshufh.w.128(<8 x i16> %x0, i32 3, <8 x i16> %x2, i8 %x3)
  %res1 = call <8 x i16> @llvm.x86.avx512.mask.pshufh.w.128(<8 x i16> %x0, i32 3, <8 x i16> zeroinitializer, i8 %x3)
  %res2 = call <8 x i16> @llvm.x86.avx512.mask.pshufh.w.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.pshufh.w.256(<16 x i16>, i32, <16 x i16>, i16)

define <16 x i16>@test_int_x86_avx512_mask_pshufh_w_256(<16 x i16> %x0, i32 %x1, <16 x i16> %x2, i16 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_pshufh_w_256:
; X86:       # %bb.0:
; X86-NEXT:    vpshufhw $3, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xfe,0x70,0xd0,0x03]
; X86-NEXT:    # ymm2 = ymm0[0,1,2,3,7,4,4,4,8,9,10,11,15,12,12,12]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpshufhw $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7e,0x29,0x70,0xc8,0x03]
; X86-NEXT:    # ymm1 {%k1} = ymm0[0,1,2,3,7,4,4,4,8,9,10,11,15,12,12,12]
; X86-NEXT:    vpshufhw $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7e,0xa9,0x70,0xc0,0x03]
; X86-NEXT:    # ymm0 {%k1} {z} = ymm0[0,1,2,3,7,4,4,4,8,9,10,11,15,12,12,12]
; X86-NEXT:    vpaddw %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc2]
; X86-NEXT:    vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pshufh_w_256:
; X64:       # %bb.0:
; X64-NEXT:    vpshufhw $3, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xfe,0x70,0xd0,0x03]
; X64-NEXT:    # ymm2 = ymm0[0,1,2,3,7,4,4,4,8,9,10,11,15,12,12,12]
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpshufhw $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7e,0x29,0x70,0xc8,0x03]
; X64-NEXT:    # ymm1 {%k1} = ymm0[0,1,2,3,7,4,4,4,8,9,10,11,15,12,12,12]
; X64-NEXT:    vpshufhw $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7e,0xa9,0x70,0xc0,0x03]
; X64-NEXT:    # ymm0 {%k1} {z} = ymm0[0,1,2,3,7,4,4,4,8,9,10,11,15,12,12,12]
; X64-NEXT:    vpaddw %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc2]
; X64-NEXT:    vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.mask.pshufh.w.256(<16 x i16> %x0, i32 3, <16 x i16> %x2, i16 %x3)
  %res1 = call <16 x i16> @llvm.x86.avx512.mask.pshufh.w.256(<16 x i16> %x0, i32 3, <16 x i16> zeroinitializer, i16 %x3)
  %res2 = call <16 x i16> @llvm.x86.avx512.mask.pshufh.w.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 <8 x i16> @llvm.x86.avx512.mask.pshufl.w.128(<8 x i16>, i32, <8 x i16>, i8)

define <8 x i16>@test_int_x86_avx512_mask_pshufl_w_128(<8 x i16> %x0, i32 %x1, <8 x i16> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_pshufl_w_128:
; X86:       # %bb.0:
; X86-NEXT:    vpshuflw $3, %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xfb,0x70,0xd0,0x03]
; X86-NEXT:    # xmm2 = xmm0[3,0,0,0,4,5,6,7]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpshuflw $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7f,0x09,0x70,0xc8,0x03]
; X86-NEXT:    # xmm1 {%k1} = xmm0[3,0,0,0,4,5,6,7]
; X86-NEXT:    vpshuflw $3, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7f,0x89,0x70,0xc0,0x03]
; X86-NEXT:    # xmm0 {%k1} {z} = xmm0[3,0,0,0,4,5,6,7]
; X86-NEXT:    vpaddw %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc2]
; X86-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pshufl_w_128:
; X64:       # %bb.0:
; X64-NEXT:    vpshuflw $3, %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xfb,0x70,0xd0,0x03]
; X64-NEXT:    # xmm2 = xmm0[3,0,0,0,4,5,6,7]
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpshuflw $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7f,0x09,0x70,0xc8,0x03]
; X64-NEXT:    # xmm1 {%k1} = xmm0[3,0,0,0,4,5,6,7]
; X64-NEXT:    vpshuflw $3, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7f,0x89,0x70,0xc0,0x03]
; X64-NEXT:    # xmm0 {%k1} {z} = xmm0[3,0,0,0,4,5,6,7]
; X64-NEXT:    vpaddw %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc2]
; X64-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.mask.pshufl.w.128(<8 x i16> %x0, i32 3, <8 x i16> %x2, i8 %x3)
  %res1 = call <8 x i16> @llvm.x86.avx512.mask.pshufl.w.128(<8 x i16> %x0, i32 3, <8 x i16> zeroinitializer, i8 %x3)
  %res2 = call <8 x i16> @llvm.x86.avx512.mask.pshufl.w.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.pshufl.w.256(<16 x i16>, i32, <16 x i16>, i16)

define <16 x i16>@test_int_x86_avx512_mask_pshufl_w_256(<16 x i16> %x0, i32 %x1, <16 x i16> %x2, i16 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_pshufl_w_256:
; X86:       # %bb.0:
; X86-NEXT:    vpshuflw $3, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xff,0x70,0xd0,0x03]
; X86-NEXT:    # ymm2 = ymm0[3,0,0,0,4,5,6,7,11,8,8,8,12,13,14,15]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpshuflw $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7f,0x29,0x70,0xc8,0x03]
; X86-NEXT:    # ymm1 {%k1} = ymm0[3,0,0,0,4,5,6,7,11,8,8,8,12,13,14,15]
; X86-NEXT:    vpshuflw $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7f,0xa9,0x70,0xc0,0x03]
; X86-NEXT:    # ymm0 {%k1} {z} = ymm0[3,0,0,0,4,5,6,7,11,8,8,8,12,13,14,15]
; X86-NEXT:    vpaddw %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc2]
; X86-NEXT:    vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pshufl_w_256:
; X64:       # %bb.0:
; X64-NEXT:    vpshuflw $3, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xff,0x70,0xd0,0x03]
; X64-NEXT:    # ymm2 = ymm0[3,0,0,0,4,5,6,7,11,8,8,8,12,13,14,15]
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpshuflw $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7f,0x29,0x70,0xc8,0x03]
; X64-NEXT:    # ymm1 {%k1} = ymm0[3,0,0,0,4,5,6,7,11,8,8,8,12,13,14,15]
; X64-NEXT:    vpshuflw $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7f,0xa9,0x70,0xc0,0x03]
; X64-NEXT:    # ymm0 {%k1} {z} = ymm0[3,0,0,0,4,5,6,7,11,8,8,8,12,13,14,15]
; X64-NEXT:    vpaddw %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc2]
; X64-NEXT:    vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.mask.pshufl.w.256(<16 x i16> %x0, i32 3, <16 x i16> %x2, i16 %x3)
  %res1 = call <16 x i16> @llvm.x86.avx512.mask.pshufl.w.256(<16 x i16> %x0, i32 3, <16 x i16> zeroinitializer, i16 %x3)
  %res2 = call <16 x i16> @llvm.x86.avx512.mask.pshufl.w.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
}

define i32 @test_pcmpeq_b_256(<32 x i8> %a, <32 x i8> %b) {
; CHECK-LABEL: test_pcmpeq_b_256:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vpcmpeqb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x74,0xc1]
; CHECK-NEXT:    kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
; CHECK-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call i32 @llvm.x86.avx512.mask.pcmpeq.b.256(<32 x i8> %a, <32 x i8> %b, i32 -1)
  ret i32 %res
}

define i32 @test_mask_pcmpeq_b_256(<32 x i8> %a, <32 x i8> %b, i32 %mask) {
; X86-LABEL: test_mask_pcmpeq_b_256:
; X86:       # %bb.0:
; X86-NEXT:    vpcmpeqb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x74,0xc1]
; X86-NEXT:    kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
; X86-NEXT:    andl {{[0-9]+}}(%esp), %eax # encoding: [0x23,0x44,0x24,0x04]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_pcmpeq_b_256:
; X64:       # %bb.0:
; X64-NEXT:    vpcmpeqb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x74,0xc1]
; X64-NEXT:    kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
; X64-NEXT:    andl %edi, %eax # encoding: [0x21,0xf8]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call i32 @llvm.x86.avx512.mask.pcmpeq.b.256(<32 x i8> %a, <32 x i8> %b, i32 %mask)
  ret i32 %res
}

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

define i16 @test_pcmpeq_w_256(<16 x i16> %a, <16 x i16> %b) {
; CHECK-LABEL: test_pcmpeq_w_256:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vpcmpeqw %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x75,0xc1]
; CHECK-NEXT:    kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
; CHECK-NEXT:    # kill: def $ax killed $ax killed $eax
; CHECK-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call i16 @llvm.x86.avx512.mask.pcmpeq.w.256(<16 x i16> %a, <16 x i16> %b, i16 -1)
  ret i16 %res
}

define i16 @test_mask_pcmpeq_w_256(<16 x i16> %a, <16 x i16> %b, i16 %mask) {
; X86-LABEL: test_mask_pcmpeq_w_256:
; X86:       # %bb.0:
; X86-NEXT:    vpcmpeqw %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x75,0xc1]
; X86-NEXT:    kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
; X86-NEXT:    andw {{[0-9]+}}(%esp), %ax # encoding: [0x66,0x23,0x44,0x24,0x04]
; X86-NEXT:    # kill: def $ax killed $ax killed $eax
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_pcmpeq_w_256:
; X64:       # %bb.0:
; X64-NEXT:    vpcmpeqw %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x75,0xc1]
; X64-NEXT:    kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
; X64-NEXT:    andl %edi, %eax # encoding: [0x21,0xf8]
; X64-NEXT:    # kill: def $ax killed $ax killed $eax
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call i16 @llvm.x86.avx512.mask.pcmpeq.w.256(<16 x i16> %a, <16 x i16> %b, i16 %mask)
  ret i16 %res
}

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

define i32 @test_pcmpgt_b_256(<32 x i8> %a, <32 x i8> %b) {
; CHECK-LABEL: test_pcmpgt_b_256:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vpcmpgtb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x64,0xc1]
; CHECK-NEXT:    kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
; CHECK-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call i32 @llvm.x86.avx512.mask.pcmpgt.b.256(<32 x i8> %a, <32 x i8> %b, i32 -1)
  ret i32 %res
}

define i32 @test_mask_pcmpgt_b_256(<32 x i8> %a, <32 x i8> %b, i32 %mask) {
; X86-LABEL: test_mask_pcmpgt_b_256:
; X86:       # %bb.0:
; X86-NEXT:    vpcmpgtb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x64,0xc1]
; X86-NEXT:    kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
; X86-NEXT:    andl {{[0-9]+}}(%esp), %eax # encoding: [0x23,0x44,0x24,0x04]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_pcmpgt_b_256:
; X64:       # %bb.0:
; X64-NEXT:    vpcmpgtb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x64,0xc1]
; X64-NEXT:    kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
; X64-NEXT:    andl %edi, %eax # encoding: [0x21,0xf8]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call i32 @llvm.x86.avx512.mask.pcmpgt.b.256(<32 x i8> %a, <32 x i8> %b, i32 %mask)
  ret i32 %res
}

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

define i16 @test_pcmpgt_w_256(<16 x i16> %a, <16 x i16> %b) {
; CHECK-LABEL: test_pcmpgt_w_256:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vpcmpgtw %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x65,0xc1]
; CHECK-NEXT:    kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
; CHECK-NEXT:    # kill: def $ax killed $ax killed $eax
; CHECK-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call i16 @llvm.x86.avx512.mask.pcmpgt.w.256(<16 x i16> %a, <16 x i16> %b, i16 -1)
  ret i16 %res
}

define i16 @test_mask_pcmpgt_w_256(<16 x i16> %a, <16 x i16> %b, i16 %mask) {
; X86-LABEL: test_mask_pcmpgt_w_256:
; X86:       # %bb.0:
; X86-NEXT:    vpcmpgtw %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x65,0xc1]
; X86-NEXT:    kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
; X86-NEXT:    andw {{[0-9]+}}(%esp), %ax # encoding: [0x66,0x23,0x44,0x24,0x04]
; X86-NEXT:    # kill: def $ax killed $ax killed $eax
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_pcmpgt_w_256:
; X64:       # %bb.0:
; X64-NEXT:    vpcmpgtw %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x65,0xc1]
; X64-NEXT:    kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
; X64-NEXT:    andl %edi, %eax # encoding: [0x21,0xf8]
; X64-NEXT:    # kill: def $ax killed $ax killed $eax
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call i16 @llvm.x86.avx512.mask.pcmpgt.w.256(<16 x i16> %a, <16 x i16> %b, i16 %mask)
  ret i16 %res
}

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

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:    kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
; CHECK-NEXT:    # kill: def $ax killed $ax killed $eax
; CHECK-NEXT:    ret{{[l|q]}} # 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) {
; X86-LABEL: test_mask_pcmpeq_b_128:
; X86:       # %bb.0:
; X86-NEXT:    vpcmpeqb %xmm1, %xmm0, %k0 # encoding: [0x62,0xf1,0x7d,0x08,0x74,0xc1]
; X86-NEXT:    kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
; X86-NEXT:    andw {{[0-9]+}}(%esp), %ax # encoding: [0x66,0x23,0x44,0x24,0x04]
; X86-NEXT:    # kill: def $ax killed $ax killed $eax
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_pcmpeq_b_128:
; X64:       # %bb.0:
; X64-NEXT:    vpcmpeqb %xmm1, %xmm0, %k0 # encoding: [0x62,0xf1,0x7d,0x08,0x74,0xc1]
; X64-NEXT:    kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
; X64-NEXT:    andl %edi, %eax # encoding: [0x21,0xf8]
; X64-NEXT:    # kill: def $ax killed $ax killed $eax
; X64-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:    kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
; CHECK-NEXT:    # kill: def $al killed $al killed $eax
; CHECK-NEXT:    ret{{[l|q]}} # 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) {
; X86-LABEL: test_mask_pcmpeq_w_128:
; X86:       # %bb.0:
; X86-NEXT:    vpcmpeqw %xmm1, %xmm0, %k0 # encoding: [0x62,0xf1,0x7d,0x08,0x75,0xc1]
; X86-NEXT:    kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
; X86-NEXT:    andb {{[0-9]+}}(%esp), %al # encoding: [0x22,0x44,0x24,0x04]
; X86-NEXT:    # kill: def $al killed $al killed $eax
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_pcmpeq_w_128:
; X64:       # %bb.0:
; X64-NEXT:    vpcmpeqw %xmm1, %xmm0, %k0 # encoding: [0x62,0xf1,0x7d,0x08,0x75,0xc1]
; X64-NEXT:    kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
; X64-NEXT:    andb %dil, %al # encoding: [0x40,0x20,0xf8]
; X64-NEXT:    # kill: def $al killed $al killed $eax
; X64-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:    kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
; CHECK-NEXT:    # kill: def $ax killed $ax killed $eax
; CHECK-NEXT:    ret{{[l|q]}} # 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) {
; X86-LABEL: test_mask_pcmpgt_b_128:
; X86:       # %bb.0:
; X86-NEXT:    vpcmpgtb %xmm1, %xmm0, %k0 # encoding: [0x62,0xf1,0x7d,0x08,0x64,0xc1]
; X86-NEXT:    kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
; X86-NEXT:    andw {{[0-9]+}}(%esp), %ax # encoding: [0x66,0x23,0x44,0x24,0x04]
; X86-NEXT:    # kill: def $ax killed $ax killed $eax
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_pcmpgt_b_128:
; X64:       # %bb.0:
; X64-NEXT:    vpcmpgtb %xmm1, %xmm0, %k0 # encoding: [0x62,0xf1,0x7d,0x08,0x64,0xc1]
; X64-NEXT:    kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
; X64-NEXT:    andl %edi, %eax # encoding: [0x21,0xf8]
; X64-NEXT:    # kill: def $ax killed $ax killed $eax
; X64-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:    kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
; CHECK-NEXT:    # kill: def $al killed $al killed $eax
; CHECK-NEXT:    ret{{[l|q]}} # 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) {
; X86-LABEL: test_mask_pcmpgt_w_128:
; X86:       # %bb.0:
; X86-NEXT:    vpcmpgtw %xmm1, %xmm0, %k0 # encoding: [0x62,0xf1,0x7d,0x08,0x65,0xc1]
; X86-NEXT:    kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
; X86-NEXT:    andb {{[0-9]+}}(%esp), %al # encoding: [0x22,0x44,0x24,0x04]
; X86-NEXT:    # kill: def $al killed $al killed $eax
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_pcmpgt_w_128:
; X64:       # %bb.0:
; X64-NEXT:    vpcmpgtw %xmm1, %xmm0, %k0 # encoding: [0x62,0xf1,0x7d,0x08,0x65,0xc1]
; X64-NEXT:    kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
; X64-NEXT:    andb %dil, %al # encoding: [0x40,0x20,0xf8]
; X64-NEXT:    # kill: def $al killed $al killed $eax
; X64-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)

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

define <16 x i8>@test_int_x86_avx512_mask_punpckhb_w_128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %x2, i16 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_punpckhb_w_128:
; X86:       # %bb.0:
; X86-NEXT:    vpunpckhbw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x68,0xd9]
; X86-NEXT:    # xmm3 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpunpckhbw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x68,0xd1]
; X86-NEXT:    # xmm2 {%k1} = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
; X86-NEXT:    vpaddb %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_punpckhb_w_128:
; X64:       # %bb.0:
; X64-NEXT:    vpunpckhbw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x68,0xd9]
; X64-NEXT:    # xmm3 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpunpckhbw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x68,0xd1]
; X64-NEXT:    # xmm2 {%k1} = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
; X64-NEXT:    vpaddb %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <16 x i8> @llvm.x86.avx512.mask.punpckhb.w.128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %x2, i16 %x3)
  %res1 = call <16 x i8> @llvm.x86.avx512.mask.punpckhb.w.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 <16 x i8> @llvm.x86.avx512.mask.punpcklb.w.128(<16 x i8>, <16 x i8>, <16 x i8>, i16)

define <16 x i8>@test_int_x86_avx512_mask_punpcklb_w_128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %x2, i16 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_punpcklb_w_128:
; X86:       # %bb.0:
; X86-NEXT:    vpunpcklbw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x60,0xd9]
; X86-NEXT:    # xmm3 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpunpcklbw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x60,0xd1]
; X86-NEXT:    # xmm2 {%k1} = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
; X86-NEXT:    vpaddb %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_punpcklb_w_128:
; X64:       # %bb.0:
; X64-NEXT:    vpunpcklbw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x60,0xd9]
; X64-NEXT:    # xmm3 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpunpcklbw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x60,0xd1]
; X64-NEXT:    # xmm2 {%k1} = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
; X64-NEXT:    vpaddb %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <16 x i8> @llvm.x86.avx512.mask.punpcklb.w.128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %x2, i16 %x3)
  %res1 = call <16 x i8> @llvm.x86.avx512.mask.punpcklb.w.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.punpckhb.w.256(<32 x i8>, <32 x i8>, <32 x i8>, i32)

define <32 x i8>@test_int_x86_avx512_mask_punpckhb_w_256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_punpckhb_w_256:
; X86:       # %bb.0:
; X86-NEXT:    vpunpckhbw %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x68,0xd9]
; X86-NEXT:    # ymm3 = ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11],ymm0[12],ymm1[12],ymm0[13],ymm1[13],ymm0[14],ymm1[14],ymm0[15],ymm1[15],ymm0[24],ymm1[24],ymm0[25],ymm1[25],ymm0[26],ymm1[26],ymm0[27],ymm1[27],ymm0[28],ymm1[28],ymm0[29],ymm1[29],ymm0[30],ymm1[30],ymm0[31],ymm1[31]
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpunpckhbw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x68,0xd1]
; X86-NEXT:    # ymm2 {%k1} = ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11],ymm0[12],ymm1[12],ymm0[13],ymm1[13],ymm0[14],ymm1[14],ymm0[15],ymm1[15],ymm0[24],ymm1[24],ymm0[25],ymm1[25],ymm0[26],ymm1[26],ymm0[27],ymm1[27],ymm0[28],ymm1[28],ymm0[29],ymm1[29],ymm0[30],ymm1[30],ymm0[31],ymm1[31]
; X86-NEXT:    vpaddb %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_punpckhb_w_256:
; X64:       # %bb.0:
; X64-NEXT:    vpunpckhbw %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x68,0xd9]
; X64-NEXT:    # ymm3 = ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11],ymm0[12],ymm1[12],ymm0[13],ymm1[13],ymm0[14],ymm1[14],ymm0[15],ymm1[15],ymm0[24],ymm1[24],ymm0[25],ymm1[25],ymm0[26],ymm1[26],ymm0[27],ymm1[27],ymm0[28],ymm1[28],ymm0[29],ymm1[29],ymm0[30],ymm1[30],ymm0[31],ymm1[31]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpunpckhbw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x68,0xd1]
; X64-NEXT:    # ymm2 {%k1} = ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11],ymm0[12],ymm1[12],ymm0[13],ymm1[13],ymm0[14],ymm1[14],ymm0[15],ymm1[15],ymm0[24],ymm1[24],ymm0[25],ymm1[25],ymm0[26],ymm1[26],ymm0[27],ymm1[27],ymm0[28],ymm1[28],ymm0[29],ymm1[29],ymm0[30],ymm1[30],ymm0[31],ymm1[31]
; X64-NEXT:    vpaddb %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <32 x i8> @llvm.x86.avx512.mask.punpckhb.w.256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 %x3)
  %res1 = call <32 x i8> @llvm.x86.avx512.mask.punpckhb.w.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 <32 x i8> @llvm.x86.avx512.mask.punpcklb.w.256(<32 x i8>, <32 x i8>, <32 x i8>, i32)

define <32 x i8>@test_int_x86_avx512_mask_punpcklb_w_256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_punpcklb_w_256:
; X86:       # %bb.0:
; X86-NEXT:    vpunpcklbw %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x60,0xd9]
; X86-NEXT:    # ymm3 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[4],ymm1[4],ymm0[5],ymm1[5],ymm0[6],ymm1[6],ymm0[7],ymm1[7],ymm0[16],ymm1[16],ymm0[17],ymm1[17],ymm0[18],ymm1[18],ymm0[19],ymm1[19],ymm0[20],ymm1[20],ymm0[21],ymm1[21],ymm0[22],ymm1[22],ymm0[23],ymm1[23]
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpunpcklbw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x60,0xd1]
; X86-NEXT:    # ymm2 {%k1} = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[4],ymm1[4],ymm0[5],ymm1[5],ymm0[6],ymm1[6],ymm0[7],ymm1[7],ymm0[16],ymm1[16],ymm0[17],ymm1[17],ymm0[18],ymm1[18],ymm0[19],ymm1[19],ymm0[20],ymm1[20],ymm0[21],ymm1[21],ymm0[22],ymm1[22],ymm0[23],ymm1[23]
; X86-NEXT:    vpaddb %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_punpcklb_w_256:
; X64:       # %bb.0:
; X64-NEXT:    vpunpcklbw %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x60,0xd9]
; X64-NEXT:    # ymm3 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[4],ymm1[4],ymm0[5],ymm1[5],ymm0[6],ymm1[6],ymm0[7],ymm1[7],ymm0[16],ymm1[16],ymm0[17],ymm1[17],ymm0[18],ymm1[18],ymm0[19],ymm1[19],ymm0[20],ymm1[20],ymm0[21],ymm1[21],ymm0[22],ymm1[22],ymm0[23],ymm1[23]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpunpcklbw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x60,0xd1]
; X64-NEXT:    # ymm2 {%k1} = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[4],ymm1[4],ymm0[5],ymm1[5],ymm0[6],ymm1[6],ymm0[7],ymm1[7],ymm0[16],ymm1[16],ymm0[17],ymm1[17],ymm0[18],ymm1[18],ymm0[19],ymm1[19],ymm0[20],ymm1[20],ymm0[21],ymm1[21],ymm0[22],ymm1[22],ymm0[23],ymm1[23]
; X64-NEXT:    vpaddb %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <32 x i8> @llvm.x86.avx512.mask.punpcklb.w.256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 %x3)
  %res1 = call <32 x i8> @llvm.x86.avx512.mask.punpcklb.w.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.punpcklw.d.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)

define <8 x i16>@test_int_x86_avx512_mask_punpcklw_d_128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_punpcklw_d_128:
; X86:       # %bb.0:
; X86-NEXT:    vpunpcklwd %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x61,0xd9]
; X86-NEXT:    # xmm3 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpunpcklwd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x61,0xd1]
; X86-NEXT:    # xmm2 {%k1} = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
; X86-NEXT:    vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_punpcklw_d_128:
; X64:       # %bb.0:
; X64-NEXT:    vpunpcklwd %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x61,0xd9]
; X64-NEXT:    # xmm3 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpunpcklwd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x61,0xd1]
; X64-NEXT:    # xmm2 {%k1} = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
; X64-NEXT:    vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.mask.punpcklw.d.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3)
  %res1 = call <8 x i16> @llvm.x86.avx512.mask.punpcklw.d.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.mask.punpckhw.d.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)

define <8 x i16>@test_int_x86_avx512_mask_punpckhw_d_128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_punpckhw_d_128:
; X86:       # %bb.0:
; X86-NEXT:    vpunpckhwd %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x69,0xd9]
; X86-NEXT:    # xmm3 = xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpunpckhwd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x69,0xd1]
; X86-NEXT:    # xmm2 {%k1} = xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
; X86-NEXT:    vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_punpckhw_d_128:
; X64:       # %bb.0:
; X64-NEXT:    vpunpckhwd %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x69,0xd9]
; X64-NEXT:    # xmm3 = xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpunpckhwd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x69,0xd1]
; X64-NEXT:    # xmm2 {%k1} = xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
; X64-NEXT:    vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x i16> @llvm.x86.avx512.mask.punpckhw.d.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3)
  %res1 = call <8 x i16> @llvm.x86.avx512.mask.punpckhw.d.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.punpcklw.d.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)

define <16 x i16>@test_int_x86_avx512_mask_punpcklw_d_256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_punpcklw_d_256:
; X86:       # %bb.0:
; X86-NEXT:    vpunpcklwd %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x61,0xd9]
; X86-NEXT:    # ymm3 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpunpcklwd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x61,0xd1]
; X86-NEXT:    # ymm2 {%k1} = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11]
; X86-NEXT:    vpaddw %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_punpcklw_d_256:
; X64:       # %bb.0:
; X64-NEXT:    vpunpcklwd %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x61,0xd9]
; X64-NEXT:    # ymm3 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpunpcklwd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x61,0xd1]
; X64-NEXT:    # ymm2 {%k1} = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11]
; X64-NEXT:    vpaddw %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.mask.punpcklw.d.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %x3)
  %res1 = call <16 x i16> @llvm.x86.avx512.mask.punpcklw.d.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.mask.punpckhw.d.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)

define <16 x i16>@test_int_x86_avx512_mask_punpckhw_d_256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_punpckhw_d_256:
; X86:       # %bb.0:
; X86-NEXT:    vpunpckhwd %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x69,0xd9]
; X86-NEXT:    # ymm3 = ymm0[4],ymm1[4],ymm0[5],ymm1[5],ymm0[6],ymm1[6],ymm0[7],ymm1[7],ymm0[12],ymm1[12],ymm0[13],ymm1[13],ymm0[14],ymm1[14],ymm0[15],ymm1[15]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpunpckhwd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x69,0xd1]
; X86-NEXT:    # ymm2 {%k1} = ymm0[4],ymm1[4],ymm0[5],ymm1[5],ymm0[6],ymm1[6],ymm0[7],ymm1[7],ymm0[12],ymm1[12],ymm0[13],ymm1[13],ymm0[14],ymm1[14],ymm0[15],ymm1[15]
; X86-NEXT:    vpaddw %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_punpckhw_d_256:
; X64:       # %bb.0:
; X64-NEXT:    vpunpckhwd %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x69,0xd9]
; X64-NEXT:    # ymm3 = ymm0[4],ymm1[4],ymm0[5],ymm1[5],ymm0[6],ymm1[6],ymm0[7],ymm1[7],ymm0[12],ymm1[12],ymm0[13],ymm1[13],ymm0[14],ymm1[14],ymm0[15],ymm1[15]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpunpckhwd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x69,0xd1]
; X64-NEXT:    # ymm2 {%k1} = ymm0[4],ymm1[4],ymm0[5],ymm1[5],ymm0[6],ymm1[6],ymm0[7],ymm1[7],ymm0[12],ymm1[12],ymm0[13],ymm1[13],ymm0[14],ymm1[14],ymm0[15],ymm1[15]
; X64-NEXT:    vpaddw %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.mask.punpckhw.d.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %x3)
  %res1 = call <16 x i16> @llvm.x86.avx512.mask.punpckhw.d.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
}

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 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # 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) {
; X86-LABEL: test_mask_add_epi16_rrk_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpaddw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xfd,0xd1]
; X86-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_add_epi16_rrk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpaddw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xfd,0xd1]
; X64-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
; X64-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) {
; X86-LABEL: test_mask_add_epi16_rrkz_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpaddw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xfd,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_add_epi16_rrkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpaddw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xfd,0xc1]
; X64-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) {
; X86-LABEL: test_mask_add_epi16_rm_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpaddw (%eax), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_add_epi16_rm_128:
; X64:       # %bb.0:
; X64-NEXT:    vpaddw (%rdi), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0x07]
; X64-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) {
; X86-LABEL: test_mask_add_epi16_rmk_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
; X86-NEXT:    kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
; X86-NEXT:    vpaddw (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xfd,0x08]
; X86-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_add_epi16_rmk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpaddw (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xfd,0x0f]
; X64-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X64-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) {
; X86-LABEL: test_mask_add_epi16_rmkz_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
; X86-NEXT:    kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
; X86-NEXT:    vpaddw (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xfd,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_add_epi16_rmkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpaddw (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xfd,0x07]
; X64-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 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # 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) {
; X86-LABEL: test_mask_add_epi16_rrk_256:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpaddw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xfd,0xd1]
; X86-NEXT:    vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_add_epi16_rrk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpaddw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xfd,0xd1]
; X64-NEXT:    vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
; X64-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) {
; X86-LABEL: test_mask_add_epi16_rrkz_256:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpaddw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xfd,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_add_epi16_rrkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpaddw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xfd,0xc1]
; X64-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) {
; X86-LABEL: test_mask_add_epi16_rm_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpaddw (%eax), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_add_epi16_rm_256:
; X64:       # %bb.0:
; X64-NEXT:    vpaddw (%rdi), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0x07]
; X64-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) {
; X86-LABEL: test_mask_add_epi16_rmk_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpaddw (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xfd,0x08]
; X86-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_add_epi16_rmk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpaddw (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xfd,0x0f]
; X64-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X64-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) {
; X86-LABEL: test_mask_add_epi16_rmkz_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpaddw (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xfd,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_add_epi16_rmkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpaddw (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xfd,0x07]
; X64-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 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xf9,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # 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) {
; X86-LABEL: test_mask_sub_epi16_rrk_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpsubw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xf9,0xd1]
; X86-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_sub_epi16_rrk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpsubw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xf9,0xd1]
; X64-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
; X64-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) {
; X86-LABEL: test_mask_sub_epi16_rrkz_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpsubw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xf9,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_sub_epi16_rrkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpsubw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xf9,0xc1]
; X64-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) {
; X86-LABEL: test_mask_sub_epi16_rm_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpsubw (%eax), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xf9,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_sub_epi16_rm_128:
; X64:       # %bb.0:
; X64-NEXT:    vpsubw (%rdi), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xf9,0x07]
; X64-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) {
; X86-LABEL: test_mask_sub_epi16_rmk_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
; X86-NEXT:    kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
; X86-NEXT:    vpsubw (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xf9,0x08]
; X86-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_sub_epi16_rmk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpsubw (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xf9,0x0f]
; X64-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X64-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) {
; X86-LABEL: test_mask_sub_epi16_rmkz_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
; X86-NEXT:    kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
; X86-NEXT:    vpsubw (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xf9,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_sub_epi16_rmkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpsubw (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xf9,0x07]
; X64-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 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xf9,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # 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) {
; X86-LABEL: test_mask_sub_epi16_rrk_256:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpsubw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xf9,0xd1]
; X86-NEXT:    vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_sub_epi16_rrk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpsubw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xf9,0xd1]
; X64-NEXT:    vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
; X64-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) {
; X86-LABEL: test_mask_sub_epi16_rrkz_256:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpsubw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xf9,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_sub_epi16_rrkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpsubw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xf9,0xc1]
; X64-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) {
; X86-LABEL: test_mask_sub_epi16_rm_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpsubw (%eax), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xf9,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_sub_epi16_rm_256:
; X64:       # %bb.0:
; X64-NEXT:    vpsubw (%rdi), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xf9,0x07]
; X64-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) {
; X86-LABEL: test_mask_sub_epi16_rmk_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpsubw (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xf9,0x08]
; X86-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_sub_epi16_rmk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpsubw (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xf9,0x0f]
; X64-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X64-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) {
; X86-LABEL: test_mask_sub_epi16_rmkz_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpsubw (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xf9,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_sub_epi16_rmkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpsubw (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xf9,0x07]
; X64-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:    ret{{[l|q]}} # 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) {
; X86-LABEL: test_mask_add_epi16_rrk_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpaddw %zmm1, %zmm0, %zmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x49,0xfd,0xd1]
; X86-NEXT:    vmovdqa64 %zmm2, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_add_epi16_rrk_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpaddw %zmm1, %zmm0, %zmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x49,0xfd,0xd1]
; X64-NEXT:    vmovdqa64 %zmm2, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc2]
; X64-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) {
; X86-LABEL: test_mask_add_epi16_rrkz_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpaddw %zmm1, %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xc9,0xfd,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_add_epi16_rrkz_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpaddw %zmm1, %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xc9,0xfd,0xc1]
; X64-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) {
; X86-LABEL: test_mask_add_epi16_rm_512:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpaddw (%eax), %zmm0, %zmm0 # encoding: [0x62,0xf1,0x7d,0x48,0xfd,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_add_epi16_rm_512:
; X64:       # %bb.0:
; X64-NEXT:    vpaddw (%rdi), %zmm0, %zmm0 # encoding: [0x62,0xf1,0x7d,0x48,0xfd,0x07]
; X64-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) {
; X86-LABEL: test_mask_add_epi16_rmk_512:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpaddw (%eax), %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x49,0xfd,0x08]
; X86-NEXT:    vmovdqa64 %zmm1, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_add_epi16_rmk_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpaddw (%rdi), %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x49,0xfd,0x0f]
; X64-NEXT:    vmovdqa64 %zmm1, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc1]
; X64-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) {
; X86-LABEL: test_mask_add_epi16_rmkz_512:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpaddw (%eax), %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xc9,0xfd,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_add_epi16_rmkz_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpaddw (%rdi), %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xc9,0xfd,0x07]
; X64-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:    ret{{[l|q]}} # 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) {
; X86-LABEL: test_mask_sub_epi16_rrk_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpsubw %zmm1, %zmm0, %zmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x49,0xf9,0xd1]
; X86-NEXT:    vmovdqa64 %zmm2, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_sub_epi16_rrk_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpsubw %zmm1, %zmm0, %zmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x49,0xf9,0xd1]
; X64-NEXT:    vmovdqa64 %zmm2, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc2]
; X64-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) {
; X86-LABEL: test_mask_sub_epi16_rrkz_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpsubw %zmm1, %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xc9,0xf9,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_sub_epi16_rrkz_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpsubw %zmm1, %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xc9,0xf9,0xc1]
; X64-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) {
; X86-LABEL: test_mask_sub_epi16_rm_512:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpsubw (%eax), %zmm0, %zmm0 # encoding: [0x62,0xf1,0x7d,0x48,0xf9,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_sub_epi16_rm_512:
; X64:       # %bb.0:
; X64-NEXT:    vpsubw (%rdi), %zmm0, %zmm0 # encoding: [0x62,0xf1,0x7d,0x48,0xf9,0x07]
; X64-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) {
; X86-LABEL: test_mask_sub_epi16_rmk_512:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpsubw (%eax), %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x49,0xf9,0x08]
; X86-NEXT:    vmovdqa64 %zmm1, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_sub_epi16_rmk_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpsubw (%rdi), %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x49,0xf9,0x0f]
; X64-NEXT:    vmovdqa64 %zmm1, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc1]
; X64-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) {
; X86-LABEL: test_mask_sub_epi16_rmkz_512:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpsubw (%eax), %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xc9,0xf9,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_sub_epi16_rmkz_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpsubw (%rdi), %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xc9,0xf9,0x07]
; X64-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:    ret{{[l|q]}} # 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) {
; X86-LABEL: test_mask_mullo_epi16_rrk_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpmullw %zmm1, %zmm0, %zmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x49,0xd5,0xd1]
; X86-NEXT:    vmovdqa64 %zmm2, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_mullo_epi16_rrk_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpmullw %zmm1, %zmm0, %zmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x49,0xd5,0xd1]
; X64-NEXT:    vmovdqa64 %zmm2, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc2]
; X64-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) {
; X86-LABEL: test_mask_mullo_epi16_rrkz_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpmullw %zmm1, %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xc9,0xd5,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_mullo_epi16_rrkz_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpmullw %zmm1, %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xc9,0xd5,0xc1]
; X64-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) {
; X86-LABEL: test_mask_mullo_epi16_rm_512:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpmullw (%eax), %zmm0, %zmm0 # encoding: [0x62,0xf1,0x7d,0x48,0xd5,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_mullo_epi16_rm_512:
; X64:       # %bb.0:
; X64-NEXT:    vpmullw (%rdi), %zmm0, %zmm0 # encoding: [0x62,0xf1,0x7d,0x48,0xd5,0x07]
; X64-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) {
; X86-LABEL: test_mask_mullo_epi16_rmk_512:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpmullw (%eax), %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x49,0xd5,0x08]
; X86-NEXT:    vmovdqa64 %zmm1, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_mullo_epi16_rmk_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpmullw (%rdi), %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x49,0xd5,0x0f]
; X64-NEXT:    vmovdqa64 %zmm1, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc1]
; X64-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) {
; X86-LABEL: test_mask_mullo_epi16_rmkz_512:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpmullw (%eax), %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xc9,0xd5,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_mullo_epi16_rmkz_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpmullw (%rdi), %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xc9,0xd5,0x07]
; X64-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 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd5,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # 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) {
; X86-LABEL: test_mask_mullo_epi16_rrk_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpmullw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xd5,0xd1]
; X86-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_mullo_epi16_rrk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpmullw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xd5,0xd1]
; X64-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
; X64-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) {
; X86-LABEL: test_mask_mullo_epi16_rrkz_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpmullw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xd5,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_mullo_epi16_rrkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpmullw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xd5,0xc1]
; X64-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) {
; X86-LABEL: test_mask_mullo_epi16_rm_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpmullw (%eax), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd5,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_mullo_epi16_rm_128:
; X64:       # %bb.0:
; X64-NEXT:    vpmullw (%rdi), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd5,0x07]
; X64-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) {
; X86-LABEL: test_mask_mullo_epi16_rmk_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
; X86-NEXT:    kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
; X86-NEXT:    vpmullw (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xd5,0x08]
; X86-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_mullo_epi16_rmk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpmullw (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xd5,0x0f]
; X64-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X64-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) {
; X86-LABEL: test_mask_mullo_epi16_rmkz_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
; X86-NEXT:    kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
; X86-NEXT:    vpmullw (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xd5,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_mullo_epi16_rmkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpmullw (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xd5,0x07]
; X64-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 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd5,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # 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) {
; X86-LABEL: test_mask_mullo_epi16_rrk_256:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpmullw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xd5,0xd1]
; X86-NEXT:    vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_mullo_epi16_rrk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpmullw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xd5,0xd1]
; X64-NEXT:    vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
; X64-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) {
; X86-LABEL: test_mask_mullo_epi16_rrkz_256:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpmullw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xd5,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_mullo_epi16_rrkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpmullw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xd5,0xc1]
; X64-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) {
; X86-LABEL: test_mask_mullo_epi16_rm_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpmullw (%eax), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd5,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_mullo_epi16_rm_256:
; X64:       # %bb.0:
; X64-NEXT:    vpmullw (%rdi), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd5,0x07]
; X64-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) {
; X86-LABEL: test_mask_mullo_epi16_rmk_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpmullw (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xd5,0x08]
; X86-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_mullo_epi16_rmk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpmullw (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xd5,0x0f]
; X64-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X64-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) {
; X86-LABEL: test_mask_mullo_epi16_rmkz_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpmullw (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xd5,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_mullo_epi16_rmkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpmullw (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xd5,0x07]
; X64-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)

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) {
; X86-LABEL: test_int_x86_avx512_mask_pmaxs_b_128:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpmaxsb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x3c,0xd1]
; X86-NEXT:    vpmaxsb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x3c,0xc1]
; X86-NEXT:    vpaddb %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmaxs_b_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpmaxsb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x3c,0xd1]
; X64-NEXT:    vpmaxsb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x3c,0xc1]
; X64-NEXT:    vpaddb %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc0]
; X64-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) {
; X86-LABEL: test_int_x86_avx512_mask_pmaxs_b_256:
; X86:       # %bb.0:
; X86-NEXT:    vpmaxsb %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x3c,0xd9]
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpmaxsb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x3c,0xd1]
; X86-NEXT:    vpaddb %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmaxs_b_256:
; X64:       # %bb.0:
; X64-NEXT:    vpmaxsb %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x3c,0xd9]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpmaxsb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x3c,0xd1]
; X64-NEXT:    vpaddb %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc3]
; X64-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) {
; X86-LABEL: test_int_x86_avx512_mask_pmaxs_w_128:
; X86:       # %bb.0:
; X86-NEXT:    vpmaxsw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xee,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpmaxsw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xee,0xd1]
; X86-NEXT:    vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmaxs_w_128:
; X64:       # %bb.0:
; X64-NEXT:    vpmaxsw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xee,0xd9]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpmaxsw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xee,0xd1]
; X64-NEXT:    vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
; X64-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) {
; X86-LABEL: test_int_x86_avx512_mask_pmaxs_w_256:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpmaxsw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xee,0xd1]
; X86-NEXT:    vpmaxsw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xee,0xc1]
; X86-NEXT:    vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmaxs_w_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpmaxsw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xee,0xd1]
; X64-NEXT:    vpmaxsw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xee,0xc1]
; X64-NEXT:    vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
; X64-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) {
; X86-LABEL: test_int_x86_avx512_mask_pmaxu_b_128:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpmaxub %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xde,0xd1]
; X86-NEXT:    vpmaxub %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xde,0xc1]
; X86-NEXT:    vpaddb %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmaxu_b_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpmaxub %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xde,0xd1]
; X64-NEXT:    vpmaxub %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xde,0xc1]
; X64-NEXT:    vpaddb %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc0]
; X64-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) {
; X86-LABEL: test_int_x86_avx512_mask_pmaxu_b_256:
; X86:       # %bb.0:
; X86-NEXT:    vpmaxub %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xde,0xd9]
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpmaxub %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xde,0xd1]
; X86-NEXT:    vpaddb %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmaxu_b_256:
; X64:       # %bb.0:
; X64-NEXT:    vpmaxub %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xde,0xd9]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpmaxub %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xde,0xd1]
; X64-NEXT:    vpaddb %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc3]
; X64-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) {
; X86-LABEL: test_int_x86_avx512_mask_pmaxu_w_128:
; X86:       # %bb.0:
; X86-NEXT:    vpmaxuw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x3e,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpmaxuw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x3e,0xd1]
; X86-NEXT:    vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmaxu_w_128:
; X64:       # %bb.0:
; X64-NEXT:    vpmaxuw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x3e,0xd9]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpmaxuw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x3e,0xd1]
; X64-NEXT:    vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
; X64-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) {
; X86-LABEL: test_int_x86_avx512_mask_pmaxu_w_256:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpmaxuw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x3e,0xd1]
; X86-NEXT:    vpmaxuw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x3e,0xc1]
; X86-NEXT:    vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmaxu_w_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpmaxuw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x3e,0xd1]
; X64-NEXT:    vpmaxuw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x3e,0xc1]
; X64-NEXT:    vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
; X64-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) {
; X86-LABEL: test_int_x86_avx512_mask_pmins_b_128:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpminsb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x38,0xd1]
; X86-NEXT:    vpminsb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x38,0xc1]
; X86-NEXT:    vpaddb %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmins_b_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpminsb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x38,0xd1]
; X64-NEXT:    vpminsb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x38,0xc1]
; X64-NEXT:    vpaddb %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc0]
; X64-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) {
; X86-LABEL: test_int_x86_avx512_mask_pmins_b_256:
; X86:       # %bb.0:
; X86-NEXT:    vpminsb %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x38,0xd9]
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpminsb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x38,0xd1]
; X86-NEXT:    vpaddb %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmins_b_256:
; X64:       # %bb.0:
; X64-NEXT:    vpminsb %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x38,0xd9]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpminsb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x38,0xd1]
; X64-NEXT:    vpaddb %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc3]
; X64-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) {
; X86-LABEL: test_int_x86_avx512_mask_pmins_w_128:
; X86:       # %bb.0:
; X86-NEXT:    vpminsw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xea,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpminsw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xea,0xd1]
; X86-NEXT:    vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmins_w_128:
; X64:       # %bb.0:
; X64-NEXT:    vpminsw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xea,0xd9]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpminsw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xea,0xd1]
; X64-NEXT:    vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
; X64-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) {
; X86-LABEL: test_int_x86_avx512_mask_pmins_w_256:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpminsw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xea,0xd1]
; X86-NEXT:    vpminsw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xea,0xc1]
; X86-NEXT:    vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmins_w_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpminsw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xea,0xd1]
; X64-NEXT:    vpminsw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xea,0xc1]
; X64-NEXT:    vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
; X64-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) {
; X86-LABEL: test_int_x86_avx512_mask_pminu_b_128:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpminub %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xda,0xd1]
; X86-NEXT:    vpminub %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xda,0xc1]
; X86-NEXT:    vpaddb %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pminu_b_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpminub %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xda,0xd1]
; X64-NEXT:    vpminub %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xda,0xc1]
; X64-NEXT:    vpaddb %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc0]
; X64-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) {
; X86-LABEL: test_int_x86_avx512_mask_pminu_b_256:
; X86:       # %bb.0:
; X86-NEXT:    vpminub %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xda,0xd9]
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpminub %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xda,0xd1]
; X86-NEXT:    vpaddb %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pminu_b_256:
; X64:       # %bb.0:
; X64-NEXT:    vpminub %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xda,0xd9]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpminub %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xda,0xd1]
; X64-NEXT:    vpaddb %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc3]
; X64-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) {
; X86-LABEL: test_int_x86_avx512_mask_pminu_w_128:
; X86:       # %bb.0:
; X86-NEXT:    vpminuw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x3a,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpminuw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x3a,0xd1]
; X86-NEXT:    vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pminu_w_128:
; X64:       # %bb.0:
; X64-NEXT:    vpminuw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x3a,0xd9]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpminuw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x3a,0xd1]
; X64-NEXT:    vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
; X64-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) {
; X86-LABEL: test_int_x86_avx512_mask_pminu_w_256:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpminuw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x3a,0xd1]
; X86-NEXT:    vpminuw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x3a,0xc1]
; X86-NEXT:    vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pminu_w_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpminuw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x3a,0xd1]
; X64-NEXT:    vpminuw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x3a,0xc1]
; X64-NEXT:    vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
; X64-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.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) {
; X86-LABEL: test_int_x86_avx512_mask_psrl_w_128:
; X86:       # %bb.0:
; X86-NEXT:    vpsrlw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd1,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpsrlw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xd1,0xd1]
; X86-NEXT:    vpsrlw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xd1,0xc1]
; X86-NEXT:    vpaddw %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xfd,0xc0]
; X86-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_psrl_w_128:
; X64:       # %bb.0:
; X64-NEXT:    vpsrlw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd1,0xd9]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpsrlw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xd1,0xd1]
; X64-NEXT:    vpsrlw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xd1,0xc1]
; X64-NEXT:    vpaddw %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xfd,0xc0]
; X64-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
; X64-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) {
; X86-LABEL: test_int_x86_avx512_mask_psrl_w_256:
; X86:       # %bb.0:
; X86-NEXT:    vpsrlw %xmm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd1,0xd9]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpsrlw %xmm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xd1,0xd1]
; X86-NEXT:    vpsrlw %xmm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xd1,0xc1]
; X86-NEXT:    vpaddw %ymm0, %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0xfd,0xc0]
; X86-NEXT:    vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_psrl_w_256:
; X64:       # %bb.0:
; X64-NEXT:    vpsrlw %xmm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd1,0xd9]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpsrlw %xmm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xd1,0xd1]
; X64-NEXT:    vpsrlw %xmm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xd1,0xc1]
; X64-NEXT:    vpaddw %ymm0, %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0xfd,0xc0]
; X64-NEXT:    vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
; X64-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.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) {
; X86-LABEL: test_int_x86_avx512_mask_psra_w_128:
; X86:       # %bb.0:
; X86-NEXT:    vpsraw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xe1,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpsraw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe1,0xd1]
; X86-NEXT:    vpsraw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xe1,0xc1]
; X86-NEXT:    vpaddw %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc3]
; X86-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_psra_w_128:
; X64:       # %bb.0:
; X64-NEXT:    vpsraw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xe1,0xd9]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpsraw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe1,0xd1]
; X64-NEXT:    vpsraw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xe1,0xc1]
; X64-NEXT:    vpaddw %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc3]
; X64-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
; X64-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 <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) {
; X86-LABEL: test_int_x86_avx512_mask_psra_w_256:
; X86:       # %bb.0:
; X86-NEXT:    vpsraw %xmm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xe1,0xd9]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpsraw %xmm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe1,0xd1]
; X86-NEXT:    vpsraw %xmm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xe1,0xc1]
; X86-NEXT:    vpaddw %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc3]
; X86-NEXT:    vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_psra_w_256:
; X64:       # %bb.0:
; X64-NEXT:    vpsraw %xmm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xe1,0xd9]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpsraw %xmm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe1,0xd1]
; X64-NEXT:    vpsraw %xmm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xe1,0xc1]
; X64-NEXT:    vpaddw %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc3]
; X64-NEXT:    vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
; X64-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 <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) {
; X86-LABEL: test_int_x86_avx512_mask_psll_w_128:
; X86:       # %bb.0:
; X86-NEXT:    vpsllw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xf1,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpsllw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xf1,0xd1]
; X86-NEXT:    vpsllw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xf1,0xc1]
; X86-NEXT:    vpaddw %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc3]
; X86-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_psll_w_128:
; X64:       # %bb.0:
; X64-NEXT:    vpsllw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xf1,0xd9]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpsllw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xf1,0xd1]
; X64-NEXT:    vpsllw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xf1,0xc1]
; X64-NEXT:    vpaddw %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc3]
; X64-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
; X64-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) {
; X86-LABEL: test_int_x86_avx512_mask_psll_w_256:
; X86:       # %bb.0:
; X86-NEXT:    vpsllw %xmm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xf1,0xd9]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpsllw %xmm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xf1,0xd1]
; X86-NEXT:    vpsllw %xmm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xf1,0xc1]
; X86-NEXT:    vpaddw %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc3]
; X86-NEXT:    vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_psll_w_256:
; X64:       # %bb.0:
; X64-NEXT:    vpsllw %xmm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xf1,0xd9]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpsllw %xmm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xf1,0xd1]
; X64-NEXT:    vpsllw %xmm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xf1,0xc1]
; X64-NEXT:    vpaddw %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc3]
; X64-NEXT:    vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
; X64-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.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) {
; X86-LABEL: test_int_x86_avx512_mask_psrl_wi_128:
; X86:       # %bb.0:
; X86-NEXT:    vpsrlw $3, %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x71,0xd0,0x03]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpsrlw $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x71,0xd0,0x03]
; X86-NEXT:    vpsrlw $3, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x71,0xd0,0x03]
; X86-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
; X86-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_psrl_wi_128:
; X64:       # %bb.0:
; X64-NEXT:    vpsrlw $3, %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x71,0xd0,0x03]
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpsrlw $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x71,0xd0,0x03]
; X64-NEXT:    vpsrlw $3, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x71,0xd0,0x03]
; X64-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
; X64-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
; X64-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) {
; X86-LABEL: test_int_x86_avx512_mask_psrl_wi_256:
; X86:       # %bb.0:
; X86-NEXT:    vpsrlw $3, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xed,0x71,0xd0,0x03]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpsrlw $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x71,0xd0,0x03]
; X86-NEXT:    vpsrlw $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x71,0xd0,0x03]
; X86-NEXT:    vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
; X86-NEXT:    vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_psrl_wi_256:
; X64:       # %bb.0:
; X64-NEXT:    vpsrlw $3, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xed,0x71,0xd0,0x03]
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpsrlw $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x71,0xd0,0x03]
; X64-NEXT:    vpsrlw $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x71,0xd0,0x03]
; X64-NEXT:    vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
; X64-NEXT:    vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
; X64-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 <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) {
; X86-LABEL: test_int_x86_avx512_mask_psra_wi_128:
; X86:       # %bb.0:
; X86-NEXT:    vpsraw $3, %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x71,0xe0,0x03]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpsraw $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x71,0xe0,0x03]
; X86-NEXT:    vpsraw $3, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x71,0xe0,0x03]
; X86-NEXT:    vpaddw %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc2]
; X86-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_psra_wi_128:
; X64:       # %bb.0:
; X64-NEXT:    vpsraw $3, %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x71,0xe0,0x03]
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpsraw $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x71,0xe0,0x03]
; X64-NEXT:    vpsraw $3, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x71,0xe0,0x03]
; X64-NEXT:    vpaddw %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc2]
; X64-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
; X64-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.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) {
; X86-LABEL: test_int_x86_avx512_mask_psra_wi_256:
; X86:       # %bb.0:
; X86-NEXT:    vpsraw $3, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xed,0x71,0xe0,0x03]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpsraw $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x71,0xe0,0x03]
; X86-NEXT:    vpsraw $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x71,0xe0,0x03]
; X86-NEXT:    vpaddw %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc2]
; X86-NEXT:    vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_psra_wi_256:
; X64:       # %bb.0:
; X64-NEXT:    vpsraw $3, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xed,0x71,0xe0,0x03]
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpsraw $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x71,0xe0,0x03]
; X64-NEXT:    vpsraw $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x71,0xe0,0x03]
; X64-NEXT:    vpaddw %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc2]
; X64-NEXT:    vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
; X64-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 <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) {
; X86-LABEL: test_int_x86_avx512_mask_psll_wi_128:
; X86:       # %bb.0:
; X86-NEXT:    vpsllw $3, %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x71,0xf0,0x03]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpsllw $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x71,0xf0,0x03]
; X86-NEXT:    vpsllw $3, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x71,0xf0,0x03]
; X86-NEXT:    vpaddw %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc2]
; X86-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_psll_wi_128:
; X64:       # %bb.0:
; X64-NEXT:    vpsllw $3, %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x71,0xf0,0x03]
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpsllw $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x71,0xf0,0x03]
; X64-NEXT:    vpsllw $3, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x71,0xf0,0x03]
; X64-NEXT:    vpaddw %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc2]
; X64-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
; X64-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) {
; X86-LABEL: test_int_x86_avx512_mask_psll_wi_256:
; X86:       # %bb.0:
; X86-NEXT:    vpsllw $3, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xed,0x71,0xf0,0x03]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpsllw $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x71,0xf0,0x03]
; X86-NEXT:    vpsllw $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x71,0xf0,0x03]
; X86-NEXT:    vpaddw %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc2]
; X86-NEXT:    vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_psll_wi_256:
; X64:       # %bb.0:
; X64-NEXT:    vpsllw $3, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xed,0x71,0xf0,0x03]
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpsllw $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x71,0xf0,0x03]
; X64-NEXT:    vpsllw $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x71,0xf0,0x03]
; X64-NEXT:    vpaddw %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc2]
; X64-NEXT:    vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
; X64-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 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) {
; X86-LABEL: test_int_x86_avx512_mask_pshuf_b_128:
; X86:       # %bb.0:
; X86-NEXT:    vpshufb %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x00,0xd9]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpshufb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x00,0xd1]
; X86-NEXT:    vpaddb %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pshuf_b_128:
; X64:       # %bb.0:
; X64-NEXT:    vpshufb %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x00,0xd9]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpshufb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x00,0xd1]
; X64-NEXT:    vpaddb %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc3]
; X64-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) {
; X86-LABEL: test_int_x86_avx512_mask_pshuf_b_256:
; X86:       # %bb.0:
; X86-NEXT:    vpshufb %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x00,0xd9]
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpshufb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x00,0xd1]
; X86-NEXT:    vpaddb %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pshuf_b_256:
; X64:       # %bb.0:
; X64-NEXT:    vpshufb %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x00,0xd9]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpshufb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x00,0xd1]
; X64-NEXT:    vpaddb %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc3]
; X64-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 <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) {
; X86-LABEL: test_int_x86_avx512_mask_pmovzxb_w_128:
; X86:       # %bb.0:
; X86-NEXT:    vpmovzxbw %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x30,0xd0]
; X86-NEXT:    # xmm2 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpmovzxbw %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x30,0xc8]
; X86-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
; X86-NEXT:    vpmovzxbw %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x30,0xc0]
; X86-NEXT:    # xmm0 {%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
; X86-NEXT:    vpaddw %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc2]
; X86-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmovzxb_w_128:
; X64:       # %bb.0:
; X64-NEXT:    vpmovzxbw %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x30,0xd0]
; X64-NEXT:    # xmm2 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpmovzxbw %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x30,0xc8]
; X64-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
; X64-NEXT:    vpmovzxbw %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x30,0xc0]
; X64-NEXT:    # xmm0 {%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
; X64-NEXT:    vpaddw %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc2]
; X64-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
; X64-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) {
; X86-LABEL: test_int_x86_avx512_mask_pmovzxb_w_256:
; X86:       # %bb.0:
; X86-NEXT:    vpmovzxbw %xmm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x30,0xd0]
; X86-NEXT:    # ymm2 = 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
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpmovzxbw %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x30,0xc8]
; X86-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
; X86-NEXT:    vpmovzxbw %xmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x30,0xc0]
; X86-NEXT:    # ymm0 {%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
; X86-NEXT:    vpaddw %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc2]
; X86-NEXT:    vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmovzxb_w_256:
; X64:       # %bb.0:
; X64-NEXT:    vpmovzxbw %xmm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x30,0xd0]
; X64-NEXT:    # ymm2 = 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
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpmovzxbw %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x30,0xc8]
; X64-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
; X64-NEXT:    vpmovzxbw %xmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x30,0xc0]
; X64-NEXT:    # ymm0 {%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
; X64-NEXT:    vpaddw %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc2]
; X64-NEXT:    vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
; X64-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) {
; X86-LABEL: test_int_x86_avx512_mask_pmovsxb_w_128:
; X86:       # %bb.0:
; X86-NEXT:    vpmovsxbw %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x20,0xd0]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpmovsxbw %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x20,0xc8]
; X86-NEXT:    vpmovsxbw %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x20,0xc0]
; X86-NEXT:    vpaddw %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc2]
; X86-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmovsxb_w_128:
; X64:       # %bb.0:
; X64-NEXT:    vpmovsxbw %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x20,0xd0]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpmovsxbw %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x20,0xc8]
; X64-NEXT:    vpmovsxbw %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x20,0xc0]
; X64-NEXT:    vpaddw %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc2]
; X64-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
; X64-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) {
; X86-LABEL: test_int_x86_avx512_mask_pmovsxb_w_256:
; X86:       # %bb.0:
; X86-NEXT:    vpmovsxbw %xmm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x20,0xd0]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpmovsxbw %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x20,0xc8]
; X86-NEXT:    vpmovsxbw %xmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x20,0xc0]
; X86-NEXT:    vpaddw %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc2]
; X86-NEXT:    vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmovsxb_w_256:
; X64:       # %bb.0:
; X64-NEXT:    vpmovsxbw %xmm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x20,0xd0]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpmovsxbw %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x20,0xc8]
; X64-NEXT:    vpmovsxbw %xmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x20,0xc0]
; X64-NEXT:    vpaddw %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc2]
; X64-NEXT:    vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
; X64-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) {
; X86-LABEL: test_int_x86_avx512_mask_pmovsxd_q_128:
; X86:       # %bb.0:
; X86-NEXT:    vpmovsxdq %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x25,0xd0]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpmovsxdq %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x25,0xc8]
; X86-NEXT:    vpmovsxdq %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x25,0xc0]
; X86-NEXT:    vpaddq %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd4,0xc2]
; X86-NEXT:    vpaddq %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xd4,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmovsxd_q_128:
; X64:       # %bb.0:
; X64-NEXT:    vpmovsxdq %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x25,0xd0]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpmovsxdq %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x25,0xc8]
; X64-NEXT:    vpmovsxdq %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x25,0xc0]
; X64-NEXT:    vpaddq %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd4,0xc2]
; X64-NEXT:    vpaddq %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xd4,0xc0]
; X64-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) {
; X86-LABEL: test_int_x86_avx512_mask_pmovsxd_q_256:
; X86:       # %bb.0:
; X86-NEXT:    vpmovsxdq %xmm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x25,0xd0]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpmovsxdq %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x25,0xc8]
; X86-NEXT:    vpmovsxdq %xmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x25,0xc0]
; X86-NEXT:    vpaddq %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd4,0xc2]
; X86-NEXT:    vpaddq %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xd4,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmovsxd_q_256:
; X64:       # %bb.0:
; X64-NEXT:    vpmovsxdq %xmm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x25,0xd0]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpmovsxdq %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x25,0xc8]
; X64-NEXT:    vpmovsxdq %xmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x25,0xc0]
; X64-NEXT:    vpaddq %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd4,0xc2]
; X64-NEXT:    vpaddq %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xd4,0xc0]
; X64-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 <16 x i8> @llvm.x86.avx512.cvtmask2b.128(i16)

define <16 x i8>@test_int_x86_avx512_cvtmask2b_128(i16 %x0) {
; X86-LABEL: test_int_x86_avx512_cvtmask2b_128:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k0 # encoding: [0xc5,0xf8,0x90,0x44,0x24,0x04]
; X86-NEXT:    vpmovm2b %k0, %xmm0 # encoding: [0x62,0xf2,0x7e,0x08,0x28,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_cvtmask2b_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k0 # encoding: [0xc5,0xfb,0x92,0xc7]
; X64-NEXT:    vpmovm2b %k0, %xmm0 # encoding: [0x62,0xf2,0x7e,0x08,0x28,0xc0]
; X64-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) {
; X86-LABEL: test_int_x86_avx512_cvtmask2b_256:
; X86:       # %bb.0:
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k0 # encoding: [0xc4,0xe1,0xf9,0x90,0x44,0x24,0x04]
; X86-NEXT:    vpmovm2b %k0, %ymm0 # encoding: [0x62,0xf2,0x7e,0x28,0x28,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_cvtmask2b_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k0 # encoding: [0xc5,0xfb,0x92,0xc7]
; X64-NEXT:    vpmovm2b %k0, %ymm0 # encoding: [0x62,0xf2,0x7e,0x28,0x28,0xc0]
; X64-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) {
; X86-LABEL: test_int_x86_avx512_cvtmask2w_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k0 # encoding: [0xc5,0xfb,0x92,0xc0]
; X86-NEXT:    vpmovm2w %k0, %xmm0 # encoding: [0x62,0xf2,0xfe,0x08,0x28,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_cvtmask2w_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k0 # encoding: [0xc5,0xfb,0x92,0xc7]
; X64-NEXT:    vpmovm2w %k0, %xmm0 # encoding: [0x62,0xf2,0xfe,0x08,0x28,0xc0]
; X64-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) {
; X86-LABEL: test_int_x86_avx512_cvtmask2w_256:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k0 # encoding: [0xc5,0xf8,0x90,0x44,0x24,0x04]
; X86-NEXT:    vpmovm2w %k0, %ymm0 # encoding: [0x62,0xf2,0xfe,0x28,0x28,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_cvtmask2w_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k0 # encoding: [0xc5,0xfb,0x92,0xc7]
; X64-NEXT:    vpmovm2w %k0, %ymm0 # encoding: [0x62,0xf2,0xfe,0x28,0x28,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <16 x i16> @llvm.x86.avx512.cvtmask2w.256(i16 %x0)
  ret <16 x i16> %res
}
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 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6b,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # 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) {
; X86-LABEL: test_mask_packs_epi32_rrk_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpackssdw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x6b,0xd1]
; X86-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_packs_epi32_rrk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpackssdw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x6b,0xd1]
; X64-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
; X64-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) {
; X86-LABEL: test_mask_packs_epi32_rrkz_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpackssdw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x6b,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_packs_epi32_rrkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpackssdw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x6b,0xc1]
; X64-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) {
; X86-LABEL: test_mask_packs_epi32_rm_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpackssdw (%eax), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6b,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_packs_epi32_rm_128:
; X64:       # %bb.0:
; X64-NEXT:    vpackssdw (%rdi), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6b,0x07]
; X64-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) {
; X86-LABEL: test_mask_packs_epi32_rmk_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
; X86-NEXT:    kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
; X86-NEXT:    vpackssdw (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x6b,0x08]
; X86-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_packs_epi32_rmk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpackssdw (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x6b,0x0f]
; X64-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X64-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) {
; X86-LABEL: test_mask_packs_epi32_rmkz_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
; X86-NEXT:    kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
; X86-NEXT:    vpackssdw (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x6b,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_packs_epi32_rmkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpackssdw (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x6b,0x07]
; X64-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) {
; X86-LABEL: test_mask_packs_epi32_rmb_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpackssdw (%eax){1to4}, %xmm0, %xmm0 # encoding: [0x62,0xf1,0x7d,0x18,0x6b,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_packs_epi32_rmb_128:
; X64:       # %bb.0:
; X64-NEXT:    vpackssdw (%rdi){1to4}, %xmm0, %xmm0 # encoding: [0x62,0xf1,0x7d,0x18,0x6b,0x07]
; X64-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) {
; X86-LABEL: test_mask_packs_epi32_rmbk_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
; X86-NEXT:    kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
; X86-NEXT:    vpackssdw (%eax){1to4}, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x19,0x6b,0x08]
; X86-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_packs_epi32_rmbk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpackssdw (%rdi){1to4}, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x19,0x6b,0x0f]
; X64-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X64-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) {
; X86-LABEL: test_mask_packs_epi32_rmbkz_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
; X86-NEXT:    kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
; X86-NEXT:    vpackssdw (%eax){1to4}, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x99,0x6b,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_packs_epi32_rmbkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpackssdw (%rdi){1to4}, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x99,0x6b,0x07]
; X64-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 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6b,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # 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) {
; X86-LABEL: test_mask_packs_epi32_rrk_256:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpackssdw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x6b,0xd1]
; X86-NEXT:    vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_packs_epi32_rrk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpackssdw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x6b,0xd1]
; X64-NEXT:    vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
; X64-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) {
; X86-LABEL: test_mask_packs_epi32_rrkz_256:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpackssdw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x6b,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_packs_epi32_rrkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpackssdw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x6b,0xc1]
; X64-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) {
; X86-LABEL: test_mask_packs_epi32_rm_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpackssdw (%eax), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6b,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_packs_epi32_rm_256:
; X64:       # %bb.0:
; X64-NEXT:    vpackssdw (%rdi), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6b,0x07]
; X64-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) {
; X86-LABEL: test_mask_packs_epi32_rmk_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpackssdw (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x6b,0x08]
; X86-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_packs_epi32_rmk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpackssdw (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x6b,0x0f]
; X64-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X64-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) {
; X86-LABEL: test_mask_packs_epi32_rmkz_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpackssdw (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x6b,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_packs_epi32_rmkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpackssdw (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x6b,0x07]
; X64-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) {
; X86-LABEL: test_mask_packs_epi32_rmb_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpackssdw (%eax){1to8}, %ymm0, %ymm0 # encoding: [0x62,0xf1,0x7d,0x38,0x6b,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_packs_epi32_rmb_256:
; X64:       # %bb.0:
; X64-NEXT:    vpackssdw (%rdi){1to8}, %ymm0, %ymm0 # encoding: [0x62,0xf1,0x7d,0x38,0x6b,0x07]
; X64-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) {
; X86-LABEL: test_mask_packs_epi32_rmbk_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpackssdw (%eax){1to8}, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x39,0x6b,0x08]
; X86-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_packs_epi32_rmbk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpackssdw (%rdi){1to8}, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x39,0x6b,0x0f]
; X64-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X64-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) {
; X86-LABEL: test_mask_packs_epi32_rmbkz_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpackssdw (%eax){1to8}, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xb9,0x6b,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_packs_epi32_rmbkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpackssdw (%rdi){1to8}, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xb9,0x6b,0x07]
; X64-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 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x63,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # 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) {
; X86-LABEL: test_mask_packs_epi16_rrk_128:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpacksswb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x63,0xd1]
; X86-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_packs_epi16_rrk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpacksswb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x63,0xd1]
; X64-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
; X64-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) {
; X86-LABEL: test_mask_packs_epi16_rrkz_128:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x63,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_packs_epi16_rrkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x63,0xc1]
; X64-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) {
; X86-LABEL: test_mask_packs_epi16_rm_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpacksswb (%eax), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x63,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_packs_epi16_rm_128:
; X64:       # %bb.0:
; X64-NEXT:    vpacksswb (%rdi), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x63,0x07]
; X64-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) {
; X86-LABEL: test_mask_packs_epi16_rmk_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpacksswb (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x63,0x08]
; X86-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_packs_epi16_rmk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpacksswb (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x63,0x0f]
; X64-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X64-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) {
; X86-LABEL: test_mask_packs_epi16_rmkz_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpacksswb (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x63,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_packs_epi16_rmkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpacksswb (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x63,0x07]
; X64-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 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x63,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # 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) {
; X86-LABEL: test_mask_packs_epi16_rrk_256:
; X86:       # %bb.0:
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpacksswb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x63,0xd1]
; X86-NEXT:    vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_packs_epi16_rrk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpacksswb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x63,0xd1]
; X64-NEXT:    vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
; X64-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) {
; X86-LABEL: test_mask_packs_epi16_rrkz_256:
; X86:       # %bb.0:
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpacksswb %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x63,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_packs_epi16_rrkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpacksswb %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x63,0xc1]
; X64-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) {
; X86-LABEL: test_mask_packs_epi16_rm_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpacksswb (%eax), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x63,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_packs_epi16_rm_256:
; X64:       # %bb.0:
; X64-NEXT:    vpacksswb (%rdi), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x63,0x07]
; X64-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) {
; X86-LABEL: test_mask_packs_epi16_rmk_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpacksswb (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x63,0x08]
; X86-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_packs_epi16_rmk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpacksswb (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x63,0x0f]
; X64-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X64-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) {
; X86-LABEL: test_mask_packs_epi16_rmkz_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpacksswb (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x63,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_packs_epi16_rmkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpacksswb (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x63,0x07]
; X64-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 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x2b,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # 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) {
; X86-LABEL: test_mask_packus_epi32_rrk_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpackusdw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x2b,0xd1]
; X86-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_packus_epi32_rrk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpackusdw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x2b,0xd1]
; X64-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
; X64-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) {
; X86-LABEL: test_mask_packus_epi32_rrkz_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x2b,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_packus_epi32_rrkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x2b,0xc1]
; X64-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) {
; X86-LABEL: test_mask_packus_epi32_rm_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpackusdw (%eax), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x2b,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_packus_epi32_rm_128:
; X64:       # %bb.0:
; X64-NEXT:    vpackusdw (%rdi), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x2b,0x07]
; X64-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) {
; X86-LABEL: test_mask_packus_epi32_rmk_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
; X86-NEXT:    kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
; X86-NEXT:    vpackusdw (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x2b,0x08]
; X86-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_packus_epi32_rmk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpackusdw (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x2b,0x0f]
; X64-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X64-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) {
; X86-LABEL: test_mask_packus_epi32_rmkz_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
; X86-NEXT:    kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
; X86-NEXT:    vpackusdw (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x2b,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_packus_epi32_rmkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpackusdw (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x2b,0x07]
; X64-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) {
; X86-LABEL: test_mask_packus_epi32_rmb_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpackusdw (%eax){1to4}, %xmm0, %xmm0 # encoding: [0x62,0xf2,0x7d,0x18,0x2b,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_packus_epi32_rmb_128:
; X64:       # %bb.0:
; X64-NEXT:    vpackusdw (%rdi){1to4}, %xmm0, %xmm0 # encoding: [0x62,0xf2,0x7d,0x18,0x2b,0x07]
; X64-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) {
; X86-LABEL: test_mask_packus_epi32_rmbk_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
; X86-NEXT:    kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
; X86-NEXT:    vpackusdw (%eax){1to4}, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x19,0x2b,0x08]
; X86-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_packus_epi32_rmbk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpackusdw (%rdi){1to4}, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x19,0x2b,0x0f]
; X64-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X64-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) {
; X86-LABEL: test_mask_packus_epi32_rmbkz_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
; X86-NEXT:    kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
; X86-NEXT:    vpackusdw (%eax){1to4}, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x99,0x2b,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_packus_epi32_rmbkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpackusdw (%rdi){1to4}, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x99,0x2b,0x07]
; X64-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 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x2b,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # 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) {
; X86-LABEL: test_mask_packus_epi32_rrk_256:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpackusdw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x2b,0xd1]
; X86-NEXT:    vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_packus_epi32_rrk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpackusdw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x2b,0xd1]
; X64-NEXT:    vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
; X64-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) {
; X86-LABEL: test_mask_packus_epi32_rrkz_256:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpackusdw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x2b,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_packus_epi32_rrkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpackusdw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x2b,0xc1]
; X64-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) {
; X86-LABEL: test_mask_packus_epi32_rm_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpackusdw (%eax), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x2b,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_packus_epi32_rm_256:
; X64:       # %bb.0:
; X64-NEXT:    vpackusdw (%rdi), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x2b,0x07]
; X64-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) {
; X86-LABEL: test_mask_packus_epi32_rmk_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpackusdw (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x2b,0x08]
; X86-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_packus_epi32_rmk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpackusdw (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x2b,0x0f]
; X64-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X64-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) {
; X86-LABEL: test_mask_packus_epi32_rmkz_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpackusdw (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x2b,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_packus_epi32_rmkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpackusdw (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x2b,0x07]
; X64-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) {
; X86-LABEL: test_mask_packus_epi32_rmb_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpackusdw (%eax){1to8}, %ymm0, %ymm0 # encoding: [0x62,0xf2,0x7d,0x38,0x2b,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_packus_epi32_rmb_256:
; X64:       # %bb.0:
; X64-NEXT:    vpackusdw (%rdi){1to8}, %ymm0, %ymm0 # encoding: [0x62,0xf2,0x7d,0x38,0x2b,0x07]
; X64-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) {
; X86-LABEL: test_mask_packus_epi32_rmbk_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpackusdw (%eax){1to8}, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x39,0x2b,0x08]
; X86-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_packus_epi32_rmbk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpackusdw (%rdi){1to8}, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x39,0x2b,0x0f]
; X64-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X64-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) {
; X86-LABEL: test_mask_packus_epi32_rmbkz_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpackusdw (%eax){1to8}, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xb9,0x2b,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_packus_epi32_rmbkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpackusdw (%rdi){1to8}, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xb9,0x2b,0x07]
; X64-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 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x67,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # 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) {
; X86-LABEL: test_mask_packus_epi16_rrk_128:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpackuswb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x67,0xd1]
; X86-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_packus_epi16_rrk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpackuswb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x67,0xd1]
; X64-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
; X64-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) {
; X86-LABEL: test_mask_packus_epi16_rrkz_128:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x67,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_packus_epi16_rrkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x67,0xc1]
; X64-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) {
; X86-LABEL: test_mask_packus_epi16_rm_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpackuswb (%eax), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x67,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_packus_epi16_rm_128:
; X64:       # %bb.0:
; X64-NEXT:    vpackuswb (%rdi), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x67,0x07]
; X64-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) {
; X86-LABEL: test_mask_packus_epi16_rmk_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpackuswb (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x67,0x08]
; X86-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_packus_epi16_rmk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpackuswb (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x67,0x0f]
; X64-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X64-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) {
; X86-LABEL: test_mask_packus_epi16_rmkz_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpackuswb (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x67,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_packus_epi16_rmkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpackuswb (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x67,0x07]
; X64-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 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x67,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # 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) {
; X86-LABEL: test_mask_packus_epi16_rrk_256:
; X86:       # %bb.0:
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpackuswb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x67,0xd1]
; X86-NEXT:    vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_packus_epi16_rrk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpackuswb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x67,0xd1]
; X64-NEXT:    vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
; X64-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) {
; X86-LABEL: test_mask_packus_epi16_rrkz_256:
; X86:       # %bb.0:
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpackuswb %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x67,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_packus_epi16_rrkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpackuswb %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x67,0xc1]
; X64-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) {
; X86-LABEL: test_mask_packus_epi16_rm_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpackuswb (%eax), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x67,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_packus_epi16_rm_256:
; X64:       # %bb.0:
; X64-NEXT:    vpackuswb (%rdi), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x67,0x07]
; X64-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) {
; X86-LABEL: test_mask_packus_epi16_rmk_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpackuswb (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x67,0x08]
; X86-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_packus_epi16_rmk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpackuswb (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x67,0x0f]
; X64-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X64-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) {
; X86-LABEL: test_mask_packus_epi16_rmkz_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpackuswb (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x67,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_packus_epi16_rmkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpackuswb (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x67,0x07]
; X64-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 i32> @test_cmp_b_256(<32 x i8> %a0, <32 x i8> %a1) {
; X86-LABEL: test_cmp_b_256:
; X86:       # %bb.0:
; X86-NEXT:    pushl %ebx # encoding: [0x53]
; X86-NEXT:    .cfi_def_cfa_offset 8
; X86-NEXT:    pushl %edi # encoding: [0x57]
; X86-NEXT:    .cfi_def_cfa_offset 12
; X86-NEXT:    pushl %esi # encoding: [0x56]
; X86-NEXT:    .cfi_def_cfa_offset 16
; X86-NEXT:    .cfi_offset %esi, -16
; X86-NEXT:    .cfi_offset %edi, -12
; X86-NEXT:    .cfi_offset %ebx, -8
; X86-NEXT:    vpcmpeqb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x74,0xc1]
; X86-NEXT:    kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
; X86-NEXT:    vpcmpgtb %ymm0, %ymm1, %k0 # encoding: [0x62,0xf1,0x75,0x28,0x64,0xc0]
; X86-NEXT:    kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
; X86-NEXT:    vpcmpleb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x3f,0xc1,0x02]
; X86-NEXT:    kmovd %k0, %edx # encoding: [0xc5,0xfb,0x93,0xd0]
; X86-NEXT:    vpcmpneqb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x3f,0xc1,0x04]
; X86-NEXT:    kmovd %k0, %esi # encoding: [0xc5,0xfb,0x93,0xf0]
; X86-NEXT:    vpcmpnltb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x3f,0xc1,0x05]
; X86-NEXT:    kmovd %k0, %edi # encoding: [0xc5,0xfb,0x93,0xf8]
; X86-NEXT:    vpcmpgtb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x64,0xc1]
; X86-NEXT:    kmovd %k0, %ebx # encoding: [0xc5,0xfb,0x93,0xd8]
; X86-NEXT:    vmovd %esi, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc6]
; X86-NEXT:    vpinsrd $1, %edi, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc7,0x01]
; X86-NEXT:    vpinsrd $2, %ebx, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc3,0x02]
; X86-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1 # encoding: [0xc5,0xf1,0x76,0xc9]
; X86-NEXT:    vpblendd $8, %xmm1, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x02,0xc1,0x08]
; X86-NEXT:    # xmm0 = xmm0[0,1,2],xmm1[3]
; X86-NEXT:    vmovd %ecx, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc9]
; X86-NEXT:    vmovd %eax, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xd0]
; X86-NEXT:    vpunpckldq %xmm1, %xmm2, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x62,0xc9]
; X86-NEXT:    # xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; X86-NEXT:    vmovd %edx, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xd2]
; X86-NEXT:    vpunpcklqdq %xmm2, %xmm1, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0x6c,0xca]
; X86-NEXT:    # xmm1 = xmm1[0],xmm2[0]
; X86-NEXT:    vinserti128 $1, %xmm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x75,0x38,0xc0,0x01]
; X86-NEXT:    popl %esi # encoding: [0x5e]
; X86-NEXT:    .cfi_def_cfa_offset 12
; X86-NEXT:    popl %edi # encoding: [0x5f]
; X86-NEXT:    .cfi_def_cfa_offset 8
; X86-NEXT:    popl %ebx # encoding: [0x5b]
; X86-NEXT:    .cfi_def_cfa_offset 4
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_cmp_b_256:
; X64:       # %bb.0:
; X64-NEXT:    vpcmpeqb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x74,0xc1]
; X64-NEXT:    kmovd %k0, %r8d # encoding: [0xc5,0x7b,0x93,0xc0]
; X64-NEXT:    vpcmpgtb %ymm0, %ymm1, %k0 # encoding: [0x62,0xf1,0x75,0x28,0x64,0xc0]
; X64-NEXT:    kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
; X64-NEXT:    vpcmpleb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x3f,0xc1,0x02]
; X64-NEXT:    kmovd %k0, %edx # encoding: [0xc5,0xfb,0x93,0xd0]
; X64-NEXT:    vpcmpneqb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x3f,0xc1,0x04]
; X64-NEXT:    kmovd %k0, %esi # encoding: [0xc5,0xfb,0x93,0xf0]
; X64-NEXT:    vpcmpnltb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x3f,0xc1,0x05]
; X64-NEXT:    kmovd %k0, %edi # encoding: [0xc5,0xfb,0x93,0xf8]
; X64-NEXT:    vpcmpgtb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x64,0xc1]
; X64-NEXT:    kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
; X64-NEXT:    vmovd %esi, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc6]
; X64-NEXT:    vpinsrd $1, %edi, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc7,0x01]
; X64-NEXT:    vpinsrd $2, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc0,0x02]
; X64-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1 # encoding: [0xc5,0xf1,0x76,0xc9]
; X64-NEXT:    vpblendd $8, %xmm1, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x02,0xc1,0x08]
; X64-NEXT:    # xmm0 = xmm0[0,1,2],xmm1[3]
; X64-NEXT:    vmovd %ecx, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc9]
; X64-NEXT:    vmovd %r8d, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xc1,0x79,0x6e,0xd0]
; X64-NEXT:    vpunpckldq %xmm1, %xmm2, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x62,0xc9]
; X64-NEXT:    # xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; X64-NEXT:    vmovd %edx, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xd2]
; X64-NEXT:    vpunpcklqdq %xmm2, %xmm1, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0x6c,0xca]
; X64-NEXT:    # xmm1 = xmm1[0],xmm2[0]
; X64-NEXT:    vinserti128 $1, %xmm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x75,0x38,0xc0,0x01]
; X64-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) {
; X86-LABEL: test_mask_cmp_b_256:
; X86:       # %bb.0:
; X86-NEXT:    pushl %ebp # encoding: [0x55]
; X86-NEXT:    .cfi_def_cfa_offset 8
; X86-NEXT:    pushl %ebx # encoding: [0x53]
; X86-NEXT:    .cfi_def_cfa_offset 12
; X86-NEXT:    pushl %edi # encoding: [0x57]
; X86-NEXT:    .cfi_def_cfa_offset 16
; X86-NEXT:    pushl %esi # encoding: [0x56]
; X86-NEXT:    .cfi_def_cfa_offset 20
; X86-NEXT:    .cfi_offset %esi, -20
; X86-NEXT:    .cfi_offset %edi, -16
; X86-NEXT:    .cfi_offset %ebx, -12
; X86-NEXT:    .cfi_offset %ebp, -8
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x14]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpcmpeqb %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x74,0xc1]
; X86-NEXT:    kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
; X86-NEXT:    vpcmpgtb %ymm0, %ymm1, %k0 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x64,0xc0]
; X86-NEXT:    kmovd %k0, %edx # encoding: [0xc5,0xfb,0x93,0xd0]
; X86-NEXT:    vpcmpleb %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x3f,0xc1,0x02]
; X86-NEXT:    kmovd %k0, %esi # encoding: [0xc5,0xfb,0x93,0xf0]
; X86-NEXT:    vpcmpneqb %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x3f,0xc1,0x04]
; X86-NEXT:    kmovd %k0, %edi # encoding: [0xc5,0xfb,0x93,0xf8]
; X86-NEXT:    vpcmpnltb %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x3f,0xc1,0x05]
; X86-NEXT:    kmovd %k0, %ebx # encoding: [0xc5,0xfb,0x93,0xd8]
; X86-NEXT:    vpcmpgtb %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x64,0xc1]
; X86-NEXT:    kmovd %k0, %ebp # encoding: [0xc5,0xfb,0x93,0xe8]
; X86-NEXT:    vmovd %edi, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc7]
; X86-NEXT:    vpinsrd $1, %ebx, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc3,0x01]
; X86-NEXT:    vpinsrd $2, %ebp, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc5,0x02]
; X86-NEXT:    vpinsrd $3, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc0,0x03]
; X86-NEXT:    vmovd %edx, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xca]
; X86-NEXT:    vmovd %ecx, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xd1]
; X86-NEXT:    vpunpckldq %xmm1, %xmm2, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x62,0xc9]
; X86-NEXT:    # xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; X86-NEXT:    vmovd %esi, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xd6]
; X86-NEXT:    vpunpcklqdq %xmm2, %xmm1, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0x6c,0xca]
; X86-NEXT:    # xmm1 = xmm1[0],xmm2[0]
; X86-NEXT:    vinserti128 $1, %xmm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x75,0x38,0xc0,0x01]
; X86-NEXT:    popl %esi # encoding: [0x5e]
; X86-NEXT:    .cfi_def_cfa_offset 16
; X86-NEXT:    popl %edi # encoding: [0x5f]
; X86-NEXT:    .cfi_def_cfa_offset 12
; X86-NEXT:    popl %ebx # encoding: [0x5b]
; X86-NEXT:    .cfi_def_cfa_offset 8
; X86-NEXT:    popl %ebp # encoding: [0x5d]
; X86-NEXT:    .cfi_def_cfa_offset 4
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_cmp_b_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpcmpeqb %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x74,0xc1]
; X64-NEXT:    kmovd %k0, %r8d # encoding: [0xc5,0x7b,0x93,0xc0]
; X64-NEXT:    vpcmpgtb %ymm0, %ymm1, %k0 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x64,0xc0]
; X64-NEXT:    kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
; X64-NEXT:    vpcmpleb %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x3f,0xc1,0x02]
; X64-NEXT:    kmovd %k0, %r9d # encoding: [0xc5,0x7b,0x93,0xc8]
; X64-NEXT:    vpcmpneqb %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x3f,0xc1,0x04]
; X64-NEXT:    kmovd %k0, %esi # encoding: [0xc5,0xfb,0x93,0xf0]
; X64-NEXT:    vpcmpnltb %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x3f,0xc1,0x05]
; X64-NEXT:    kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
; X64-NEXT:    vpcmpgtb %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x64,0xc1]
; X64-NEXT:    kmovd %k0, %edx # encoding: [0xc5,0xfb,0x93,0xd0]
; X64-NEXT:    vmovd %esi, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc6]
; X64-NEXT:    vpinsrd $1, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc0,0x01]
; X64-NEXT:    vpinsrd $2, %edx, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc2,0x02]
; X64-NEXT:    vpinsrd $3, %edi, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc7,0x03]
; X64-NEXT:    vmovd %ecx, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc9]
; X64-NEXT:    vmovd %r8d, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xc1,0x79,0x6e,0xd0]
; X64-NEXT:    vpunpckldq %xmm1, %xmm2, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x62,0xc9]
; X64-NEXT:    # xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; X64-NEXT:    vmovd %r9d, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xc1,0x79,0x6e,0xd1]
; X64-NEXT:    vpunpcklqdq %xmm2, %xmm1, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0x6c,0xca]
; X64-NEXT:    # xmm1 = xmm1[0],xmm2[0]
; X64-NEXT:    vinserti128 $1, %xmm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x75,0x38,0xc0,0x01]
; X64-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) {
; X86-LABEL: test_ucmp_b_256:
; X86:       # %bb.0:
; X86-NEXT:    pushl %ebx # encoding: [0x53]
; X86-NEXT:    .cfi_def_cfa_offset 8
; X86-NEXT:    pushl %edi # encoding: [0x57]
; X86-NEXT:    .cfi_def_cfa_offset 12
; X86-NEXT:    pushl %esi # encoding: [0x56]
; X86-NEXT:    .cfi_def_cfa_offset 16
; X86-NEXT:    .cfi_offset %esi, -16
; X86-NEXT:    .cfi_offset %edi, -12
; X86-NEXT:    .cfi_offset %ebx, -8
; X86-NEXT:    vpcmpeqb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x74,0xc1]
; X86-NEXT:    kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
; X86-NEXT:    vpcmpltub %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x3e,0xc1,0x01]
; X86-NEXT:    kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
; X86-NEXT:    vpcmpleub %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x3e,0xc1,0x02]
; X86-NEXT:    kmovd %k0, %edx # encoding: [0xc5,0xfb,0x93,0xd0]
; X86-NEXT:    vpcmpneqb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x3f,0xc1,0x04]
; X86-NEXT:    kmovd %k0, %esi # encoding: [0xc5,0xfb,0x93,0xf0]
; X86-NEXT:    vpcmpnltub %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x3e,0xc1,0x05]
; X86-NEXT:    kmovd %k0, %edi # encoding: [0xc5,0xfb,0x93,0xf8]
; X86-NEXT:    vpcmpnleub %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x3e,0xc1,0x06]
; X86-NEXT:    kmovd %k0, %ebx # encoding: [0xc5,0xfb,0x93,0xd8]
; X86-NEXT:    vmovd %esi, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc6]
; X86-NEXT:    vpinsrd $1, %edi, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc7,0x01]
; X86-NEXT:    vpinsrd $2, %ebx, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc3,0x02]
; X86-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1 # encoding: [0xc5,0xf1,0x76,0xc9]
; X86-NEXT:    vpblendd $8, %xmm1, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x02,0xc1,0x08]
; X86-NEXT:    # xmm0 = xmm0[0,1,2],xmm1[3]
; X86-NEXT:    vmovd %ecx, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc9]
; X86-NEXT:    vmovd %eax, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xd0]
; X86-NEXT:    vpunpckldq %xmm1, %xmm2, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x62,0xc9]
; X86-NEXT:    # xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; X86-NEXT:    vmovd %edx, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xd2]
; X86-NEXT:    vpunpcklqdq %xmm2, %xmm1, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0x6c,0xca]
; X86-NEXT:    # xmm1 = xmm1[0],xmm2[0]
; X86-NEXT:    vinserti128 $1, %xmm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x75,0x38,0xc0,0x01]
; X86-NEXT:    popl %esi # encoding: [0x5e]
; X86-NEXT:    .cfi_def_cfa_offset 12
; X86-NEXT:    popl %edi # encoding: [0x5f]
; X86-NEXT:    .cfi_def_cfa_offset 8
; X86-NEXT:    popl %ebx # encoding: [0x5b]
; X86-NEXT:    .cfi_def_cfa_offset 4
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_ucmp_b_256:
; X64:       # %bb.0:
; X64-NEXT:    vpcmpeqb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x74,0xc1]
; X64-NEXT:    kmovd %k0, %r8d # encoding: [0xc5,0x7b,0x93,0xc0]
; X64-NEXT:    vpcmpltub %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x3e,0xc1,0x01]
; X64-NEXT:    kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
; X64-NEXT:    vpcmpleub %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x3e,0xc1,0x02]
; X64-NEXT:    kmovd %k0, %edx # encoding: [0xc5,0xfb,0x93,0xd0]
; X64-NEXT:    vpcmpneqb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x3f,0xc1,0x04]
; X64-NEXT:    kmovd %k0, %esi # encoding: [0xc5,0xfb,0x93,0xf0]
; X64-NEXT:    vpcmpnltub %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x3e,0xc1,0x05]
; X64-NEXT:    kmovd %k0, %edi # encoding: [0xc5,0xfb,0x93,0xf8]
; X64-NEXT:    vpcmpnleub %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x3e,0xc1,0x06]
; X64-NEXT:    kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
; X64-NEXT:    vmovd %esi, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc6]
; X64-NEXT:    vpinsrd $1, %edi, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc7,0x01]
; X64-NEXT:    vpinsrd $2, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc0,0x02]
; X64-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1 # encoding: [0xc5,0xf1,0x76,0xc9]
; X64-NEXT:    vpblendd $8, %xmm1, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x02,0xc1,0x08]
; X64-NEXT:    # xmm0 = xmm0[0,1,2],xmm1[3]
; X64-NEXT:    vmovd %ecx, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc9]
; X64-NEXT:    vmovd %r8d, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xc1,0x79,0x6e,0xd0]
; X64-NEXT:    vpunpckldq %xmm1, %xmm2, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x62,0xc9]
; X64-NEXT:    # xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; X64-NEXT:    vmovd %edx, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xd2]
; X64-NEXT:    vpunpcklqdq %xmm2, %xmm1, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0x6c,0xca]
; X64-NEXT:    # xmm1 = xmm1[0],xmm2[0]
; X64-NEXT:    vinserti128 $1, %xmm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x75,0x38,0xc0,0x01]
; X64-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) {
; X86-LABEL: test_mask_ucmp_b_256:
; X86:       # %bb.0:
; X86-NEXT:    pushl %ebp # encoding: [0x55]
; X86-NEXT:    .cfi_def_cfa_offset 8
; X86-NEXT:    pushl %ebx # encoding: [0x53]
; X86-NEXT:    .cfi_def_cfa_offset 12
; X86-NEXT:    pushl %edi # encoding: [0x57]
; X86-NEXT:    .cfi_def_cfa_offset 16
; X86-NEXT:    pushl %esi # encoding: [0x56]
; X86-NEXT:    .cfi_def_cfa_offset 20
; X86-NEXT:    .cfi_offset %esi, -20
; X86-NEXT:    .cfi_offset %edi, -16
; X86-NEXT:    .cfi_offset %ebx, -12
; X86-NEXT:    .cfi_offset %ebp, -8
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x14]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpcmpeqb %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x74,0xc1]
; X86-NEXT:    kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
; X86-NEXT:    vpcmpltub %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x01]
; X86-NEXT:    kmovd %k0, %edx # encoding: [0xc5,0xfb,0x93,0xd0]
; X86-NEXT:    vpcmpleub %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x02]
; X86-NEXT:    kmovd %k0, %esi # encoding: [0xc5,0xfb,0x93,0xf0]
; X86-NEXT:    vpcmpneqb %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x3f,0xc1,0x04]
; X86-NEXT:    kmovd %k0, %edi # encoding: [0xc5,0xfb,0x93,0xf8]
; X86-NEXT:    vpcmpnltub %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x05]
; X86-NEXT:    kmovd %k0, %ebx # encoding: [0xc5,0xfb,0x93,0xd8]
; X86-NEXT:    vpcmpnleub %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x06]
; X86-NEXT:    kmovd %k0, %ebp # encoding: [0xc5,0xfb,0x93,0xe8]
; X86-NEXT:    vmovd %edi, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc7]
; X86-NEXT:    vpinsrd $1, %ebx, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc3,0x01]
; X86-NEXT:    vpinsrd $2, %ebp, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc5,0x02]
; X86-NEXT:    vpinsrd $3, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc0,0x03]
; X86-NEXT:    vmovd %edx, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xca]
; X86-NEXT:    vmovd %ecx, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xd1]
; X86-NEXT:    vpunpckldq %xmm1, %xmm2, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x62,0xc9]
; X86-NEXT:    # xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; X86-NEXT:    vmovd %esi, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xd6]
; X86-NEXT:    vpunpcklqdq %xmm2, %xmm1, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0x6c,0xca]
; X86-NEXT:    # xmm1 = xmm1[0],xmm2[0]
; X86-NEXT:    vinserti128 $1, %xmm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x75,0x38,0xc0,0x01]
; X86-NEXT:    popl %esi # encoding: [0x5e]
; X86-NEXT:    .cfi_def_cfa_offset 16
; X86-NEXT:    popl %edi # encoding: [0x5f]
; X86-NEXT:    .cfi_def_cfa_offset 12
; X86-NEXT:    popl %ebx # encoding: [0x5b]
; X86-NEXT:    .cfi_def_cfa_offset 8
; X86-NEXT:    popl %ebp # encoding: [0x5d]
; X86-NEXT:    .cfi_def_cfa_offset 4
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_ucmp_b_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpcmpeqb %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x74,0xc1]
; X64-NEXT:    kmovd %k0, %r8d # encoding: [0xc5,0x7b,0x93,0xc0]
; X64-NEXT:    vpcmpltub %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x01]
; X64-NEXT:    kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
; X64-NEXT:    vpcmpleub %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x02]
; X64-NEXT:    kmovd %k0, %r9d # encoding: [0xc5,0x7b,0x93,0xc8]
; X64-NEXT:    vpcmpneqb %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x3f,0xc1,0x04]
; X64-NEXT:    kmovd %k0, %esi # encoding: [0xc5,0xfb,0x93,0xf0]
; X64-NEXT:    vpcmpnltub %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x05]
; X64-NEXT:    kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
; X64-NEXT:    vpcmpnleub %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x06]
; X64-NEXT:    kmovd %k0, %edx # encoding: [0xc5,0xfb,0x93,0xd0]
; X64-NEXT:    vmovd %esi, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc6]
; X64-NEXT:    vpinsrd $1, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc0,0x01]
; X64-NEXT:    vpinsrd $2, %edx, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc2,0x02]
; X64-NEXT:    vpinsrd $3, %edi, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc7,0x03]
; X64-NEXT:    vmovd %ecx, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc9]
; X64-NEXT:    vmovd %r8d, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xc1,0x79,0x6e,0xd0]
; X64-NEXT:    vpunpckldq %xmm1, %xmm2, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x62,0xc9]
; X64-NEXT:    # xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
; X64-NEXT:    vmovd %r9d, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xc1,0x79,0x6e,0xd1]
; X64-NEXT:    vpunpcklqdq %xmm2, %xmm1, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0x6c,0xca]
; X64-NEXT:    # xmm1 = xmm1[0],xmm2[0]
; X64-NEXT:    vinserti128 $1, %xmm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x75,0x38,0xc0,0x01]
; X64-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: [0x62,0xf1,0x7d,0x28,0x75,0xc1]
; CHECK-NEXT:    vpcmpgtw %ymm0, %ymm1, %k1 # encoding: [0x62,0xf1,0x75,0x28,0x65,0xc8]
; CHECK-NEXT:    vpcmplew %ymm1, %ymm0, %k2 # encoding: [0x62,0xf3,0xfd,0x28,0x3f,0xd1,0x02]
; CHECK-NEXT:    vpcmpneqw %ymm1, %ymm0, %k3 # encoding: [0x62,0xf3,0xfd,0x28,0x3f,0xd9,0x04]
; CHECK-NEXT:    vpcmpnltw %ymm1, %ymm0, %k4 # encoding: [0x62,0xf3,0xfd,0x28,0x3f,0xe1,0x05]
; CHECK-NEXT:    vpcmpgtw %ymm1, %ymm0, %k5 # encoding: [0x62,0xf1,0x7d,0x28,0x65,0xe9]
; CHECK-NEXT:    kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
; CHECK-NEXT:    vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
; CHECK-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
; CHECK-NEXT:    kmovd %k1, %eax # encoding: [0xc5,0xfb,0x93,0xc1]
; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
; CHECK-NEXT:    kmovd %k2, %eax # encoding: [0xc5,0xfb,0x93,0xc2]
; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
; CHECK-NEXT:    kmovd %k3, %eax # encoding: [0xc5,0xfb,0x93,0xc3]
; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
; CHECK-NEXT:    kmovd %k4, %eax # encoding: [0xc5,0xfb,0x93,0xc4]
; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
; CHECK-NEXT:    kmovd %k5, %eax # encoding: [0xc5,0xfb,0x93,0xc5]
; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
; CHECK-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1 # encoding: [0xc5,0xf1,0x76,0xc9]
; CHECK-NEXT:    vpblendw $128, %xmm1, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x0e,0xc1,0x80]
; CHECK-NEXT:    # xmm0 = xmm0[0,1,2,3,4,5,6],xmm1[7]
; CHECK-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; CHECK-NEXT:    ret{{[l|q]}} # 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) {
; X86-LABEL: test_mask_cmp_w_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpcmpeqw %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x75,0xc1]
; X86-NEXT:    vpcmpgtw %ymm0, %ymm1, %k2 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x65,0xd0]
; X86-NEXT:    vpcmplew %ymm1, %ymm0, %k3 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x3f,0xd9,0x02]
; X86-NEXT:    vpcmpneqw %ymm1, %ymm0, %k4 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x3f,0xe1,0x04]
; X86-NEXT:    vpcmpnltw %ymm1, %ymm0, %k5 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x3f,0xe9,0x05]
; X86-NEXT:    vpcmpgtw %ymm1, %ymm0, %k1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x65,0xc9]
; X86-NEXT:    kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
; X86-NEXT:    vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
; X86-NEXT:    vpinsrw $0, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x00]
; X86-NEXT:    kmovd %k2, %ecx # encoding: [0xc5,0xfb,0x93,0xca]
; X86-NEXT:    vpinsrw $1, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x01]
; X86-NEXT:    kmovd %k3, %ecx # encoding: [0xc5,0xfb,0x93,0xcb]
; X86-NEXT:    vpinsrw $2, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x02]
; X86-NEXT:    kmovd %k4, %ecx # encoding: [0xc5,0xfb,0x93,0xcc]
; X86-NEXT:    vpinsrw $4, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x04]
; X86-NEXT:    kmovd %k5, %ecx # encoding: [0xc5,0xfb,0x93,0xcd]
; X86-NEXT:    vpinsrw $5, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x05]
; X86-NEXT:    kmovd %k1, %ecx # encoding: [0xc5,0xfb,0x93,0xc9]
; X86-NEXT:    vpinsrw $6, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x06]
; X86-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_cmp_w_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpcmpeqw %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x75,0xc1]
; X64-NEXT:    vpcmpgtw %ymm0, %ymm1, %k2 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x65,0xd0]
; X64-NEXT:    vpcmplew %ymm1, %ymm0, %k3 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x3f,0xd9,0x02]
; X64-NEXT:    vpcmpneqw %ymm1, %ymm0, %k4 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x3f,0xe1,0x04]
; X64-NEXT:    vpcmpnltw %ymm1, %ymm0, %k5 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x3f,0xe9,0x05]
; X64-NEXT:    vpcmpgtw %ymm1, %ymm0, %k1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x65,0xc9]
; X64-NEXT:    kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
; X64-NEXT:    vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
; X64-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
; X64-NEXT:    kmovd %k2, %eax # encoding: [0xc5,0xfb,0x93,0xc2]
; X64-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
; X64-NEXT:    kmovd %k3, %eax # encoding: [0xc5,0xfb,0x93,0xc3]
; X64-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
; X64-NEXT:    kmovd %k4, %eax # encoding: [0xc5,0xfb,0x93,0xc4]
; X64-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
; X64-NEXT:    kmovd %k5, %eax # encoding: [0xc5,0xfb,0x93,0xc5]
; X64-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
; X64-NEXT:    kmovd %k1, %eax # encoding: [0xc5,0xfb,0x93,0xc1]
; X64-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
; X64-NEXT:    vpinsrw $7, %edi, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc7,0x07]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-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:    vpcmpeqw %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x75,0xc1]
; 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:    vpcmpneqw %ymm1, %ymm0, %k3 # encoding: [0x62,0xf3,0xfd,0x28,0x3f,0xd9,0x04]
; CHECK-NEXT:    vpcmpnltuw %ymm1, %ymm0, %k4 # encoding: [0x62,0xf3,0xfd,0x28,0x3e,0xe1,0x05]
; CHECK-NEXT:    vpcmpnleuw %ymm1, %ymm0, %k5 # encoding: [0x62,0xf3,0xfd,0x28,0x3e,0xe9,0x06]
; CHECK-NEXT:    kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
; CHECK-NEXT:    vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
; CHECK-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
; CHECK-NEXT:    kmovd %k1, %eax # encoding: [0xc5,0xfb,0x93,0xc1]
; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
; CHECK-NEXT:    kmovd %k2, %eax # encoding: [0xc5,0xfb,0x93,0xc2]
; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
; CHECK-NEXT:    kmovd %k3, %eax # encoding: [0xc5,0xfb,0x93,0xc3]
; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
; CHECK-NEXT:    kmovd %k4, %eax # encoding: [0xc5,0xfb,0x93,0xc4]
; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
; CHECK-NEXT:    kmovd %k5, %eax # encoding: [0xc5,0xfb,0x93,0xc5]
; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
; CHECK-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1 # encoding: [0xc5,0xf1,0x76,0xc9]
; CHECK-NEXT:    vpblendw $128, %xmm1, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x0e,0xc1,0x80]
; CHECK-NEXT:    # xmm0 = xmm0[0,1,2,3,4,5,6],xmm1[7]
; CHECK-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; CHECK-NEXT:    ret{{[l|q]}} # 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) {
; X86-LABEL: test_mask_ucmp_w_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpcmpeqw %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x75,0xc1]
; X86-NEXT:    vpcmpltuw %ymm1, %ymm0, %k2 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x3e,0xd1,0x01]
; X86-NEXT:    vpcmpleuw %ymm1, %ymm0, %k3 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x3e,0xd9,0x02]
; X86-NEXT:    vpcmpneqw %ymm1, %ymm0, %k4 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x3f,0xe1,0x04]
; X86-NEXT:    vpcmpnltuw %ymm1, %ymm0, %k5 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x3e,0xe9,0x05]
; X86-NEXT:    vpcmpnleuw %ymm1, %ymm0, %k1 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x3e,0xc9,0x06]
; X86-NEXT:    kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
; X86-NEXT:    vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
; X86-NEXT:    vpinsrw $0, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x00]
; X86-NEXT:    kmovd %k2, %ecx # encoding: [0xc5,0xfb,0x93,0xca]
; X86-NEXT:    vpinsrw $1, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x01]
; X86-NEXT:    kmovd %k3, %ecx # encoding: [0xc5,0xfb,0x93,0xcb]
; X86-NEXT:    vpinsrw $2, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x02]
; X86-NEXT:    kmovd %k4, %ecx # encoding: [0xc5,0xfb,0x93,0xcc]
; X86-NEXT:    vpinsrw $4, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x04]
; X86-NEXT:    kmovd %k5, %ecx # encoding: [0xc5,0xfb,0x93,0xcd]
; X86-NEXT:    vpinsrw $5, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x05]
; X86-NEXT:    kmovd %k1, %ecx # encoding: [0xc5,0xfb,0x93,0xc9]
; X86-NEXT:    vpinsrw $6, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x06]
; X86-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_ucmp_w_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpcmpeqw %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x75,0xc1]
; X64-NEXT:    vpcmpltuw %ymm1, %ymm0, %k2 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x3e,0xd1,0x01]
; X64-NEXT:    vpcmpleuw %ymm1, %ymm0, %k3 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x3e,0xd9,0x02]
; X64-NEXT:    vpcmpneqw %ymm1, %ymm0, %k4 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x3f,0xe1,0x04]
; X64-NEXT:    vpcmpnltuw %ymm1, %ymm0, %k5 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x3e,0xe9,0x05]
; X64-NEXT:    vpcmpnleuw %ymm1, %ymm0, %k1 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x3e,0xc9,0x06]
; X64-NEXT:    kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
; X64-NEXT:    vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
; X64-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
; X64-NEXT:    kmovd %k2, %eax # encoding: [0xc5,0xfb,0x93,0xc2]
; X64-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
; X64-NEXT:    kmovd %k3, %eax # encoding: [0xc5,0xfb,0x93,0xc3]
; X64-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
; X64-NEXT:    kmovd %k4, %eax # encoding: [0xc5,0xfb,0x93,0xc4]
; X64-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
; X64-NEXT:    kmovd %k5, %eax # encoding: [0xc5,0xfb,0x93,0xc5]
; X64-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
; X64-NEXT:    kmovd %k1, %eax # encoding: [0xc5,0xfb,0x93,0xc1]
; X64-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
; X64-NEXT:    vpinsrw $7, %edi, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc7,0x07]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-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

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: [0x62,0xf1,0x7d,0x08,0x74,0xc1]
; CHECK-NEXT:    vpcmpgtb %xmm0, %xmm1, %k1 # encoding: [0x62,0xf1,0x75,0x08,0x64,0xc8]
; CHECK-NEXT:    vpcmpleb %xmm1, %xmm0, %k2 # encoding: [0x62,0xf3,0x7d,0x08,0x3f,0xd1,0x02]
; CHECK-NEXT:    vpcmpneqb %xmm1, %xmm0, %k3 # encoding: [0x62,0xf3,0x7d,0x08,0x3f,0xd9,0x04]
; CHECK-NEXT:    vpcmpnltb %xmm1, %xmm0, %k4 # encoding: [0x62,0xf3,0x7d,0x08,0x3f,0xe1,0x05]
; CHECK-NEXT:    vpcmpgtb %xmm1, %xmm0, %k5 # encoding: [0x62,0xf1,0x7d,0x08,0x64,0xe9]
; CHECK-NEXT:    kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
; CHECK-NEXT:    vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
; CHECK-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
; CHECK-NEXT:    kmovd %k1, %eax # encoding: [0xc5,0xfb,0x93,0xc1]
; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
; CHECK-NEXT:    kmovd %k2, %eax # encoding: [0xc5,0xfb,0x93,0xc2]
; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
; CHECK-NEXT:    kmovd %k3, %eax # encoding: [0xc5,0xfb,0x93,0xc3]
; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
; CHECK-NEXT:    kmovd %k4, %eax # encoding: [0xc5,0xfb,0x93,0xc4]
; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
; CHECK-NEXT:    kmovd %k5, %eax # encoding: [0xc5,0xfb,0x93,0xc5]
; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
; CHECK-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1 # encoding: [0xc5,0xf1,0x76,0xc9]
; CHECK-NEXT:    vpblendw $128, %xmm1, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x0e,0xc1,0x80]
; CHECK-NEXT:    # xmm0 = xmm0[0,1,2,3,4,5,6],xmm1[7]
; CHECK-NEXT:    ret{{[l|q]}} # 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) {
; X86-LABEL: test_mask_cmp_b_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpcmpeqb %xmm1, %xmm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x74,0xc1]
; X86-NEXT:    vpcmpgtb %xmm0, %xmm1, %k2 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x64,0xd0]
; X86-NEXT:    vpcmpleb %xmm1, %xmm0, %k3 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x3f,0xd9,0x02]
; X86-NEXT:    vpcmpneqb %xmm1, %xmm0, %k4 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x3f,0xe1,0x04]
; X86-NEXT:    vpcmpnltb %xmm1, %xmm0, %k5 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x3f,0xe9,0x05]
; X86-NEXT:    vpcmpgtb %xmm1, %xmm0, %k1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x64,0xc9]
; X86-NEXT:    kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
; X86-NEXT:    vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
; X86-NEXT:    vpinsrw $0, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x00]
; X86-NEXT:    kmovd %k2, %ecx # encoding: [0xc5,0xfb,0x93,0xca]
; X86-NEXT:    vpinsrw $1, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x01]
; X86-NEXT:    kmovd %k3, %ecx # encoding: [0xc5,0xfb,0x93,0xcb]
; X86-NEXT:    vpinsrw $2, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x02]
; X86-NEXT:    kmovd %k4, %ecx # encoding: [0xc5,0xfb,0x93,0xcc]
; X86-NEXT:    vpinsrw $4, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x04]
; X86-NEXT:    kmovd %k5, %ecx # encoding: [0xc5,0xfb,0x93,0xcd]
; X86-NEXT:    vpinsrw $5, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x05]
; X86-NEXT:    kmovd %k1, %ecx # encoding: [0xc5,0xfb,0x93,0xc9]
; X86-NEXT:    vpinsrw $6, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x06]
; X86-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_cmp_b_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpcmpeqb %xmm1, %xmm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x74,0xc1]
; X64-NEXT:    vpcmpgtb %xmm0, %xmm1, %k2 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x64,0xd0]
; X64-NEXT:    vpcmpleb %xmm1, %xmm0, %k3 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x3f,0xd9,0x02]
; X64-NEXT:    vpcmpneqb %xmm1, %xmm0, %k4 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x3f,0xe1,0x04]
; X64-NEXT:    vpcmpnltb %xmm1, %xmm0, %k5 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x3f,0xe9,0x05]
; X64-NEXT:    vpcmpgtb %xmm1, %xmm0, %k1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x64,0xc9]
; X64-NEXT:    kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
; X64-NEXT:    vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
; X64-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
; X64-NEXT:    kmovd %k2, %eax # encoding: [0xc5,0xfb,0x93,0xc2]
; X64-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
; X64-NEXT:    kmovd %k3, %eax # encoding: [0xc5,0xfb,0x93,0xc3]
; X64-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
; X64-NEXT:    kmovd %k4, %eax # encoding: [0xc5,0xfb,0x93,0xc4]
; X64-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
; X64-NEXT:    kmovd %k5, %eax # encoding: [0xc5,0xfb,0x93,0xc5]
; X64-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
; X64-NEXT:    kmovd %k1, %eax # encoding: [0xc5,0xfb,0x93,0xc1]
; X64-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
; X64-NEXT:    vpinsrw $7, %edi, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc7,0x07]
; X64-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:    vpcmpeqb %xmm1, %xmm0, %k0 # encoding: [0x62,0xf1,0x7d,0x08,0x74,0xc1]
; 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:    vpcmpneqb %xmm1, %xmm0, %k3 # encoding: [0x62,0xf3,0x7d,0x08,0x3f,0xd9,0x04]
; CHECK-NEXT:    vpcmpnltub %xmm1, %xmm0, %k4 # encoding: [0x62,0xf3,0x7d,0x08,0x3e,0xe1,0x05]
; CHECK-NEXT:    vpcmpnleub %xmm1, %xmm0, %k5 # encoding: [0x62,0xf3,0x7d,0x08,0x3e,0xe9,0x06]
; CHECK-NEXT:    kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
; CHECK-NEXT:    vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
; CHECK-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
; CHECK-NEXT:    kmovd %k1, %eax # encoding: [0xc5,0xfb,0x93,0xc1]
; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
; CHECK-NEXT:    kmovd %k2, %eax # encoding: [0xc5,0xfb,0x93,0xc2]
; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
; CHECK-NEXT:    kmovd %k3, %eax # encoding: [0xc5,0xfb,0x93,0xc3]
; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
; CHECK-NEXT:    kmovd %k4, %eax # encoding: [0xc5,0xfb,0x93,0xc4]
; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
; CHECK-NEXT:    kmovd %k5, %eax # encoding: [0xc5,0xfb,0x93,0xc5]
; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
; CHECK-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1 # encoding: [0xc5,0xf1,0x76,0xc9]
; CHECK-NEXT:    vpblendw $128, %xmm1, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x0e,0xc1,0x80]
; CHECK-NEXT:    # xmm0 = xmm0[0,1,2,3,4,5,6],xmm1[7]
; CHECK-NEXT:    ret{{[l|q]}} # 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) {
; X86-LABEL: test_mask_ucmp_b_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpcmpeqb %xmm1, %xmm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x74,0xc1]
; X86-NEXT:    vpcmpltub %xmm1, %xmm0, %k2 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x3e,0xd1,0x01]
; X86-NEXT:    vpcmpleub %xmm1, %xmm0, %k3 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x3e,0xd9,0x02]
; X86-NEXT:    vpcmpneqb %xmm1, %xmm0, %k4 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x3f,0xe1,0x04]
; X86-NEXT:    vpcmpnltub %xmm1, %xmm0, %k5 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x3e,0xe9,0x05]
; X86-NEXT:    vpcmpnleub %xmm1, %xmm0, %k1 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x3e,0xc9,0x06]
; X86-NEXT:    kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
; X86-NEXT:    vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
; X86-NEXT:    vpinsrw $0, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x00]
; X86-NEXT:    kmovd %k2, %ecx # encoding: [0xc5,0xfb,0x93,0xca]
; X86-NEXT:    vpinsrw $1, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x01]
; X86-NEXT:    kmovd %k3, %ecx # encoding: [0xc5,0xfb,0x93,0xcb]
; X86-NEXT:    vpinsrw $2, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x02]
; X86-NEXT:    kmovd %k4, %ecx # encoding: [0xc5,0xfb,0x93,0xcc]
; X86-NEXT:    vpinsrw $4, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x04]
; X86-NEXT:    kmovd %k5, %ecx # encoding: [0xc5,0xfb,0x93,0xcd]
; X86-NEXT:    vpinsrw $5, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x05]
; X86-NEXT:    kmovd %k1, %ecx # encoding: [0xc5,0xfb,0x93,0xc9]
; X86-NEXT:    vpinsrw $6, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x06]
; X86-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_ucmp_b_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpcmpeqb %xmm1, %xmm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x74,0xc1]
; X64-NEXT:    vpcmpltub %xmm1, %xmm0, %k2 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x3e,0xd1,0x01]
; X64-NEXT:    vpcmpleub %xmm1, %xmm0, %k3 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x3e,0xd9,0x02]
; X64-NEXT:    vpcmpneqb %xmm1, %xmm0, %k4 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x3f,0xe1,0x04]
; X64-NEXT:    vpcmpnltub %xmm1, %xmm0, %k5 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x3e,0xe9,0x05]
; X64-NEXT:    vpcmpnleub %xmm1, %xmm0, %k1 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x3e,0xc9,0x06]
; X64-NEXT:    kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
; X64-NEXT:    vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
; X64-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
; X64-NEXT:    kmovd %k2, %eax # encoding: [0xc5,0xfb,0x93,0xc2]
; X64-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
; X64-NEXT:    kmovd %k3, %eax # encoding: [0xc5,0xfb,0x93,0xc3]
; X64-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
; X64-NEXT:    kmovd %k4, %eax # encoding: [0xc5,0xfb,0x93,0xc4]
; X64-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
; X64-NEXT:    kmovd %k5, %eax # encoding: [0xc5,0xfb,0x93,0xc5]
; X64-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
; X64-NEXT:    kmovd %k1, %eax # encoding: [0xc5,0xfb,0x93,0xc1]
; X64-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
; X64-NEXT:    vpinsrw $7, %edi, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc7,0x07]
; X64-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: [0x62,0xf1,0x7d,0x08,0x75,0xc1]
; CHECK-NEXT:    vpcmpgtw %xmm0, %xmm1, %k1 # encoding: [0x62,0xf1,0x75,0x08,0x65,0xc8]
; CHECK-NEXT:    vpcmplew %xmm1, %xmm0, %k2 # encoding: [0x62,0xf3,0xfd,0x08,0x3f,0xd1,0x02]
; CHECK-NEXT:    vpcmpneqw %xmm1, %xmm0, %k3 # encoding: [0x62,0xf3,0xfd,0x08,0x3f,0xd9,0x04]
; CHECK-NEXT:    vpcmpnltw %xmm1, %xmm0, %k4 # encoding: [0x62,0xf3,0xfd,0x08,0x3f,0xe1,0x05]
; CHECK-NEXT:    vpcmpgtw %xmm1, %xmm0, %k5 # encoding: [0x62,0xf1,0x7d,0x08,0x65,0xe9]
; CHECK-NEXT:    kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
; CHECK-NEXT:    vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
; CHECK-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
; CHECK-NEXT:    kmovd %k1, %eax # encoding: [0xc5,0xfb,0x93,0xc1]
; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
; CHECK-NEXT:    kmovd %k2, %eax # encoding: [0xc5,0xfb,0x93,0xc2]
; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
; CHECK-NEXT:    kmovd %k3, %eax # encoding: [0xc5,0xfb,0x93,0xc3]
; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
; CHECK-NEXT:    kmovd %k4, %eax # encoding: [0xc5,0xfb,0x93,0xc4]
; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
; CHECK-NEXT:    kmovd %k5, %eax # encoding: [0xc5,0xfb,0x93,0xc5]
; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
; CHECK-NEXT:    movl $255, %eax # encoding: [0xb8,0xff,0x00,0x00,0x00]
; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
; CHECK-NEXT:    ret{{[l|q]}} # 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) {
; X86-LABEL: test_mask_cmp_w_128:
; X86:       # %bb.0:
; X86-NEXT:    movzwl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb7,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpcmpeqw %xmm1, %xmm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x75,0xc1]
; X86-NEXT:    vpcmpgtw %xmm0, %xmm1, %k2 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x65,0xd0]
; X86-NEXT:    vpcmplew %xmm1, %xmm0, %k3 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x3f,0xd9,0x02]
; X86-NEXT:    vpcmpneqw %xmm1, %xmm0, %k4 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x3f,0xe1,0x04]
; X86-NEXT:    vpcmpnltw %xmm1, %xmm0, %k5 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x3f,0xe9,0x05]
; X86-NEXT:    vpcmpgtw %xmm1, %xmm0, %k1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x65,0xc9]
; X86-NEXT:    kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
; X86-NEXT:    vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
; X86-NEXT:    vpinsrw $0, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x00]
; X86-NEXT:    kmovd %k2, %ecx # encoding: [0xc5,0xfb,0x93,0xca]
; X86-NEXT:    vpinsrw $1, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x01]
; X86-NEXT:    kmovd %k3, %ecx # encoding: [0xc5,0xfb,0x93,0xcb]
; X86-NEXT:    vpinsrw $2, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x02]
; X86-NEXT:    kmovd %k4, %ecx # encoding: [0xc5,0xfb,0x93,0xcc]
; X86-NEXT:    vpinsrw $4, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x04]
; X86-NEXT:    kmovd %k5, %ecx # encoding: [0xc5,0xfb,0x93,0xcd]
; X86-NEXT:    vpinsrw $5, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x05]
; X86-NEXT:    kmovd %k1, %ecx # encoding: [0xc5,0xfb,0x93,0xc9]
; X86-NEXT:    vpinsrw $6, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x06]
; X86-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_cmp_w_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpcmpeqw %xmm1, %xmm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x75,0xc1]
; X64-NEXT:    vpcmpgtw %xmm0, %xmm1, %k2 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x65,0xd0]
; X64-NEXT:    vpcmplew %xmm1, %xmm0, %k3 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x3f,0xd9,0x02]
; X64-NEXT:    vpcmpneqw %xmm1, %xmm0, %k4 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x3f,0xe1,0x04]
; X64-NEXT:    vpcmpnltw %xmm1, %xmm0, %k5 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x3f,0xe9,0x05]
; X64-NEXT:    vpcmpgtw %xmm1, %xmm0, %k1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x65,0xc9]
; X64-NEXT:    kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
; X64-NEXT:    vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
; X64-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
; X64-NEXT:    kmovd %k2, %eax # encoding: [0xc5,0xfb,0x93,0xc2]
; X64-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
; X64-NEXT:    kmovd %k3, %eax # encoding: [0xc5,0xfb,0x93,0xc3]
; X64-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
; X64-NEXT:    kmovd %k4, %eax # encoding: [0xc5,0xfb,0x93,0xc4]
; X64-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
; X64-NEXT:    kmovd %k5, %eax # encoding: [0xc5,0xfb,0x93,0xc5]
; X64-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
; X64-NEXT:    kmovd %k1, %eax # encoding: [0xc5,0xfb,0x93,0xc1]
; X64-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
; X64-NEXT:    vpinsrw $7, %edi, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc7,0x07]
; X64-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:    vpcmpeqw %xmm1, %xmm0, %k0 # encoding: [0x62,0xf1,0x7d,0x08,0x75,0xc1]
; 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:    vpcmpneqw %xmm1, %xmm0, %k3 # encoding: [0x62,0xf3,0xfd,0x08,0x3f,0xd9,0x04]
; CHECK-NEXT:    vpcmpnltuw %xmm1, %xmm0, %k4 # encoding: [0x62,0xf3,0xfd,0x08,0x3e,0xe1,0x05]
; CHECK-NEXT:    vpcmpnleuw %xmm1, %xmm0, %k5 # encoding: [0x62,0xf3,0xfd,0x08,0x3e,0xe9,0x06]
; CHECK-NEXT:    kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
; CHECK-NEXT:    vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
; CHECK-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
; CHECK-NEXT:    kmovd %k1, %eax # encoding: [0xc5,0xfb,0x93,0xc1]
; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
; CHECK-NEXT:    kmovd %k2, %eax # encoding: [0xc5,0xfb,0x93,0xc2]
; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
; CHECK-NEXT:    kmovd %k3, %eax # encoding: [0xc5,0xfb,0x93,0xc3]
; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
; CHECK-NEXT:    kmovd %k4, %eax # encoding: [0xc5,0xfb,0x93,0xc4]
; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
; CHECK-NEXT:    kmovd %k5, %eax # encoding: [0xc5,0xfb,0x93,0xc5]
; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
; CHECK-NEXT:    movl $255, %eax # encoding: [0xb8,0xff,0x00,0x00,0x00]
; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
; CHECK-NEXT:    ret{{[l|q]}} # 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) {
; X86-LABEL: test_mask_ucmp_w_128:
; X86:       # %bb.0:
; X86-NEXT:    movzwl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb7,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpcmpeqw %xmm1, %xmm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x75,0xc1]
; X86-NEXT:    vpcmpltuw %xmm1, %xmm0, %k2 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x3e,0xd1,0x01]
; X86-NEXT:    vpcmpleuw %xmm1, %xmm0, %k3 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x3e,0xd9,0x02]
; X86-NEXT:    vpcmpneqw %xmm1, %xmm0, %k4 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x3f,0xe1,0x04]
; X86-NEXT:    vpcmpnltuw %xmm1, %xmm0, %k5 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x3e,0xe9,0x05]
; X86-NEXT:    vpcmpnleuw %xmm1, %xmm0, %k1 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x3e,0xc9,0x06]
; X86-NEXT:    kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
; X86-NEXT:    vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
; X86-NEXT:    vpinsrw $0, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x00]
; X86-NEXT:    kmovd %k2, %ecx # encoding: [0xc5,0xfb,0x93,0xca]
; X86-NEXT:    vpinsrw $1, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x01]
; X86-NEXT:    kmovd %k3, %ecx # encoding: [0xc5,0xfb,0x93,0xcb]
; X86-NEXT:    vpinsrw $2, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x02]
; X86-NEXT:    kmovd %k4, %ecx # encoding: [0xc5,0xfb,0x93,0xcc]
; X86-NEXT:    vpinsrw $4, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x04]
; X86-NEXT:    kmovd %k5, %ecx # encoding: [0xc5,0xfb,0x93,0xcd]
; X86-NEXT:    vpinsrw $5, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x05]
; X86-NEXT:    kmovd %k1, %ecx # encoding: [0xc5,0xfb,0x93,0xc9]
; X86-NEXT:    vpinsrw $6, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x06]
; X86-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_ucmp_w_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpcmpeqw %xmm1, %xmm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x75,0xc1]
; X64-NEXT:    vpcmpltuw %xmm1, %xmm0, %k2 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x3e,0xd1,0x01]
; X64-NEXT:    vpcmpleuw %xmm1, %xmm0, %k3 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x3e,0xd9,0x02]
; X64-NEXT:    vpcmpneqw %xmm1, %xmm0, %k4 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x3f,0xe1,0x04]
; X64-NEXT:    vpcmpnltuw %xmm1, %xmm0, %k5 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x3e,0xe9,0x05]
; X64-NEXT:    vpcmpnleuw %xmm1, %xmm0, %k1 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x3e,0xc9,0x06]
; X64-NEXT:    kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
; X64-NEXT:    vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
; X64-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
; X64-NEXT:    kmovd %k2, %eax # encoding: [0xc5,0xfb,0x93,0xc2]
; X64-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
; X64-NEXT:    kmovd %k3, %eax # encoding: [0xc5,0xfb,0x93,0xc3]
; X64-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
; X64-NEXT:    kmovd %k4, %eax # encoding: [0xc5,0xfb,0x93,0xc4]
; X64-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
; X64-NEXT:    kmovd %k5, %eax # encoding: [0xc5,0xfb,0x93,0xc5]
; X64-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
; X64-NEXT:    kmovd %k1, %eax # encoding: [0xc5,0xfb,0x93,0xc1]
; X64-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
; X64-NEXT:    vpinsrw $7, %edi, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc7,0x07]
; X64-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

define <16 x i8>@mm_mask_avg_epu8(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %x2, i16 %x3) {
; X86-LABEL: mm_mask_avg_epu8:
; X86:       # %bb.0:
; X86-NEXT:    vpavgb %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xe0,0xd9]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpavgb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe0,0xd1]
; X86-NEXT:    vpaddb %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: mm_mask_avg_epu8:
; X64:       # %bb.0:
; X64-NEXT:    vpavgb %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xe0,0xd9]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpavgb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe0,0xd1]
; X64-NEXT:    vpaddb %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc3]
; X64-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 <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) {
; X86-LABEL: test_int_x86_avx512_mask_pabs_b_128:
; X86:       # %bb.0:
; X86-NEXT:    vpabsb %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x1c,0xd0]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpabsb %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x1c,0xc8]
; X86-NEXT:    vpaddb %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfc,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pabs_b_128:
; X64:       # %bb.0:
; X64-NEXT:    vpabsb %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x1c,0xd0]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpabsb %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x1c,0xc8]
; X64-NEXT:    vpaddb %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfc,0xc2]
; X64-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 <16 x i8> @llvm.x86.avx512.mask.pavg.b.128(<16 x i8>, <16 x i8>, <16 x i8>, i16)

define <32 x i8>@mm256_mask_avg_epu8(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 %x3) {
; X86-LABEL: mm256_mask_avg_epu8:
; X86:       # %bb.0:
; X86-NEXT:    vpavgb %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xe0,0xd9]
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpavgb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe0,0xd1]
; X86-NEXT:    vpaddb %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: mm256_mask_avg_epu8:
; X64:       # %bb.0:
; X64-NEXT:    vpavgb %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xe0,0xd9]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpavgb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe0,0xd1]
; X64-NEXT:    vpaddb %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc3]
; X64-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 <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) {
; X86-LABEL: test_int_x86_avx512_mask_pabs_b_256:
; X86:       # %bb.0:
; X86-NEXT:    vpabsb %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x1c,0xd0]
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpabsb %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x1c,0xc8]
; X86-NEXT:    vpaddb %ymm2, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfc,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pabs_b_256:
; X64:       # %bb.0:
; X64-NEXT:    vpabsb %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x1c,0xd0]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpabsb %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x1c,0xc8]
; X64-NEXT:    vpaddb %ymm2, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfc,0xc2]
; X64-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 <32 x i8> @llvm.x86.avx512.mask.pavg.b.256(<32 x i8>, <32 x i8>, <32 x i8>, i32)

define <8 x i16>@mm_mask_avg_epu16(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3) {
; X86-LABEL: mm_mask_avg_epu16:
; X86:       # %bb.0:
; X86-NEXT:    vpavgw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xe3,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpavgw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe3,0xd1]
; X86-NEXT:    vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: mm_mask_avg_epu16:
; X64:       # %bb.0:
; X64-NEXT:    vpavgw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xe3,0xd9]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpavgw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe3,0xd1]
; X64-NEXT:    vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
; X64-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 <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) {
; X86-LABEL: test_int_x86_avx512_mask_pabs_w_128:
; X86:       # %bb.0:
; X86-NEXT:    vpabsw %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x1d,0xd0]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpabsw %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x1d,0xc8]
; X86-NEXT:    vpaddw %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pabs_w_128:
; X64:       # %bb.0:
; X64-NEXT:    vpabsw %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x1d,0xd0]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpabsw %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x1d,0xc8]
; X64-NEXT:    vpaddw %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc2]
; X64-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 <8 x i16> @llvm.x86.avx512.mask.pavg.w.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)

define <16 x i16>@mm256_mask_avg_epu16(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %x3) {
; X86-LABEL: mm256_mask_avg_epu16:
; X86:       # %bb.0:
; X86-NEXT:    vpavgw %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xe3,0xd9]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpavgw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe3,0xd1]
; X86-NEXT:    vpaddw %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: mm256_mask_avg_epu16:
; X64:       # %bb.0:
; X64-NEXT:    vpavgw %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xe3,0xd9]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpavgw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe3,0xd1]
; X64-NEXT:    vpaddw %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc3]
; X64-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 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) {
; X86-LABEL: test_int_x86_avx512_mask_pabs_w_256:
; X86:       # %bb.0:
; X86-NEXT:    vpabsw %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x1d,0xd0]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpabsw %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x1d,0xc8]
; X86-NEXT:    vpaddw %ymm2, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pabs_w_256:
; X64:       # %bb.0:
; X64-NEXT:    vpabsw %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x1d,0xd0]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpabsw %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x1d,0xc8]
; X64-NEXT:    vpaddw %ymm2, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc2]
; X64-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 <16 x i16> @llvm.x86.avx512.mask.pavg.w.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)

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) {
; X86-LABEL: test_int_x86_avx512_ptestm_b_128:
; X86:       # %bb.0:
; X86-NEXT:    vptestmb %xmm1, %xmm0, %k0 # encoding: [0x62,0xf2,0x7d,0x08,0x26,0xc1]
; X86-NEXT:    kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
; X86-NEXT:    movzwl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb7,0x44,0x24,0x04]
; X86-NEXT:    andw %cx, %ax # encoding: [0x66,0x21,0xc8]
; X86-NEXT:    addl %ecx, %eax # encoding: [0x01,0xc8]
; X86-NEXT:    # kill: def $ax killed $ax killed $eax
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_ptestm_b_128:
; X64:       # %bb.0:
; X64-NEXT:    vptestmb %xmm1, %xmm0, %k0 # encoding: [0x62,0xf2,0x7d,0x08,0x26,0xc1]
; X64-NEXT:    kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
; X64-NEXT:    andl %eax, %edi # encoding: [0x21,0xc7]
; X64-NEXT:    addl %edi, %eax # encoding: [0x01,0xf8]
; X64-NEXT:    # kill: def $ax killed $ax killed $eax
; X64-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) {
; X86-LABEL: test_int_x86_avx512_ptestm_b_256:
; X86:       # %bb.0:
; X86-NEXT:    vptestmb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf2,0x7d,0x28,0x26,0xc1]
; X86-NEXT:    kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    andl %ecx, %eax # encoding: [0x21,0xc8]
; X86-NEXT:    addl %ecx, %eax # encoding: [0x01,0xc8]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_ptestm_b_256:
; X64:       # %bb.0:
; X64-NEXT:    vptestmb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf2,0x7d,0x28,0x26,0xc1]
; X64-NEXT:    kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
; X64-NEXT:    andl %eax, %edi # encoding: [0x21,0xc7]
; X64-NEXT:    addl %edi, %eax # encoding: [0x01,0xf8]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-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) {
; X86-LABEL: test_int_x86_avx512_ptestm_w_128:
; X86:       # %bb.0:
; X86-NEXT:    vptestmw %xmm1, %xmm0, %k0 # encoding: [0x62,0xf2,0xfd,0x08,0x26,0xc1]
; X86-NEXT:    kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
; X86-NEXT:    movb {{[0-9]+}}(%esp), %al # encoding: [0x8a,0x44,0x24,0x04]
; X86-NEXT:    andb %cl, %al # encoding: [0x20,0xc8]
; X86-NEXT:    addb %cl, %al # encoding: [0x00,0xc8]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_ptestm_w_128:
; X64:       # %bb.0:
; X64-NEXT:    vptestmw %xmm1, %xmm0, %k0 # encoding: [0x62,0xf2,0xfd,0x08,0x26,0xc1]
; X64-NEXT:    kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
; X64-NEXT:    andb %al, %dil # encoding: [0x40,0x20,0xc7]
; X64-NEXT:    addb %dil, %al # encoding: [0x40,0x00,0xf8]
; X64-NEXT:    # kill: def $al killed $al killed $eax
; X64-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) {
; X86-LABEL: test_int_x86_avx512_ptestm_w_256:
; X86:       # %bb.0:
; X86-NEXT:    vptestmw %ymm1, %ymm0, %k0 # encoding: [0x62,0xf2,0xfd,0x28,0x26,0xc1]
; X86-NEXT:    kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
; X86-NEXT:    movzwl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb7,0x44,0x24,0x04]
; X86-NEXT:    andw %cx, %ax # encoding: [0x66,0x21,0xc8]
; X86-NEXT:    addl %ecx, %eax # encoding: [0x01,0xc8]
; X86-NEXT:    # kill: def $ax killed $ax killed $eax
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_ptestm_w_256:
; X64:       # %bb.0:
; X64-NEXT:    vptestmw %ymm1, %ymm0, %k0 # encoding: [0x62,0xf2,0xfd,0x28,0x26,0xc1]
; X64-NEXT:    kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
; X64-NEXT:    andl %eax, %edi # encoding: [0x21,0xc7]
; X64-NEXT:    addl %edi, %eax # encoding: [0x01,0xf8]
; X64-NEXT:    # kill: def $ax killed $ax killed $eax
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-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) {
; X86-LABEL: test_int_x86_avx512_ptestnm_b_128:
; X86:       # %bb.0:
; X86-NEXT:    vptestnmb %xmm1, %xmm0, %k0 # encoding: [0x62,0xf2,0x7e,0x08,0x26,0xc1]
; X86-NEXT:    kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
; X86-NEXT:    movzwl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb7,0x44,0x24,0x04]
; X86-NEXT:    andw %cx, %ax # encoding: [0x66,0x21,0xc8]
; X86-NEXT:    addl %ecx, %eax # encoding: [0x01,0xc8]
; X86-NEXT:    # kill: def $ax killed $ax killed $eax
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_ptestnm_b_128:
; X64:       # %bb.0:
; X64-NEXT:    vptestnmb %xmm1, %xmm0, %k0 # encoding: [0x62,0xf2,0x7e,0x08,0x26,0xc1]
; X64-NEXT:    kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
; X64-NEXT:    andl %eax, %edi # encoding: [0x21,0xc7]
; X64-NEXT:    addl %edi, %eax # encoding: [0x01,0xf8]
; X64-NEXT:    # kill: def $ax killed $ax killed $eax
; X64-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) {
; X86-LABEL: test_int_x86_avx512_ptestnm_b_256:
; X86:       # %bb.0:
; X86-NEXT:    vptestnmb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf2,0x7e,0x28,0x26,0xc1]
; X86-NEXT:    kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    andl %ecx, %eax # encoding: [0x21,0xc8]
; X86-NEXT:    addl %ecx, %eax # encoding: [0x01,0xc8]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_ptestnm_b_256:
; X64:       # %bb.0:
; X64-NEXT:    vptestnmb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf2,0x7e,0x28,0x26,0xc1]
; X64-NEXT:    kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
; X64-NEXT:    andl %eax, %edi # encoding: [0x21,0xc7]
; X64-NEXT:    addl %edi, %eax # encoding: [0x01,0xf8]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-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) {
; X86-LABEL: test_int_x86_avx512_ptestnm_w_128:
; X86:       # %bb.0:
; X86-NEXT:    vptestnmw %xmm1, %xmm0, %k0 # encoding: [0x62,0xf2,0xfe,0x08,0x26,0xc1]
; X86-NEXT:    kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
; X86-NEXT:    movb {{[0-9]+}}(%esp), %al # encoding: [0x8a,0x44,0x24,0x04]
; X86-NEXT:    andb %cl, %al # encoding: [0x20,0xc8]
; X86-NEXT:    addb %cl, %al # encoding: [0x00,0xc8]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_ptestnm_w_128:
; X64:       # %bb.0:
; X64-NEXT:    vptestnmw %xmm1, %xmm0, %k0 # encoding: [0x62,0xf2,0xfe,0x08,0x26,0xc1]
; X64-NEXT:    kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
; X64-NEXT:    andb %al, %dil # encoding: [0x40,0x20,0xc7]
; X64-NEXT:    addb %dil, %al # encoding: [0x40,0x00,0xf8]
; X64-NEXT:    # kill: def $al killed $al killed $eax
; X64-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) {
; X86-LABEL: test_int_x86_avx512_ptestnm_w_256:
; X86:       # %bb.0:
; X86-NEXT:    vptestnmw %ymm1, %ymm0, %k0 # encoding: [0x62,0xf2,0xfe,0x28,0x26,0xc1]
; X86-NEXT:    kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
; X86-NEXT:    movzwl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb7,0x44,0x24,0x04]
; X86-NEXT:    andw %cx, %ax # encoding: [0x66,0x21,0xc8]
; X86-NEXT:    addl %ecx, %eax # encoding: [0x01,0xc8]
; X86-NEXT:    # kill: def $ax killed $ax killed $eax
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_ptestnm_w_256:
; X64:       # %bb.0:
; X64-NEXT:    vptestnmw %ymm1, %ymm0, %k0 # encoding: [0x62,0xf2,0xfe,0x28,0x26,0xc1]
; X64-NEXT:    kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
; X64-NEXT:    andl %eax, %edi # encoding: [0x21,0xc7]
; X64-NEXT:    addl %edi, %eax # encoding: [0x01,0xf8]
; X64-NEXT:    # kill: def $ax killed $ax killed $eax
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-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 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:    kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
; CHECK-NEXT:    # kill: def $ax killed $ax killed $eax
; CHECK-NEXT:    ret{{[l|q]}} # 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:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; CHECK-NEXT:    ret{{[l|q]}} # 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:    kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
; CHECK-NEXT:    # kill: def $al killed $al killed $eax
; CHECK-NEXT:    ret{{[l|q]}} # 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:    kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
; CHECK-NEXT:    # kill: def $ax killed $ax killed $eax
; CHECK-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
    %res = call i16 @llvm.x86.avx512.cvtw2mask.256(<16 x i16> %x0)
    ret i16 %res
}

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) {
; X86-LABEL: test_int_x86_avx512_mask_pmulhu_w_128:
; X86:       # %bb.0:
; X86-NEXT:    vpmulhuw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xe4,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpmulhuw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe4,0xd1]
; X86-NEXT:    vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmulhu_w_128:
; X64:       # %bb.0:
; X64-NEXT:    vpmulhuw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xe4,0xd9]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpmulhuw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe4,0xd1]
; X64-NEXT:    vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
; X64-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) {
; X86-LABEL: test_int_x86_avx512_mask_pmulhu_w_256:
; X86:       # %bb.0:
; X86-NEXT:    vpmulhuw %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xe4,0xd9]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpmulhuw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe4,0xd1]
; X86-NEXT:    vpaddw %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmulhu_w_256:
; X64:       # %bb.0:
; X64-NEXT:    vpmulhuw %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xe4,0xd9]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpmulhuw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe4,0xd1]
; X64-NEXT:    vpaddw %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc3]
; X64-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) {
; X86-LABEL: test_int_x86_avx512_mask_pmulh_w_128:
; X86:       # %bb.0:
; X86-NEXT:    vpmulhw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xe5,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpmulhw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe5,0xd1]
; X86-NEXT:    vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmulh_w_128:
; X64:       # %bb.0:
; X64-NEXT:    vpmulhw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xe5,0xd9]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpmulhw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe5,0xd1]
; X64-NEXT:    vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
; X64-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) {
; X86-LABEL: test_int_x86_avx512_mask_pmulh_w_256:
; X86:       # %bb.0:
; X86-NEXT:    vpmulhw %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xe5,0xd9]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpmulhw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe5,0xd1]
; X86-NEXT:    vpaddw %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmulh_w_256:
; X64:       # %bb.0:
; X64-NEXT:    vpmulhw %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xe5,0xd9]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpmulhw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe5,0xd1]
; X64-NEXT:    vpaddw %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc3]
; X64-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) {
; X86-LABEL: test_int_x86_avx512_mask_pmulhr_sw_128:
; X86:       # %bb.0:
; X86-NEXT:    vpmulhrsw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x0b,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpmulhrsw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x0b,0xd1]
; X86-NEXT:    vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmulhr_sw_128:
; X64:       # %bb.0:
; X64-NEXT:    vpmulhrsw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x0b,0xd9]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpmulhrsw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x0b,0xd1]
; X64-NEXT:    vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
; X64-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) {
; X86-LABEL: test_int_x86_avx512_mask_pmulhr_sw_256:
; X86:       # %bb.0:
; X86-NEXT:    vpmulhrsw %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x0b,0xd9]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpmulhrsw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x0b,0xd1]
; X86-NEXT:    vpaddw %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmulhr_sw_256:
; X64:       # %bb.0:
; X64-NEXT:    vpmulhrsw %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x0b,0xd9]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpmulhrsw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x0b,0xd1]
; X64-NEXT:    vpaddw %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc3]
; X64-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 <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) {
; X86-LABEL: test_int_x86_avx512_mask_pmaddubs_w_128:
; X86:       # %bb.0:
; X86-NEXT:    vpmaddubsw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x04,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpmaddubsw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x04,0xd1]
; X86-NEXT:    vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmaddubs_w_128:
; X64:       # %bb.0:
; X64-NEXT:    vpmaddubsw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x04,0xd9]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpmaddubsw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x04,0xd1]
; X64-NEXT:    vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
; X64-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) {
; X86-LABEL: test_int_x86_avx512_mask_pmaddubs_w_256:
; X86:       # %bb.0:
; X86-NEXT:    vpmaddubsw %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x04,0xd9]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpmaddubsw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x04,0xd1]
; X86-NEXT:    vpaddw %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmaddubs_w_256:
; X64:       # %bb.0:
; X64-NEXT:    vpmaddubsw %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x04,0xd9]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpmaddubsw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x04,0xd1]
; X64-NEXT:    vpaddw %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc3]
; X64-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 <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) {
; X86-LABEL: test_int_x86_avx512_mask_pmaddw_d_128:
; X86:       # %bb.0:
; X86-NEXT:    vpmaddwd %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xf5,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpmaddwd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xf5,0xd1]
; X86-NEXT:    vpaddd %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmaddw_d_128:
; X64:       # %bb.0:
; X64-NEXT:    vpmaddwd %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xf5,0xd9]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpmaddwd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xf5,0xd1]
; X64-NEXT:    vpaddd %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc3]
; X64-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) {
; X86-LABEL: test_int_x86_avx512_mask_pmaddw_d_256:
; X86:       # %bb.0:
; X86-NEXT:    vpmaddwd %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xf5,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpmaddwd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xf5,0xd1]
; X86-NEXT:    vpaddd %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfe,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmaddw_d_256:
; X64:       # %bb.0:
; X64-NEXT:    vpmaddwd %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xf5,0xd9]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpmaddwd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xf5,0xd1]
; X64-NEXT:    vpaddd %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfe,0xc3]
; X64-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.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) {
; X86-LABEL: test_int_x86_avx512_mask_permvar_hi_128:
; X86:       # %bb.0:
; X86-NEXT:    vpermw %xmm0, %xmm1, %xmm3 # encoding: [0x62,0xf2,0xf5,0x08,0x8d,0xd8]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpermw %xmm0, %xmm1, %xmm2 {%k1} # encoding: [0x62,0xf2,0xf5,0x09,0x8d,0xd0]
; X86-NEXT:    vpermw %xmm0, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0x89,0x8d,0xc0]
; X86-NEXT:    vpaddw %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc3]
; X86-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_permvar_hi_128:
; X64:       # %bb.0:
; X64-NEXT:    vpermw %xmm0, %xmm1, %xmm3 # encoding: [0x62,0xf2,0xf5,0x08,0x8d,0xd8]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpermw %xmm0, %xmm1, %xmm2 {%k1} # encoding: [0x62,0xf2,0xf5,0x09,0x8d,0xd0]
; X64-NEXT:    vpermw %xmm0, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0x89,0x8d,0xc0]
; X64-NEXT:    vpaddw %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc3]
; X64-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
; X64-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) {
; X86-LABEL: test_int_x86_avx512_mask_permvar_hi_256:
; X86:       # %bb.0:
; X86-NEXT:    vpermw %ymm0, %ymm1, %ymm3 # encoding: [0x62,0xf2,0xf5,0x28,0x8d,0xd8]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpermw %ymm0, %ymm1, %ymm2 {%k1} # encoding: [0x62,0xf2,0xf5,0x29,0x8d,0xd0]
; X86-NEXT:    vpermw %ymm0, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0xa9,0x8d,0xc0]
; X86-NEXT:    vpaddw %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc3]
; X86-NEXT:    vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_permvar_hi_256:
; X64:       # %bb.0:
; X64-NEXT:    vpermw %ymm0, %ymm1, %ymm3 # encoding: [0x62,0xf2,0xf5,0x28,0x8d,0xd8]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpermw %ymm0, %ymm1, %ymm2 {%k1} # encoding: [0x62,0xf2,0xf5,0x29,0x8d,0xd0]
; X64-NEXT:    vpermw %ymm0, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0xa9,0x8d,0xc0]
; X64-NEXT:    vpaddw %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc3]
; X64-NEXT:    vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
; X64-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 <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) {
; X86-LABEL: test_int_x86_avx512_mask_vpermt2var_hi_128:
; X86:       # %bb.0:
; X86-NEXT:    vmovdqa %xmm1, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd9]
; X86-NEXT:    vpermt2w %xmm2, %xmm0, %xmm3 # encoding: [0x62,0xf2,0xfd,0x08,0x7d,0xda]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpermt2w %xmm2, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x7d,0xca]
; X86-NEXT:    vpaddw %xmm3, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_vpermt2var_hi_128:
; X64:       # %bb.0:
; X64-NEXT:    vmovdqa %xmm1, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd9]
; X64-NEXT:    vpermt2w %xmm2, %xmm0, %xmm3 # encoding: [0x62,0xf2,0xfd,0x08,0x7d,0xda]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpermt2w %xmm2, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x7d,0xca]
; X64-NEXT:    vpaddw %xmm3, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc3]
; X64-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) {
; X86-LABEL: test_int_x86_avx512_maskz_vpermt2var_hi_128:
; X86:       # %bb.0:
; X86-NEXT:    vmovdqa %xmm1, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd9]
; X86-NEXT:    vpermt2w %xmm2, %xmm0, %xmm3 # encoding: [0x62,0xf2,0xfd,0x08,0x7d,0xda]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpermt2w %xmm2, %xmm0, %xmm1 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x7d,0xca]
; X86-NEXT:    vpaddw %xmm3, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_maskz_vpermt2var_hi_128:
; X64:       # %bb.0:
; X64-NEXT:    vmovdqa %xmm1, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd9]
; X64-NEXT:    vpermt2w %xmm2, %xmm0, %xmm3 # encoding: [0x62,0xf2,0xfd,0x08,0x7d,0xda]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpermt2w %xmm2, %xmm0, %xmm1 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x7d,0xca]
; X64-NEXT:    vpaddw %xmm3, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc3]
; X64-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) {
; X86-LABEL: test_int_x86_avx512_mask_vpermt2var_hi_256:
; X86:       # %bb.0:
; X86-NEXT:    vmovdqa %ymm1, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd9]
; X86-NEXT:    vpermt2w %ymm2, %ymm0, %ymm3 # encoding: [0x62,0xf2,0xfd,0x28,0x7d,0xda]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpermt2w %ymm2, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x7d,0xca]
; X86-NEXT:    vpaddw %ymm3, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_vpermt2var_hi_256:
; X64:       # %bb.0:
; X64-NEXT:    vmovdqa %ymm1, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd9]
; X64-NEXT:    vpermt2w %ymm2, %ymm0, %ymm3 # encoding: [0x62,0xf2,0xfd,0x28,0x7d,0xda]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpermt2w %ymm2, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x7d,0xca]
; X64-NEXT:    vpaddw %ymm3, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc3]
; X64-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) {
; X86-LABEL: test_int_x86_avx512_maskz_vpermt2var_hi_256:
; X86:       # %bb.0:
; X86-NEXT:    vmovdqa %ymm1, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd9]
; X86-NEXT:    vpermt2w %ymm2, %ymm0, %ymm3 # encoding: [0x62,0xf2,0xfd,0x28,0x7d,0xda]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpermt2w %ymm2, %ymm0, %ymm1 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x7d,0xca]
; X86-NEXT:    vpaddw %ymm3, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_maskz_vpermt2var_hi_256:
; X64:       # %bb.0:
; X64-NEXT:    vmovdqa %ymm1, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd9]
; X64-NEXT:    vpermt2w %ymm2, %ymm0, %ymm3 # encoding: [0x62,0xf2,0xfd,0x28,0x7d,0xda]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpermt2w %ymm2, %ymm0, %ymm1 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x7d,0xca]
; X64-NEXT:    vpaddw %ymm3, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc3]
; X64-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) {
; X86-LABEL: test_int_x86_avx512_mask_vpermi2var_hi_128:
; X86:       # %bb.0:
; X86-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
; X86-NEXT:    vpermt2w %xmm2, %xmm1, %xmm3 # encoding: [0x62,0xf2,0xf5,0x08,0x7d,0xda]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpermi2w %xmm2, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x75,0xca]
; X86-NEXT:    vpaddw %xmm3, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_vpermi2var_hi_128:
; X64:       # %bb.0:
; X64-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
; X64-NEXT:    vpermt2w %xmm2, %xmm1, %xmm3 # encoding: [0x62,0xf2,0xf5,0x08,0x7d,0xda]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpermi2w %xmm2, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x75,0xca]
; X64-NEXT:    vpaddw %xmm3, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc3]
; X64-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) {
; X86-LABEL: test_int_x86_avx512_mask_vpermi2var_hi_256:
; X86:       # %bb.0:
; X86-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
; X86-NEXT:    vpermt2w %ymm2, %ymm1, %ymm3 # encoding: [0x62,0xf2,0xf5,0x28,0x7d,0xda]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpermi2w %ymm2, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x75,0xca]
; X86-NEXT:    vpaddw %ymm3, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_vpermi2var_hi_256:
; X64:       # %bb.0:
; X64-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
; X64-NEXT:    vpermt2w %ymm2, %ymm1, %ymm3 # encoding: [0x62,0xf2,0xf5,0x28,0x7d,0xda]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpermi2w %ymm2, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x75,0xca]
; X64-NEXT:    vpaddw %ymm3, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc3]
; X64-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 <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) {
; X86-LABEL: test_int_x86_avx512_mask_dbpsadbw_128:
; X86:       # %bb.0:
; X86-NEXT:    vdbpsadbw $2, %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf3,0x7d,0x08,0x42,0xd9,0x02]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vdbpsadbw $2, %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x42,0xd1,0x02]
; X86-NEXT:    vdbpsadbw $2, %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0x89,0x42,0xc1,0x02]
; X86-NEXT:    vpaddw %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc3]
; X86-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_dbpsadbw_128:
; X64:       # %bb.0:
; X64-NEXT:    vdbpsadbw $2, %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf3,0x7d,0x08,0x42,0xd9,0x02]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vdbpsadbw $2, %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x42,0xd1,0x02]
; X64-NEXT:    vdbpsadbw $2, %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0x89,0x42,0xc1,0x02]
; X64-NEXT:    vpaddw %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc3]
; X64-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
; X64-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) {
; X86-LABEL: test_int_x86_avx512_mask_dbpsadbw_256:
; X86:       # %bb.0:
; X86-NEXT:    vdbpsadbw $2, %ymm1, %ymm0, %ymm3 # encoding: [0x62,0xf3,0x7d,0x28,0x42,0xd9,0x02]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vdbpsadbw $2, %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x42,0xd1,0x02]
; X86-NEXT:    vdbpsadbw $2, %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0xa9,0x42,0xc1,0x02]
; X86-NEXT:    vpaddw %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc3]
; X86-NEXT:    vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_dbpsadbw_256:
; X64:       # %bb.0:
; X64-NEXT:    vdbpsadbw $2, %ymm1, %ymm0, %ymm3 # encoding: [0x62,0xf3,0x7d,0x28,0x42,0xd9,0x02]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vdbpsadbw $2, %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x42,0xd1,0x02]
; X64-NEXT:    vdbpsadbw $2, %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0xa9,0x42,0xc1,0x02]
; X64-NEXT:    vpaddw %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc3]
; X64-NEXT:    vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
; X64-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
}

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 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xdd,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # 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) {
; X86-LABEL: test_mask_adds_epu16_rrk_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpaddusw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xdd,0xd1]
; X86-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_adds_epu16_rrk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpaddusw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xdd,0xd1]
; X64-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
; X64-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) {
; X86-LABEL: test_mask_adds_epu16_rrkz_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpaddusw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xdd,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_adds_epu16_rrkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpaddusw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xdd,0xc1]
; X64-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) {
; X86-LABEL: test_mask_adds_epu16_rm_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpaddusw (%eax), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xdd,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_adds_epu16_rm_128:
; X64:       # %bb.0:
; X64-NEXT:    vpaddusw (%rdi), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xdd,0x07]
; X64-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) {
; X86-LABEL: test_mask_adds_epu16_rmk_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
; X86-NEXT:    kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
; X86-NEXT:    vpaddusw (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xdd,0x08]
; X86-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_adds_epu16_rmk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpaddusw (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xdd,0x0f]
; X64-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X64-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) {
; X86-LABEL: test_mask_adds_epu16_rmkz_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
; X86-NEXT:    kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
; X86-NEXT:    vpaddusw (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xdd,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_adds_epu16_rmkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpaddusw (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xdd,0x07]
; X64-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 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xdd,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # 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) {
; X86-LABEL: test_mask_adds_epu16_rrk_256:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpaddusw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xdd,0xd1]
; X86-NEXT:    vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_adds_epu16_rrk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpaddusw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xdd,0xd1]
; X64-NEXT:    vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
; X64-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) {
; X86-LABEL: test_mask_adds_epu16_rrkz_256:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpaddusw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xdd,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_adds_epu16_rrkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpaddusw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xdd,0xc1]
; X64-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) {
; X86-LABEL: test_mask_adds_epu16_rm_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpaddusw (%eax), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xdd,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_adds_epu16_rm_256:
; X64:       # %bb.0:
; X64-NEXT:    vpaddusw (%rdi), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xdd,0x07]
; X64-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) {
; X86-LABEL: test_mask_adds_epu16_rmk_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpaddusw (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xdd,0x08]
; X86-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_adds_epu16_rmk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpaddusw (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xdd,0x0f]
; X64-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X64-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) {
; X86-LABEL: test_mask_adds_epu16_rmkz_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpaddusw (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xdd,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_adds_epu16_rmkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpaddusw (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xdd,0x07]
; X64-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 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd9,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # 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) {
; X86-LABEL: test_mask_subs_epu16_rrk_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpsubusw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xd9,0xd1]
; X86-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_subs_epu16_rrk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpsubusw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xd9,0xd1]
; X64-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
; X64-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) {
; X86-LABEL: test_mask_subs_epu16_rrkz_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpsubusw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xd9,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_subs_epu16_rrkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpsubusw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xd9,0xc1]
; X64-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) {
; X86-LABEL: test_mask_subs_epu16_rm_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpsubusw (%eax), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd9,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_subs_epu16_rm_128:
; X64:       # %bb.0:
; X64-NEXT:    vpsubusw (%rdi), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd9,0x07]
; X64-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) {
; X86-LABEL: test_mask_subs_epu16_rmk_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
; X86-NEXT:    kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
; X86-NEXT:    vpsubusw (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xd9,0x08]
; X86-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_subs_epu16_rmk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpsubusw (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xd9,0x0f]
; X64-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X64-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) {
; X86-LABEL: test_mask_subs_epu16_rmkz_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
; X86-NEXT:    kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
; X86-NEXT:    vpsubusw (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xd9,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_subs_epu16_rmkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpsubusw (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xd9,0x07]
; X64-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 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd9,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # 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) {
; X86-LABEL: test_mask_subs_epu16_rrk_256:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpsubusw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xd9,0xd1]
; X86-NEXT:    vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_subs_epu16_rrk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpsubusw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xd9,0xd1]
; X64-NEXT:    vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
; X64-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) {
; X86-LABEL: test_mask_subs_epu16_rrkz_256:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpsubusw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xd9,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_subs_epu16_rrkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpsubusw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xd9,0xc1]
; X64-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) {
; X86-LABEL: test_mask_subs_epu16_rm_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpsubusw (%eax), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd9,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_subs_epu16_rm_256:
; X64:       # %bb.0:
; X64-NEXT:    vpsubusw (%rdi), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd9,0x07]
; X64-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) {
; X86-LABEL: test_mask_subs_epu16_rmk_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpsubusw (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xd9,0x08]
; X86-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_subs_epu16_rmk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpsubusw (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xd9,0x0f]
; X64-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X64-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) {
; X86-LABEL: test_mask_subs_epu16_rmkz_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpsubusw (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xd9,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_subs_epu16_rmkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpsubusw (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xd9,0x07]
; X64-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_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 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xdc,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # 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) {
; X86-LABEL: test_mask_adds_epu8_rrk_128:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpaddusb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xdc,0xd1]
; X86-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_adds_epu8_rrk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpaddusb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xdc,0xd1]
; X64-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
; X64-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) {
; X86-LABEL: test_mask_adds_epu8_rrkz_128:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpaddusb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xdc,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_adds_epu8_rrkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpaddusb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xdc,0xc1]
; X64-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) {
; X86-LABEL: test_mask_adds_epu8_rm_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpaddusb (%eax), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xdc,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_adds_epu8_rm_128:
; X64:       # %bb.0:
; X64-NEXT:    vpaddusb (%rdi), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xdc,0x07]
; X64-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) {
; X86-LABEL: test_mask_adds_epu8_rmk_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpaddusb (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xdc,0x08]
; X86-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_adds_epu8_rmk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpaddusb (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xdc,0x0f]
; X64-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X64-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) {
; X86-LABEL: test_mask_adds_epu8_rmkz_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpaddusb (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xdc,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_adds_epu8_rmkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpaddusb (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xdc,0x07]
; X64-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 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xdc,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # 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) {
; X86-LABEL: test_mask_adds_epu8_rrk_256:
; X86:       # %bb.0:
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpaddusb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xdc,0xd1]
; X86-NEXT:    vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_adds_epu8_rrk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpaddusb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xdc,0xd1]
; X64-NEXT:    vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
; X64-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) {
; X86-LABEL: test_mask_adds_epu8_rrkz_256:
; X86:       # %bb.0:
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpaddusb %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xdc,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_adds_epu8_rrkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpaddusb %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xdc,0xc1]
; X64-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) {
; X86-LABEL: test_mask_adds_epu8_rm_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpaddusb (%eax), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xdc,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_adds_epu8_rm_256:
; X64:       # %bb.0:
; X64-NEXT:    vpaddusb (%rdi), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xdc,0x07]
; X64-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) {
; X86-LABEL: test_mask_adds_epu8_rmk_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpaddusb (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xdc,0x08]
; X86-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_adds_epu8_rmk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpaddusb (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xdc,0x0f]
; X64-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X64-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) {
; X86-LABEL: test_mask_adds_epu8_rmkz_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpaddusb (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xdc,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_adds_epu8_rmkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpaddusb (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xdc,0x07]
; X64-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 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd8,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # 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) {
; X86-LABEL: test_mask_subs_epu8_rrk_128:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpsubusb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xd8,0xd1]
; X86-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_subs_epu8_rrk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpsubusb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xd8,0xd1]
; X64-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
; X64-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) {
; X86-LABEL: test_mask_subs_epu8_rrkz_128:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpsubusb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xd8,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_subs_epu8_rrkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpsubusb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xd8,0xc1]
; X64-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) {
; X86-LABEL: test_mask_subs_epu8_rm_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpsubusb (%eax), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd8,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_subs_epu8_rm_128:
; X64:       # %bb.0:
; X64-NEXT:    vpsubusb (%rdi), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd8,0x07]
; X64-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) {
; X86-LABEL: test_mask_subs_epu8_rmk_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpsubusb (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xd8,0x08]
; X86-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_subs_epu8_rmk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpsubusb (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xd8,0x0f]
; X64-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X64-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) {
; X86-LABEL: test_mask_subs_epu8_rmkz_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpsubusb (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xd8,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_subs_epu8_rmkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpsubusb (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xd8,0x07]
; X64-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 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd8,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # 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) {
; X86-LABEL: test_mask_subs_epu8_rrk_256:
; X86:       # %bb.0:
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpsubusb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xd8,0xd1]
; X86-NEXT:    vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_subs_epu8_rrk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpsubusb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xd8,0xd1]
; X64-NEXT:    vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
; X64-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) {
; X86-LABEL: test_mask_subs_epu8_rrkz_256:
; X86:       # %bb.0:
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpsubusb %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xd8,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_subs_epu8_rrkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpsubusb %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xd8,0xc1]
; X64-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) {
; X86-LABEL: test_mask_subs_epu8_rm_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpsubusb (%eax), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd8,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_subs_epu8_rm_256:
; X64:       # %bb.0:
; X64-NEXT:    vpsubusb (%rdi), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd8,0x07]
; X64-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) {
; X86-LABEL: test_mask_subs_epu8_rmk_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpsubusb (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xd8,0x08]
; X86-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_subs_epu8_rmk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpsubusb (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xd8,0x0f]
; X64-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X64-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) {
; X86-LABEL: test_mask_subs_epu8_rmkz_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpsubusb (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xd8,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_subs_epu8_rmkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpsubusb (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xd8,0x07]
; X64-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)

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 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xed,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # 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) {
; X86-LABEL: test_mask_adds_epi16_rrk_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpaddsw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xed,0xd1]
; X86-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_adds_epi16_rrk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpaddsw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xed,0xd1]
; X64-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
; X64-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) {
; X86-LABEL: test_mask_adds_epi16_rrkz_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpaddsw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xed,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_adds_epi16_rrkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpaddsw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xed,0xc1]
; X64-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) {
; X86-LABEL: test_mask_adds_epi16_rm_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpaddsw (%eax), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xed,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_adds_epi16_rm_128:
; X64:       # %bb.0:
; X64-NEXT:    vpaddsw (%rdi), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xed,0x07]
; X64-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) {
; X86-LABEL: test_mask_adds_epi16_rmk_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
; X86-NEXT:    kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
; X86-NEXT:    vpaddsw (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xed,0x08]
; X86-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_adds_epi16_rmk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpaddsw (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xed,0x0f]
; X64-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X64-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) {
; X86-LABEL: test_mask_adds_epi16_rmkz_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
; X86-NEXT:    kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
; X86-NEXT:    vpaddsw (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xed,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_adds_epi16_rmkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpaddsw (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xed,0x07]
; X64-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 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xed,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # 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) {
; X86-LABEL: test_mask_adds_epi16_rrk_256:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpaddsw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xed,0xd1]
; X86-NEXT:    vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_adds_epi16_rrk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpaddsw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xed,0xd1]
; X64-NEXT:    vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
; X64-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) {
; X86-LABEL: test_mask_adds_epi16_rrkz_256:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpaddsw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xed,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_adds_epi16_rrkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpaddsw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xed,0xc1]
; X64-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) {
; X86-LABEL: test_mask_adds_epi16_rm_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpaddsw (%eax), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xed,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_adds_epi16_rm_256:
; X64:       # %bb.0:
; X64-NEXT:    vpaddsw (%rdi), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xed,0x07]
; X64-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) {
; X86-LABEL: test_mask_adds_epi16_rmk_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpaddsw (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xed,0x08]
; X86-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_adds_epi16_rmk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpaddsw (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xed,0x0f]
; X64-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X64-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) {
; X86-LABEL: test_mask_adds_epi16_rmkz_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpaddsw (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xed,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_adds_epi16_rmkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpaddsw (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xed,0x07]
; X64-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)

declare <8 x i16> @llvm.x86.sse2.psubs.w(<8 x i16>, <8 x i16>) #0

define <8 x i16> @test_test_subs_epi16_rrk_128(<8 x i16> %a, <8 x i16> %b, <8 x i16> %passThru, i8 %mask) {
; X86-LABEL: test_test_subs_epi16_rrk_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpsubsw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe9,0xd1]
; X86-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_test_subs_epi16_rrk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpsubsw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe9,0xd1]
; X64-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = call <8 x i16> @llvm.x86.sse2.psubs.w(<8 x i16> %a, <8 x i16> %b)
  %2 = bitcast i8 %mask to <8 x i1>
  %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> %passThru
  ret <8 x i16> %3
}

define <8 x i16> @test_test_subs_epi16_rrkz_128(<8 x i16> %a, <8 x i16> %b, i8 %mask) {
; X86-LABEL: test_test_subs_epi16_rrkz_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpsubsw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xe9,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_test_subs_epi16_rrkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpsubsw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xe9,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = call <8 x i16> @llvm.x86.sse2.psubs.w(<8 x i16> %a, <8 x i16> %b)
  %2 = bitcast i8 %mask to <8 x i1>
  %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> zeroinitializer
  ret <8 x i16> %3
}

define <8 x i16> @test_test_subs_epi16_rmk_128(<8 x i16> %a, <8 x i16>* %ptr_b, <8 x i16> %passThru, i8 %mask) {
; X86-LABEL: test_test_subs_epi16_rmk_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
; X86-NEXT:    kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
; X86-NEXT:    vpsubsw (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe9,0x08]
; X86-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_test_subs_epi16_rmk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpsubsw (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe9,0x0f]
; X64-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %b = load <8 x i16>, <8 x i16>* %ptr_b
  %1 = call <8 x i16> @llvm.x86.sse2.psubs.w(<8 x i16> %a, <8 x i16> %b)
  %2 = bitcast i8 %mask to <8 x i1>
  %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> %passThru
  ret <8 x i16> %3
}

define <8 x i16> @test_test_subs_epi16_rmkz_128(<8 x i16> %a, <8 x i16>* %ptr_b, i8 %mask) {
; X86-LABEL: test_test_subs_epi16_rmkz_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
; X86-NEXT:    kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
; X86-NEXT:    vpsubsw (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xe9,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_test_subs_epi16_rmkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpsubsw (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xe9,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %b = load <8 x i16>, <8 x i16>* %ptr_b
  %1 = call <8 x i16> @llvm.x86.sse2.psubs.w(<8 x i16> %a, <8 x i16> %b)
  %2 = bitcast i8 %mask to <8 x i1>
  %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> zeroinitializer
  ret <8 x i16> %3
}

declare <16 x i16> @llvm.x86.avx2.psubs.w(<16 x i16>, <16 x i16>) #0

define <16 x i16> @test_test_subs_epi16_rrk_256(<16 x i16> %a, <16 x i16> %b, <16 x i16> %passThru, i16 %mask) {
; X86-LABEL: test_test_subs_epi16_rrk_256:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpsubsw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe9,0xd1]
; X86-NEXT:    vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_test_subs_epi16_rrk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpsubsw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe9,0xd1]
; X64-NEXT:    vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = call <16 x i16> @llvm.x86.avx2.psubs.w(<16 x i16> %a, <16 x i16> %b)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> %passThru
  ret <16 x i16> %3
}

define <16 x i16> @test_test_subs_epi16_rrkz_256(<16 x i16> %a, <16 x i16> %b, i16 %mask) {
; X86-LABEL: test_test_subs_epi16_rrkz_256:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpsubsw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xe9,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_test_subs_epi16_rrkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpsubsw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xe9,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = call <16 x i16> @llvm.x86.avx2.psubs.w(<16 x i16> %a, <16 x i16> %b)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> zeroinitializer
  ret <16 x i16> %3
}

define <16 x i16> @test_test_subs_epi16_rmk_256(<16 x i16> %a, <16 x i16>* %ptr_b, <16 x i16> %passThru, i16 %mask) {
; X86-LABEL: test_test_subs_epi16_rmk_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpsubsw (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe9,0x08]
; X86-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_test_subs_epi16_rmk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpsubsw (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe9,0x0f]
; X64-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %b = load <16 x i16>, <16 x i16>* %ptr_b
  %1 = call <16 x i16> @llvm.x86.avx2.psubs.w(<16 x i16> %a, <16 x i16> %b)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> %passThru
  ret <16 x i16> %3
}

define <16 x i16> @test_test_subs_epi16_rmkz_256(<16 x i16> %a, <16 x i16>* %ptr_b, i16 %mask) {
; X86-LABEL: test_test_subs_epi16_rmkz_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpsubsw (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xe9,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_test_subs_epi16_rmkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpsubsw (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xe9,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %b = load <16 x i16>, <16 x i16>* %ptr_b
  %1 = call <16 x i16> @llvm.x86.avx2.psubs.w(<16 x i16> %a, <16 x i16> %b)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> zeroinitializer
  ret <16 x i16> %3
}

declare <16 x i8> @llvm.x86.sse2.psubs.b(<16 x i8>, <16 x i8>) #0

define <16 x i8> @test_test_subs_epi8_rrk_128(<16 x i8> %a, <16 x i8> %b, <16 x i8> %passThru, i16 %mask) {
; X86-LABEL: test_test_subs_epi8_rrk_128:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpsubsb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe8,0xd1]
; X86-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_test_subs_epi8_rrk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpsubsb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe8,0xd1]
; X64-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = call <16 x i8> @llvm.x86.sse2.psubs.b(<16 x i8> %a, <16 x i8> %b)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x i8> %1, <16 x i8> %passThru
  ret <16 x i8> %3
}

define <16 x i8> @test_test_subs_epi8_rrkz_128(<16 x i8> %a, <16 x i8> %b, i16 %mask) {
; X86-LABEL: test_test_subs_epi8_rrkz_128:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpsubsb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xe8,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_test_subs_epi8_rrkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpsubsb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xe8,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = call <16 x i8> @llvm.x86.sse2.psubs.b(<16 x i8> %a, <16 x i8> %b)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x i8> %1, <16 x i8> zeroinitializer
  ret <16 x i8> %3
}

define <16 x i8> @test_test_subs_epi8_rmk_128(<16 x i8> %a, <16 x i8>* %ptr_b, <16 x i8> %passThru, i16 %mask) {
; X86-LABEL: test_test_subs_epi8_rmk_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpsubsb (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe8,0x08]
; X86-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_test_subs_epi8_rmk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpsubsb (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe8,0x0f]
; X64-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %b = load <16 x i8>, <16 x i8>* %ptr_b
  %1 = call <16 x i8> @llvm.x86.sse2.psubs.b(<16 x i8> %a, <16 x i8> %b)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x i8> %1, <16 x i8> %passThru
  ret <16 x i8> %3
}

define <16 x i8> @test_test_subs_epi8_rmkz_128(<16 x i8> %a, <16 x i8>* %ptr_b, i16 %mask) {
; X86-LABEL: test_test_subs_epi8_rmkz_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpsubsb (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xe8,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_test_subs_epi8_rmkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpsubsb (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xe8,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %b = load <16 x i8>, <16 x i8>* %ptr_b
  %1 = call <16 x i8> @llvm.x86.sse2.psubs.b(<16 x i8> %a, <16 x i8> %b)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x i8> %1, <16 x i8> zeroinitializer
  ret <16 x i8> %3
}

declare <32 x i8> @llvm.x86.avx2.psubs.b(<32 x i8>, <32 x i8>) #0

define <32 x i8> @test_test_subs_epi8_rrk_256(<32 x i8> %a, <32 x i8> %b, <32 x i8> %passThru, i32 %mask) {
; X86-LABEL: test_test_subs_epi8_rrk_256:
; X86:       # %bb.0:
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpsubsb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe8,0xd1]
; X86-NEXT:    vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_test_subs_epi8_rrk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpsubsb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe8,0xd1]
; X64-NEXT:    vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = call <32 x i8> @llvm.x86.avx2.psubs.b(<32 x i8> %a, <32 x i8> %b)
  %2 = bitcast i32 %mask to <32 x i1>
  %3 = select <32 x i1> %2, <32 x i8> %1, <32 x i8> %passThru
  ret <32 x i8> %3
}

define <32 x i8> @test_test_subs_epi8_rrkz_256(<32 x i8> %a, <32 x i8> %b, i32 %mask) {
; X86-LABEL: test_test_subs_epi8_rrkz_256:
; X86:       # %bb.0:
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpsubsb %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xe8,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_test_subs_epi8_rrkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpsubsb %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xe8,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = call <32 x i8> @llvm.x86.avx2.psubs.b(<32 x i8> %a, <32 x i8> %b)
  %2 = bitcast i32 %mask to <32 x i1>
  %3 = select <32 x i1> %2, <32 x i8> %1, <32 x i8> zeroinitializer
  ret <32 x i8> %3
}

define <32 x i8> @test_test_subs_epi8_rmk_256(<32 x i8> %a, <32 x i8>* %ptr_b, <32 x i8> %passThru, i32 %mask) {
; X86-LABEL: test_test_subs_epi8_rmk_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpsubsb (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe8,0x08]
; X86-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_test_subs_epi8_rmk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpsubsb (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe8,0x0f]
; X64-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %b = load <32 x i8>, <32 x i8>* %ptr_b
  %1 = call <32 x i8> @llvm.x86.avx2.psubs.b(<32 x i8> %a, <32 x i8> %b)
  %2 = bitcast i32 %mask to <32 x i1>
  %3 = select <32 x i1> %2, <32 x i8> %1, <32 x i8> %passThru
  ret <32 x i8> %3
}

define <32 x i8> @test_test_subs_epi8_rmkz_256(<32 x i8> %a, <32 x i8>* %ptr_b, i32 %mask) {
; X86-LABEL: test_test_subs_epi8_rmkz_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpsubsb (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xe8,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_test_subs_epi8_rmkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpsubsb (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xe8,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %b = load <32 x i8>, <32 x i8>* %ptr_b
  %1 = call <32 x i8> @llvm.x86.avx2.psubs.b(<32 x i8> %a, <32 x i8> %b)
  %2 = bitcast i32 %mask to <32 x i1>
  %3 = select <32 x i1> %2, <32 x i8> %1, <32 x i8> zeroinitializer
  ret <32 x i8> %3
}

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 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xe9,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # 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) {
; X86-LABEL: test_mask_subs_epi16_rrk_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpsubsw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe9,0xd1]
; X86-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_subs_epi16_rrk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpsubsw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe9,0xd1]
; X64-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
; X64-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) {
; X86-LABEL: test_mask_subs_epi16_rrkz_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpsubsw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xe9,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_subs_epi16_rrkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpsubsw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xe9,0xc1]
; X64-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) {
; X86-LABEL: test_mask_subs_epi16_rm_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpsubsw (%eax), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xe9,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_subs_epi16_rm_128:
; X64:       # %bb.0:
; X64-NEXT:    vpsubsw (%rdi), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xe9,0x07]
; X64-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) {
; X86-LABEL: test_mask_subs_epi16_rmk_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
; X86-NEXT:    kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
; X86-NEXT:    vpsubsw (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe9,0x08]
; X86-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_subs_epi16_rmk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpsubsw (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe9,0x0f]
; X64-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X64-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) {
; X86-LABEL: test_mask_subs_epi16_rmkz_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
; X86-NEXT:    kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
; X86-NEXT:    vpsubsw (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xe9,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_subs_epi16_rmkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpsubsw (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xe9,0x07]
; X64-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 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xe9,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # 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) {
; X86-LABEL: test_mask_subs_epi16_rrk_256:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpsubsw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe9,0xd1]
; X86-NEXT:    vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_subs_epi16_rrk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpsubsw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe9,0xd1]
; X64-NEXT:    vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
; X64-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) {
; X86-LABEL: test_mask_subs_epi16_rrkz_256:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpsubsw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xe9,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_subs_epi16_rrkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpsubsw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xe9,0xc1]
; X64-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) {
; X86-LABEL: test_mask_subs_epi16_rm_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpsubsw (%eax), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xe9,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_subs_epi16_rm_256:
; X64:       # %bb.0:
; X64-NEXT:    vpsubsw (%rdi), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xe9,0x07]
; X64-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) {
; X86-LABEL: test_mask_subs_epi16_rmk_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpsubsw (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe9,0x08]
; X86-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_subs_epi16_rmk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpsubsw (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe9,0x0f]
; X64-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X64-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) {
; X86-LABEL: test_mask_subs_epi16_rmkz_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpsubsw (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xe9,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_subs_epi16_rmkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpsubsw (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xe9,0x07]
; X64-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)

declare <8 x i16> @llvm.x86.sse2.padds.w(<8 x i16>, <8 x i16>) #0

define <8 x i16> @test_adds_epi16_rrk_128(<8 x i16> %a, <8 x i16> %b, <8 x i16> %passThru, i8 %mask) {
; X86-LABEL: test_adds_epi16_rrk_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpaddsw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xed,0xd1]
; X86-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_adds_epi16_rrk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpaddsw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xed,0xd1]
; X64-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = call <8 x i16> @llvm.x86.sse2.padds.w(<8 x i16> %a, <8 x i16> %b)
  %2 = bitcast i8 %mask to <8 x i1>
  %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> %passThru
  ret <8 x i16> %3
}

define <8 x i16> @test_adds_epi16_rrkz_128(<8 x i16> %a, <8 x i16> %b, i8 %mask) {
; X86-LABEL: test_adds_epi16_rrkz_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpaddsw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xed,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_adds_epi16_rrkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpaddsw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xed,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = call <8 x i16> @llvm.x86.sse2.padds.w(<8 x i16> %a, <8 x i16> %b)
  %2 = bitcast i8 %mask to <8 x i1>
  %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> zeroinitializer
  ret <8 x i16> %3
}

define <8 x i16> @test_adds_epi16_rmk_128(<8 x i16> %a, <8 x i16>* %ptr_b, <8 x i16> %passThru, i8 %mask) {
; X86-LABEL: test_adds_epi16_rmk_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
; X86-NEXT:    kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
; X86-NEXT:    vpaddsw (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xed,0x08]
; X86-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_adds_epi16_rmk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpaddsw (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xed,0x0f]
; X64-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %b = load <8 x i16>, <8 x i16>* %ptr_b
  %1 = call <8 x i16> @llvm.x86.sse2.padds.w(<8 x i16> %a, <8 x i16> %b)
  %2 = bitcast i8 %mask to <8 x i1>
  %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> %passThru
  ret <8 x i16> %3
}

define <8 x i16> @test_adds_epi16_rmkz_128(<8 x i16> %a, <8 x i16>* %ptr_b, i8 %mask) {
; X86-LABEL: test_adds_epi16_rmkz_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
; X86-NEXT:    kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
; X86-NEXT:    vpaddsw (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xed,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_adds_epi16_rmkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpaddsw (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xed,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %b = load <8 x i16>, <8 x i16>* %ptr_b
  %1 = call <8 x i16> @llvm.x86.sse2.padds.w(<8 x i16> %a, <8 x i16> %b)
  %2 = bitcast i8 %mask to <8 x i1>
  %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> zeroinitializer
  ret <8 x i16> %3
}

declare <16 x i16> @llvm.x86.avx2.padds.w(<16 x i16>, <16 x i16>) #0

define <16 x i16> @test_adds_epi16_rrk_256(<16 x i16> %a, <16 x i16> %b, <16 x i16> %passThru, i16 %mask) {
; X86-LABEL: test_adds_epi16_rrk_256:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpaddsw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xed,0xd1]
; X86-NEXT:    vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_adds_epi16_rrk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpaddsw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xed,0xd1]
; X64-NEXT:    vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = call <16 x i16> @llvm.x86.avx2.padds.w(<16 x i16> %a, <16 x i16> %b)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> %passThru
  ret <16 x i16> %3
}

define <16 x i16> @test_adds_epi16_rrkz_256(<16 x i16> %a, <16 x i16> %b, i16 %mask) {
; X86-LABEL: test_adds_epi16_rrkz_256:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpaddsw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xed,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_adds_epi16_rrkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpaddsw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xed,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = call <16 x i16> @llvm.x86.avx2.padds.w(<16 x i16> %a, <16 x i16> %b)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> zeroinitializer
  ret <16 x i16> %3
}

define <16 x i16> @test_adds_epi16_rmk_256(<16 x i16> %a, <16 x i16>* %ptr_b, <16 x i16> %passThru, i16 %mask) {
; X86-LABEL: test_adds_epi16_rmk_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpaddsw (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xed,0x08]
; X86-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_adds_epi16_rmk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpaddsw (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xed,0x0f]
; X64-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %b = load <16 x i16>, <16 x i16>* %ptr_b
  %1 = call <16 x i16> @llvm.x86.avx2.padds.w(<16 x i16> %a, <16 x i16> %b)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> %passThru
  ret <16 x i16> %3
}

define <16 x i16> @test_adds_epi16_rmkz_256(<16 x i16> %a, <16 x i16>* %ptr_b, i16 %mask) {
; X86-LABEL: test_adds_epi16_rmkz_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpaddsw (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xed,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_adds_epi16_rmkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpaddsw (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xed,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %b = load <16 x i16>, <16 x i16>* %ptr_b
  %1 = call <16 x i16> @llvm.x86.avx2.padds.w(<16 x i16> %a, <16 x i16> %b)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> zeroinitializer
  ret <16 x i16> %3
}

declare <16 x i8> @llvm.x86.sse2.padds.b(<16 x i8>, <16 x i8>) #0

define <16 x i8> @test_adds_epi8_rrk_128(<16 x i8> %a, <16 x i8> %b, <16 x i8> %passThru, i16 %mask) {
; X86-LABEL: test_adds_epi8_rrk_128:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpaddsb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xec,0xd1]
; X86-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_adds_epi8_rrk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpaddsb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xec,0xd1]
; X64-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = call <16 x i8> @llvm.x86.sse2.padds.b(<16 x i8> %a, <16 x i8> %b)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x i8> %1, <16 x i8> %passThru
  ret <16 x i8> %3
}

define <16 x i8> @test_adds_epi8_rrkz_128(<16 x i8> %a, <16 x i8> %b, i16 %mask) {
; X86-LABEL: test_adds_epi8_rrkz_128:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpaddsb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xec,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_adds_epi8_rrkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpaddsb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xec,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = call <16 x i8> @llvm.x86.sse2.padds.b(<16 x i8> %a, <16 x i8> %b)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x i8> %1, <16 x i8> zeroinitializer
  ret <16 x i8> %3
}

define <16 x i8> @test_adds_epi8_rm_128(<16 x i8> %a, <16 x i8>* %ptr_b) {
; X86-LABEL: test_adds_epi8_rm_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpaddsb (%eax), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xec,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_adds_epi8_rm_128:
; X64:       # %bb.0:
; X64-NEXT:    vpaddsb (%rdi), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xec,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %b = load <16 x i8>, <16 x i8>* %ptr_b
  %1 = call <16 x i8> @llvm.x86.sse2.padds.b(<16 x i8> %a, <16 x i8> %b)
  ret <16 x i8> %1
}

define <16 x i8> @test_adds_epi8_rmk_128(<16 x i8> %a, <16 x i8>* %ptr_b, <16 x i8> %passThru, i16 %mask) {
; X86-LABEL: test_adds_epi8_rmk_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpaddsb (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xec,0x08]
; X86-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_adds_epi8_rmk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpaddsb (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xec,0x0f]
; X64-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %b = load <16 x i8>, <16 x i8>* %ptr_b
  %1 = call <16 x i8> @llvm.x86.sse2.padds.b(<16 x i8> %a, <16 x i8> %b)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x i8> %1, <16 x i8> %passThru
  ret <16 x i8> %3
}

define <16 x i8> @test_adds_epi8_rmkz_128(<16 x i8> %a, <16 x i8>* %ptr_b, i16 %mask) {
; X86-LABEL: test_adds_epi8_rmkz_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpaddsb (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xec,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_adds_epi8_rmkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpaddsb (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xec,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %b = load <16 x i8>, <16 x i8>* %ptr_b
  %1 = call <16 x i8> @llvm.x86.sse2.padds.b(<16 x i8> %a, <16 x i8> %b)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x i8> %1, <16 x i8> zeroinitializer
  ret <16 x i8> %3
}

declare <32 x i8> @llvm.x86.avx2.padds.b(<32 x i8>, <32 x i8>) #0

define <32 x i8> @test_adds_epi8_rrk_256(<32 x i8> %a, <32 x i8> %b, <32 x i8> %passThru, i32 %mask) {
; X86-LABEL: test_adds_epi8_rrk_256:
; X86:       # %bb.0:
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpaddsb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xec,0xd1]
; X86-NEXT:    vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_adds_epi8_rrk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpaddsb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xec,0xd1]
; X64-NEXT:    vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = call <32 x i8> @llvm.x86.avx2.padds.b(<32 x i8> %a, <32 x i8> %b)
  %2 = bitcast i32 %mask to <32 x i1>
  %3 = select <32 x i1> %2, <32 x i8> %1, <32 x i8> %passThru
  ret <32 x i8> %3
}

define <32 x i8> @test_adds_epi8_rrkz_256(<32 x i8> %a, <32 x i8> %b, i32 %mask) {
; X86-LABEL: test_adds_epi8_rrkz_256:
; X86:       # %bb.0:
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpaddsb %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xec,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_adds_epi8_rrkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpaddsb %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xec,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = call <32 x i8> @llvm.x86.avx2.padds.b(<32 x i8> %a, <32 x i8> %b)
  %2 = bitcast i32 %mask to <32 x i1>
  %3 = select <32 x i1> %2, <32 x i8> %1, <32 x i8> zeroinitializer
  ret <32 x i8> %3
}

define <32 x i8> @test_adds_epi8_rmk_256(<32 x i8> %a, <32 x i8>* %ptr_b, <32 x i8> %passThru, i32 %mask) {
; X86-LABEL: test_adds_epi8_rmk_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpaddsb (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xec,0x08]
; X86-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_adds_epi8_rmk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpaddsb (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xec,0x0f]
; X64-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %b = load <32 x i8>, <32 x i8>* %ptr_b
  %1 = call <32 x i8> @llvm.x86.avx2.padds.b(<32 x i8> %a, <32 x i8> %b)
  %2 = bitcast i32 %mask to <32 x i1>
  %3 = select <32 x i1> %2, <32 x i8> %1, <32 x i8> %passThru
  ret <32 x i8> %3
}

define <32 x i8> @test_adds_epi8_rmkz_256(<32 x i8> %a, <32 x i8>* %ptr_b, i32 %mask) {
; X86-LABEL: test_adds_epi8_rmkz_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpaddsb (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xec,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_adds_epi8_rmkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpaddsb (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xec,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %b = load <32 x i8>, <32 x i8>* %ptr_b
  %1 = call <32 x i8> @llvm.x86.avx2.padds.b(<32 x i8> %a, <32 x i8> %b)
  %2 = bitcast i32 %mask to <32 x i1>
  %3 = select <32 x i1> %2, <32 x i8> %1, <32 x i8> zeroinitializer
  ret <32 x i8> %3
}

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 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xec,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # 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) {
; X86-LABEL: test_mask_adds_epi8_rrk_128:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpaddsb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xec,0xd1]
; X86-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_adds_epi8_rrk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpaddsb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xec,0xd1]
; X64-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
; X64-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) {
; X86-LABEL: test_mask_adds_epi8_rrkz_128:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpaddsb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xec,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_adds_epi8_rrkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpaddsb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xec,0xc1]
; X64-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) {
; X86-LABEL: test_mask_adds_epi8_rm_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpaddsb (%eax), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xec,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_adds_epi8_rm_128:
; X64:       # %bb.0:
; X64-NEXT:    vpaddsb (%rdi), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xec,0x07]
; X64-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) {
; X86-LABEL: test_mask_adds_epi8_rmk_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpaddsb (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xec,0x08]
; X86-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_adds_epi8_rmk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpaddsb (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xec,0x0f]
; X64-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X64-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) {
; X86-LABEL: test_mask_adds_epi8_rmkz_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpaddsb (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xec,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_adds_epi8_rmkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpaddsb (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xec,0x07]
; X64-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 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xec,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # 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) {
; X86-LABEL: test_mask_adds_epi8_rrk_256:
; X86:       # %bb.0:
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpaddsb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xec,0xd1]
; X86-NEXT:    vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_adds_epi8_rrk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpaddsb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xec,0xd1]
; X64-NEXT:    vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
; X64-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) {
; X86-LABEL: test_mask_adds_epi8_rrkz_256:
; X86:       # %bb.0:
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpaddsb %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xec,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_adds_epi8_rrkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpaddsb %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xec,0xc1]
; X64-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) {
; X86-LABEL: test_mask_adds_epi8_rm_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpaddsb (%eax), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xec,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_adds_epi8_rm_256:
; X64:       # %bb.0:
; X64-NEXT:    vpaddsb (%rdi), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xec,0x07]
; X64-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) {
; X86-LABEL: test_mask_adds_epi8_rmk_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpaddsb (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xec,0x08]
; X86-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_adds_epi8_rmk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpaddsb (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xec,0x0f]
; X64-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X64-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) {
; X86-LABEL: test_mask_adds_epi8_rmkz_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpaddsb (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xec,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_adds_epi8_rmkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpaddsb (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xec,0x07]
; X64-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 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xe8,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # 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) {
; X86-LABEL: test_mask_subs_epi8_rrk_128:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpsubsb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe8,0xd1]
; X86-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_subs_epi8_rrk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpsubsb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe8,0xd1]
; X64-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
; X64-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) {
; X86-LABEL: test_mask_subs_epi8_rrkz_128:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpsubsb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xe8,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_subs_epi8_rrkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpsubsb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xe8,0xc1]
; X64-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) {
; X86-LABEL: test_mask_subs_epi8_rm_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpsubsb (%eax), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xe8,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_subs_epi8_rm_128:
; X64:       # %bb.0:
; X64-NEXT:    vpsubsb (%rdi), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xe8,0x07]
; X64-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) {
; X86-LABEL: test_mask_subs_epi8_rmk_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpsubsb (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe8,0x08]
; X86-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_subs_epi8_rmk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpsubsb (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe8,0x0f]
; X64-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X64-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) {
; X86-LABEL: test_mask_subs_epi8_rmkz_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpsubsb (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xe8,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_subs_epi8_rmkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpsubsb (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xe8,0x07]
; X64-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 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xe8,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # 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) {
; X86-LABEL: test_mask_subs_epi8_rrk_256:
; X86:       # %bb.0:
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpsubsb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe8,0xd1]
; X86-NEXT:    vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_subs_epi8_rrk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpsubsb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe8,0xd1]
; X64-NEXT:    vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
; X64-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) {
; X86-LABEL: test_mask_subs_epi8_rrkz_256:
; X86:       # %bb.0:
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpsubsb %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xe8,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_subs_epi8_rrkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpsubsb %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xe8,0xc1]
; X64-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) {
; X86-LABEL: test_mask_subs_epi8_rm_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpsubsb (%eax), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xe8,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_subs_epi8_rm_256:
; X64:       # %bb.0:
; X64-NEXT:    vpsubsb (%rdi), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xe8,0x07]
; X64-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) {
; X86-LABEL: test_mask_subs_epi8_rmk_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpsubsb (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe8,0x08]
; X86-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_subs_epi8_rmk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpsubsb (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe8,0x0f]
; X64-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X64-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) {
; X86-LABEL: test_mask_subs_epi8_rmkz_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpsubsb (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xe8,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_subs_epi8_rmkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpsubsb (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xe8,0x07]
; X64-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)