test/CodeGen/X86/avx512vl-intrinsics-upgrade.ll

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

declare <4 x i32> @llvm.x86.avx512.mask.pbroadcast.d.gpr.128(i32, <4 x i32>, i8)

define <4 x i32>@test_int_x86_avx512_mask_pbroadcast_d_gpr_128(i32 %x0, <4 x i32> %x1, i8 %mask) {
; X86-LABEL: test_int_x86_avx512_mask_pbroadcast_d_gpr_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpbroadcastd %eax, %xmm1 # encoding: [0x62,0xf2,0x7d,0x08,0x7c,0xc8]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
; X86-NEXT:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpbroadcastd %eax, %xmm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x7c,0xc0]
; X86-NEXT:    vpbroadcastd %eax, %xmm2 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x7c,0xd0]
; X86-NEXT:    vpaddd %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfe,0xc2]
; X86-NEXT:    vpaddd %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pbroadcast_d_gpr_128:
; X64:       # %bb.0:
; X64-NEXT:    vpbroadcastd %edi, %xmm1 # encoding: [0x62,0xf2,0x7d,0x08,0x7c,0xcf]
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpbroadcastd %edi, %xmm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x7c,0xc7]
; X64-NEXT:    vpbroadcastd %edi, %xmm2 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x7c,0xd7]
; X64-NEXT:    vpaddd %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfe,0xc2]
; X64-NEXT:    vpaddd %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i32> @llvm.x86.avx512.mask.pbroadcast.d.gpr.128(i32 %x0, <4 x i32> %x1, i8 -1)
  %res1 = call <4 x i32> @llvm.x86.avx512.mask.pbroadcast.d.gpr.128(i32 %x0, <4 x i32> %x1, i8 %mask)
  %res2 = call <4 x i32> @llvm.x86.avx512.mask.pbroadcast.d.gpr.128(i32 %x0, <4 x i32> zeroinitializer, i8 %mask)
  %res3 = add <4 x i32> %res, %res1
  %res4 = add <4 x i32> %res2, %res3
  ret <4 x i32> %res4
}


declare <2 x i64> @llvm.x86.avx512.mask.pbroadcast.q.gpr.128(i64, <2 x i64>, i8)

define <2 x i64>@test_int_x86_avx512_mask_pbroadcast_q_gpr_128(i64 %x0, <2 x i64> %x1, i8 %mask) {
; X86-LABEL: test_int_x86_avx512_mask_pbroadcast_q_gpr_128:
; X86:       # %bb.0:
; X86-NEXT:    vmovq {{[0-9]+}}(%esp), %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xfa,0x7e,0x4c,0x24,0x04]
; X86-NEXT:    # xmm1 = mem[0],zero
; X86-NEXT:    vpbroadcastq %xmm1, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x59,0xd1]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x0c]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpbroadcastq %xmm1, %xmm0 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x59,0xc1]
; X86-NEXT:    vpbroadcastq %xmm1, %xmm1 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x59,0xc9]
; X86-NEXT:    vpaddq %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd4,0xc1]
; X86-NEXT:    vpaddq %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xd4,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pbroadcast_q_gpr_128:
; X64:       # %bb.0:
; X64-NEXT:    vpbroadcastq %rdi, %xmm1 # encoding: [0x62,0xf2,0xfd,0x08,0x7c,0xcf]
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpbroadcastq %rdi, %xmm0 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x7c,0xc7]
; X64-NEXT:    vpbroadcastq %rdi, %xmm2 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x7c,0xd7]
; 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.pbroadcast.q.gpr.128(i64 %x0, <2 x i64> %x1,i8 -1)
  %res1 = call <2 x i64> @llvm.x86.avx512.mask.pbroadcast.q.gpr.128(i64 %x0, <2 x i64> %x1,i8 %mask)
  %res2 = call <2 x i64> @llvm.x86.avx512.mask.pbroadcast.q.gpr.128(i64 %x0, <2 x i64> zeroinitializer,i8 %mask)
  %res3 = add <2 x i64> %res, %res1
  %res4 = add <2 x i64> %res2, %res3
  ret <2 x i64> %res4
}


 declare <8 x i32> @llvm.x86.avx512.mask.pbroadcast.d.gpr.256(i32, <8 x i32>, i8)

  define <8 x i32>@test_int_x86_avx512_mask_pbroadcast_d_gpr_256(i32 %x0, <8 x i32> %x1, i8 %mask) {
; X86-LABEL: test_int_x86_avx512_mask_pbroadcast_d_gpr_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpbroadcastd %eax, %ymm1 # encoding: [0x62,0xf2,0x7d,0x28,0x7c,0xc8]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
; X86-NEXT:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpbroadcastd %eax, %ymm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x7c,0xc0]
; X86-NEXT:    vpbroadcastd %eax, %ymm2 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x7c,0xd0]
; X86-NEXT:    vpaddd %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfe,0xc2]
; X86-NEXT:    vpaddd %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pbroadcast_d_gpr_256:
; X64:       # %bb.0:
; X64-NEXT:    vpbroadcastd %edi, %ymm1 # encoding: [0x62,0xf2,0x7d,0x28,0x7c,0xcf]
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpbroadcastd %edi, %ymm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x7c,0xc7]
; X64-NEXT:    vpbroadcastd %edi, %ymm2 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x7c,0xd7]
; X64-NEXT:    vpaddd %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfe,0xc2]
; X64-NEXT:    vpaddd %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
    %res = call <8 x i32> @llvm.x86.avx512.mask.pbroadcast.d.gpr.256(i32 %x0, <8 x i32> %x1, i8 -1)
    %res1 = call <8 x i32> @llvm.x86.avx512.mask.pbroadcast.d.gpr.256(i32 %x0, <8 x i32> %x1, i8 %mask)
    %res2 = call <8 x i32> @llvm.x86.avx512.mask.pbroadcast.d.gpr.256(i32 %x0, <8 x i32> zeroinitializer, i8 %mask)
    %res3 = add <8 x i32> %res, %res1
    %res4 = add <8 x i32> %res2, %res3
    ret <8 x i32> %res4
  }

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

  define <4 x i64>@test_int_x86_avx512_mask_pbroadcast_q_gpr_256(i64 %x0, <4 x i64> %x1, i8 %mask) {
; X86-LABEL: test_int_x86_avx512_mask_pbroadcast_q_gpr_256:
; X86:       # %bb.0:
; X86-NEXT:    vmovq {{[0-9]+}}(%esp), %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xfa,0x7e,0x4c,0x24,0x04]
; X86-NEXT:    # xmm1 = mem[0],zero
; X86-NEXT:    vpbroadcastq %xmm1, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x59,0xd1]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x0c]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpbroadcastq %xmm1, %ymm0 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x59,0xc1]
; X86-NEXT:    vpbroadcastq %xmm1, %ymm1 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x59,0xc9]
; X86-NEXT:    vpaddq %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd4,0xc1]
; X86-NEXT:    vpaddq %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xd4,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pbroadcast_q_gpr_256:
; X64:       # %bb.0:
; X64-NEXT:    vpbroadcastq %rdi, %ymm1 # encoding: [0x62,0xf2,0xfd,0x28,0x7c,0xcf]
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpbroadcastq %rdi, %ymm0 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x7c,0xc7]
; X64-NEXT:    vpbroadcastq %rdi, %ymm2 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x7c,0xd7]
; 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.pbroadcast.q.gpr.256(i64 %x0, <4 x i64> %x1,i8 -1)
    %res1 = call <4 x i64> @llvm.x86.avx512.mask.pbroadcast.q.gpr.256(i64 %x0, <4 x i64> %x1,i8 %mask)
    %res2 = call <4 x i64> @llvm.x86.avx512.mask.pbroadcast.q.gpr.256(i64 %x0, <4 x i64> zeroinitializer,i8 %mask)
    %res3 = add <4 x i64> %res, %res1
    %res4 = add <4 x i64> %res2, %res3
    ret <4 x i64> %res4
  }



declare <8 x i32> @llvm.x86.avx512.pbroadcastd.256(<4 x i32>, <8 x i32>, i8)

define <8 x i32>@test_int_x86_avx512_pbroadcastd_256(<4 x i32> %x0, <8 x i32> %x1, i8 %mask, i32 * %y_ptr) {
; X86-LABEL: test_int_x86_avx512_pbroadcastd_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x08]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x04]
; X86-NEXT:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpbroadcastd %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x58,0xc8]
; X86-NEXT:    vpbroadcastd %xmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x58,0xc0]
; X86-NEXT:    vpaddd (%eax){1to8}, %ymm1, %ymm1 # encoding: [0x62,0xf1,0x75,0x38,0xfe,0x08]
; X86-NEXT:    vpaddd %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfe,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_pbroadcastd_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpbroadcastd %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x58,0xc8]
; X64-NEXT:    vpbroadcastd %xmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x58,0xc0]
; X64-NEXT:    vpaddd (%rsi){1to8}, %ymm1, %ymm1 # encoding: [0x62,0xf1,0x75,0x38,0xfe,0x0e]
; X64-NEXT:    vpaddd %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfe,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %y_32  = load i32, i32 * %y_ptr
  %y = insertelement <4 x i32> undef, i32 %y_32, i32 0
  %res = call <8 x i32> @llvm.x86.avx512.pbroadcastd.256(<4 x i32> %y, <8 x i32> %x1, i8 -1)
  %res1 = call <8 x i32> @llvm.x86.avx512.pbroadcastd.256(<4 x i32> %x0, <8 x i32> %x1, i8 %mask)
  %res2 = call <8 x i32> @llvm.x86.avx512.pbroadcastd.256(<4 x i32> %x0, <8 x i32> zeroinitializer, i8 %mask)
  %res3 = add <8 x i32> %res, %res1
  %res4 = add <8 x i32> %res2, %res3
  ret <8 x i32> %res4
}

declare <4 x i32> @llvm.x86.avx512.pbroadcastd.128(<4 x i32>, <4 x i32>, i8)

define <4 x i32>@test_int_x86_avx512_pbroadcastd_128(<4 x i32> %x0, <4 x i32> %x1, i8 %mask) {
; X86-LABEL: test_int_x86_avx512_pbroadcastd_128:
; X86:       # %bb.0:
; X86-NEXT:    vpbroadcastd %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x58,0xd0]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpbroadcastd %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x58,0xc8]
; X86-NEXT:    vpbroadcastd %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x58,0xc0]
; X86-NEXT:    vpaddd %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc0]
; X86-NEXT:    vpaddd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_pbroadcastd_128:
; X64:       # %bb.0:
; X64-NEXT:    vpbroadcastd %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x58,0xd0]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpbroadcastd %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x58,0xc8]
; X64-NEXT:    vpbroadcastd %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x58,0xc0]
; X64-NEXT:    vpaddd %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc0]
; X64-NEXT:    vpaddd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i32> @llvm.x86.avx512.pbroadcastd.128(<4 x i32> %x0, <4 x i32> %x1, i8 -1)
  %res1 = call <4 x i32> @llvm.x86.avx512.pbroadcastd.128(<4 x i32> %x0, <4 x i32> %x1, i8 %mask)
  %res2 = call <4 x i32> @llvm.x86.avx512.pbroadcastd.128(<4 x i32> %x0, <4 x i32> zeroinitializer, i8 %mask)
  %res3 = add <4 x i32> %res, %res1
  %res4 = add <4 x i32> %res2, %res3
  ret <4 x i32> %res4
}

declare <4 x i64> @llvm.x86.avx512.pbroadcastq.256(<2 x i64>, <4 x i64>, i8)

define <4 x i64>@test_int_x86_avx512_pbroadcastq_256(<2 x i64> %x0, <4 x i64> %x1, i8 %mask) {
; X86-LABEL: test_int_x86_avx512_pbroadcastq_256:
; X86:       # %bb.0:
; X86-NEXT:    vpbroadcastq %xmm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x59,0xd0]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpbroadcastq %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x59,0xc8]
; X86-NEXT:    vpbroadcastq %xmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x59,0xc0]
; X86-NEXT:    vpaddq %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xd4,0xc0]
; X86-NEXT:    vpaddq %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xd4,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_pbroadcastq_256:
; X64:       # %bb.0:
; X64-NEXT:    vpbroadcastq %xmm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x59,0xd0]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpbroadcastq %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x59,0xc8]
; X64-NEXT:    vpbroadcastq %xmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x59,0xc0]
; X64-NEXT:    vpaddq %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xd4,0xc0]
; X64-NEXT:    vpaddq %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xd4,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i64> @llvm.x86.avx512.pbroadcastq.256(<2 x i64> %x0, <4 x i64> %x1,i8 -1)
  %res1 = call <4 x i64> @llvm.x86.avx512.pbroadcastq.256(<2 x i64> %x0, <4 x i64> %x1,i8 %mask)
  %res2 = call <4 x i64> @llvm.x86.avx512.pbroadcastq.256(<2 x i64> %x0, <4 x i64> zeroinitializer,i8 %mask)
  %res3 = add <4 x i64> %res, %res1
  %res4 = add <4 x i64> %res2, %res3
  ret <4 x i64> %res4
}

declare <2 x i64> @llvm.x86.avx512.pbroadcastq.128(<2 x i64>, <2 x i64>, i8)

define <2 x i64>@test_int_x86_avx512_pbroadcastq_128(<2 x i64> %x0, <2 x i64> %x1, i8 %mask) {
; X86-LABEL: test_int_x86_avx512_pbroadcastq_128:
; X86:       # %bb.0:
; X86-NEXT:    vpbroadcastq %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x59,0xd0]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpbroadcastq %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x59,0xc8]
; X86-NEXT:    vpbroadcastq %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x59,0xc0]
; X86-NEXT:    vpaddq %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xd4,0xc0]
; X86-NEXT:    vpaddq %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xd4,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_pbroadcastq_128:
; X64:       # %bb.0:
; X64-NEXT:    vpbroadcastq %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x59,0xd0]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpbroadcastq %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x59,0xc8]
; X64-NEXT:    vpbroadcastq %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x59,0xc0]
; X64-NEXT:    vpaddq %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xd4,0xc0]
; X64-NEXT:    vpaddq %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xd4,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x i64> @llvm.x86.avx512.pbroadcastq.128(<2 x i64> %x0, <2 x i64> %x1,i8 -1)
  %res1 = call <2 x i64> @llvm.x86.avx512.pbroadcastq.128(<2 x i64> %x0, <2 x i64> %x1,i8 %mask)
  %res2 = call <2 x i64> @llvm.x86.avx512.pbroadcastq.128(<2 x i64> %x0, <2 x i64> zeroinitializer,i8 %mask)
  %res3 = add <2 x i64> %res, %res1
  %res4 = add <2 x i64> %res2, %res3
  ret <2 x i64> %res4
}

declare <4 x double> @llvm.x86.avx512.mask.broadcast.sd.pd.256(<2 x double>, <4 x double>, i8) nounwind readonly

define <4 x double> @test_x86_vbroadcast_sd_pd_256(<2 x double> %a0, <4 x double> %a1, i8 %mask ) {
; X86-LABEL: test_x86_vbroadcast_sd_pd_256:
; X86:       # %bb.0:
; X86-NEXT:    vbroadcastsd %xmm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x19,0xd0]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vbroadcastsd %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x19,0xc8]
; X86-NEXT:    vaddpd %ymm1, %ymm2, %ymm1 # EVEX TO VEX Compression encoding: [0xc5,0xed,0x58,0xc9]
; X86-NEXT:    vbroadcastsd %xmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x19,0xc0]
; X86-NEXT:    vaddpd %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x58,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_x86_vbroadcast_sd_pd_256:
; X64:       # %bb.0:
; X64-NEXT:    vbroadcastsd %xmm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x19,0xd0]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vbroadcastsd %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x19,0xc8]
; X64-NEXT:    vaddpd %ymm1, %ymm2, %ymm1 # EVEX TO VEX Compression encoding: [0xc5,0xed,0x58,0xc9]
; X64-NEXT:    vbroadcastsd %xmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x19,0xc0]
; X64-NEXT:    vaddpd %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x58,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x double> @llvm.x86.avx512.mask.broadcast.sd.pd.256(<2 x double> %a0, <4 x double> zeroinitializer, i8 -1)
  %res1 = call <4 x double> @llvm.x86.avx512.mask.broadcast.sd.pd.256(<2 x double> %a0, <4 x double> %a1, i8 %mask)
  %res2 = call <4 x double> @llvm.x86.avx512.mask.broadcast.sd.pd.256(<2 x double> %a0, <4 x double> zeroinitializer, i8 %mask)
  %res3 = fadd <4 x double> %res, %res1
  %res4 = fadd <4 x double> %res2, %res3
  ret <4 x double> %res4
}

declare <8 x float> @llvm.x86.avx512.mask.broadcast.ss.ps.256(<4 x float>, <8 x float>, i8) nounwind readonly

define <8 x float> @test_x86_vbroadcast_ss_ps_256(<4 x float> %a0, <8 x float> %a1, i8 %mask ) {
; X86-LABEL: test_x86_vbroadcast_ss_ps_256:
; X86:       # %bb.0:
; X86-NEXT:    vbroadcastss %xmm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x18,0xd0]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vbroadcastss %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x18,0xc8]
; X86-NEXT:    vaddps %ymm1, %ymm2, %ymm1 # EVEX TO VEX Compression encoding: [0xc5,0xec,0x58,0xc9]
; X86-NEXT:    vbroadcastss %xmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x18,0xc0]
; X86-NEXT:    vaddps %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x58,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_x86_vbroadcast_ss_ps_256:
; X64:       # %bb.0:
; X64-NEXT:    vbroadcastss %xmm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x18,0xd0]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vbroadcastss %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x18,0xc8]
; X64-NEXT:    vaddps %ymm1, %ymm2, %ymm1 # EVEX TO VEX Compression encoding: [0xc5,0xec,0x58,0xc9]
; X64-NEXT:    vbroadcastss %xmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x18,0xc0]
; X64-NEXT:    vaddps %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x58,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask.broadcast.ss.ps.256(<4 x float> %a0, <8 x float> zeroinitializer, i8 -1)
  %res1 = call <8 x float> @llvm.x86.avx512.mask.broadcast.ss.ps.256(<4 x float> %a0, <8 x float> %a1, i8 %mask)
  %res2 = call <8 x float> @llvm.x86.avx512.mask.broadcast.ss.ps.256(<4 x float> %a0, <8 x float> zeroinitializer, i8 %mask)
  %res3 = fadd <8 x float> %res, %res1
  %res4 = fadd <8 x float> %res2, %res3
  ret <8 x float> %res4
}

declare <4 x float> @llvm.x86.avx512.mask.broadcast.ss.ps.128(<4 x float>, <4 x float>, i8) nounwind readonly

define <4 x float> @test_x86_vbroadcast_ss_ps_128(<4 x float> %a0, <4 x float> %a1, i8 %mask ) {
; X86-LABEL: test_x86_vbroadcast_ss_ps_128:
; X86:       # %bb.0:
; X86-NEXT:    vbroadcastss %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x18,0xd0]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vbroadcastss %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x18,0xc8]
; X86-NEXT:    vaddps %xmm1, %xmm2, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xe8,0x58,0xc9]
; X86-NEXT:    vbroadcastss %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x18,0xc0]
; X86-NEXT:    vaddps %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x58,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_x86_vbroadcast_ss_ps_128:
; X64:       # %bb.0:
; X64-NEXT:    vbroadcastss %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x18,0xd0]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vbroadcastss %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x18,0xc8]
; X64-NEXT:    vaddps %xmm1, %xmm2, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xe8,0x58,0xc9]
; X64-NEXT:    vbroadcastss %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x18,0xc0]
; X64-NEXT:    vaddps %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x58,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask.broadcast.ss.ps.128(<4 x float> %a0, <4 x float> zeroinitializer, i8 -1)
  %res1 = call <4 x float> @llvm.x86.avx512.mask.broadcast.ss.ps.128(<4 x float> %a0, <4 x float> %a1, i8 %mask)
  %res2 = call <4 x float> @llvm.x86.avx512.mask.broadcast.ss.ps.128(<4 x float> %a0, <4 x float> zeroinitializer, i8 %mask)
  %res3 = fadd <4 x float> %res, %res1
  %res4 = fadd <4 x float> %res2, %res3
  ret <4 x float> %res4
}

declare <4 x float> @llvm.x86.avx512.mask.movsldup.128(<4 x float>, <4 x float>, i8)

define <4 x float>@test_int_x86_avx512_mask_movsldup_128(<4 x float> %x0, <4 x float> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_movsldup_128:
; X86:       # %bb.0:
; X86-NEXT:    vmovsldup %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xfa,0x12,0xd0]
; X86-NEXT:    # xmm2 = xmm0[0,0,2,2]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vmovsldup %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7e,0x09,0x12,0xc8]
; X86-NEXT:    # xmm1 {%k1} = xmm0[0,0,2,2]
; X86-NEXT:    vaddps %xmm2, %xmm1, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf0,0x58,0xca]
; X86-NEXT:    vmovsldup %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7e,0x89,0x12,0xc0]
; X86-NEXT:    # xmm0 {%k1} {z} = xmm0[0,0,2,2]
; X86-NEXT:    vaddps %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x58,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_movsldup_128:
; X64:       # %bb.0:
; X64-NEXT:    vmovsldup %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xfa,0x12,0xd0]
; X64-NEXT:    # xmm2 = xmm0[0,0,2,2]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vmovsldup %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7e,0x09,0x12,0xc8]
; X64-NEXT:    # xmm1 {%k1} = xmm0[0,0,2,2]
; X64-NEXT:    vaddps %xmm2, %xmm1, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf0,0x58,0xca]
; X64-NEXT:    vmovsldup %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7e,0x89,0x12,0xc0]
; X64-NEXT:    # xmm0 {%k1} {z} = xmm0[0,0,2,2]
; X64-NEXT:    vaddps %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x58,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask.movsldup.128(<4 x float> %x0, <4 x float> %x1, i8 %x2)
  %res1 = call <4 x float> @llvm.x86.avx512.mask.movsldup.128(<4 x float> %x0, <4 x float> %x1, i8 -1)
  %res2 = call <4 x float> @llvm.x86.avx512.mask.movsldup.128(<4 x float> %x0, <4 x float> zeroinitializer, i8 %x2)
  %res3 = fadd <4 x float> %res, %res1
  %res4 = fadd <4 x float> %res2, %res3
  ret <4 x float> %res4
}

declare <8 x float> @llvm.x86.avx512.mask.movsldup.256(<8 x float>, <8 x float>, i8)

define <8 x float>@test_int_x86_avx512_mask_movsldup_256(<8 x float> %x0, <8 x float> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_movsldup_256:
; X86:       # %bb.0:
; X86-NEXT:    vmovsldup %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xfe,0x12,0xd0]
; X86-NEXT:    # ymm2 = ymm0[0,0,2,2,4,4,6,6]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vmovsldup %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7e,0x29,0x12,0xc8]
; X86-NEXT:    # ymm1 {%k1} = ymm0[0,0,2,2,4,4,6,6]
; X86-NEXT:    vaddps %ymm2, %ymm1, %ymm1 # EVEX TO VEX Compression encoding: [0xc5,0xf4,0x58,0xca]
; X86-NEXT:    vmovsldup %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7e,0xa9,0x12,0xc0]
; X86-NEXT:    # ymm0 {%k1} {z} = ymm0[0,0,2,2,4,4,6,6]
; X86-NEXT:    vaddps %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x58,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_movsldup_256:
; X64:       # %bb.0:
; X64-NEXT:    vmovsldup %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xfe,0x12,0xd0]
; X64-NEXT:    # ymm2 = ymm0[0,0,2,2,4,4,6,6]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vmovsldup %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7e,0x29,0x12,0xc8]
; X64-NEXT:    # ymm1 {%k1} = ymm0[0,0,2,2,4,4,6,6]
; X64-NEXT:    vaddps %ymm2, %ymm1, %ymm1 # EVEX TO VEX Compression encoding: [0xc5,0xf4,0x58,0xca]
; X64-NEXT:    vmovsldup %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7e,0xa9,0x12,0xc0]
; X64-NEXT:    # ymm0 {%k1} {z} = ymm0[0,0,2,2,4,4,6,6]
; X64-NEXT:    vaddps %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x58,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask.movsldup.256(<8 x float> %x0, <8 x float> %x1, i8 %x2)
  %res1 = call <8 x float> @llvm.x86.avx512.mask.movsldup.256(<8 x float> %x0, <8 x float> %x1, i8 -1)
  %res2 = call <8 x float> @llvm.x86.avx512.mask.movsldup.256(<8 x float> %x0, <8 x float> zeroinitializer, i8 %x2)
  %res3 = fadd <8 x float> %res, %res1
  %res4 = fadd <8 x float> %res2, %res3
  ret <8 x float> %res4
}

declare <4 x float> @llvm.x86.avx512.mask.movshdup.128(<4 x float>, <4 x float>, i8)

define <4 x float>@test_int_x86_avx512_mask_movshdup_128(<4 x float> %x0, <4 x float> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_movshdup_128:
; X86:       # %bb.0:
; X86-NEXT:    vmovshdup %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xfa,0x16,0xd0]
; X86-NEXT:    # xmm2 = xmm0[1,1,3,3]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vmovshdup %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7e,0x09,0x16,0xc8]
; X86-NEXT:    # xmm1 {%k1} = xmm0[1,1,3,3]
; X86-NEXT:    vaddps %xmm2, %xmm1, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf0,0x58,0xca]
; X86-NEXT:    vmovshdup %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7e,0x89,0x16,0xc0]
; X86-NEXT:    # xmm0 {%k1} {z} = xmm0[1,1,3,3]
; X86-NEXT:    vaddps %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x58,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_movshdup_128:
; X64:       # %bb.0:
; X64-NEXT:    vmovshdup %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xfa,0x16,0xd0]
; X64-NEXT:    # xmm2 = xmm0[1,1,3,3]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vmovshdup %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7e,0x09,0x16,0xc8]
; X64-NEXT:    # xmm1 {%k1} = xmm0[1,1,3,3]
; X64-NEXT:    vaddps %xmm2, %xmm1, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf0,0x58,0xca]
; X64-NEXT:    vmovshdup %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7e,0x89,0x16,0xc0]
; X64-NEXT:    # xmm0 {%k1} {z} = xmm0[1,1,3,3]
; X64-NEXT:    vaddps %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x58,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask.movshdup.128(<4 x float> %x0, <4 x float> %x1, i8 %x2)
  %res1 = call <4 x float> @llvm.x86.avx512.mask.movshdup.128(<4 x float> %x0, <4 x float> %x1, i8 -1)
  %res2 = call <4 x float> @llvm.x86.avx512.mask.movshdup.128(<4 x float> %x0, <4 x float> zeroinitializer, i8 %x2)
  %res3 = fadd <4 x float> %res, %res1
  %res4 = fadd <4 x float> %res2, %res3
  ret <4 x float> %res4
}

declare <8 x float> @llvm.x86.avx512.mask.movshdup.256(<8 x float>, <8 x float>, i8)

define <8 x float>@test_int_x86_avx512_mask_movshdup_256(<8 x float> %x0, <8 x float> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_movshdup_256:
; X86:       # %bb.0:
; X86-NEXT:    vmovshdup %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xfe,0x16,0xd0]
; X86-NEXT:    # ymm2 = ymm0[1,1,3,3,5,5,7,7]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vmovshdup %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7e,0x29,0x16,0xc8]
; X86-NEXT:    # ymm1 {%k1} = ymm0[1,1,3,3,5,5,7,7]
; X86-NEXT:    vaddps %ymm2, %ymm1, %ymm1 # EVEX TO VEX Compression encoding: [0xc5,0xf4,0x58,0xca]
; X86-NEXT:    vmovshdup %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7e,0xa9,0x16,0xc0]
; X86-NEXT:    # ymm0 {%k1} {z} = ymm0[1,1,3,3,5,5,7,7]
; X86-NEXT:    vaddps %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x58,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_movshdup_256:
; X64:       # %bb.0:
; X64-NEXT:    vmovshdup %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xfe,0x16,0xd0]
; X64-NEXT:    # ymm2 = ymm0[1,1,3,3,5,5,7,7]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vmovshdup %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7e,0x29,0x16,0xc8]
; X64-NEXT:    # ymm1 {%k1} = ymm0[1,1,3,3,5,5,7,7]
; X64-NEXT:    vaddps %ymm2, %ymm1, %ymm1 # EVEX TO VEX Compression encoding: [0xc5,0xf4,0x58,0xca]
; X64-NEXT:    vmovshdup %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7e,0xa9,0x16,0xc0]
; X64-NEXT:    # ymm0 {%k1} {z} = ymm0[1,1,3,3,5,5,7,7]
; X64-NEXT:    vaddps %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x58,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask.movshdup.256(<8 x float> %x0, <8 x float> %x1, i8 %x2)
  %res1 = call <8 x float> @llvm.x86.avx512.mask.movshdup.256(<8 x float> %x0, <8 x float> %x1, i8 -1)
  %res2 = call <8 x float> @llvm.x86.avx512.mask.movshdup.256(<8 x float> %x0, <8 x float> zeroinitializer, i8 %x2)
  %res3 = fadd <8 x float> %res, %res1
  %res4 = fadd <8 x float> %res2, %res3
  ret <8 x float> %res4
}
declare <2 x double> @llvm.x86.avx512.mask.movddup.128(<2 x double>, <2 x double>, i8)

define <2 x double>@test_int_x86_avx512_mask_movddup_128(<2 x double> %x0, <2 x double> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_movddup_128:
; X86:       # %bb.0:
; X86-NEXT:    vmovddup %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xfb,0x12,0xd0]
; X86-NEXT:    # xmm2 = xmm0[0,0]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vmovddup %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xff,0x09,0x12,0xc8]
; X86-NEXT:    # xmm1 {%k1} = xmm0[0,0]
; X86-NEXT:    vaddpd %xmm2, %xmm1, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0x58,0xca]
; X86-NEXT:    vmovddup %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0xff,0x89,0x12,0xc0]
; X86-NEXT:    # xmm0 {%k1} {z} = xmm0[0,0]
; X86-NEXT:    vaddpd %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x58,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_movddup_128:
; X64:       # %bb.0:
; X64-NEXT:    vmovddup %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xfb,0x12,0xd0]
; X64-NEXT:    # xmm2 = xmm0[0,0]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vmovddup %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xff,0x09,0x12,0xc8]
; X64-NEXT:    # xmm1 {%k1} = xmm0[0,0]
; X64-NEXT:    vaddpd %xmm2, %xmm1, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0x58,0xca]
; X64-NEXT:    vmovddup %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0xff,0x89,0x12,0xc0]
; X64-NEXT:    # xmm0 {%k1} {z} = xmm0[0,0]
; X64-NEXT:    vaddpd %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x58,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x double> @llvm.x86.avx512.mask.movddup.128(<2 x double> %x0, <2 x double> %x1, i8 %x2)
  %res1 = call <2 x double> @llvm.x86.avx512.mask.movddup.128(<2 x double> %x0, <2 x double> %x1, i8 -1)
  %res2 = call <2 x double> @llvm.x86.avx512.mask.movddup.128(<2 x double> %x0, <2 x double> zeroinitializer, i8 %x2)
  %res3 = fadd <2 x double> %res, %res1
  %res4 = fadd <2 x double> %res2, %res3
  ret <2 x double> %res4
}

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

define <4 x double>@test_int_x86_avx512_mask_movddup_256(<4 x double> %x0, <4 x double> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_movddup_256:
; X86:       # %bb.0:
; X86-NEXT:    vmovddup %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xff,0x12,0xd0]
; X86-NEXT:    # ymm2 = ymm0[0,0,2,2]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vmovddup %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0xff,0x29,0x12,0xc8]
; X86-NEXT:    # ymm1 {%k1} = ymm0[0,0,2,2]
; X86-NEXT:    vaddpd %ymm2, %ymm1, %ymm1 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0x58,0xca]
; X86-NEXT:    vmovddup %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0xff,0xa9,0x12,0xc0]
; X86-NEXT:    # ymm0 {%k1} {z} = ymm0[0,0,2,2]
; X86-NEXT:    vaddpd %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x58,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_movddup_256:
; X64:       # %bb.0:
; X64-NEXT:    vmovddup %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xff,0x12,0xd0]
; X64-NEXT:    # ymm2 = ymm0[0,0,2,2]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vmovddup %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0xff,0x29,0x12,0xc8]
; X64-NEXT:    # ymm1 {%k1} = ymm0[0,0,2,2]
; X64-NEXT:    vaddpd %ymm2, %ymm1, %ymm1 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0x58,0xca]
; X64-NEXT:    vmovddup %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0xff,0xa9,0x12,0xc0]
; X64-NEXT:    # ymm0 {%k1} {z} = ymm0[0,0,2,2]
; X64-NEXT:    vaddpd %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x58,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x double> @llvm.x86.avx512.mask.movddup.256(<4 x double> %x0, <4 x double> %x1, i8 %x2)
  %res1 = call <4 x double> @llvm.x86.avx512.mask.movddup.256(<4 x double> %x0, <4 x double> %x1, i8 -1)
  %res2 = call <4 x double> @llvm.x86.avx512.mask.movddup.256(<4 x double> %x0, <4 x double> zeroinitializer, i8 %x2)
  %res3 = fadd <4 x double> %res, %res1
  %res4 = fadd <4 x double> %res2, %res3
  ret <4 x double> %res4
}

declare <4 x double> @llvm.x86.avx512.mask.vpermil.pd.256(<4 x double>, i32, <4 x double>, i8)

define <4 x double>@test_int_x86_avx512_mask_vpermil_pd_256(<4 x double> %x0, <4 x double> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_vpermil_pd_256:
; X86:       # %bb.0:
; X86-NEXT:    vpermilpd $6, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x7d,0x05,0xd0,0x06]
; X86-NEXT:    # ymm2 = ymm0[0,1,3,2]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpermilpd $6, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x05,0xc8,0x06]
; X86-NEXT:    # ymm1 {%k1} = ymm0[0,1,3,2]
; X86-NEXT:    vpermilpd $6, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf3,0xfd,0xa9,0x05,0xc0,0x06]
; X86-NEXT:    # ymm0 {%k1} {z} = ymm0[0,1,3,2]
; X86-NEXT:    vaddpd %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0x58,0xc0]
; X86-NEXT:    vaddpd %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0x58,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_vpermil_pd_256:
; X64:       # %bb.0:
; X64-NEXT:    vpermilpd $6, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x7d,0x05,0xd0,0x06]
; X64-NEXT:    # ymm2 = ymm0[0,1,3,2]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpermilpd $6, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x05,0xc8,0x06]
; X64-NEXT:    # ymm1 {%k1} = ymm0[0,1,3,2]
; X64-NEXT:    vpermilpd $6, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf3,0xfd,0xa9,0x05,0xc0,0x06]
; X64-NEXT:    # ymm0 {%k1} {z} = ymm0[0,1,3,2]
; X64-NEXT:    vaddpd %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0x58,0xc0]
; X64-NEXT:    vaddpd %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0x58,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x double> @llvm.x86.avx512.mask.vpermil.pd.256(<4 x double> %x0, i32 22, <4 x double> %x2, i8 %x3)
  %res1 = call <4 x double> @llvm.x86.avx512.mask.vpermil.pd.256(<4 x double> %x0, i32 22, <4 x double> zeroinitializer, i8 %x3)
  %res2 = call <4 x double> @llvm.x86.avx512.mask.vpermil.pd.256(<4 x double> %x0, i32 22, <4 x double> %x2, i8 -1)
  %res3 = fadd <4 x double> %res, %res1
  %res4 = fadd <4 x double> %res2, %res3
  ret <4 x double> %res4
}

declare <2 x double> @llvm.x86.avx512.mask.vpermil.pd.128(<2 x double>, i32, <2 x double>, i8)

define <2 x double>@test_int_x86_avx512_mask_vpermil_pd_128(<2 x double> %x0, <2 x double> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_vpermil_pd_128:
; X86:       # %bb.0:
; X86-NEXT:    vpermilpd $1, %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x05,0xd0,0x01]
; X86-NEXT:    # xmm2 = xmm0[1,0]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpermilpd $1, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x05,0xc8,0x01]
; X86-NEXT:    # xmm1 {%k1} = xmm0[1,0]
; X86-NEXT:    vpermilpd $1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf3,0xfd,0x89,0x05,0xc0,0x01]
; X86-NEXT:    # xmm0 {%k1} {z} = xmm0[1,0]
; X86-NEXT:    vaddpd %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0x58,0xc0]
; X86-NEXT:    vaddpd %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x58,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_vpermil_pd_128:
; X64:       # %bb.0:
; X64-NEXT:    vpermilpd $1, %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x05,0xd0,0x01]
; X64-NEXT:    # xmm2 = xmm0[1,0]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpermilpd $1, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x05,0xc8,0x01]
; X64-NEXT:    # xmm1 {%k1} = xmm0[1,0]
; X64-NEXT:    vpermilpd $1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf3,0xfd,0x89,0x05,0xc0,0x01]
; X64-NEXT:    # xmm0 {%k1} {z} = xmm0[1,0]
; X64-NEXT:    vaddpd %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0x58,0xc0]
; X64-NEXT:    vaddpd %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x58,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x double> @llvm.x86.avx512.mask.vpermil.pd.128(<2 x double> %x0, i32 1, <2 x double> %x2, i8 %x3)
  %res1 = call <2 x double> @llvm.x86.avx512.mask.vpermil.pd.128(<2 x double> %x0, i32 1, <2 x double> zeroinitializer, i8 %x3)
  %res2 = call <2 x double> @llvm.x86.avx512.mask.vpermil.pd.128(<2 x double> %x0, i32 1, <2 x double> %x2, i8 -1)
  %res3 = fadd <2 x double> %res, %res1
  %res4 = fadd <2 x double> %res3, %res2
  ret <2 x double> %res4
}

declare <8 x float> @llvm.x86.avx512.mask.vpermil.ps.256(<8 x float>, i32, <8 x float>, i8)

define <8 x float>@test_int_x86_avx512_mask_vpermil_ps_256(<8 x float> %x0, <8 x float> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_vpermil_ps_256:
; X86:       # %bb.0:
; X86-NEXT:    vpermilps $22, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x7d,0x04,0xd0,0x16]
; X86-NEXT:    # ymm2 = ymm0[2,1,1,0,6,5,5,4]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpermilps $22, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x04,0xc8,0x16]
; X86-NEXT:    # ymm1 {%k1} = ymm0[2,1,1,0,6,5,5,4]
; X86-NEXT:    vpermilps $22, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0xa9,0x04,0xc0,0x16]
; X86-NEXT:    # ymm0 {%k1} {z} = ymm0[2,1,1,0,6,5,5,4]
; X86-NEXT:    vaddps %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf4,0x58,0xc0]
; X86-NEXT:    vaddps %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x58,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_vpermil_ps_256:
; X64:       # %bb.0:
; X64-NEXT:    vpermilps $22, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x7d,0x04,0xd0,0x16]
; X64-NEXT:    # ymm2 = ymm0[2,1,1,0,6,5,5,4]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpermilps $22, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x04,0xc8,0x16]
; X64-NEXT:    # ymm1 {%k1} = ymm0[2,1,1,0,6,5,5,4]
; X64-NEXT:    vpermilps $22, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0xa9,0x04,0xc0,0x16]
; X64-NEXT:    # ymm0 {%k1} {z} = ymm0[2,1,1,0,6,5,5,4]
; X64-NEXT:    vaddps %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf4,0x58,0xc0]
; X64-NEXT:    vaddps %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x58,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask.vpermil.ps.256(<8 x float> %x0, i32 22, <8 x float> %x2, i8 %x3)
  %res1 = call <8 x float> @llvm.x86.avx512.mask.vpermil.ps.256(<8 x float> %x0, i32 22, <8 x float> zeroinitializer, i8 %x3)
  %res2 = call <8 x float> @llvm.x86.avx512.mask.vpermil.ps.256(<8 x float> %x0, i32 22, <8 x float> %x2, i8 -1)
  %res3 = fadd <8 x float> %res, %res1
  %res4 = fadd <8 x float> %res3, %res2
  ret <8 x float> %res4
}

declare <4 x float> @llvm.x86.avx512.mask.vpermil.ps.128(<4 x float>, i32, <4 x float>, i8)

define <4 x float>@test_int_x86_avx512_mask_vpermil_ps_128(<4 x float> %x0, <4 x float> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_vpermil_ps_128:
; X86:       # %bb.0:
; X86-NEXT:    vpermilps $22, %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x04,0xd0,0x16]
; X86-NEXT:    # xmm2 = xmm0[2,1,1,0]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpermilps $22, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x04,0xc8,0x16]
; X86-NEXT:    # xmm1 {%k1} = xmm0[2,1,1,0]
; X86-NEXT:    vpermilps $22, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0x89,0x04,0xc0,0x16]
; X86-NEXT:    # xmm0 {%k1} {z} = xmm0[2,1,1,0]
; X86-NEXT:    vaddps %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf0,0x58,0xc0]
; X86-NEXT:    vaddps %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe8,0x58,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_vpermil_ps_128:
; X64:       # %bb.0:
; X64-NEXT:    vpermilps $22, %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x04,0xd0,0x16]
; X64-NEXT:    # xmm2 = xmm0[2,1,1,0]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpermilps $22, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x04,0xc8,0x16]
; X64-NEXT:    # xmm1 {%k1} = xmm0[2,1,1,0]
; X64-NEXT:    vpermilps $22, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0x89,0x04,0xc0,0x16]
; X64-NEXT:    # xmm0 {%k1} {z} = xmm0[2,1,1,0]
; X64-NEXT:    vaddps %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf0,0x58,0xc0]
; X64-NEXT:    vaddps %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe8,0x58,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask.vpermil.ps.128(<4 x float> %x0, i32 22, <4 x float> %x2, i8 %x3)
  %res1 = call <4 x float> @llvm.x86.avx512.mask.vpermil.ps.128(<4 x float> %x0, i32 22, <4 x float> zeroinitializer, i8 %x3)
  %res2 = call <4 x float> @llvm.x86.avx512.mask.vpermil.ps.128(<4 x float> %x0, i32 22, <4 x float> %x2, i8 -1)
  %res3 = fadd <4 x float> %res, %res1
  %res4 = fadd <4 x float> %res2, %res3
  ret <4 x float> %res4
}

declare <4 x double> @llvm.x86.avx512.mask.perm.df.256(<4 x double>, i32, <4 x double>, i8)

define <4 x double>@test_int_x86_avx512_mask_perm_df_256(<4 x double> %x0, i32 %x1, <4 x double> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_perm_df_256:
; X86:       # %bb.0:
; X86-NEXT:    vpermpd $3, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0xfd,0x01,0xd0,0x03]
; X86-NEXT:    # ymm2 = ymm0[3,0,0,0]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpermpd $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x01,0xc8,0x03]
; X86-NEXT:    # ymm1 {%k1} = ymm0[3,0,0,0]
; X86-NEXT:    vpermpd $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf3,0xfd,0xa9,0x01,0xc0,0x03]
; X86-NEXT:    # ymm0 {%k1} {z} = ymm0[3,0,0,0]
; X86-NEXT:    vaddpd %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0x58,0xc0]
; X86-NEXT:    vaddpd %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x58,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_perm_df_256:
; X64:       # %bb.0:
; X64-NEXT:    vpermpd $3, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0xfd,0x01,0xd0,0x03]
; X64-NEXT:    # ymm2 = ymm0[3,0,0,0]
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpermpd $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x01,0xc8,0x03]
; X64-NEXT:    # ymm1 {%k1} = ymm0[3,0,0,0]
; X64-NEXT:    vpermpd $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf3,0xfd,0xa9,0x01,0xc0,0x03]
; X64-NEXT:    # ymm0 {%k1} {z} = ymm0[3,0,0,0]
; X64-NEXT:    vaddpd %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0x58,0xc0]
; X64-NEXT:    vaddpd %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x58,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x double> @llvm.x86.avx512.mask.perm.df.256(<4 x double> %x0, i32 3, <4 x double> %x2, i8 %x3)
  %res1 = call <4 x double> @llvm.x86.avx512.mask.perm.df.256(<4 x double> %x0, i32 3, <4 x double> zeroinitializer, i8 %x3)
  %res2 = call <4 x double> @llvm.x86.avx512.mask.perm.df.256(<4 x double> %x0, i32 3, <4 x double> %x2, i8 -1)
  %res3 = fadd <4 x double> %res, %res1
  %res4 = fadd <4 x double> %res3, %res2
  ret <4 x double> %res4
}

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

define <4 x i64>@test_int_x86_avx512_mask_perm_di_256(<4 x i64> %x0, i32 %x1, <4 x i64> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_perm_di_256:
; X86:       # %bb.0:
; X86-NEXT:    vpermq $3, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0xfd,0x00,0xd0,0x03]
; X86-NEXT:    # ymm2 = ymm0[3,0,0,0]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpermq $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x00,0xc8,0x03]
; X86-NEXT:    # ymm1 {%k1} = ymm0[3,0,0,0]
; X86-NEXT:    vpermq $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf3,0xfd,0xa9,0x00,0xc0,0x03]
; X86-NEXT:    # ymm0 {%k1} {z} = ymm0[3,0,0,0]
; 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_perm_di_256:
; X64:       # %bb.0:
; X64-NEXT:    vpermq $3, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0xfd,0x00,0xd0,0x03]
; X64-NEXT:    # ymm2 = ymm0[3,0,0,0]
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpermq $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x00,0xc8,0x03]
; X64-NEXT:    # ymm1 {%k1} = ymm0[3,0,0,0]
; X64-NEXT:    vpermq $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf3,0xfd,0xa9,0x00,0xc0,0x03]
; X64-NEXT:    # ymm0 {%k1} {z} = ymm0[3,0,0,0]
; 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.perm.di.256(<4 x i64> %x0, i32 3, <4 x i64> %x2, i8 %x3)
  %res1 = call <4 x i64> @llvm.x86.avx512.mask.perm.di.256(<4 x i64> %x0, i32 3, <4 x i64> zeroinitializer, i8 %x3)
  %res2 = call <4 x i64> @llvm.x86.avx512.mask.perm.di.256(<4 x i64> %x0, i32 3, <4 x i64> %x2, i8 -1)
  %res3 = add <4 x i64> %res, %res1
  %res4 = add <4 x i64> %res3, %res2
  ret <4 x i64> %res4
}

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

define void@test_int_x86_avx512_mask_store_pd_128(i8* %ptr1, i8* %ptr2, <2 x double> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_store_pd_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:    kmovw %edx, %k1 # encoding: [0xc5,0xf8,0x92,0xca]
; X86-NEXT:    vmovapd %xmm0, (%ecx) {%k1} # encoding: [0x62,0xf1,0xfd,0x09,0x29,0x01]
; X86-NEXT:    vmovapd %xmm0, (%eax) # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x29,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_store_pd_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edx, %k1 # encoding: [0xc5,0xf8,0x92,0xca]
; X64-NEXT:    vmovapd %xmm0, (%rdi) {%k1} # encoding: [0x62,0xf1,0xfd,0x09,0x29,0x07]
; X64-NEXT:    vmovapd %xmm0, (%rsi) # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x29,0x06]
; X64-NEXT:    retq # encoding: [0xc3]
  call void @llvm.x86.avx512.mask.store.pd.128(i8* %ptr1, <2 x double> %x1, i8 %x2)
  call void @llvm.x86.avx512.mask.store.pd.128(i8* %ptr2, <2 x double> %x1, i8 -1)
  ret void
}

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

define void@test_int_x86_avx512_mask_store_pd_256(i8* %ptr1, i8* %ptr2, <4 x double> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_store_pd_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:    movzbl {{[0-9]+}}(%esp), %edx # encoding: [0x0f,0xb6,0x54,0x24,0x0c]
; X86-NEXT:    kmovw %edx, %k1 # encoding: [0xc5,0xf8,0x92,0xca]
; X86-NEXT:    vmovapd %ymm0, (%ecx) {%k1} # encoding: [0x62,0xf1,0xfd,0x29,0x29,0x01]
; X86-NEXT:    vmovapd %ymm0, (%eax) # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x29,0x00]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_store_pd_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edx, %k1 # encoding: [0xc5,0xf8,0x92,0xca]
; X64-NEXT:    vmovapd %ymm0, (%rdi) {%k1} # encoding: [0x62,0xf1,0xfd,0x29,0x29,0x07]
; X64-NEXT:    vmovapd %ymm0, (%rsi) # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x29,0x06]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
  call void @llvm.x86.avx512.mask.store.pd.256(i8* %ptr1, <4 x double> %x1, i8 %x2)
  call void @llvm.x86.avx512.mask.store.pd.256(i8* %ptr2, <4 x double> %x1, i8 -1)
  ret void
}

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

define void@test_int_x86_avx512_mask_storeu_pd_128(i8* %ptr1, i8* %ptr2, <2 x double> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_storeu_pd_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:    kmovw %edx, %k1 # encoding: [0xc5,0xf8,0x92,0xca]
; X86-NEXT:    vmovupd %xmm0, (%ecx) {%k1} # encoding: [0x62,0xf1,0xfd,0x09,0x11,0x01]
; X86-NEXT:    vmovupd %xmm0, (%eax) # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x11,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_storeu_pd_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edx, %k1 # encoding: [0xc5,0xf8,0x92,0xca]
; X64-NEXT:    vmovupd %xmm0, (%rdi) {%k1} # encoding: [0x62,0xf1,0xfd,0x09,0x11,0x07]
; X64-NEXT:    vmovupd %xmm0, (%rsi) # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x11,0x06]
; X64-NEXT:    retq # encoding: [0xc3]
  call void @llvm.x86.avx512.mask.storeu.pd.128(i8* %ptr1, <2 x double> %x1, i8 %x2)
  call void @llvm.x86.avx512.mask.storeu.pd.128(i8* %ptr2, <2 x double> %x1, i8 -1)
  ret void
}

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

define void@test_int_x86_avx512_mask_storeu_pd_256(i8* %ptr1, i8* %ptr2, <4 x double> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_storeu_pd_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:    movzbl {{[0-9]+}}(%esp), %edx # encoding: [0x0f,0xb6,0x54,0x24,0x0c]
; X86-NEXT:    kmovw %edx, %k1 # encoding: [0xc5,0xf8,0x92,0xca]
; X86-NEXT:    vmovupd %ymm0, (%ecx) {%k1} # encoding: [0x62,0xf1,0xfd,0x29,0x11,0x01]
; X86-NEXT:    vmovupd %ymm0, (%eax) # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x11,0x00]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_storeu_pd_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edx, %k1 # encoding: [0xc5,0xf8,0x92,0xca]
; X64-NEXT:    vmovupd %ymm0, (%rdi) {%k1} # encoding: [0x62,0xf1,0xfd,0x29,0x11,0x07]
; X64-NEXT:    vmovupd %ymm0, (%rsi) # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x11,0x06]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
  call void @llvm.x86.avx512.mask.storeu.pd.256(i8* %ptr1, <4 x double> %x1, i8 %x2)
  call void @llvm.x86.avx512.mask.storeu.pd.256(i8* %ptr2, <4 x double> %x1, i8 -1)
  ret void
}

declare void @llvm.x86.avx512.mask.store.ps.128(i8*, <4 x float>, i8)

define void@test_int_x86_avx512_mask_store_ps_128(i8* %ptr1, i8* %ptr2, <4 x float> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_store_ps_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:    kmovw %edx, %k1 # encoding: [0xc5,0xf8,0x92,0xca]
; X86-NEXT:    vmovaps %xmm0, (%ecx) {%k1} # encoding: [0x62,0xf1,0x7c,0x09,0x29,0x01]
; X86-NEXT:    vmovaps %xmm0, (%eax) # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x29,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_store_ps_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edx, %k1 # encoding: [0xc5,0xf8,0x92,0xca]
; X64-NEXT:    vmovaps %xmm0, (%rdi) {%k1} # encoding: [0x62,0xf1,0x7c,0x09,0x29,0x07]
; X64-NEXT:    vmovaps %xmm0, (%rsi) # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x29,0x06]
; X64-NEXT:    retq # encoding: [0xc3]
    call void @llvm.x86.avx512.mask.store.ps.128(i8* %ptr1, <4 x float> %x1, i8 %x2)
    call void @llvm.x86.avx512.mask.store.ps.128(i8* %ptr2, <4 x float> %x1, i8 -1)
    ret void
}

declare void @llvm.x86.avx512.mask.store.ps.256(i8*, <8 x float>, i8)

define void@test_int_x86_avx512_mask_store_ps_256(i8* %ptr1, i8* %ptr2, <8 x float> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_store_ps_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:    movzbl {{[0-9]+}}(%esp), %edx # encoding: [0x0f,0xb6,0x54,0x24,0x0c]
; X86-NEXT:    kmovw %edx, %k1 # encoding: [0xc5,0xf8,0x92,0xca]
; X86-NEXT:    vmovaps %ymm0, (%ecx) {%k1} # encoding: [0x62,0xf1,0x7c,0x29,0x29,0x01]
; X86-NEXT:    vmovaps %ymm0, (%eax) # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x29,0x00]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_store_ps_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edx, %k1 # encoding: [0xc5,0xf8,0x92,0xca]
; X64-NEXT:    vmovaps %ymm0, (%rdi) {%k1} # encoding: [0x62,0xf1,0x7c,0x29,0x29,0x07]
; X64-NEXT:    vmovaps %ymm0, (%rsi) # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x29,0x06]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
    call void @llvm.x86.avx512.mask.store.ps.256(i8* %ptr1, <8 x float> %x1, i8 %x2)
    call void @llvm.x86.avx512.mask.store.ps.256(i8* %ptr2, <8 x float> %x1, i8 -1)
    ret void
}

declare void @llvm.x86.avx512.mask.storeu.ps.128(i8*, <4 x float>, i8)

define void@test_int_x86_avx512_mask_storeu_ps_128(i8* %ptr1, i8* %ptr2, <4 x float> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_storeu_ps_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:    kmovw %edx, %k1 # encoding: [0xc5,0xf8,0x92,0xca]
; X86-NEXT:    vmovups %xmm0, (%ecx) {%k1} # encoding: [0x62,0xf1,0x7c,0x09,0x11,0x01]
; X86-NEXT:    vmovups %xmm0, (%eax) # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x11,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_storeu_ps_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edx, %k1 # encoding: [0xc5,0xf8,0x92,0xca]
; X64-NEXT:    vmovups %xmm0, (%rdi) {%k1} # encoding: [0x62,0xf1,0x7c,0x09,0x11,0x07]
; X64-NEXT:    vmovups %xmm0, (%rsi) # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x11,0x06]
; X64-NEXT:    retq # encoding: [0xc3]
    call void @llvm.x86.avx512.mask.storeu.ps.128(i8* %ptr1, <4 x float> %x1, i8 %x2)
    call void @llvm.x86.avx512.mask.storeu.ps.128(i8* %ptr2, <4 x float> %x1, i8 -1)
    ret void
}

declare void @llvm.x86.avx512.mask.storeu.ps.256(i8*, <8 x float>, i8)

define void@test_int_x86_avx512_mask_storeu_ps_256(i8* %ptr1, i8* %ptr2, <8 x float> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_storeu_ps_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:    movzbl {{[0-9]+}}(%esp), %edx # encoding: [0x0f,0xb6,0x54,0x24,0x0c]
; X86-NEXT:    kmovw %edx, %k1 # encoding: [0xc5,0xf8,0x92,0xca]
; X86-NEXT:    vmovups %ymm0, (%ecx) {%k1} # encoding: [0x62,0xf1,0x7c,0x29,0x11,0x01]
; X86-NEXT:    vmovups %ymm0, (%eax) # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x11,0x00]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_storeu_ps_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edx, %k1 # encoding: [0xc5,0xf8,0x92,0xca]
; X64-NEXT:    vmovups %ymm0, (%rdi) {%k1} # encoding: [0x62,0xf1,0x7c,0x29,0x11,0x07]
; X64-NEXT:    vmovups %ymm0, (%rsi) # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x11,0x06]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
    call void @llvm.x86.avx512.mask.storeu.ps.256(i8* %ptr1, <8 x float> %x1, i8 %x2)
    call void @llvm.x86.avx512.mask.storeu.ps.256(i8* %ptr2, <8 x float> %x1, i8 -1)
    ret void
}

declare void @llvm.x86.avx512.mask.storeu.q.128(i8*, <2 x i64>, i8)

define void@test_int_x86_avx512_mask_storeu_q_128(i8* %ptr1, i8* %ptr2, <2 x i64> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_storeu_q_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:    kmovw %edx, %k1 # encoding: [0xc5,0xf8,0x92,0xca]
; X86-NEXT:    vmovdqu64 %xmm0, (%ecx) {%k1} # encoding: [0x62,0xf1,0xfe,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_q_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edx, %k1 # encoding: [0xc5,0xf8,0x92,0xca]
; X64-NEXT:    vmovdqu64 %xmm0, (%rdi) {%k1} # encoding: [0x62,0xf1,0xfe,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.q.128(i8* %ptr1, <2 x i64> %x1, i8 %x2)
  call void @llvm.x86.avx512.mask.storeu.q.128(i8* %ptr2, <2 x i64> %x1, i8 -1)
  ret void
}

declare void @llvm.x86.avx512.mask.storeu.q.256(i8*, <4 x i64>, i8)

define void@test_int_x86_avx512_mask_storeu_q_256(i8* %ptr1, i8* %ptr2, <4 x i64> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_storeu_q_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:    movzbl {{[0-9]+}}(%esp), %edx # encoding: [0x0f,0xb6,0x54,0x24,0x0c]
; X86-NEXT:    kmovw %edx, %k1 # encoding: [0xc5,0xf8,0x92,0xca]
; X86-NEXT:    vmovdqu64 %ymm0, (%ecx) {%k1} # encoding: [0x62,0xf1,0xfe,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_q_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edx, %k1 # encoding: [0xc5,0xf8,0x92,0xca]
; X64-NEXT:    vmovdqu64 %ymm0, (%rdi) {%k1} # encoding: [0x62,0xf1,0xfe,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.q.256(i8* %ptr1, <4 x i64> %x1, i8 %x2)
  call void @llvm.x86.avx512.mask.storeu.q.256(i8* %ptr2, <4 x i64> %x1, i8 -1)
  ret void
}

declare void @llvm.x86.avx512.mask.storeu.d.128(i8*, <4 x i32>, i8)

define void@test_int_x86_avx512_mask_storeu_d_128(i8* %ptr1, i8* %ptr2, <4 x i32> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_storeu_d_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:    kmovw %edx, %k1 # encoding: [0xc5,0xf8,0x92,0xca]
; X86-NEXT:    vmovdqu32 %xmm0, (%ecx) {%k1} # encoding: [0x62,0xf1,0x7e,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_d_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edx, %k1 # encoding: [0xc5,0xf8,0x92,0xca]
; X64-NEXT:    vmovdqu32 %xmm0, (%rdi) {%k1} # encoding: [0x62,0xf1,0x7e,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.d.128(i8* %ptr1, <4 x i32> %x1, i8 %x2)
  call void @llvm.x86.avx512.mask.storeu.d.128(i8* %ptr2, <4 x i32> %x1, i8 -1)
  ret void
}

declare void @llvm.x86.avx512.mask.storeu.d.256(i8*, <8 x i32>, i8)

define void@test_int_x86_avx512_mask_storeu_d_256(i8* %ptr1, i8* %ptr2, <8 x i32> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_storeu_d_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:    movzbl {{[0-9]+}}(%esp), %edx # encoding: [0x0f,0xb6,0x54,0x24,0x0c]
; X86-NEXT:    kmovw %edx, %k1 # encoding: [0xc5,0xf8,0x92,0xca]
; X86-NEXT:    vmovdqu32 %ymm0, (%ecx) {%k1} # encoding: [0x62,0xf1,0x7e,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_d_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edx, %k1 # encoding: [0xc5,0xf8,0x92,0xca]
; X64-NEXT:    vmovdqu32 %ymm0, (%rdi) {%k1} # encoding: [0x62,0xf1,0x7e,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.d.256(i8* %ptr1, <8 x i32> %x1, i8 %x2)
  call void @llvm.x86.avx512.mask.storeu.d.256(i8* %ptr2, <8 x i32> %x1, i8 -1)
  ret void
}

declare void @llvm.x86.avx512.mask.store.q.128(i8*, <2 x i64>, i8)

define void@test_int_x86_avx512_mask_store_q_128(i8* %ptr1, i8* %ptr2, <2 x i64> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_store_q_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:    kmovw %edx, %k1 # encoding: [0xc5,0xf8,0x92,0xca]
; X86-NEXT:    vmovdqa64 %xmm0, (%ecx) {%k1} # encoding: [0x62,0xf1,0xfd,0x09,0x7f,0x01]
; X86-NEXT:    vmovdqa %xmm0, (%eax) # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x7f,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_store_q_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edx, %k1 # encoding: [0xc5,0xf8,0x92,0xca]
; X64-NEXT:    vmovdqa64 %xmm0, (%rdi) {%k1} # encoding: [0x62,0xf1,0xfd,0x09,0x7f,0x07]
; X64-NEXT:    vmovdqa %xmm0, (%rsi) # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x7f,0x06]
; X64-NEXT:    retq # encoding: [0xc3]
  call void @llvm.x86.avx512.mask.store.q.128(i8* %ptr1, <2 x i64> %x1, i8 %x2)
  call void @llvm.x86.avx512.mask.store.q.128(i8* %ptr2, <2 x i64> %x1, i8 -1)
  ret void
}

declare void @llvm.x86.avx512.mask.store.q.256(i8*, <4 x i64>, i8)

define void@test_int_x86_avx512_mask_store_q_256(i8* %ptr1, i8* %ptr2, <4 x i64> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_store_q_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:    movzbl {{[0-9]+}}(%esp), %edx # encoding: [0x0f,0xb6,0x54,0x24,0x0c]
; X86-NEXT:    kmovw %edx, %k1 # encoding: [0xc5,0xf8,0x92,0xca]
; X86-NEXT:    vmovdqa64 %ymm0, (%ecx) {%k1} # encoding: [0x62,0xf1,0xfd,0x29,0x7f,0x01]
; X86-NEXT:    vmovdqa %ymm0, (%eax) # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x7f,0x00]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_store_q_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edx, %k1 # encoding: [0xc5,0xf8,0x92,0xca]
; X64-NEXT:    vmovdqa64 %ymm0, (%rdi) {%k1} # encoding: [0x62,0xf1,0xfd,0x29,0x7f,0x07]
; X64-NEXT:    vmovdqa %ymm0, (%rsi) # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x7f,0x06]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
  call void @llvm.x86.avx512.mask.store.q.256(i8* %ptr1, <4 x i64> %x1, i8 %x2)
  call void @llvm.x86.avx512.mask.store.q.256(i8* %ptr2, <4 x i64> %x1, i8 -1)
  ret void
}

declare void @llvm.x86.avx512.mask.store.d.128(i8*, <4 x i32>, i8)

define void@test_int_x86_avx512_mask_store_d_128(i8* %ptr1, i8* %ptr2, <4 x i32> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_store_d_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:    kmovw %edx, %k1 # encoding: [0xc5,0xf8,0x92,0xca]
; X86-NEXT:    vmovdqa32 %xmm0, (%ecx) {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x7f,0x01]
; X86-NEXT:    vmovdqa %xmm0, (%eax) # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x7f,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_store_d_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edx, %k1 # encoding: [0xc5,0xf8,0x92,0xca]
; X64-NEXT:    vmovdqa32 %xmm0, (%rdi) {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x7f,0x07]
; X64-NEXT:    vmovdqa %xmm0, (%rsi) # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x7f,0x06]
; X64-NEXT:    retq # encoding: [0xc3]
  call void @llvm.x86.avx512.mask.store.d.128(i8* %ptr1, <4 x i32> %x1, i8 %x2)
  call void @llvm.x86.avx512.mask.store.d.128(i8* %ptr2, <4 x i32> %x1, i8 -1)
  ret void
}

declare void @llvm.x86.avx512.mask.store.d.256(i8*, <8 x i32>, i8)

define void@test_int_x86_avx512_mask_store_d_256(i8* %ptr1, i8* %ptr2, <8 x i32> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_store_d_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:    movzbl {{[0-9]+}}(%esp), %edx # encoding: [0x0f,0xb6,0x54,0x24,0x0c]
; X86-NEXT:    kmovw %edx, %k1 # encoding: [0xc5,0xf8,0x92,0xca]
; X86-NEXT:    vmovdqa32 %ymm0, (%ecx) {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x7f,0x01]
; X86-NEXT:    vmovdqa %ymm0, (%eax) # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x7f,0x00]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_store_d_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edx, %k1 # encoding: [0xc5,0xf8,0x92,0xca]
; X64-NEXT:    vmovdqa32 %ymm0, (%rdi) {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x7f,0x07]
; X64-NEXT:    vmovdqa %ymm0, (%rsi) # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x7f,0x06]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
  call void @llvm.x86.avx512.mask.store.d.256(i8* %ptr1, <8 x i32> %x1, i8 %x2)
  call void @llvm.x86.avx512.mask.store.d.256(i8* %ptr2, <8 x i32> %x1, i8 -1)
  ret void
}

define <8 x float> @test_mask_load_aligned_ps_256(<8 x float> %data, i8* %ptr, i8 %mask) {
; X86-LABEL: test_mask_load_aligned_ps_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vmovaps (%eax), %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x28,0x00]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
; X86-NEXT:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vmovaps (%eax), %ymm0 {%k1} # encoding: [0x62,0xf1,0x7c,0x29,0x28,0x00]
; X86-NEXT:    vmovaps (%eax), %ymm1 {%k1} {z} # encoding: [0x62,0xf1,0x7c,0xa9,0x28,0x08]
; X86-NEXT:    vaddps %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf4,0x58,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_load_aligned_ps_256:
; X64:       # %bb.0:
; X64-NEXT:    vmovaps (%rdi), %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x28,0x07]
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vmovaps (%rdi), %ymm0 {%k1} # encoding: [0x62,0xf1,0x7c,0x29,0x28,0x07]
; X64-NEXT:    vmovaps (%rdi), %ymm1 {%k1} {z} # encoding: [0x62,0xf1,0x7c,0xa9,0x28,0x0f]
; X64-NEXT:    vaddps %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf4,0x58,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask.load.ps.256(i8* %ptr, <8 x float> zeroinitializer, i8 -1)
  %res1 = call <8 x float> @llvm.x86.avx512.mask.load.ps.256(i8* %ptr, <8 x float> %res, i8 %mask)
  %res2 = call <8 x float> @llvm.x86.avx512.mask.load.ps.256(i8* %ptr, <8 x float> zeroinitializer, i8 %mask)
  %res4 = fadd <8 x float> %res2, %res1
  ret <8 x float> %res4
}

declare <8 x float> @llvm.x86.avx512.mask.load.ps.256(i8*, <8 x float>, i8)

define <8 x float> @test_mask_load_unaligned_ps_256(<8 x float> %data, i8* %ptr, i8 %mask) {
; X86-LABEL: test_mask_load_unaligned_ps_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vmovups (%eax), %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x10,0x00]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
; X86-NEXT:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vmovups (%eax), %ymm0 {%k1} # encoding: [0x62,0xf1,0x7c,0x29,0x10,0x00]
; X86-NEXT:    vmovups (%eax), %ymm1 {%k1} {z} # encoding: [0x62,0xf1,0x7c,0xa9,0x10,0x08]
; X86-NEXT:    vaddps %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf4,0x58,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_load_unaligned_ps_256:
; X64:       # %bb.0:
; X64-NEXT:    vmovups (%rdi), %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x10,0x07]
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vmovups (%rdi), %ymm0 {%k1} # encoding: [0x62,0xf1,0x7c,0x29,0x10,0x07]
; X64-NEXT:    vmovups (%rdi), %ymm1 {%k1} {z} # encoding: [0x62,0xf1,0x7c,0xa9,0x10,0x0f]
; X64-NEXT:    vaddps %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf4,0x58,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask.loadu.ps.256(i8* %ptr, <8 x float> zeroinitializer, i8 -1)
  %res1 = call <8 x float> @llvm.x86.avx512.mask.loadu.ps.256(i8* %ptr, <8 x float> %res, i8 %mask)
  %res2 = call <8 x float> @llvm.x86.avx512.mask.loadu.ps.256(i8* %ptr, <8 x float> zeroinitializer, i8 %mask)
  %res4 = fadd <8 x float> %res2, %res1
  ret <8 x float> %res4
}

declare <8 x float> @llvm.x86.avx512.mask.loadu.ps.256(i8*, <8 x float>, i8)

define <4 x double> @test_mask_load_aligned_pd_256(<4 x double> %data, i8* %ptr, i8 %mask) {
; X86-LABEL: test_mask_load_aligned_pd_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vmovapd (%eax), %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x28,0x00]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
; X86-NEXT:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vmovapd (%eax), %ymm0 {%k1} # encoding: [0x62,0xf1,0xfd,0x29,0x28,0x00]
; X86-NEXT:    vmovapd (%eax), %ymm1 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0xa9,0x28,0x08]
; X86-NEXT:    vaddpd %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0x58,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_load_aligned_pd_256:
; X64:       # %bb.0:
; X64-NEXT:    vmovapd (%rdi), %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x28,0x07]
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vmovapd (%rdi), %ymm0 {%k1} # encoding: [0x62,0xf1,0xfd,0x29,0x28,0x07]
; X64-NEXT:    vmovapd (%rdi), %ymm1 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0xa9,0x28,0x0f]
; X64-NEXT:    vaddpd %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0x58,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x double> @llvm.x86.avx512.mask.load.pd.256(i8* %ptr, <4 x double> zeroinitializer, i8 -1)
  %res1 = call <4 x double> @llvm.x86.avx512.mask.load.pd.256(i8* %ptr, <4 x double> %res, i8 %mask)
  %res2 = call <4 x double> @llvm.x86.avx512.mask.load.pd.256(i8* %ptr, <4 x double> zeroinitializer, i8 %mask)
  %res4 = fadd <4 x double> %res2, %res1
  ret <4 x double> %res4
}

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

define <4 x double> @test_mask_load_unaligned_pd_256(<4 x double> %data, i8* %ptr, i8 %mask) {
; X86-LABEL: test_mask_load_unaligned_pd_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vmovupd (%eax), %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x10,0x00]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
; X86-NEXT:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vmovupd (%eax), %ymm0 {%k1} # encoding: [0x62,0xf1,0xfd,0x29,0x10,0x00]
; X86-NEXT:    vmovupd (%eax), %ymm1 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0xa9,0x10,0x08]
; X86-NEXT:    vaddpd %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0x58,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_load_unaligned_pd_256:
; X64:       # %bb.0:
; X64-NEXT:    vmovupd (%rdi), %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x10,0x07]
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vmovupd (%rdi), %ymm0 {%k1} # encoding: [0x62,0xf1,0xfd,0x29,0x10,0x07]
; X64-NEXT:    vmovupd (%rdi), %ymm1 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0xa9,0x10,0x0f]
; X64-NEXT:    vaddpd %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0x58,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x double> @llvm.x86.avx512.mask.loadu.pd.256(i8* %ptr, <4 x double> zeroinitializer, i8 -1)
  %res1 = call <4 x double> @llvm.x86.avx512.mask.loadu.pd.256(i8* %ptr, <4 x double> %res, i8 %mask)
  %res2 = call <4 x double> @llvm.x86.avx512.mask.loadu.pd.256(i8* %ptr, <4 x double> zeroinitializer, i8 %mask)
  %res4 = fadd <4 x double> %res2, %res1
  ret <4 x double> %res4
}

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

define <4 x float> @test_mask_load_aligned_ps_128(<4 x float> %data, i8* %ptr, i8 %mask) {
; X86-LABEL: test_mask_load_aligned_ps_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vmovaps (%eax), %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x28,0x00]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
; X86-NEXT:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vmovaps (%eax), %xmm0 {%k1} # encoding: [0x62,0xf1,0x7c,0x09,0x28,0x00]
; X86-NEXT:    vmovaps (%eax), %xmm1 {%k1} {z} # encoding: [0x62,0xf1,0x7c,0x89,0x28,0x08]
; X86-NEXT:    vaddps %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf0,0x58,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_load_aligned_ps_128:
; X64:       # %bb.0:
; X64-NEXT:    vmovaps (%rdi), %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x28,0x07]
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vmovaps (%rdi), %xmm0 {%k1} # encoding: [0x62,0xf1,0x7c,0x09,0x28,0x07]
; X64-NEXT:    vmovaps (%rdi), %xmm1 {%k1} {z} # encoding: [0x62,0xf1,0x7c,0x89,0x28,0x0f]
; X64-NEXT:    vaddps %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf0,0x58,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask.load.ps.128(i8* %ptr, <4 x float> zeroinitializer, i8 -1)
  %res1 = call <4 x float> @llvm.x86.avx512.mask.load.ps.128(i8* %ptr, <4 x float> %res, i8 %mask)
  %res2 = call <4 x float> @llvm.x86.avx512.mask.load.ps.128(i8* %ptr, <4 x float> zeroinitializer, i8 %mask)
  %res4 = fadd <4 x float> %res2, %res1
  ret <4 x float> %res4
}

declare <4 x float> @llvm.x86.avx512.mask.load.ps.128(i8*, <4 x float>, i8)

define <4 x float> @test_mask_load_unaligned_ps_128(<4 x float> %data, i8* %ptr, i8 %mask) {
; X86-LABEL: test_mask_load_unaligned_ps_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vmovups (%eax), %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x10,0x00]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
; X86-NEXT:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vmovups (%eax), %xmm0 {%k1} # encoding: [0x62,0xf1,0x7c,0x09,0x10,0x00]
; X86-NEXT:    vmovups (%eax), %xmm1 {%k1} {z} # encoding: [0x62,0xf1,0x7c,0x89,0x10,0x08]
; X86-NEXT:    vaddps %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf0,0x58,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_load_unaligned_ps_128:
; X64:       # %bb.0:
; X64-NEXT:    vmovups (%rdi), %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x10,0x07]
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vmovups (%rdi), %xmm0 {%k1} # encoding: [0x62,0xf1,0x7c,0x09,0x10,0x07]
; X64-NEXT:    vmovups (%rdi), %xmm1 {%k1} {z} # encoding: [0x62,0xf1,0x7c,0x89,0x10,0x0f]
; X64-NEXT:    vaddps %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf0,0x58,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask.loadu.ps.128(i8* %ptr, <4 x float> zeroinitializer, i8 -1)
  %res1 = call <4 x float> @llvm.x86.avx512.mask.loadu.ps.128(i8* %ptr, <4 x float> %res, i8 %mask)
  %res2 = call <4 x float> @llvm.x86.avx512.mask.loadu.ps.128(i8* %ptr, <4 x float> zeroinitializer, i8 %mask)
  %res4 = fadd <4 x float> %res2, %res1
  ret <4 x float> %res4
}

declare <4 x float> @llvm.x86.avx512.mask.loadu.ps.128(i8*, <4 x float>, i8)

define <2 x double> @test_mask_load_aligned_pd_128(<2 x double> %data, i8* %ptr, i8 %mask) {
; X86-LABEL: test_mask_load_aligned_pd_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vmovapd (%eax), %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x28,0x00]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
; X86-NEXT:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vmovapd (%eax), %xmm0 {%k1} # encoding: [0x62,0xf1,0xfd,0x09,0x28,0x00]
; X86-NEXT:    vmovapd (%eax), %xmm1 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0x89,0x28,0x08]
; X86-NEXT:    vaddpd %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0x58,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_load_aligned_pd_128:
; X64:       # %bb.0:
; X64-NEXT:    vmovapd (%rdi), %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x28,0x07]
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vmovapd (%rdi), %xmm0 {%k1} # encoding: [0x62,0xf1,0xfd,0x09,0x28,0x07]
; X64-NEXT:    vmovapd (%rdi), %xmm1 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0x89,0x28,0x0f]
; X64-NEXT:    vaddpd %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0x58,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x double> @llvm.x86.avx512.mask.load.pd.128(i8* %ptr, <2 x double> zeroinitializer, i8 -1)
  %res1 = call <2 x double> @llvm.x86.avx512.mask.load.pd.128(i8* %ptr, <2 x double> %res, i8 %mask)
  %res2 = call <2 x double> @llvm.x86.avx512.mask.load.pd.128(i8* %ptr, <2 x double> zeroinitializer, i8 %mask)
  %res4 = fadd <2 x double> %res2, %res1
  ret <2 x double> %res4
}

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

define <2 x double> @test_mask_load_unaligned_pd_128(<2 x double> %data, i8* %ptr, i8 %mask) {
; X86-LABEL: test_mask_load_unaligned_pd_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vmovupd (%eax), %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x10,0x00]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
; X86-NEXT:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vmovupd (%eax), %xmm0 {%k1} # encoding: [0x62,0xf1,0xfd,0x09,0x10,0x00]
; X86-NEXT:    vmovupd (%eax), %xmm1 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0x89,0x10,0x08]
; X86-NEXT:    vaddpd %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0x58,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_load_unaligned_pd_128:
; X64:       # %bb.0:
; X64-NEXT:    vmovupd (%rdi), %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x10,0x07]
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vmovupd (%rdi), %xmm0 {%k1} # encoding: [0x62,0xf1,0xfd,0x09,0x10,0x07]
; X64-NEXT:    vmovupd (%rdi), %xmm1 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0x89,0x10,0x0f]
; X64-NEXT:    vaddpd %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0x58,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x double> @llvm.x86.avx512.mask.loadu.pd.128(i8* %ptr, <2 x double> zeroinitializer, i8 -1)
  %res1 = call <2 x double> @llvm.x86.avx512.mask.loadu.pd.128(i8* %ptr, <2 x double> %res, i8 %mask)
  %res2 = call <2 x double> @llvm.x86.avx512.mask.loadu.pd.128(i8* %ptr, <2 x double> zeroinitializer, i8 %mask)
  %res4 = fadd <2 x double> %res2, %res1
  ret <2 x double> %res4
}

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

declare <4 x i32> @llvm.x86.avx512.mask.loadu.d.128(i8*, <4 x i32>, i8)

define <4 x i32> @test_mask_load_unaligned_d_128(i8* %ptr, i8* %ptr2, <4 x i32> %data, i8 %mask) {
; X86-LABEL: test_mask_load_unaligned_d_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:    kmovw %edx, %k1 # encoding: [0xc5,0xf8,0x92,0xca]
; X86-NEXT:    vmovdqu32 (%eax), %xmm0 {%k1} # encoding: [0x62,0xf1,0x7e,0x09,0x6f,0x00]
; X86-NEXT:    vmovdqu32 (%ecx), %xmm1 {%k1} {z} # encoding: [0x62,0xf1,0x7e,0x89,0x6f,0x09]
; X86-NEXT:    vpaddd %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_load_unaligned_d_128:
; X64:       # %bb.0:
; X64-NEXT:    vmovdqu (%rdi), %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xfa,0x6f,0x07]
; X64-NEXT:    kmovw %edx, %k1 # encoding: [0xc5,0xf8,0x92,0xca]
; X64-NEXT:    vmovdqu32 (%rsi), %xmm0 {%k1} # encoding: [0x62,0xf1,0x7e,0x09,0x6f,0x06]
; X64-NEXT:    vmovdqu32 (%rdi), %xmm1 {%k1} {z} # encoding: [0x62,0xf1,0x7e,0x89,0x6f,0x0f]
; X64-NEXT:    vpaddd %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i32> @llvm.x86.avx512.mask.loadu.d.128(i8* %ptr, <4 x i32> zeroinitializer, i8 -1)
  %res1 = call <4 x i32> @llvm.x86.avx512.mask.loadu.d.128(i8* %ptr2, <4 x i32> %res, i8 %mask)
  %res2 = call <4 x i32> @llvm.x86.avx512.mask.loadu.d.128(i8* %ptr, <4 x i32> zeroinitializer, i8 %mask)
  %res4 = add <4 x i32> %res2, %res1
  ret <4 x i32> %res4
}

declare <8 x i32> @llvm.x86.avx512.mask.loadu.d.256(i8*, <8 x i32>, i8)

define <8 x i32> @test_mask_load_unaligned_d_256(i8* %ptr, i8* %ptr2, <8 x i32> %data, i8 %mask) {
; X86-LABEL: test_mask_load_unaligned_d_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:    movzbl {{[0-9]+}}(%esp), %edx # encoding: [0x0f,0xb6,0x54,0x24,0x0c]
; X86-NEXT:    kmovw %edx, %k1 # encoding: [0xc5,0xf8,0x92,0xca]
; X86-NEXT:    vmovdqu32 (%eax), %ymm0 {%k1} # encoding: [0x62,0xf1,0x7e,0x29,0x6f,0x00]
; X86-NEXT:    vmovdqu32 (%ecx), %ymm1 {%k1} {z} # encoding: [0x62,0xf1,0x7e,0xa9,0x6f,0x09]
; X86-NEXT:    vpaddd %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_load_unaligned_d_256:
; X64:       # %bb.0:
; X64-NEXT:    vmovdqu (%rdi), %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfe,0x6f,0x07]
; X64-NEXT:    kmovw %edx, %k1 # encoding: [0xc5,0xf8,0x92,0xca]
; X64-NEXT:    vmovdqu32 (%rsi), %ymm0 {%k1} # encoding: [0x62,0xf1,0x7e,0x29,0x6f,0x06]
; X64-NEXT:    vmovdqu32 (%rdi), %ymm1 {%k1} {z} # encoding: [0x62,0xf1,0x7e,0xa9,0x6f,0x0f]
; X64-NEXT:    vpaddd %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x i32> @llvm.x86.avx512.mask.loadu.d.256(i8* %ptr, <8 x i32> zeroinitializer, i8 -1)
  %res1 = call <8 x i32> @llvm.x86.avx512.mask.loadu.d.256(i8* %ptr2, <8 x i32> %res, i8 %mask)
  %res2 = call <8 x i32> @llvm.x86.avx512.mask.loadu.d.256(i8* %ptr, <8 x i32> zeroinitializer, i8 %mask)
  %res4 = add <8 x i32> %res2, %res1
  ret <8 x i32> %res4
}

declare <2 x i64> @llvm.x86.avx512.mask.loadu.q.128(i8*, <2 x i64>, i8)

define <2 x i64> @test_mask_load_unaligned_q_128(i8* %ptr, i8* %ptr2, <2 x i64> %data, i8 %mask) {
; X86-LABEL: test_mask_load_unaligned_q_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:    kmovw %edx, %k1 # encoding: [0xc5,0xf8,0x92,0xca]
; X86-NEXT:    vmovdqu64 (%eax), %xmm0 {%k1} # encoding: [0x62,0xf1,0xfe,0x09,0x6f,0x00]
; X86-NEXT:    vmovdqu64 (%ecx), %xmm1 {%k1} {z} # encoding: [0x62,0xf1,0xfe,0x89,0x6f,0x09]
; X86-NEXT:    vpaddq %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xd4,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_load_unaligned_q_128:
; X64:       # %bb.0:
; X64-NEXT:    vmovdqu (%rdi), %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xfa,0x6f,0x07]
; X64-NEXT:    kmovw %edx, %k1 # encoding: [0xc5,0xf8,0x92,0xca]
; X64-NEXT:    vmovdqu64 (%rsi), %xmm0 {%k1} # encoding: [0x62,0xf1,0xfe,0x09,0x6f,0x06]
; X64-NEXT:    vmovdqu64 (%rdi), %xmm1 {%k1} {z} # encoding: [0x62,0xf1,0xfe,0x89,0x6f,0x0f]
; 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.loadu.q.128(i8* %ptr, <2 x i64> zeroinitializer, i8 -1)
  %res1 = call <2 x i64> @llvm.x86.avx512.mask.loadu.q.128(i8* %ptr2, <2 x i64> %res, i8 %mask)
  %res2 = call <2 x i64> @llvm.x86.avx512.mask.loadu.q.128(i8* %ptr, <2 x i64> zeroinitializer, i8 %mask)
  %res4 = add <2 x i64> %res2, %res1
  ret <2 x i64> %res4
}

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

define <4 x i64> @test_mask_load_unaligned_q_256(i8* %ptr, i8* %ptr2, <4 x i64> %data, i8 %mask) {
; X86-LABEL: test_mask_load_unaligned_q_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:    movzbl {{[0-9]+}}(%esp), %edx # encoding: [0x0f,0xb6,0x54,0x24,0x0c]
; X86-NEXT:    kmovw %edx, %k1 # encoding: [0xc5,0xf8,0x92,0xca]
; X86-NEXT:    vmovdqu64 (%eax), %ymm0 {%k1} # encoding: [0x62,0xf1,0xfe,0x29,0x6f,0x00]
; X86-NEXT:    vmovdqu64 (%ecx), %ymm1 {%k1} {z} # encoding: [0x62,0xf1,0xfe,0xa9,0x6f,0x09]
; X86-NEXT:    vpaddq %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xd4,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_load_unaligned_q_256:
; X64:       # %bb.0:
; X64-NEXT:    vmovdqu (%rdi), %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfe,0x6f,0x07]
; X64-NEXT:    kmovw %edx, %k1 # encoding: [0xc5,0xf8,0x92,0xca]
; X64-NEXT:    vmovdqu64 (%rsi), %ymm0 {%k1} # encoding: [0x62,0xf1,0xfe,0x29,0x6f,0x06]
; X64-NEXT:    vmovdqu64 (%rdi), %ymm1 {%k1} {z} # encoding: [0x62,0xf1,0xfe,0xa9,0x6f,0x0f]
; 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.loadu.q.256(i8* %ptr, <4 x i64> zeroinitializer, i8 -1)
  %res1 = call <4 x i64> @llvm.x86.avx512.mask.loadu.q.256(i8* %ptr2, <4 x i64> %res, i8 %mask)
  %res2 = call <4 x i64> @llvm.x86.avx512.mask.loadu.q.256(i8* %ptr, <4 x i64> zeroinitializer, i8 %mask)
  %res4 = add <4 x i64> %res2, %res1
  ret <4 x i64> %res4
}

declare <4 x i32> @llvm.x86.avx512.mask.load.d.128(i8*, <4 x i32>, i8)

define <4 x i32> @test_mask_load_aligned_d_128(<4 x i32> %data, i8* %ptr, i8 %mask) {
; X86-LABEL: test_mask_load_aligned_d_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vmovdqa (%eax), %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0x00]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
; X86-NEXT:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vmovdqa32 (%eax), %xmm0 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x6f,0x00]
; X86-NEXT:    vmovdqa32 (%eax), %xmm1 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x6f,0x08]
; X86-NEXT:    vpaddd %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_load_aligned_d_128:
; X64:       # %bb.0:
; X64-NEXT:    vmovdqa (%rdi), %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0x07]
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vmovdqa32 (%rdi), %xmm0 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x6f,0x07]
; X64-NEXT:    vmovdqa32 (%rdi), %xmm1 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x6f,0x0f]
; X64-NEXT:    vpaddd %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i32> @llvm.x86.avx512.mask.load.d.128(i8* %ptr, <4 x i32> zeroinitializer, i8 -1)
  %res1 = call <4 x i32> @llvm.x86.avx512.mask.load.d.128(i8* %ptr, <4 x i32> %res, i8 %mask)
  %res2 = call <4 x i32> @llvm.x86.avx512.mask.load.d.128(i8* %ptr, <4 x i32> zeroinitializer, i8 %mask)
  %res4 = add <4 x i32> %res2, %res1
  ret <4 x i32> %res4
}

declare <8 x i32> @llvm.x86.avx512.mask.load.d.256(i8*, <8 x i32>, i8)

define <8 x i32> @test_mask_load_aligned_d_256(<8 x i32> %data, i8* %ptr, i8 %mask) {
; X86-LABEL: test_mask_load_aligned_d_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vmovdqa (%eax), %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0x00]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
; X86-NEXT:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vmovdqa32 (%eax), %ymm0 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x6f,0x00]
; X86-NEXT:    vmovdqa32 (%eax), %ymm1 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x6f,0x08]
; X86-NEXT:    vpaddd %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_load_aligned_d_256:
; X64:       # %bb.0:
; X64-NEXT:    vmovdqa (%rdi), %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0x07]
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vmovdqa32 (%rdi), %ymm0 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x6f,0x07]
; X64-NEXT:    vmovdqa32 (%rdi), %ymm1 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x6f,0x0f]
; X64-NEXT:    vpaddd %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x i32> @llvm.x86.avx512.mask.load.d.256(i8* %ptr, <8 x i32> zeroinitializer, i8 -1)
  %res1 = call <8 x i32> @llvm.x86.avx512.mask.load.d.256(i8* %ptr, <8 x i32> %res, i8 %mask)
  %res2 = call <8 x i32> @llvm.x86.avx512.mask.load.d.256(i8* %ptr, <8 x i32> zeroinitializer, i8 %mask)
  %res4 = add <8 x i32> %res2, %res1
  ret <8 x i32> %res4
}

declare <2 x i64> @llvm.x86.avx512.mask.load.q.128(i8*, <2 x i64>, i8)

define <2 x i64> @test_mask_load_aligned_q_128(<2 x i64> %data, i8* %ptr, i8 %mask) {
; X86-LABEL: test_mask_load_aligned_q_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vmovdqa (%eax), %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0x00]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
; X86-NEXT:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vmovdqa64 (%eax), %xmm0 {%k1} # encoding: [0x62,0xf1,0xfd,0x09,0x6f,0x00]
; X86-NEXT:    vmovdqa64 (%eax), %xmm1 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0x89,0x6f,0x08]
; X86-NEXT:    vpaddq %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xd4,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_load_aligned_q_128:
; X64:       # %bb.0:
; X64-NEXT:    vmovdqa (%rdi), %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0x07]
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vmovdqa64 (%rdi), %xmm0 {%k1} # encoding: [0x62,0xf1,0xfd,0x09,0x6f,0x07]
; X64-NEXT:    vmovdqa64 (%rdi), %xmm1 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0x89,0x6f,0x0f]
; 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.load.q.128(i8* %ptr, <2 x i64> zeroinitializer, i8 -1)
  %res1 = call <2 x i64> @llvm.x86.avx512.mask.load.q.128(i8* %ptr, <2 x i64> %res, i8 %mask)
  %res2 = call <2 x i64> @llvm.x86.avx512.mask.load.q.128(i8* %ptr, <2 x i64> zeroinitializer, i8 %mask)
  %res4 = add <2 x i64> %res2, %res1
  ret <2 x i64> %res4
}

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

define <4 x i64> @test_mask_load_aligned_q_256(<4 x i64> %data, i8* %ptr, i8 %mask) {
; X86-LABEL: test_mask_load_aligned_q_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vmovdqa (%eax), %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0x00]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
; X86-NEXT:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vmovdqa64 (%eax), %ymm0 {%k1} # encoding: [0x62,0xf1,0xfd,0x29,0x6f,0x00]
; X86-NEXT:    vmovdqa64 (%eax), %ymm1 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0xa9,0x6f,0x08]
; X86-NEXT:    vpaddq %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xd4,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_load_aligned_q_256:
; X64:       # %bb.0:
; X64-NEXT:    vmovdqa (%rdi), %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0x07]
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vmovdqa64 (%rdi), %ymm0 {%k1} # encoding: [0x62,0xf1,0xfd,0x29,0x6f,0x07]
; X64-NEXT:    vmovdqa64 (%rdi), %ymm1 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0xa9,0x6f,0x0f]
; 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.load.q.256(i8* %ptr, <4 x i64> zeroinitializer, i8 -1)
  %res1 = call <4 x i64> @llvm.x86.avx512.mask.load.q.256(i8* %ptr, <4 x i64> %res, i8 %mask)
  %res2 = call <4 x i64> @llvm.x86.avx512.mask.load.q.256(i8* %ptr, <4 x i64> zeroinitializer, i8 %mask)
  %res4 = add <4 x i64> %res2, %res1
  ret <4 x i64> %res4
}

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

define <4 x i32>@test_int_x86_avx512_mask_pshuf_d_128(<4 x i32> %x0, i32 %x1, <4 x i32> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_pshuf_d_128:
; X86:       # %bb.0:
; X86-NEXT:    vpshufd $3, %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x70,0xd0,0x03]
; X86-NEXT:    # xmm2 = xmm0[3,0,0,0]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpshufd $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x70,0xc8,0x03]
; X86-NEXT:    # xmm1 {%k1} = xmm0[3,0,0,0]
; X86-NEXT:    vpshufd $3, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x70,0xc0,0x03]
; X86-NEXT:    # xmm0 {%k1} {z} = xmm0[3,0,0,0]
; X86-NEXT:    vpaddd %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfe,0xc2]
; X86-NEXT:    vpaddd %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pshuf_d_128:
; X64:       # %bb.0:
; X64-NEXT:    vpshufd $3, %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x70,0xd0,0x03]
; X64-NEXT:    # xmm2 = xmm0[3,0,0,0]
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpshufd $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x70,0xc8,0x03]
; X64-NEXT:    # xmm1 {%k1} = xmm0[3,0,0,0]
; X64-NEXT:    vpshufd $3, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x70,0xc0,0x03]
; X64-NEXT:    # xmm0 {%k1} {z} = xmm0[3,0,0,0]
; X64-NEXT:    vpaddd %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfe,0xc2]
; X64-NEXT:    vpaddd %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
	%res = call <4 x i32> @llvm.x86.avx512.mask.pshuf.d.128(<4 x i32> %x0, i32 3, <4 x i32> %x2, i8 %x3)
	%res1 = call <4 x i32> @llvm.x86.avx512.mask.pshuf.d.128(<4 x i32> %x0, i32 3, <4 x i32> zeroinitializer, i8 %x3)
	%res2 = call <4 x i32> @llvm.x86.avx512.mask.pshuf.d.128(<4 x i32> %x0, i32 3, <4 x i32> %x2, i8 -1)
	%res3 = add <4 x i32> %res, %res1
	%res4 = add <4 x i32> %res3, %res2
	ret <4 x i32> %res4
}

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

define <8 x i32>@test_int_x86_avx512_mask_pshuf_d_256(<8 x i32> %x0, i32 %x1, <8 x i32> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_pshuf_d_256:
; X86:       # %bb.0:
; X86-NEXT:    vpshufd $3, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x70,0xd0,0x03]
; X86-NEXT:    # ymm2 = ymm0[3,0,0,0,7,4,4,4]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpshufd $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x70,0xc8,0x03]
; X86-NEXT:    # ymm1 {%k1} = ymm0[3,0,0,0,7,4,4,4]
; X86-NEXT:    vpshufd $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x70,0xc0,0x03]
; X86-NEXT:    # ymm0 {%k1} {z} = ymm0[3,0,0,0,7,4,4,4]
; X86-NEXT:    vpaddd %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfe,0xc2]
; X86-NEXT:    vpaddd %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pshuf_d_256:
; X64:       # %bb.0:
; X64-NEXT:    vpshufd $3, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x70,0xd0,0x03]
; X64-NEXT:    # ymm2 = ymm0[3,0,0,0,7,4,4,4]
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpshufd $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x70,0xc8,0x03]
; X64-NEXT:    # ymm1 {%k1} = ymm0[3,0,0,0,7,4,4,4]
; X64-NEXT:    vpshufd $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x70,0xc0,0x03]
; X64-NEXT:    # ymm0 {%k1} {z} = ymm0[3,0,0,0,7,4,4,4]
; X64-NEXT:    vpaddd %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfe,0xc2]
; X64-NEXT:    vpaddd %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
	%res = call <8 x i32> @llvm.x86.avx512.mask.pshuf.d.256(<8 x i32> %x0, i32 3, <8 x i32> %x2, i8 %x3)
	%res1 = call <8 x i32> @llvm.x86.avx512.mask.pshuf.d.256(<8 x i32> %x0, i32 3, <8 x i32> zeroinitializer, i8 %x3)
	%res2 = call <8 x i32> @llvm.x86.avx512.mask.pshuf.d.256(<8 x i32> %x0, i32 3, <8 x i32> %x2, i8 -1)
	%res3 = add <8 x i32> %res, %res1
	%res4 = add <8 x i32> %res3, %res2
	ret <8 x i32> %res4
}

define i8 @test_pcmpeq_d_256(<8 x i32> %a, <8 x i32> %b) {
; CHECK-LABEL: test_pcmpeq_d_256:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vpcmpeqd %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x76,0xc1]
; CHECK-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; CHECK-NEXT:    # kill: def $al killed $al killed $eax
; CHECK-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call i8 @llvm.x86.avx512.mask.pcmpeq.d.256(<8 x i32> %a, <8 x i32> %b, i8 -1)
  ret i8 %res
}

define i8 @test_mask_pcmpeq_d_256(<8 x i32> %a, <8 x i32> %b, i8 %mask) {
; X86-LABEL: test_mask_pcmpeq_d_256:
; X86:       # %bb.0:
; X86-NEXT:    vpcmpeqd %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x76,0xc1]
; X86-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; X86-NEXT:    andb {{[0-9]+}}(%esp), %al # encoding: [0x22,0x44,0x24,0x04]
; X86-NEXT:    # kill: def $al killed $al killed $eax
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_pcmpeq_d_256:
; X64:       # %bb.0:
; X64-NEXT:    vpcmpeqd %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x76,0xc1]
; X64-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; X64-NEXT:    andb %dil, %al # encoding: [0x40,0x20,0xf8]
; X64-NEXT:    # kill: def $al killed $al killed $eax
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call i8 @llvm.x86.avx512.mask.pcmpeq.d.256(<8 x i32> %a, <8 x i32> %b, i8 %mask)
  ret i8 %res
}

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

define i8 @test_pcmpeq_q_256(<4 x i64> %a, <4 x i64> %b) {
; CHECK-LABEL: test_pcmpeq_q_256:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vpcmpeqq %ymm1, %ymm0, %k0 # encoding: [0x62,0xf2,0xfd,0x28,0x29,0xc1]
; CHECK-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; CHECK-NEXT:    # kill: def $al killed $al killed $eax
; CHECK-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call i8 @llvm.x86.avx512.mask.pcmpeq.q.256(<4 x i64> %a, <4 x i64> %b, i8 -1)
  ret i8 %res
}

define i8 @test_mask_pcmpeq_q_256(<4 x i64> %a, <4 x i64> %b, i8 %mask) {
; X86-LABEL: test_mask_pcmpeq_q_256:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpcmpeqq %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x29,0xc1]
; X86-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; X86-NEXT:    # kill: def $al killed $al killed $eax
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_pcmpeq_q_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpcmpeqq %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x29,0xc1]
; X64-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; X64-NEXT:    # kill: def $al killed $al killed $eax
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call i8 @llvm.x86.avx512.mask.pcmpeq.q.256(<4 x i64> %a, <4 x i64> %b, i8 %mask)
  ret i8 %res
}

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

define i8 @test_pcmpgt_d_256(<8 x i32> %a, <8 x i32> %b) {
; CHECK-LABEL: test_pcmpgt_d_256:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vpcmpgtd %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x66,0xc1]
; CHECK-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; CHECK-NEXT:    # kill: def $al killed $al killed $eax
; CHECK-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call i8 @llvm.x86.avx512.mask.pcmpgt.d.256(<8 x i32> %a, <8 x i32> %b, i8 -1)
  ret i8 %res
}

define i8 @test_mask_pcmpgt_d_256(<8 x i32> %a, <8 x i32> %b, i8 %mask) {
; X86-LABEL: test_mask_pcmpgt_d_256:
; X86:       # %bb.0:
; X86-NEXT:    vpcmpgtd %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x66,0xc1]
; X86-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; X86-NEXT:    andb {{[0-9]+}}(%esp), %al # encoding: [0x22,0x44,0x24,0x04]
; X86-NEXT:    # kill: def $al killed $al killed $eax
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_pcmpgt_d_256:
; X64:       # %bb.0:
; X64-NEXT:    vpcmpgtd %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x66,0xc1]
; X64-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; X64-NEXT:    andb %dil, %al # encoding: [0x40,0x20,0xf8]
; X64-NEXT:    # kill: def $al killed $al killed $eax
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call i8 @llvm.x86.avx512.mask.pcmpgt.d.256(<8 x i32> %a, <8 x i32> %b, i8 %mask)
  ret i8 %res
}

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

define i8 @test_pcmpgt_q_256(<4 x i64> %a, <4 x i64> %b) {
; CHECK-LABEL: test_pcmpgt_q_256:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vpcmpgtq %ymm1, %ymm0, %k0 # encoding: [0x62,0xf2,0xfd,0x28,0x37,0xc1]
; CHECK-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; CHECK-NEXT:    # kill: def $al killed $al killed $eax
; CHECK-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call i8 @llvm.x86.avx512.mask.pcmpgt.q.256(<4 x i64> %a, <4 x i64> %b, i8 -1)
  ret i8 %res
}

define i8 @test_mask_pcmpgt_q_256(<4 x i64> %a, <4 x i64> %b, i8 %mask) {
; X86-LABEL: test_mask_pcmpgt_q_256:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpcmpgtq %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x37,0xc1]
; X86-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; X86-NEXT:    # kill: def $al killed $al killed $eax
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_pcmpgt_q_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpcmpgtq %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x37,0xc1]
; X64-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; X64-NEXT:    # kill: def $al killed $al killed $eax
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call i8 @llvm.x86.avx512.mask.pcmpgt.q.256(<4 x i64> %a, <4 x i64> %b, i8 %mask)
  ret i8 %res
}

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

define i8 @test_pcmpeq_d_128(<4 x i32> %a, <4 x i32> %b) {
; CHECK-LABEL: test_pcmpeq_d_128:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vpcmpeqd %xmm1, %xmm0, %k0 # encoding: [0x62,0xf1,0x7d,0x08,0x76,0xc1]
; CHECK-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; CHECK-NEXT:    # kill: def $al killed $al killed $eax
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call i8 @llvm.x86.avx512.mask.pcmpeq.d.128(<4 x i32> %a, <4 x i32> %b, i8 -1)
  ret i8 %res
}

define i8 @test_mask_pcmpeq_d_128(<4 x i32> %a, <4 x i32> %b, i8 %mask) {
; X86-LABEL: test_mask_pcmpeq_d_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpcmpeqd %xmm1, %xmm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x76,0xc1]
; X86-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; X86-NEXT:    # kill: def $al killed $al killed $eax
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_pcmpeq_d_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpcmpeqd %xmm1, %xmm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x76,0xc1]
; X64-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; X64-NEXT:    # kill: def $al killed $al killed $eax
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call i8 @llvm.x86.avx512.mask.pcmpeq.d.128(<4 x i32> %a, <4 x i32> %b, i8 %mask)
  ret i8 %res
}

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

define i8 @test_pcmpeq_q_128(<2 x i64> %a, <2 x i64> %b) {
; CHECK-LABEL: test_pcmpeq_q_128:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vpcmpeqq %xmm1, %xmm0, %k0 # encoding: [0x62,0xf2,0xfd,0x08,0x29,0xc1]
; CHECK-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; CHECK-NEXT:    # kill: def $al killed $al killed $eax
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call i8 @llvm.x86.avx512.mask.pcmpeq.q.128(<2 x i64> %a, <2 x i64> %b, i8 -1)
  ret i8 %res
}

define i8 @test_mask_pcmpeq_q_128(<2 x i64> %a, <2 x i64> %b, i8 %mask) {
; X86-LABEL: test_mask_pcmpeq_q_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpcmpeqq %xmm1, %xmm0, %k0 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x29,0xc1]
; X86-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; X86-NEXT:    # kill: def $al killed $al killed $eax
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_pcmpeq_q_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpcmpeqq %xmm1, %xmm0, %k0 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x29,0xc1]
; X64-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; X64-NEXT:    # kill: def $al killed $al killed $eax
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call i8 @llvm.x86.avx512.mask.pcmpeq.q.128(<2 x i64> %a, <2 x i64> %b, i8 %mask)
  ret i8 %res
}

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

define i8 @test_pcmpgt_d_128(<4 x i32> %a, <4 x i32> %b) {
; CHECK-LABEL: test_pcmpgt_d_128:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vpcmpgtd %xmm1, %xmm0, %k0 # encoding: [0x62,0xf1,0x7d,0x08,0x66,0xc1]
; CHECK-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; CHECK-NEXT:    # kill: def $al killed $al killed $eax
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call i8 @llvm.x86.avx512.mask.pcmpgt.d.128(<4 x i32> %a, <4 x i32> %b, i8 -1)
  ret i8 %res
}

define i8 @test_mask_pcmpgt_d_128(<4 x i32> %a, <4 x i32> %b, i8 %mask) {
; X86-LABEL: test_mask_pcmpgt_d_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpcmpgtd %xmm1, %xmm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x66,0xc1]
; X86-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; X86-NEXT:    # kill: def $al killed $al killed $eax
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_pcmpgt_d_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpcmpgtd %xmm1, %xmm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x66,0xc1]
; X64-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; X64-NEXT:    # kill: def $al killed $al killed $eax
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call i8 @llvm.x86.avx512.mask.pcmpgt.d.128(<4 x i32> %a, <4 x i32> %b, i8 %mask)
  ret i8 %res
}

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

define i8 @test_pcmpgt_q_128(<2 x i64> %a, <2 x i64> %b) {
; CHECK-LABEL: test_pcmpgt_q_128:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vpcmpgtq %xmm1, %xmm0, %k0 # encoding: [0x62,0xf2,0xfd,0x08,0x37,0xc1]
; CHECK-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; CHECK-NEXT:    # kill: def $al killed $al killed $eax
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call i8 @llvm.x86.avx512.mask.pcmpgt.q.128(<2 x i64> %a, <2 x i64> %b, i8 -1)
  ret i8 %res
}

define i8 @test_mask_pcmpgt_q_128(<2 x i64> %a, <2 x i64> %b, i8 %mask) {
; X86-LABEL: test_mask_pcmpgt_q_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpcmpgtq %xmm1, %xmm0, %k0 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x37,0xc1]
; X86-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; X86-NEXT:    # kill: def $al killed $al killed $eax
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_pcmpgt_q_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpcmpgtq %xmm1, %xmm0, %k0 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x37,0xc1]
; X64-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; X64-NEXT:    # kill: def $al killed $al killed $eax
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call i8 @llvm.x86.avx512.mask.pcmpgt.q.128(<2 x i64> %a, <2 x i64> %b, i8 %mask)
  ret i8 %res
}

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

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

define <2 x double>@test_int_x86_avx512_mask_unpckh_pd_128(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_unpckh_pd_128:
; X86:       # %bb.0:
; X86-NEXT:    vunpckhpd %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x15,0xd9]
; X86-NEXT:    # xmm3 = xmm0[1],xmm1[1]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vunpckhpd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0xfd,0x09,0x15,0xd1]
; X86-NEXT:    # xmm2 {%k1} = xmm0[1],xmm1[1]
; X86-NEXT:    vaddpd %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x58,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_unpckh_pd_128:
; X64:       # %bb.0:
; X64-NEXT:    vunpckhpd %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x15,0xd9]
; X64-NEXT:    # xmm3 = xmm0[1],xmm1[1]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vunpckhpd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0xfd,0x09,0x15,0xd1]
; X64-NEXT:    # xmm2 {%k1} = xmm0[1],xmm1[1]
; X64-NEXT:    vaddpd %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x58,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x double> @llvm.x86.avx512.mask.unpckh.pd.128(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 %x3)
  %res1 = call <2 x double> @llvm.x86.avx512.mask.unpckh.pd.128(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 -1)
  %res2 = fadd <2 x double> %res, %res1
  ret <2 x double> %res2
}

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

define <4 x double>@test_int_x86_avx512_mask_unpckh_pd_256(<4 x double> %x0, <4 x double> %x1, <4 x double> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_unpckh_pd_256:
; X86:       # %bb.0:
; X86-NEXT:    vunpckhpd %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x15,0xd9]
; X86-NEXT:    # ymm3 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vunpckhpd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0xfd,0x29,0x15,0xd1]
; X86-NEXT:    # ymm2 {%k1} = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
; X86-NEXT:    vaddpd %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0x58,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_unpckh_pd_256:
; X64:       # %bb.0:
; X64-NEXT:    vunpckhpd %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x15,0xd9]
; X64-NEXT:    # ymm3 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vunpckhpd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0xfd,0x29,0x15,0xd1]
; X64-NEXT:    # ymm2 {%k1} = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
; X64-NEXT:    vaddpd %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0x58,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x double> @llvm.x86.avx512.mask.unpckh.pd.256(<4 x double> %x0, <4 x double> %x1, <4 x double> %x2, i8 %x3)
  %res1 = call <4 x double> @llvm.x86.avx512.mask.unpckh.pd.256(<4 x double> %x0, <4 x double> %x1, <4 x double> %x2, i8 -1)
  %res2 = fadd <4 x double> %res, %res1
  ret <4 x double> %res2
}

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

define <4 x float>@test_int_x86_avx512_mask_unpckh_ps_128(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_unpckh_ps_128:
; X86:       # %bb.0:
; X86-NEXT:    vunpckhps %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x15,0xd9]
; X86-NEXT:    # xmm3 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vunpckhps %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7c,0x09,0x15,0xd1]
; X86-NEXT:    # xmm2 {%k1} = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
; X86-NEXT:    vaddps %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe8,0x58,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_unpckh_ps_128:
; X64:       # %bb.0:
; X64-NEXT:    vunpckhps %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x15,0xd9]
; X64-NEXT:    # xmm3 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vunpckhps %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7c,0x09,0x15,0xd1]
; X64-NEXT:    # xmm2 {%k1} = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
; X64-NEXT:    vaddps %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe8,0x58,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask.unpckh.ps.128(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 %x3)
  %res1 = call <4 x float> @llvm.x86.avx512.mask.unpckh.ps.128(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 -1)
  %res2 = fadd <4 x float> %res, %res1
  ret <4 x float> %res2
}

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

define <8 x float>@test_int_x86_avx512_mask_unpckh_ps_256(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_unpckh_ps_256:
; X86:       # %bb.0:
; X86-NEXT:    vunpckhps %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x15,0xd9]
; X86-NEXT:    # ymm3 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vunpckhps %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7c,0x29,0x15,0xd1]
; X86-NEXT:    # ymm2 {%k1} = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
; X86-NEXT:    vaddps %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xec,0x58,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_unpckh_ps_256:
; X64:       # %bb.0:
; X64-NEXT:    vunpckhps %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x15,0xd9]
; X64-NEXT:    # ymm3 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vunpckhps %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7c,0x29,0x15,0xd1]
; X64-NEXT:    # ymm2 {%k1} = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
; X64-NEXT:    vaddps %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xec,0x58,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask.unpckh.ps.256(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2, i8 %x3)
  %res1 = call <8 x float> @llvm.x86.avx512.mask.unpckh.ps.256(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2, i8 -1)
  %res2 = fadd <8 x float> %res, %res1
  ret <8 x float> %res2
}

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

define <2 x double>@test_int_x86_avx512_mask_unpckl_pd_128(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_unpckl_pd_128:
; X86:       # %bb.0:
; X86-NEXT:    vunpcklpd %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x14,0xd9]
; X86-NEXT:    # xmm3 = xmm0[0],xmm1[0]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vunpcklpd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0xfd,0x09,0x14,0xd1]
; X86-NEXT:    # xmm2 {%k1} = xmm0[0],xmm1[0]
; X86-NEXT:    vaddpd %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x58,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_unpckl_pd_128:
; X64:       # %bb.0:
; X64-NEXT:    vunpcklpd %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x14,0xd9]
; X64-NEXT:    # xmm3 = xmm0[0],xmm1[0]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vunpcklpd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0xfd,0x09,0x14,0xd1]
; X64-NEXT:    # xmm2 {%k1} = xmm0[0],xmm1[0]
; X64-NEXT:    vaddpd %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x58,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x double> @llvm.x86.avx512.mask.unpckl.pd.128(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 %x3)
  %res1 = call <2 x double> @llvm.x86.avx512.mask.unpckl.pd.128(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 -1)
  %res2 = fadd <2 x double> %res, %res1
  ret <2 x double> %res2
}

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

define <4 x double>@test_int_x86_avx512_mask_unpckl_pd_256(<4 x double> %x0, <4 x double> %x1, <4 x double> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_unpckl_pd_256:
; X86:       # %bb.0:
; X86-NEXT:    vunpcklpd %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x14,0xd9]
; X86-NEXT:    # ymm3 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vunpcklpd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0xfd,0x29,0x14,0xd1]
; X86-NEXT:    # ymm2 {%k1} = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; X86-NEXT:    vaddpd %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0x58,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_unpckl_pd_256:
; X64:       # %bb.0:
; X64-NEXT:    vunpcklpd %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x14,0xd9]
; X64-NEXT:    # ymm3 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vunpcklpd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0xfd,0x29,0x14,0xd1]
; X64-NEXT:    # ymm2 {%k1} = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; X64-NEXT:    vaddpd %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0x58,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x double> @llvm.x86.avx512.mask.unpckl.pd.256(<4 x double> %x0, <4 x double> %x1, <4 x double> %x2, i8 %x3)
  %res1 = call <4 x double> @llvm.x86.avx512.mask.unpckl.pd.256(<4 x double> %x0, <4 x double> %x1, <4 x double> %x2, i8 -1)
  %res2 = fadd <4 x double> %res, %res1
  ret <4 x double> %res2
}

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

define <4 x float>@test_int_x86_avx512_mask_unpckl_ps_128(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_unpckl_ps_128:
; X86:       # %bb.0:
; X86-NEXT:    vunpcklps %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x14,0xd9]
; X86-NEXT:    # xmm3 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vunpcklps %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7c,0x09,0x14,0xd1]
; X86-NEXT:    # xmm2 {%k1} = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; X86-NEXT:    vaddps %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe8,0x58,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_unpckl_ps_128:
; X64:       # %bb.0:
; X64-NEXT:    vunpcklps %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x14,0xd9]
; X64-NEXT:    # xmm3 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vunpcklps %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7c,0x09,0x14,0xd1]
; X64-NEXT:    # xmm2 {%k1} = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; X64-NEXT:    vaddps %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe8,0x58,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask.unpckl.ps.128(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 %x3)
  %res1 = call <4 x float> @llvm.x86.avx512.mask.unpckl.ps.128(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 -1)
  %res2 = fadd <4 x float> %res, %res1
  ret <4 x float> %res2
}

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

define <8 x float>@test_int_x86_avx512_mask_unpckl_ps_256(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_unpckl_ps_256:
; X86:       # %bb.0:
; X86-NEXT:    vunpcklps %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x14,0xd9]
; X86-NEXT:    # ymm3 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vunpcklps %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7c,0x29,0x14,0xd1]
; X86-NEXT:    # ymm2 {%k1} = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
; X86-NEXT:    vaddps %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xec,0x58,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_unpckl_ps_256:
; X64:       # %bb.0:
; X64-NEXT:    vunpcklps %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x14,0xd9]
; X64-NEXT:    # ymm3 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vunpcklps %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7c,0x29,0x14,0xd1]
; X64-NEXT:    # ymm2 {%k1} = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
; X64-NEXT:    vaddps %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xec,0x58,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask.unpckl.ps.256(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2, i8 %x3)
  %res1 = call <8 x float> @llvm.x86.avx512.mask.unpckl.ps.256(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2, i8 -1)
  %res2 = fadd <8 x float> %res, %res1
  ret <8 x float> %res2
}

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

define <4 x i32>@test_int_x86_avx512_mask_punpckhd_q_128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_punpckhd_q_128:
; X86:       # %bb.0:
; X86-NEXT:    vpunpckhdq %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6a,0xd9]
; X86-NEXT:    # xmm3 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpunpckhdq %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x6a,0xd1]
; X86-NEXT:    # xmm2 {%k1} = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
; 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_punpckhd_q_128:
; X64:       # %bb.0:
; X64-NEXT:    vpunpckhdq %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6a,0xd9]
; X64-NEXT:    # xmm3 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpunpckhdq %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x6a,0xd1]
; X64-NEXT:    # xmm2 {%k1} = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
; 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.punpckhd.q.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 %x3)
  %res1 = call <4 x i32> @llvm.x86.avx512.mask.punpckhd.q.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 -1)
  %res2 = add <4 x i32> %res, %res1
  ret <4 x i32> %res2
}

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

define <4 x i32>@test_int_x86_avx512_mask_punpckld_q_128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_punpckld_q_128:
; X86:       # %bb.0:
; X86-NEXT:    vpunpckldq %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x62,0xd9]
; X86-NEXT:    # xmm3 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpunpckldq %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x62,0xd1]
; X86-NEXT:    # xmm2 {%k1} = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; 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_punpckld_q_128:
; X64:       # %bb.0:
; X64-NEXT:    vpunpckldq %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x62,0xd9]
; X64-NEXT:    # xmm3 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpunpckldq %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x62,0xd1]
; X64-NEXT:    # xmm2 {%k1} = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
; 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.punpckld.q.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 %x3)
  %res1 = call <4 x i32> @llvm.x86.avx512.mask.punpckld.q.128(<4 x i32> %x0, <4 x i32> %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.punpckhd.q.256(<8 x i32>, <8 x i32>, <8 x i32>, i8)

define <8 x i32>@test_int_x86_avx512_mask_punpckhd_q_256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_punpckhd_q_256:
; X86:       # %bb.0:
; X86-NEXT:    vpunpckhdq %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6a,0xd9]
; X86-NEXT:    # ymm3 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpunpckhdq %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x6a,0xd1]
; X86-NEXT:    # ymm2 {%k1} = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
; 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_punpckhd_q_256:
; X64:       # %bb.0:
; X64-NEXT:    vpunpckhdq %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6a,0xd9]
; X64-NEXT:    # ymm3 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpunpckhdq %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x6a,0xd1]
; X64-NEXT:    # ymm2 {%k1} = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
; 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.punpckhd.q.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 %x3)
  %res1 = call <8 x i32> @llvm.x86.avx512.mask.punpckhd.q.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 -1)
  %res2 = add <8 x i32> %res, %res1
  ret <8 x i32> %res2
}

declare <8 x i32> @llvm.x86.avx512.mask.punpckld.q.256(<8 x i32>, <8 x i32>, <8 x i32>, i8)

define <8 x i32>@test_int_x86_avx512_mask_punpckld_q_256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_punpckld_q_256:
; X86:       # %bb.0:
; X86-NEXT:    vpunpckldq %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x62,0xd9]
; X86-NEXT:    # ymm3 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpunpckldq %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x62,0xd1]
; X86-NEXT:    # ymm2 {%k1} = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
; 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_punpckld_q_256:
; X64:       # %bb.0:
; X64-NEXT:    vpunpckldq %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x62,0xd9]
; X64-NEXT:    # ymm3 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpunpckldq %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x62,0xd1]
; X64-NEXT:    # ymm2 {%k1} = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
; 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.punpckld.q.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 %x3)
  %res1 = call <8 x i32> @llvm.x86.avx512.mask.punpckld.q.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 -1)
  %res2 = add <8 x i32> %res, %res1
  ret <8 x i32> %res2
}

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

define <2 x i64>@test_int_x86_avx512_mask_punpckhqd_q_128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_punpckhqd_q_128:
; X86:       # %bb.0:
; X86-NEXT:    vpunpckhqdq %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6d,0xd9]
; X86-NEXT:    # xmm3 = xmm0[1],xmm1[1]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpunpckhqdq %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0xfd,0x09,0x6d,0xd1]
; X86-NEXT:    # xmm2 {%k1} = xmm0[1],xmm1[1]
; X86-NEXT:    vpaddq %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xd4,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_punpckhqd_q_128:
; X64:       # %bb.0:
; X64-NEXT:    vpunpckhqdq %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6d,0xd9]
; X64-NEXT:    # xmm3 = xmm0[1],xmm1[1]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpunpckhqdq %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0xfd,0x09,0x6d,0xd1]
; X64-NEXT:    # xmm2 {%k1} = xmm0[1],xmm1[1]
; X64-NEXT:    vpaddq %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xd4,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x i64> @llvm.x86.avx512.mask.punpckhqd.q.128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i8 %x3)
  %res1 = call <2 x i64> @llvm.x86.avx512.mask.punpckhqd.q.128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i8 -1)
  %res2 = add <2 x i64> %res, %res1
  ret <2 x i64> %res2
}

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

define <2 x i64>@test_int_x86_avx512_mask_punpcklqd_q_128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_punpcklqd_q_128:
; X86:       # %bb.0:
; X86-NEXT:    vpunpcklqdq %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6c,0xd9]
; X86-NEXT:    # xmm3 = xmm0[0],xmm1[0]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpunpcklqdq %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0xfd,0x09,0x6c,0xd1]
; X86-NEXT:    # xmm2 {%k1} = xmm0[0],xmm1[0]
; X86-NEXT:    vpaddq %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xd4,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_punpcklqd_q_128:
; X64:       # %bb.0:
; X64-NEXT:    vpunpcklqdq %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6c,0xd9]
; X64-NEXT:    # xmm3 = xmm0[0],xmm1[0]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpunpcklqdq %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0xfd,0x09,0x6c,0xd1]
; X64-NEXT:    # xmm2 {%k1} = xmm0[0],xmm1[0]
; X64-NEXT:    vpaddq %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xd4,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x i64> @llvm.x86.avx512.mask.punpcklqd.q.128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i8 %x3)
  %res1 = call <2 x i64> @llvm.x86.avx512.mask.punpcklqd.q.128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i8 -1)
  %res2 = add <2 x i64> %res, %res1
  ret <2 x i64> %res2
}

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

define <4 x i64>@test_int_x86_avx512_mask_punpcklqd_q_256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_punpcklqd_q_256:
; X86:       # %bb.0:
; X86-NEXT:    vpunpcklqdq %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6c,0xd9]
; X86-NEXT:    # ymm3 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpunpcklqdq %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0xfd,0x29,0x6c,0xd1]
; X86-NEXT:    # ymm2 {%k1} = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; X86-NEXT:    vpaddq %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xd4,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_punpcklqd_q_256:
; X64:       # %bb.0:
; X64-NEXT:    vpunpcklqdq %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6c,0xd9]
; X64-NEXT:    # ymm3 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpunpcklqdq %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0xfd,0x29,0x6c,0xd1]
; X64-NEXT:    # ymm2 {%k1} = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; X64-NEXT:    vpaddq %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xd4,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i64> @llvm.x86.avx512.mask.punpcklqd.q.256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2, i8 %x3)
  %res1 = call <4 x i64> @llvm.x86.avx512.mask.punpcklqd.q.256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2, i8 -1)
  %res2 = add <4 x i64> %res, %res1
  ret <4 x i64> %res2
}

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

define <4 x i64>@test_int_x86_avx512_mask_punpckhqd_q_256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_punpckhqd_q_256:
; X86:       # %bb.0:
; X86-NEXT:    vpunpckhqdq %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6d,0xd9]
; X86-NEXT:    # ymm3 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpunpckhqdq %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0xfd,0x29,0x6d,0xd1]
; X86-NEXT:    # ymm2 {%k1} = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
; X86-NEXT:    vpaddq %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xd4,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_punpckhqd_q_256:
; X64:       # %bb.0:
; X64-NEXT:    vpunpckhqdq %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6d,0xd9]
; X64-NEXT:    # ymm3 = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpunpckhqdq %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0xfd,0x29,0x6d,0xd1]
; X64-NEXT:    # ymm2 {%k1} = ymm0[1],ymm1[1],ymm0[3],ymm1[3]
; X64-NEXT:    vpaddq %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xd4,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i64> @llvm.x86.avx512.mask.punpckhqd.q.256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2, i8 %x3)
  %res1 = call <4 x i64> @llvm.x86.avx512.mask.punpckhqd.q.256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2, i8 -1)
  %res2 = add <4 x i64> %res, %res1
  ret <4 x i64> %res2
}

define <4 x i32> @test_mask_and_epi32_rr_128(<4 x i32> %a, <4 x i32> %b) {
; CHECK-LABEL: test_mask_and_epi32_rr_128:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vandps %xmm1, %xmm0, %xmm0 # encoding: [0xc5,0xf8,0x54,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <4 x i32> @llvm.x86.avx512.mask.pand.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> zeroinitializer, i8 -1)
  ret <4 x i32> %res
}

define <4 x i32> @test_mask_and_epi32_rrk_128(<4 x i32> %a, <4 x i32> %b, <4 x i32> %passThru, i8 %mask) {
; X86-LABEL: test_mask_and_epi32_rrk_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpandd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xdb,0xd1]
; X86-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_and_epi32_rrk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpandd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xdb,0xd1]
; X64-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i32> @llvm.x86.avx512.mask.pand.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> %passThru, i8 %mask)
  ret <4 x i32> %res
}

define <4 x i32> @test_mask_and_epi32_rrkz_128(<4 x i32> %a, <4 x i32> %b, i8 %mask) {
; X86-LABEL: test_mask_and_epi32_rrkz_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpandd %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xdb,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_and_epi32_rrkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpandd %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xdb,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i32> @llvm.x86.avx512.mask.pand.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> zeroinitializer, i8 %mask)
  ret <4 x i32> %res
}

define <4 x i32> @test_mask_and_epi32_rm_128(<4 x i32> %a, <4 x i32>* %ptr_b) {
; X86-LABEL: test_mask_and_epi32_rm_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vandps (%eax), %xmm0, %xmm0 # encoding: [0xc5,0xf8,0x54,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_and_epi32_rm_128:
; X64:       # %bb.0:
; X64-NEXT:    vandps (%rdi), %xmm0, %xmm0 # encoding: [0xc5,0xf8,0x54,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %b = load <4 x i32>, <4 x i32>* %ptr_b
  %res = call <4 x i32> @llvm.x86.avx512.mask.pand.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> zeroinitializer, i8 -1)
  ret <4 x i32> %res
}

define <4 x i32> @test_mask_and_epi32_rmk_128(<4 x i32> %a, <4 x i32>* %ptr_b, <4 x i32> %passThru, i8 %mask) {
; X86-LABEL: test_mask_and_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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpandd (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xdb,0x08]
; X86-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_and_epi32_rmk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpandd (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xdb,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 <4 x i32> @llvm.x86.avx512.mask.pand.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> %passThru, i8 %mask)
  ret <4 x i32> %res
}

define <4 x i32> @test_mask_and_epi32_rmkz_128(<4 x i32> %a, <4 x i32>* %ptr_b, i8 %mask) {
; X86-LABEL: test_mask_and_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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpandd (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xdb,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_and_epi32_rmkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpandd (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xdb,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %b = load <4 x i32>, <4 x i32>* %ptr_b
  %res = call <4 x i32> @llvm.x86.avx512.mask.pand.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> zeroinitializer, i8 %mask)
  ret <4 x i32> %res
}

define <4 x i32> @test_mask_and_epi32_rmb_128(<4 x i32> %a, i32* %ptr_b) {
; X86-LABEL: test_mask_and_epi32_rmb_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpandd (%eax){1to4}, %xmm0, %xmm0 # encoding: [0x62,0xf1,0x7d,0x18,0xdb,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_and_epi32_rmb_128:
; X64:       # %bb.0:
; X64-NEXT:    vpandd (%rdi){1to4}, %xmm0, %xmm0 # encoding: [0x62,0xf1,0x7d,0x18,0xdb,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 <4 x i32> @llvm.x86.avx512.mask.pand.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> zeroinitializer, i8 -1)
  ret <4 x i32> %res
}

define <4 x i32> @test_mask_and_epi32_rmbk_128(<4 x i32> %a, i32* %ptr_b, <4 x i32> %passThru, i8 %mask) {
; X86-LABEL: test_mask_and_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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpandd (%eax){1to4}, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x19,0xdb,0x08]
; X86-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_and_epi32_rmbk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpandd (%rdi){1to4}, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x19,0xdb,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 <4 x i32> @llvm.x86.avx512.mask.pand.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> %passThru, i8 %mask)
  ret <4 x i32> %res
}

define <4 x i32> @test_mask_and_epi32_rmbkz_128(<4 x i32> %a, i32* %ptr_b, i8 %mask) {
; X86-LABEL: test_mask_and_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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpandd (%eax){1to4}, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x99,0xdb,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_and_epi32_rmbkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpandd (%rdi){1to4}, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x99,0xdb,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 <4 x i32> @llvm.x86.avx512.mask.pand.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> zeroinitializer, i8 %mask)
  ret <4 x i32> %res
}

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

define <8 x i32> @test_mask_and_epi32_rr_256(<8 x i32> %a, <8 x i32> %b) {
; CHECK-LABEL: test_mask_and_epi32_rr_256:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vandps %ymm1, %ymm0, %ymm0 # encoding: [0xc5,0xfc,0x54,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <8 x i32> @llvm.x86.avx512.mask.pand.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> zeroinitializer, i8 -1)
  ret <8 x i32> %res
}

define <8 x i32> @test_mask_and_epi32_rrk_256(<8 x i32> %a, <8 x i32> %b, <8 x i32> %passThru, i8 %mask) {
; X86-LABEL: test_mask_and_epi32_rrk_256:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpandd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xdb,0xd1]
; X86-NEXT:    vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_and_epi32_rrk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpandd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xdb,0xd1]
; X64-NEXT:    vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x i32> @llvm.x86.avx512.mask.pand.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> %passThru, i8 %mask)
  ret <8 x i32> %res
}

define <8 x i32> @test_mask_and_epi32_rrkz_256(<8 x i32> %a, <8 x i32> %b, i8 %mask) {
; X86-LABEL: test_mask_and_epi32_rrkz_256:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpandd %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xdb,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_and_epi32_rrkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpandd %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xdb,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x i32> @llvm.x86.avx512.mask.pand.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> zeroinitializer, i8 %mask)
  ret <8 x i32> %res
}

define <8 x i32> @test_mask_and_epi32_rm_256(<8 x i32> %a, <8 x i32>* %ptr_b) {
; X86-LABEL: test_mask_and_epi32_rm_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vandps (%eax), %ymm0, %ymm0 # encoding: [0xc5,0xfc,0x54,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_and_epi32_rm_256:
; X64:       # %bb.0:
; X64-NEXT:    vandps (%rdi), %ymm0, %ymm0 # encoding: [0xc5,0xfc,0x54,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %b = load <8 x i32>, <8 x i32>* %ptr_b
  %res = call <8 x i32> @llvm.x86.avx512.mask.pand.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> zeroinitializer, i8 -1)
  ret <8 x i32> %res
}

define <8 x i32> @test_mask_and_epi32_rmk_256(<8 x i32> %a, <8 x i32>* %ptr_b, <8 x i32> %passThru, i8 %mask) {
; X86-LABEL: test_mask_and_epi32_rmk_256:
; 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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpandd (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xdb,0x08]
; X86-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_and_epi32_rmk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpandd (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xdb,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 <8 x i32> @llvm.x86.avx512.mask.pand.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> %passThru, i8 %mask)
  ret <8 x i32> %res
}

define <8 x i32> @test_mask_and_epi32_rmkz_256(<8 x i32> %a, <8 x i32>* %ptr_b, i8 %mask) {
; X86-LABEL: test_mask_and_epi32_rmkz_256:
; 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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpandd (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xdb,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_and_epi32_rmkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpandd (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xdb,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %b = load <8 x i32>, <8 x i32>* %ptr_b
  %res = call <8 x i32> @llvm.x86.avx512.mask.pand.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> zeroinitializer, i8 %mask)
  ret <8 x i32> %res
}

define <8 x i32> @test_mask_and_epi32_rmb_256(<8 x i32> %a, i32* %ptr_b) {
; X86-LABEL: test_mask_and_epi32_rmb_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpandd (%eax){1to8}, %ymm0, %ymm0 # encoding: [0x62,0xf1,0x7d,0x38,0xdb,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_and_epi32_rmb_256:
; X64:       # %bb.0:
; X64-NEXT:    vpandd (%rdi){1to8}, %ymm0, %ymm0 # encoding: [0x62,0xf1,0x7d,0x38,0xdb,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 <8 x i32> @llvm.x86.avx512.mask.pand.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> zeroinitializer, i8 -1)
  ret <8 x i32> %res
}

define <8 x i32> @test_mask_and_epi32_rmbk_256(<8 x i32> %a, i32* %ptr_b, <8 x i32> %passThru, i8 %mask) {
; X86-LABEL: test_mask_and_epi32_rmbk_256:
; 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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpandd (%eax){1to8}, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x39,0xdb,0x08]
; X86-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_and_epi32_rmbk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpandd (%rdi){1to8}, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x39,0xdb,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 <8 x i32> @llvm.x86.avx512.mask.pand.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> %passThru, i8 %mask)
  ret <8 x i32> %res
}

define <8 x i32> @test_mask_and_epi32_rmbkz_256(<8 x i32> %a, i32* %ptr_b, i8 %mask) {
; X86-LABEL: test_mask_and_epi32_rmbkz_256:
; 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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpandd (%eax){1to8}, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xb9,0xdb,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_and_epi32_rmbkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpandd (%rdi){1to8}, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xb9,0xdb,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 <8 x i32> @llvm.x86.avx512.mask.pand.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> zeroinitializer, i8 %mask)
  ret <8 x i32> %res
}

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

define <4 x i32> @test_mask_or_epi32_rr_128(<4 x i32> %a, <4 x i32> %b) {
; CHECK-LABEL: test_mask_or_epi32_rr_128:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vorps %xmm1, %xmm0, %xmm0 # encoding: [0xc5,0xf8,0x56,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <4 x i32> @llvm.x86.avx512.mask.por.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> zeroinitializer, i8 -1)
  ret <4 x i32> %res
}

define <4 x i32> @test_mask_or_epi32_rrk_128(<4 x i32> %a, <4 x i32> %b, <4 x i32> %passThru, i8 %mask) {
; X86-LABEL: test_mask_or_epi32_rrk_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpord %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xeb,0xd1]
; X86-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_or_epi32_rrk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpord %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xeb,0xd1]
; X64-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i32> @llvm.x86.avx512.mask.por.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> %passThru, i8 %mask)
  ret <4 x i32> %res
}

define <4 x i32> @test_mask_or_epi32_rrkz_128(<4 x i32> %a, <4 x i32> %b, i8 %mask) {
; X86-LABEL: test_mask_or_epi32_rrkz_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpord %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xeb,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_or_epi32_rrkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpord %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xeb,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i32> @llvm.x86.avx512.mask.por.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> zeroinitializer, i8 %mask)
  ret <4 x i32> %res
}

define <4 x i32> @test_mask_or_epi32_rm_128(<4 x i32> %a, <4 x i32>* %ptr_b) {
; X86-LABEL: test_mask_or_epi32_rm_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vorps (%eax), %xmm0, %xmm0 # encoding: [0xc5,0xf8,0x56,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_or_epi32_rm_128:
; X64:       # %bb.0:
; X64-NEXT:    vorps (%rdi), %xmm0, %xmm0 # encoding: [0xc5,0xf8,0x56,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %b = load <4 x i32>, <4 x i32>* %ptr_b
  %res = call <4 x i32> @llvm.x86.avx512.mask.por.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> zeroinitializer, i8 -1)
  ret <4 x i32> %res
}

define <4 x i32> @test_mask_or_epi32_rmk_128(<4 x i32> %a, <4 x i32>* %ptr_b, <4 x i32> %passThru, i8 %mask) {
; X86-LABEL: test_mask_or_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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpord (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xeb,0x08]
; X86-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_or_epi32_rmk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpord (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xeb,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 <4 x i32> @llvm.x86.avx512.mask.por.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> %passThru, i8 %mask)
  ret <4 x i32> %res
}

define <4 x i32> @test_mask_or_epi32_rmkz_128(<4 x i32> %a, <4 x i32>* %ptr_b, i8 %mask) {
; X86-LABEL: test_mask_or_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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpord (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xeb,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_or_epi32_rmkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpord (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xeb,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %b = load <4 x i32>, <4 x i32>* %ptr_b
  %res = call <4 x i32> @llvm.x86.avx512.mask.por.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> zeroinitializer, i8 %mask)
  ret <4 x i32> %res
}

define <4 x i32> @test_mask_or_epi32_rmb_128(<4 x i32> %a, i32* %ptr_b) {
; X86-LABEL: test_mask_or_epi32_rmb_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpord (%eax){1to4}, %xmm0, %xmm0 # encoding: [0x62,0xf1,0x7d,0x18,0xeb,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_or_epi32_rmb_128:
; X64:       # %bb.0:
; X64-NEXT:    vpord (%rdi){1to4}, %xmm0, %xmm0 # encoding: [0x62,0xf1,0x7d,0x18,0xeb,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 <4 x i32> @llvm.x86.avx512.mask.por.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> zeroinitializer, i8 -1)
  ret <4 x i32> %res
}

define <4 x i32> @test_mask_or_epi32_rmbk_128(<4 x i32> %a, i32* %ptr_b, <4 x i32> %passThru, i8 %mask) {
; X86-LABEL: test_mask_or_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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpord (%eax){1to4}, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x19,0xeb,0x08]
; X86-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_or_epi32_rmbk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpord (%rdi){1to4}, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x19,0xeb,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 <4 x i32> @llvm.x86.avx512.mask.por.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> %passThru, i8 %mask)
  ret <4 x i32> %res
}

define <4 x i32> @test_mask_or_epi32_rmbkz_128(<4 x i32> %a, i32* %ptr_b, i8 %mask) {
; X86-LABEL: test_mask_or_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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpord (%eax){1to4}, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x99,0xeb,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_or_epi32_rmbkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpord (%rdi){1to4}, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x99,0xeb,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 <4 x i32> @llvm.x86.avx512.mask.por.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> zeroinitializer, i8 %mask)
  ret <4 x i32> %res
}

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

define <8 x i32> @test_mask_or_epi32_rr_256(<8 x i32> %a, <8 x i32> %b) {
; CHECK-LABEL: test_mask_or_epi32_rr_256:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vorps %ymm1, %ymm0, %ymm0 # encoding: [0xc5,0xfc,0x56,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <8 x i32> @llvm.x86.avx512.mask.por.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> zeroinitializer, i8 -1)
  ret <8 x i32> %res
}

define <8 x i32> @test_mask_or_epi32_rrk_256(<8 x i32> %a, <8 x i32> %b, <8 x i32> %passThru, i8 %mask) {
; X86-LABEL: test_mask_or_epi32_rrk_256:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpord %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xeb,0xd1]
; X86-NEXT:    vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_or_epi32_rrk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpord %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xeb,0xd1]
; X64-NEXT:    vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x i32> @llvm.x86.avx512.mask.por.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> %passThru, i8 %mask)
  ret <8 x i32> %res
}

define <8 x i32> @test_mask_or_epi32_rrkz_256(<8 x i32> %a, <8 x i32> %b, i8 %mask) {
; X86-LABEL: test_mask_or_epi32_rrkz_256:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpord %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xeb,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_or_epi32_rrkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpord %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xeb,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x i32> @llvm.x86.avx512.mask.por.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> zeroinitializer, i8 %mask)
  ret <8 x i32> %res
}

define <8 x i32> @test_mask_or_epi32_rm_256(<8 x i32> %a, <8 x i32>* %ptr_b) {
; X86-LABEL: test_mask_or_epi32_rm_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vorps (%eax), %ymm0, %ymm0 # encoding: [0xc5,0xfc,0x56,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_or_epi32_rm_256:
; X64:       # %bb.0:
; X64-NEXT:    vorps (%rdi), %ymm0, %ymm0 # encoding: [0xc5,0xfc,0x56,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %b = load <8 x i32>, <8 x i32>* %ptr_b
  %res = call <8 x i32> @llvm.x86.avx512.mask.por.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> zeroinitializer, i8 -1)
  ret <8 x i32> %res
}

define <8 x i32> @test_mask_or_epi32_rmk_256(<8 x i32> %a, <8 x i32>* %ptr_b, <8 x i32> %passThru, i8 %mask) {
; X86-LABEL: test_mask_or_epi32_rmk_256:
; 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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpord (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xeb,0x08]
; X86-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_or_epi32_rmk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpord (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xeb,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 <8 x i32> @llvm.x86.avx512.mask.por.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> %passThru, i8 %mask)
  ret <8 x i32> %res
}

define <8 x i32> @test_mask_or_epi32_rmkz_256(<8 x i32> %a, <8 x i32>* %ptr_b, i8 %mask) {
; X86-LABEL: test_mask_or_epi32_rmkz_256:
; 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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpord (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xeb,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_or_epi32_rmkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpord (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xeb,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %b = load <8 x i32>, <8 x i32>* %ptr_b
  %res = call <8 x i32> @llvm.x86.avx512.mask.por.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> zeroinitializer, i8 %mask)
  ret <8 x i32> %res
}

define <8 x i32> @test_mask_or_epi32_rmb_256(<8 x i32> %a, i32* %ptr_b) {
; X86-LABEL: test_mask_or_epi32_rmb_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpord (%eax){1to8}, %ymm0, %ymm0 # encoding: [0x62,0xf1,0x7d,0x38,0xeb,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_or_epi32_rmb_256:
; X64:       # %bb.0:
; X64-NEXT:    vpord (%rdi){1to8}, %ymm0, %ymm0 # encoding: [0x62,0xf1,0x7d,0x38,0xeb,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 <8 x i32> @llvm.x86.avx512.mask.por.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> zeroinitializer, i8 -1)
  ret <8 x i32> %res
}

define <8 x i32> @test_mask_or_epi32_rmbk_256(<8 x i32> %a, i32* %ptr_b, <8 x i32> %passThru, i8 %mask) {
; X86-LABEL: test_mask_or_epi32_rmbk_256:
; 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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpord (%eax){1to8}, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x39,0xeb,0x08]
; X86-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_or_epi32_rmbk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpord (%rdi){1to8}, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x39,0xeb,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 <8 x i32> @llvm.x86.avx512.mask.por.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> %passThru, i8 %mask)
  ret <8 x i32> %res
}

define <8 x i32> @test_mask_or_epi32_rmbkz_256(<8 x i32> %a, i32* %ptr_b, i8 %mask) {
; X86-LABEL: test_mask_or_epi32_rmbkz_256:
; 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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpord (%eax){1to8}, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xb9,0xeb,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_or_epi32_rmbkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpord (%rdi){1to8}, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xb9,0xeb,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 <8 x i32> @llvm.x86.avx512.mask.por.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> zeroinitializer, i8 %mask)
  ret <8 x i32> %res
}

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

define <4 x i32> @test_mask_xor_epi32_rr_128(<4 x i32> %a, <4 x i32> %b) {
; CHECK-LABEL: test_mask_xor_epi32_rr_128:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vxorps %xmm1, %xmm0, %xmm0 # encoding: [0xc5,0xf8,0x57,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <4 x i32> @llvm.x86.avx512.mask.pxor.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> zeroinitializer, i8 -1)
  ret <4 x i32> %res
}

define <4 x i32> @test_mask_xor_epi32_rrk_128(<4 x i32> %a, <4 x i32> %b, <4 x i32> %passThru, i8 %mask) {
; X86-LABEL: test_mask_xor_epi32_rrk_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpxord %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xef,0xd1]
; X86-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_xor_epi32_rrk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpxord %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xef,0xd1]
; X64-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i32> @llvm.x86.avx512.mask.pxor.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> %passThru, i8 %mask)
  ret <4 x i32> %res
}

define <4 x i32> @test_mask_xor_epi32_rrkz_128(<4 x i32> %a, <4 x i32> %b, i8 %mask) {
; X86-LABEL: test_mask_xor_epi32_rrkz_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpxord %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xef,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_xor_epi32_rrkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpxord %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xef,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i32> @llvm.x86.avx512.mask.pxor.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> zeroinitializer, i8 %mask)
  ret <4 x i32> %res
}

define <4 x i32> @test_mask_xor_epi32_rm_128(<4 x i32> %a, <4 x i32>* %ptr_b) {
; X86-LABEL: test_mask_xor_epi32_rm_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vxorps (%eax), %xmm0, %xmm0 # encoding: [0xc5,0xf8,0x57,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_xor_epi32_rm_128:
; X64:       # %bb.0:
; X64-NEXT:    vxorps (%rdi), %xmm0, %xmm0 # encoding: [0xc5,0xf8,0x57,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %b = load <4 x i32>, <4 x i32>* %ptr_b
  %res = call <4 x i32> @llvm.x86.avx512.mask.pxor.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> zeroinitializer, i8 -1)
  ret <4 x i32> %res
}

define <4 x i32> @test_mask_xor_epi32_rmk_128(<4 x i32> %a, <4 x i32>* %ptr_b, <4 x i32> %passThru, i8 %mask) {
; X86-LABEL: test_mask_xor_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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpxord (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xef,0x08]
; X86-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_xor_epi32_rmk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpxord (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xef,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 <4 x i32> @llvm.x86.avx512.mask.pxor.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> %passThru, i8 %mask)
  ret <4 x i32> %res
}

define <4 x i32> @test_mask_xor_epi32_rmkz_128(<4 x i32> %a, <4 x i32>* %ptr_b, i8 %mask) {
; X86-LABEL: test_mask_xor_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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpxord (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xef,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_xor_epi32_rmkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpxord (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xef,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %b = load <4 x i32>, <4 x i32>* %ptr_b
  %res = call <4 x i32> @llvm.x86.avx512.mask.pxor.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> zeroinitializer, i8 %mask)
  ret <4 x i32> %res
}

define <4 x i32> @test_mask_xor_epi32_rmb_128(<4 x i32> %a, i32* %ptr_b) {
; X86-LABEL: test_mask_xor_epi32_rmb_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpxord (%eax){1to4}, %xmm0, %xmm0 # encoding: [0x62,0xf1,0x7d,0x18,0xef,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_xor_epi32_rmb_128:
; X64:       # %bb.0:
; X64-NEXT:    vpxord (%rdi){1to4}, %xmm0, %xmm0 # encoding: [0x62,0xf1,0x7d,0x18,0xef,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 <4 x i32> @llvm.x86.avx512.mask.pxor.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> zeroinitializer, i8 -1)
  ret <4 x i32> %res
}

define <4 x i32> @test_mask_xor_epi32_rmbk_128(<4 x i32> %a, i32* %ptr_b, <4 x i32> %passThru, i8 %mask) {
; X86-LABEL: test_mask_xor_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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpxord (%eax){1to4}, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x19,0xef,0x08]
; X86-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_xor_epi32_rmbk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpxord (%rdi){1to4}, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x19,0xef,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 <4 x i32> @llvm.x86.avx512.mask.pxor.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> %passThru, i8 %mask)
  ret <4 x i32> %res
}

define <4 x i32> @test_mask_xor_epi32_rmbkz_128(<4 x i32> %a, i32* %ptr_b, i8 %mask) {
; X86-LABEL: test_mask_xor_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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpxord (%eax){1to4}, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x99,0xef,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_xor_epi32_rmbkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpxord (%rdi){1to4}, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x99,0xef,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 <4 x i32> @llvm.x86.avx512.mask.pxor.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> zeroinitializer, i8 %mask)
  ret <4 x i32> %res
}

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

define <8 x i32> @test_mask_xor_epi32_rr_256(<8 x i32> %a, <8 x i32> %b) {
; CHECK-LABEL: test_mask_xor_epi32_rr_256:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vxorps %ymm1, %ymm0, %ymm0 # encoding: [0xc5,0xfc,0x57,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <8 x i32> @llvm.x86.avx512.mask.pxor.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> zeroinitializer, i8 -1)
  ret <8 x i32> %res
}

define <8 x i32> @test_mask_xor_epi32_rrk_256(<8 x i32> %a, <8 x i32> %b, <8 x i32> %passThru, i8 %mask) {
; X86-LABEL: test_mask_xor_epi32_rrk_256:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpxord %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xef,0xd1]
; X86-NEXT:    vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_xor_epi32_rrk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpxord %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xef,0xd1]
; X64-NEXT:    vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x i32> @llvm.x86.avx512.mask.pxor.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> %passThru, i8 %mask)
  ret <8 x i32> %res
}

define <8 x i32> @test_mask_xor_epi32_rrkz_256(<8 x i32> %a, <8 x i32> %b, i8 %mask) {
; X86-LABEL: test_mask_xor_epi32_rrkz_256:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpxord %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xef,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_xor_epi32_rrkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpxord %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xef,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x i32> @llvm.x86.avx512.mask.pxor.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> zeroinitializer, i8 %mask)
  ret <8 x i32> %res
}

define <8 x i32> @test_mask_xor_epi32_rm_256(<8 x i32> %a, <8 x i32>* %ptr_b) {
; X86-LABEL: test_mask_xor_epi32_rm_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vxorps (%eax), %ymm0, %ymm0 # encoding: [0xc5,0xfc,0x57,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_xor_epi32_rm_256:
; X64:       # %bb.0:
; X64-NEXT:    vxorps (%rdi), %ymm0, %ymm0 # encoding: [0xc5,0xfc,0x57,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %b = load <8 x i32>, <8 x i32>* %ptr_b
  %res = call <8 x i32> @llvm.x86.avx512.mask.pxor.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> zeroinitializer, i8 -1)
  ret <8 x i32> %res
}

define <8 x i32> @test_mask_xor_epi32_rmk_256(<8 x i32> %a, <8 x i32>* %ptr_b, <8 x i32> %passThru, i8 %mask) {
; X86-LABEL: test_mask_xor_epi32_rmk_256:
; 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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpxord (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xef,0x08]
; X86-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_xor_epi32_rmk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpxord (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xef,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 <8 x i32> @llvm.x86.avx512.mask.pxor.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> %passThru, i8 %mask)
  ret <8 x i32> %res
}

define <8 x i32> @test_mask_xor_epi32_rmkz_256(<8 x i32> %a, <8 x i32>* %ptr_b, i8 %mask) {
; X86-LABEL: test_mask_xor_epi32_rmkz_256:
; 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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpxord (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xef,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_xor_epi32_rmkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpxord (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xef,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %b = load <8 x i32>, <8 x i32>* %ptr_b
  %res = call <8 x i32> @llvm.x86.avx512.mask.pxor.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> zeroinitializer, i8 %mask)
  ret <8 x i32> %res
}

define <8 x i32> @test_mask_xor_epi32_rmb_256(<8 x i32> %a, i32* %ptr_b) {
; X86-LABEL: test_mask_xor_epi32_rmb_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpxord (%eax){1to8}, %ymm0, %ymm0 # encoding: [0x62,0xf1,0x7d,0x38,0xef,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_xor_epi32_rmb_256:
; X64:       # %bb.0:
; X64-NEXT:    vpxord (%rdi){1to8}, %ymm0, %ymm0 # encoding: [0x62,0xf1,0x7d,0x38,0xef,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 <8 x i32> @llvm.x86.avx512.mask.pxor.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> zeroinitializer, i8 -1)
  ret <8 x i32> %res
}

define <8 x i32> @test_mask_xor_epi32_rmbk_256(<8 x i32> %a, i32* %ptr_b, <8 x i32> %passThru, i8 %mask) {
; X86-LABEL: test_mask_xor_epi32_rmbk_256:
; 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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpxord (%eax){1to8}, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x39,0xef,0x08]
; X86-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_xor_epi32_rmbk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpxord (%rdi){1to8}, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x39,0xef,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 <8 x i32> @llvm.x86.avx512.mask.pxor.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> %passThru, i8 %mask)
  ret <8 x i32> %res
}

define <8 x i32> @test_mask_xor_epi32_rmbkz_256(<8 x i32> %a, i32* %ptr_b, i8 %mask) {
; X86-LABEL: test_mask_xor_epi32_rmbkz_256:
; 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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpxord (%eax){1to8}, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xb9,0xef,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_xor_epi32_rmbkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpxord (%rdi){1to8}, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xb9,0xef,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 <8 x i32> @llvm.x86.avx512.mask.pxor.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> zeroinitializer, i8 %mask)
  ret <8 x i32> %res
}

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

define <4 x i32> @test_mask_andnot_epi32_rr_128(<4 x i32> %a, <4 x i32> %b) {
; CHECK-LABEL: test_mask_andnot_epi32_rr_128:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vandnps %xmm1, %xmm0, %xmm0 # encoding: [0xc5,0xf8,0x55,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <4 x i32> @llvm.x86.avx512.mask.pandn.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> zeroinitializer, i8 -1)
  ret <4 x i32> %res
}

define <4 x i32> @test_mask_andnot_epi32_rrk_128(<4 x i32> %a, <4 x i32> %b, <4 x i32> %passThru, i8 %mask) {
; X86-LABEL: test_mask_andnot_epi32_rrk_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpandnd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xdf,0xd1]
; X86-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_andnot_epi32_rrk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpandnd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xdf,0xd1]
; X64-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i32> @llvm.x86.avx512.mask.pandn.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> %passThru, i8 %mask)
  ret <4 x i32> %res
}

define <4 x i32> @test_mask_andnot_epi32_rrkz_128(<4 x i32> %a, <4 x i32> %b, i8 %mask) {
; X86-LABEL: test_mask_andnot_epi32_rrkz_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpandnd %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xdf,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_andnot_epi32_rrkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpandnd %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xdf,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i32> @llvm.x86.avx512.mask.pandn.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> zeroinitializer, i8 %mask)
  ret <4 x i32> %res
}

define <4 x i32> @test_mask_andnot_epi32_rm_128(<4 x i32> %a, <4 x i32>* %ptr_b) {
; X86-LABEL: test_mask_andnot_epi32_rm_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vandnps (%eax), %xmm0, %xmm0 # encoding: [0xc5,0xf8,0x55,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_andnot_epi32_rm_128:
; X64:       # %bb.0:
; X64-NEXT:    vandnps (%rdi), %xmm0, %xmm0 # encoding: [0xc5,0xf8,0x55,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %b = load <4 x i32>, <4 x i32>* %ptr_b
  %res = call <4 x i32> @llvm.x86.avx512.mask.pandn.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> zeroinitializer, i8 -1)
  ret <4 x i32> %res
}

define <4 x i32> @test_mask_andnot_epi32_rmk_128(<4 x i32> %a, <4 x i32>* %ptr_b, <4 x i32> %passThru, i8 %mask) {
; X86-LABEL: test_mask_andnot_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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpandnd (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xdf,0x08]
; X86-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_andnot_epi32_rmk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpandnd (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xdf,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 <4 x i32> @llvm.x86.avx512.mask.pandn.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> %passThru, i8 %mask)
  ret <4 x i32> %res
}

define <4 x i32> @test_mask_andnot_epi32_rmkz_128(<4 x i32> %a, <4 x i32>* %ptr_b, i8 %mask) {
; X86-LABEL: test_mask_andnot_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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpandnd (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xdf,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_andnot_epi32_rmkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpandnd (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xdf,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %b = load <4 x i32>, <4 x i32>* %ptr_b
  %res = call <4 x i32> @llvm.x86.avx512.mask.pandn.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> zeroinitializer, i8 %mask)
  ret <4 x i32> %res
}

define <4 x i32> @test_mask_andnot_epi32_rmb_128(<4 x i32> %a, i32* %ptr_b) {
; X86-LABEL: test_mask_andnot_epi32_rmb_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpandnd (%eax){1to4}, %xmm0, %xmm0 # encoding: [0x62,0xf1,0x7d,0x18,0xdf,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_andnot_epi32_rmb_128:
; X64:       # %bb.0:
; X64-NEXT:    vpandnd (%rdi){1to4}, %xmm0, %xmm0 # encoding: [0x62,0xf1,0x7d,0x18,0xdf,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 <4 x i32> @llvm.x86.avx512.mask.pandn.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> zeroinitializer, i8 -1)
  ret <4 x i32> %res
}

define <4 x i32> @test_mask_andnot_epi32_rmbk_128(<4 x i32> %a, i32* %ptr_b, <4 x i32> %passThru, i8 %mask) {
; X86-LABEL: test_mask_andnot_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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpandnd (%eax){1to4}, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x19,0xdf,0x08]
; X86-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_andnot_epi32_rmbk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpandnd (%rdi){1to4}, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x19,0xdf,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 <4 x i32> @llvm.x86.avx512.mask.pandn.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> %passThru, i8 %mask)
  ret <4 x i32> %res
}

define <4 x i32> @test_mask_andnot_epi32_rmbkz_128(<4 x i32> %a, i32* %ptr_b, i8 %mask) {
; X86-LABEL: test_mask_andnot_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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpandnd (%eax){1to4}, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x99,0xdf,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_andnot_epi32_rmbkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpandnd (%rdi){1to4}, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x99,0xdf,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 <4 x i32> @llvm.x86.avx512.mask.pandn.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> zeroinitializer, i8 %mask)
  ret <4 x i32> %res
}

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

define <8 x i32> @test_mask_andnot_epi32_rr_256(<8 x i32> %a, <8 x i32> %b) {
; CHECK-LABEL: test_mask_andnot_epi32_rr_256:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vandnps %ymm1, %ymm0, %ymm0 # encoding: [0xc5,0xfc,0x55,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <8 x i32> @llvm.x86.avx512.mask.pandn.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> zeroinitializer, i8 -1)
  ret <8 x i32> %res
}

define <8 x i32> @test_mask_andnot_epi32_rrk_256(<8 x i32> %a, <8 x i32> %b, <8 x i32> %passThru, i8 %mask) {
; X86-LABEL: test_mask_andnot_epi32_rrk_256:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpandnd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xdf,0xd1]
; X86-NEXT:    vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_andnot_epi32_rrk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpandnd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xdf,0xd1]
; X64-NEXT:    vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x i32> @llvm.x86.avx512.mask.pandn.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> %passThru, i8 %mask)
  ret <8 x i32> %res
}

define <8 x i32> @test_mask_andnot_epi32_rrkz_256(<8 x i32> %a, <8 x i32> %b, i8 %mask) {
; X86-LABEL: test_mask_andnot_epi32_rrkz_256:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpandnd %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xdf,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_andnot_epi32_rrkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpandnd %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xdf,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x i32> @llvm.x86.avx512.mask.pandn.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> zeroinitializer, i8 %mask)
  ret <8 x i32> %res
}

define <8 x i32> @test_mask_andnot_epi32_rm_256(<8 x i32> %a, <8 x i32>* %ptr_b) {
; X86-LABEL: test_mask_andnot_epi32_rm_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vandnps (%eax), %ymm0, %ymm0 # encoding: [0xc5,0xfc,0x55,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_andnot_epi32_rm_256:
; X64:       # %bb.0:
; X64-NEXT:    vandnps (%rdi), %ymm0, %ymm0 # encoding: [0xc5,0xfc,0x55,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %b = load <8 x i32>, <8 x i32>* %ptr_b
  %res = call <8 x i32> @llvm.x86.avx512.mask.pandn.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> zeroinitializer, i8 -1)
  ret <8 x i32> %res
}

define <8 x i32> @test_mask_andnot_epi32_rmk_256(<8 x i32> %a, <8 x i32>* %ptr_b, <8 x i32> %passThru, i8 %mask) {
; X86-LABEL: test_mask_andnot_epi32_rmk_256:
; 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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpandnd (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xdf,0x08]
; X86-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_andnot_epi32_rmk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpandnd (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xdf,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 <8 x i32> @llvm.x86.avx512.mask.pandn.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> %passThru, i8 %mask)
  ret <8 x i32> %res
}

define <8 x i32> @test_mask_andnot_epi32_rmkz_256(<8 x i32> %a, <8 x i32>* %ptr_b, i8 %mask) {
; X86-LABEL: test_mask_andnot_epi32_rmkz_256:
; 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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpandnd (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xdf,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_andnot_epi32_rmkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpandnd (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xdf,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %b = load <8 x i32>, <8 x i32>* %ptr_b
  %res = call <8 x i32> @llvm.x86.avx512.mask.pandn.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> zeroinitializer, i8 %mask)
  ret <8 x i32> %res
}

define <8 x i32> @test_mask_andnot_epi32_rmb_256(<8 x i32> %a, i32* %ptr_b) {
; X86-LABEL: test_mask_andnot_epi32_rmb_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpandnd (%eax){1to8}, %ymm0, %ymm0 # encoding: [0x62,0xf1,0x7d,0x38,0xdf,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_andnot_epi32_rmb_256:
; X64:       # %bb.0:
; X64-NEXT:    vpandnd (%rdi){1to8}, %ymm0, %ymm0 # encoding: [0x62,0xf1,0x7d,0x38,0xdf,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 <8 x i32> @llvm.x86.avx512.mask.pandn.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> zeroinitializer, i8 -1)
  ret <8 x i32> %res
}

define <8 x i32> @test_mask_andnot_epi32_rmbk_256(<8 x i32> %a, i32* %ptr_b, <8 x i32> %passThru, i8 %mask) {
; X86-LABEL: test_mask_andnot_epi32_rmbk_256:
; 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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpandnd (%eax){1to8}, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x39,0xdf,0x08]
; X86-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_andnot_epi32_rmbk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpandnd (%rdi){1to8}, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x39,0xdf,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 <8 x i32> @llvm.x86.avx512.mask.pandn.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> %passThru, i8 %mask)
  ret <8 x i32> %res
}

define <8 x i32> @test_mask_andnot_epi32_rmbkz_256(<8 x i32> %a, i32* %ptr_b, i8 %mask) {
; X86-LABEL: test_mask_andnot_epi32_rmbkz_256:
; 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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpandnd (%eax){1to8}, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xb9,0xdf,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_andnot_epi32_rmbkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpandnd (%rdi){1to8}, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xb9,0xdf,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 <8 x i32> @llvm.x86.avx512.mask.pandn.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> zeroinitializer, i8 %mask)
  ret <8 x i32> %res
}

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

define <2 x i64> @test_mask_andnot_epi64_rr_128(<2 x i64> %a, <2 x i64> %b) {
; CHECK-LABEL: test_mask_andnot_epi64_rr_128:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vandnps %xmm1, %xmm0, %xmm0 # encoding: [0xc5,0xf8,0x55,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <2 x i64> @llvm.x86.avx512.mask.pandn.q.128(<2 x i64> %a, <2 x i64> %b, <2 x i64> zeroinitializer, i8 -1)
  ret <2 x i64> %res
}

define <2 x i64> @test_mask_andnot_epi64_rrk_128(<2 x i64> %a, <2 x i64> %b, <2 x i64> %passThru, i8 %mask) {
; X86-LABEL: test_mask_andnot_epi64_rrk_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpandnq %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0xfd,0x09,0xdf,0xd1]
; X86-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_andnot_epi64_rrk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpandnq %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0xfd,0x09,0xdf,0xd1]
; X64-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x i64> @llvm.x86.avx512.mask.pandn.q.128(<2 x i64> %a, <2 x i64> %b, <2 x i64> %passThru, i8 %mask)
  ret <2 x i64> %res
}

define <2 x i64> @test_mask_andnot_epi64_rrkz_128(<2 x i64> %a, <2 x i64> %b, i8 %mask) {
; X86-LABEL: test_mask_andnot_epi64_rrkz_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpandnq %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0x89,0xdf,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_andnot_epi64_rrkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpandnq %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0x89,0xdf,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x i64> @llvm.x86.avx512.mask.pandn.q.128(<2 x i64> %a, <2 x i64> %b, <2 x i64> zeroinitializer, i8 %mask)
  ret <2 x i64> %res
}

define <2 x i64> @test_mask_andnot_epi64_rm_128(<2 x i64> %a, <2 x i64>* %ptr_b) {
; X86-LABEL: test_mask_andnot_epi64_rm_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vandnps (%eax), %xmm0, %xmm0 # encoding: [0xc5,0xf8,0x55,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_andnot_epi64_rm_128:
; X64:       # %bb.0:
; X64-NEXT:    vandnps (%rdi), %xmm0, %xmm0 # encoding: [0xc5,0xf8,0x55,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %b = load <2 x i64>, <2 x i64>* %ptr_b
  %res = call <2 x i64> @llvm.x86.avx512.mask.pandn.q.128(<2 x i64> %a, <2 x i64> %b, <2 x i64> zeroinitializer, i8 -1)
  ret <2 x i64> %res
}

define <2 x i64> @test_mask_andnot_epi64_rmk_128(<2 x i64> %a, <2 x i64>* %ptr_b, <2 x i64> %passThru, i8 %mask) {
; X86-LABEL: test_mask_andnot_epi64_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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpandnq (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xfd,0x09,0xdf,0x08]
; X86-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_andnot_epi64_rmk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpandnq (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xfd,0x09,0xdf,0x0f]
; X64-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %b = load <2 x i64>, <2 x i64>* %ptr_b
  %res = call <2 x i64> @llvm.x86.avx512.mask.pandn.q.128(<2 x i64> %a, <2 x i64> %b, <2 x i64> %passThru, i8 %mask)
  ret <2 x i64> %res
}

define <2 x i64> @test_mask_andnot_epi64_rmkz_128(<2 x i64> %a, <2 x i64>* %ptr_b, i8 %mask) {
; X86-LABEL: test_mask_andnot_epi64_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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpandnq (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0x89,0xdf,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_andnot_epi64_rmkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpandnq (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0x89,0xdf,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %b = load <2 x i64>, <2 x i64>* %ptr_b
  %res = call <2 x i64> @llvm.x86.avx512.mask.pandn.q.128(<2 x i64> %a, <2 x i64> %b, <2 x i64> zeroinitializer, i8 %mask)
  ret <2 x i64> %res
}

define <2 x i64> @test_mask_andnot_epi64_rmb_128(<2 x i64> %a, i64* %ptr_b) {
; X86-LABEL: test_mask_andnot_epi64_rmb_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpbroadcastq (%eax), %xmm1 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x59,0x08]
; X86-NEXT:    vpandn %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xdf,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_andnot_epi64_rmb_128:
; X64:       # %bb.0:
; X64-NEXT:    vpandnq (%rdi){1to2}, %xmm0, %xmm0 # encoding: [0x62,0xf1,0xfd,0x18,0xdf,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %q = load i64, i64* %ptr_b
  %vecinit.i = insertelement <2 x i64> undef, i64 %q, i32 0
  %b = shufflevector <2 x i64> %vecinit.i, <2 x i64> undef, <2 x i32> zeroinitializer
  %res = call <2 x i64> @llvm.x86.avx512.mask.pandn.q.128(<2 x i64> %a, <2 x i64> %b, <2 x i64> zeroinitializer, i8 -1)
  ret <2 x i64> %res
}

define <2 x i64> @test_mask_andnot_epi64_rmbk_128(<2 x i64> %a, i64* %ptr_b, <2 x i64> %passThru, i8 %mask) {
; X86-LABEL: test_mask_andnot_epi64_rmbk_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpbroadcastq (%eax), %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x59,0x10]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpandnq %xmm2, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xfd,0x09,0xdf,0xca]
; X86-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_andnot_epi64_rmbk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpandnq (%rdi){1to2}, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xfd,0x19,0xdf,0x0f]
; X64-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %q = load i64, i64* %ptr_b
  %vecinit.i = insertelement <2 x i64> undef, i64 %q, i32 0
  %b = shufflevector <2 x i64> %vecinit.i, <2 x i64> undef, <2 x i32> zeroinitializer
  %res = call <2 x i64> @llvm.x86.avx512.mask.pandn.q.128(<2 x i64> %a, <2 x i64> %b, <2 x i64> %passThru, i8 %mask)
  ret <2 x i64> %res
}

define <2 x i64> @test_mask_andnot_epi64_rmbkz_128(<2 x i64> %a, i64* %ptr_b, i8 %mask) {
; X86-LABEL: test_mask_andnot_epi64_rmbkz_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpbroadcastq (%eax), %xmm1 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x59,0x08]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpandnq %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0x89,0xdf,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_andnot_epi64_rmbkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpandnq (%rdi){1to2}, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0x99,0xdf,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %q = load i64, i64* %ptr_b
  %vecinit.i = insertelement <2 x i64> undef, i64 %q, i32 0
  %b = shufflevector <2 x i64> %vecinit.i, <2 x i64> undef, <2 x i32> zeroinitializer
  %res = call <2 x i64> @llvm.x86.avx512.mask.pandn.q.128(<2 x i64> %a, <2 x i64> %b, <2 x i64> zeroinitializer, i8 %mask)
  ret <2 x i64> %res
}

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

define <4 x i64> @test_mask_andnot_epi64_rr_256(<4 x i64> %a, <4 x i64> %b) {
; CHECK-LABEL: test_mask_andnot_epi64_rr_256:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vandnps %ymm1, %ymm0, %ymm0 # encoding: [0xc5,0xfc,0x55,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <4 x i64> @llvm.x86.avx512.mask.pandn.q.256(<4 x i64> %a, <4 x i64> %b, <4 x i64> zeroinitializer, i8 -1)
  ret <4 x i64> %res
}

define <4 x i64> @test_mask_andnot_epi64_rrk_256(<4 x i64> %a, <4 x i64> %b, <4 x i64> %passThru, i8 %mask) {
; X86-LABEL: test_mask_andnot_epi64_rrk_256:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpandnq %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0xfd,0x29,0xdf,0xd1]
; X86-NEXT:    vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_andnot_epi64_rrk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpandnq %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0xfd,0x29,0xdf,0xd1]
; X64-NEXT:    vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i64> @llvm.x86.avx512.mask.pandn.q.256(<4 x i64> %a, <4 x i64> %b, <4 x i64> %passThru, i8 %mask)
  ret <4 x i64> %res
}

define <4 x i64> @test_mask_andnot_epi64_rrkz_256(<4 x i64> %a, <4 x i64> %b, i8 %mask) {
; X86-LABEL: test_mask_andnot_epi64_rrkz_256:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpandnq %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0xa9,0xdf,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_andnot_epi64_rrkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpandnq %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0xa9,0xdf,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i64> @llvm.x86.avx512.mask.pandn.q.256(<4 x i64> %a, <4 x i64> %b, <4 x i64> zeroinitializer, i8 %mask)
  ret <4 x i64> %res
}

define <4 x i64> @test_mask_andnot_epi64_rm_256(<4 x i64> %a, <4 x i64>* %ptr_b) {
; X86-LABEL: test_mask_andnot_epi64_rm_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vandnps (%eax), %ymm0, %ymm0 # encoding: [0xc5,0xfc,0x55,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_andnot_epi64_rm_256:
; X64:       # %bb.0:
; X64-NEXT:    vandnps (%rdi), %ymm0, %ymm0 # encoding: [0xc5,0xfc,0x55,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %b = load <4 x i64>, <4 x i64>* %ptr_b
  %res = call <4 x i64> @llvm.x86.avx512.mask.pandn.q.256(<4 x i64> %a, <4 x i64> %b, <4 x i64> zeroinitializer, i8 -1)
  ret <4 x i64> %res
}

define <4 x i64> @test_mask_andnot_epi64_rmk_256(<4 x i64> %a, <4 x i64>* %ptr_b, <4 x i64> %passThru, i8 %mask) {
; X86-LABEL: test_mask_andnot_epi64_rmk_256:
; 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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpandnq (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0xfd,0x29,0xdf,0x08]
; X86-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_andnot_epi64_rmk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpandnq (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0xfd,0x29,0xdf,0x0f]
; X64-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %b = load <4 x i64>, <4 x i64>* %ptr_b
  %res = call <4 x i64> @llvm.x86.avx512.mask.pandn.q.256(<4 x i64> %a, <4 x i64> %b, <4 x i64> %passThru, i8 %mask)
  ret <4 x i64> %res
}

define <4 x i64> @test_mask_andnot_epi64_rmkz_256(<4 x i64> %a, <4 x i64>* %ptr_b, i8 %mask) {
; X86-LABEL: test_mask_andnot_epi64_rmkz_256:
; 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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpandnq (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0xa9,0xdf,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_andnot_epi64_rmkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpandnq (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0xa9,0xdf,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %b = load <4 x i64>, <4 x i64>* %ptr_b
  %res = call <4 x i64> @llvm.x86.avx512.mask.pandn.q.256(<4 x i64> %a, <4 x i64> %b, <4 x i64> zeroinitializer, i8 %mask)
  ret <4 x i64> %res
}

define <4 x i64> @test_mask_andnot_epi64_rmb_256(<4 x i64> %a, i64* %ptr_b) {
; X86-LABEL: test_mask_andnot_epi64_rmb_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vmovsd (%eax), %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xfb,0x10,0x08]
; X86-NEXT:    # xmm1 = mem[0],zero
; X86-NEXT:    vbroadcastsd %xmm1, %ymm1 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x19,0xc9]
; X86-NEXT:    vandnps %ymm1, %ymm0, %ymm0 # encoding: [0xc5,0xfc,0x55,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_andnot_epi64_rmb_256:
; X64:       # %bb.0:
; X64-NEXT:    vpandnq (%rdi){1to4}, %ymm0, %ymm0 # encoding: [0x62,0xf1,0xfd,0x38,0xdf,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %q = load i64, i64* %ptr_b
  %vecinit.i = insertelement <4 x i64> undef, i64 %q, i32 0
  %b = shufflevector <4 x i64> %vecinit.i, <4 x i64> undef, <4 x i32> zeroinitializer
  %res = call <4 x i64> @llvm.x86.avx512.mask.pandn.q.256(<4 x i64> %a, <4 x i64> %b, <4 x i64> zeroinitializer, i8 -1)
  ret <4 x i64> %res
}

define <4 x i64> @test_mask_andnot_epi64_rmbk_256(<4 x i64> %a, i64* %ptr_b, <4 x i64> %passThru, i8 %mask) {
; X86-LABEL: test_mask_andnot_epi64_rmbk_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vmovq (%eax), %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xfa,0x7e,0x10]
; X86-NEXT:    # xmm2 = mem[0],zero
; X86-NEXT:    vpbroadcastq %xmm2, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x59,0xd2]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpandnq %ymm2, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0xfd,0x29,0xdf,0xca]
; X86-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_andnot_epi64_rmbk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpandnq (%rdi){1to4}, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0xfd,0x39,0xdf,0x0f]
; X64-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %q = load i64, i64* %ptr_b
  %vecinit.i = insertelement <4 x i64> undef, i64 %q, i32 0
  %b = shufflevector <4 x i64> %vecinit.i, <4 x i64> undef, <4 x i32> zeroinitializer
  %res = call <4 x i64> @llvm.x86.avx512.mask.pandn.q.256(<4 x i64> %a, <4 x i64> %b, <4 x i64> %passThru, i8 %mask)
  ret <4 x i64> %res
}

define <4 x i64> @test_mask_andnot_epi64_rmbkz_256(<4 x i64> %a, i64* %ptr_b, i8 %mask) {
; X86-LABEL: test_mask_andnot_epi64_rmbkz_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vmovq (%eax), %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xfa,0x7e,0x08]
; X86-NEXT:    # xmm1 = mem[0],zero
; X86-NEXT:    vpbroadcastq %xmm1, %ymm1 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x59,0xc9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpandnq %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0xa9,0xdf,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_andnot_epi64_rmbkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpandnq (%rdi){1to4}, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0xb9,0xdf,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %q = load i64, i64* %ptr_b
  %vecinit.i = insertelement <4 x i64> undef, i64 %q, i32 0
  %b = shufflevector <4 x i64> %vecinit.i, <4 x i64> undef, <4 x i32> zeroinitializer
  %res = call <4 x i64> @llvm.x86.avx512.mask.pandn.q.256(<4 x i64> %a, <4 x i64> %b, <4 x i64> zeroinitializer, i8 %mask)
  ret <4 x i64> %res
}

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

define <4 x i32> @test_mask_add_epi32_rr_128(<4 x i32> %a, <4 x i32> %b) {
; CHECK-LABEL: test_mask_add_epi32_rr_128:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vpaddd %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfe,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <4 x i32> @llvm.x86.avx512.mask.padd.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> zeroinitializer, i8 -1)
  ret <4 x i32> %res
}

define <4 x i32> @test_mask_add_epi32_rrk_128(<4 x i32> %a, <4 x i32> %b, <4 x i32> %passThru, i8 %mask) {
; X86-LABEL: test_mask_add_epi32_rrk_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpaddd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xfe,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_epi32_rrk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpaddd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xfe,0xd1]
; X64-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i32> @llvm.x86.avx512.mask.padd.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> %passThru, i8 %mask)
  ret <4 x i32> %res
}

define <4 x i32> @test_mask_add_epi32_rrkz_128(<4 x i32> %a, <4 x i32> %b, i8 %mask) {
; X86-LABEL: test_mask_add_epi32_rrkz_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpaddd %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xfe,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_add_epi32_rrkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpaddd %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xfe,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i32> @llvm.x86.avx512.mask.padd.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> zeroinitializer, i8 %mask)
  ret <4 x i32> %res
}

define <4 x i32> @test_mask_add_epi32_rm_128(<4 x i32> %a, <4 x i32>* %ptr_b) {
; X86-LABEL: test_mask_add_epi32_rm_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpaddd (%eax), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfe,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_add_epi32_rm_128:
; X64:       # %bb.0:
; X64-NEXT:    vpaddd (%rdi), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfe,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %b = load <4 x i32>, <4 x i32>* %ptr_b
  %res = call <4 x i32> @llvm.x86.avx512.mask.padd.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> zeroinitializer, i8 -1)
  ret <4 x i32> %res
}

define <4 x i32> @test_mask_add_epi32_rmk_128(<4 x i32> %a, <4 x i32>* %ptr_b, <4 x i32> %passThru, i8 %mask) {
; X86-LABEL: test_mask_add_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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpaddd (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xfe,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_epi32_rmk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpaddd (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xfe,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 <4 x i32> @llvm.x86.avx512.mask.padd.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> %passThru, i8 %mask)
  ret <4 x i32> %res
}

define <4 x i32> @test_mask_add_epi32_rmkz_128(<4 x i32> %a, <4 x i32>* %ptr_b, i8 %mask) {
; X86-LABEL: test_mask_add_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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpaddd (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xfe,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_add_epi32_rmkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpaddd (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xfe,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %b = load <4 x i32>, <4 x i32>* %ptr_b
  %res = call <4 x i32> @llvm.x86.avx512.mask.padd.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> zeroinitializer, i8 %mask)
  ret <4 x i32> %res
}

define <4 x i32> @test_mask_add_epi32_rmb_128(<4 x i32> %a, i32* %ptr_b) {
; X86-LABEL: test_mask_add_epi32_rmb_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpaddd (%eax){1to4}, %xmm0, %xmm0 # encoding: [0x62,0xf1,0x7d,0x18,0xfe,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_add_epi32_rmb_128:
; X64:       # %bb.0:
; X64-NEXT:    vpaddd (%rdi){1to4}, %xmm0, %xmm0 # encoding: [0x62,0xf1,0x7d,0x18,0xfe,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 <4 x i32> @llvm.x86.avx512.mask.padd.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> zeroinitializer, i8 -1)
  ret <4 x i32> %res
}

define <4 x i32> @test_mask_add_epi32_rmbk_128(<4 x i32> %a, i32* %ptr_b, <4 x i32> %passThru, i8 %mask) {
; X86-LABEL: test_mask_add_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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpaddd (%eax){1to4}, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x19,0xfe,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_epi32_rmbk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpaddd (%rdi){1to4}, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x19,0xfe,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 <4 x i32> @llvm.x86.avx512.mask.padd.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> %passThru, i8 %mask)
  ret <4 x i32> %res
}

define <4 x i32> @test_mask_add_epi32_rmbkz_128(<4 x i32> %a, i32* %ptr_b, i8 %mask) {
; X86-LABEL: test_mask_add_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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpaddd (%eax){1to4}, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x99,0xfe,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_add_epi32_rmbkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpaddd (%rdi){1to4}, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x99,0xfe,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 <4 x i32> @llvm.x86.avx512.mask.padd.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> zeroinitializer, i8 %mask)
  ret <4 x i32> %res
}

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

define <4 x i32> @test_mask_sub_epi32_rr_128(<4 x i32> %a, <4 x i32> %b) {
; CHECK-LABEL: test_mask_sub_epi32_rr_128:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vpsubd %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfa,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <4 x i32> @llvm.x86.avx512.mask.psub.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> zeroinitializer, i8 -1)
  ret <4 x i32> %res
}

define <4 x i32> @test_mask_sub_epi32_rrk_128(<4 x i32> %a, <4 x i32> %b, <4 x i32> %passThru, i8 %mask) {
; X86-LABEL: test_mask_sub_epi32_rrk_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpsubd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xfa,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_epi32_rrk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpsubd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xfa,0xd1]
; X64-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i32> @llvm.x86.avx512.mask.psub.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> %passThru, i8 %mask)
  ret <4 x i32> %res
}

define <4 x i32> @test_mask_sub_epi32_rrkz_128(<4 x i32> %a, <4 x i32> %b, i8 %mask) {
; X86-LABEL: test_mask_sub_epi32_rrkz_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpsubd %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xfa,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_sub_epi32_rrkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpsubd %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xfa,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i32> @llvm.x86.avx512.mask.psub.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> zeroinitializer, i8 %mask)
  ret <4 x i32> %res
}

define <4 x i32> @test_mask_sub_epi32_rm_128(<4 x i32> %a, <4 x i32>* %ptr_b) {
; X86-LABEL: test_mask_sub_epi32_rm_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpsubd (%eax), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfa,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_sub_epi32_rm_128:
; X64:       # %bb.0:
; X64-NEXT:    vpsubd (%rdi), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfa,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %b = load <4 x i32>, <4 x i32>* %ptr_b
  %res = call <4 x i32> @llvm.x86.avx512.mask.psub.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> zeroinitializer, i8 -1)
  ret <4 x i32> %res
}

define <4 x i32> @test_mask_sub_epi32_rmk_128(<4 x i32> %a, <4 x i32>* %ptr_b, <4 x i32> %passThru, i8 %mask) {
; X86-LABEL: test_mask_sub_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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpsubd (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xfa,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_epi32_rmk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpsubd (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xfa,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 <4 x i32> @llvm.x86.avx512.mask.psub.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> %passThru, i8 %mask)
  ret <4 x i32> %res
}

define <4 x i32> @test_mask_sub_epi32_rmkz_128(<4 x i32> %a, <4 x i32>* %ptr_b, i8 %mask) {
; X86-LABEL: test_mask_sub_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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpsubd (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xfa,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_sub_epi32_rmkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpsubd (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xfa,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %b = load <4 x i32>, <4 x i32>* %ptr_b
  %res = call <4 x i32> @llvm.x86.avx512.mask.psub.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> zeroinitializer, i8 %mask)
  ret <4 x i32> %res
}

define <4 x i32> @test_mask_sub_epi32_rmb_128(<4 x i32> %a, i32* %ptr_b) {
; X86-LABEL: test_mask_sub_epi32_rmb_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpsubd (%eax){1to4}, %xmm0, %xmm0 # encoding: [0x62,0xf1,0x7d,0x18,0xfa,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_sub_epi32_rmb_128:
; X64:       # %bb.0:
; X64-NEXT:    vpsubd (%rdi){1to4}, %xmm0, %xmm0 # encoding: [0x62,0xf1,0x7d,0x18,0xfa,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 <4 x i32> @llvm.x86.avx512.mask.psub.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> zeroinitializer, i8 -1)
  ret <4 x i32> %res
}

define <4 x i32> @test_mask_sub_epi32_rmbk_128(<4 x i32> %a, i32* %ptr_b, <4 x i32> %passThru, i8 %mask) {
; X86-LABEL: test_mask_sub_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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpsubd (%eax){1to4}, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x19,0xfa,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_epi32_rmbk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpsubd (%rdi){1to4}, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x19,0xfa,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 <4 x i32> @llvm.x86.avx512.mask.psub.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> %passThru, i8 %mask)
  ret <4 x i32> %res
}

define <4 x i32> @test_mask_sub_epi32_rmbkz_128(<4 x i32> %a, i32* %ptr_b, i8 %mask) {
; X86-LABEL: test_mask_sub_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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpsubd (%eax){1to4}, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x99,0xfa,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_sub_epi32_rmbkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpsubd (%rdi){1to4}, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x99,0xfa,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 <4 x i32> @llvm.x86.avx512.mask.psub.d.128(<4 x i32> %a, <4 x i32> %b, <4 x i32> zeroinitializer, i8 %mask)
  ret <4 x i32> %res
}

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

define <8 x i32> @test_mask_sub_epi32_rr_256(<8 x i32> %a, <8 x i32> %b) {
; CHECK-LABEL: test_mask_sub_epi32_rr_256:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vpsubd %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfa,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <8 x i32> @llvm.x86.avx512.mask.psub.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> zeroinitializer, i8 -1)
  ret <8 x i32> %res
}

define <8 x i32> @test_mask_sub_epi32_rrk_256(<8 x i32> %a, <8 x i32> %b, <8 x i32> %passThru, i8 %mask) {
; X86-LABEL: test_mask_sub_epi32_rrk_256:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpsubd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xfa,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_epi32_rrk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpsubd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xfa,0xd1]
; X64-NEXT:    vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x i32> @llvm.x86.avx512.mask.psub.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> %passThru, i8 %mask)
  ret <8 x i32> %res
}

define <8 x i32> @test_mask_sub_epi32_rrkz_256(<8 x i32> %a, <8 x i32> %b, i8 %mask) {
; X86-LABEL: test_mask_sub_epi32_rrkz_256:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpsubd %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xfa,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_sub_epi32_rrkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpsubd %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xfa,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x i32> @llvm.x86.avx512.mask.psub.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> zeroinitializer, i8 %mask)
  ret <8 x i32> %res
}

define <8 x i32> @test_mask_sub_epi32_rm_256(<8 x i32> %a, <8 x i32>* %ptr_b) {
; X86-LABEL: test_mask_sub_epi32_rm_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpsubd (%eax), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfa,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_sub_epi32_rm_256:
; X64:       # %bb.0:
; X64-NEXT:    vpsubd (%rdi), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfa,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %b = load <8 x i32>, <8 x i32>* %ptr_b
  %res = call <8 x i32> @llvm.x86.avx512.mask.psub.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> zeroinitializer, i8 -1)
  ret <8 x i32> %res
}

define <8 x i32> @test_mask_sub_epi32_rmk_256(<8 x i32> %a, <8 x i32>* %ptr_b, <8 x i32> %passThru, i8 %mask) {
; X86-LABEL: test_mask_sub_epi32_rmk_256:
; 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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpsubd (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xfa,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_epi32_rmk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpsubd (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xfa,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 <8 x i32> @llvm.x86.avx512.mask.psub.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> %passThru, i8 %mask)
  ret <8 x i32> %res
}

define <8 x i32> @test_mask_sub_epi32_rmkz_256(<8 x i32> %a, <8 x i32>* %ptr_b, i8 %mask) {
; X86-LABEL: test_mask_sub_epi32_rmkz_256:
; 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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpsubd (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xfa,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_sub_epi32_rmkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpsubd (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xfa,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %b = load <8 x i32>, <8 x i32>* %ptr_b
  %res = call <8 x i32> @llvm.x86.avx512.mask.psub.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> zeroinitializer, i8 %mask)
  ret <8 x i32> %res
}

define <8 x i32> @test_mask_sub_epi32_rmb_256(<8 x i32> %a, i32* %ptr_b) {
; X86-LABEL: test_mask_sub_epi32_rmb_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpsubd (%eax){1to8}, %ymm0, %ymm0 # encoding: [0x62,0xf1,0x7d,0x38,0xfa,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_sub_epi32_rmb_256:
; X64:       # %bb.0:
; X64-NEXT:    vpsubd (%rdi){1to8}, %ymm0, %ymm0 # encoding: [0x62,0xf1,0x7d,0x38,0xfa,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 <8 x i32> @llvm.x86.avx512.mask.psub.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> zeroinitializer, i8 -1)
  ret <8 x i32> %res
}

define <8 x i32> @test_mask_sub_epi32_rmbk_256(<8 x i32> %a, i32* %ptr_b, <8 x i32> %passThru, i8 %mask) {
; X86-LABEL: test_mask_sub_epi32_rmbk_256:
; 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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpsubd (%eax){1to8}, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x39,0xfa,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_epi32_rmbk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpsubd (%rdi){1to8}, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x39,0xfa,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 <8 x i32> @llvm.x86.avx512.mask.psub.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> %passThru, i8 %mask)
  ret <8 x i32> %res
}

define <8 x i32> @test_mask_sub_epi32_rmbkz_256(<8 x i32> %a, i32* %ptr_b, i8 %mask) {
; X86-LABEL: test_mask_sub_epi32_rmbkz_256:
; 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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpsubd (%eax){1to8}, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xb9,0xfa,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_sub_epi32_rmbkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpsubd (%rdi){1to8}, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xb9,0xfa,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 <8 x i32> @llvm.x86.avx512.mask.psub.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> zeroinitializer, i8 %mask)
  ret <8 x i32> %res
}

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

define <8 x i32> @test_mask_add_epi32_rr_256(<8 x i32> %a, <8 x i32> %b) {
; CHECK-LABEL: test_mask_add_epi32_rr_256:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vpaddd %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfe,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <8 x i32> @llvm.x86.avx512.mask.padd.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> zeroinitializer, i8 -1)
  ret <8 x i32> %res
}

define <8 x i32> @test_mask_add_epi32_rrk_256(<8 x i32> %a, <8 x i32> %b, <8 x i32> %passThru, i8 %mask) {
; X86-LABEL: test_mask_add_epi32_rrk_256:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpaddd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xfe,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_epi32_rrk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpaddd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xfe,0xd1]
; X64-NEXT:    vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x i32> @llvm.x86.avx512.mask.padd.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> %passThru, i8 %mask)
  ret <8 x i32> %res
}

define <8 x i32> @test_mask_add_epi32_rrkz_256(<8 x i32> %a, <8 x i32> %b, i8 %mask) {
; X86-LABEL: test_mask_add_epi32_rrkz_256:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpaddd %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xfe,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_add_epi32_rrkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpaddd %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xfe,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x i32> @llvm.x86.avx512.mask.padd.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> zeroinitializer, i8 %mask)
  ret <8 x i32> %res
}

define <8 x i32> @test_mask_add_epi32_rm_256(<8 x i32> %a, <8 x i32>* %ptr_b) {
; X86-LABEL: test_mask_add_epi32_rm_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpaddd (%eax), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfe,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_add_epi32_rm_256:
; X64:       # %bb.0:
; X64-NEXT:    vpaddd (%rdi), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfe,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %b = load <8 x i32>, <8 x i32>* %ptr_b
  %res = call <8 x i32> @llvm.x86.avx512.mask.padd.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> zeroinitializer, i8 -1)
  ret <8 x i32> %res
}

define <8 x i32> @test_mask_add_epi32_rmk_256(<8 x i32> %a, <8 x i32>* %ptr_b, <8 x i32> %passThru, i8 %mask) {
; X86-LABEL: test_mask_add_epi32_rmk_256:
; 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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpaddd (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xfe,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_epi32_rmk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpaddd (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xfe,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 <8 x i32> @llvm.x86.avx512.mask.padd.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> %passThru, i8 %mask)
  ret <8 x i32> %res
}

define <8 x i32> @test_mask_add_epi32_rmkz_256(<8 x i32> %a, <8 x i32>* %ptr_b, i8 %mask) {
; X86-LABEL: test_mask_add_epi32_rmkz_256:
; 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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpaddd (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xfe,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_add_epi32_rmkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpaddd (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xfe,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %b = load <8 x i32>, <8 x i32>* %ptr_b
  %res = call <8 x i32> @llvm.x86.avx512.mask.padd.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> zeroinitializer, i8 %mask)
  ret <8 x i32> %res
}

define <8 x i32> @test_mask_add_epi32_rmb_256(<8 x i32> %a, i32* %ptr_b) {
; X86-LABEL: test_mask_add_epi32_rmb_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpaddd (%eax){1to8}, %ymm0, %ymm0 # encoding: [0x62,0xf1,0x7d,0x38,0xfe,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_add_epi32_rmb_256:
; X64:       # %bb.0:
; X64-NEXT:    vpaddd (%rdi){1to8}, %ymm0, %ymm0 # encoding: [0x62,0xf1,0x7d,0x38,0xfe,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 <8 x i32> @llvm.x86.avx512.mask.padd.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> zeroinitializer, i8 -1)
  ret <8 x i32> %res
}

define <8 x i32> @test_mask_add_epi32_rmbk_256(<8 x i32> %a, i32* %ptr_b, <8 x i32> %passThru, i8 %mask) {
; X86-LABEL: test_mask_add_epi32_rmbk_256:
; 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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpaddd (%eax){1to8}, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x39,0xfe,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_epi32_rmbk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpaddd (%rdi){1to8}, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x39,0xfe,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 <8 x i32> @llvm.x86.avx512.mask.padd.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> %passThru, i8 %mask)
  ret <8 x i32> %res
}

define <8 x i32> @test_mask_add_epi32_rmbkz_256(<8 x i32> %a, i32* %ptr_b, i8 %mask) {
; X86-LABEL: test_mask_add_epi32_rmbkz_256:
; 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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpaddd (%eax){1to8}, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xb9,0xfe,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_add_epi32_rmbkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpaddd (%rdi){1to8}, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xb9,0xfe,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 <8 x i32> @llvm.x86.avx512.mask.padd.d.256(<8 x i32> %a, <8 x i32> %b, <8 x i32> zeroinitializer, i8 %mask)
  ret <8 x i32> %res
}

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

define <8 x float> @test_mm512_maskz_add_ps_256(<8 x float> %a0, <8 x float> %a1, i8 %mask) {
; X86-LABEL: test_mm512_maskz_add_ps_256:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vaddps %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7c,0xa9,0x58,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mm512_maskz_add_ps_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vaddps %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7c,0xa9,0x58,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask.add.ps.256(<8 x float> %a0, <8 x float> %a1, <8 x float>zeroinitializer, i8 %mask)
  ret <8 x float> %res
}

define <8 x float> @test_mm512_mask_add_ps_256(<8 x float> %a0, <8 x float> %a1, <8 x float> %src, i8 %mask) {
; X86-LABEL: test_mm512_mask_add_ps_256:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vaddps %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7c,0x29,0x58,0xd1]
; X86-NEXT:    vmovaps %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x28,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mm512_mask_add_ps_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vaddps %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7c,0x29,0x58,0xd1]
; X64-NEXT:    vmovaps %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x28,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask.add.ps.256(<8 x float> %a0, <8 x float> %a1, <8 x float> %src, i8 %mask)
  ret <8 x float> %res
}

define <8 x float> @test_mm512_add_ps_256(<8 x float> %a0, <8 x float> %a1, i8 %mask) {
; CHECK-LABEL: test_mm512_add_ps_256:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vaddps %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x58,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask.add.ps.256(<8 x float> %a0, <8 x float> %a1, <8 x float>zeroinitializer, i8 -1)
  ret <8 x float> %res
}
declare <8 x float> @llvm.x86.avx512.mask.add.ps.256(<8 x float>, <8 x float>, <8 x float>, i8)

define <4 x float> @test_mm512_maskz_add_ps_128(<4 x float> %a0, <4 x float> %a1, i8 %mask) {
; X86-LABEL: test_mm512_maskz_add_ps_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vaddps %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7c,0x89,0x58,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mm512_maskz_add_ps_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vaddps %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7c,0x89,0x58,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask.add.ps.128(<4 x float> %a0, <4 x float> %a1, <4 x float>zeroinitializer, i8 %mask)
  ret <4 x float> %res
}

define <4 x float> @test_mm512_mask_add_ps_128(<4 x float> %a0, <4 x float> %a1, <4 x float> %src, i8 %mask) {
; X86-LABEL: test_mm512_mask_add_ps_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vaddps %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7c,0x09,0x58,0xd1]
; X86-NEXT:    vmovaps %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x28,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mm512_mask_add_ps_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vaddps %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7c,0x09,0x58,0xd1]
; X64-NEXT:    vmovaps %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x28,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask.add.ps.128(<4 x float> %a0, <4 x float> %a1, <4 x float> %src, i8 %mask)
  ret <4 x float> %res
}

define <4 x float> @test_mm512_add_ps_128(<4 x float> %a0, <4 x float> %a1, i8 %mask) {
; CHECK-LABEL: test_mm512_add_ps_128:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vaddps %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x58,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask.add.ps.128(<4 x float> %a0, <4 x float> %a1, <4 x float>zeroinitializer, i8 -1)
  ret <4 x float> %res
}
declare <4 x float> @llvm.x86.avx512.mask.add.ps.128(<4 x float>, <4 x float>, <4 x float>, i8)

define <8 x float> @test_mm512_maskz_sub_ps_256(<8 x float> %a0, <8 x float> %a1, i8 %mask) {
; X86-LABEL: test_mm512_maskz_sub_ps_256:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vsubps %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7c,0xa9,0x5c,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mm512_maskz_sub_ps_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vsubps %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7c,0xa9,0x5c,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask.sub.ps.256(<8 x float> %a0, <8 x float> %a1, <8 x float>zeroinitializer, i8 %mask)
  ret <8 x float> %res
}

define <8 x float> @test_mm512_mask_sub_ps_256(<8 x float> %a0, <8 x float> %a1, <8 x float> %src, i8 %mask) {
; X86-LABEL: test_mm512_mask_sub_ps_256:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vsubps %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7c,0x29,0x5c,0xd1]
; X86-NEXT:    vmovaps %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x28,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mm512_mask_sub_ps_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vsubps %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7c,0x29,0x5c,0xd1]
; X64-NEXT:    vmovaps %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x28,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask.sub.ps.256(<8 x float> %a0, <8 x float> %a1, <8 x float> %src, i8 %mask)
  ret <8 x float> %res
}

define <8 x float> @test_mm512_sub_ps_256(<8 x float> %a0, <8 x float> %a1, i8 %mask) {
; CHECK-LABEL: test_mm512_sub_ps_256:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vsubps %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x5c,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask.sub.ps.256(<8 x float> %a0, <8 x float> %a1, <8 x float>zeroinitializer, i8 -1)
  ret <8 x float> %res
}
declare <8 x float> @llvm.x86.avx512.mask.sub.ps.256(<8 x float>, <8 x float>, <8 x float>, i8)

define <4 x float> @test_mm512_maskz_sub_ps_128(<4 x float> %a0, <4 x float> %a1, i8 %mask) {
; X86-LABEL: test_mm512_maskz_sub_ps_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vsubps %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7c,0x89,0x5c,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mm512_maskz_sub_ps_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vsubps %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7c,0x89,0x5c,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask.sub.ps.128(<4 x float> %a0, <4 x float> %a1, <4 x float>zeroinitializer, i8 %mask)
  ret <4 x float> %res
}

define <4 x float> @test_mm512_mask_sub_ps_128(<4 x float> %a0, <4 x float> %a1, <4 x float> %src, i8 %mask) {
; X86-LABEL: test_mm512_mask_sub_ps_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vsubps %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7c,0x09,0x5c,0xd1]
; X86-NEXT:    vmovaps %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x28,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mm512_mask_sub_ps_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vsubps %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7c,0x09,0x5c,0xd1]
; X64-NEXT:    vmovaps %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x28,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask.sub.ps.128(<4 x float> %a0, <4 x float> %a1, <4 x float> %src, i8 %mask)
  ret <4 x float> %res
}

define <4 x float> @test_mm512_sub_ps_128(<4 x float> %a0, <4 x float> %a1, i8 %mask) {
; CHECK-LABEL: test_mm512_sub_ps_128:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vsubps %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x5c,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask.sub.ps.128(<4 x float> %a0, <4 x float> %a1, <4 x float>zeroinitializer, i8 -1)
  ret <4 x float> %res
}
declare <4 x float> @llvm.x86.avx512.mask.sub.ps.128(<4 x float>, <4 x float>, <4 x float>, i8)

define <8 x float> @test_mm512_maskz_mul_ps_256(<8 x float> %a0, <8 x float> %a1, i8 %mask) {
; X86-LABEL: test_mm512_maskz_mul_ps_256:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vmulps %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7c,0xa9,0x59,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mm512_maskz_mul_ps_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vmulps %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7c,0xa9,0x59,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask.mul.ps.256(<8 x float> %a0, <8 x float> %a1, <8 x float>zeroinitializer, i8 %mask)
  ret <8 x float> %res
}

define <8 x float> @test_mm512_mask_mul_ps_256(<8 x float> %a0, <8 x float> %a1, <8 x float> %src, i8 %mask) {
; X86-LABEL: test_mm512_mask_mul_ps_256:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vmulps %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7c,0x29,0x59,0xd1]
; X86-NEXT:    vmovaps %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x28,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mm512_mask_mul_ps_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vmulps %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7c,0x29,0x59,0xd1]
; X64-NEXT:    vmovaps %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x28,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask.mul.ps.256(<8 x float> %a0, <8 x float> %a1, <8 x float> %src, i8 %mask)
  ret <8 x float> %res
}

define <8 x float> @test_mm512_mul_ps_256(<8 x float> %a0, <8 x float> %a1, i8 %mask) {
; CHECK-LABEL: test_mm512_mul_ps_256:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vmulps %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x59,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask.mul.ps.256(<8 x float> %a0, <8 x float> %a1, <8 x float>zeroinitializer, i8 -1)
  ret <8 x float> %res
}
declare <8 x float> @llvm.x86.avx512.mask.mul.ps.256(<8 x float>, <8 x float>, <8 x float>, i8)

define <4 x float> @test_mm512_maskz_mul_ps_128(<4 x float> %a0, <4 x float> %a1, i8 %mask) {
; X86-LABEL: test_mm512_maskz_mul_ps_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vmulps %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7c,0x89,0x59,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mm512_maskz_mul_ps_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vmulps %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7c,0x89,0x59,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask.mul.ps.128(<4 x float> %a0, <4 x float> %a1, <4 x float>zeroinitializer, i8 %mask)
  ret <4 x float> %res
}

define <4 x float> @test_mm512_mask_mul_ps_128(<4 x float> %a0, <4 x float> %a1, <4 x float> %src, i8 %mask) {
; X86-LABEL: test_mm512_mask_mul_ps_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vmulps %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7c,0x09,0x59,0xd1]
; X86-NEXT:    vmovaps %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x28,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mm512_mask_mul_ps_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vmulps %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7c,0x09,0x59,0xd1]
; X64-NEXT:    vmovaps %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x28,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask.mul.ps.128(<4 x float> %a0, <4 x float> %a1, <4 x float> %src, i8 %mask)
  ret <4 x float> %res
}

define <4 x float> @test_mm512_mul_ps_128(<4 x float> %a0, <4 x float> %a1, i8 %mask) {
; CHECK-LABEL: test_mm512_mul_ps_128:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vmulps %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x59,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask.mul.ps.128(<4 x float> %a0, <4 x float> %a1, <4 x float>zeroinitializer, i8 -1)
  ret <4 x float> %res
}
declare <4 x float> @llvm.x86.avx512.mask.mul.ps.128(<4 x float>, <4 x float>, <4 x float>, i8)

define <8 x float> @test_mm512_maskz_div_ps_256(<8 x float> %a0, <8 x float> %a1, i8 %mask) {
; X86-LABEL: test_mm512_maskz_div_ps_256:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vdivps %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7c,0xa9,0x5e,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mm512_maskz_div_ps_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vdivps %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7c,0xa9,0x5e,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask.div.ps.256(<8 x float> %a0, <8 x float> %a1, <8 x float>zeroinitializer, i8 %mask)
  ret <8 x float> %res
}

define <8 x float> @test_mm512_mask_div_ps_256(<8 x float> %a0, <8 x float> %a1, <8 x float> %src, i8 %mask) {
; X86-LABEL: test_mm512_mask_div_ps_256:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vdivps %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7c,0x29,0x5e,0xd1]
; X86-NEXT:    vmovaps %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x28,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mm512_mask_div_ps_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vdivps %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7c,0x29,0x5e,0xd1]
; X64-NEXT:    vmovaps %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x28,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask.div.ps.256(<8 x float> %a0, <8 x float> %a1, <8 x float> %src, i8 %mask)
  ret <8 x float> %res
}

define <8 x float> @test_mm512_div_ps_256(<8 x float> %a0, <8 x float> %a1, i8 %mask) {
; CHECK-LABEL: test_mm512_div_ps_256:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vdivps %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x5e,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask.div.ps.256(<8 x float> %a0, <8 x float> %a1, <8 x float>zeroinitializer, i8 -1)
  ret <8 x float> %res
}
declare <8 x float> @llvm.x86.avx512.mask.div.ps.256(<8 x float>, <8 x float>, <8 x float>, i8)

define <4 x float> @test_mm512_maskz_div_ps_128(<4 x float> %a0, <4 x float> %a1, i8 %mask) {
; X86-LABEL: test_mm512_maskz_div_ps_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vdivps %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7c,0x89,0x5e,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mm512_maskz_div_ps_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vdivps %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7c,0x89,0x5e,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask.div.ps.128(<4 x float> %a0, <4 x float> %a1, <4 x float>zeroinitializer, i8 %mask)
  ret <4 x float> %res
}

define <4 x float> @test_mm512_mask_div_ps_128(<4 x float> %a0, <4 x float> %a1, <4 x float> %src, i8 %mask) {
; X86-LABEL: test_mm512_mask_div_ps_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vdivps %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7c,0x09,0x5e,0xd1]
; X86-NEXT:    vmovaps %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x28,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mm512_mask_div_ps_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vdivps %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7c,0x09,0x5e,0xd1]
; X64-NEXT:    vmovaps %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x28,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask.div.ps.128(<4 x float> %a0, <4 x float> %a1, <4 x float> %src, i8 %mask)
  ret <4 x float> %res
}

define <4 x float> @test_mm512_div_ps_128(<4 x float> %a0, <4 x float> %a1, i8 %mask) {
; CHECK-LABEL: test_mm512_div_ps_128:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vdivps %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x5e,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask.div.ps.128(<4 x float> %a0, <4 x float> %a1, <4 x float>zeroinitializer, i8 -1)
  ret <4 x float> %res
}
declare <4 x float> @llvm.x86.avx512.mask.div.ps.128(<4 x float>, <4 x float>, <4 x float>, i8)

declare <8 x float> @llvm.x86.avx512.mask.shuf.f32x4.256(<8 x float>, <8 x float>, i32, <8 x float>, i8)

define <8 x float>@test_int_x86_avx512_mask_shuf_f32x4_256(<8 x float> %x0, <8 x float> %x1, <8 x float> %x3, i8 %x4) {
; X86-LABEL: test_int_x86_avx512_mask_shuf_f32x4_256:
; X86:       # %bb.0:
; X86-NEXT:    vblendps $240, %ymm1, %ymm0, %ymm0 # encoding: [0xc4,0xe3,0x7d,0x0c,0xc1,0xf0]
; X86-NEXT:    # ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vmovaps %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7c,0x29,0x28,0xd0]
; X86-NEXT:    vmovaps %ymm0, %ymm1 {%k1} {z} # encoding: [0x62,0xf1,0x7c,0xa9,0x28,0xc8]
; X86-NEXT:    vaddps %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xec,0x58,0xc0]
; X86-NEXT:    vaddps %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf4,0x58,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_shuf_f32x4_256:
; X64:       # %bb.0:
; X64-NEXT:    vblendps $240, %ymm1, %ymm0, %ymm0 # encoding: [0xc4,0xe3,0x7d,0x0c,0xc1,0xf0]
; X64-NEXT:    # ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vmovaps %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7c,0x29,0x28,0xd0]
; X64-NEXT:    vmovaps %ymm0, %ymm1 {%k1} {z} # encoding: [0x62,0xf1,0x7c,0xa9,0x28,0xc8]
; X64-NEXT:    vaddps %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xec,0x58,0xc0]
; X64-NEXT:    vaddps %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf4,0x58,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
    %res = call <8 x float> @llvm.x86.avx512.mask.shuf.f32x4.256(<8 x float> %x0, <8 x float> %x1, i32 22, <8 x float> %x3, i8 %x4)
    %res1 = call <8 x float> @llvm.x86.avx512.mask.shuf.f32x4.256(<8 x float> %x0, <8 x float> %x1, i32 22, <8 x float> %x3, i8 -1)
    %res2 = call <8 x float> @llvm.x86.avx512.mask.shuf.f32x4.256(<8 x float> %x0, <8 x float> %x1, i32 22, <8 x float> zeroinitializer, i8 %x4)
    %res3 = fadd <8 x float> %res, %res1
    %res4 = fadd <8 x float> %res2, %res3
    ret <8 x float> %res4
}

declare <4 x double> @llvm.x86.avx512.mask.shuf.f64x2.256(<4 x double>, <4 x double>, i32, <4 x double>, i8)

define <4 x double>@test_int_x86_avx512_mask_shuf_f64x2_256(<4 x double> %x0, <4 x double> %x1, <4 x double> %x3, i8 %x4) {
; X86-LABEL: test_int_x86_avx512_mask_shuf_f64x2_256:
; X86:       # %bb.0:
; X86-NEXT:    vblendpd $12, %ymm1, %ymm0, %ymm0 # encoding: [0xc4,0xe3,0x7d,0x0d,0xc1,0x0c]
; X86-NEXT:    # ymm0 = ymm0[0,1],ymm1[2,3]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vmovapd %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0xfd,0x29,0x28,0xd0]
; X86-NEXT:    vmovapd %ymm0, %ymm1 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0xa9,0x28,0xc8]
; X86-NEXT:    vaddpd %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0x58,0xc0]
; X86-NEXT:    vaddpd %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0x58,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_shuf_f64x2_256:
; X64:       # %bb.0:
; X64-NEXT:    vblendpd $12, %ymm1, %ymm0, %ymm0 # encoding: [0xc4,0xe3,0x7d,0x0d,0xc1,0x0c]
; X64-NEXT:    # ymm0 = ymm0[0,1],ymm1[2,3]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vmovapd %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0xfd,0x29,0x28,0xd0]
; X64-NEXT:    vmovapd %ymm0, %ymm1 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0xa9,0x28,0xc8]
; X64-NEXT:    vaddpd %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0x58,0xc0]
; X64-NEXT:    vaddpd %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0x58,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
    %res = call <4 x double> @llvm.x86.avx512.mask.shuf.f64x2.256(<4 x double> %x0, <4 x double> %x1, i32 22, <4 x double> %x3, i8 %x4)
    %res1 = call <4 x double> @llvm.x86.avx512.mask.shuf.f64x2.256(<4 x double> %x0, <4 x double> %x1, i32 22, <4 x double> %x3, i8 -1)
    %res2 = call <4 x double> @llvm.x86.avx512.mask.shuf.f64x2.256(<4 x double> %x0, <4 x double> %x1, i32 22, <4 x double> zeroinitializer, i8 %x4)
    %res3 = fadd <4 x double> %res, %res1
    %res4 = fadd <4 x double> %res2, %res3
    ret <4 x double> %res4
}

declare <8 x i32> @llvm.x86.avx512.mask.shuf.i32x4.256(<8 x i32>, <8 x i32>, i32, <8 x i32>, i8)

define <8 x i32>@test_int_x86_avx512_mask_shuf_i32x4_256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x3, i8 %x4) {
; X86-LABEL: test_int_x86_avx512_mask_shuf_i32x4_256:
; X86:       # %bb.0:
; X86-NEXT:    vpblendd $240, %ymm1, %ymm0, %ymm0 # encoding: [0xc4,0xe3,0x7d,0x02,0xc1,0xf0]
; X86-NEXT:    # ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vmovdqa32 %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x6f,0xd0]
; X86-NEXT:    vpaddd %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_shuf_i32x4_256:
; X64:       # %bb.0:
; X64-NEXT:    vpblendd $240, %ymm1, %ymm0, %ymm0 # encoding: [0xc4,0xe3,0x7d,0x02,0xc1,0xf0]
; X64-NEXT:    # ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vmovdqa32 %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x6f,0xd0]
; X64-NEXT:    vpaddd %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
    %res = call <8 x i32> @llvm.x86.avx512.mask.shuf.i32x4.256(<8 x i32> %x0, <8 x i32> %x1, i32 22, <8 x i32> %x3, i8 %x4)
    %res1 = call <8 x i32> @llvm.x86.avx512.mask.shuf.i32x4.256(<8 x i32> %x0, <8 x i32> %x1, i32 22, <8 x i32> %x3, i8 -1)
    %res2 = add <8 x i32> %res, %res1
    ret <8 x i32> %res2
}

declare <4 x i64> @llvm.x86.avx512.mask.shuf.i64x2.256(<4 x i64>, <4 x i64>, i32, <4 x i64>, i8)

define <4 x i64>@test_int_x86_avx512_mask_shuf_i64x2_256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x3, i8 %x4) {
; X86-LABEL: test_int_x86_avx512_mask_shuf_i64x2_256:
; X86:       # %bb.0:
; X86-NEXT:    vpblendd $240, %ymm1, %ymm0, %ymm0 # encoding: [0xc4,0xe3,0x7d,0x02,0xc1,0xf0]
; X86-NEXT:    # ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vmovdqa64 %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0xfd,0x29,0x6f,0xd0]
; X86-NEXT:    vpaddq %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xd4,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_shuf_i64x2_256:
; X64:       # %bb.0:
; X64-NEXT:    vpblendd $240, %ymm1, %ymm0, %ymm0 # encoding: [0xc4,0xe3,0x7d,0x02,0xc1,0xf0]
; X64-NEXT:    # ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vmovdqa64 %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0xfd,0x29,0x6f,0xd0]
; X64-NEXT:    vpaddq %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xd4,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
    %res = call <4 x i64> @llvm.x86.avx512.mask.shuf.i64x2.256(<4 x i64> %x0, <4 x i64> %x1, i32 22, <4 x i64> %x3, i8 %x4)
    %res1 = call <4 x i64> @llvm.x86.avx512.mask.shuf.i64x2.256(<4 x i64> %x0, <4 x i64> %x1, i32 22, <4 x i64> %x3, i8 -1)
    %res2 = add <4 x i64> %res, %res1
    ret <4 x i64> %res2
}

declare <2 x double> @llvm.x86.avx512.mask.shuf.pd.128(<2 x double>, <2 x double>, i32, <2 x double>, i8)

define <2 x double>@test_int_x86_avx512_mask_shuf_pd_128(<2 x double> %x0, <2 x double> %x1, <2 x double> %x3, i8 %x4) {
; X86-LABEL: test_int_x86_avx512_mask_shuf_pd_128:
; X86:       # %bb.0:
; X86-NEXT:    vshufpd $1, %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc6,0xd9,0x01]
; X86-NEXT:    # xmm3 = xmm0[1],xmm1[0]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vshufpd $1, %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0xfd,0x09,0xc6,0xd1,0x01]
; X86-NEXT:    # xmm2 {%k1} = xmm0[1],xmm1[0]
; X86-NEXT:    vaddpd %xmm3, %xmm2, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x58,0xd3]
; X86-NEXT:    vshufpd $1, %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0x89,0xc6,0xc1,0x01]
; X86-NEXT:    # xmm0 {%k1} {z} = xmm0[1],xmm1[0]
; X86-NEXT:    vaddpd %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x58,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_shuf_pd_128:
; X64:       # %bb.0:
; X64-NEXT:    vshufpd $1, %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc6,0xd9,0x01]
; X64-NEXT:    # xmm3 = xmm0[1],xmm1[0]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vshufpd $1, %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0xfd,0x09,0xc6,0xd1,0x01]
; X64-NEXT:    # xmm2 {%k1} = xmm0[1],xmm1[0]
; X64-NEXT:    vaddpd %xmm3, %xmm2, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x58,0xd3]
; X64-NEXT:    vshufpd $1, %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0x89,0xc6,0xc1,0x01]
; X64-NEXT:    # xmm0 {%k1} {z} = xmm0[1],xmm1[0]
; X64-NEXT:    vaddpd %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x58,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x double> @llvm.x86.avx512.mask.shuf.pd.128(<2 x double> %x0, <2 x double> %x1, i32 1, <2 x double> %x3, i8 %x4)
  %res1 = call <2 x double> @llvm.x86.avx512.mask.shuf.pd.128(<2 x double> %x0, <2 x double> %x1, i32 1, <2 x double> %x3, i8 -1)
  %res2 = call <2 x double> @llvm.x86.avx512.mask.shuf.pd.128(<2 x double> %x0, <2 x double> %x1, i32 1, <2 x double> zeroinitializer, i8 %x4)
  %res3 = fadd <2 x double> %res, %res1
  %res4 = fadd <2 x double> %res2, %res3
  ret <2 x double> %res4
}

declare <4 x double> @llvm.x86.avx512.mask.shuf.pd.256(<4 x double>, <4 x double>, i32, <4 x double>, i8)

define <4 x double>@test_int_x86_avx512_mask_shuf_pd_256(<4 x double> %x0, <4 x double> %x1, <4 x double> %x3, i8 %x4) {
; X86-LABEL: test_int_x86_avx512_mask_shuf_pd_256:
; X86:       # %bb.0:
; X86-NEXT:    vshufpd $6, %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xc6,0xd9,0x06]
; X86-NEXT:    # ymm3 = ymm0[0],ymm1[1],ymm0[3],ymm1[2]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vshufpd $6, %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0xfd,0x29,0xc6,0xd1,0x06]
; X86-NEXT:    # ymm2 {%k1} = ymm0[0],ymm1[1],ymm0[3],ymm1[2]
; X86-NEXT:    vaddpd %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0x58,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_shuf_pd_256:
; X64:       # %bb.0:
; X64-NEXT:    vshufpd $6, %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xc6,0xd9,0x06]
; X64-NEXT:    # ymm3 = ymm0[0],ymm1[1],ymm0[3],ymm1[2]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vshufpd $6, %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0xfd,0x29,0xc6,0xd1,0x06]
; X64-NEXT:    # ymm2 {%k1} = ymm0[0],ymm1[1],ymm0[3],ymm1[2]
; X64-NEXT:    vaddpd %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0x58,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x double> @llvm.x86.avx512.mask.shuf.pd.256(<4 x double> %x0, <4 x double> %x1, i32 6, <4 x double> %x3, i8 %x4)
  %res1 = call <4 x double> @llvm.x86.avx512.mask.shuf.pd.256(<4 x double> %x0, <4 x double> %x1, i32 6, <4 x double> %x3, i8 -1)
  %res2 = fadd <4 x double> %res, %res1
  ret <4 x double> %res2
}

declare <4 x float> @llvm.x86.avx512.mask.shuf.ps.128(<4 x float>, <4 x float>, i32, <4 x float>, i8)

define <4 x float>@test_int_x86_avx512_mask_shuf_ps_128(<4 x float> %x0, <4 x float> %x1, <4 x float> %x3, i8 %x4) {
; X86-LABEL: test_int_x86_avx512_mask_shuf_ps_128:
; X86:       # %bb.0:
; X86-NEXT:    vshufps $22, %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0xc6,0xd9,0x16]
; X86-NEXT:    # xmm3 = xmm0[2,1],xmm1[1,0]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vshufps $22, %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7c,0x09,0xc6,0xd1,0x16]
; X86-NEXT:    # xmm2 {%k1} = xmm0[2,1],xmm1[1,0]
; X86-NEXT:    vaddps %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe8,0x58,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_shuf_ps_128:
; X64:       # %bb.0:
; X64-NEXT:    vshufps $22, %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0xc6,0xd9,0x16]
; X64-NEXT:    # xmm3 = xmm0[2,1],xmm1[1,0]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vshufps $22, %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7c,0x09,0xc6,0xd1,0x16]
; X64-NEXT:    # xmm2 {%k1} = xmm0[2,1],xmm1[1,0]
; X64-NEXT:    vaddps %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe8,0x58,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask.shuf.ps.128(<4 x float> %x0, <4 x float> %x1, i32 22, <4 x float> %x3, i8 %x4)
  %res1 = call <4 x float> @llvm.x86.avx512.mask.shuf.ps.128(<4 x float> %x0, <4 x float> %x1, i32 22, <4 x float> %x3, i8 -1)
  %res2 = fadd <4 x float> %res, %res1
  ret <4 x float> %res2
}

declare <8 x float> @llvm.x86.avx512.mask.shuf.ps.256(<8 x float>, <8 x float>, i32, <8 x float>, i8)

define <8 x float>@test_int_x86_avx512_mask_shuf_ps_256(<8 x float> %x0, <8 x float> %x1, <8 x float> %x3, i8 %x4) {
; X86-LABEL: test_int_x86_avx512_mask_shuf_ps_256:
; X86:       # %bb.0:
; X86-NEXT:    vshufps $22, %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0xc6,0xd9,0x16]
; X86-NEXT:    # ymm3 = ymm0[2,1],ymm1[1,0],ymm0[6,5],ymm1[5,4]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vshufps $22, %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7c,0x29,0xc6,0xd1,0x16]
; X86-NEXT:    # ymm2 {%k1} = ymm0[2,1],ymm1[1,0],ymm0[6,5],ymm1[5,4]
; X86-NEXT:    vaddps %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xec,0x58,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_shuf_ps_256:
; X64:       # %bb.0:
; X64-NEXT:    vshufps $22, %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0xc6,0xd9,0x16]
; X64-NEXT:    # ymm3 = ymm0[2,1],ymm1[1,0],ymm0[6,5],ymm1[5,4]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vshufps $22, %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7c,0x29,0xc6,0xd1,0x16]
; X64-NEXT:    # ymm2 {%k1} = ymm0[2,1],ymm1[1,0],ymm0[6,5],ymm1[5,4]
; X64-NEXT:    vaddps %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xec,0x58,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask.shuf.ps.256(<8 x float> %x0, <8 x float> %x1, i32 22, <8 x float> %x3, i8 %x4)
  %res1 = call <8 x float> @llvm.x86.avx512.mask.shuf.ps.256(<8 x float> %x0, <8 x float> %x1, i32 22, <8 x float> %x3, i8 -1)
  %res2 = fadd <8 x float> %res, %res1
  ret <8 x float> %res2
}

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

define <4 x i32>@test_int_x86_avx512_mask_pmaxs_d_128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 %mask) {
; X86-LABEL: test_int_x86_avx512_mask_pmaxs_d_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpmaxsd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x3d,0xd1]
; X86-NEXT:    vpmaxsd %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x3d,0xc1]
; X86-NEXT:    vpaddd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmaxs_d_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmaxsd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x3d,0xd1]
; X64-NEXT:    vpmaxsd %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x3d,0xc1]
; X64-NEXT:    vpaddd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i32> @llvm.x86.avx512.mask.pmaxs.d.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2 ,i8 %mask)
  %res1 = call <4 x i32> @llvm.x86.avx512.mask.pmaxs.d.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> zeroinitializer, i8 %mask)
  %res2 = add <4 x i32> %res, %res1
  ret <4 x i32> %res2
}

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

define <8 x i32>@test_int_x86_avx512_mask_pmaxs_d_256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_pmaxs_d_256:
; X86:       # %bb.0:
; X86-NEXT:    vpmaxsd %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x3d,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpmaxsd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x3d,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_pmaxs_d_256:
; X64:       # %bb.0:
; X64-NEXT:    vpmaxsd %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x3d,0xd9]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmaxsd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x3d,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.pmaxs.d.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 %x3)
  %res1 = call <8 x i32> @llvm.x86.avx512.mask.pmaxs.d.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 -1)
  %res2 = add <8 x i32> %res, %res1
  ret <8 x i32> %res2
}

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

define <2 x i64>@test_int_x86_avx512_mask_pmaxs_q_128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_pmaxs_q_128:
; X86:       # %bb.0:
; X86-NEXT:    vpmaxsq %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf2,0xfd,0x08,0x3d,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpmaxsq %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x3d,0xd1]
; X86-NEXT:    vpaddq %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xd4,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmaxs_q_128:
; X64:       # %bb.0:
; X64-NEXT:    vpmaxsq %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf2,0xfd,0x08,0x3d,0xd9]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmaxsq %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x3d,0xd1]
; X64-NEXT:    vpaddq %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xd4,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x i64> @llvm.x86.avx512.mask.pmaxs.q.128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i8 %x3)
  %res1 = call <2 x i64> @llvm.x86.avx512.mask.pmaxs.q.128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i8 -1)
  %res2 = add <2 x i64> %res, %res1
  ret <2 x i64> %res2
}

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

define <4 x i64>@test_int_x86_avx512_mask_pmaxs_q_256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2, i8 %mask) {
; X86-LABEL: test_int_x86_avx512_mask_pmaxs_q_256:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpmaxsq %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x3d,0xd1]
; X86-NEXT:    vpmaxsq %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x3d,0xc1]
; X86-NEXT:    vpaddq %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xd4,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmaxs_q_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmaxsq %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x3d,0xd1]
; X64-NEXT:    vpmaxsq %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x3d,0xc1]
; X64-NEXT:    vpaddq %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xd4,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i64> @llvm.x86.avx512.mask.pmaxs.q.256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2, i8 %mask)
  %res1 = call <4 x i64> @llvm.x86.avx512.mask.pmaxs.q.256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> zeroinitializer, i8 %mask)
  %res2 = add <4 x i64> %res, %res1
  ret <4 x i64> %res2
}

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

define <4 x i32>@test_int_x86_avx512_mask_pmaxu_d_128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2,i8 %mask) {
; X86-LABEL: test_int_x86_avx512_mask_pmaxu_d_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpmaxud %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x3f,0xd1]
; X86-NEXT:    vpmaxud %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x3f,0xc1]
; X86-NEXT:    vpaddd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmaxu_d_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmaxud %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x3f,0xd1]
; X64-NEXT:    vpmaxud %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x3f,0xc1]
; X64-NEXT:    vpaddd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i32> @llvm.x86.avx512.mask.pmaxu.d.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 %mask)
  %res1 = call <4 x i32> @llvm.x86.avx512.mask.pmaxu.d.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> zeroinitializer, i8 %mask)
  %res2 = add <4 x i32> %res, %res1
  ret <4 x i32> %res2
}

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

define <8 x i32>@test_int_x86_avx512_mask_pmaxu_d_256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_pmaxu_d_256:
; X86:       # %bb.0:
; X86-NEXT:    vpmaxud %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x3f,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpmaxud %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x3f,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_pmaxu_d_256:
; X64:       # %bb.0:
; X64-NEXT:    vpmaxud %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x3f,0xd9]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmaxud %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x3f,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.pmaxu.d.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 %x3)
  %res1 = call <8 x i32> @llvm.x86.avx512.mask.pmaxu.d.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 -1)
  %res2 = add <8 x i32> %res, %res1
  ret <8 x i32> %res2
}

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

define <2 x i64>@test_int_x86_avx512_mask_pmaxu_q_128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_pmaxu_q_128:
; X86:       # %bb.0:
; X86-NEXT:    vpmaxuq %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf2,0xfd,0x08,0x3f,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpmaxuq %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x3f,0xd1]
; X86-NEXT:    vpaddq %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xd4,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmaxu_q_128:
; X64:       # %bb.0:
; X64-NEXT:    vpmaxuq %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf2,0xfd,0x08,0x3f,0xd9]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmaxuq %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x3f,0xd1]
; X64-NEXT:    vpaddq %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xd4,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x i64> @llvm.x86.avx512.mask.pmaxu.q.128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i8 %x3)
  %res1 = call <2 x i64> @llvm.x86.avx512.mask.pmaxu.q.128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i8 -1)
  %res2 = add <2 x i64> %res, %res1
  ret <2 x i64> %res2
}

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

define <4 x i64>@test_int_x86_avx512_mask_pmaxu_q_256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2, i8 %mask) {
; X86-LABEL: test_int_x86_avx512_mask_pmaxu_q_256:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpmaxuq %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x3f,0xd1]
; X86-NEXT:    vpmaxuq %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x3f,0xc1]
; X86-NEXT:    vpaddq %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xd4,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmaxu_q_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmaxuq %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x3f,0xd1]
; X64-NEXT:    vpmaxuq %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x3f,0xc1]
; X64-NEXT:    vpaddq %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xd4,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i64> @llvm.x86.avx512.mask.pmaxu.q.256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2, i8 %mask)
  %res1 = call <4 x i64> @llvm.x86.avx512.mask.pmaxu.q.256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> zeroinitializer, i8 %mask)
  %res2 = add <4 x i64> %res, %res1
  ret <4 x i64> %res2
}

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

define <4 x i32>@test_int_x86_avx512_mask_pmins_d_128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 %mask) {
; X86-LABEL: test_int_x86_avx512_mask_pmins_d_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpminsd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x39,0xd1]
; X86-NEXT:    vpminsd %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x39,0xc1]
; X86-NEXT:    vpaddd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmins_d_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpminsd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x39,0xd1]
; X64-NEXT:    vpminsd %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x39,0xc1]
; X64-NEXT:    vpaddd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i32> @llvm.x86.avx512.mask.pmins.d.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 %mask)
  %res1 = call <4 x i32> @llvm.x86.avx512.mask.pmins.d.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> zeroinitializer, i8 %mask)
  %res2 = add <4 x i32> %res, %res1
  ret <4 x i32> %res2
}

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

define <8 x i32>@test_int_x86_avx512_mask_pmins_d_256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_pmins_d_256:
; X86:       # %bb.0:
; X86-NEXT:    vpminsd %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x39,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpminsd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x39,0xd1]
; 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_pmins_d_256:
; X64:       # %bb.0:
; X64-NEXT:    vpminsd %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x39,0xd9]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpminsd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x39,0xd1]
; 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.pmins.d.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 %x3)
  %res1 = call <8 x i32> @llvm.x86.avx512.mask.pmins.d.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 -1)
  %res2 = add <8 x i32> %res, %res1
  ret <8 x i32> %res2
}

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

define <2 x i64>@test_int_x86_avx512_mask_pmins_q_128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_pmins_q_128:
; X86:       # %bb.0:
; X86-NEXT:    vpminsq %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf2,0xfd,0x08,0x39,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpminsq %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x39,0xd1]
; X86-NEXT:    vpaddq %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xd4,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmins_q_128:
; X64:       # %bb.0:
; X64-NEXT:    vpminsq %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf2,0xfd,0x08,0x39,0xd9]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpminsq %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x39,0xd1]
; X64-NEXT:    vpaddq %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xd4,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x i64> @llvm.x86.avx512.mask.pmins.q.128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i8 %x3)
  %res1 = call <2 x i64> @llvm.x86.avx512.mask.pmins.q.128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i8 -1)
  %res2 = add <2 x i64> %res, %res1
  ret <2 x i64> %res2
}

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

define <4 x i64>@test_int_x86_avx512_mask_pmins_q_256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2, i8 %mask) {
; X86-LABEL: test_int_x86_avx512_mask_pmins_q_256:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpminsq %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x39,0xd1]
; X86-NEXT:    vpminsq %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x39,0xc1]
; X86-NEXT:    vpaddq %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xd4,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmins_q_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpminsq %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x39,0xd1]
; X64-NEXT:    vpminsq %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x39,0xc1]
; X64-NEXT:    vpaddq %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xd4,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i64> @llvm.x86.avx512.mask.pmins.q.256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2, i8 %mask)
  %res1 = call <4 x i64> @llvm.x86.avx512.mask.pmins.q.256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> zeroinitializer, i8 %mask)
  %res2 = add <4 x i64> %res, %res1
  ret <4 x i64> %res2
}

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

define <4 x i32>@test_int_x86_avx512_mask_pminu_d_128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 %mask) {
; X86-LABEL: test_int_x86_avx512_mask_pminu_d_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpminud %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x3b,0xd1]
; X86-NEXT:    vpminud %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x3b,0xc1]
; X86-NEXT:    vpaddd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pminu_d_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpminud %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x3b,0xd1]
; X64-NEXT:    vpminud %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x3b,0xc1]
; X64-NEXT:    vpaddd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i32> @llvm.x86.avx512.mask.pminu.d.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 %mask)
  %res1 = call <4 x i32> @llvm.x86.avx512.mask.pminu.d.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> zeroinitializer, i8 %mask)
  %res2 = add <4 x i32> %res, %res1
  ret <4 x i32> %res2
}

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

define <8 x i32>@test_int_x86_avx512_mask_pminu_d_256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_pminu_d_256:
; X86:       # %bb.0:
; X86-NEXT:    vpminud %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x3b,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpminud %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x3b,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_pminu_d_256:
; X64:       # %bb.0:
; X64-NEXT:    vpminud %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x3b,0xd9]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpminud %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x3b,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.pminu.d.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 %x3)
  %res1 = call <8 x i32> @llvm.x86.avx512.mask.pminu.d.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 -1)
  %res2 = add <8 x i32> %res, %res1
  ret <8 x i32> %res2
}

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

define <2 x i64>@test_int_x86_avx512_mask_pminu_q_128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_pminu_q_128:
; X86:       # %bb.0:
; X86-NEXT:    vpminuq %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf2,0xfd,0x08,0x3b,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpminuq %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x3b,0xd1]
; X86-NEXT:    vpaddq %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xd4,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pminu_q_128:
; X64:       # %bb.0:
; X64-NEXT:    vpminuq %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf2,0xfd,0x08,0x3b,0xd9]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpminuq %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x3b,0xd1]
; X64-NEXT:    vpaddq %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xd4,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x i64> @llvm.x86.avx512.mask.pminu.q.128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i8 %x3)
  %res1 = call <2 x i64> @llvm.x86.avx512.mask.pminu.q.128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i8 -1)
  %res2 = add <2 x i64> %res, %res1
  ret <2 x i64> %res2
}

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

define <4 x i64>@test_int_x86_avx512_mask_pminu_q_256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2, i8 %mask) {
; X86-LABEL: test_int_x86_avx512_mask_pminu_q_256:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpminuq %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x3b,0xd1]
; X86-NEXT:    vpminuq %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x3b,0xc1]
; X86-NEXT:    vpaddq %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xd4,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pminu_q_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpminuq %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x3b,0xd1]
; X64-NEXT:    vpminuq %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x3b,0xc1]
; X64-NEXT:    vpaddq %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xd4,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i64> @llvm.x86.avx512.mask.pminu.q.256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2, i8 %mask)
  %res1 = call <4 x i64> @llvm.x86.avx512.mask.pminu.q.256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> zeroinitializer, i8 %mask)
  %res2 = add <4 x i64> %res, %res1
  ret <4 x i64> %res2
}

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

define <2 x i64>@test_int_x86_avx512_mask_psrl_q_128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_psrl_q_128:
; X86:       # %bb.0:
; X86-NEXT:    vpsrlq %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd3,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpsrlq %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0xfd,0x09,0xd3,0xd1]
; X86-NEXT:    vpsrlq %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0x89,0xd3,0xc1]
; X86-NEXT:    vpaddq %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xd4,0xc0]
; X86-NEXT:    vpaddq %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xd4,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_psrl_q_128:
; X64:       # %bb.0:
; X64-NEXT:    vpsrlq %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd3,0xd9]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpsrlq %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0xfd,0x09,0xd3,0xd1]
; X64-NEXT:    vpsrlq %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0x89,0xd3,0xc1]
; X64-NEXT:    vpaddq %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xd4,0xc0]
; X64-NEXT:    vpaddq %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xd4,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x i64> @llvm.x86.avx512.mask.psrl.q.128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i8 %x3)
  %res1 = call <2 x i64> @llvm.x86.avx512.mask.psrl.q.128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i8 -1)
  %res2 = call <2 x i64> @llvm.x86.avx512.mask.psrl.q.128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> zeroinitializer, i8 %x3)
  %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.psrl.q.256(<4 x i64>, <2 x i64>, <4 x i64>, i8)

define <4 x i64>@test_int_x86_avx512_mask_psrl_q_256(<4 x i64> %x0, <2 x i64> %x1, <4 x i64> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_psrl_q_256:
; X86:       # %bb.0:
; X86-NEXT:    vpsrlq %xmm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd3,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpsrlq %xmm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0xfd,0x29,0xd3,0xd1]
; X86-NEXT:    vpsrlq %xmm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0xa9,0xd3,0xc1]
; X86-NEXT:    vpaddq %ymm0, %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0xd4,0xc0]
; X86-NEXT:    vpaddq %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xd4,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_psrl_q_256:
; X64:       # %bb.0:
; X64-NEXT:    vpsrlq %xmm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd3,0xd9]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpsrlq %xmm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0xfd,0x29,0xd3,0xd1]
; X64-NEXT:    vpsrlq %xmm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0xa9,0xd3,0xc1]
; X64-NEXT:    vpaddq %ymm0, %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0xd4,0xc0]
; X64-NEXT:    vpaddq %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xd4,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i64> @llvm.x86.avx512.mask.psrl.q.256(<4 x i64> %x0, <2 x i64> %x1, <4 x i64> %x2, i8 %x3)
  %res1 = call <4 x i64> @llvm.x86.avx512.mask.psrl.q.256(<4 x i64> %x0, <2 x i64> %x1, <4 x i64> %x2, i8 -1)
  %res2 = call <4 x i64> @llvm.x86.avx512.mask.psrl.q.256(<4 x i64> %x0, <2 x i64> %x1, <4 x i64> zeroinitializer, i8 %x3)
  %res3 = add <4 x i64> %res, %res1
  %res4 = add <4 x i64> %res3, %res2
  ret <4 x i64> %res4
}

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

define <4 x i32>@test_int_x86_avx512_mask_psrl_d_128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_psrl_d_128:
; X86:       # %bb.0:
; X86-NEXT:    vpsrld %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd2,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpsrld %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xd2,0xd1]
; X86-NEXT:    vpsrld %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xd2,0xc1]
; X86-NEXT:    vpaddd %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xfe,0xc0]
; X86-NEXT:    vpaddd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_psrl_d_128:
; X64:       # %bb.0:
; X64-NEXT:    vpsrld %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd2,0xd9]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpsrld %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xd2,0xd1]
; X64-NEXT:    vpsrld %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xd2,0xc1]
; X64-NEXT:    vpaddd %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xfe,0xc0]
; X64-NEXT:    vpaddd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i32> @llvm.x86.avx512.mask.psrl.d.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 %x3)
  %res1 = call <4 x i32> @llvm.x86.avx512.mask.psrl.d.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 -1)
  %res2 = call <4 x i32> @llvm.x86.avx512.mask.psrl.d.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> zeroinitializer, i8 %x3)
  %res3 = add <4 x i32> %res, %res1
  %res4 = add <4 x i32> %res3, %res2
  ret <4 x i32> %res4
}

declare <8 x i32> @llvm.x86.avx512.mask.psrl.d.256(<8 x i32>, <4 x i32>, <8 x i32>, i8)

define <8 x i32>@test_int_x86_avx512_mask_psrl_d_256(<8 x i32> %x0, <4 x i32> %x1, <8 x i32> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_psrl_d_256:
; X86:       # %bb.0:
; X86-NEXT:    vpsrld %xmm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd2,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpsrld %xmm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xd2,0xd1]
; X86-NEXT:    vpsrld %xmm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xd2,0xc1]
; X86-NEXT:    vpaddd %ymm0, %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0xfe,0xc0]
; X86-NEXT:    vpaddd %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_psrl_d_256:
; X64:       # %bb.0:
; X64-NEXT:    vpsrld %xmm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd2,0xd9]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpsrld %xmm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xd2,0xd1]
; X64-NEXT:    vpsrld %xmm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xd2,0xc1]
; X64-NEXT:    vpaddd %ymm0, %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0xfe,0xc0]
; X64-NEXT:    vpaddd %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x i32> @llvm.x86.avx512.mask.psrl.d.256(<8 x i32> %x0, <4 x i32> %x1, <8 x i32> %x2, i8 %x3)
  %res1 = call <8 x i32> @llvm.x86.avx512.mask.psrl.d.256(<8 x i32> %x0, <4 x i32> %x1, <8 x i32> %x2, i8 -1)
  %res2 = call <8 x i32> @llvm.x86.avx512.mask.psrl.d.256(<8 x i32> %x0, <4 x i32> %x1, <8 x i32> zeroinitializer, i8 %x3)
  %res3 = add <8 x i32> %res, %res1
  %res4 = add <8 x i32> %res2, %res3
  ret <8 x i32> %res4
}

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

define <4 x i32>@test_int_x86_avx512_mask_psra_d_128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_psra_d_128:
; X86:       # %bb.0:
; X86-NEXT:    vpsrad %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xe2,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpsrad %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe2,0xd1]
; X86-NEXT:    vpsrad %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xe2,0xc1]
; X86-NEXT:    vpaddd %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfe,0xc3]
; X86-NEXT:    vpaddd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_psra_d_128:
; X64:       # %bb.0:
; X64-NEXT:    vpsrad %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xe2,0xd9]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpsrad %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe2,0xd1]
; X64-NEXT:    vpsrad %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xe2,0xc1]
; X64-NEXT:    vpaddd %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfe,0xc3]
; X64-NEXT:    vpaddd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i32> @llvm.x86.avx512.mask.psra.d.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 %x3)
  %res1 = call <4 x i32> @llvm.x86.avx512.mask.psra.d.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> zeroinitializer, i8 %x3)
  %res2 = call <4 x i32> @llvm.x86.avx512.mask.psra.d.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 -1)
  %res3 = add <4 x i32> %res, %res1
  %res4 = add <4 x i32> %res3, %res2
  ret <4 x i32> %res4
}

declare <8 x i32> @llvm.x86.avx512.mask.psra.d.256(<8 x i32>, <4 x i32>, <8 x i32>, i8)

define <8 x i32>@test_int_x86_avx512_mask_psra_d_256(<8 x i32> %x0, <4 x i32> %x1, <8 x i32> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_psra_d_256:
; X86:       # %bb.0:
; X86-NEXT:    vpsrad %xmm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xe2,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpsrad %xmm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe2,0xd1]
; X86-NEXT:    vpsrad %xmm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xe2,0xc1]
; X86-NEXT:    vpaddd %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfe,0xc3]
; X86-NEXT:    vpaddd %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_psra_d_256:
; X64:       # %bb.0:
; X64-NEXT:    vpsrad %xmm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xe2,0xd9]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpsrad %xmm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe2,0xd1]
; X64-NEXT:    vpsrad %xmm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xe2,0xc1]
; X64-NEXT:    vpaddd %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfe,0xc3]
; X64-NEXT:    vpaddd %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x i32> @llvm.x86.avx512.mask.psra.d.256(<8 x i32> %x0, <4 x i32> %x1, <8 x i32> %x2, i8 %x3)
  %res1 = call <8 x i32> @llvm.x86.avx512.mask.psra.d.256(<8 x i32> %x0, <4 x i32> %x1, <8 x i32> zeroinitializer, i8 %x3)
  %res2 = call <8 x i32> @llvm.x86.avx512.mask.psra.d.256(<8 x i32> %x0, <4 x i32> %x1, <8 x i32> %x2, i8 -1)
  %res3 = add <8 x i32> %res, %res1
  %res4 = add <8 x i32> %res3, %res2
  ret <8 x i32> %res4
}

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

define <4 x i32>@test_int_x86_avx512_mask_psll_d_128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_psll_d_128:
; X86:       # %bb.0:
; X86-NEXT:    vpslld %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xf2,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpslld %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xf2,0xd1]
; X86-NEXT:    vpslld %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xf2,0xc1]
; X86-NEXT:    vpaddd %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfe,0xc3]
; X86-NEXT:    vpaddd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_psll_d_128:
; X64:       # %bb.0:
; X64-NEXT:    vpslld %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xf2,0xd9]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpslld %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xf2,0xd1]
; X64-NEXT:    vpslld %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xf2,0xc1]
; X64-NEXT:    vpaddd %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfe,0xc3]
; X64-NEXT:    vpaddd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i32> @llvm.x86.avx512.mask.psll.d.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 %x3)
  %res1 = call <4 x i32> @llvm.x86.avx512.mask.psll.d.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> zeroinitializer, i8 %x3)
  %res2 = call <4 x i32> @llvm.x86.avx512.mask.psll.d.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 -1)
  %res3 = add <4 x i32> %res, %res1
  %res4 = add <4 x i32> %res3, %res2
  ret <4 x i32> %res4
}

declare <8 x i32> @llvm.x86.avx512.mask.psll.d.256(<8 x i32>, <4 x i32>, <8 x i32>, i8)

define <8 x i32>@test_int_x86_avx512_mask_psll_d_256(<8 x i32> %x0, <4 x i32> %x1, <8 x i32> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_psll_d_256:
; X86:       # %bb.0:
; X86-NEXT:    vpslld %xmm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xf2,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpslld %xmm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xf2,0xd1]
; X86-NEXT:    vpslld %xmm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xf2,0xc1]
; X86-NEXT:    vpaddd %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfe,0xc3]
; X86-NEXT:    vpaddd %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_psll_d_256:
; X64:       # %bb.0:
; X64-NEXT:    vpslld %xmm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xf2,0xd9]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpslld %xmm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xf2,0xd1]
; X64-NEXT:    vpslld %xmm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xf2,0xc1]
; X64-NEXT:    vpaddd %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfe,0xc3]
; X64-NEXT:    vpaddd %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x i32> @llvm.x86.avx512.mask.psll.d.256(<8 x i32> %x0, <4 x i32> %x1, <8 x i32> %x2, i8 %x3)
  %res1 = call <8 x i32> @llvm.x86.avx512.mask.psll.d.256(<8 x i32> %x0, <4 x i32> %x1, <8 x i32> zeroinitializer, i8 %x3)
  %res2 = call <8 x i32> @llvm.x86.avx512.mask.psll.d.256(<8 x i32> %x0, <4 x i32> %x1, <8 x i32> %x2, i8 -1)
  %res3 = add <8 x i32> %res, %res1
  %res4 = add <8 x i32> %res3, %res2
  ret <8 x i32> %res4
}

declare <4 x i64> @llvm.x86.avx512.mask.psll.q.256(<4 x i64>, <2 x i64>, <4 x i64>, i8)

define <4 x i64>@test_int_x86_avx512_mask_psll_q_256(<4 x i64> %x0, <2 x i64> %x1, <4 x i64> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_psll_q_256:
; X86:       # %bb.0:
; X86-NEXT:    vpsllq %xmm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xf3,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpsllq %xmm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0xfd,0x29,0xf3,0xd1]
; X86-NEXT:    vpsllq %xmm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0xa9,0xf3,0xc1]
; X86-NEXT:    vpaddq %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd4,0xc3]
; X86-NEXT:    vpaddq %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xd4,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_psll_q_256:
; X64:       # %bb.0:
; X64-NEXT:    vpsllq %xmm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xf3,0xd9]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpsllq %xmm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0xfd,0x29,0xf3,0xd1]
; X64-NEXT:    vpsllq %xmm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0xa9,0xf3,0xc1]
; X64-NEXT:    vpaddq %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd4,0xc3]
; X64-NEXT:    vpaddq %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xd4,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i64> @llvm.x86.avx512.mask.psll.q.256(<4 x i64> %x0, <2 x i64> %x1, <4 x i64> %x2, i8 %x3)
  %res1 = call <4 x i64> @llvm.x86.avx512.mask.psll.q.256(<4 x i64> %x0, <2 x i64> %x1, <4 x i64> zeroinitializer, i8 %x3)
  %res2 = call <4 x i64> @llvm.x86.avx512.mask.psll.q.256(<4 x i64> %x0, <2 x i64> %x1, <4 x i64> %x2, i8 -1)
  %res3 = add <4 x i64> %res, %res1
  %res4 = add <4 x i64> %res3, %res2
  ret <4 x i64> %res4
}

declare <2 x i64> @llvm.x86.avx512.mask.psrl.qi.128(<2 x i64>, i32, <2 x i64>, i8)

define <2 x i64>@test_int_x86_avx512_mask_psrl_qi_128(<2 x i64> %x0, i32 %x1, <2 x i64> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_psrl_qi_128:
; X86:       # %bb.0:
; X86-NEXT:    vpsrlq $3, %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x73,0xd0,0x03]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpsrlq $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x09,0x73,0xd0,0x03]
; X86-NEXT:    vpsrlq $3, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0x89,0x73,0xd0,0x03]
; X86-NEXT:    vpaddq %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xd4,0xc0]
; 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_psrl_qi_128:
; X64:       # %bb.0:
; X64-NEXT:    vpsrlq $3, %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x73,0xd0,0x03]
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpsrlq $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x09,0x73,0xd0,0x03]
; X64-NEXT:    vpsrlq $3, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0x89,0x73,0xd0,0x03]
; X64-NEXT:    vpaddq %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xd4,0xc0]
; 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.psrl.qi.128(<2 x i64> %x0, i32 3, <2 x i64> %x2, i8 %x3)
  %res1 = call <2 x i64> @llvm.x86.avx512.mask.psrl.qi.128(<2 x i64> %x0, i32 3, <2 x i64> %x2, i8 -1)
  %res2 = call <2 x i64> @llvm.x86.avx512.mask.psrl.qi.128(<2 x i64> %x0, i32 3, <2 x i64> zeroinitializer, i8 %x3)
  %res3 = add <2 x i64> %res, %res1
  %res4 = add <2 x i64> %res2, %res3
  ret <2 x i64> %res4
}

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

define <4 x i64>@test_int_x86_avx512_mask_psrl_qi_256(<4 x i64> %x0, i32 %x1, <4 x i64> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_psrl_qi_256:
; X86:       # %bb.0:
; X86-NEXT:    vpsrlq $3, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xed,0x73,0xd0,0x03]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpsrlq $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x29,0x73,0xd0,0x03]
; X86-NEXT:    vpsrlq $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0xa9,0x73,0xd0,0x03]
; X86-NEXT:    vpaddq %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xd4,0xc0]
; 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_psrl_qi_256:
; X64:       # %bb.0:
; X64-NEXT:    vpsrlq $3, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xed,0x73,0xd0,0x03]
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpsrlq $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x29,0x73,0xd0,0x03]
; X64-NEXT:    vpsrlq $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0xa9,0x73,0xd0,0x03]
; X64-NEXT:    vpaddq %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xd4,0xc0]
; 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.psrl.qi.256(<4 x i64> %x0, i32 3, <4 x i64> %x2, i8 %x3)
  %res1 = call <4 x i64> @llvm.x86.avx512.mask.psrl.qi.256(<4 x i64> %x0, i32 3, <4 x i64> %x2, i8 -1)
  %res2 = call <4 x i64> @llvm.x86.avx512.mask.psrl.qi.256(<4 x i64> %x0, i32 3, <4 x i64> zeroinitializer, i8 %x3)
  %res3 = add <4 x i64> %res, %res1
  %res4 = add <4 x i64> %res2, %res3
  ret <4 x i64> %res4
}

declare <4 x i32> @llvm.x86.avx512.mask.psrl.di.128(<4 x i32>, i32, <4 x i32>, i8)

define <4 x i32>@test_int_x86_avx512_mask_psrl_di_128(<4 x i32> %x0, i32 %x1, <4 x i32> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_psrl_di_128:
; X86:       # %bb.0:
; X86-NEXT:    vpsrld $3, %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x72,0xd0,0x03]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpsrld $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x72,0xd0,0x03]
; X86-NEXT:    vpsrld $3, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x72,0xd0,0x03]
; X86-NEXT:    vpaddd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc0]
; X86-NEXT:    vpaddd %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_psrl_di_128:
; X64:       # %bb.0:
; X64-NEXT:    vpsrld $3, %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x72,0xd0,0x03]
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpsrld $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x72,0xd0,0x03]
; X64-NEXT:    vpsrld $3, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x72,0xd0,0x03]
; X64-NEXT:    vpaddd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc0]
; X64-NEXT:    vpaddd %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i32> @llvm.x86.avx512.mask.psrl.di.128(<4 x i32> %x0, i32 3, <4 x i32> %x2, i8 %x3)
  %res1 = call <4 x i32> @llvm.x86.avx512.mask.psrl.di.128(<4 x i32> %x0, i32 3, <4 x i32> %x2, i8 -1)
  %res2 = call <4 x i32> @llvm.x86.avx512.mask.psrl.di.128(<4 x i32> %x0, i32 3, <4 x i32> zeroinitializer, i8 %x3)
  %res3 = add <4 x i32> %res, %res1
  %res4 = add <4 x i32> %res2, %res3
  ret <4 x i32> %res4
}

declare <8 x i32> @llvm.x86.avx512.mask.psrl.di.256(<8 x i32>, i32, <8 x i32>, i8)

define <8 x i32>@test_int_x86_avx512_mask_psrl_di_256(<8 x i32> %x0, i32 %x1, <8 x i32> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_psrl_di_256:
; X86:       # %bb.0:
; X86-NEXT:    vpsrld $3, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xed,0x72,0xd0,0x03]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpsrld $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x72,0xd0,0x03]
; X86-NEXT:    vpsrld $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x72,0xd0,0x03]
; X86-NEXT:    vpaddd %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfe,0xc0]
; X86-NEXT:    vpaddd %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_psrl_di_256:
; X64:       # %bb.0:
; X64-NEXT:    vpsrld $3, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xed,0x72,0xd0,0x03]
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpsrld $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x72,0xd0,0x03]
; X64-NEXT:    vpsrld $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x72,0xd0,0x03]
; X64-NEXT:    vpaddd %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfe,0xc0]
; X64-NEXT:    vpaddd %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x i32> @llvm.x86.avx512.mask.psrl.di.256(<8 x i32> %x0, i32 3, <8 x i32> %x2, i8 %x3)
  %res1 = call <8 x i32> @llvm.x86.avx512.mask.psrl.di.256(<8 x i32> %x0, i32 3, <8 x i32> %x2, i8 -1)
  %res2 = call <8 x i32> @llvm.x86.avx512.mask.psrl.di.256(<8 x i32> %x0, i32 3, <8 x i32> zeroinitializer, i8 %x3)
  %res3 = add <8 x i32> %res, %res1
  %res4 = add <8 x i32> %res2, %res3
  ret <8 x i32> %res4
}

declare <4 x i32> @llvm.x86.avx512.mask.psll.di.128(<4 x i32>, i32, <4 x i32>, i8)

define <4 x i32>@test_int_x86_avx512_mask_psll_di_128(<4 x i32> %x0, i32 %x1, <4 x i32> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_psll_di_128:
; X86:       # %bb.0:
; X86-NEXT:    vpslld $3, %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x72,0xf0,0x03]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpslld $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x72,0xf0,0x03]
; X86-NEXT:    vpslld $3, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x72,0xf0,0x03]
; X86-NEXT:    vpaddd %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfe,0xc2]
; X86-NEXT:    vpaddd %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_psll_di_128:
; X64:       # %bb.0:
; X64-NEXT:    vpslld $3, %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x72,0xf0,0x03]
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpslld $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x72,0xf0,0x03]
; X64-NEXT:    vpslld $3, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x72,0xf0,0x03]
; X64-NEXT:    vpaddd %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfe,0xc2]
; X64-NEXT:    vpaddd %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i32> @llvm.x86.avx512.mask.psll.di.128(<4 x i32> %x0, i32 3, <4 x i32> %x2, i8 %x3)
  %res1 = call <4 x i32> @llvm.x86.avx512.mask.psll.di.128(<4 x i32> %x0, i32 3, <4 x i32> zeroinitializer, i8 %x3)
  %res2 = call <4 x i32> @llvm.x86.avx512.mask.psll.di.128(<4 x i32> %x0, i32 3, <4 x i32> %x2, i8 -1)
  %res3 = add <4 x i32> %res, %res1
  %res4 = add <4 x i32> %res3, %res2
  ret <4 x i32> %res4
}

declare <8 x i32> @llvm.x86.avx512.mask.psll.di.256(<8 x i32>, i32, <8 x i32>, i8)

define <8 x i32>@test_int_x86_avx512_mask_psll_di_256(<8 x i32> %x0, i32 %x1, <8 x i32> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_psll_di_256:
; X86:       # %bb.0:
; X86-NEXT:    vpslld $3, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xed,0x72,0xf0,0x03]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpslld $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x72,0xf0,0x03]
; X86-NEXT:    vpslld $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x72,0xf0,0x03]
; X86-NEXT:    vpaddd %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfe,0xc2]
; X86-NEXT:    vpaddd %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_psll_di_256:
; X64:       # %bb.0:
; X64-NEXT:    vpslld $3, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xed,0x72,0xf0,0x03]
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpslld $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x72,0xf0,0x03]
; X64-NEXT:    vpslld $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x72,0xf0,0x03]
; X64-NEXT:    vpaddd %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfe,0xc2]
; X64-NEXT:    vpaddd %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x i32> @llvm.x86.avx512.mask.psll.di.256(<8 x i32> %x0, i32 3, <8 x i32> %x2, i8 %x3)
  %res1 = call <8 x i32> @llvm.x86.avx512.mask.psll.di.256(<8 x i32> %x0, i32 3, <8 x i32> zeroinitializer, i8 %x3)
  %res2 = call <8 x i32> @llvm.x86.avx512.mask.psll.di.256(<8 x i32> %x0, i32 3, <8 x i32> %x2, i8 -1)
  %res3 = add <8 x i32> %res, %res1
  %res4 = add <8 x i32> %res3, %res2
  ret <8 x i32> %res4
}

declare <2 x i64> @llvm.x86.avx512.mask.psrlv2.di(<2 x i64>, <2 x i64>, <2 x i64>, i8)

define <2 x i64>@test_int_x86_avx512_mask_psrlv2_di(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_psrlv2_di:
; X86:       # %bb.0:
; X86-NEXT:    vpsrlvq %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0xf9,0x45,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpsrlvq %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x45,0xd1]
; X86-NEXT:    vpsrlvq %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x45,0xc1]
; X86-NEXT:    vpaddq %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd4,0xc3]
; X86-NEXT:    vpaddq %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xd4,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_psrlv2_di:
; X64:       # %bb.0:
; X64-NEXT:    vpsrlvq %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0xf9,0x45,0xd9]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpsrlvq %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x45,0xd1]
; X64-NEXT:    vpsrlvq %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x45,0xc1]
; X64-NEXT:    vpaddq %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd4,0xc3]
; X64-NEXT:    vpaddq %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xd4,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x i64> @llvm.x86.avx512.mask.psrlv2.di(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i8 %x3)
  %res1 = call <2 x i64> @llvm.x86.avx512.mask.psrlv2.di(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> zeroinitializer, i8 %x3)
  %res2 = call <2 x i64> @llvm.x86.avx512.mask.psrlv2.di(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, 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.psrlv4.di(<4 x i64>, <4 x i64>, <4 x i64>, i8)

define <4 x i64>@test_int_x86_avx512_mask_psrlv4_di(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_psrlv4_di:
; X86:       # %bb.0:
; X86-NEXT:    vpsrlvq %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0xfd,0x45,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpsrlvq %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x45,0xd1]
; X86-NEXT:    vpsrlvq %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x45,0xc1]
; X86-NEXT:    vpaddq %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd4,0xc3]
; X86-NEXT:    vpaddq %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xd4,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_psrlv4_di:
; X64:       # %bb.0:
; X64-NEXT:    vpsrlvq %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0xfd,0x45,0xd9]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpsrlvq %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x45,0xd1]
; X64-NEXT:    vpsrlvq %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x45,0xc1]
; X64-NEXT:    vpaddq %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd4,0xc3]
; X64-NEXT:    vpaddq %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xd4,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i64> @llvm.x86.avx512.mask.psrlv4.di(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2, i8 %x3)
  %res1 = call <4 x i64> @llvm.x86.avx512.mask.psrlv4.di(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> zeroinitializer, i8 %x3)
  %res2 = call <4 x i64> @llvm.x86.avx512.mask.psrlv4.di(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2, i8 -1)
  %res3 = add <4 x i64> %res, %res1
  %res4 = add <4 x i64> %res3, %res2
  ret <4 x i64> %res4
}

declare <4 x i32> @llvm.x86.avx512.mask.psrlv4.si(<4 x i32>, <4 x i32>, <4 x i32>, i8)

define <4 x i32>@test_int_x86_avx512_mask_psrlv4_si(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_psrlv4_si:
; X86:       # %bb.0:
; X86-NEXT:    vpsrlvd %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x45,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpsrlvd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x45,0xd1]
; X86-NEXT:    vpsrlvd %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x45,0xc1]
; X86-NEXT:    vpaddd %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfe,0xc3]
; X86-NEXT:    vpaddd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_psrlv4_si:
; X64:       # %bb.0:
; X64-NEXT:    vpsrlvd %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x45,0xd9]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpsrlvd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x45,0xd1]
; X64-NEXT:    vpsrlvd %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x45,0xc1]
; X64-NEXT:    vpaddd %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfe,0xc3]
; X64-NEXT:    vpaddd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i32> @llvm.x86.avx512.mask.psrlv4.si(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 %x3)
  %res1 = call <4 x i32> @llvm.x86.avx512.mask.psrlv4.si(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> zeroinitializer, i8 %x3)
  %res2 = call <4 x i32> @llvm.x86.avx512.mask.psrlv4.si(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 -1)
  %res3 = add <4 x i32> %res, %res1
  %res4 = add <4 x i32> %res3, %res2
  ret <4 x i32> %res4
}

declare <8 x i32> @llvm.x86.avx512.mask.psrlv8.si(<8 x i32>, <8 x i32>, <8 x i32>, i8)

define <8 x i32>@test_int_x86_avx512_mask_psrlv8_si(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_psrlv8_si:
; X86:       # %bb.0:
; X86-NEXT:    vpsrlvd %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x45,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpsrlvd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x45,0xd1]
; X86-NEXT:    vpsrlvd %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x45,0xc1]
; X86-NEXT:    vpaddd %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfe,0xc3]
; X86-NEXT:    vpaddd %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_psrlv8_si:
; X64:       # %bb.0:
; X64-NEXT:    vpsrlvd %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x45,0xd9]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpsrlvd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x45,0xd1]
; X64-NEXT:    vpsrlvd %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x45,0xc1]
; X64-NEXT:    vpaddd %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfe,0xc3]
; X64-NEXT:    vpaddd %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x i32> @llvm.x86.avx512.mask.psrlv8.si(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 %x3)
  %res1 = call <8 x i32> @llvm.x86.avx512.mask.psrlv8.si(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> zeroinitializer, i8 %x3)
  %res2 = call <8 x i32> @llvm.x86.avx512.mask.psrlv8.si(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 -1)
  %res3 = add <8 x i32> %res, %res1
  %res4 = add <8 x i32> %res3, %res2
  ret <8 x i32> %res4
}

declare <4 x i32> @llvm.x86.avx512.mask.psrav4.si(<4 x i32>, <4 x i32>, <4 x i32>, i8)

define <4 x i32>@test_int_x86_avx512_mask_psrav4_si(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_psrav4_si:
; X86:       # %bb.0:
; X86-NEXT:    vpsravd %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x46,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpsravd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x46,0xd1]
; X86-NEXT:    vpsravd %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x46,0xc1]
; X86-NEXT:    vpaddd %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfe,0xc3]
; X86-NEXT:    vpaddd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_psrav4_si:
; X64:       # %bb.0:
; X64-NEXT:    vpsravd %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x46,0xd9]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpsravd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x46,0xd1]
; X64-NEXT:    vpsravd %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x46,0xc1]
; X64-NEXT:    vpaddd %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfe,0xc3]
; X64-NEXT:    vpaddd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i32> @llvm.x86.avx512.mask.psrav4.si(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 %x3)
  %res1 = call <4 x i32> @llvm.x86.avx512.mask.psrav4.si(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> zeroinitializer, i8 %x3)
  %res2 = call <4 x i32> @llvm.x86.avx512.mask.psrav4.si(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 -1)
  %res3 = add <4 x i32> %res, %res1
  %res4 = add <4 x i32> %res3, %res2
  ret <4 x i32> %res4
}

declare <8 x i32> @llvm.x86.avx512.mask.psrav8.si(<8 x i32>, <8 x i32>, <8 x i32>, i8)

define <8 x i32>@test_int_x86_avx512_mask_psrav8_si(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_psrav8_si:
; X86:       # %bb.0:
; X86-NEXT:    vpsravd %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x46,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpsravd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x46,0xd1]
; X86-NEXT:    vpsravd %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x46,0xc1]
; X86-NEXT:    vpaddd %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfe,0xc3]
; X86-NEXT:    vpaddd %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_psrav8_si:
; X64:       # %bb.0:
; X64-NEXT:    vpsravd %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x46,0xd9]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpsravd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x46,0xd1]
; X64-NEXT:    vpsravd %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x46,0xc1]
; X64-NEXT:    vpaddd %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfe,0xc3]
; X64-NEXT:    vpaddd %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x i32> @llvm.x86.avx512.mask.psrav8.si(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 %x3)
  %res1 = call <8 x i32> @llvm.x86.avx512.mask.psrav8.si(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> zeroinitializer, i8 %x3)
  %res2 = call <8 x i32> @llvm.x86.avx512.mask.psrav8.si(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 -1)
  %res3 = add <8 x i32> %res, %res1
  %res4 = add <8 x i32> %res3, %res2
  ret <8 x i32> %res4
}

define <8 x i32>@test_int_x86_avx512_mask_psrav8_si_const() {
; X86-LABEL: test_int_x86_avx512_mask_psrav8_si_const:
; X86:       # %bb.0:
; X86-NEXT:    vmovdqa {{\.LCPI.*}}, %ymm0 # EVEX TO VEX Compression ymm0 = [2,9,4294967284,23,4294967270,37,4294967256,51]
; X86-NEXT:    # encoding: [0xc5,0xfd,0x6f,0x05,A,A,A,A]
; X86-NEXT:    # fixup A - offset: 4, value: {{\.LCPI.*}}, kind: FK_Data_4
; X86-NEXT:    vpsravd {{\.LCPI.*}}, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x46,0x05,A,A,A,A]
; X86-NEXT:    # fixup A - offset: 5, value: {{\.LCPI.*}}, kind: FK_Data_4
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_psrav8_si_const:
; X64:       # %bb.0:
; X64-NEXT:    vmovdqa {{.*}}(%rip), %ymm0 # EVEX TO VEX Compression ymm0 = [2,9,4294967284,23,4294967270,37,4294967256,51]
; X64-NEXT:    # encoding: [0xc5,0xfd,0x6f,0x05,A,A,A,A]
; X64-NEXT:    # fixup A - offset: 4, value: {{\.LCPI.*}}-4, kind: reloc_riprel_4byte
; X64-NEXT:    vpsravd {{.*}}(%rip), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x46,0x05,A,A,A,A]
; X64-NEXT:    # fixup A - offset: 5, value: {{\.LCPI.*}}-4, kind: reloc_riprel_4byte
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x i32> @llvm.x86.avx512.mask.psrav8.si(<8 x i32> <i32 2, i32 9, i32 -12, i32 23, i32 -26, i32 37, i32 -40, i32 51>, <8 x i32> <i32 1, i32 18, i32 35, i32 52, i32 69, i32 15, i32 32, i32 49>, <8 x i32> zeroinitializer, i8 -1)
  ret <8 x i32> %res
}

declare <2 x i64> @llvm.x86.avx512.mask.psllv2.di(<2 x i64>, <2 x i64>, <2 x i64>, i8)

define <2 x i64>@test_int_x86_avx512_mask_psllv2_di(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_psllv2_di:
; X86:       # %bb.0:
; X86-NEXT:    vpsllvq %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0xf9,0x47,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpsllvq %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x47,0xd1]
; X86-NEXT:    vpsllvq %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x47,0xc1]
; X86-NEXT:    vpaddq %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd4,0xc3]
; X86-NEXT:    vpaddq %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xd4,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_psllv2_di:
; X64:       # %bb.0:
; X64-NEXT:    vpsllvq %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0xf9,0x47,0xd9]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpsllvq %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x47,0xd1]
; X64-NEXT:    vpsllvq %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x47,0xc1]
; X64-NEXT:    vpaddq %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd4,0xc3]
; X64-NEXT:    vpaddq %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xd4,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x i64> @llvm.x86.avx512.mask.psllv2.di(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i8 %x3)
  %res1 = call <2 x i64> @llvm.x86.avx512.mask.psllv2.di(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> zeroinitializer, i8 %x3)
  %res2 = call <2 x i64> @llvm.x86.avx512.mask.psllv2.di(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, 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.psllv4.di(<4 x i64>, <4 x i64>, <4 x i64>, i8)

define <4 x i64>@test_int_x86_avx512_mask_psllv4_di(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_psllv4_di:
; X86:       # %bb.0:
; X86-NEXT:    vpsllvq %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0xfd,0x47,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpsllvq %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x47,0xd1]
; X86-NEXT:    vpsllvq %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x47,0xc1]
; X86-NEXT:    vpaddq %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd4,0xc3]
; X86-NEXT:    vpaddq %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xd4,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_psllv4_di:
; X64:       # %bb.0:
; X64-NEXT:    vpsllvq %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0xfd,0x47,0xd9]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpsllvq %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x47,0xd1]
; X64-NEXT:    vpsllvq %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x47,0xc1]
; X64-NEXT:    vpaddq %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd4,0xc3]
; X64-NEXT:    vpaddq %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xd4,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i64> @llvm.x86.avx512.mask.psllv4.di(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2, i8 %x3)
  %res1 = call <4 x i64> @llvm.x86.avx512.mask.psllv4.di(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> zeroinitializer, i8 %x3)
  %res2 = call <4 x i64> @llvm.x86.avx512.mask.psllv4.di(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2, i8 -1)
  %res3 = add <4 x i64> %res, %res1
  %res4 = add <4 x i64> %res3, %res2
  ret <4 x i64> %res4
}

declare <4 x i32> @llvm.x86.avx512.mask.psllv4.si(<4 x i32>, <4 x i32>, <4 x i32>, i8)

define <4 x i32>@test_int_x86_avx512_mask_psllv4_si(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_psllv4_si:
; X86:       # %bb.0:
; X86-NEXT:    vpsllvd %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x47,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpsllvd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x47,0xd1]
; X86-NEXT:    vpsllvd %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x47,0xc1]
; X86-NEXT:    vpaddd %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfe,0xc3]
; X86-NEXT:    vpaddd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_psllv4_si:
; X64:       # %bb.0:
; X64-NEXT:    vpsllvd %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x47,0xd9]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpsllvd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x47,0xd1]
; X64-NEXT:    vpsllvd %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x47,0xc1]
; X64-NEXT:    vpaddd %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfe,0xc3]
; X64-NEXT:    vpaddd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i32> @llvm.x86.avx512.mask.psllv4.si(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 %x3)
  %res1 = call <4 x i32> @llvm.x86.avx512.mask.psllv4.si(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> zeroinitializer, i8 %x3)
  %res2 = call <4 x i32> @llvm.x86.avx512.mask.psllv4.si(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 -1)
  %res3 = add <4 x i32> %res, %res1
  %res4 = add <4 x i32> %res3, %res2
  ret <4 x i32> %res4
}

declare <8 x i32> @llvm.x86.avx512.mask.psllv8.si(<8 x i32>, <8 x i32>, <8 x i32>, i8)

define <8 x i32>@test_int_x86_avx512_mask_psllv8_si(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_psllv8_si:
; X86:       # %bb.0:
; X86-NEXT:    vpsllvd %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x47,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpsllvd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x47,0xd1]
; X86-NEXT:    vpsllvd %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x47,0xc1]
; X86-NEXT:    vpaddd %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfe,0xc3]
; X86-NEXT:    vpaddd %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_psllv8_si:
; X64:       # %bb.0:
; X64-NEXT:    vpsllvd %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x47,0xd9]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpsllvd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x47,0xd1]
; X64-NEXT:    vpsllvd %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x47,0xc1]
; X64-NEXT:    vpaddd %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfe,0xc3]
; X64-NEXT:    vpaddd %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x i32> @llvm.x86.avx512.mask.psllv8.si(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 %x3)
  %res1 = call <8 x i32> @llvm.x86.avx512.mask.psllv8.si(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> zeroinitializer, i8 %x3)
  %res2 = call <8 x i32> @llvm.x86.avx512.mask.psllv8.si(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 -1)
  %res3 = add <8 x i32> %res, %res1
  %res4 = add <8 x i32> %res3, %res2
  ret <8 x i32> %res4
}

declare <4 x i32> @llvm.x86.avx512.mask.pmovzxb.d.128(<16 x i8>, <4 x i32>, i8)

define <4 x i32>@test_int_x86_avx512_mask_pmovzxb_d_128(<16 x i8> %x0, <4 x i32> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmovzxb_d_128:
; X86:       # %bb.0:
; X86-NEXT:    vpmovzxbd %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x31,0xd0]
; X86-NEXT:    # xmm2 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpmovzxbd %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x31,0xc8]
; X86-NEXT:    # xmm1 {%k1} = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero
; X86-NEXT:    vpmovzxbd %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x31,0xc0]
; X86-NEXT:    # xmm0 {%k1} {z} = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero
; X86-NEXT:    vpaddd %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfe,0xc2]
; X86-NEXT:    vpaddd %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmovzxb_d_128:
; X64:       # %bb.0:
; X64-NEXT:    vpmovzxbd %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x31,0xd0]
; X64-NEXT:    # xmm2 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmovzxbd %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x31,0xc8]
; X64-NEXT:    # xmm1 {%k1} = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero
; X64-NEXT:    vpmovzxbd %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x31,0xc0]
; X64-NEXT:    # xmm0 {%k1} {z} = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero
; X64-NEXT:    vpaddd %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfe,0xc2]
; X64-NEXT:    vpaddd %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i32> @llvm.x86.avx512.mask.pmovzxb.d.128(<16 x i8> %x0, <4 x i32> %x1, i8 %x2)
  %res1 = call <4 x i32> @llvm.x86.avx512.mask.pmovzxb.d.128(<16 x i8> %x0, <4 x i32> zeroinitializer, i8 %x2)
  %res2 = call <4 x i32> @llvm.x86.avx512.mask.pmovzxb.d.128(<16 x i8> %x0, <4 x i32> %x1, i8 -1)
  %res3 = add <4 x i32> %res, %res1
  %res4 = add <4 x i32> %res3, %res2
  ret <4 x i32> %res4
}

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

define <8 x i32>@test_int_x86_avx512_mask_pmovzxb_d_256(<16 x i8> %x0, <8 x i32> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmovzxb_d_256:
; X86:       # %bb.0:
; X86-NEXT:    vpmovzxbd %xmm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x31,0xd0]
; X86-NEXT:    # ymm2 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpmovzxbd %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x31,0xc8]
; X86-NEXT:    # ymm1 {%k1} = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero
; X86-NEXT:    vpmovzxbd %xmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x31,0xc0]
; X86-NEXT:    # ymm0 {%k1} {z} = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero
; X86-NEXT:    vpaddd %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfe,0xc2]
; X86-NEXT:    vpaddd %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmovzxb_d_256:
; X64:       # %bb.0:
; X64-NEXT:    vpmovzxbd %xmm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x31,0xd0]
; X64-NEXT:    # ymm2 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmovzxbd %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x31,0xc8]
; X64-NEXT:    # ymm1 {%k1} = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero
; X64-NEXT:    vpmovzxbd %xmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x31,0xc0]
; X64-NEXT:    # ymm0 {%k1} {z} = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero
; X64-NEXT:    vpaddd %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfe,0xc2]
; X64-NEXT:    vpaddd %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x i32> @llvm.x86.avx512.mask.pmovzxb.d.256(<16 x i8> %x0, <8 x i32> %x1, i8 %x2)
  %res1 = call <8 x i32> @llvm.x86.avx512.mask.pmovzxb.d.256(<16 x i8> %x0, <8 x i32> zeroinitializer, i8 %x2)
  %res2 = call <8 x i32> @llvm.x86.avx512.mask.pmovzxb.d.256(<16 x i8> %x0, <8 x i32> %x1, i8 -1)
  %res3 = add <8 x i32> %res, %res1
  %res4 = add <8 x i32> %res3, %res2
  ret <8 x i32> %res4
}

declare <2 x i64> @llvm.x86.avx512.mask.pmovzxb.q.128(<16 x i8>, <2 x i64>, i8)

define <2 x i64>@test_int_x86_avx512_mask_pmovzxb_q_128(<16 x i8> %x0, <2 x i64> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmovzxb_q_128:
; X86:       # %bb.0:
; X86-NEXT:    vpmovzxbq %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x32,0xd0]
; X86-NEXT:    # xmm2 = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpmovzxbq %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x32,0xc8]
; X86-NEXT:    # xmm1 {%k1} = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero
; X86-NEXT:    vpmovzxbq %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x32,0xc0]
; X86-NEXT:    # xmm0 {%k1} {z} = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero
; 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_pmovzxb_q_128:
; X64:       # %bb.0:
; X64-NEXT:    vpmovzxbq %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x32,0xd0]
; X64-NEXT:    # xmm2 = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmovzxbq %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x32,0xc8]
; X64-NEXT:    # xmm1 {%k1} = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero
; X64-NEXT:    vpmovzxbq %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x32,0xc0]
; X64-NEXT:    # xmm0 {%k1} {z} = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero
; 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.pmovzxb.q.128(<16 x i8> %x0, <2 x i64> %x1, i8 %x2)
  %res1 = call <2 x i64> @llvm.x86.avx512.mask.pmovzxb.q.128(<16 x i8> %x0, <2 x i64> zeroinitializer, i8 %x2)
  %res2 = call <2 x i64> @llvm.x86.avx512.mask.pmovzxb.q.128(<16 x i8> %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.pmovzxb.q.256(<16 x i8>, <4 x i64>, i8)

define <4 x i64>@test_int_x86_avx512_mask_pmovzxb_q_256(<16 x i8> %x0, <4 x i64> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmovzxb_q_256:
; X86:       # %bb.0:
; X86-NEXT:    vpmovzxbq %xmm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x32,0xd0]
; X86-NEXT:    # ymm2 = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero,xmm0[2],zero,zero,zero,zero,zero,zero,zero,xmm0[3],zero,zero,zero,zero,zero,zero,zero
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpmovzxbq %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x32,0xc8]
; X86-NEXT:    # ymm1 {%k1} = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero,xmm0[2],zero,zero,zero,zero,zero,zero,zero,xmm0[3],zero,zero,zero,zero,zero,zero,zero
; X86-NEXT:    vpmovzxbq %xmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x32,0xc0]
; X86-NEXT:    # ymm0 {%k1} {z} = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero,xmm0[2],zero,zero,zero,zero,zero,zero,zero,xmm0[3],zero,zero,zero,zero,zero,zero,zero
; 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_pmovzxb_q_256:
; X64:       # %bb.0:
; X64-NEXT:    vpmovzxbq %xmm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x32,0xd0]
; X64-NEXT:    # ymm2 = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero,xmm0[2],zero,zero,zero,zero,zero,zero,zero,xmm0[3],zero,zero,zero,zero,zero,zero,zero
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmovzxbq %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x32,0xc8]
; X64-NEXT:    # ymm1 {%k1} = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero,xmm0[2],zero,zero,zero,zero,zero,zero,zero,xmm0[3],zero,zero,zero,zero,zero,zero,zero
; X64-NEXT:    vpmovzxbq %xmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x32,0xc0]
; X64-NEXT:    # ymm0 {%k1} {z} = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero,xmm0[2],zero,zero,zero,zero,zero,zero,zero,xmm0[3],zero,zero,zero,zero,zero,zero,zero
; 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.pmovzxb.q.256(<16 x i8> %x0, <4 x i64> %x1, i8 %x2)
  %res1 = call <4 x i64> @llvm.x86.avx512.mask.pmovzxb.q.256(<16 x i8> %x0, <4 x i64> zeroinitializer, i8 %x2)
  %res2 = call <4 x i64> @llvm.x86.avx512.mask.pmovzxb.q.256(<16 x i8> %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 <2 x i64> @llvm.x86.avx512.mask.pmovzxd.q.128(<4 x i32>, <2 x i64>, i8)

define <2 x i64>@test_int_x86_avx512_mask_pmovzxd_q_128(<4 x i32> %x0, <2 x i64> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmovzxd_q_128:
; X86:       # %bb.0:
; X86-NEXT:    vpmovzxdq %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x35,0xd0]
; X86-NEXT:    # xmm2 = xmm0[0],zero,xmm0[1],zero
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpmovzxdq %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x35,0xc8]
; X86-NEXT:    # xmm1 {%k1} = xmm0[0],zero,xmm0[1],zero
; X86-NEXT:    vpmovzxdq %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x35,0xc0]
; X86-NEXT:    # xmm0 {%k1} {z} = xmm0[0],zero,xmm0[1],zero
; 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_pmovzxd_q_128:
; X64:       # %bb.0:
; X64-NEXT:    vpmovzxdq %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x35,0xd0]
; X64-NEXT:    # xmm2 = xmm0[0],zero,xmm0[1],zero
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmovzxdq %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x35,0xc8]
; X64-NEXT:    # xmm1 {%k1} = xmm0[0],zero,xmm0[1],zero
; X64-NEXT:    vpmovzxdq %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x35,0xc0]
; X64-NEXT:    # xmm0 {%k1} {z} = xmm0[0],zero,xmm0[1],zero
; 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.pmovzxd.q.128(<4 x i32> %x0, <2 x i64> %x1, i8 %x2)
  %res1 = call <2 x i64> @llvm.x86.avx512.mask.pmovzxd.q.128(<4 x i32> %x0, <2 x i64> zeroinitializer, i8 %x2)
  %res2 = call <2 x i64> @llvm.x86.avx512.mask.pmovzxd.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.pmovzxd.q.256(<4 x i32>, <4 x i64>, i8)

define <4 x i64>@test_int_x86_avx512_mask_pmovzxd_q_256(<4 x i32> %x0, <4 x i64> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmovzxd_q_256:
; X86:       # %bb.0:
; X86-NEXT:    vpmovzxdq %xmm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x35,0xd0]
; X86-NEXT:    # ymm2 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpmovzxdq %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x35,0xc8]
; X86-NEXT:    # ymm1 {%k1} = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero
; X86-NEXT:    vpmovzxdq %xmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x35,0xc0]
; X86-NEXT:    # ymm0 {%k1} {z} = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero
; 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_pmovzxd_q_256:
; X64:       # %bb.0:
; X64-NEXT:    vpmovzxdq %xmm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x35,0xd0]
; X64-NEXT:    # ymm2 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmovzxdq %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x35,0xc8]
; X64-NEXT:    # ymm1 {%k1} = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero
; X64-NEXT:    vpmovzxdq %xmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x35,0xc0]
; X64-NEXT:    # ymm0 {%k1} {z} = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero
; 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.pmovzxd.q.256(<4 x i32> %x0, <4 x i64> %x1, i8 %x2)
  %res1 = call <4 x i64> @llvm.x86.avx512.mask.pmovzxd.q.256(<4 x i32> %x0, <4 x i64> zeroinitializer, i8 %x2)
  %res2 = call <4 x i64> @llvm.x86.avx512.mask.pmovzxd.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 <4 x i32> @llvm.x86.avx512.mask.pmovzxw.d.128(<8 x i16>, <4 x i32>, i8)

define <4 x i32>@test_int_x86_avx512_mask_pmovzxw_d_128(<8 x i16> %x0, <4 x i32> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmovzxw_d_128:
; X86:       # %bb.0:
; X86-NEXT:    vpmovzxwd %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x33,0xd0]
; X86-NEXT:    # xmm2 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpmovzxwd %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x33,0xc8]
; X86-NEXT:    # xmm1 {%k1} = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero
; X86-NEXT:    vpmovzxwd %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x33,0xc0]
; X86-NEXT:    # xmm0 {%k1} {z} = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero
; X86-NEXT:    vpaddd %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfe,0xc2]
; X86-NEXT:    vpaddd %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmovzxw_d_128:
; X64:       # %bb.0:
; X64-NEXT:    vpmovzxwd %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x33,0xd0]
; X64-NEXT:    # xmm2 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmovzxwd %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x33,0xc8]
; X64-NEXT:    # xmm1 {%k1} = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero
; X64-NEXT:    vpmovzxwd %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x33,0xc0]
; X64-NEXT:    # xmm0 {%k1} {z} = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero
; X64-NEXT:    vpaddd %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfe,0xc2]
; X64-NEXT:    vpaddd %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i32> @llvm.x86.avx512.mask.pmovzxw.d.128(<8 x i16> %x0, <4 x i32> %x1, i8 %x2)
  %res1 = call <4 x i32> @llvm.x86.avx512.mask.pmovzxw.d.128(<8 x i16> %x0, <4 x i32> zeroinitializer, i8 %x2)
  %res2 = call <4 x i32> @llvm.x86.avx512.mask.pmovzxw.d.128(<8 x i16> %x0, <4 x i32> %x1, i8 -1)
  %res3 = add <4 x i32> %res, %res1
  %res4 = add <4 x i32> %res3, %res2
  ret <4 x i32> %res4
}

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

define <8 x i32>@test_int_x86_avx512_mask_pmovzxw_d_256(<8 x i16> %x0, <8 x i32> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmovzxw_d_256:
; X86:       # %bb.0:
; X86-NEXT:    vpmovzxwd %xmm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x33,0xd0]
; 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
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpmovzxwd %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x33,0xc8]
; 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
; X86-NEXT:    vpmovzxwd %xmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x33,0xc0]
; 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
; X86-NEXT:    vpaddd %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfe,0xc2]
; X86-NEXT:    vpaddd %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmovzxw_d_256:
; X64:       # %bb.0:
; X64-NEXT:    vpmovzxwd %xmm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x33,0xd0]
; 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
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmovzxwd %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x33,0xc8]
; 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
; X64-NEXT:    vpmovzxwd %xmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x33,0xc0]
; 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
; X64-NEXT:    vpaddd %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfe,0xc2]
; X64-NEXT:    vpaddd %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x i32> @llvm.x86.avx512.mask.pmovzxw.d.256(<8 x i16> %x0, <8 x i32> %x1, i8 %x2)
  %res1 = call <8 x i32> @llvm.x86.avx512.mask.pmovzxw.d.256(<8 x i16> %x0, <8 x i32> zeroinitializer, i8 %x2)
  %res2 = call <8 x i32> @llvm.x86.avx512.mask.pmovzxw.d.256(<8 x i16> %x0, <8 x i32> %x1, i8 -1)
  %res3 = add <8 x i32> %res, %res1
  %res4 = add <8 x i32> %res3, %res2
  ret <8 x i32> %res4
}

declare <2 x i64> @llvm.x86.avx512.mask.pmovzxw.q.128(<8 x i16>, <2 x i64>, i8)

define <2 x i64>@test_int_x86_avx512_mask_pmovzxw_q_128(<8 x i16> %x0, <2 x i64> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmovzxw_q_128:
; X86:       # %bb.0:
; X86-NEXT:    vpmovzxwq %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x34,0xd0]
; X86-NEXT:    # xmm2 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpmovzxwq %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x34,0xc8]
; X86-NEXT:    # xmm1 {%k1} = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero
; X86-NEXT:    vpmovzxwq %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x34,0xc0]
; X86-NEXT:    # xmm0 {%k1} {z} = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero
; 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_pmovzxw_q_128:
; X64:       # %bb.0:
; X64-NEXT:    vpmovzxwq %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x34,0xd0]
; X64-NEXT:    # xmm2 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmovzxwq %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x34,0xc8]
; X64-NEXT:    # xmm1 {%k1} = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero
; X64-NEXT:    vpmovzxwq %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x34,0xc0]
; X64-NEXT:    # xmm0 {%k1} {z} = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero
; 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.pmovzxw.q.128(<8 x i16> %x0, <2 x i64> %x1, i8 %x2)
  %res1 = call <2 x i64> @llvm.x86.avx512.mask.pmovzxw.q.128(<8 x i16> %x0, <2 x i64> zeroinitializer, i8 %x2)
  %res2 = call <2 x i64> @llvm.x86.avx512.mask.pmovzxw.q.128(<8 x i16> %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.pmovzxw.q.256(<8 x i16>, <4 x i64>, i8)

define <4 x i64>@test_int_x86_avx512_mask_pmovzxw_q_256(<8 x i16> %x0, <4 x i64> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmovzxw_q_256:
; X86:       # %bb.0:
; X86-NEXT:    vpmovzxwq %xmm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x34,0xd0]
; X86-NEXT:    # ymm2 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpmovzxwq %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x34,0xc8]
; X86-NEXT:    # ymm1 {%k1} = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero
; X86-NEXT:    vpmovzxwq %xmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x34,0xc0]
; X86-NEXT:    # ymm0 {%k1} {z} = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero
; 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_pmovzxw_q_256:
; X64:       # %bb.0:
; X64-NEXT:    vpmovzxwq %xmm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x34,0xd0]
; X64-NEXT:    # ymm2 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmovzxwq %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x34,0xc8]
; X64-NEXT:    # ymm1 {%k1} = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero
; X64-NEXT:    vpmovzxwq %xmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x34,0xc0]
; X64-NEXT:    # ymm0 {%k1} {z} = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero
; 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.pmovzxw.q.256(<8 x i16> %x0, <4 x i64> %x1, i8 %x2)
  %res1 = call <4 x i64> @llvm.x86.avx512.mask.pmovzxw.q.256(<8 x i16> %x0, <4 x i64> zeroinitializer, i8 %x2)
  %res2 = call <4 x i64> @llvm.x86.avx512.mask.pmovzxw.q.256(<8 x i16> %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 <4 x i32> @llvm.x86.avx512.mask.pmovsxb.d.128(<16 x i8>, <4 x i32>, i8)

define <4 x i32>@test_int_x86_avx512_mask_pmovsxb_d_128(<16 x i8> %x0, <4 x i32> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmovsxb_d_128:
; X86:       # %bb.0:
; X86-NEXT:    vpmovsxbd %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x21,0xd0]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpmovsxbd %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x21,0xc8]
; X86-NEXT:    vpmovsxbd %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x21,0xc0]
; X86-NEXT:    vpaddd %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfe,0xc2]
; X86-NEXT:    vpaddd %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmovsxb_d_128:
; X64:       # %bb.0:
; X64-NEXT:    vpmovsxbd %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x21,0xd0]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmovsxbd %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x21,0xc8]
; X64-NEXT:    vpmovsxbd %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x21,0xc0]
; X64-NEXT:    vpaddd %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfe,0xc2]
; X64-NEXT:    vpaddd %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i32> @llvm.x86.avx512.mask.pmovsxb.d.128(<16 x i8> %x0, <4 x i32> %x1, i8 %x2)
  %res1 = call <4 x i32> @llvm.x86.avx512.mask.pmovsxb.d.128(<16 x i8> %x0, <4 x i32> zeroinitializer, i8 %x2)
  %res2 = call <4 x i32> @llvm.x86.avx512.mask.pmovsxb.d.128(<16 x i8> %x0, <4 x i32> %x1, i8 -1)
  %res3 = add <4 x i32> %res, %res1
  %res4 = add <4 x i32> %res3, %res2
  ret <4 x i32> %res4
}

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

define <8 x i32>@test_int_x86_avx512_mask_pmovsxb_d_256(<16 x i8> %x0, <8 x i32> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmovsxb_d_256:
; X86:       # %bb.0:
; X86-NEXT:    vpmovsxbd %xmm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x21,0xd0]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpmovsxbd %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x21,0xc8]
; X86-NEXT:    vpmovsxbd %xmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x21,0xc0]
; X86-NEXT:    vpaddd %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfe,0xc2]
; X86-NEXT:    vpaddd %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmovsxb_d_256:
; X64:       # %bb.0:
; X64-NEXT:    vpmovsxbd %xmm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x21,0xd0]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmovsxbd %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x21,0xc8]
; X64-NEXT:    vpmovsxbd %xmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x21,0xc0]
; X64-NEXT:    vpaddd %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfe,0xc2]
; X64-NEXT:    vpaddd %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x i32> @llvm.x86.avx512.mask.pmovsxb.d.256(<16 x i8> %x0, <8 x i32> %x1, i8 %x2)
  %res1 = call <8 x i32> @llvm.x86.avx512.mask.pmovsxb.d.256(<16 x i8> %x0, <8 x i32> zeroinitializer, i8 %x2)
  %res2 = call <8 x i32> @llvm.x86.avx512.mask.pmovsxb.d.256(<16 x i8> %x0, <8 x i32> %x1, i8 -1)
  %res3 = add <8 x i32> %res, %res1
  %res4 = add <8 x i32> %res3, %res2
  ret <8 x i32> %res4
}

declare <2 x i64> @llvm.x86.avx512.mask.pmovsxb.q.128(<16 x i8>, <2 x i64>, i8)

define <2 x i64>@test_int_x86_avx512_mask_pmovsxb_q_128(<16 x i8> %x0, <2 x i64> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmovsxb_q_128:
; X86:       # %bb.0:
; X86-NEXT:    vpmovsxbq %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x22,0xd0]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpmovsxbq %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x22,0xc8]
; X86-NEXT:    vpmovsxbq %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x22,0xc0]
; 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_pmovsxb_q_128:
; X64:       # %bb.0:
; X64-NEXT:    vpmovsxbq %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x22,0xd0]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmovsxbq %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x22,0xc8]
; X64-NEXT:    vpmovsxbq %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x22,0xc0]
; 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.pmovsxb.q.128(<16 x i8> %x0, <2 x i64> %x1, i8 %x2)
  %res1 = call <2 x i64> @llvm.x86.avx512.mask.pmovsxb.q.128(<16 x i8> %x0, <2 x i64> zeroinitializer, i8 %x2)
  %res2 = call <2 x i64> @llvm.x86.avx512.mask.pmovsxb.q.128(<16 x i8> %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.pmovsxb.q.256(<16 x i8>, <4 x i64>, i8)

define <4 x i64>@test_int_x86_avx512_mask_pmovsxb_q_256(<16 x i8> %x0, <4 x i64> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmovsxb_q_256:
; X86:       # %bb.0:
; X86-NEXT:    vpmovsxbq %xmm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x22,0xd0]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpmovsxbq %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x22,0xc8]
; X86-NEXT:    vpmovsxbq %xmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x22,0xc0]
; 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_pmovsxb_q_256:
; X64:       # %bb.0:
; X64-NEXT:    vpmovsxbq %xmm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x22,0xd0]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmovsxbq %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x22,0xc8]
; X64-NEXT:    vpmovsxbq %xmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x22,0xc0]
; 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.pmovsxb.q.256(<16 x i8> %x0, <4 x i64> %x1, i8 %x2)
  %res1 = call <4 x i64> @llvm.x86.avx512.mask.pmovsxb.q.256(<16 x i8> %x0, <4 x i64> zeroinitializer, i8 %x2)
  %res2 = call <4 x i64> @llvm.x86.avx512.mask.pmovsxb.q.256(<16 x i8> %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 <4 x i32> @llvm.x86.avx512.mask.pmovsxw.d.128(<8 x i16>, <4 x i32>, i8)

define <4 x i32>@test_int_x86_avx512_mask_pmovsxw_d_128(<8 x i16> %x0, <4 x i32> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmovsxw_d_128:
; X86:       # %bb.0:
; X86-NEXT:    vpmovsxwd %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x23,0xd0]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpmovsxwd %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x23,0xc8]
; X86-NEXT:    vpmovsxwd %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x23,0xc0]
; X86-NEXT:    vpaddd %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfe,0xc2]
; X86-NEXT:    vpaddd %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmovsxw_d_128:
; X64:       # %bb.0:
; X64-NEXT:    vpmovsxwd %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x23,0xd0]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmovsxwd %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x23,0xc8]
; X64-NEXT:    vpmovsxwd %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x23,0xc0]
; X64-NEXT:    vpaddd %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfe,0xc2]
; X64-NEXT:    vpaddd %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i32> @llvm.x86.avx512.mask.pmovsxw.d.128(<8 x i16> %x0, <4 x i32> %x1, i8 %x2)
  %res1 = call <4 x i32> @llvm.x86.avx512.mask.pmovsxw.d.128(<8 x i16> %x0, <4 x i32> zeroinitializer, i8 %x2)
  %res2 = call <4 x i32> @llvm.x86.avx512.mask.pmovsxw.d.128(<8 x i16> %x0, <4 x i32> %x1, i8 -1)
  %res3 = add <4 x i32> %res, %res1
  %res4 = add <4 x i32> %res3, %res2
  ret <4 x i32> %res4
}

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

define <8 x i32>@test_int_x86_avx512_mask_pmovsxw_d_256(<8 x i16> %x0, <8 x i32> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmovsxw_d_256:
; X86:       # %bb.0:
; X86-NEXT:    vpmovsxwd %xmm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x23,0xd0]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpmovsxwd %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x23,0xc8]
; X86-NEXT:    vpmovsxwd %xmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x23,0xc0]
; X86-NEXT:    vpaddd %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfe,0xc2]
; X86-NEXT:    vpaddd %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmovsxw_d_256:
; X64:       # %bb.0:
; X64-NEXT:    vpmovsxwd %xmm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x23,0xd0]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmovsxwd %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x23,0xc8]
; X64-NEXT:    vpmovsxwd %xmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x23,0xc0]
; X64-NEXT:    vpaddd %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfe,0xc2]
; X64-NEXT:    vpaddd %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x i32> @llvm.x86.avx512.mask.pmovsxw.d.256(<8 x i16> %x0, <8 x i32> %x1, i8 %x2)
  %res1 = call <8 x i32> @llvm.x86.avx512.mask.pmovsxw.d.256(<8 x i16> %x0, <8 x i32> zeroinitializer, i8 %x2)
  %res2 = call <8 x i32> @llvm.x86.avx512.mask.pmovsxw.d.256(<8 x i16> %x0, <8 x i32> %x1, i8 -1)
  %res3 = add <8 x i32> %res, %res1
  %res4 = add <8 x i32> %res3, %res2
  ret <8 x i32> %res4
}

declare <2 x i64> @llvm.x86.avx512.mask.pmovsxw.q.128(<8 x i16>, <2 x i64>, i8)

define <2 x i64>@test_int_x86_avx512_mask_pmovsxw_q_128(<8 x i16> %x0, <2 x i64> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmovsxw_q_128:
; X86:       # %bb.0:
; X86-NEXT:    vpmovsxwq %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x24,0xd0]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpmovsxwq %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x24,0xc8]
; X86-NEXT:    vpmovsxwq %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x24,0xc0]
; 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_pmovsxw_q_128:
; X64:       # %bb.0:
; X64-NEXT:    vpmovsxwq %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x24,0xd0]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmovsxwq %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x24,0xc8]
; X64-NEXT:    vpmovsxwq %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x24,0xc0]
; 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.pmovsxw.q.128(<8 x i16> %x0, <2 x i64> %x1, i8 %x2)
  %res1 = call <2 x i64> @llvm.x86.avx512.mask.pmovsxw.q.128(<8 x i16> %x0, <2 x i64> zeroinitializer, i8 %x2)
  %res2 = call <2 x i64> @llvm.x86.avx512.mask.pmovsxw.q.128(<8 x i16> %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.pmovsxw.q.256(<8 x i16>, <4 x i64>, i8)

define <4 x i64>@test_int_x86_avx512_mask_pmovsxw_q_256(<8 x i16> %x0, <4 x i64> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmovsxw_q_256:
; X86:       # %bb.0:
; X86-NEXT:    vpmovsxwq %xmm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x24,0xd0]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpmovsxwq %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x24,0xc8]
; X86-NEXT:    vpmovsxwq %xmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x24,0xc0]
; 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_pmovsxw_q_256:
; X64:       # %bb.0:
; X64-NEXT:    vpmovsxwq %xmm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x24,0xd0]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmovsxwq %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x24,0xc8]
; X64-NEXT:    vpmovsxwq %xmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x24,0xc0]
; 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.pmovsxw.q.256(<8 x i16> %x0, <4 x i64> %x1, i8 %x2)
  %res1 = call <4 x i64> @llvm.x86.avx512.mask.pmovsxw.q.256(<8 x i16> %x0, <4 x i64> zeroinitializer, i8 %x2)
  %res2 = call <4 x i64> @llvm.x86.avx512.mask.pmovsxw.q.256(<8 x i16> %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 <2 x i64> @llvm.x86.avx512.mask.psra.q.128(<2 x i64>, <2 x i64>, <2 x i64>, i8)

define <2 x i64>@test_int_x86_avx512_mask_psra_q_128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_psra_q_128:
; X86:       # %bb.0:
; X86-NEXT:    vpsraq %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf1,0xfd,0x08,0xe2,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpsraq %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0xfd,0x09,0xe2,0xd1]
; X86-NEXT:    vpsraq %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0x89,0xe2,0xc1]
; X86-NEXT:    vpaddq %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd4,0xc3]
; X86-NEXT:    vpaddq %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xd4,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_psra_q_128:
; X64:       # %bb.0:
; X64-NEXT:    vpsraq %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf1,0xfd,0x08,0xe2,0xd9]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpsraq %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0xfd,0x09,0xe2,0xd1]
; X64-NEXT:    vpsraq %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0x89,0xe2,0xc1]
; X64-NEXT:    vpaddq %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd4,0xc3]
; X64-NEXT:    vpaddq %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xd4,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x i64> @llvm.x86.avx512.mask.psra.q.128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i8 %x3)
  %res1 = call <2 x i64> @llvm.x86.avx512.mask.psra.q.128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> zeroinitializer, i8 %x3)
  %res2 = call <2 x i64> @llvm.x86.avx512.mask.psra.q.128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, 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.psra.q.256(<4 x i64>, <2 x i64>, <4 x i64>, i8)

define <4 x i64>@test_int_x86_avx512_mask_psra_q_256(<4 x i64> %x0, <2 x i64> %x1, <4 x i64> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_psra_q_256:
; X86:       # %bb.0:
; X86-NEXT:    vpsraq %xmm1, %ymm0, %ymm3 # encoding: [0x62,0xf1,0xfd,0x28,0xe2,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpsraq %xmm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0xfd,0x29,0xe2,0xd1]
; X86-NEXT:    vpsraq %xmm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0xa9,0xe2,0xc1]
; X86-NEXT:    vpaddq %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd4,0xc3]
; X86-NEXT:    vpaddq %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xd4,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_psra_q_256:
; X64:       # %bb.0:
; X64-NEXT:    vpsraq %xmm1, %ymm0, %ymm3 # encoding: [0x62,0xf1,0xfd,0x28,0xe2,0xd9]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpsraq %xmm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0xfd,0x29,0xe2,0xd1]
; X64-NEXT:    vpsraq %xmm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0xa9,0xe2,0xc1]
; X64-NEXT:    vpaddq %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd4,0xc3]
; X64-NEXT:    vpaddq %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xd4,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i64> @llvm.x86.avx512.mask.psra.q.256(<4 x i64> %x0, <2 x i64> %x1, <4 x i64> %x2, i8 %x3)
  %res1 = call <4 x i64> @llvm.x86.avx512.mask.psra.q.256(<4 x i64> %x0, <2 x i64> %x1, <4 x i64> zeroinitializer, i8 %x3)
  %res2 = call <4 x i64> @llvm.x86.avx512.mask.psra.q.256(<4 x i64> %x0, <2 x i64> %x1, <4 x i64> %x2, i8 -1)
  %res3 = add <4 x i64> %res, %res1
  %res4 = add <4 x i64> %res3, %res2
  ret <4 x i64> %res4
}

declare <2 x i64> @llvm.x86.avx512.mask.psra.qi.128(<2 x i64>, i32, <2 x i64>, i8)

define <2 x i64>@test_int_x86_avx512_mask_psra_qi_128(<2 x i64> %x0, i32 %x1, <2 x i64> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_psra_qi_128:
; X86:       # %bb.0:
; X86-NEXT:    vpsraq $3, %xmm0, %xmm2 # encoding: [0x62,0xf1,0xed,0x08,0x72,0xe0,0x03]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpsraq $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x09,0x72,0xe0,0x03]
; X86-NEXT:    vpsraq $3, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0x89,0x72,0xe0,0x03]
; 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_psra_qi_128:
; X64:       # %bb.0:
; X64-NEXT:    vpsraq $3, %xmm0, %xmm2 # encoding: [0x62,0xf1,0xed,0x08,0x72,0xe0,0x03]
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpsraq $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x09,0x72,0xe0,0x03]
; X64-NEXT:    vpsraq $3, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0x89,0x72,0xe0,0x03]
; 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.psra.qi.128(<2 x i64> %x0, i32 3, <2 x i64> %x2, i8 %x3)
  %res1 = call <2 x i64> @llvm.x86.avx512.mask.psra.qi.128(<2 x i64> %x0, i32 3, <2 x i64> zeroinitializer, i8 %x3)
  %res2 = call <2 x i64> @llvm.x86.avx512.mask.psra.qi.128(<2 x i64> %x0, i32 3, <2 x i64> %x2, 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.psra.qi.256(<4 x i64>, i32, <4 x i64>, i8)

define <4 x i64>@test_int_x86_avx512_mask_psra_qi_256(<4 x i64> %x0, i32 %x1, <4 x i64> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_psra_qi_256:
; X86:       # %bb.0:
; X86-NEXT:    vpsraq $3, %ymm0, %ymm2 # encoding: [0x62,0xf1,0xed,0x28,0x72,0xe0,0x03]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpsraq $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x29,0x72,0xe0,0x03]
; X86-NEXT:    vpsraq $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0xa9,0x72,0xe0,0x03]
; 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_psra_qi_256:
; X64:       # %bb.0:
; X64-NEXT:    vpsraq $3, %ymm0, %ymm2 # encoding: [0x62,0xf1,0xed,0x28,0x72,0xe0,0x03]
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpsraq $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x29,0x72,0xe0,0x03]
; X64-NEXT:    vpsraq $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0xa9,0x72,0xe0,0x03]
; 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.psra.qi.256(<4 x i64> %x0, i32 3, <4 x i64> %x2, i8 %x3)
  %res1 = call <4 x i64> @llvm.x86.avx512.mask.psra.qi.256(<4 x i64> %x0, i32 3, <4 x i64> zeroinitializer, i8 %x3)
  %res2 = call <4 x i64> @llvm.x86.avx512.mask.psra.qi.256(<4 x i64> %x0, i32 3, <4 x i64> %x2, i8 -1)
  %res3 = add <4 x i64> %res, %res1
  %res4 = add <4 x i64> %res3, %res2
  ret <4 x i64> %res4
}

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

define <2 x i64>@test_int_x86_avx512_mask_psrav_q_128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_psrav_q_128:
; X86:       # %bb.0:
; X86-NEXT:    vpsravq %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf2,0xfd,0x08,0x46,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpsravq %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x46,0xd1]
; X86-NEXT:    vpsravq %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x46,0xc1]
; X86-NEXT:    vpaddq %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd4,0xc3]
; X86-NEXT:    vpaddq %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xd4,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_psrav_q_128:
; X64:       # %bb.0:
; X64-NEXT:    vpsravq %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf2,0xfd,0x08,0x46,0xd9]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpsravq %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x46,0xd1]
; X64-NEXT:    vpsravq %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x46,0xc1]
; X64-NEXT:    vpaddq %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd4,0xc3]
; X64-NEXT:    vpaddq %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xd4,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x i64> @llvm.x86.avx512.mask.psrav.q.128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i8 %x3)
  %res1 = call <2 x i64> @llvm.x86.avx512.mask.psrav.q.128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> zeroinitializer, i8 %x3)
  %res2 = call <2 x i64> @llvm.x86.avx512.mask.psrav.q.128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i8 -1)
  %res3 = add <2 x i64> %res, %res1
  %res4 = add <2 x i64> %res3, %res2
  ret <2 x i64> %res4
}

define <2 x i64>@test_int_x86_avx512_mask_psrav_q_128_const(i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_psrav_q_128_const:
; X86:       # %bb.0:
; X86-NEXT:    vmovdqa {{\.LCPI.*}}, %xmm0 # EVEX TO VEX Compression xmm0 = [2,0,4294967287,4294967295]
; X86-NEXT:    # encoding: [0xc5,0xf9,0x6f,0x05,A,A,A,A]
; X86-NEXT:    # fixup A - offset: 4, value: {{\.LCPI.*}}, kind: FK_Data_4
; X86-NEXT:    vpsravq {{\.LCPI.*}}, %xmm0, %xmm0 # encoding: [0x62,0xf2,0xfd,0x08,0x46,0x05,A,A,A,A]
; X86-NEXT:    # fixup A - offset: 6, value: {{\.LCPI.*}}, kind: FK_Data_4
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_psrav_q_128_const:
; X64:       # %bb.0:
; X64-NEXT:    vmovdqa {{.*}}(%rip), %xmm0 # EVEX TO VEX Compression xmm0 = [2,18446744073709551607]
; X64-NEXT:    # encoding: [0xc5,0xf9,0x6f,0x05,A,A,A,A]
; X64-NEXT:    # fixup A - offset: 4, value: {{\.LCPI.*}}-4, kind: reloc_riprel_4byte
; X64-NEXT:    vpsravq {{.*}}(%rip), %xmm0, %xmm0 # encoding: [0x62,0xf2,0xfd,0x08,0x46,0x05,A,A,A,A]
; X64-NEXT:    # fixup A - offset: 6, value: {{\.LCPI.*}}-4, kind: reloc_riprel_4byte
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x i64> @llvm.x86.avx512.mask.psrav.q.128(<2 x i64> <i64 2, i64 -9>, <2 x i64> <i64 1, i64 90>, <2 x i64> zeroinitializer, i8 -1)
  ret <2 x i64> %res
}

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

define <4 x i64>@test_int_x86_avx512_mask_psrav_q_256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_psrav_q_256:
; X86:       # %bb.0:
; X86-NEXT:    vpsravq %ymm1, %ymm0, %ymm3 # encoding: [0x62,0xf2,0xfd,0x28,0x46,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpsravq %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x46,0xd1]
; X86-NEXT:    vpsravq %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x46,0xc1]
; X86-NEXT:    vpaddq %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd4,0xc3]
; X86-NEXT:    vpaddq %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xd4,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_psrav_q_256:
; X64:       # %bb.0:
; X64-NEXT:    vpsravq %ymm1, %ymm0, %ymm3 # encoding: [0x62,0xf2,0xfd,0x28,0x46,0xd9]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpsravq %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x46,0xd1]
; X64-NEXT:    vpsravq %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x46,0xc1]
; X64-NEXT:    vpaddq %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd4,0xc3]
; X64-NEXT:    vpaddq %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xd4,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i64> @llvm.x86.avx512.mask.psrav.q.256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2, i8 %x3)
  %res1 = call <4 x i64> @llvm.x86.avx512.mask.psrav.q.256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> zeroinitializer, i8 %x3)
  %res2 = call <4 x i64> @llvm.x86.avx512.mask.psrav.q.256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2, i8 -1)
  %res3 = add <4 x i64> %res, %res1
  %res4 = add <4 x i64> %res3, %res2
  ret <4 x i64> %res4
}

declare <2 x double> @llvm.x86.avx512.mask.cvtdq2pd.128(<4 x i32>, <2 x double>, i8)

define <2 x double>@test_int_x86_avx512_mask_cvt_dq2pd_128(<4 x i32> %x0, <2 x double> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_cvt_dq2pd_128:
; X86:       # %bb.0:
; X86-NEXT:    vcvtdq2pd %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xfa,0xe6,0xd0]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vcvtdq2pd %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7e,0x09,0xe6,0xc8]
; X86-NEXT:    vaddpd %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0x58,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_cvt_dq2pd_128:
; X64:       # %bb.0:
; X64-NEXT:    vcvtdq2pd %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xfa,0xe6,0xd0]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vcvtdq2pd %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7e,0x09,0xe6,0xc8]
; X64-NEXT:    vaddpd %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0x58,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x double> @llvm.x86.avx512.mask.cvtdq2pd.128(<4 x i32> %x0, <2 x double> %x1, i8 %x2)
  %res1 = call <2 x double> @llvm.x86.avx512.mask.cvtdq2pd.128(<4 x i32> %x0, <2 x double> %x1, i8 -1)
  %res2 = fadd <2 x double> %res, %res1
  ret <2 x double> %res2
}

declare <4 x double> @llvm.x86.avx512.mask.cvtdq2pd.256(<4 x i32>, <4 x double>, i8)

define <4 x double>@test_int_x86_avx512_mask_cvt_dq2pd_256(<4 x i32> %x0, <4 x double> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_cvt_dq2pd_256:
; X86:       # %bb.0:
; X86-NEXT:    vcvtdq2pd %xmm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xfe,0xe6,0xd0]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vcvtdq2pd %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7e,0x29,0xe6,0xc8]
; X86-NEXT:    vaddpd %ymm2, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0x58,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_cvt_dq2pd_256:
; X64:       # %bb.0:
; X64-NEXT:    vcvtdq2pd %xmm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xfe,0xe6,0xd0]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vcvtdq2pd %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7e,0x29,0xe6,0xc8]
; X64-NEXT:    vaddpd %ymm2, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0x58,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x double> @llvm.x86.avx512.mask.cvtdq2pd.256(<4 x i32> %x0, <4 x double> %x1, i8 %x2)
  %res1 = call <4 x double> @llvm.x86.avx512.mask.cvtdq2pd.256(<4 x i32> %x0, <4 x double> %x1, i8 -1)
  %res2 = fadd <4 x double> %res, %res1
  ret <4 x double> %res2
}

declare <2 x double> @llvm.x86.avx512.mask.cvtudq2pd.128(<4 x i32>, <2 x double>, i8)

define <2 x double>@test_int_x86_avx512_mask_cvt_udq2pd_128(<4 x i32> %x0, <2 x double> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_cvt_udq2pd_128:
; X86:       # %bb.0:
; X86-NEXT:    vcvtudq2pd %xmm0, %xmm2 # encoding: [0x62,0xf1,0x7e,0x08,0x7a,0xd0]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vcvtudq2pd %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7e,0x09,0x7a,0xc8]
; X86-NEXT:    vaddpd %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0x58,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_cvt_udq2pd_128:
; X64:       # %bb.0:
; X64-NEXT:    vcvtudq2pd %xmm0, %xmm2 # encoding: [0x62,0xf1,0x7e,0x08,0x7a,0xd0]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vcvtudq2pd %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7e,0x09,0x7a,0xc8]
; X64-NEXT:    vaddpd %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0x58,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x double> @llvm.x86.avx512.mask.cvtudq2pd.128(<4 x i32> %x0, <2 x double> %x1, i8 %x2)
  %res1 = call <2 x double> @llvm.x86.avx512.mask.cvtudq2pd.128(<4 x i32> %x0, <2 x double> %x1, i8 -1)
  %res2 = fadd <2 x double> %res, %res1
  ret <2 x double> %res2
}

declare <4 x double> @llvm.x86.avx512.mask.cvtudq2pd.256(<4 x i32>, <4 x double>, i8)

define <4 x double>@test_int_x86_avx512_mask_cvt_udq2pd_256(<4 x i32> %x0, <4 x double> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_cvt_udq2pd_256:
; X86:       # %bb.0:
; X86-NEXT:    vcvtudq2pd %xmm0, %ymm2 # encoding: [0x62,0xf1,0x7e,0x28,0x7a,0xd0]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vcvtudq2pd %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7e,0x29,0x7a,0xc8]
; X86-NEXT:    vaddpd %ymm2, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0x58,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_cvt_udq2pd_256:
; X64:       # %bb.0:
; X64-NEXT:    vcvtudq2pd %xmm0, %ymm2 # encoding: [0x62,0xf1,0x7e,0x28,0x7a,0xd0]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vcvtudq2pd %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7e,0x29,0x7a,0xc8]
; X64-NEXT:    vaddpd %ymm2, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0x58,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x double> @llvm.x86.avx512.mask.cvtudq2pd.256(<4 x i32> %x0, <4 x double> %x1, i8 %x2)
  %res1 = call <4 x double> @llvm.x86.avx512.mask.cvtudq2pd.256(<4 x i32> %x0, <4 x double> %x1, i8 -1)
  %res2 = fadd <4 x double> %res, %res1
  ret <4 x double> %res2
}

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

define <4 x i32>@test_int_x86_avx512_mask_valign_d_128(<4 x i32> %x0, <4 x i32> %x1,<4 x i32> %x3, i8 %x4) {
; X86-LABEL: test_int_x86_avx512_mask_valign_d_128:
; X86:       # %bb.0:
; X86-NEXT:    vpalignr $8, %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x0f,0xd9,0x08]
; X86-NEXT:    # xmm3 = xmm1[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    valignd $2, %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x03,0xd1,0x02]
; X86-NEXT:    # xmm2 {%k1} = xmm1[2,3],xmm0[0,1]
; X86-NEXT:    valignd $2, %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0x89,0x03,0xc1,0x02]
; X86-NEXT:    # xmm0 {%k1} {z} = xmm1[2,3],xmm0[0,1]
; X86-NEXT:    vpaddd %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xfe,0xc0]
; X86-NEXT:    vpaddd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_valign_d_128:
; X64:       # %bb.0:
; X64-NEXT:    vpalignr $8, %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x0f,0xd9,0x08]
; X64-NEXT:    # xmm3 = xmm1[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    valignd $2, %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x03,0xd1,0x02]
; X64-NEXT:    # xmm2 {%k1} = xmm1[2,3],xmm0[0,1]
; X64-NEXT:    valignd $2, %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0x89,0x03,0xc1,0x02]
; X64-NEXT:    # xmm0 {%k1} {z} = xmm1[2,3],xmm0[0,1]
; X64-NEXT:    vpaddd %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xfe,0xc0]
; X64-NEXT:    vpaddd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i32> @llvm.x86.avx512.mask.valign.d.128(<4 x i32> %x0, <4 x i32> %x1, i32 2, <4 x i32> %x3, i8 %x4)
  %res1 = call <4 x i32> @llvm.x86.avx512.mask.valign.d.128(<4 x i32> %x0, <4 x i32> %x1, i32 2, <4 x i32> %x3, i8 -1)
  %res2 = call <4 x i32> @llvm.x86.avx512.mask.valign.d.128(<4 x i32> %x0, <4 x i32> %x1, i32 2, <4 x i32> zeroinitializer,i8 %x4)
  %res3 = add <4 x i32> %res, %res1
  %res4 = add <4 x i32> %res3, %res2
  ret <4 x i32> %res4
}

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

define <8 x i32>@test_int_x86_avx512_mask_valign_d_256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x3, i8 %x4) {
; X86-LABEL: test_int_x86_avx512_mask_valign_d_256:
; X86:       # %bb.0:
; X86-NEXT:    valignq $3, %ymm1, %ymm0, %ymm3 # encoding: [0x62,0xf3,0xfd,0x28,0x03,0xd9,0x03]
; X86-NEXT:    # ymm3 = ymm1[3],ymm0[0,1,2]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    valignd $6, %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x03,0xd1,0x06]
; X86-NEXT:    # ymm2 {%k1} = ymm1[6,7],ymm0[0,1,2,3,4,5]
; 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_valign_d_256:
; X64:       # %bb.0:
; X64-NEXT:    valignq $3, %ymm1, %ymm0, %ymm3 # encoding: [0x62,0xf3,0xfd,0x28,0x03,0xd9,0x03]
; X64-NEXT:    # ymm3 = ymm1[3],ymm0[0,1,2]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    valignd $6, %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x03,0xd1,0x06]
; X64-NEXT:    # ymm2 {%k1} = ymm1[6,7],ymm0[0,1,2,3,4,5]
; 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.valign.d.256(<8 x i32> %x0, <8 x i32> %x1, i32 6, <8 x i32> %x3, i8 %x4)
  %res1 = call <8 x i32> @llvm.x86.avx512.mask.valign.d.256(<8 x i32> %x0, <8 x i32> %x1, i32 6, <8 x i32> %x3, i8 -1)
  %res2 = add <8 x i32> %res, %res1
  ret <8 x i32> %res2
}

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

define <2 x i64>@test_int_x86_avx512_mask_valign_q_128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x3, i8 %x4) {
; X86-LABEL: test_int_x86_avx512_mask_valign_q_128:
; X86:       # %bb.0:
; X86-NEXT:    vpalignr $8, %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x0f,0xd9,0x08]
; X86-NEXT:    # xmm3 = xmm1[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    valignq $1, %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x03,0xd1,0x01]
; X86-NEXT:    # xmm2 {%k1} = xmm1[1],xmm0[0]
; X86-NEXT:    vpaddq %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xd4,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_valign_q_128:
; X64:       # %bb.0:
; X64-NEXT:    vpalignr $8, %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x0f,0xd9,0x08]
; X64-NEXT:    # xmm3 = xmm1[8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4,5,6,7]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    valignq $1, %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x03,0xd1,0x01]
; X64-NEXT:    # xmm2 {%k1} = xmm1[1],xmm0[0]
; X64-NEXT:    vpaddq %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xd4,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x i64> @llvm.x86.avx512.mask.valign.q.128(<2 x i64> %x0, <2 x i64> %x1, i32 1, <2 x i64> %x3, i8 %x4)
  %res1 = call <2 x i64> @llvm.x86.avx512.mask.valign.q.128(<2 x i64> %x0, <2 x i64> %x1, i32 1, <2 x i64> %x3, i8 -1)
  %res2 = add <2 x i64> %res, %res1
  ret <2 x i64> %res2
}

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

define <4 x i64>@test_int_x86_avx512_mask_valign_q_256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x3, i8 %x4) {
; X86-LABEL: test_int_x86_avx512_mask_valign_q_256:
; X86:       # %bb.0:
; X86-NEXT:    valignq $3, %ymm1, %ymm0, %ymm3 # encoding: [0x62,0xf3,0xfd,0x28,0x03,0xd9,0x03]
; X86-NEXT:    # ymm3 = ymm1[3],ymm0[0,1,2]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    valignq $3, %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x03,0xd1,0x03]
; X86-NEXT:    # ymm2 {%k1} = ymm1[3],ymm0[0,1,2]
; X86-NEXT:    vpaddq %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xd4,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_valign_q_256:
; X64:       # %bb.0:
; X64-NEXT:    valignq $3, %ymm1, %ymm0, %ymm3 # encoding: [0x62,0xf3,0xfd,0x28,0x03,0xd9,0x03]
; X64-NEXT:    # ymm3 = ymm1[3],ymm0[0,1,2]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    valignq $3, %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x03,0xd1,0x03]
; X64-NEXT:    # ymm2 {%k1} = ymm1[3],ymm0[0,1,2]
; X64-NEXT:    vpaddq %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xd4,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i64> @llvm.x86.avx512.mask.valign.q.256(<4 x i64> %x0, <4 x i64> %x1, i32 3, <4 x i64> %x3, i8 %x4)
  %res1 = call <4 x i64> @llvm.x86.avx512.mask.valign.q.256(<4 x i64> %x0, <4 x i64> %x1, i32 3, <4 x i64> %x3, i8 -1)
  %res2 = add <4 x i64> %res, %res1
  ret <4 x i64> %res2
}

declare <4 x double> @llvm.x86.avx512.mask.vpermilvar.pd.256(<4 x double>, <4 x i64>, <4 x double>, i8)

define <4 x double>@test_int_x86_avx512_mask_vpermilvar_pd_256(<4 x double> %x0, <4 x i64> %x1, <4 x double> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_vpermilvar_pd_256:
; X86:       # %bb.0:
; X86-NEXT:    vpermilpd %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x0d,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpermilpd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x0d,0xd1]
; X86-NEXT:    vpermilpd %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x0d,0xc1]
; X86-NEXT:    vaddpd %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0x58,0xc0]
; X86-NEXT:    vaddpd %ymm0, %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0x58,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_vpermilvar_pd_256:
; X64:       # %bb.0:
; X64-NEXT:    vpermilpd %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x0d,0xd9]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpermilpd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x0d,0xd1]
; X64-NEXT:    vpermilpd %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x0d,0xc1]
; X64-NEXT:    vaddpd %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0x58,0xc0]
; X64-NEXT:    vaddpd %ymm0, %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0x58,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x double> @llvm.x86.avx512.mask.vpermilvar.pd.256(<4 x double> %x0, <4 x i64> %x1, <4 x double> %x2, i8 %x3)
  %res1 = call <4 x double> @llvm.x86.avx512.mask.vpermilvar.pd.256(<4 x double> %x0, <4 x i64> %x1, <4 x double> zeroinitializer, i8 %x3)
  %res2 = call <4 x double> @llvm.x86.avx512.mask.vpermilvar.pd.256(<4 x double> %x0, <4 x i64> %x1, <4 x double> %x2, i8 -1)
  %res3 = fadd <4 x double> %res, %res1
  %res4 = fadd <4 x double> %res2, %res3
  ret <4 x double> %res4
}

declare <2 x double> @llvm.x86.avx512.mask.vpermilvar.pd.128(<2 x double>, <2 x i64>, <2 x double>, i8)

define <2 x double>@test_int_x86_avx512_mask_vpermilvar_pd_128(<2 x double> %x0, <2 x i64> %x1, <2 x double> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_vpermilvar_pd_128:
; X86:       # %bb.0:
; X86-NEXT:    vpermilpd %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x0d,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpermilpd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x0d,0xd1]
; X86-NEXT:    vpermilpd %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x0d,0xc1]
; X86-NEXT:    vaddpd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x58,0xc0]
; X86-NEXT:    vaddpd %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x58,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_vpermilvar_pd_128:
; X64:       # %bb.0:
; X64-NEXT:    vpermilpd %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x0d,0xd9]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpermilpd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x0d,0xd1]
; X64-NEXT:    vpermilpd %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x0d,0xc1]
; X64-NEXT:    vaddpd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x58,0xc0]
; X64-NEXT:    vaddpd %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x58,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x double> @llvm.x86.avx512.mask.vpermilvar.pd.128(<2 x double> %x0, <2 x i64> %x1, <2 x double> %x2, i8 %x3)
  %res1 = call <2 x double> @llvm.x86.avx512.mask.vpermilvar.pd.128(<2 x double> %x0, <2 x i64> %x1, <2 x double> zeroinitializer, i8 %x3)
  %res2 = call <2 x double> @llvm.x86.avx512.mask.vpermilvar.pd.128(<2 x double> %x0, <2 x i64> %x1, <2 x double> %x2, i8 -1)
  %res3 = fadd <2 x double> %res, %res1
  %res4 = fadd <2 x double> %res3, %res2
  ret <2 x double> %res4
}

declare <8 x float> @llvm.x86.avx512.mask.vpermilvar.ps.256(<8 x float>, <8 x i32>, <8 x float>, i8)

define <8 x float>@test_int_x86_avx512_mask_vpermilvar_ps_256(<8 x float> %x0, <8 x i32> %x1, <8 x float> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_vpermilvar_ps_256:
; X86:       # %bb.0:
; X86-NEXT:    vpermilps %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x0c,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpermilps %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x0c,0xd1]
; X86-NEXT:    vpermilps %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x0c,0xc1]
; X86-NEXT:    vaddps %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xec,0x58,0xc0]
; X86-NEXT:    vaddps %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x58,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_vpermilvar_ps_256:
; X64:       # %bb.0:
; X64-NEXT:    vpermilps %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x0c,0xd9]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpermilps %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x0c,0xd1]
; X64-NEXT:    vpermilps %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x0c,0xc1]
; X64-NEXT:    vaddps %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xec,0x58,0xc0]
; X64-NEXT:    vaddps %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x58,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask.vpermilvar.ps.256(<8 x float> %x0, <8 x i32> %x1, <8 x float> %x2, i8 %x3)
  %res1 = call <8 x float> @llvm.x86.avx512.mask.vpermilvar.ps.256(<8 x float> %x0, <8 x i32> %x1, <8 x float> zeroinitializer, i8 %x3)
  %res2 = call <8 x float> @llvm.x86.avx512.mask.vpermilvar.ps.256(<8 x float> %x0, <8 x i32> %x1, <8 x float> %x2, i8 -1)
  %res3 = fadd <8 x float> %res, %res1
  %res4 = fadd <8 x float> %res3, %res2
  ret <8 x float> %res4
}

declare <4 x float> @llvm.x86.avx512.mask.vpermilvar.ps.128(<4 x float>, <4 x i32>, <4 x float>, i8)

define <4 x float>@test_int_x86_avx512_mask_vpermilvar_ps_128(<4 x float> %x0, <4 x i32> %x1, <4 x float> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_vpermilvar_ps_128:
; X86:       # %bb.0:
; X86-NEXT:    vpermilps %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x0c,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpermilps %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x0c,0xd1]
; X86-NEXT:    vpermilps %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x0c,0xc1]
; X86-NEXT:    vaddps %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe8,0x58,0xc0]
; X86-NEXT:    vaddps %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe0,0x58,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_vpermilvar_ps_128:
; X64:       # %bb.0:
; X64-NEXT:    vpermilps %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x0c,0xd9]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpermilps %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x0c,0xd1]
; X64-NEXT:    vpermilps %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x0c,0xc1]
; X64-NEXT:    vaddps %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe8,0x58,0xc0]
; X64-NEXT:    vaddps %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe0,0x58,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask.vpermilvar.ps.128(<4 x float> %x0, <4 x i32> %x1, <4 x float> %x2, i8 %x3)
  %res1 = call <4 x float> @llvm.x86.avx512.mask.vpermilvar.ps.128(<4 x float> %x0, <4 x i32> %x1, <4 x float> zeroinitializer, i8 %x3)
  %res2 = call <4 x float> @llvm.x86.avx512.mask.vpermilvar.ps.128(<4 x float> %x0, <4 x i32> %x1, <4 x float> %x2, i8 -1)
  %res3 = fadd <4 x float> %res, %res1
  %res4 = fadd <4 x float> %res2, %res3
  ret <4 x float> %res4
}

declare <4 x float> @llvm.x86.avx512.mask.vextractf32x4.256(<8 x float>, i32, <4 x float>, i8)

define <4 x float>@test_int_x86_avx512_mask_vextractf32x4_256(<8 x float> %x0, <4 x float> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_vextractf32x4_256:
; X86:       # %bb.0:
; X86-NEXT:    vextractf128 $1, %ymm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x7d,0x19,0xc2,0x01]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vextractf32x4 $1, %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x19,0xc1,0x01]
; X86-NEXT:    vextractf32x4 $1, %ymm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0xa9,0x19,0xc0,0x01]
; X86-NEXT:    vaddps %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf0,0x58,0xc0]
; X86-NEXT:    vaddps %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe8,0x58,0xc0]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_vextractf32x4_256:
; X64:       # %bb.0:
; X64-NEXT:    vextractf128 $1, %ymm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x7d,0x19,0xc2,0x01]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vextractf32x4 $1, %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x19,0xc1,0x01]
; X64-NEXT:    vextractf32x4 $1, %ymm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0xa9,0x19,0xc0,0x01]
; X64-NEXT:    vaddps %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf0,0x58,0xc0]
; X64-NEXT:    vaddps %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe8,0x58,0xc0]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
  %res  = call <4 x float> @llvm.x86.avx512.mask.vextractf32x4.256(<8 x float> %x0, i32 1, <4 x float> %x2, i8 %x3)
  %res1 = call <4 x float> @llvm.x86.avx512.mask.vextractf32x4.256(<8 x float> %x0, i32 1, <4 x float> zeroinitializer, i8 %x3)
  %res2 = call <4 x float> @llvm.x86.avx512.mask.vextractf32x4.256(<8 x float> %x0, i32 1, <4 x float> zeroinitializer, i8 -1)
  %res3 = fadd <4 x float> %res, %res1
  %res4 = fadd <4 x float> %res2, %res3
  ret <4 x float> %res4
}

declare <8 x float> @llvm.x86.avx512.mask.insertf32x4.256(<8 x float>, <4 x float>, i32, <8 x float>, i8)

define <8 x float>@test_int_x86_avx512_mask_insertf32x4_256(<8 x float> %x0, <4 x float> %x1, <8 x float> %x3, i8 %x4) {
; X86-LABEL: test_int_x86_avx512_mask_insertf32x4_256:
; X86:       # %bb.0:
; X86-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x7d,0x18,0xd9,0x01]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vinsertf32x4 $1, %xmm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x18,0xd1,0x01]
; X86-NEXT:    vaddps %ymm3, %ymm2, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xec,0x58,0xd3]
; X86-NEXT:    vinsertf32x4 $1, %xmm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0xa9,0x18,0xc1,0x01]
; X86-NEXT:    vaddps %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x58,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_insertf32x4_256:
; X64:       # %bb.0:
; X64-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x7d,0x18,0xd9,0x01]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vinsertf32x4 $1, %xmm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x18,0xd1,0x01]
; X64-NEXT:    vaddps %ymm3, %ymm2, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xec,0x58,0xd3]
; X64-NEXT:    vinsertf32x4 $1, %xmm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0xa9,0x18,0xc1,0x01]
; X64-NEXT:    vaddps %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x58,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask.insertf32x4.256(<8 x float> %x0, <4 x float> %x1, i32 1, <8 x float> %x3, i8 %x4)
  %res1 = call <8 x float> @llvm.x86.avx512.mask.insertf32x4.256(<8 x float> %x0, <4 x float> %x1, i32 1, <8 x float> %x3, i8 -1)
  %res2 = call <8 x float> @llvm.x86.avx512.mask.insertf32x4.256(<8 x float> %x0, <4 x float> %x1, i32 1, <8 x float> zeroinitializer, i8 %x4)
  %res3 = fadd <8 x float> %res, %res1
  %res4 = fadd <8 x float> %res2, %res3
  ret <8 x float> %res4
}

declare <8 x i32> @llvm.x86.avx512.mask.inserti32x4.256(<8 x i32>, <4 x i32>, i32, <8 x i32>, i8)

define <8 x i32>@test_int_x86_avx512_mask_inserti32x4_256(<8 x i32> %x0, <4 x i32> %x1, <8 x i32> %x3, i8 %x4) {
; X86-LABEL: test_int_x86_avx512_mask_inserti32x4_256:
; X86:       # %bb.0:
; X86-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x7d,0x38,0xd9,0x01]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vinserti32x4 $1, %xmm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x38,0xd1,0x01]
; X86-NEXT:    vinserti32x4 $1, %xmm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0xa9,0x38,0xc1,0x01]
; X86-NEXT:    vpaddd %ymm0, %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0xfe,0xc0]
; X86-NEXT:    vpaddd %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_inserti32x4_256:
; X64:       # %bb.0:
; X64-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x7d,0x38,0xd9,0x01]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vinserti32x4 $1, %xmm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x38,0xd1,0x01]
; X64-NEXT:    vinserti32x4 $1, %xmm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0xa9,0x38,0xc1,0x01]
; X64-NEXT:    vpaddd %ymm0, %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0xfe,0xc0]
; X64-NEXT:    vpaddd %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]

  %res = call <8 x i32> @llvm.x86.avx512.mask.inserti32x4.256(<8 x i32> %x0, <4 x i32> %x1, i32 1, <8 x i32> %x3, i8 %x4)
  %res1 = call <8 x i32> @llvm.x86.avx512.mask.inserti32x4.256(<8 x i32> %x0, <4 x i32> %x1, i32 1, <8 x i32> %x3, i8 -1)
  %res2 = call <8 x i32> @llvm.x86.avx512.mask.inserti32x4.256(<8 x i32> %x0, <4 x i32> %x1, i32 1, <8 x i32> zeroinitializer, i8 %x4)
  %res3 = add <8 x i32> %res, %res1
  %res4 = add <8 x i32> %res2, %res3
  ret <8 x i32> %res4
}

define <8 x float> @test_mm512_maskz_max_ps_256(<8 x float> %a0, <8 x float> %a1, i8 %mask) {
; X86-LABEL: test_mm512_maskz_max_ps_256:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vmaxps %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7c,0xa9,0x5f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mm512_maskz_max_ps_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vmaxps %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7c,0xa9,0x5f,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask.max.ps.256(<8 x float> %a0, <8 x float> %a1, <8 x float>zeroinitializer, i8 %mask)
  ret <8 x float> %res
}

define <8 x float> @test_mm512_mask_max_ps_256(<8 x float> %a0, <8 x float> %a1, <8 x float> %src, i8 %mask) {
; X86-LABEL: test_mm512_mask_max_ps_256:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vmaxps %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7c,0x29,0x5f,0xd1]
; X86-NEXT:    vmovaps %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x28,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mm512_mask_max_ps_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vmaxps %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7c,0x29,0x5f,0xd1]
; X64-NEXT:    vmovaps %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x28,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask.max.ps.256(<8 x float> %a0, <8 x float> %a1, <8 x float> %src, i8 %mask)
  ret <8 x float> %res
}

define <8 x float> @test_mm512_max_ps_256(<8 x float> %a0, <8 x float> %a1, i8 %mask) {
; CHECK-LABEL: test_mm512_max_ps_256:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vmaxps %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x5f,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask.max.ps.256(<8 x float> %a0, <8 x float> %a1, <8 x float>zeroinitializer, i8 -1)
  ret <8 x float> %res
}
declare <8 x float> @llvm.x86.avx512.mask.max.ps.256(<8 x float>, <8 x float>, <8 x float>, i8)

define <4 x float> @test_mm512_maskz_max_ps_128(<4 x float> %a0, <4 x float> %a1, i8 %mask) {
; X86-LABEL: test_mm512_maskz_max_ps_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vmaxps %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7c,0x89,0x5f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mm512_maskz_max_ps_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vmaxps %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7c,0x89,0x5f,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask.max.ps.128(<4 x float> %a0, <4 x float> %a1, <4 x float>zeroinitializer, i8 %mask)
  ret <4 x float> %res
}

define <4 x float> @test_mm512_mask_max_ps_128(<4 x float> %a0, <4 x float> %a1, <4 x float> %src, i8 %mask) {
; X86-LABEL: test_mm512_mask_max_ps_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vmaxps %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7c,0x09,0x5f,0xd1]
; X86-NEXT:    vmovaps %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x28,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mm512_mask_max_ps_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vmaxps %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7c,0x09,0x5f,0xd1]
; X64-NEXT:    vmovaps %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x28,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask.max.ps.128(<4 x float> %a0, <4 x float> %a1, <4 x float> %src, i8 %mask)
  ret <4 x float> %res
}

define <4 x float> @test_mm512_max_ps_128(<4 x float> %a0, <4 x float> %a1, i8 %mask) {
; CHECK-LABEL: test_mm512_max_ps_128:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vmaxps %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x5f,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask.max.ps.128(<4 x float> %a0, <4 x float> %a1, <4 x float>zeroinitializer, i8 -1)
  ret <4 x float> %res
}
declare <4 x float> @llvm.x86.avx512.mask.max.ps.128(<4 x float>, <4 x float>, <4 x float>, i8)

define <8 x float> @test_mm512_maskz_min_ps_256(<8 x float> %a0, <8 x float> %a1, i8 %mask) {
; X86-LABEL: test_mm512_maskz_min_ps_256:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vminps %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7c,0xa9,0x5d,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mm512_maskz_min_ps_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vminps %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7c,0xa9,0x5d,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask.min.ps.256(<8 x float> %a0, <8 x float> %a1, <8 x float>zeroinitializer, i8 %mask)
  ret <8 x float> %res
}

define <8 x float> @test_mm512_mask_min_ps_256(<8 x float> %a0, <8 x float> %a1, <8 x float> %src, i8 %mask) {
; X86-LABEL: test_mm512_mask_min_ps_256:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vminps %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7c,0x29,0x5d,0xd1]
; X86-NEXT:    vmovaps %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x28,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mm512_mask_min_ps_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vminps %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7c,0x29,0x5d,0xd1]
; X64-NEXT:    vmovaps %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x28,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask.min.ps.256(<8 x float> %a0, <8 x float> %a1, <8 x float> %src, i8 %mask)
  ret <8 x float> %res
}

define <8 x float> @test_mm512_min_ps_256(<8 x float> %a0, <8 x float> %a1, i8 %mask) {
; CHECK-LABEL: test_mm512_min_ps_256:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vminps %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x5d,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask.min.ps.256(<8 x float> %a0, <8 x float> %a1, <8 x float>zeroinitializer, i8 -1)
  ret <8 x float> %res
}
declare <8 x float> @llvm.x86.avx512.mask.min.ps.256(<8 x float>, <8 x float>, <8 x float>, i8)

define <4 x float> @test_mm512_maskz_min_ps_128(<4 x float> %a0, <4 x float> %a1, i8 %mask) {
; X86-LABEL: test_mm512_maskz_min_ps_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vminps %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7c,0x89,0x5d,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mm512_maskz_min_ps_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vminps %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7c,0x89,0x5d,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask.min.ps.128(<4 x float> %a0, <4 x float> %a1, <4 x float>zeroinitializer, i8 %mask)
  ret <4 x float> %res
}

define <4 x float> @test_mm512_mask_min_ps_128(<4 x float> %a0, <4 x float> %a1, <4 x float> %src, i8 %mask) {
; X86-LABEL: test_mm512_mask_min_ps_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vminps %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7c,0x09,0x5d,0xd1]
; X86-NEXT:    vmovaps %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x28,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mm512_mask_min_ps_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vminps %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7c,0x09,0x5d,0xd1]
; X64-NEXT:    vmovaps %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x28,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask.min.ps.128(<4 x float> %a0, <4 x float> %a1, <4 x float> %src, i8 %mask)
  ret <4 x float> %res
}

define <4 x float> @test_mm512_min_ps_128(<4 x float> %a0, <4 x float> %a1, i8 %mask) {
; CHECK-LABEL: test_mm512_min_ps_128:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vminps %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x5d,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask.min.ps.128(<4 x float> %a0, <4 x float> %a1, <4 x float>zeroinitializer, i8 -1)
  ret <4 x float> %res
}
declare <4 x float> @llvm.x86.avx512.mask.min.ps.128(<4 x float>, <4 x float>, <4 x float>, i8)

define <8 x i8> @test_cmp_d_256(<8 x i32> %a0, <8 x i32> %a1) {
; CHECK-LABEL: test_cmp_d_256:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vpcmpeqd %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x76,0xc1]
; CHECK-NEXT:    vpcmpgtd %ymm0, %ymm1, %k1 # encoding: [0x62,0xf1,0x75,0x28,0x66,0xc8]
; CHECK-NEXT:    vpcmpled %ymm1, %ymm0, %k2 # encoding: [0x62,0xf3,0x7d,0x28,0x1f,0xd1,0x02]
; CHECK-NEXT:    vpcmpneqd %ymm1, %ymm0, %k3 # encoding: [0x62,0xf3,0x7d,0x28,0x1f,0xd9,0x04]
; CHECK-NEXT:    vpcmpnltd %ymm1, %ymm0, %k4 # encoding: [0x62,0xf3,0x7d,0x28,0x1f,0xe1,0x05]
; CHECK-NEXT:    vpcmpgtd %ymm1, %ymm0, %k5 # encoding: [0x62,0xf1,0x7d,0x28,0x66,0xe9]
; CHECK-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; CHECK-NEXT:    vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
; CHECK-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
; CHECK-NEXT:    kmovw %k1, %eax # encoding: [0xc5,0xf8,0x93,0xc1]
; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
; CHECK-NEXT:    kmovw %k2, %eax # encoding: [0xc5,0xf8,0x93,0xc2]
; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
; CHECK-NEXT:    kmovw %k3, %eax # encoding: [0xc5,0xf8,0x93,0xc3]
; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
; CHECK-NEXT:    kmovw %k4, %eax # encoding: [0xc5,0xf8,0x93,0xc4]
; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
; CHECK-NEXT:    kmovw %k5, %eax # encoding: [0xc5,0xf8,0x93,0xc5]
; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
; CHECK-NEXT:    movl $255, %eax # encoding: [0xb8,0xff,0x00,0x00,0x00]
; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
; CHECK-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res0 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 0, i8 -1)
  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
  %res1 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 1, i8 -1)
  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
  %res2 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 2, i8 -1)
  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
  %res3 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 3, i8 -1)
  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
  %res4 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 4, i8 -1)
  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
  %res5 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 5, i8 -1)
  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
  %res6 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 6, i8 -1)
  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
  %res7 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %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_d_256(<8 x i32> %a0, <8 x i32> %a1, i8 %mask) {
; X86-LABEL: test_mask_cmp_d_256:
; X86:       # %bb.0:
; X86-NEXT:    movzwl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb7,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpcmpeqd %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x76,0xc1]
; X86-NEXT:    vpcmpgtd %ymm0, %ymm1, %k2 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x66,0xd0]
; X86-NEXT:    vpcmpled %ymm1, %ymm0, %k3 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x1f,0xd9,0x02]
; X86-NEXT:    vpcmpneqd %ymm1, %ymm0, %k4 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x1f,0xe1,0x04]
; X86-NEXT:    vpcmpnltd %ymm1, %ymm0, %k5 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x1f,0xe9,0x05]
; X86-NEXT:    vpcmpgtd %ymm1, %ymm0, %k1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x66,0xc9]
; X86-NEXT:    kmovw %k0, %ecx # encoding: [0xc5,0xf8,0x93,0xc8]
; X86-NEXT:    vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
; X86-NEXT:    vpinsrw $0, %ecx, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc1,0x00]
; X86-NEXT:    kmovw %k2, %ecx # encoding: [0xc5,0xf8,0x93,0xca]
; X86-NEXT:    vpinsrw $1, %ecx, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc1,0x01]
; X86-NEXT:    kmovw %k3, %ecx # encoding: [0xc5,0xf8,0x93,0xcb]
; X86-NEXT:    vpinsrw $2, %ecx, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc1,0x02]
; X86-NEXT:    kmovw %k4, %ecx # encoding: [0xc5,0xf8,0x93,0xcc]
; X86-NEXT:    vpinsrw $4, %ecx, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc1,0x04]
; X86-NEXT:    kmovw %k5, %ecx # encoding: [0xc5,0xf8,0x93,0xcd]
; X86-NEXT:    vpinsrw $5, %ecx, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc1,0x05]
; X86-NEXT:    kmovw %k1, %ecx # encoding: [0xc5,0xf8,0x93,0xc9]
; X86-NEXT:    vpinsrw $6, %ecx, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc1,0x06]
; X86-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_cmp_d_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpcmpeqd %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x76,0xc1]
; X64-NEXT:    vpcmpgtd %ymm0, %ymm1, %k2 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x66,0xd0]
; X64-NEXT:    vpcmpled %ymm1, %ymm0, %k3 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x1f,0xd9,0x02]
; X64-NEXT:    vpcmpneqd %ymm1, %ymm0, %k4 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x1f,0xe1,0x04]
; X64-NEXT:    vpcmpnltd %ymm1, %ymm0, %k5 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x1f,0xe9,0x05]
; X64-NEXT:    vpcmpgtd %ymm1, %ymm0, %k1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x66,0xc9]
; X64-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; X64-NEXT:    vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
; X64-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
; X64-NEXT:    kmovw %k2, %eax # encoding: [0xc5,0xf8,0x93,0xc2]
; X64-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
; X64-NEXT:    kmovw %k3, %eax # encoding: [0xc5,0xf8,0x93,0xc3]
; X64-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
; X64-NEXT:    kmovw %k4, %eax # encoding: [0xc5,0xf8,0x93,0xc4]
; X64-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
; X64-NEXT:    kmovw %k5, %eax # encoding: [0xc5,0xf8,0x93,0xc5]
; X64-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
; X64-NEXT:    kmovw %k1, %eax # encoding: [0xc5,0xf8,0x93,0xc1]
; X64-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
; X64-NEXT:    vpinsrw $7, %edi, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc7,0x07]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
  %res0 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 0, i8 %mask)
  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
  %res1 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 1, i8 %mask)
  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
  %res2 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 2, i8 %mask)
  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
  %res3 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 3, i8 %mask)
  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
  %res4 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 4, i8 %mask)
  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
  %res5 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 5, i8 %mask)
  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
  %res6 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 6, i8 %mask)
  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
  %res7 = call i8 @llvm.x86.avx512.mask.cmp.d.256(<8 x i32> %a0, <8 x i32> %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.d.256(<8 x i32>, <8 x i32>, i32, i8) nounwind readnone

define <8 x i8> @test_ucmp_d_256(<8 x i32> %a0, <8 x i32> %a1) {
; CHECK-LABEL: test_ucmp_d_256:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vpcmpeqd %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x76,0xc1]
; CHECK-NEXT:    vpcmpltud %ymm1, %ymm0, %k1 # encoding: [0x62,0xf3,0x7d,0x28,0x1e,0xc9,0x01]
; CHECK-NEXT:    vpcmpleud %ymm1, %ymm0, %k2 # encoding: [0x62,0xf3,0x7d,0x28,0x1e,0xd1,0x02]
; CHECK-NEXT:    vpcmpneqd %ymm1, %ymm0, %k3 # encoding: [0x62,0xf3,0x7d,0x28,0x1f,0xd9,0x04]
; CHECK-NEXT:    vpcmpnltud %ymm1, %ymm0, %k4 # encoding: [0x62,0xf3,0x7d,0x28,0x1e,0xe1,0x05]
; CHECK-NEXT:    vpcmpnleud %ymm1, %ymm0, %k5 # encoding: [0x62,0xf3,0x7d,0x28,0x1e,0xe9,0x06]
; CHECK-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; CHECK-NEXT:    vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
; CHECK-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
; CHECK-NEXT:    kmovw %k1, %eax # encoding: [0xc5,0xf8,0x93,0xc1]
; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
; CHECK-NEXT:    kmovw %k2, %eax # encoding: [0xc5,0xf8,0x93,0xc2]
; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
; CHECK-NEXT:    kmovw %k3, %eax # encoding: [0xc5,0xf8,0x93,0xc3]
; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
; CHECK-NEXT:    kmovw %k4, %eax # encoding: [0xc5,0xf8,0x93,0xc4]
; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
; CHECK-NEXT:    kmovw %k5, %eax # encoding: [0xc5,0xf8,0x93,0xc5]
; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
; CHECK-NEXT:    movl $255, %eax # encoding: [0xb8,0xff,0x00,0x00,0x00]
; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
; CHECK-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res0 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 0, i8 -1)
  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
  %res1 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 1, i8 -1)
  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
  %res2 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 2, i8 -1)
  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
  %res3 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 3, i8 -1)
  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
  %res4 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 4, i8 -1)
  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
  %res5 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 5, i8 -1)
  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
  %res6 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 6, i8 -1)
  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
  %res7 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %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_d_256(<8 x i32> %a0, <8 x i32> %a1, i8 %mask) {
; X86-LABEL: test_mask_ucmp_d_256:
; X86:       # %bb.0:
; X86-NEXT:    movzwl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb7,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpcmpeqd %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x76,0xc1]
; X86-NEXT:    vpcmpltud %ymm1, %ymm0, %k2 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x1e,0xd1,0x01]
; X86-NEXT:    vpcmpleud %ymm1, %ymm0, %k3 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x1e,0xd9,0x02]
; X86-NEXT:    vpcmpneqd %ymm1, %ymm0, %k4 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x1f,0xe1,0x04]
; X86-NEXT:    vpcmpnltud %ymm1, %ymm0, %k5 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x1e,0xe9,0x05]
; X86-NEXT:    vpcmpnleud %ymm1, %ymm0, %k1 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x1e,0xc9,0x06]
; X86-NEXT:    kmovw %k0, %ecx # encoding: [0xc5,0xf8,0x93,0xc8]
; X86-NEXT:    vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
; X86-NEXT:    vpinsrw $0, %ecx, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc1,0x00]
; X86-NEXT:    kmovw %k2, %ecx # encoding: [0xc5,0xf8,0x93,0xca]
; X86-NEXT:    vpinsrw $1, %ecx, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc1,0x01]
; X86-NEXT:    kmovw %k3, %ecx # encoding: [0xc5,0xf8,0x93,0xcb]
; X86-NEXT:    vpinsrw $2, %ecx, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc1,0x02]
; X86-NEXT:    kmovw %k4, %ecx # encoding: [0xc5,0xf8,0x93,0xcc]
; X86-NEXT:    vpinsrw $4, %ecx, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc1,0x04]
; X86-NEXT:    kmovw %k5, %ecx # encoding: [0xc5,0xf8,0x93,0xcd]
; X86-NEXT:    vpinsrw $5, %ecx, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc1,0x05]
; X86-NEXT:    kmovw %k1, %ecx # encoding: [0xc5,0xf8,0x93,0xc9]
; X86-NEXT:    vpinsrw $6, %ecx, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc1,0x06]
; X86-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_ucmp_d_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpcmpeqd %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x76,0xc1]
; X64-NEXT:    vpcmpltud %ymm1, %ymm0, %k2 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x1e,0xd1,0x01]
; X64-NEXT:    vpcmpleud %ymm1, %ymm0, %k3 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x1e,0xd9,0x02]
; X64-NEXT:    vpcmpneqd %ymm1, %ymm0, %k4 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x1f,0xe1,0x04]
; X64-NEXT:    vpcmpnltud %ymm1, %ymm0, %k5 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x1e,0xe9,0x05]
; X64-NEXT:    vpcmpnleud %ymm1, %ymm0, %k1 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x1e,0xc9,0x06]
; X64-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; X64-NEXT:    vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
; X64-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
; X64-NEXT:    kmovw %k2, %eax # encoding: [0xc5,0xf8,0x93,0xc2]
; X64-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
; X64-NEXT:    kmovw %k3, %eax # encoding: [0xc5,0xf8,0x93,0xc3]
; X64-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
; X64-NEXT:    kmovw %k4, %eax # encoding: [0xc5,0xf8,0x93,0xc4]
; X64-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
; X64-NEXT:    kmovw %k5, %eax # encoding: [0xc5,0xf8,0x93,0xc5]
; X64-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
; X64-NEXT:    kmovw %k1, %eax # encoding: [0xc5,0xf8,0x93,0xc1]
; X64-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
; X64-NEXT:    vpinsrw $7, %edi, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc7,0x07]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
  %res0 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 0, i8 %mask)
  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
  %res1 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 1, i8 %mask)
  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
  %res2 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 2, i8 %mask)
  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
  %res3 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 3, i8 %mask)
  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
  %res4 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 4, i8 %mask)
  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
  %res5 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 5, i8 %mask)
  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
  %res6 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %a1, i32 6, i8 %mask)
  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
  %res7 = call i8 @llvm.x86.avx512.mask.ucmp.d.256(<8 x i32> %a0, <8 x i32> %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.d.256(<8 x i32>, <8 x i32>, i32, i8) nounwind readnone

define <8 x i8> @test_cmp_q_256(<4 x i64> %a0, <4 x i64> %a1) {
; CHECK-LABEL: test_cmp_q_256:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vpcmpeqq %ymm1, %ymm0, %k0 # encoding: [0x62,0xf2,0xfd,0x28,0x29,0xc1]
; CHECK-NEXT:    vpcmpgtq %ymm0, %ymm1, %k1 # encoding: [0x62,0xf2,0xf5,0x28,0x37,0xc8]
; CHECK-NEXT:    vpcmpleq %ymm1, %ymm0, %k2 # encoding: [0x62,0xf3,0xfd,0x28,0x1f,0xd1,0x02]
; CHECK-NEXT:    vpcmpneqq %ymm1, %ymm0, %k3 # encoding: [0x62,0xf3,0xfd,0x28,0x1f,0xd9,0x04]
; CHECK-NEXT:    vpcmpnltq %ymm1, %ymm0, %k4 # encoding: [0x62,0xf3,0xfd,0x28,0x1f,0xe1,0x05]
; CHECK-NEXT:    vpcmpgtq %ymm1, %ymm0, %k5 # encoding: [0x62,0xf2,0xfd,0x28,0x37,0xe9]
; CHECK-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; CHECK-NEXT:    vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
; CHECK-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
; CHECK-NEXT:    kmovw %k1, %eax # encoding: [0xc5,0xf8,0x93,0xc1]
; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
; CHECK-NEXT:    kmovw %k2, %eax # encoding: [0xc5,0xf8,0x93,0xc2]
; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
; CHECK-NEXT:    kmovw %k3, %eax # encoding: [0xc5,0xf8,0x93,0xc3]
; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
; CHECK-NEXT:    kmovw %k4, %eax # encoding: [0xc5,0xf8,0x93,0xc4]
; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
; CHECK-NEXT:    kmovw %k5, %eax # encoding: [0xc5,0xf8,0x93,0xc5]
; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
; CHECK-NEXT:    movl $15, %eax # encoding: [0xb8,0x0f,0x00,0x00,0x00]
; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
; CHECK-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res0 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 0, i8 -1)
  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
  %res1 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 1, i8 -1)
  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
  %res2 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 2, i8 -1)
  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
  %res3 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 3, i8 -1)
  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
  %res4 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 4, i8 -1)
  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
  %res5 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 5, i8 -1)
  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
  %res6 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 6, i8 -1)
  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
  %res7 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %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_q_256(<4 x i64> %a0, <4 x i64> %a1, i8 %mask) {
; X86-LABEL: test_mask_cmp_q_256:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k2 # encoding: [0xc5,0xf8,0x92,0xd0]
; X86-NEXT:    vpcmpeqq %ymm1, %ymm0, %k0 {%k2} # encoding: [0x62,0xf2,0xfd,0x2a,0x29,0xc1]
; X86-NEXT:    vpcmpgtq %ymm0, %ymm1, %k1 {%k2} # encoding: [0x62,0xf2,0xf5,0x2a,0x37,0xc8]
; X86-NEXT:    vpcmpleq %ymm1, %ymm0, %k3 {%k2} # encoding: [0x62,0xf3,0xfd,0x2a,0x1f,0xd9,0x02]
; X86-NEXT:    vpcmpneqq %ymm1, %ymm0, %k4 {%k2} # encoding: [0x62,0xf3,0xfd,0x2a,0x1f,0xe1,0x04]
; X86-NEXT:    vpcmpnltq %ymm1, %ymm0, %k5 {%k2} # encoding: [0x62,0xf3,0xfd,0x2a,0x1f,0xe9,0x05]
; X86-NEXT:    vpcmpgtq %ymm1, %ymm0, %k6 {%k2} # encoding: [0x62,0xf2,0xfd,0x2a,0x37,0xf1]
; X86-NEXT:    kshiftlw $12, %k2, %k2 # encoding: [0xc4,0xe3,0xf9,0x32,0xd2,0x0c]
; X86-NEXT:    kshiftrw $12, %k2, %k2 # encoding: [0xc4,0xe3,0xf9,0x30,0xd2,0x0c]
; X86-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; X86-NEXT:    vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
; X86-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
; X86-NEXT:    kmovw %k1, %eax # encoding: [0xc5,0xf8,0x93,0xc1]
; X86-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
; X86-NEXT:    kmovw %k3, %eax # encoding: [0xc5,0xf8,0x93,0xc3]
; X86-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
; X86-NEXT:    kmovw %k4, %eax # encoding: [0xc5,0xf8,0x93,0xc4]
; X86-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
; X86-NEXT:    kmovw %k5, %eax # encoding: [0xc5,0xf8,0x93,0xc5]
; X86-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
; X86-NEXT:    kmovw %k6, %eax # encoding: [0xc5,0xf8,0x93,0xc6]
; X86-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
; X86-NEXT:    kmovw %k2, %eax # encoding: [0xc5,0xf8,0x93,0xc2]
; X86-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_cmp_q_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k2 # encoding: [0xc5,0xf8,0x92,0xd7]
; X64-NEXT:    vpcmpeqq %ymm1, %ymm0, %k0 {%k2} # encoding: [0x62,0xf2,0xfd,0x2a,0x29,0xc1]
; X64-NEXT:    vpcmpgtq %ymm0, %ymm1, %k1 {%k2} # encoding: [0x62,0xf2,0xf5,0x2a,0x37,0xc8]
; X64-NEXT:    vpcmpleq %ymm1, %ymm0, %k3 {%k2} # encoding: [0x62,0xf3,0xfd,0x2a,0x1f,0xd9,0x02]
; X64-NEXT:    vpcmpneqq %ymm1, %ymm0, %k4 {%k2} # encoding: [0x62,0xf3,0xfd,0x2a,0x1f,0xe1,0x04]
; X64-NEXT:    vpcmpnltq %ymm1, %ymm0, %k5 {%k2} # encoding: [0x62,0xf3,0xfd,0x2a,0x1f,0xe9,0x05]
; X64-NEXT:    vpcmpgtq %ymm1, %ymm0, %k6 {%k2} # encoding: [0x62,0xf2,0xfd,0x2a,0x37,0xf1]
; X64-NEXT:    kshiftlw $12, %k2, %k2 # encoding: [0xc4,0xe3,0xf9,0x32,0xd2,0x0c]
; X64-NEXT:    kshiftrw $12, %k2, %k2 # encoding: [0xc4,0xe3,0xf9,0x30,0xd2,0x0c]
; X64-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; X64-NEXT:    vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
; X64-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
; X64-NEXT:    kmovw %k1, %eax # encoding: [0xc5,0xf8,0x93,0xc1]
; X64-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
; X64-NEXT:    kmovw %k3, %eax # encoding: [0xc5,0xf8,0x93,0xc3]
; X64-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
; X64-NEXT:    kmovw %k4, %eax # encoding: [0xc5,0xf8,0x93,0xc4]
; X64-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
; X64-NEXT:    kmovw %k5, %eax # encoding: [0xc5,0xf8,0x93,0xc5]
; X64-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
; X64-NEXT:    kmovw %k6, %eax # encoding: [0xc5,0xf8,0x93,0xc6]
; X64-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
; X64-NEXT:    kmovw %k2, %eax # encoding: [0xc5,0xf8,0x93,0xc2]
; X64-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
  %res0 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 0, i8 %mask)
  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
  %res1 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 1, i8 %mask)
  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
  %res2 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 2, i8 %mask)
  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
  %res3 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 3, i8 %mask)
  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
  %res4 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 4, i8 %mask)
  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
  %res5 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 5, i8 %mask)
  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
  %res6 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 6, i8 %mask)
  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
  %res7 = call i8 @llvm.x86.avx512.mask.cmp.q.256(<4 x i64> %a0, <4 x i64> %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.q.256(<4 x i64>, <4 x i64>, i32, i8) nounwind readnone

define <8 x i8> @test_ucmp_q_256(<4 x i64> %a0, <4 x i64> %a1) {
; CHECK-LABEL: test_ucmp_q_256:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vpcmpeqq %ymm1, %ymm0, %k0 # encoding: [0x62,0xf2,0xfd,0x28,0x29,0xc1]
; CHECK-NEXT:    vpcmpltuq %ymm1, %ymm0, %k1 # encoding: [0x62,0xf3,0xfd,0x28,0x1e,0xc9,0x01]
; CHECK-NEXT:    vpcmpleuq %ymm1, %ymm0, %k2 # encoding: [0x62,0xf3,0xfd,0x28,0x1e,0xd1,0x02]
; CHECK-NEXT:    vpcmpneqq %ymm1, %ymm0, %k3 # encoding: [0x62,0xf3,0xfd,0x28,0x1f,0xd9,0x04]
; CHECK-NEXT:    vpcmpnltuq %ymm1, %ymm0, %k4 # encoding: [0x62,0xf3,0xfd,0x28,0x1e,0xe1,0x05]
; CHECK-NEXT:    vpcmpnleuq %ymm1, %ymm0, %k5 # encoding: [0x62,0xf3,0xfd,0x28,0x1e,0xe9,0x06]
; CHECK-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; CHECK-NEXT:    vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
; CHECK-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
; CHECK-NEXT:    kmovw %k1, %eax # encoding: [0xc5,0xf8,0x93,0xc1]
; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
; CHECK-NEXT:    kmovw %k2, %eax # encoding: [0xc5,0xf8,0x93,0xc2]
; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
; CHECK-NEXT:    kmovw %k3, %eax # encoding: [0xc5,0xf8,0x93,0xc3]
; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
; CHECK-NEXT:    kmovw %k4, %eax # encoding: [0xc5,0xf8,0x93,0xc4]
; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
; CHECK-NEXT:    kmovw %k5, %eax # encoding: [0xc5,0xf8,0x93,0xc5]
; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
; CHECK-NEXT:    movl $15, %eax # encoding: [0xb8,0x0f,0x00,0x00,0x00]
; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
; CHECK-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res0 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 0, i8 -1)
  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
  %res1 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 1, i8 -1)
  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
  %res2 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 2, i8 -1)
  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
  %res3 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 3, i8 -1)
  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
  %res4 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 4, i8 -1)
  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
  %res5 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 5, i8 -1)
  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
  %res6 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 6, i8 -1)
  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
  %res7 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %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_q_256(<4 x i64> %a0, <4 x i64> %a1, i8 %mask) {
; X86-LABEL: test_mask_ucmp_q_256:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k2 # encoding: [0xc5,0xf8,0x92,0xd0]
; X86-NEXT:    vpcmpeqq %ymm1, %ymm0, %k0 {%k2} # encoding: [0x62,0xf2,0xfd,0x2a,0x29,0xc1]
; X86-NEXT:    vpcmpltuq %ymm1, %ymm0, %k1 {%k2} # encoding: [0x62,0xf3,0xfd,0x2a,0x1e,0xc9,0x01]
; X86-NEXT:    vpcmpleuq %ymm1, %ymm0, %k3 {%k2} # encoding: [0x62,0xf3,0xfd,0x2a,0x1e,0xd9,0x02]
; X86-NEXT:    vpcmpneqq %ymm1, %ymm0, %k4 {%k2} # encoding: [0x62,0xf3,0xfd,0x2a,0x1f,0xe1,0x04]
; X86-NEXT:    vpcmpnltuq %ymm1, %ymm0, %k5 {%k2} # encoding: [0x62,0xf3,0xfd,0x2a,0x1e,0xe9,0x05]
; X86-NEXT:    vpcmpnleuq %ymm1, %ymm0, %k6 {%k2} # encoding: [0x62,0xf3,0xfd,0x2a,0x1e,0xf1,0x06]
; X86-NEXT:    kshiftlw $12, %k2, %k2 # encoding: [0xc4,0xe3,0xf9,0x32,0xd2,0x0c]
; X86-NEXT:    kshiftrw $12, %k2, %k2 # encoding: [0xc4,0xe3,0xf9,0x30,0xd2,0x0c]
; X86-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; X86-NEXT:    vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
; X86-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
; X86-NEXT:    kmovw %k1, %eax # encoding: [0xc5,0xf8,0x93,0xc1]
; X86-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
; X86-NEXT:    kmovw %k3, %eax # encoding: [0xc5,0xf8,0x93,0xc3]
; X86-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
; X86-NEXT:    kmovw %k4, %eax # encoding: [0xc5,0xf8,0x93,0xc4]
; X86-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
; X86-NEXT:    kmovw %k5, %eax # encoding: [0xc5,0xf8,0x93,0xc5]
; X86-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
; X86-NEXT:    kmovw %k6, %eax # encoding: [0xc5,0xf8,0x93,0xc6]
; X86-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
; X86-NEXT:    kmovw %k2, %eax # encoding: [0xc5,0xf8,0x93,0xc2]
; X86-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_ucmp_q_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k2 # encoding: [0xc5,0xf8,0x92,0xd7]
; X64-NEXT:    vpcmpeqq %ymm1, %ymm0, %k0 {%k2} # encoding: [0x62,0xf2,0xfd,0x2a,0x29,0xc1]
; X64-NEXT:    vpcmpltuq %ymm1, %ymm0, %k1 {%k2} # encoding: [0x62,0xf3,0xfd,0x2a,0x1e,0xc9,0x01]
; X64-NEXT:    vpcmpleuq %ymm1, %ymm0, %k3 {%k2} # encoding: [0x62,0xf3,0xfd,0x2a,0x1e,0xd9,0x02]
; X64-NEXT:    vpcmpneqq %ymm1, %ymm0, %k4 {%k2} # encoding: [0x62,0xf3,0xfd,0x2a,0x1f,0xe1,0x04]
; X64-NEXT:    vpcmpnltuq %ymm1, %ymm0, %k5 {%k2} # encoding: [0x62,0xf3,0xfd,0x2a,0x1e,0xe9,0x05]
; X64-NEXT:    vpcmpnleuq %ymm1, %ymm0, %k6 {%k2} # encoding: [0x62,0xf3,0xfd,0x2a,0x1e,0xf1,0x06]
; X64-NEXT:    kshiftlw $12, %k2, %k2 # encoding: [0xc4,0xe3,0xf9,0x32,0xd2,0x0c]
; X64-NEXT:    kshiftrw $12, %k2, %k2 # encoding: [0xc4,0xe3,0xf9,0x30,0xd2,0x0c]
; X64-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; X64-NEXT:    vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
; X64-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
; X64-NEXT:    kmovw %k1, %eax # encoding: [0xc5,0xf8,0x93,0xc1]
; X64-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
; X64-NEXT:    kmovw %k3, %eax # encoding: [0xc5,0xf8,0x93,0xc3]
; X64-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
; X64-NEXT:    kmovw %k4, %eax # encoding: [0xc5,0xf8,0x93,0xc4]
; X64-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
; X64-NEXT:    kmovw %k5, %eax # encoding: [0xc5,0xf8,0x93,0xc5]
; X64-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
; X64-NEXT:    kmovw %k6, %eax # encoding: [0xc5,0xf8,0x93,0xc6]
; X64-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
; X64-NEXT:    kmovw %k2, %eax # encoding: [0xc5,0xf8,0x93,0xc2]
; X64-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
  %res0 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 0, i8 %mask)
  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
  %res1 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 1, i8 %mask)
  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
  %res2 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 2, i8 %mask)
  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
  %res3 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 3, i8 %mask)
  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
  %res4 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 4, i8 %mask)
  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
  %res5 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 5, i8 %mask)
  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
  %res6 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %a1, i32 6, i8 %mask)
  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
  %res7 = call i8 @llvm.x86.avx512.mask.ucmp.q.256(<4 x i64> %a0, <4 x i64> %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.q.256(<4 x i64>, <4 x i64>, i32, i8) nounwind readnone

define <8 x i8> @test_cmp_d_128(<4 x i32> %a0, <4 x i32> %a1) {
; CHECK-LABEL: test_cmp_d_128:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vpcmpeqd %xmm1, %xmm0, %k0 # encoding: [0x62,0xf1,0x7d,0x08,0x76,0xc1]
; CHECK-NEXT:    vpcmpgtd %xmm0, %xmm1, %k1 # encoding: [0x62,0xf1,0x75,0x08,0x66,0xc8]
; CHECK-NEXT:    vpcmpled %xmm1, %xmm0, %k2 # encoding: [0x62,0xf3,0x7d,0x08,0x1f,0xd1,0x02]
; CHECK-NEXT:    vpcmpneqd %xmm1, %xmm0, %k3 # encoding: [0x62,0xf3,0x7d,0x08,0x1f,0xd9,0x04]
; CHECK-NEXT:    vpcmpnltd %xmm1, %xmm0, %k4 # encoding: [0x62,0xf3,0x7d,0x08,0x1f,0xe1,0x05]
; CHECK-NEXT:    vpcmpgtd %xmm1, %xmm0, %k5 # encoding: [0x62,0xf1,0x7d,0x08,0x66,0xe9]
; CHECK-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; CHECK-NEXT:    vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
; CHECK-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
; CHECK-NEXT:    kmovw %k1, %eax # encoding: [0xc5,0xf8,0x93,0xc1]
; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
; CHECK-NEXT:    kmovw %k2, %eax # encoding: [0xc5,0xf8,0x93,0xc2]
; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
; CHECK-NEXT:    kmovw %k3, %eax # encoding: [0xc5,0xf8,0x93,0xc3]
; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
; CHECK-NEXT:    kmovw %k4, %eax # encoding: [0xc5,0xf8,0x93,0xc4]
; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
; CHECK-NEXT:    kmovw %k5, %eax # encoding: [0xc5,0xf8,0x93,0xc5]
; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
; CHECK-NEXT:    movl $15, %eax # encoding: [0xb8,0x0f,0x00,0x00,0x00]
; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res0 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 0, i8 -1)
  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
  %res1 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 1, i8 -1)
  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
  %res2 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 2, i8 -1)
  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
  %res3 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 3, i8 -1)
  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
  %res4 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 4, i8 -1)
  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
  %res5 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 5, i8 -1)
  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
  %res6 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 6, i8 -1)
  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
  %res7 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %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_d_128(<4 x i32> %a0, <4 x i32> %a1, i8 %mask) {
; X86-LABEL: test_mask_cmp_d_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k2 # encoding: [0xc5,0xf8,0x92,0xd0]
; X86-NEXT:    vpcmpeqd %xmm1, %xmm0, %k0 {%k2} # encoding: [0x62,0xf1,0x7d,0x0a,0x76,0xc1]
; X86-NEXT:    vpcmpgtd %xmm0, %xmm1, %k1 {%k2} # encoding: [0x62,0xf1,0x75,0x0a,0x66,0xc8]
; X86-NEXT:    vpcmpled %xmm1, %xmm0, %k3 {%k2} # encoding: [0x62,0xf3,0x7d,0x0a,0x1f,0xd9,0x02]
; X86-NEXT:    vpcmpneqd %xmm1, %xmm0, %k4 {%k2} # encoding: [0x62,0xf3,0x7d,0x0a,0x1f,0xe1,0x04]
; X86-NEXT:    vpcmpnltd %xmm1, %xmm0, %k5 {%k2} # encoding: [0x62,0xf3,0x7d,0x0a,0x1f,0xe9,0x05]
; X86-NEXT:    vpcmpgtd %xmm1, %xmm0, %k6 {%k2} # encoding: [0x62,0xf1,0x7d,0x0a,0x66,0xf1]
; X86-NEXT:    kshiftlw $12, %k2, %k2 # encoding: [0xc4,0xe3,0xf9,0x32,0xd2,0x0c]
; X86-NEXT:    kshiftrw $12, %k2, %k2 # encoding: [0xc4,0xe3,0xf9,0x30,0xd2,0x0c]
; X86-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; X86-NEXT:    vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
; X86-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
; X86-NEXT:    kmovw %k1, %eax # encoding: [0xc5,0xf8,0x93,0xc1]
; X86-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
; X86-NEXT:    kmovw %k3, %eax # encoding: [0xc5,0xf8,0x93,0xc3]
; X86-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
; X86-NEXT:    kmovw %k4, %eax # encoding: [0xc5,0xf8,0x93,0xc4]
; X86-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
; X86-NEXT:    kmovw %k5, %eax # encoding: [0xc5,0xf8,0x93,0xc5]
; X86-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
; X86-NEXT:    kmovw %k6, %eax # encoding: [0xc5,0xf8,0x93,0xc6]
; X86-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
; X86-NEXT:    kmovw %k2, %eax # encoding: [0xc5,0xf8,0x93,0xc2]
; X86-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_cmp_d_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k2 # encoding: [0xc5,0xf8,0x92,0xd7]
; X64-NEXT:    vpcmpeqd %xmm1, %xmm0, %k0 {%k2} # encoding: [0x62,0xf1,0x7d,0x0a,0x76,0xc1]
; X64-NEXT:    vpcmpgtd %xmm0, %xmm1, %k1 {%k2} # encoding: [0x62,0xf1,0x75,0x0a,0x66,0xc8]
; X64-NEXT:    vpcmpled %xmm1, %xmm0, %k3 {%k2} # encoding: [0x62,0xf3,0x7d,0x0a,0x1f,0xd9,0x02]
; X64-NEXT:    vpcmpneqd %xmm1, %xmm0, %k4 {%k2} # encoding: [0x62,0xf3,0x7d,0x0a,0x1f,0xe1,0x04]
; X64-NEXT:    vpcmpnltd %xmm1, %xmm0, %k5 {%k2} # encoding: [0x62,0xf3,0x7d,0x0a,0x1f,0xe9,0x05]
; X64-NEXT:    vpcmpgtd %xmm1, %xmm0, %k6 {%k2} # encoding: [0x62,0xf1,0x7d,0x0a,0x66,0xf1]
; X64-NEXT:    kshiftlw $12, %k2, %k2 # encoding: [0xc4,0xe3,0xf9,0x32,0xd2,0x0c]
; X64-NEXT:    kshiftrw $12, %k2, %k2 # encoding: [0xc4,0xe3,0xf9,0x30,0xd2,0x0c]
; X64-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; X64-NEXT:    vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
; X64-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
; X64-NEXT:    kmovw %k1, %eax # encoding: [0xc5,0xf8,0x93,0xc1]
; X64-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
; X64-NEXT:    kmovw %k3, %eax # encoding: [0xc5,0xf8,0x93,0xc3]
; X64-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
; X64-NEXT:    kmovw %k4, %eax # encoding: [0xc5,0xf8,0x93,0xc4]
; X64-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
; X64-NEXT:    kmovw %k5, %eax # encoding: [0xc5,0xf8,0x93,0xc5]
; X64-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
; X64-NEXT:    kmovw %k6, %eax # encoding: [0xc5,0xf8,0x93,0xc6]
; X64-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
; X64-NEXT:    kmovw %k2, %eax # encoding: [0xc5,0xf8,0x93,0xc2]
; X64-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %res0 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 0, i8 %mask)
  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
  %res1 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 1, i8 %mask)
  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
  %res2 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 2, i8 %mask)
  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
  %res3 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 3, i8 %mask)
  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
  %res4 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 4, i8 %mask)
  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
  %res5 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 5, i8 %mask)
  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
  %res6 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 6, i8 %mask)
  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
  %res7 = call i8 @llvm.x86.avx512.mask.cmp.d.128(<4 x i32> %a0, <4 x i32> %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.d.128(<4 x i32>, <4 x i32>, i32, i8) nounwind readnone

define <8 x i8> @test_ucmp_d_128(<4 x i32> %a0, <4 x i32> %a1) {
; CHECK-LABEL: test_ucmp_d_128:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vpcmpeqd %xmm1, %xmm0, %k0 # encoding: [0x62,0xf1,0x7d,0x08,0x76,0xc1]
; CHECK-NEXT:    vpcmpltud %xmm1, %xmm0, %k1 # encoding: [0x62,0xf3,0x7d,0x08,0x1e,0xc9,0x01]
; CHECK-NEXT:    vpcmpleud %xmm1, %xmm0, %k2 # encoding: [0x62,0xf3,0x7d,0x08,0x1e,0xd1,0x02]
; CHECK-NEXT:    vpcmpneqd %xmm1, %xmm0, %k3 # encoding: [0x62,0xf3,0x7d,0x08,0x1f,0xd9,0x04]
; CHECK-NEXT:    vpcmpnltud %xmm1, %xmm0, %k4 # encoding: [0x62,0xf3,0x7d,0x08,0x1e,0xe1,0x05]
; CHECK-NEXT:    vpcmpnleud %xmm1, %xmm0, %k5 # encoding: [0x62,0xf3,0x7d,0x08,0x1e,0xe9,0x06]
; CHECK-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; CHECK-NEXT:    vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
; CHECK-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
; CHECK-NEXT:    kmovw %k1, %eax # encoding: [0xc5,0xf8,0x93,0xc1]
; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
; CHECK-NEXT:    kmovw %k2, %eax # encoding: [0xc5,0xf8,0x93,0xc2]
; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
; CHECK-NEXT:    kmovw %k3, %eax # encoding: [0xc5,0xf8,0x93,0xc3]
; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
; CHECK-NEXT:    kmovw %k4, %eax # encoding: [0xc5,0xf8,0x93,0xc4]
; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
; CHECK-NEXT:    kmovw %k5, %eax # encoding: [0xc5,0xf8,0x93,0xc5]
; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
; CHECK-NEXT:    movl $15, %eax # encoding: [0xb8,0x0f,0x00,0x00,0x00]
; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res0 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 0, i8 -1)
  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
  %res1 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 1, i8 -1)
  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
  %res2 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 2, i8 -1)
  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
  %res3 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 3, i8 -1)
  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
  %res4 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 4, i8 -1)
  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
  %res5 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 5, i8 -1)
  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
  %res6 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 6, i8 -1)
  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
  %res7 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %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_d_128(<4 x i32> %a0, <4 x i32> %a1, i8 %mask) {
; X86-LABEL: test_mask_ucmp_d_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k2 # encoding: [0xc5,0xf8,0x92,0xd0]
; X86-NEXT:    vpcmpeqd %xmm1, %xmm0, %k0 {%k2} # encoding: [0x62,0xf1,0x7d,0x0a,0x76,0xc1]
; X86-NEXT:    vpcmpltud %xmm1, %xmm0, %k1 {%k2} # encoding: [0x62,0xf3,0x7d,0x0a,0x1e,0xc9,0x01]
; X86-NEXT:    vpcmpleud %xmm1, %xmm0, %k3 {%k2} # encoding: [0x62,0xf3,0x7d,0x0a,0x1e,0xd9,0x02]
; X86-NEXT:    vpcmpneqd %xmm1, %xmm0, %k4 {%k2} # encoding: [0x62,0xf3,0x7d,0x0a,0x1f,0xe1,0x04]
; X86-NEXT:    vpcmpnltud %xmm1, %xmm0, %k5 {%k2} # encoding: [0x62,0xf3,0x7d,0x0a,0x1e,0xe9,0x05]
; X86-NEXT:    vpcmpnleud %xmm1, %xmm0, %k6 {%k2} # encoding: [0x62,0xf3,0x7d,0x0a,0x1e,0xf1,0x06]
; X86-NEXT:    kshiftlw $12, %k2, %k2 # encoding: [0xc4,0xe3,0xf9,0x32,0xd2,0x0c]
; X86-NEXT:    kshiftrw $12, %k2, %k2 # encoding: [0xc4,0xe3,0xf9,0x30,0xd2,0x0c]
; X86-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; X86-NEXT:    vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
; X86-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
; X86-NEXT:    kmovw %k1, %eax # encoding: [0xc5,0xf8,0x93,0xc1]
; X86-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
; X86-NEXT:    kmovw %k3, %eax # encoding: [0xc5,0xf8,0x93,0xc3]
; X86-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
; X86-NEXT:    kmovw %k4, %eax # encoding: [0xc5,0xf8,0x93,0xc4]
; X86-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
; X86-NEXT:    kmovw %k5, %eax # encoding: [0xc5,0xf8,0x93,0xc5]
; X86-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
; X86-NEXT:    kmovw %k6, %eax # encoding: [0xc5,0xf8,0x93,0xc6]
; X86-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
; X86-NEXT:    kmovw %k2, %eax # encoding: [0xc5,0xf8,0x93,0xc2]
; X86-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_ucmp_d_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k2 # encoding: [0xc5,0xf8,0x92,0xd7]
; X64-NEXT:    vpcmpeqd %xmm1, %xmm0, %k0 {%k2} # encoding: [0x62,0xf1,0x7d,0x0a,0x76,0xc1]
; X64-NEXT:    vpcmpltud %xmm1, %xmm0, %k1 {%k2} # encoding: [0x62,0xf3,0x7d,0x0a,0x1e,0xc9,0x01]
; X64-NEXT:    vpcmpleud %xmm1, %xmm0, %k3 {%k2} # encoding: [0x62,0xf3,0x7d,0x0a,0x1e,0xd9,0x02]
; X64-NEXT:    vpcmpneqd %xmm1, %xmm0, %k4 {%k2} # encoding: [0x62,0xf3,0x7d,0x0a,0x1f,0xe1,0x04]
; X64-NEXT:    vpcmpnltud %xmm1, %xmm0, %k5 {%k2} # encoding: [0x62,0xf3,0x7d,0x0a,0x1e,0xe9,0x05]
; X64-NEXT:    vpcmpnleud %xmm1, %xmm0, %k6 {%k2} # encoding: [0x62,0xf3,0x7d,0x0a,0x1e,0xf1,0x06]
; X64-NEXT:    kshiftlw $12, %k2, %k2 # encoding: [0xc4,0xe3,0xf9,0x32,0xd2,0x0c]
; X64-NEXT:    kshiftrw $12, %k2, %k2 # encoding: [0xc4,0xe3,0xf9,0x30,0xd2,0x0c]
; X64-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; X64-NEXT:    vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
; X64-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
; X64-NEXT:    kmovw %k1, %eax # encoding: [0xc5,0xf8,0x93,0xc1]
; X64-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
; X64-NEXT:    kmovw %k3, %eax # encoding: [0xc5,0xf8,0x93,0xc3]
; X64-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
; X64-NEXT:    kmovw %k4, %eax # encoding: [0xc5,0xf8,0x93,0xc4]
; X64-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
; X64-NEXT:    kmovw %k5, %eax # encoding: [0xc5,0xf8,0x93,0xc5]
; X64-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
; X64-NEXT:    kmovw %k6, %eax # encoding: [0xc5,0xf8,0x93,0xc6]
; X64-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
; X64-NEXT:    kmovw %k2, %eax # encoding: [0xc5,0xf8,0x93,0xc2]
; X64-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %res0 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 0, i8 %mask)
  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
  %res1 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 1, i8 %mask)
  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
  %res2 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 2, i8 %mask)
  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
  %res3 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 3, i8 %mask)
  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
  %res4 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 4, i8 %mask)
  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
  %res5 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 5, i8 %mask)
  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
  %res6 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %a1, i32 6, i8 %mask)
  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
  %res7 = call i8 @llvm.x86.avx512.mask.ucmp.d.128(<4 x i32> %a0, <4 x i32> %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.d.128(<4 x i32>, <4 x i32>, i32, i8) nounwind readnone

define <8 x i8> @test_cmp_q_128(<2 x i64> %a0, <2 x i64> %a1) {
; CHECK-LABEL: test_cmp_q_128:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vpcmpeqq %xmm1, %xmm0, %k0 # encoding: [0x62,0xf2,0xfd,0x08,0x29,0xc1]
; CHECK-NEXT:    vpcmpgtq %xmm0, %xmm1, %k1 # encoding: [0x62,0xf2,0xf5,0x08,0x37,0xc8]
; CHECK-NEXT:    vpcmpleq %xmm1, %xmm0, %k2 # encoding: [0x62,0xf3,0xfd,0x08,0x1f,0xd1,0x02]
; CHECK-NEXT:    vpcmpneqq %xmm1, %xmm0, %k3 # encoding: [0x62,0xf3,0xfd,0x08,0x1f,0xd9,0x04]
; CHECK-NEXT:    vpcmpnltq %xmm1, %xmm0, %k4 # encoding: [0x62,0xf3,0xfd,0x08,0x1f,0xe1,0x05]
; CHECK-NEXT:    vpcmpgtq %xmm1, %xmm0, %k5 # encoding: [0x62,0xf2,0xfd,0x08,0x37,0xe9]
; CHECK-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; CHECK-NEXT:    vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
; CHECK-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
; CHECK-NEXT:    kmovw %k1, %eax # encoding: [0xc5,0xf8,0x93,0xc1]
; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
; CHECK-NEXT:    kmovw %k2, %eax # encoding: [0xc5,0xf8,0x93,0xc2]
; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
; CHECK-NEXT:    kmovw %k3, %eax # encoding: [0xc5,0xf8,0x93,0xc3]
; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
; CHECK-NEXT:    kmovw %k4, %eax # encoding: [0xc5,0xf8,0x93,0xc4]
; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
; CHECK-NEXT:    kmovw %k5, %eax # encoding: [0xc5,0xf8,0x93,0xc5]
; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
; CHECK-NEXT:    movl $3, %eax # encoding: [0xb8,0x03,0x00,0x00,0x00]
; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res0 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 0, i8 -1)
  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
  %res1 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 1, i8 -1)
  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
  %res2 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 2, i8 -1)
  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
  %res3 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 3, i8 -1)
  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
  %res4 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 4, i8 -1)
  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
  %res5 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 5, i8 -1)
  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
  %res6 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 6, i8 -1)
  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
  %res7 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %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_q_128(<2 x i64> %a0, <2 x i64> %a1, i8 %mask) {
; X86-LABEL: test_mask_cmp_q_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k2 # encoding: [0xc5,0xf8,0x92,0xd0]
; X86-NEXT:    vpcmpeqq %xmm1, %xmm0, %k0 {%k2} # encoding: [0x62,0xf2,0xfd,0x0a,0x29,0xc1]
; X86-NEXT:    vpcmpgtq %xmm0, %xmm1, %k1 {%k2} # encoding: [0x62,0xf2,0xf5,0x0a,0x37,0xc8]
; X86-NEXT:    vpcmpleq %xmm1, %xmm0, %k3 {%k2} # encoding: [0x62,0xf3,0xfd,0x0a,0x1f,0xd9,0x02]
; X86-NEXT:    vpcmpneqq %xmm1, %xmm0, %k4 {%k2} # encoding: [0x62,0xf3,0xfd,0x0a,0x1f,0xe1,0x04]
; X86-NEXT:    vpcmpnltq %xmm1, %xmm0, %k5 {%k2} # encoding: [0x62,0xf3,0xfd,0x0a,0x1f,0xe9,0x05]
; X86-NEXT:    vpcmpgtq %xmm1, %xmm0, %k6 {%k2} # encoding: [0x62,0xf2,0xfd,0x0a,0x37,0xf1]
; X86-NEXT:    kshiftlw $14, %k2, %k2 # encoding: [0xc4,0xe3,0xf9,0x32,0xd2,0x0e]
; X86-NEXT:    kshiftrw $14, %k2, %k2 # encoding: [0xc4,0xe3,0xf9,0x30,0xd2,0x0e]
; X86-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; X86-NEXT:    vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
; X86-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
; X86-NEXT:    kmovw %k1, %eax # encoding: [0xc5,0xf8,0x93,0xc1]
; X86-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
; X86-NEXT:    kmovw %k3, %eax # encoding: [0xc5,0xf8,0x93,0xc3]
; X86-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
; X86-NEXT:    kmovw %k4, %eax # encoding: [0xc5,0xf8,0x93,0xc4]
; X86-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
; X86-NEXT:    kmovw %k5, %eax # encoding: [0xc5,0xf8,0x93,0xc5]
; X86-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
; X86-NEXT:    kmovw %k6, %eax # encoding: [0xc5,0xf8,0x93,0xc6]
; X86-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
; X86-NEXT:    kmovw %k2, %eax # encoding: [0xc5,0xf8,0x93,0xc2]
; X86-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_cmp_q_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k2 # encoding: [0xc5,0xf8,0x92,0xd7]
; X64-NEXT:    vpcmpeqq %xmm1, %xmm0, %k0 {%k2} # encoding: [0x62,0xf2,0xfd,0x0a,0x29,0xc1]
; X64-NEXT:    vpcmpgtq %xmm0, %xmm1, %k1 {%k2} # encoding: [0x62,0xf2,0xf5,0x0a,0x37,0xc8]
; X64-NEXT:    vpcmpleq %xmm1, %xmm0, %k3 {%k2} # encoding: [0x62,0xf3,0xfd,0x0a,0x1f,0xd9,0x02]
; X64-NEXT:    vpcmpneqq %xmm1, %xmm0, %k4 {%k2} # encoding: [0x62,0xf3,0xfd,0x0a,0x1f,0xe1,0x04]
; X64-NEXT:    vpcmpnltq %xmm1, %xmm0, %k5 {%k2} # encoding: [0x62,0xf3,0xfd,0x0a,0x1f,0xe9,0x05]
; X64-NEXT:    vpcmpgtq %xmm1, %xmm0, %k6 {%k2} # encoding: [0x62,0xf2,0xfd,0x0a,0x37,0xf1]
; X64-NEXT:    kshiftlw $14, %k2, %k2 # encoding: [0xc4,0xe3,0xf9,0x32,0xd2,0x0e]
; X64-NEXT:    kshiftrw $14, %k2, %k2 # encoding: [0xc4,0xe3,0xf9,0x30,0xd2,0x0e]
; X64-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; X64-NEXT:    vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
; X64-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
; X64-NEXT:    kmovw %k1, %eax # encoding: [0xc5,0xf8,0x93,0xc1]
; X64-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
; X64-NEXT:    kmovw %k3, %eax # encoding: [0xc5,0xf8,0x93,0xc3]
; X64-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
; X64-NEXT:    kmovw %k4, %eax # encoding: [0xc5,0xf8,0x93,0xc4]
; X64-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
; X64-NEXT:    kmovw %k5, %eax # encoding: [0xc5,0xf8,0x93,0xc5]
; X64-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
; X64-NEXT:    kmovw %k6, %eax # encoding: [0xc5,0xf8,0x93,0xc6]
; X64-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
; X64-NEXT:    kmovw %k2, %eax # encoding: [0xc5,0xf8,0x93,0xc2]
; X64-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %res0 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 0, i8 %mask)
  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
  %res1 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 1, i8 %mask)
  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
  %res2 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 2, i8 %mask)
  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
  %res3 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 3, i8 %mask)
  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
  %res4 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 4, i8 %mask)
  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
  %res5 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 5, i8 %mask)
  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
  %res6 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 6, i8 %mask)
  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
  %res7 = call i8 @llvm.x86.avx512.mask.cmp.q.128(<2 x i64> %a0, <2 x i64> %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.q.128(<2 x i64>, <2 x i64>, i32, i8) nounwind readnone

define <8 x i8> @test_ucmp_q_128(<2 x i64> %a0, <2 x i64> %a1) {
; CHECK-LABEL: test_ucmp_q_128:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vpcmpeqq %xmm1, %xmm0, %k0 # encoding: [0x62,0xf2,0xfd,0x08,0x29,0xc1]
; CHECK-NEXT:    vpcmpltuq %xmm1, %xmm0, %k1 # encoding: [0x62,0xf3,0xfd,0x08,0x1e,0xc9,0x01]
; CHECK-NEXT:    vpcmpleuq %xmm1, %xmm0, %k2 # encoding: [0x62,0xf3,0xfd,0x08,0x1e,0xd1,0x02]
; CHECK-NEXT:    vpcmpneqq %xmm1, %xmm0, %k3 # encoding: [0x62,0xf3,0xfd,0x08,0x1f,0xd9,0x04]
; CHECK-NEXT:    vpcmpnltuq %xmm1, %xmm0, %k4 # encoding: [0x62,0xf3,0xfd,0x08,0x1e,0xe1,0x05]
; CHECK-NEXT:    vpcmpnleuq %xmm1, %xmm0, %k5 # encoding: [0x62,0xf3,0xfd,0x08,0x1e,0xe9,0x06]
; CHECK-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; CHECK-NEXT:    vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
; CHECK-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
; CHECK-NEXT:    kmovw %k1, %eax # encoding: [0xc5,0xf8,0x93,0xc1]
; CHECK-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
; CHECK-NEXT:    kmovw %k2, %eax # encoding: [0xc5,0xf8,0x93,0xc2]
; CHECK-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
; CHECK-NEXT:    kmovw %k3, %eax # encoding: [0xc5,0xf8,0x93,0xc3]
; CHECK-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
; CHECK-NEXT:    kmovw %k4, %eax # encoding: [0xc5,0xf8,0x93,0xc4]
; CHECK-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
; CHECK-NEXT:    kmovw %k5, %eax # encoding: [0xc5,0xf8,0x93,0xc5]
; CHECK-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
; CHECK-NEXT:    movl $3, %eax # encoding: [0xb8,0x03,0x00,0x00,0x00]
; CHECK-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res0 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 0, i8 -1)
  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
  %res1 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 1, i8 -1)
  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
  %res2 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 2, i8 -1)
  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
  %res3 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 3, i8 -1)
  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
  %res4 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 4, i8 -1)
  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
  %res5 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 5, i8 -1)
  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
  %res6 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 6, i8 -1)
  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
  %res7 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %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_q_128(<2 x i64> %a0, <2 x i64> %a1, i8 %mask) {
; X86-LABEL: test_mask_ucmp_q_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k2 # encoding: [0xc5,0xf8,0x92,0xd0]
; X86-NEXT:    vpcmpeqq %xmm1, %xmm0, %k0 {%k2} # encoding: [0x62,0xf2,0xfd,0x0a,0x29,0xc1]
; X86-NEXT:    vpcmpltuq %xmm1, %xmm0, %k1 {%k2} # encoding: [0x62,0xf3,0xfd,0x0a,0x1e,0xc9,0x01]
; X86-NEXT:    vpcmpleuq %xmm1, %xmm0, %k3 {%k2} # encoding: [0x62,0xf3,0xfd,0x0a,0x1e,0xd9,0x02]
; X86-NEXT:    vpcmpneqq %xmm1, %xmm0, %k4 {%k2} # encoding: [0x62,0xf3,0xfd,0x0a,0x1f,0xe1,0x04]
; X86-NEXT:    vpcmpnltuq %xmm1, %xmm0, %k5 {%k2} # encoding: [0x62,0xf3,0xfd,0x0a,0x1e,0xe9,0x05]
; X86-NEXT:    vpcmpnleuq %xmm1, %xmm0, %k6 {%k2} # encoding: [0x62,0xf3,0xfd,0x0a,0x1e,0xf1,0x06]
; X86-NEXT:    kshiftlw $14, %k2, %k2 # encoding: [0xc4,0xe3,0xf9,0x32,0xd2,0x0e]
; X86-NEXT:    kshiftrw $14, %k2, %k2 # encoding: [0xc4,0xe3,0xf9,0x30,0xd2,0x0e]
; X86-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; X86-NEXT:    vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
; X86-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
; X86-NEXT:    kmovw %k1, %eax # encoding: [0xc5,0xf8,0x93,0xc1]
; X86-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
; X86-NEXT:    kmovw %k3, %eax # encoding: [0xc5,0xf8,0x93,0xc3]
; X86-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
; X86-NEXT:    kmovw %k4, %eax # encoding: [0xc5,0xf8,0x93,0xc4]
; X86-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
; X86-NEXT:    kmovw %k5, %eax # encoding: [0xc5,0xf8,0x93,0xc5]
; X86-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
; X86-NEXT:    kmovw %k6, %eax # encoding: [0xc5,0xf8,0x93,0xc6]
; X86-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
; X86-NEXT:    kmovw %k2, %eax # encoding: [0xc5,0xf8,0x93,0xc2]
; X86-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_ucmp_q_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k2 # encoding: [0xc5,0xf8,0x92,0xd7]
; X64-NEXT:    vpcmpeqq %xmm1, %xmm0, %k0 {%k2} # encoding: [0x62,0xf2,0xfd,0x0a,0x29,0xc1]
; X64-NEXT:    vpcmpltuq %xmm1, %xmm0, %k1 {%k2} # encoding: [0x62,0xf3,0xfd,0x0a,0x1e,0xc9,0x01]
; X64-NEXT:    vpcmpleuq %xmm1, %xmm0, %k3 {%k2} # encoding: [0x62,0xf3,0xfd,0x0a,0x1e,0xd9,0x02]
; X64-NEXT:    vpcmpneqq %xmm1, %xmm0, %k4 {%k2} # encoding: [0x62,0xf3,0xfd,0x0a,0x1f,0xe1,0x04]
; X64-NEXT:    vpcmpnltuq %xmm1, %xmm0, %k5 {%k2} # encoding: [0x62,0xf3,0xfd,0x0a,0x1e,0xe9,0x05]
; X64-NEXT:    vpcmpnleuq %xmm1, %xmm0, %k6 {%k2} # encoding: [0x62,0xf3,0xfd,0x0a,0x1e,0xf1,0x06]
; X64-NEXT:    kshiftlw $14, %k2, %k2 # encoding: [0xc4,0xe3,0xf9,0x32,0xd2,0x0e]
; X64-NEXT:    kshiftrw $14, %k2, %k2 # encoding: [0xc4,0xe3,0xf9,0x30,0xd2,0x0e]
; X64-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; X64-NEXT:    vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
; X64-NEXT:    vpinsrw $0, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
; X64-NEXT:    kmovw %k1, %eax # encoding: [0xc5,0xf8,0x93,0xc1]
; X64-NEXT:    vpinsrw $1, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
; X64-NEXT:    kmovw %k3, %eax # encoding: [0xc5,0xf8,0x93,0xc3]
; X64-NEXT:    vpinsrw $2, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
; X64-NEXT:    kmovw %k4, %eax # encoding: [0xc5,0xf8,0x93,0xc4]
; X64-NEXT:    vpinsrw $4, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
; X64-NEXT:    kmovw %k5, %eax # encoding: [0xc5,0xf8,0x93,0xc5]
; X64-NEXT:    vpinsrw $5, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
; X64-NEXT:    kmovw %k6, %eax # encoding: [0xc5,0xf8,0x93,0xc6]
; X64-NEXT:    vpinsrw $6, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
; X64-NEXT:    kmovw %k2, %eax # encoding: [0xc5,0xf8,0x93,0xc2]
; X64-NEXT:    vpinsrw $7, %eax, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %res0 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 0, i8 %mask)
  %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
  %res1 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 1, i8 %mask)
  %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
  %res2 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 2, i8 %mask)
  %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
  %res3 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 3, i8 %mask)
  %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
  %res4 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 4, i8 %mask)
  %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
  %res5 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 5, i8 %mask)
  %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
  %res6 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %a1, i32 6, i8 %mask)
  %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
  %res7 = call i8 @llvm.x86.avx512.mask.ucmp.q.128(<2 x i64> %a0, <2 x i64> %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.q.128(<2 x i64>, <2 x i64>, i32, i8) nounwind readnone

declare <8 x float> @llvm.x86.avx512.mask.broadcastf32x4.256(<4 x float>, <8 x float>, i8)

define <8 x float>@test_int_x86_avx512_mask_broadcastf32x4_256(<4 x float> %x0, <8 x float> %x2, i8 %mask) {
; X86-LABEL: test_int_x86_avx512_mask_broadcastf32x4_256:
; X86:       # %bb.0:
; X86-NEXT:    # kill: def $xmm0 killed $xmm0 def $ymm0
; X86-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x7d,0x18,0xd0,0x01]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vinsertf32x4 $1, %xmm0, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x18,0xc8,0x01]
; X86-NEXT:    vaddps %ymm1, %ymm2, %ymm1 # EVEX TO VEX Compression encoding: [0xc5,0xec,0x58,0xc9]
; X86-NEXT:    vinsertf32x4 $1, %xmm0, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0xa9,0x18,0xc0,0x01]
; X86-NEXT:    vaddps %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x58,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_broadcastf32x4_256:
; X64:       # %bb.0:
; X64-NEXT:    # kill: def $xmm0 killed $xmm0 def $ymm0
; X64-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x7d,0x18,0xd0,0x01]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vinsertf32x4 $1, %xmm0, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x18,0xc8,0x01]
; X64-NEXT:    vaddps %ymm1, %ymm2, %ymm1 # EVEX TO VEX Compression encoding: [0xc5,0xec,0x58,0xc9]
; X64-NEXT:    vinsertf32x4 $1, %xmm0, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0xa9,0x18,0xc0,0x01]
; X64-NEXT:    vaddps %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x58,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %res1 = call <8 x float> @llvm.x86.avx512.mask.broadcastf32x4.256(<4 x float> %x0, <8 x float> %x2, i8 -1)
  %res2 = call <8 x float> @llvm.x86.avx512.mask.broadcastf32x4.256(<4 x float> %x0, <8 x float> %x2, i8 %mask)
  %res3 = call <8 x float> @llvm.x86.avx512.mask.broadcastf32x4.256(<4 x float> %x0, <8 x float> zeroinitializer, i8 %mask)
  %res4 = fadd <8 x float> %res1, %res2
  %res5 = fadd <8 x float> %res3, %res4
  ret <8 x float> %res5
}

define <8 x float>@test_int_x86_avx512_mask_broadcastf32x4_256_load(<4 x float>* %x0ptr, <8 x float> %x2, i8 %mask) {
; X86-LABEL: test_int_x86_avx512_mask_broadcastf32x4_256_load:
; 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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vbroadcastf32x4 (%eax), %ymm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x1a,0x00]
; X86-NEXT:    # ymm0 {%k1} = mem[0,1,2,3,0,1,2,3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_broadcastf32x4_256_load:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vbroadcastf32x4 (%rdi), %ymm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x1a,0x07]
; X64-NEXT:    # ymm0 {%k1} = mem[0,1,2,3,0,1,2,3]
; X64-NEXT:    retq # encoding: [0xc3]
  %x0 = load <4 x float>, <4 x float>* %x0ptr
  %res = call <8 x float> @llvm.x86.avx512.mask.broadcastf32x4.256(<4 x float> %x0, <8 x float> %x2, i8 %mask)
  ret <8 x float> %res
}

declare <8 x i32> @llvm.x86.avx512.mask.broadcasti32x4.256(<4 x i32>, <8 x i32>, i8)

define <8 x i32>@test_int_x86_avx512_mask_broadcasti32x4_256(<4 x i32> %x0, <8 x i32> %x2, i8 %mask) {
; X86-LABEL: test_int_x86_avx512_mask_broadcasti32x4_256:
; X86:       # %bb.0:
; X86-NEXT:    # kill: def $xmm0 killed $xmm0 def $ymm0
; X86-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x7d,0x38,0xd0,0x01]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vinserti32x4 $1, %xmm0, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x38,0xc8,0x01]
; X86-NEXT:    vinserti32x4 $1, %xmm0, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0xa9,0x38,0xc0,0x01]
; X86-NEXT:    vpaddd %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfe,0xc0]
; X86-NEXT:    vpaddd %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_broadcasti32x4_256:
; X64:       # %bb.0:
; X64-NEXT:    # kill: def $xmm0 killed $xmm0 def $ymm0
; X64-NEXT:    vinserti128 $1, %xmm0, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x7d,0x38,0xd0,0x01]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vinserti32x4 $1, %xmm0, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x38,0xc8,0x01]
; X64-NEXT:    vinserti32x4 $1, %xmm0, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0xa9,0x38,0xc0,0x01]
; X64-NEXT:    vpaddd %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfe,0xc0]
; X64-NEXT:    vpaddd %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res1 = call <8 x i32> @llvm.x86.avx512.mask.broadcasti32x4.256(<4 x i32> %x0, <8 x i32> %x2, i8 -1)
  %res2 = call <8 x i32> @llvm.x86.avx512.mask.broadcasti32x4.256(<4 x i32> %x0, <8 x i32> %x2, i8 %mask)
  %res3 = call <8 x i32> @llvm.x86.avx512.mask.broadcasti32x4.256(<4 x i32> %x0, <8 x i32> zeroinitializer, i8 %mask)
  %res4 = add <8 x i32> %res1, %res2
  %res5 = add <8 x i32> %res3, %res4
  ret <8 x i32> %res5
}

define <8 x i32>@test_int_x86_avx512_mask_broadcasti32x4_256_load(<4 x i32>* %x0ptr, <8 x i32> %x2, i8 %mask) {
; X86-LABEL: test_int_x86_avx512_mask_broadcasti32x4_256_load:
; 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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vbroadcasti32x4 (%eax), %ymm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x5a,0x00]
; X86-NEXT:    # ymm0 {%k1} = mem[0,1,2,3,0,1,2,3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_broadcasti32x4_256_load:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vbroadcasti32x4 (%rdi), %ymm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x5a,0x07]
; X64-NEXT:    # ymm0 {%k1} = mem[0,1,2,3,0,1,2,3]
; X64-NEXT:    retq # encoding: [0xc3]
  %x0 = load <4 x i32>, <4 x i32>* %x0ptr
  %res = call <8 x i32> @llvm.x86.avx512.mask.broadcasti32x4.256(<4 x i32> %x0, <8 x i32> %x2, i8 %mask)
  ret <8 x i32> %res
}

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

define <2 x i64>@test_int_x86_avx512_mask_pabs_q_128(<2 x i64> %x0, <2 x i64> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pabs_q_128:
; X86:       # %bb.0:
; X86-NEXT:    vpabsq %xmm0, %xmm2 # encoding: [0x62,0xf2,0xfd,0x08,0x1f,0xd0]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpabsq %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x1f,0xc8]
; X86-NEXT:    vpaddq %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xd4,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pabs_q_128:
; X64:       # %bb.0:
; X64-NEXT:    vpabsq %xmm0, %xmm2 # encoding: [0x62,0xf2,0xfd,0x08,0x1f,0xd0]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpabsq %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x1f,0xc8]
; X64-NEXT:    vpaddq %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xd4,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x i64> @llvm.x86.avx512.mask.pabs.q.128(<2 x i64> %x0, <2 x i64> %x1, i8 %x2)
  %res1 = call <2 x i64> @llvm.x86.avx512.mask.pabs.q.128(<2 x i64> %x0, <2 x i64> %x1, i8 -1)
  %res2 = add <2 x i64> %res, %res1
  ret <2 x i64> %res2
}

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

define <4 x i64>@test_int_x86_avx512_mask_pabs_q_256(<4 x i64> %x0, <4 x i64> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pabs_q_256:
; X86:       # %bb.0:
; X86-NEXT:    vpabsq %ymm0, %ymm2 # encoding: [0x62,0xf2,0xfd,0x28,0x1f,0xd0]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpabsq %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x1f,0xc8]
; X86-NEXT:    vpaddq %ymm2, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xd4,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pabs_q_256:
; X64:       # %bb.0:
; X64-NEXT:    vpabsq %ymm0, %ymm2 # encoding: [0x62,0xf2,0xfd,0x28,0x1f,0xd0]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpabsq %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x1f,0xc8]
; X64-NEXT:    vpaddq %ymm2, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xd4,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i64> @llvm.x86.avx512.mask.pabs.q.256(<4 x i64> %x0, <4 x i64> %x1, i8 %x2)
  %res1 = call <4 x i64> @llvm.x86.avx512.mask.pabs.q.256(<4 x i64> %x0, <4 x i64> %x1, i8 -1)
  %res2 = add <4 x i64> %res, %res1
  ret <4 x i64> %res2
}

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

define <4 x i32>@test_int_x86_avx512_mask_pabs_d_128(<4 x i32> %x0, <4 x i32> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pabs_d_128:
; X86:       # %bb.0:
; X86-NEXT:    vpabsd %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x1e,0xd0]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpabsd %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x1e,0xc8]
; X86-NEXT:    vpaddd %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pabs_d_128:
; X64:       # %bb.0:
; X64-NEXT:    vpabsd %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x1e,0xd0]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpabsd %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x1e,0xc8]
; X64-NEXT:    vpaddd %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i32> @llvm.x86.avx512.mask.pabs.d.128(<4 x i32> %x0, <4 x i32> %x1, i8 %x2)
  %res1 = call <4 x i32> @llvm.x86.avx512.mask.pabs.d.128(<4 x i32> %x0, <4 x i32> %x1, i8 -1)
  %res2 = add <4 x i32> %res, %res1
  ret <4 x i32> %res2
}

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

define <8 x i32>@test_int_x86_avx512_mask_pabs_d_256(<8 x i32> %x0, <8 x i32> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pabs_d_256:
; X86:       # %bb.0:
; X86-NEXT:    vpabsd %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x1e,0xd0]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpabsd %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x1e,0xc8]
; X86-NEXT:    vpaddd %ymm2, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfe,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pabs_d_256:
; X64:       # %bb.0:
; X64-NEXT:    vpabsd %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x1e,0xd0]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpabsd %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x1e,0xc8]
; X64-NEXT:    vpaddd %ymm2, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfe,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x i32> @llvm.x86.avx512.mask.pabs.d.256(<8 x i32> %x0, <8 x i32> %x1, i8 %x2)
  %res1 = call <8 x i32> @llvm.x86.avx512.mask.pabs.d.256(<8 x i32> %x0, <8 x i32> %x1, i8 -1)
  %res2 = add <8 x i32> %res, %res1
  ret <8 x i32> %res2
}

declare i8 @llvm.x86.avx512.ptestm.d.128(<4 x i32>, <4 x i32>,i8)

define i8@test_int_x86_avx512_ptestm_d_128(<4 x i32> %x0, <4 x i32> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_ptestm_d_128:
; X86:       # %bb.0:
; X86-NEXT:    vptestmd %xmm1, %xmm0, %k0 # encoding: [0x62,0xf2,0x7d,0x08,0x27,0xc1]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vptestmd %xmm1, %xmm0, %k1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x27,0xc9]
; X86-NEXT:    kmovw %k1, %ecx # encoding: [0xc5,0xf8,0x93,0xc9]
; X86-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; X86-NEXT:    addb %cl, %al # encoding: [0x00,0xc8]
; X86-NEXT:    # kill: def $al killed $al killed $eax
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_ptestm_d_128:
; X64:       # %bb.0:
; X64-NEXT:    vptestmd %xmm1, %xmm0, %k0 # encoding: [0x62,0xf2,0x7d,0x08,0x27,0xc1]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vptestmd %xmm1, %xmm0, %k1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x27,0xc9]
; X64-NEXT:    kmovw %k1, %ecx # encoding: [0xc5,0xf8,0x93,0xc9]
; X64-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; X64-NEXT:    addb %cl, %al # encoding: [0x00,0xc8]
; X64-NEXT:    # kill: def $al killed $al killed $eax
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call i8 @llvm.x86.avx512.ptestm.d.128(<4 x i32> %x0, <4 x i32> %x1, i8 %x2)
  %res1 = call i8 @llvm.x86.avx512.ptestm.d.128(<4 x i32> %x0, <4 x i32> %x1, i8-1)
  %res2 = add i8 %res, %res1
  ret i8 %res2
}

declare i8 @llvm.x86.avx512.ptestm.d.256(<8 x i32>, <8 x i32>, i8)

define i8@test_int_x86_avx512_ptestm_d_256(<8 x i32> %x0, <8 x i32> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_ptestm_d_256:
; X86:       # %bb.0:
; X86-NEXT:    vptestmd %ymm1, %ymm0, %k0 # encoding: [0x62,0xf2,0x7d,0x28,0x27,0xc1]
; X86-NEXT:    kmovw %k0, %ecx # encoding: [0xc5,0xf8,0x93,0xc8]
; 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:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_ptestm_d_256:
; X64:       # %bb.0:
; X64-NEXT:    vptestmd %ymm1, %ymm0, %k0 # encoding: [0x62,0xf2,0x7d,0x28,0x27,0xc1]
; X64-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; 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:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call i8 @llvm.x86.avx512.ptestm.d.256(<8 x i32> %x0, <8 x i32> %x1, i8 %x2)
  %res1 = call i8 @llvm.x86.avx512.ptestm.d.256(<8 x i32> %x0, <8 x i32> %x1, i8-1)
  %res2 = add i8 %res, %res1
  ret i8 %res2
}

declare i8 @llvm.x86.avx512.ptestm.q.128(<2 x i64>, <2 x i64>, i8)

define i8@test_int_x86_avx512_ptestm_q_128(<2 x i64> %x0, <2 x i64> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_ptestm_q_128:
; X86:       # %bb.0:
; X86-NEXT:    vptestmq %xmm1, %xmm0, %k0 # encoding: [0x62,0xf2,0xfd,0x08,0x27,0xc1]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vptestmq %xmm1, %xmm0, %k1 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x27,0xc9]
; X86-NEXT:    kmovw %k1, %ecx # encoding: [0xc5,0xf8,0x93,0xc9]
; X86-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; X86-NEXT:    addb %cl, %al # encoding: [0x00,0xc8]
; X86-NEXT:    # kill: def $al killed $al killed $eax
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_ptestm_q_128:
; X64:       # %bb.0:
; X64-NEXT:    vptestmq %xmm1, %xmm0, %k0 # encoding: [0x62,0xf2,0xfd,0x08,0x27,0xc1]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vptestmq %xmm1, %xmm0, %k1 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x27,0xc9]
; X64-NEXT:    kmovw %k1, %ecx # encoding: [0xc5,0xf8,0x93,0xc9]
; X64-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; X64-NEXT:    addb %cl, %al # encoding: [0x00,0xc8]
; X64-NEXT:    # kill: def $al killed $al killed $eax
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call i8 @llvm.x86.avx512.ptestm.q.128(<2 x i64> %x0, <2 x i64> %x1, i8 %x2)
  %res1 = call i8 @llvm.x86.avx512.ptestm.q.128(<2 x i64> %x0, <2 x i64> %x1, i8-1)
  %res2 = add i8 %res, %res1
  ret i8 %res2
}

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

define i8@test_int_x86_avx512_ptestm_q_256(<4 x i64> %x0, <4 x i64> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_ptestm_q_256:
; X86:       # %bb.0:
; X86-NEXT:    vptestmq %ymm1, %ymm0, %k0 # encoding: [0x62,0xf2,0xfd,0x28,0x27,0xc1]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vptestmq %ymm1, %ymm0, %k1 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x27,0xc9]
; X86-NEXT:    kmovw %k1, %ecx # encoding: [0xc5,0xf8,0x93,0xc9]
; X86-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; X86-NEXT:    addb %cl, %al # encoding: [0x00,0xc8]
; X86-NEXT:    # kill: def $al killed $al killed $eax
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_ptestm_q_256:
; X64:       # %bb.0:
; X64-NEXT:    vptestmq %ymm1, %ymm0, %k0 # encoding: [0x62,0xf2,0xfd,0x28,0x27,0xc1]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vptestmq %ymm1, %ymm0, %k1 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x27,0xc9]
; X64-NEXT:    kmovw %k1, %ecx # encoding: [0xc5,0xf8,0x93,0xc9]
; X64-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; X64-NEXT:    addb %cl, %al # encoding: [0x00,0xc8]
; X64-NEXT:    # kill: def $al killed $al killed $eax
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call i8 @llvm.x86.avx512.ptestm.q.256(<4 x i64> %x0, <4 x i64> %x1, i8 %x2)
  %res1 = call i8 @llvm.x86.avx512.ptestm.q.256(<4 x i64> %x0, <4 x i64> %x1, i8-1)
  %res2 = add i8 %res, %res1
  ret i8 %res2
}

declare i8 @llvm.x86.avx512.ptestnm.d.128(<4 x i32>, <4 x i32>, i8 %x2)

define i8@test_int_x86_avx512_ptestnm_d_128(<4 x i32> %x0, <4 x i32> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_ptestnm_d_128:
; X86:       # %bb.0:
; X86-NEXT:    vptestnmd %xmm1, %xmm0, %k0 # encoding: [0x62,0xf2,0x7e,0x08,0x27,0xc1]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vptestnmd %xmm1, %xmm0, %k1 {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x27,0xc9]
; X86-NEXT:    kmovw %k1, %ecx # encoding: [0xc5,0xf8,0x93,0xc9]
; X86-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; X86-NEXT:    addb %cl, %al # encoding: [0x00,0xc8]
; X86-NEXT:    # kill: def $al killed $al killed $eax
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_ptestnm_d_128:
; X64:       # %bb.0:
; X64-NEXT:    vptestnmd %xmm1, %xmm0, %k0 # encoding: [0x62,0xf2,0x7e,0x08,0x27,0xc1]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vptestnmd %xmm1, %xmm0, %k1 {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x27,0xc9]
; X64-NEXT:    kmovw %k1, %ecx # encoding: [0xc5,0xf8,0x93,0xc9]
; X64-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; X64-NEXT:    addb %cl, %al # encoding: [0x00,0xc8]
; X64-NEXT:    # kill: def $al killed $al killed $eax
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call i8 @llvm.x86.avx512.ptestnm.d.128(<4 x i32> %x0, <4 x i32> %x1, i8 %x2)
  %res1 = call i8 @llvm.x86.avx512.ptestnm.d.128(<4 x i32> %x0, <4 x i32> %x1, i8-1)
  %res2 = add i8 %res, %res1
  ret i8 %res2
}

declare i8 @llvm.x86.avx512.ptestnm.d.256(<8 x i32>, <8 x i32>, i8 %x2)

define i8@test_int_x86_avx512_ptestnm_d_256(<8 x i32> %x0, <8 x i32> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_ptestnm_d_256:
; X86:       # %bb.0:
; X86-NEXT:    vptestnmd %ymm1, %ymm0, %k0 # encoding: [0x62,0xf2,0x7e,0x28,0x27,0xc1]
; X86-NEXT:    kmovw %k0, %ecx # encoding: [0xc5,0xf8,0x93,0xc8]
; 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:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_ptestnm_d_256:
; X64:       # %bb.0:
; X64-NEXT:    vptestnmd %ymm1, %ymm0, %k0 # encoding: [0x62,0xf2,0x7e,0x28,0x27,0xc1]
; X64-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; 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:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call i8 @llvm.x86.avx512.ptestnm.d.256(<8 x i32> %x0, <8 x i32> %x1, i8 %x2)
  %res1 = call i8 @llvm.x86.avx512.ptestnm.d.256(<8 x i32> %x0, <8 x i32> %x1, i8-1)
  %res2 = add i8 %res, %res1
  ret i8 %res2
}

declare i8 @llvm.x86.avx512.ptestnm.q.128(<2 x i64>, <2 x i64>, i8 %x2)

define i8@test_int_x86_avx512_ptestnm_q_128(<2 x i64> %x0, <2 x i64> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_ptestnm_q_128:
; X86:       # %bb.0:
; X86-NEXT:    vptestnmq %xmm1, %xmm0, %k0 # encoding: [0x62,0xf2,0xfe,0x08,0x27,0xc1]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vptestnmq %xmm1, %xmm0, %k1 {%k1} # encoding: [0x62,0xf2,0xfe,0x09,0x27,0xc9]
; X86-NEXT:    kmovw %k1, %ecx # encoding: [0xc5,0xf8,0x93,0xc9]
; X86-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; X86-NEXT:    addb %cl, %al # encoding: [0x00,0xc8]
; X86-NEXT:    # kill: def $al killed $al killed $eax
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_ptestnm_q_128:
; X64:       # %bb.0:
; X64-NEXT:    vptestnmq %xmm1, %xmm0, %k0 # encoding: [0x62,0xf2,0xfe,0x08,0x27,0xc1]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vptestnmq %xmm1, %xmm0, %k1 {%k1} # encoding: [0x62,0xf2,0xfe,0x09,0x27,0xc9]
; X64-NEXT:    kmovw %k1, %ecx # encoding: [0xc5,0xf8,0x93,0xc9]
; X64-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; X64-NEXT:    addb %cl, %al # encoding: [0x00,0xc8]
; X64-NEXT:    # kill: def $al killed $al killed $eax
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call i8 @llvm.x86.avx512.ptestnm.q.128(<2 x i64> %x0, <2 x i64> %x1, i8 %x2)
  %res1 = call i8 @llvm.x86.avx512.ptestnm.q.128(<2 x i64> %x0, <2 x i64> %x1, i8-1)
  %res2 = add i8 %res, %res1
  ret i8 %res2
}

declare i8 @llvm.x86.avx512.ptestnm.q.256(<4 x i64>, <4 x i64>, i8 %x2)

define i8@test_int_x86_avx512_ptestnm_q_256(<4 x i64> %x0, <4 x i64> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_ptestnm_q_256:
; X86:       # %bb.0:
; X86-NEXT:    vptestnmq %ymm1, %ymm0, %k0 # encoding: [0x62,0xf2,0xfe,0x28,0x27,0xc1]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vptestnmq %ymm1, %ymm0, %k1 {%k1} # encoding: [0x62,0xf2,0xfe,0x29,0x27,0xc9]
; X86-NEXT:    kmovw %k1, %ecx # encoding: [0xc5,0xf8,0x93,0xc9]
; X86-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; X86-NEXT:    addb %cl, %al # encoding: [0x00,0xc8]
; X86-NEXT:    # kill: def $al killed $al killed $eax
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_ptestnm_q_256:
; X64:       # %bb.0:
; X64-NEXT:    vptestnmq %ymm1, %ymm0, %k0 # encoding: [0x62,0xf2,0xfe,0x28,0x27,0xc1]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vptestnmq %ymm1, %ymm0, %k1 {%k1} # encoding: [0x62,0xf2,0xfe,0x29,0x27,0xc9]
; X64-NEXT:    kmovw %k1, %ecx # encoding: [0xc5,0xf8,0x93,0xc9]
; X64-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; X64-NEXT:    addb %cl, %al # encoding: [0x00,0xc8]
; X64-NEXT:    # kill: def $al killed $al killed $eax
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call i8 @llvm.x86.avx512.ptestnm.q.256(<4 x i64> %x0, <4 x i64> %x1, i8 %x2)
  %res1 = call i8 @llvm.x86.avx512.ptestnm.q.256(<4 x i64> %x0, <4 x i64> %x1, i8-1)
  %res2 = add i8 %res, %res1
  ret i8 %res2
}

define i8 @test_cmpps_256(<8 x float> %a, <8 x float> %b) {
; CHECK-LABEL: test_cmpps_256:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vcmpleps %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7c,0x28,0xc2,0xc1,0x02]
; CHECK-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; CHECK-NEXT:    # kill: def $al killed $al killed $eax
; CHECK-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call i8 @llvm.x86.avx512.mask.cmp.ps.256(<8 x float> %a, <8 x float> %b, i32 2, i8 -1)
  ret i8 %res
}
declare i8 @llvm.x86.avx512.mask.cmp.ps.256(<8 x float> , <8 x float> , i32, i8)

define i8 @test_cmpps_128(<4 x float> %a, <4 x float> %b) {
; CHECK-LABEL: test_cmpps_128:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vcmpleps %xmm1, %xmm0, %k0 # encoding: [0x62,0xf1,0x7c,0x08,0xc2,0xc1,0x02]
; CHECK-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; CHECK-NEXT:    # kill: def $al killed $al killed $eax
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call i8 @llvm.x86.avx512.mask.cmp.ps.128(<4 x float> %a, <4 x float> %b, i32 2, i8 -1)
  ret i8 %res
}
declare i8 @llvm.x86.avx512.mask.cmp.ps.128(<4 x float> , <4 x float> , i32, i8)

define i8 @test_cmppd_256(<4 x double> %a, <4 x double> %b) {
; CHECK-LABEL: test_cmppd_256:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vcmplepd %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0xfd,0x28,0xc2,0xc1,0x02]
; CHECK-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; CHECK-NEXT:    # kill: def $al killed $al killed $eax
; CHECK-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call i8 @llvm.x86.avx512.mask.cmp.pd.256(<4 x double> %a, <4 x double> %b, i32 2, i8 -1)
  ret i8 %res
}
declare i8 @llvm.x86.avx512.mask.cmp.pd.256(<4 x double> , <4 x double> , i32, i8)

define i8 @test_cmppd_128(<2 x double> %a, <2 x double> %b) {
; CHECK-LABEL: test_cmppd_128:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vcmplepd %xmm1, %xmm0, %k0 # encoding: [0x62,0xf1,0xfd,0x08,0xc2,0xc1,0x02]
; CHECK-NEXT:    kmovw %k0, %eax # encoding: [0xc5,0xf8,0x93,0xc0]
; CHECK-NEXT:    # kill: def $al killed $al killed $eax
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call i8 @llvm.x86.avx512.mask.cmp.pd.128(<2 x double> %a, <2 x double> %b, i32 2, i8 -1)
  ret i8 %res
}
declare i8 @llvm.x86.avx512.mask.cmp.pd.128(<2 x double> , <2 x double> , i32, i8)

define < 2 x i64> @test_mask_mul_epi32_rr_128(< 4 x i32> %a, < 4 x i32> %b) {
; CHECK-LABEL: test_mask_mul_epi32_rr_128:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vpmuldq %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x28,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call < 2 x i64> @llvm.x86.avx512.mask.pmul.dq.128(< 4 x i32> %a, < 4 x i32> %b, < 2 x i64> zeroinitializer, i8 -1)
  ret < 2 x i64> %res
}

define < 2 x i64> @test_mask_mul_epi32_rrk_128(< 4 x i32> %a, < 4 x i32> %b, < 2 x i64> %passThru, i8 %mask) {
; X86-LABEL: test_mask_mul_epi32_rrk_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpmuldq %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x28,0xd1]
; X86-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_mul_epi32_rrk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmuldq %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x28,0xd1]
; X64-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call < 2 x i64> @llvm.x86.avx512.mask.pmul.dq.128(< 4 x i32> %a, < 4 x i32> %b, < 2 x i64> %passThru, i8 %mask)
  ret < 2 x i64> %res
}

define < 2 x i64> @test_mask_mul_epi32_rrkz_128(< 4 x i32> %a, < 4 x i32> %b, i8 %mask) {
; X86-LABEL: test_mask_mul_epi32_rrkz_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpmuldq %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x28,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_mul_epi32_rrkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmuldq %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x28,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call < 2 x i64> @llvm.x86.avx512.mask.pmul.dq.128(< 4 x i32> %a, < 4 x i32> %b, < 2 x i64> zeroinitializer, i8 %mask)
  ret < 2 x i64> %res
}

define < 2 x i64> @test_mask_mul_epi32_rm_128(< 4 x i32> %a, < 4 x i32>* %ptr_b) {
; X86-LABEL: test_mask_mul_epi32_rm_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpmuldq (%eax), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x28,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_mul_epi32_rm_128:
; X64:       # %bb.0:
; X64-NEXT:    vpmuldq (%rdi), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x28,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %b = load < 4 x i32>, < 4 x i32>* %ptr_b
  %res = call < 2 x i64> @llvm.x86.avx512.mask.pmul.dq.128(< 4 x i32> %a, < 4 x i32> %b, < 2 x i64> zeroinitializer, i8 -1)
  ret < 2 x i64> %res
}

define < 2 x i64> @test_mask_mul_epi32_rmk_128(< 4 x i32> %a, < 4 x i32>* %ptr_b, < 2 x i64> %passThru, i8 %mask) {
; X86-LABEL: test_mask_mul_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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpmuldq (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x28,0x08]
; X86-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_mul_epi32_rmk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpmuldq (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x28,0x0f]
; 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 < 2 x i64> @llvm.x86.avx512.mask.pmul.dq.128(< 4 x i32> %a, < 4 x i32> %b, < 2 x i64> %passThru, i8 %mask)
  ret < 2 x i64> %res
}

define < 2 x i64> @test_mask_mul_epi32_rmkz_128(< 4 x i32> %a, < 4 x i32>* %ptr_b, i8 %mask) {
; X86-LABEL: test_mask_mul_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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpmuldq (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x28,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_mul_epi32_rmkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpmuldq (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x28,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %b = load < 4 x i32>, < 4 x i32>* %ptr_b
  %res = call < 2 x i64> @llvm.x86.avx512.mask.pmul.dq.128(< 4 x i32> %a, < 4 x i32> %b, < 2 x i64> zeroinitializer, i8 %mask)
  ret < 2 x i64> %res
}

define < 2 x i64> @test_mask_mul_epi32_rmb_128(< 4 x i32> %a, i64* %ptr_b) {
; X86-LABEL: test_mask_mul_epi32_rmb_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpbroadcastd (%eax), %xmm1 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x58,0x08]
; X86-NEXT:    vpmuldq %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x28,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_mul_epi32_rmb_128:
; X64:       # %bb.0:
; X64-NEXT:    vpmuldq (%rdi){1to2}, %xmm0, %xmm0 # encoding: [0x62,0xf2,0xfd,0x18,0x28,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %q = load i64, i64* %ptr_b
  %vecinit.i = insertelement < 2 x i64> undef, i64 %q, i32 0
  %b64 = shufflevector < 2 x i64> %vecinit.i, < 2 x i64> undef, <2 x i32> zeroinitializer
  %b = bitcast < 2 x i64> %b64 to < 4 x i32>
  %res = call < 2 x i64> @llvm.x86.avx512.mask.pmul.dq.128(< 4 x i32> %a, < 4 x i32> %b, < 2 x i64> zeroinitializer, i8 -1)
  ret < 2 x i64> %res
}

define < 2 x i64> @test_mask_mul_epi32_rmbk_128(< 4 x i32> %a, i64* %ptr_b, < 2 x i64> %passThru, i8 %mask) {
; X86-LABEL: test_mask_mul_epi32_rmbk_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpbroadcastd (%eax), %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x58,0x10]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpmuldq %xmm2, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x28,0xca]
; X86-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_mul_epi32_rmbk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpmuldq (%rdi){1to2}, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x19,0x28,0x0f]
; X64-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %q = load i64, i64* %ptr_b
  %vecinit.i = insertelement < 2 x i64> undef, i64 %q, i32 0
  %b64 = shufflevector < 2 x i64> %vecinit.i, < 2 x i64> undef, <2 x i32> zeroinitializer
  %b = bitcast < 2 x i64> %b64 to < 4 x i32>
  %res = call < 2 x i64> @llvm.x86.avx512.mask.pmul.dq.128(< 4 x i32> %a, < 4 x i32> %b, < 2 x i64> %passThru, i8 %mask)
  ret < 2 x i64> %res
}

define < 2 x i64> @test_mask_mul_epi32_rmbkz_128(< 4 x i32> %a, i64* %ptr_b, i8 %mask) {
; X86-LABEL: test_mask_mul_epi32_rmbkz_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpbroadcastd (%eax), %xmm1 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x58,0x08]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpmuldq %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x28,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_mul_epi32_rmbkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpmuldq (%rdi){1to2}, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x99,0x28,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %q = load i64, i64* %ptr_b
  %vecinit.i = insertelement < 2 x i64> undef, i64 %q, i32 0
  %b64 = shufflevector < 2 x i64> %vecinit.i, < 2 x i64> undef, < 2 x i32> zeroinitializer
  %b = bitcast < 2 x i64> %b64 to < 4 x i32>
  %res = call < 2 x i64> @llvm.x86.avx512.mask.pmul.dq.128(< 4 x i32> %a, < 4 x i32> %b, < 2 x i64> zeroinitializer, i8 %mask)
  ret < 2 x i64> %res
}

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

define < 4 x i64> @test_mask_mul_epi32_rr_256(< 8 x i32> %a, < 8 x i32> %b) {
; CHECK-LABEL: test_mask_mul_epi32_rr_256:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vpmuldq %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x28,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call < 4 x i64> @llvm.x86.avx512.mask.pmul.dq.256(< 8 x i32> %a, < 8 x i32> %b, < 4 x i64> zeroinitializer, i8 -1)
  ret < 4 x i64> %res
}

define < 4 x i64> @test_mask_mul_epi32_rrk_256(< 8 x i32> %a, < 8 x i32> %b, < 4 x i64> %passThru, i8 %mask) {
; X86-LABEL: test_mask_mul_epi32_rrk_256:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpmuldq %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x28,0xd1]
; X86-NEXT:    vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_mul_epi32_rrk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmuldq %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x28,0xd1]
; X64-NEXT:    vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call < 4 x i64> @llvm.x86.avx512.mask.pmul.dq.256(< 8 x i32> %a, < 8 x i32> %b, < 4 x i64> %passThru, i8 %mask)
  ret < 4 x i64> %res
}

define < 4 x i64> @test_mask_mul_epi32_rrkz_256(< 8 x i32> %a, < 8 x i32> %b, i8 %mask) {
; X86-LABEL: test_mask_mul_epi32_rrkz_256:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpmuldq %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x28,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_mul_epi32_rrkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmuldq %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x28,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call < 4 x i64> @llvm.x86.avx512.mask.pmul.dq.256(< 8 x i32> %a, < 8 x i32> %b, < 4 x i64> zeroinitializer, i8 %mask)
  ret < 4 x i64> %res
}

define < 4 x i64> @test_mask_mul_epi32_rm_256(< 8 x i32> %a, < 8 x i32>* %ptr_b) {
; X86-LABEL: test_mask_mul_epi32_rm_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpmuldq (%eax), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x28,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_mul_epi32_rm_256:
; X64:       # %bb.0:
; X64-NEXT:    vpmuldq (%rdi), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x28,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %b = load < 8 x i32>, < 8 x i32>* %ptr_b
  %res = call < 4 x i64> @llvm.x86.avx512.mask.pmul.dq.256(< 8 x i32> %a, < 8 x i32> %b, < 4 x i64> zeroinitializer, i8 -1)
  ret < 4 x i64> %res
}

define < 4 x i64> @test_mask_mul_epi32_rmk_256(< 8 x i32> %a, < 8 x i32>* %ptr_b, < 4 x i64> %passThru, i8 %mask) {
; X86-LABEL: test_mask_mul_epi32_rmk_256:
; 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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpmuldq (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x28,0x08]
; X86-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_mul_epi32_rmk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpmuldq (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x28,0x0f]
; 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 < 4 x i64> @llvm.x86.avx512.mask.pmul.dq.256(< 8 x i32> %a, < 8 x i32> %b, < 4 x i64> %passThru, i8 %mask)
  ret < 4 x i64> %res
}

define < 4 x i64> @test_mask_mul_epi32_rmkz_256(< 8 x i32> %a, < 8 x i32>* %ptr_b, i8 %mask) {
; X86-LABEL: test_mask_mul_epi32_rmkz_256:
; 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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpmuldq (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x28,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_mul_epi32_rmkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpmuldq (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x28,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %b = load < 8 x i32>, < 8 x i32>* %ptr_b
  %res = call < 4 x i64> @llvm.x86.avx512.mask.pmul.dq.256(< 8 x i32> %a, < 8 x i32> %b, < 4 x i64> zeroinitializer, i8 %mask)
  ret < 4 x i64> %res
}

define < 4 x i64> @test_mask_mul_epi32_rmb_256(< 8 x i32> %a, i64* %ptr_b) {
; X86-LABEL: test_mask_mul_epi32_rmb_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpbroadcastd (%eax), %ymm1 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x58,0x08]
; X86-NEXT:    vpmuldq %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x28,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_mul_epi32_rmb_256:
; X64:       # %bb.0:
; X64-NEXT:    vpmuldq (%rdi){1to4}, %ymm0, %ymm0 # encoding: [0x62,0xf2,0xfd,0x38,0x28,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %q = load i64, i64* %ptr_b
  %vecinit.i = insertelement < 4 x i64> undef, i64 %q, i32 0
  %b64 = shufflevector < 4 x i64> %vecinit.i, < 4 x i64> undef, < 4 x i32> zeroinitializer
  %b = bitcast < 4 x i64> %b64 to < 8 x i32>
  %res = call < 4 x i64> @llvm.x86.avx512.mask.pmul.dq.256(< 8 x i32> %a, < 8 x i32> %b, < 4 x i64> zeroinitializer, i8 -1)
  ret < 4 x i64> %res
}

define < 4 x i64> @test_mask_mul_epi32_rmbk_256(< 8 x i32> %a, i64* %ptr_b, < 4 x i64> %passThru, i8 %mask) {
; X86-LABEL: test_mask_mul_epi32_rmbk_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpbroadcastd (%eax), %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x58,0x10]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpmuldq %ymm2, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x28,0xca]
; X86-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_mul_epi32_rmbk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpmuldq (%rdi){1to4}, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x39,0x28,0x0f]
; X64-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %q = load i64, i64* %ptr_b
  %vecinit.i = insertelement < 4 x i64> undef, i64 %q, i32 0
  %b64 = shufflevector < 4 x i64> %vecinit.i, < 4 x i64> undef, < 4 x i32> zeroinitializer
  %b = bitcast < 4 x i64> %b64 to < 8 x i32>
  %res = call < 4 x i64> @llvm.x86.avx512.mask.pmul.dq.256(< 8 x i32> %a, < 8 x i32> %b, < 4 x i64> %passThru, i8 %mask)
  ret < 4 x i64> %res
}

define < 4 x i64> @test_mask_mul_epi32_rmbkz_256(< 8 x i32> %a, i64* %ptr_b, i8 %mask) {
; X86-LABEL: test_mask_mul_epi32_rmbkz_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpbroadcastd (%eax), %ymm1 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x58,0x08]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpmuldq %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x28,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_mul_epi32_rmbkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpmuldq (%rdi){1to4}, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xb9,0x28,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %q = load i64, i64* %ptr_b
  %vecinit.i = insertelement < 4 x i64> undef, i64 %q, i32 0
  %b64 = shufflevector < 4 x i64> %vecinit.i, < 4 x i64> undef, < 4 x i32> zeroinitializer
  %b = bitcast < 4 x i64> %b64 to < 8 x i32>
  %res = call < 4 x i64> @llvm.x86.avx512.mask.pmul.dq.256(< 8 x i32> %a, < 8 x i32> %b, < 4 x i64> zeroinitializer, i8 %mask)
  ret < 4 x i64> %res
}

declare < 4 x i64> @llvm.x86.avx512.mask.pmul.dq.256(< 8 x i32>, < 8 x i32>, < 4 x i64>, i8)

define < 2 x i64> @test_mask_mul_epu32_rr_128(< 4 x i32> %a, < 4 x i32> %b) {
; CHECK-LABEL: test_mask_mul_epu32_rr_128:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vpmuludq %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xf4,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call < 2 x i64> @llvm.x86.avx512.mask.pmulu.dq.128(< 4 x i32> %a, < 4 x i32> %b, < 2 x i64> zeroinitializer, i8 -1)
  ret < 2 x i64> %res
}

define < 2 x i64> @test_mask_mul_epu32_rrk_128(< 4 x i32> %a, < 4 x i32> %b, < 2 x i64> %passThru, i8 %mask) {
; X86-LABEL: test_mask_mul_epu32_rrk_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpmuludq %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0xfd,0x09,0xf4,0xd1]
; X86-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_mul_epu32_rrk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmuludq %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0xfd,0x09,0xf4,0xd1]
; X64-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call < 2 x i64> @llvm.x86.avx512.mask.pmulu.dq.128(< 4 x i32> %a, < 4 x i32> %b, < 2 x i64> %passThru, i8 %mask)
  ret < 2 x i64> %res
}

define < 2 x i64> @test_mask_mul_epu32_rrkz_128(< 4 x i32> %a, < 4 x i32> %b, i8 %mask) {
; X86-LABEL: test_mask_mul_epu32_rrkz_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpmuludq %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0x89,0xf4,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_mul_epu32_rrkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmuludq %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0x89,0xf4,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call < 2 x i64> @llvm.x86.avx512.mask.pmulu.dq.128(< 4 x i32> %a, < 4 x i32> %b, < 2 x i64> zeroinitializer, i8 %mask)
  ret < 2 x i64> %res
}

define < 2 x i64> @test_mask_mul_epu32_rm_128(< 4 x i32> %a, < 4 x i32>* %ptr_b) {
; X86-LABEL: test_mask_mul_epu32_rm_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpmuludq (%eax), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xf4,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_mul_epu32_rm_128:
; X64:       # %bb.0:
; X64-NEXT:    vpmuludq (%rdi), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xf4,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %b = load < 4 x i32>, < 4 x i32>* %ptr_b
  %res = call < 2 x i64> @llvm.x86.avx512.mask.pmulu.dq.128(< 4 x i32> %a, < 4 x i32> %b, < 2 x i64> zeroinitializer, i8 -1)
  ret < 2 x i64> %res
}

define < 2 x i64> @test_mask_mul_epu32_rmk_128(< 4 x i32> %a, < 4 x i32>* %ptr_b, < 2 x i64> %passThru, i8 %mask) {
; X86-LABEL: test_mask_mul_epu32_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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpmuludq (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xfd,0x09,0xf4,0x08]
; X86-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_mul_epu32_rmk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpmuludq (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xfd,0x09,0xf4,0x0f]
; 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 < 2 x i64> @llvm.x86.avx512.mask.pmulu.dq.128(< 4 x i32> %a, < 4 x i32> %b, < 2 x i64> %passThru, i8 %mask)
  ret < 2 x i64> %res
}

define < 2 x i64> @test_mask_mul_epu32_rmkz_128(< 4 x i32> %a, < 4 x i32>* %ptr_b, i8 %mask) {
; X86-LABEL: test_mask_mul_epu32_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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpmuludq (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0x89,0xf4,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_mul_epu32_rmkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpmuludq (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0x89,0xf4,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %b = load < 4 x i32>, < 4 x i32>* %ptr_b
  %res = call < 2 x i64> @llvm.x86.avx512.mask.pmulu.dq.128(< 4 x i32> %a, < 4 x i32> %b, < 2 x i64> zeroinitializer, i8 %mask)
  ret < 2 x i64> %res
}

define < 2 x i64> @test_mask_mul_epu32_rmb_128(< 4 x i32> %a, i64* %ptr_b) {
; X86-LABEL: test_mask_mul_epu32_rmb_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpbroadcastd (%eax), %xmm1 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x58,0x08]
; X86-NEXT:    vpmuludq %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xf4,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_mul_epu32_rmb_128:
; X64:       # %bb.0:
; X64-NEXT:    vpmuludq (%rdi){1to2}, %xmm0, %xmm0 # encoding: [0x62,0xf1,0xfd,0x18,0xf4,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %q = load i64, i64* %ptr_b
  %vecinit.i = insertelement < 2 x i64> undef, i64 %q, i32 0
  %b64 = shufflevector < 2 x i64> %vecinit.i, < 2 x i64> undef, <2 x i32> zeroinitializer
  %b = bitcast < 2 x i64> %b64 to < 4 x i32>
  %res = call < 2 x i64> @llvm.x86.avx512.mask.pmulu.dq.128(< 4 x i32> %a, < 4 x i32> %b, < 2 x i64> zeroinitializer, i8 -1)
  ret < 2 x i64> %res
}

define < 2 x i64> @test_mask_mul_epu32_rmbk_128(< 4 x i32> %a, i64* %ptr_b, < 2 x i64> %passThru, i8 %mask) {
; X86-LABEL: test_mask_mul_epu32_rmbk_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpbroadcastd (%eax), %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x58,0x10]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpmuludq %xmm2, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xfd,0x09,0xf4,0xca]
; X86-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_mul_epu32_rmbk_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpmuludq (%rdi){1to2}, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xfd,0x19,0xf4,0x0f]
; X64-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %q = load i64, i64* %ptr_b
  %vecinit.i = insertelement < 2 x i64> undef, i64 %q, i32 0
  %b64 = shufflevector < 2 x i64> %vecinit.i, < 2 x i64> undef, <2 x i32> zeroinitializer
  %b = bitcast < 2 x i64> %b64 to < 4 x i32>
  %res = call < 2 x i64> @llvm.x86.avx512.mask.pmulu.dq.128(< 4 x i32> %a, < 4 x i32> %b, < 2 x i64> %passThru, i8 %mask)
  ret < 2 x i64> %res
}

define < 2 x i64> @test_mask_mul_epu32_rmbkz_128(< 4 x i32> %a, i64* %ptr_b, i8 %mask) {
; X86-LABEL: test_mask_mul_epu32_rmbkz_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpbroadcastd (%eax), %xmm1 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x58,0x08]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpmuludq %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0x89,0xf4,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_mul_epu32_rmbkz_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpmuludq (%rdi){1to2}, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0x99,0xf4,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %q = load i64, i64* %ptr_b
  %vecinit.i = insertelement < 2 x i64> undef, i64 %q, i32 0
  %b64 = shufflevector < 2 x i64> %vecinit.i, < 2 x i64> undef, < 2 x i32> zeroinitializer
  %b = bitcast < 2 x i64> %b64 to < 4 x i32>
  %res = call < 2 x i64> @llvm.x86.avx512.mask.pmulu.dq.128(< 4 x i32> %a, < 4 x i32> %b, < 2 x i64> zeroinitializer, i8 %mask)
  ret < 2 x i64> %res
}

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

define < 4 x i64> @test_mask_mul_epu32_rr_256(< 8 x i32> %a, < 8 x i32> %b) {
; CHECK-LABEL: test_mask_mul_epu32_rr_256:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vpmuludq %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xf4,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call < 4 x i64> @llvm.x86.avx512.mask.pmulu.dq.256(< 8 x i32> %a, < 8 x i32> %b, < 4 x i64> zeroinitializer, i8 -1)
  ret < 4 x i64> %res
}

define < 4 x i64> @test_mask_mul_epu32_rrk_256(< 8 x i32> %a, < 8 x i32> %b, < 4 x i64> %passThru, i8 %mask) {
; X86-LABEL: test_mask_mul_epu32_rrk_256:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpmuludq %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0xfd,0x29,0xf4,0xd1]
; X86-NEXT:    vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_mul_epu32_rrk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmuludq %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0xfd,0x29,0xf4,0xd1]
; X64-NEXT:    vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call < 4 x i64> @llvm.x86.avx512.mask.pmulu.dq.256(< 8 x i32> %a, < 8 x i32> %b, < 4 x i64> %passThru, i8 %mask)
  ret < 4 x i64> %res
}

define < 4 x i64> @test_mask_mul_epu32_rrkz_256(< 8 x i32> %a, < 8 x i32> %b, i8 %mask) {
; X86-LABEL: test_mask_mul_epu32_rrkz_256:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpmuludq %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0xa9,0xf4,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_mul_epu32_rrkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmuludq %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0xa9,0xf4,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call < 4 x i64> @llvm.x86.avx512.mask.pmulu.dq.256(< 8 x i32> %a, < 8 x i32> %b, < 4 x i64> zeroinitializer, i8 %mask)
  ret < 4 x i64> %res
}

define < 4 x i64> @test_mask_mul_epu32_rm_256(< 8 x i32> %a, < 8 x i32>* %ptr_b) {
; X86-LABEL: test_mask_mul_epu32_rm_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpmuludq (%eax), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xf4,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_mul_epu32_rm_256:
; X64:       # %bb.0:
; X64-NEXT:    vpmuludq (%rdi), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xf4,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %b = load < 8 x i32>, < 8 x i32>* %ptr_b
  %res = call < 4 x i64> @llvm.x86.avx512.mask.pmulu.dq.256(< 8 x i32> %a, < 8 x i32> %b, < 4 x i64> zeroinitializer, i8 -1)
  ret < 4 x i64> %res
}

define < 4 x i64> @test_mask_mul_epu32_rmk_256(< 8 x i32> %a, < 8 x i32>* %ptr_b, < 4 x i64> %passThru, i8 %mask) {
; X86-LABEL: test_mask_mul_epu32_rmk_256:
; 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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpmuludq (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0xfd,0x29,0xf4,0x08]
; X86-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_mul_epu32_rmk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpmuludq (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0xfd,0x29,0xf4,0x0f]
; 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 < 4 x i64> @llvm.x86.avx512.mask.pmulu.dq.256(< 8 x i32> %a, < 8 x i32> %b, < 4 x i64> %passThru, i8 %mask)
  ret < 4 x i64> %res
}

define < 4 x i64> @test_mask_mul_epu32_rmkz_256(< 8 x i32> %a, < 8 x i32>* %ptr_b, i8 %mask) {
; X86-LABEL: test_mask_mul_epu32_rmkz_256:
; 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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpmuludq (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0xa9,0xf4,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_mul_epu32_rmkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpmuludq (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0xa9,0xf4,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %b = load < 8 x i32>, < 8 x i32>* %ptr_b
  %res = call < 4 x i64> @llvm.x86.avx512.mask.pmulu.dq.256(< 8 x i32> %a, < 8 x i32> %b, < 4 x i64> zeroinitializer, i8 %mask)
  ret < 4 x i64> %res
}

define < 4 x i64> @test_mask_mul_epu32_rmb_256(< 8 x i32> %a, i64* %ptr_b) {
; X86-LABEL: test_mask_mul_epu32_rmb_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpbroadcastd (%eax), %ymm1 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x58,0x08]
; X86-NEXT:    vpmuludq %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xf4,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_mul_epu32_rmb_256:
; X64:       # %bb.0:
; X64-NEXT:    vpmuludq (%rdi){1to4}, %ymm0, %ymm0 # encoding: [0x62,0xf1,0xfd,0x38,0xf4,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %q = load i64, i64* %ptr_b
  %vecinit.i = insertelement < 4 x i64> undef, i64 %q, i32 0
  %b64 = shufflevector < 4 x i64> %vecinit.i, < 4 x i64> undef, < 4 x i32> zeroinitializer
  %b = bitcast < 4 x i64> %b64 to < 8 x i32>
  %res = call < 4 x i64> @llvm.x86.avx512.mask.pmulu.dq.256(< 8 x i32> %a, < 8 x i32> %b, < 4 x i64> zeroinitializer, i8 -1)
  ret < 4 x i64> %res
}

define < 4 x i64> @test_mask_mul_epu32_rmbk_256(< 8 x i32> %a, i64* %ptr_b, < 4 x i64> %passThru, i8 %mask) {
; X86-LABEL: test_mask_mul_epu32_rmbk_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpbroadcastd (%eax), %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x58,0x10]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpmuludq %ymm2, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0xfd,0x29,0xf4,0xca]
; X86-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_mul_epu32_rmbk_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpmuludq (%rdi){1to4}, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0xfd,0x39,0xf4,0x0f]
; X64-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %q = load i64, i64* %ptr_b
  %vecinit.i = insertelement < 4 x i64> undef, i64 %q, i32 0
  %b64 = shufflevector < 4 x i64> %vecinit.i, < 4 x i64> undef, < 4 x i32> zeroinitializer
  %b = bitcast < 4 x i64> %b64 to < 8 x i32>
  %res = call < 4 x i64> @llvm.x86.avx512.mask.pmulu.dq.256(< 8 x i32> %a, < 8 x i32> %b, < 4 x i64> %passThru, i8 %mask)
  ret < 4 x i64> %res
}

define < 4 x i64> @test_mask_mul_epu32_rmbkz_256(< 8 x i32> %a, i64* %ptr_b, i8 %mask) {
; X86-LABEL: test_mask_mul_epu32_rmbkz_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpbroadcastd (%eax), %ymm1 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x58,0x08]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpmuludq %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0xa9,0xf4,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_mul_epu32_rmbkz_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpmuludq (%rdi){1to4}, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0xb9,0xf4,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %q = load i64, i64* %ptr_b
  %vecinit.i = insertelement < 4 x i64> undef, i64 %q, i32 0
  %b64 = shufflevector < 4 x i64> %vecinit.i, < 4 x i64> undef, < 4 x i32> zeroinitializer
  %b = bitcast < 4 x i64> %b64 to < 8 x i32>
  %res = call < 4 x i64> @llvm.x86.avx512.mask.pmulu.dq.256(< 8 x i32> %a, < 8 x i32> %b, < 4 x i64> zeroinitializer, i8 %mask)
  ret < 4 x i64> %res
}

declare < 4 x i64> @llvm.x86.avx512.mask.pmulu.dq.256(< 8 x i32>, < 8 x i32>, < 4 x i64>, i8)

declare <4 x float> @llvm.x86.avx512.mask.cvtdq2ps.128(<4 x i32>, <4 x float>, i8)

define <4 x float>@test_int_x86_avx512_mask_cvt_dq2ps_128(<4 x i32> %x0, <4 x float> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_cvt_dq2ps_128:
; X86:       # %bb.0:
; X86-NEXT:    vcvtdq2ps %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x5b,0xd0]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vcvtdq2ps %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7c,0x09,0x5b,0xc8]
; X86-NEXT:    vaddps %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf0,0x58,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_cvt_dq2ps_128:
; X64:       # %bb.0:
; X64-NEXT:    vcvtdq2ps %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x5b,0xd0]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vcvtdq2ps %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7c,0x09,0x5b,0xc8]
; X64-NEXT:    vaddps %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf0,0x58,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask.cvtdq2ps.128(<4 x i32> %x0, <4 x float> %x1, i8 %x2)
  %res1 = call <4 x float> @llvm.x86.avx512.mask.cvtdq2ps.128(<4 x i32> %x0, <4 x float> %x1, i8 -1)
  %res2 = fadd <4 x float> %res, %res1
  ret <4 x float> %res2
}

declare <8 x float> @llvm.x86.avx512.mask.cvtdq2ps.256(<8 x i32>, <8 x float>, i8)

define <8 x float>@test_int_x86_avx512_mask_cvt_dq2ps_256(<8 x i32> %x0, <8 x float> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_cvt_dq2ps_256:
; X86:       # %bb.0:
; X86-NEXT:    vcvtdq2ps %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x5b,0xd0]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vcvtdq2ps %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7c,0x29,0x5b,0xc8]
; X86-NEXT:    vaddps %ymm2, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf4,0x58,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_cvt_dq2ps_256:
; X64:       # %bb.0:
; X64-NEXT:    vcvtdq2ps %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x5b,0xd0]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vcvtdq2ps %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7c,0x29,0x5b,0xc8]
; X64-NEXT:    vaddps %ymm2, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf4,0x58,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask.cvtdq2ps.256(<8 x i32> %x0, <8 x float> %x1, i8 %x2)
  %res1 = call <8 x float> @llvm.x86.avx512.mask.cvtdq2ps.256(<8 x i32> %x0, <8 x float> %x1, i8 -1)
  %res2 = fadd <8 x float> %res, %res1
  ret <8 x float> %res2
}

declare <4 x i32> @llvm.x86.avx512.mask.cvtpd2dq.256(<4 x double>, <4 x i32>, i8)

define <4 x i32>@test_int_x86_avx512_mask_cvt_pd2dq_256(<4 x double> %x0, <4 x i32> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_cvt_pd2dq_256:
; X86:       # %bb.0:
; X86-NEXT:    vcvtpd2dq %ymm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xff,0xe6,0xd0]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vcvtpd2dq %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xff,0x29,0xe6,0xc8]
; X86-NEXT:    vpaddd %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc2]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_cvt_pd2dq_256:
; X64:       # %bb.0:
; X64-NEXT:    vcvtpd2dq %ymm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xff,0xe6,0xd0]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vcvtpd2dq %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xff,0x29,0xe6,0xc8]
; X64-NEXT:    vpaddd %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc2]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i32> @llvm.x86.avx512.mask.cvtpd2dq.256(<4 x double> %x0, <4 x i32> %x1, i8 %x2)
  %res1 = call <4 x i32> @llvm.x86.avx512.mask.cvtpd2dq.256(<4 x double> %x0, <4 x i32> %x1, i8 -1)
  %res2 = add <4 x i32> %res, %res1
  ret <4 x i32> %res2
}

declare <4 x float> @llvm.x86.avx512.mask.cvtpd2ps.256(<4 x double>, <4 x float>, i8)

define <4 x float>@test_int_x86_avx512_mask_cvt_pd2ps_256(<4 x double> %x0, <4 x float> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_cvt_pd2ps_256:
; X86:       # %bb.0:
; X86-NEXT:    vcvtpd2ps %ymm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x5a,0xd0]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vcvtpd2ps %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xfd,0x29,0x5a,0xc8]
; X86-NEXT:    vaddps %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf0,0x58,0xc2]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_cvt_pd2ps_256:
; X64:       # %bb.0:
; X64-NEXT:    vcvtpd2ps %ymm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x5a,0xd0]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vcvtpd2ps %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xfd,0x29,0x5a,0xc8]
; X64-NEXT:    vaddps %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf0,0x58,0xc2]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask.cvtpd2ps.256(<4 x double> %x0, <4 x float> %x1, i8 %x2)
  %res1 = call <4 x float> @llvm.x86.avx512.mask.cvtpd2ps.256(<4 x double> %x0, <4 x float> %x1, i8 -1)
  %res2 = fadd <4 x float> %res, %res1
  ret <4 x float> %res2
}

declare <4 x double> @llvm.x86.avx512.mask.cvtps2pd.256(<4 x float>, <4 x double>, i8)

define <4 x double>@test_int_x86_avx512_mask_cvt_ps2pd_256(<4 x float> %x0, <4 x double> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_cvt_ps2pd_256:
; X86:       # %bb.0:
; X86-NEXT:    vcvtps2pd %xmm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x5a,0xd0]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vcvtps2pd %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7c,0x29,0x5a,0xc8]
; X86-NEXT:    vaddpd %ymm2, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0x58,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_cvt_ps2pd_256:
; X64:       # %bb.0:
; X64-NEXT:    vcvtps2pd %xmm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x5a,0xd0]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vcvtps2pd %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7c,0x29,0x5a,0xc8]
; X64-NEXT:    vaddpd %ymm2, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0x58,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x double> @llvm.x86.avx512.mask.cvtps2pd.256(<4 x float> %x0, <4 x double> %x1, i8 %x2)
  %res1 = call <4 x double> @llvm.x86.avx512.mask.cvtps2pd.256(<4 x float> %x0, <4 x double> %x1, i8 -1)
  %res2 = fadd <4 x double> %res, %res1
  ret <4 x double> %res2
}

declare <2 x double> @llvm.x86.avx512.mask.cvtps2pd.128(<4 x float>, <2 x double>, i8)

define <2 x double>@test_int_x86_avx512_mask_cvt_ps2pd_128(<4 x float> %x0, <2 x double> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_cvt_ps2pd_128:
; X86:       # %bb.0:
; X86-NEXT:    vcvtps2pd %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x5a,0xd0]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vcvtps2pd %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7c,0x09,0x5a,0xc8]
; X86-NEXT:    vaddpd %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0x58,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_cvt_ps2pd_128:
; X64:       # %bb.0:
; X64-NEXT:    vcvtps2pd %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x5a,0xd0]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vcvtps2pd %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7c,0x09,0x5a,0xc8]
; X64-NEXT:    vaddpd %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0x58,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x double> @llvm.x86.avx512.mask.cvtps2pd.128(<4 x float> %x0, <2 x double> %x1, i8 %x2)
  %res1 = call <2 x double> @llvm.x86.avx512.mask.cvtps2pd.128(<4 x float> %x0, <2 x double> %x1, i8 -1)
  %res2 = fadd <2 x double> %res, %res1
  ret <2 x double> %res2
}

declare <4 x i32> @llvm.x86.avx512.mask.cvttpd2dq.256(<4 x double>, <4 x i32>, i8)

define <4 x i32>@test_int_x86_avx512_mask_cvtt_pd2dq_256(<4 x double> %x0, <4 x i32> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_cvtt_pd2dq_256:
; X86:       # %bb.0:
; X86-NEXT:    vcvttpd2dq %ymm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xe6,0xd0]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vcvttpd2dq %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xfd,0x29,0xe6,0xc8]
; X86-NEXT:    vpaddd %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc2]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_cvtt_pd2dq_256:
; X64:       # %bb.0:
; X64-NEXT:    vcvttpd2dq %ymm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xe6,0xd0]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vcvttpd2dq %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xfd,0x29,0xe6,0xc8]
; X64-NEXT:    vpaddd %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc2]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i32> @llvm.x86.avx512.mask.cvttpd2dq.256(<4 x double> %x0, <4 x i32> %x1, i8 %x2)
  %res1 = call <4 x i32> @llvm.x86.avx512.mask.cvttpd2dq.256(<4 x double> %x0, <4 x i32> %x1, i8 -1)
  %res2 = add <4 x i32> %res, %res1
  ret <4 x i32> %res2
}

declare <4 x i32> @llvm.x86.avx512.mask.cvttps2dq.128(<4 x float>, <4 x i32>, i8)

define <4 x i32>@test_int_x86_avx512_mask_cvtt_ps2dq_128(<4 x float> %x0, <4 x i32> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_cvtt_ps2dq_128:
; X86:       # %bb.0:
; X86-NEXT:    vcvttps2dq %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xfa,0x5b,0xd0]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vcvttps2dq %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7e,0x09,0x5b,0xc8]
; X86-NEXT:    vpaddd %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_cvtt_ps2dq_128:
; X64:       # %bb.0:
; X64-NEXT:    vcvttps2dq %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xfa,0x5b,0xd0]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vcvttps2dq %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7e,0x09,0x5b,0xc8]
; X64-NEXT:    vpaddd %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i32> @llvm.x86.avx512.mask.cvttps2dq.128(<4 x float> %x0, <4 x i32> %x1, i8 %x2)
  %res1 = call <4 x i32> @llvm.x86.avx512.mask.cvttps2dq.128(<4 x float> %x0, <4 x i32> %x1, i8 -1)
  %res2 = add <4 x i32> %res, %res1
  ret <4 x i32> %res2
}

declare <8 x i32> @llvm.x86.avx512.mask.cvttps2dq.256(<8 x float>, <8 x i32>, i8)

define <8 x i32>@test_int_x86_avx512_mask_cvtt_ps2dq_256(<8 x float> %x0, <8 x i32> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_cvtt_ps2dq_256:
; X86:       # %bb.0:
; X86-NEXT:    vcvttps2dq %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xfe,0x5b,0xd0]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vcvttps2dq %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7e,0x29,0x5b,0xc8]
; X86-NEXT:    vpaddd %ymm2, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfe,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_cvtt_ps2dq_256:
; X64:       # %bb.0:
; X64-NEXT:    vcvttps2dq %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xfe,0x5b,0xd0]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vcvttps2dq %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7e,0x29,0x5b,0xc8]
; X64-NEXT:    vpaddd %ymm2, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfe,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x i32> @llvm.x86.avx512.mask.cvttps2dq.256(<8 x float> %x0, <8 x i32> %x1, i8 %x2)
  %res1 = call <8 x i32> @llvm.x86.avx512.mask.cvttps2dq.256(<8 x float> %x0, <8 x i32> %x1, i8 -1)
  %res2 = add <8 x i32> %res, %res1
  ret <8 x i32> %res2
}

declare <8 x float> @llvm.x86.avx512.mask.permvar.sf.256(<8 x float>, <8 x i32>, <8 x float>, i8)

define <8 x float>@test_int_x86_avx512_mask_permvar_sf_256(<8 x float> %x0, <8 x i32> %x1, <8 x float> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_permvar_sf_256:
; X86:       # %bb.0:
; X86-NEXT:    vpermps %ymm0, %ymm1, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x75,0x16,0xd8]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpermps %ymm0, %ymm1, %ymm2 {%k1} # encoding: [0x62,0xf2,0x75,0x29,0x16,0xd0]
; X86-NEXT:    vpermps %ymm0, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xa9,0x16,0xc0]
; X86-NEXT:    vaddps %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xec,0x58,0xc0]
; X86-NEXT:    vaddps %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x58,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_permvar_sf_256:
; X64:       # %bb.0:
; X64-NEXT:    vpermps %ymm0, %ymm1, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x75,0x16,0xd8]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpermps %ymm0, %ymm1, %ymm2 {%k1} # encoding: [0x62,0xf2,0x75,0x29,0x16,0xd0]
; X64-NEXT:    vpermps %ymm0, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xa9,0x16,0xc0]
; X64-NEXT:    vaddps %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xec,0x58,0xc0]
; X64-NEXT:    vaddps %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x58,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask.permvar.sf.256(<8 x float> %x0, <8 x i32> %x1, <8 x float> %x2, i8 %x3)
  %res1 = call <8 x float> @llvm.x86.avx512.mask.permvar.sf.256(<8 x float> %x0, <8 x i32> %x1, <8 x float> zeroinitializer, i8 %x3)
  %res2 = call <8 x float> @llvm.x86.avx512.mask.permvar.sf.256(<8 x float> %x0, <8 x i32> %x1, <8 x float> %x2, i8 -1)
  %res3 = fadd <8 x float> %res, %res1
  %res4 = fadd <8 x float> %res3, %res2
  ret <8 x float> %res4
}

declare <8 x i32> @llvm.x86.avx512.mask.permvar.si.256(<8 x i32>, <8 x i32>, <8 x i32>, i8)

define <8 x i32>@test_int_x86_avx512_mask_permvar_si_256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_permvar_si_256:
; X86:       # %bb.0:
; X86-NEXT:    vpermd %ymm0, %ymm1, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x75,0x36,0xd8]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpermd %ymm0, %ymm1, %ymm2 {%k1} # encoding: [0x62,0xf2,0x75,0x29,0x36,0xd0]
; X86-NEXT:    vpermd %ymm0, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xa9,0x36,0xc0]
; X86-NEXT:    vpaddd %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfe,0xc3]
; X86-NEXT:    vpaddd %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_permvar_si_256:
; X64:       # %bb.0:
; X64-NEXT:    vpermd %ymm0, %ymm1, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x75,0x36,0xd8]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpermd %ymm0, %ymm1, %ymm2 {%k1} # encoding: [0x62,0xf2,0x75,0x29,0x36,0xd0]
; X64-NEXT:    vpermd %ymm0, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xa9,0x36,0xc0]
; X64-NEXT:    vpaddd %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfe,0xc3]
; X64-NEXT:    vpaddd %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x i32> @llvm.x86.avx512.mask.permvar.si.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 %x3)
  %res1 = call <8 x i32> @llvm.x86.avx512.mask.permvar.si.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> zeroinitializer, i8 %x3)
  %res2 = call <8 x i32> @llvm.x86.avx512.mask.permvar.si.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 -1)
  %res3 = add <8 x i32> %res, %res1
  %res4 = add <8 x i32> %res3, %res2
  ret <8 x i32> %res4
}

declare <4 x double> @llvm.x86.avx512.mask.permvar.df.256(<4 x double>, <4 x i64>, <4 x double>, i8)

define <4 x double>@test_int_x86_avx512_mask_permvar_df_256(<4 x double> %x0, <4 x i64> %x1, <4 x double> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_permvar_df_256:
; X86:       # %bb.0:
; X86-NEXT:    vpermpd %ymm0, %ymm1, %ymm3 # encoding: [0x62,0xf2,0xf5,0x28,0x16,0xd8]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpermpd %ymm0, %ymm1, %ymm2 {%k1} # encoding: [0x62,0xf2,0xf5,0x29,0x16,0xd0]
; X86-NEXT:    vpermpd %ymm0, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0xa9,0x16,0xc0]
; X86-NEXT:    vaddpd %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0x58,0xc0]
; X86-NEXT:    vaddpd %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x58,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_permvar_df_256:
; X64:       # %bb.0:
; X64-NEXT:    vpermpd %ymm0, %ymm1, %ymm3 # encoding: [0x62,0xf2,0xf5,0x28,0x16,0xd8]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpermpd %ymm0, %ymm1, %ymm2 {%k1} # encoding: [0x62,0xf2,0xf5,0x29,0x16,0xd0]
; X64-NEXT:    vpermpd %ymm0, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0xa9,0x16,0xc0]
; X64-NEXT:    vaddpd %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0x58,0xc0]
; X64-NEXT:    vaddpd %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x58,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x double> @llvm.x86.avx512.mask.permvar.df.256(<4 x double> %x0, <4 x i64> %x1, <4 x double> %x2, i8 %x3)
  %res1 = call <4 x double> @llvm.x86.avx512.mask.permvar.df.256(<4 x double> %x0, <4 x i64> %x1, <4 x double> zeroinitializer, i8 %x3)
  %res2 = call <4 x double> @llvm.x86.avx512.mask.permvar.df.256(<4 x double> %x0, <4 x i64> %x1, <4 x double> %x2, i8 -1)
  %res3 = fadd <4 x double> %res, %res1
  %res4 = fadd <4 x double> %res3, %res2
  ret <4 x double> %res4
}

declare <4 x i64> @llvm.x86.avx512.mask.permvar.di.256(<4 x i64>, <4 x i64>, <4 x i64>, i8)

define <4 x i64>@test_int_x86_avx512_mask_permvar_di_256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_permvar_di_256:
; X86:       # %bb.0:
; X86-NEXT:    vpermq %ymm0, %ymm1, %ymm3 # encoding: [0x62,0xf2,0xf5,0x28,0x36,0xd8]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpermq %ymm0, %ymm1, %ymm2 {%k1} # encoding: [0x62,0xf2,0xf5,0x29,0x36,0xd0]
; X86-NEXT:    vpermq %ymm0, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0xa9,0x36,0xc0]
; X86-NEXT:    vpaddq %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd4,0xc3]
; X86-NEXT:    vpaddq %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xd4,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_permvar_di_256:
; X64:       # %bb.0:
; X64-NEXT:    vpermq %ymm0, %ymm1, %ymm3 # encoding: [0x62,0xf2,0xf5,0x28,0x36,0xd8]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpermq %ymm0, %ymm1, %ymm2 {%k1} # encoding: [0x62,0xf2,0xf5,0x29,0x36,0xd0]
; X64-NEXT:    vpermq %ymm0, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0xa9,0x36,0xc0]
; X64-NEXT:    vpaddq %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd4,0xc3]
; X64-NEXT:    vpaddq %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xd4,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i64> @llvm.x86.avx512.mask.permvar.di.256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2, i8 %x3)
  %res1 = call <4 x i64> @llvm.x86.avx512.mask.permvar.di.256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> zeroinitializer, i8 %x3)
  %res2 = call <4 x i64> @llvm.x86.avx512.mask.permvar.di.256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2, i8 -1)
  %res3 = add <4 x i64> %res, %res1
  %res4 = add <4 x i64> %res3, %res2
  ret <4 x i64> %res4
}

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

define <4 x i32>@test_int_x86_avx512_mask_pternlog_d_128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 %x4) {
; X86-LABEL: test_int_x86_avx512_mask_pternlog_d_128:
; X86:       # %bb.0:
; X86-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
; X86-NEXT:    vpternlogd $33, %xmm2, %xmm1, %xmm3 # encoding: [0x62,0xf3,0x75,0x08,0x25,0xda,0x21]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpternlogd $33, %xmm2, %xmm1, %xmm0 {%k1} # encoding: [0x62,0xf3,0x75,0x09,0x25,0xc2,0x21]
; X86-NEXT:    vpaddd %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfe,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pternlog_d_128:
; X64:       # %bb.0:
; X64-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
; X64-NEXT:    vpternlogd $33, %xmm2, %xmm1, %xmm3 # encoding: [0x62,0xf3,0x75,0x08,0x25,0xda,0x21]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpternlogd $33, %xmm2, %xmm1, %xmm0 {%k1} # encoding: [0x62,0xf3,0x75,0x09,0x25,0xc2,0x21]
; X64-NEXT:    vpaddd %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfe,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i32> @llvm.x86.avx512.mask.pternlog.d.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i32 33, i8 %x4)
  %res1 = call <4 x i32> @llvm.x86.avx512.mask.pternlog.d.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i32 33, i8 -1)
  %res2 = add <4 x i32> %res, %res1
  ret <4 x i32> %res2
}

declare <4 x i32> @llvm.x86.avx512.maskz.pternlog.d.128(<4 x i32>, <4 x i32>, <4 x i32>, i32, i8)

define <4 x i32>@test_int_x86_avx512_maskz_pternlog_d_128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 %x4) {
; X86-LABEL: test_int_x86_avx512_maskz_pternlog_d_128:
; X86:       # %bb.0:
; X86-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
; X86-NEXT:    vpternlogd $33, %xmm2, %xmm1, %xmm3 # encoding: [0x62,0xf3,0x75,0x08,0x25,0xda,0x21]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpternlogd $33, %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf3,0x75,0x89,0x25,0xc2,0x21]
; X86-NEXT:    vpaddd %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfe,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_maskz_pternlog_d_128:
; X64:       # %bb.0:
; X64-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
; X64-NEXT:    vpternlogd $33, %xmm2, %xmm1, %xmm3 # encoding: [0x62,0xf3,0x75,0x08,0x25,0xda,0x21]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpternlogd $33, %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf3,0x75,0x89,0x25,0xc2,0x21]
; X64-NEXT:    vpaddd %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfe,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i32> @llvm.x86.avx512.maskz.pternlog.d.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i32 33, i8 %x4)
  %res1 = call <4 x i32> @llvm.x86.avx512.maskz.pternlog.d.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i32 33, i8 -1)
  %res2 = add <4 x i32> %res, %res1
  ret <4 x i32> %res2
}

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

define <8 x i32>@test_int_x86_avx512_mask_pternlog_d_256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 %x4) {
; X86-LABEL: test_int_x86_avx512_mask_pternlog_d_256:
; X86:       # %bb.0:
; X86-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
; X86-NEXT:    vpternlogd $33, %ymm2, %ymm1, %ymm3 # encoding: [0x62,0xf3,0x75,0x28,0x25,0xda,0x21]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpternlogd $33, %ymm2, %ymm1, %ymm0 {%k1} # encoding: [0x62,0xf3,0x75,0x29,0x25,0xc2,0x21]
; X86-NEXT:    vpaddd %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfe,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pternlog_d_256:
; X64:       # %bb.0:
; X64-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
; X64-NEXT:    vpternlogd $33, %ymm2, %ymm1, %ymm3 # encoding: [0x62,0xf3,0x75,0x28,0x25,0xda,0x21]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpternlogd $33, %ymm2, %ymm1, %ymm0 {%k1} # encoding: [0x62,0xf3,0x75,0x29,0x25,0xc2,0x21]
; X64-NEXT:    vpaddd %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfe,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x i32> @llvm.x86.avx512.mask.pternlog.d.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i32 33, i8 %x4)
  %res1 = call <8 x i32> @llvm.x86.avx512.mask.pternlog.d.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i32 33, i8 -1)
  %res2 = add <8 x i32> %res, %res1
  ret <8 x i32> %res2
}

declare <8 x i32> @llvm.x86.avx512.maskz.pternlog.d.256(<8 x i32>, <8 x i32>, <8 x i32>, i32, i8)

define <8 x i32>@test_int_x86_avx512_maskz_pternlog_d_256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 %x4) {
; X86-LABEL: test_int_x86_avx512_maskz_pternlog_d_256:
; X86:       # %bb.0:
; X86-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
; X86-NEXT:    vpternlogd $33, %ymm2, %ymm1, %ymm3 # encoding: [0x62,0xf3,0x75,0x28,0x25,0xda,0x21]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpternlogd $33, %ymm2, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf3,0x75,0xa9,0x25,0xc2,0x21]
; X86-NEXT:    vpaddd %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfe,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_maskz_pternlog_d_256:
; X64:       # %bb.0:
; X64-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
; X64-NEXT:    vpternlogd $33, %ymm2, %ymm1, %ymm3 # encoding: [0x62,0xf3,0x75,0x28,0x25,0xda,0x21]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpternlogd $33, %ymm2, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf3,0x75,0xa9,0x25,0xc2,0x21]
; X64-NEXT:    vpaddd %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfe,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x i32> @llvm.x86.avx512.maskz.pternlog.d.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i32 33, i8 %x4)
  %res1 = call <8 x i32> @llvm.x86.avx512.maskz.pternlog.d.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i32 33, i8 -1)
  %res2 = add <8 x i32> %res, %res1
  ret <8 x i32> %res2
}

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

define <2 x i64>@test_int_x86_avx512_mask_pternlog_q_128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i8 %x4) {
; X86-LABEL: test_int_x86_avx512_mask_pternlog_q_128:
; X86:       # %bb.0:
; X86-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
; X86-NEXT:    vpternlogq $33, %xmm2, %xmm1, %xmm3 # encoding: [0x62,0xf3,0xf5,0x08,0x25,0xda,0x21]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpternlogq $33, %xmm2, %xmm1, %xmm0 {%k1} # encoding: [0x62,0xf3,0xf5,0x09,0x25,0xc2,0x21]
; X86-NEXT:    vpaddq %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd4,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pternlog_q_128:
; X64:       # %bb.0:
; X64-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
; X64-NEXT:    vpternlogq $33, %xmm2, %xmm1, %xmm3 # encoding: [0x62,0xf3,0xf5,0x08,0x25,0xda,0x21]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpternlogq $33, %xmm2, %xmm1, %xmm0 {%k1} # encoding: [0x62,0xf3,0xf5,0x09,0x25,0xc2,0x21]
; X64-NEXT:    vpaddq %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd4,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x i64> @llvm.x86.avx512.mask.pternlog.q.128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i32 33, i8 %x4)
  %res1 = call <2 x i64> @llvm.x86.avx512.mask.pternlog.q.128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i32 33, i8 -1)
  %res2 = add <2 x i64> %res, %res1
  ret <2 x i64> %res2
}

declare <2 x i64> @llvm.x86.avx512.maskz.pternlog.q.128(<2 x i64>, <2 x i64>, <2 x i64>, i32, i8)

define <2 x i64>@test_int_x86_avx512_maskz_pternlog_q_128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i8 %x4) {
; X86-LABEL: test_int_x86_avx512_maskz_pternlog_q_128:
; X86:       # %bb.0:
; X86-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
; X86-NEXT:    vpternlogq $33, %xmm2, %xmm1, %xmm3 # encoding: [0x62,0xf3,0xf5,0x08,0x25,0xda,0x21]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpternlogq $33, %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf3,0xf5,0x89,0x25,0xc2,0x21]
; X86-NEXT:    vpaddq %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd4,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_maskz_pternlog_q_128:
; X64:       # %bb.0:
; X64-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
; X64-NEXT:    vpternlogq $33, %xmm2, %xmm1, %xmm3 # encoding: [0x62,0xf3,0xf5,0x08,0x25,0xda,0x21]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpternlogq $33, %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf3,0xf5,0x89,0x25,0xc2,0x21]
; X64-NEXT:    vpaddq %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd4,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x i64> @llvm.x86.avx512.maskz.pternlog.q.128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i32 33, i8 %x4)
  %res1 = call <2 x i64> @llvm.x86.avx512.maskz.pternlog.q.128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i32 33, i8 -1)
  %res2 = add <2 x i64> %res, %res1
  ret <2 x i64> %res2
}

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

define <4 x i64>@test_int_x86_avx512_mask_pternlog_q_256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2, i8 %x4) {
; X86-LABEL: test_int_x86_avx512_mask_pternlog_q_256:
; X86:       # %bb.0:
; X86-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
; X86-NEXT:    vpternlogq $33, %ymm2, %ymm1, %ymm3 # encoding: [0x62,0xf3,0xf5,0x28,0x25,0xda,0x21]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpternlogq $33, %ymm2, %ymm1, %ymm0 {%k1} # encoding: [0x62,0xf3,0xf5,0x29,0x25,0xc2,0x21]
; X86-NEXT:    vpaddq %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd4,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pternlog_q_256:
; X64:       # %bb.0:
; X64-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
; X64-NEXT:    vpternlogq $33, %ymm2, %ymm1, %ymm3 # encoding: [0x62,0xf3,0xf5,0x28,0x25,0xda,0x21]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpternlogq $33, %ymm2, %ymm1, %ymm0 {%k1} # encoding: [0x62,0xf3,0xf5,0x29,0x25,0xc2,0x21]
; X64-NEXT:    vpaddq %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd4,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i64> @llvm.x86.avx512.mask.pternlog.q.256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2, i32 33, i8 %x4)
  %res1 = call <4 x i64> @llvm.x86.avx512.mask.pternlog.q.256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2, i32 33, i8 -1)
  %res2 = add <4 x i64> %res, %res1
  ret <4 x i64> %res2
}

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

define <4 x i64>@test_int_x86_avx512_maskz_pternlog_q_256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2, i8 %x4) {
; X86-LABEL: test_int_x86_avx512_maskz_pternlog_q_256:
; X86:       # %bb.0:
; X86-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
; X86-NEXT:    vpternlogq $33, %ymm2, %ymm1, %ymm3 # encoding: [0x62,0xf3,0xf5,0x28,0x25,0xda,0x21]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpternlogq $33, %ymm2, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf3,0xf5,0xa9,0x25,0xc2,0x21]
; X86-NEXT:    vpaddq %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd4,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_maskz_pternlog_q_256:
; X64:       # %bb.0:
; X64-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
; X64-NEXT:    vpternlogq $33, %ymm2, %ymm1, %ymm3 # encoding: [0x62,0xf3,0xf5,0x28,0x25,0xda,0x21]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpternlogq $33, %ymm2, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf3,0xf5,0xa9,0x25,0xc2,0x21]
; X64-NEXT:    vpaddq %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd4,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i64> @llvm.x86.avx512.maskz.pternlog.q.256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2, i32 33, i8 %x4)
  %res1 = call <4 x i64> @llvm.x86.avx512.maskz.pternlog.q.256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2, i32 33, i8 -1)
  %res2 = add <4 x i64> %res, %res1
  ret <4 x i64> %res2
}

declare <4 x float> @llvm.x86.avx512.mask.cvtudq2ps.128(<4 x i32>, <4 x float>, i8)

define <4 x float>@test_int_x86_avx512_mask_cvt_udq2ps_128(<4 x i32> %x0, <4 x float> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_cvt_udq2ps_128:
; X86:       # %bb.0:
; X86-NEXT:    vcvtudq2ps %xmm0, %xmm2 # encoding: [0x62,0xf1,0x7f,0x08,0x7a,0xd0]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vcvtudq2ps %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7f,0x09,0x7a,0xc8]
; X86-NEXT:    vaddps %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf0,0x58,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_cvt_udq2ps_128:
; X64:       # %bb.0:
; X64-NEXT:    vcvtudq2ps %xmm0, %xmm2 # encoding: [0x62,0xf1,0x7f,0x08,0x7a,0xd0]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vcvtudq2ps %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7f,0x09,0x7a,0xc8]
; X64-NEXT:    vaddps %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf0,0x58,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask.cvtudq2ps.128(<4 x i32> %x0, <4 x float> %x1, i8 %x2)
  %res1 = call <4 x float> @llvm.x86.avx512.mask.cvtudq2ps.128(<4 x i32> %x0, <4 x float> %x1, i8 -1)
  %res2 = fadd <4 x float> %res, %res1
  ret <4 x float> %res2
}

declare <8 x float> @llvm.x86.avx512.mask.cvtudq2ps.256(<8 x i32>, <8 x float>, i8)

define <8 x float>@test_int_x86_avx512_mask_cvt_udq2ps_256(<8 x i32> %x0, <8 x float> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_cvt_udq2ps_256:
; X86:       # %bb.0:
; X86-NEXT:    vcvtudq2ps %ymm0, %ymm2 # encoding: [0x62,0xf1,0x7f,0x28,0x7a,0xd0]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vcvtudq2ps %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7f,0x29,0x7a,0xc8]
; X86-NEXT:    vaddps %ymm2, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf4,0x58,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_cvt_udq2ps_256:
; X64:       # %bb.0:
; X64-NEXT:    vcvtudq2ps %ymm0, %ymm2 # encoding: [0x62,0xf1,0x7f,0x28,0x7a,0xd0]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vcvtudq2ps %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7f,0x29,0x7a,0xc8]
; X64-NEXT:    vaddps %ymm2, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf4,0x58,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask.cvtudq2ps.256(<8 x i32> %x0, <8 x float> %x1, i8 %x2)
  %res1 = call <8 x float> @llvm.x86.avx512.mask.cvtudq2ps.256(<8 x i32> %x0, <8 x float> %x1, i8 -1)
  %res2 = fadd <8 x float> %res, %res1
  ret <8 x float> %res2
}

declare <4 x i32> @llvm.x86.avx512.mask.vpermi2var.d.128(<4 x i32>, <4 x i32>, <4 x i32>, i8)

define <4 x i32>@test_int_x86_avx512_mask_vpermi2var_d_128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_vpermi2var_d_128:
; X86:       # %bb.0:
; X86-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
; X86-NEXT:    vpermt2d %xmm2, %xmm1, %xmm3 # encoding: [0x62,0xf2,0x75,0x08,0x7e,0xda]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpermi2d %xmm2, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x76,0xca]
; X86-NEXT:    vpaddd %xmm3, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_vpermi2var_d_128:
; X64:       # %bb.0:
; X64-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
; X64-NEXT:    vpermt2d %xmm2, %xmm1, %xmm3 # encoding: [0x62,0xf2,0x75,0x08,0x7e,0xda]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpermi2d %xmm2, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x76,0xca]
; X64-NEXT:    vpaddd %xmm3, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i32> @llvm.x86.avx512.mask.vpermi2var.d.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 %x3)
  %res1 = call <4 x i32> @llvm.x86.avx512.mask.vpermi2var.d.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 -1)
  %res2 = add <4 x i32> %res, %res1
  ret <4 x i32> %res2
}

declare <4 x i32> @llvm.x86.avx512.mask.vpermt2var.d.128(<4 x i32>, <4 x i32>, <4 x i32>, i8)

define <4 x i32>@test_int_x86_avx512_mask_vpermt2var_d_128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_vpermt2var_d_128:
; X86:       # %bb.0:
; X86-NEXT:    vmovdqa %xmm1, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd9]
; X86-NEXT:    vpermt2d %xmm2, %xmm0, %xmm3 # encoding: [0x62,0xf2,0x7d,0x08,0x7e,0xda]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpermt2d %xmm2, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x7e,0xca]
; X86-NEXT:    vpaddd %xmm3, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_vpermt2var_d_128:
; X64:       # %bb.0:
; X64-NEXT:    vmovdqa %xmm1, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd9]
; X64-NEXT:    vpermt2d %xmm2, %xmm0, %xmm3 # encoding: [0x62,0xf2,0x7d,0x08,0x7e,0xda]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpermt2d %xmm2, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x7e,0xca]
; X64-NEXT:    vpaddd %xmm3, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i32> @llvm.x86.avx512.mask.vpermt2var.d.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 %x3)
  %res1 = call <4 x i32> @llvm.x86.avx512.mask.vpermt2var.d.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 -1)
  %res2 = add <4 x i32> %res, %res1
  ret <4 x i32> %res2
}

declare <4 x i32> @llvm.x86.avx512.maskz.vpermt2var.d.128(<4 x i32>, <4 x i32>, <4 x i32>, i8)

define <4 x i32>@test_int_x86_avx512_maskz_vpermt2var_d_128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_maskz_vpermt2var_d_128:
; X86:       # %bb.0:
; X86-NEXT:    vmovdqa %xmm1, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd9]
; X86-NEXT:    vpermt2d %xmm2, %xmm0, %xmm3 # encoding: [0x62,0xf2,0x7d,0x08,0x7e,0xda]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpermt2d %xmm2, %xmm0, %xmm1 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x7e,0xca]
; X86-NEXT:    vpaddd %xmm3, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_maskz_vpermt2var_d_128:
; X64:       # %bb.0:
; X64-NEXT:    vmovdqa %xmm1, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd9]
; X64-NEXT:    vpermt2d %xmm2, %xmm0, %xmm3 # encoding: [0x62,0xf2,0x7d,0x08,0x7e,0xda]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpermt2d %xmm2, %xmm0, %xmm1 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x7e,0xca]
; X64-NEXT:    vpaddd %xmm3, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i32> @llvm.x86.avx512.maskz.vpermt2var.d.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 %x3)
  %res1 = call <4 x i32> @llvm.x86.avx512.maskz.vpermt2var.d.128(<4 x i32> %x0, <4 x i32> %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.vpermi2var.d.256(<8 x i32>, <8 x i32>, <8 x i32>, i8)

define <8 x i32>@test_int_x86_avx512_mask_vpermi2var_d_256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_vpermi2var_d_256:
; X86:       # %bb.0:
; X86-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
; X86-NEXT:    vpermt2d %ymm2, %ymm1, %ymm3 # encoding: [0x62,0xf2,0x75,0x28,0x7e,0xda]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpermi2d %ymm2, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x76,0xca]
; X86-NEXT:    vpaddd %ymm3, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfe,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_vpermi2var_d_256:
; X64:       # %bb.0:
; X64-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
; X64-NEXT:    vpermt2d %ymm2, %ymm1, %ymm3 # encoding: [0x62,0xf2,0x75,0x28,0x7e,0xda]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpermi2d %ymm2, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x76,0xca]
; X64-NEXT:    vpaddd %ymm3, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfe,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x i32> @llvm.x86.avx512.mask.vpermi2var.d.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 %x3)
  %res1 = call <8 x i32> @llvm.x86.avx512.mask.vpermi2var.d.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 -1)
  %res2 = add <8 x i32> %res, %res1
  ret <8 x i32> %res2
}

declare <8 x i32> @llvm.x86.avx512.mask.vpermt2var.d.256(<8 x i32>, <8 x i32>, <8 x i32>, i8)

define <8 x i32>@test_int_x86_avx512_mask_vpermt2var_d_256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_vpermt2var_d_256:
; X86:       # %bb.0:
; X86-NEXT:    vmovdqa %ymm1, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd9]
; X86-NEXT:    vpermt2d %ymm2, %ymm0, %ymm3 # encoding: [0x62,0xf2,0x7d,0x28,0x7e,0xda]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpermt2d %ymm2, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x7e,0xca]
; X86-NEXT:    vpaddd %ymm3, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfe,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_vpermt2var_d_256:
; X64:       # %bb.0:
; X64-NEXT:    vmovdqa %ymm1, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd9]
; X64-NEXT:    vpermt2d %ymm2, %ymm0, %ymm3 # encoding: [0x62,0xf2,0x7d,0x28,0x7e,0xda]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpermt2d %ymm2, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x7e,0xca]
; X64-NEXT:    vpaddd %ymm3, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfe,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x i32> @llvm.x86.avx512.mask.vpermt2var.d.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 %x3)
  %res1 = call <8 x i32> @llvm.x86.avx512.mask.vpermt2var.d.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 -1)
  %res2 = add <8 x i32> %res, %res1
  ret <8 x i32> %res2
}

declare <8 x i32> @llvm.x86.avx512.maskz.vpermt2var.d.256(<8 x i32>, <8 x i32>, <8 x i32>, i8)

define <8 x i32>@test_int_x86_avx512_maskz_vpermt2var_d_256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_maskz_vpermt2var_d_256:
; X86:       # %bb.0:
; X86-NEXT:    vmovdqa %ymm1, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd9]
; X86-NEXT:    vpermt2d %ymm2, %ymm0, %ymm3 # encoding: [0x62,0xf2,0x7d,0x28,0x7e,0xda]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpermt2d %ymm2, %ymm0, %ymm1 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x7e,0xca]
; X86-NEXT:    vpaddd %ymm3, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfe,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_maskz_vpermt2var_d_256:
; X64:       # %bb.0:
; X64-NEXT:    vmovdqa %ymm1, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd9]
; X64-NEXT:    vpermt2d %ymm2, %ymm0, %ymm3 # encoding: [0x62,0xf2,0x7d,0x28,0x7e,0xda]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpermt2d %ymm2, %ymm0, %ymm1 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x7e,0xca]
; X64-NEXT:    vpaddd %ymm3, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfe,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x i32> @llvm.x86.avx512.maskz.vpermt2var.d.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 %x3)
  %res1 = call <8 x i32> @llvm.x86.avx512.maskz.vpermt2var.d.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 -1)
  %res2 = add <8 x i32> %res, %res1
  ret <8 x i32> %res2
}

declare <2 x double> @llvm.x86.avx512.mask.vpermi2var.pd.128(<2 x double>, <2 x i64>, <2 x double>, i8)

define <2 x double>@test_int_x86_avx512_mask_vpermi2var_pd_128(<2 x double> %x0, <2 x i64> %x1, <2 x double> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_vpermi2var_pd_128:
; X86:       # %bb.0:
; X86-NEXT:    vmovapd %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x28,0xd8]
; X86-NEXT:    vpermt2pd %xmm2, %xmm1, %xmm3 # encoding: [0x62,0xf2,0xf5,0x08,0x7f,0xda]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpermi2pd %xmm2, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x77,0xca]
; X86-NEXT:    vaddpd %xmm3, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0x58,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_vpermi2var_pd_128:
; X64:       # %bb.0:
; X64-NEXT:    vmovapd %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x28,0xd8]
; X64-NEXT:    vpermt2pd %xmm2, %xmm1, %xmm3 # encoding: [0x62,0xf2,0xf5,0x08,0x7f,0xda]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpermi2pd %xmm2, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x77,0xca]
; X64-NEXT:    vaddpd %xmm3, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0x58,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x double> @llvm.x86.avx512.mask.vpermi2var.pd.128(<2 x double> %x0, <2 x i64> %x1, <2 x double> %x2, i8 %x3)
  %res1 = call <2 x double> @llvm.x86.avx512.mask.vpermi2var.pd.128(<2 x double> %x0, <2 x i64> %x1, <2 x double> %x2, i8 -1)
  %res2 = fadd <2 x double> %res, %res1
  ret <2 x double> %res2
}

declare <4 x double> @llvm.x86.avx512.mask.vpermi2var.pd.256(<4 x double>, <4 x i64>, <4 x double>, i8)

define <4 x double>@test_int_x86_avx512_mask_vpermi2var_pd_256(<4 x double> %x0, <4 x i64> %x1, <4 x double> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_vpermi2var_pd_256:
; X86:       # %bb.0:
; X86-NEXT:    vmovapd %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x28,0xd8]
; X86-NEXT:    vpermt2pd %ymm2, %ymm1, %ymm3 # encoding: [0x62,0xf2,0xf5,0x28,0x7f,0xda]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpermi2pd %ymm2, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x77,0xca]
; X86-NEXT:    vaddpd %ymm3, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0x58,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_vpermi2var_pd_256:
; X64:       # %bb.0:
; X64-NEXT:    vmovapd %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x28,0xd8]
; X64-NEXT:    vpermt2pd %ymm2, %ymm1, %ymm3 # encoding: [0x62,0xf2,0xf5,0x28,0x7f,0xda]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpermi2pd %ymm2, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x77,0xca]
; X64-NEXT:    vaddpd %ymm3, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0x58,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x double> @llvm.x86.avx512.mask.vpermi2var.pd.256(<4 x double> %x0, <4 x i64> %x1, <4 x double> %x2, i8 %x3)
  %res1 = call <4 x double> @llvm.x86.avx512.mask.vpermi2var.pd.256(<4 x double> %x0, <4 x i64> %x1, <4 x double> %x2, i8 -1)
  %res2 = fadd <4 x double> %res, %res1
  ret <4 x double> %res2
}

declare <4 x float> @llvm.x86.avx512.mask.vpermi2var.ps.128(<4 x float>, <4 x i32>, <4 x float>, i8)

define <4 x float>@test_int_x86_avx512_mask_vpermi2var_ps_128(<4 x float> %x0, <4 x i32> %x1, <4 x float> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_vpermi2var_ps_128:
; X86:       # %bb.0:
; X86-NEXT:    vmovaps %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x28,0xd8]
; X86-NEXT:    vpermt2ps %xmm2, %xmm1, %xmm3 # encoding: [0x62,0xf2,0x75,0x08,0x7f,0xda]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpermi2ps %xmm2, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x77,0xca]
; X86-NEXT:    vaddps %xmm3, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf0,0x58,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_vpermi2var_ps_128:
; X64:       # %bb.0:
; X64-NEXT:    vmovaps %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x28,0xd8]
; X64-NEXT:    vpermt2ps %xmm2, %xmm1, %xmm3 # encoding: [0x62,0xf2,0x75,0x08,0x7f,0xda]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpermi2ps %xmm2, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x77,0xca]
; X64-NEXT:    vaddps %xmm3, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf0,0x58,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask.vpermi2var.ps.128(<4 x float> %x0, <4 x i32> %x1, <4 x float> %x2, i8 %x3)
  %res1 = call <4 x float> @llvm.x86.avx512.mask.vpermi2var.ps.128(<4 x float> %x0, <4 x i32> %x1, <4 x float> %x2, i8 -1)
  %res2 = fadd <4 x float> %res, %res1
  ret <4 x float> %res2
}

define <4 x float>@test_int_x86_avx512_mask_vpermi2var_ps_128_cast(<4 x float> %x0, <2 x i64> %x1, <4 x float> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_vpermi2var_ps_128_cast:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpermi2ps %xmm2, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x77,0xca]
; X86-NEXT:    vmovaps %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x28,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_vpermi2var_ps_128_cast:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpermi2ps %xmm2, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x77,0xca]
; X64-NEXT:    vmovaps %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x28,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %x1cast = bitcast <2 x i64> %x1 to <4 x i32>
  %res = call <4 x float> @llvm.x86.avx512.mask.vpermi2var.ps.128(<4 x float> %x0, <4 x i32> %x1cast, <4 x float> %x2, i8 %x3)
  ret <4 x float> %res
}

declare <8 x float> @llvm.x86.avx512.mask.vpermi2var.ps.256(<8 x float>, <8 x i32>, <8 x float>, i8)

define <8 x float>@test_int_x86_avx512_mask_vpermi2var_ps_256(<8 x float> %x0, <8 x i32> %x1, <8 x float> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_vpermi2var_ps_256:
; X86:       # %bb.0:
; X86-NEXT:    vmovaps %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x28,0xd8]
; X86-NEXT:    vpermt2ps %ymm2, %ymm1, %ymm3 # encoding: [0x62,0xf2,0x75,0x28,0x7f,0xda]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpermi2ps %ymm2, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x77,0xca]
; X86-NEXT:    vaddps %ymm3, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf4,0x58,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_vpermi2var_ps_256:
; X64:       # %bb.0:
; X64-NEXT:    vmovaps %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x28,0xd8]
; X64-NEXT:    vpermt2ps %ymm2, %ymm1, %ymm3 # encoding: [0x62,0xf2,0x75,0x28,0x7f,0xda]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpermi2ps %ymm2, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x77,0xca]
; X64-NEXT:    vaddps %ymm3, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf4,0x58,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask.vpermi2var.ps.256(<8 x float> %x0, <8 x i32> %x1, <8 x float> %x2, i8 %x3)
  %res1 = call <8 x float> @llvm.x86.avx512.mask.vpermi2var.ps.256(<8 x float> %x0, <8 x i32> %x1, <8 x float> %x2, i8 -1)
  %res2 = fadd <8 x float> %res, %res1
  ret <8 x float> %res2
}

declare <2 x i64> @llvm.x86.avx512.mask.vpermi2var.q.128(<2 x i64>, <2 x i64>, <2 x i64>, i8)

define <2 x i64>@test_int_x86_avx512_mask_vpermi2var_q_128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_vpermi2var_q_128:
; X86:       # %bb.0:
; X86-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
; X86-NEXT:    vpermt2q %xmm2, %xmm1, %xmm3 # encoding: [0x62,0xf2,0xf5,0x08,0x7e,0xda]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpermi2q %xmm2, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x76,0xca]
; X86-NEXT:    vpaddq %xmm3, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xd4,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_vpermi2var_q_128:
; X64:       # %bb.0:
; X64-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
; X64-NEXT:    vpermt2q %xmm2, %xmm1, %xmm3 # encoding: [0x62,0xf2,0xf5,0x08,0x7e,0xda]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpermi2q %xmm2, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x76,0xca]
; X64-NEXT:    vpaddq %xmm3, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xd4,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x i64> @llvm.x86.avx512.mask.vpermi2var.q.128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i8 %x3)
  %res1 = call <2 x i64> @llvm.x86.avx512.mask.vpermi2var.q.128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i8 -1)
  %res2 = add <2 x i64> %res, %res1
  ret <2 x i64> %res2
}

declare <2 x i64> @llvm.x86.avx512.mask.vpermt2var.q.128(<2 x i64>, <2 x i64>, <2 x i64>, i8)

define <2 x i64>@test_int_x86_avx512_mask_vpermt2var_q_128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_vpermt2var_q_128:
; X86:       # %bb.0:
; X86-NEXT:    vmovdqa %xmm1, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd9]
; X86-NEXT:    vpermt2q %xmm2, %xmm0, %xmm3 # encoding: [0x62,0xf2,0xfd,0x08,0x7e,0xda]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpermt2q %xmm2, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x7e,0xca]
; X86-NEXT:    vpaddq %xmm3, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xd4,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_vpermt2var_q_128:
; X64:       # %bb.0:
; X64-NEXT:    vmovdqa %xmm1, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd9]
; X64-NEXT:    vpermt2q %xmm2, %xmm0, %xmm3 # encoding: [0x62,0xf2,0xfd,0x08,0x7e,0xda]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpermt2q %xmm2, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x7e,0xca]
; X64-NEXT:    vpaddq %xmm3, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xd4,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x i64> @llvm.x86.avx512.mask.vpermt2var.q.128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i8 %x3)
  %res1 = call <2 x i64> @llvm.x86.avx512.mask.vpermt2var.q.128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i8 -1)
  %res2 = add <2 x i64> %res, %res1
  ret <2 x i64> %res2
}

declare <2 x i64> @llvm.x86.avx512.maskz.vpermt2var.q.128(<2 x i64>, <2 x i64>, <2 x i64>, i8)

define <2 x i64>@test_int_x86_avx512_maskz_vpermt2var_q_128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_maskz_vpermt2var_q_128:
; X86:       # %bb.0:
; X86-NEXT:    vmovdqa %xmm1, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd9]
; X86-NEXT:    vpermt2q %xmm2, %xmm0, %xmm3 # encoding: [0x62,0xf2,0xfd,0x08,0x7e,0xda]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpermt2q %xmm2, %xmm0, %xmm1 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x7e,0xca]
; X86-NEXT:    vpaddq %xmm3, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xd4,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_maskz_vpermt2var_q_128:
; X64:       # %bb.0:
; X64-NEXT:    vmovdqa %xmm1, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd9]
; X64-NEXT:    vpermt2q %xmm2, %xmm0, %xmm3 # encoding: [0x62,0xf2,0xfd,0x08,0x7e,0xda]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpermt2q %xmm2, %xmm0, %xmm1 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x7e,0xca]
; X64-NEXT:    vpaddq %xmm3, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xd4,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x i64> @llvm.x86.avx512.maskz.vpermt2var.q.128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i8 %x3)
  %res1 = call <2 x i64> @llvm.x86.avx512.maskz.vpermt2var.q.128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i8 -1)
  %res2 = add <2 x i64> %res, %res1
  ret <2 x i64> %res2
}

declare <4 x i64> @llvm.x86.avx512.mask.vpermi2var.q.256(<4 x i64>, <4 x i64>, <4 x i64>, i8)

define <4 x i64>@test_int_x86_avx512_mask_vpermi2var_q_256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_vpermi2var_q_256:
; X86:       # %bb.0:
; X86-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
; X86-NEXT:    vpermt2q %ymm2, %ymm1, %ymm3 # encoding: [0x62,0xf2,0xf5,0x28,0x7e,0xda]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpermi2q %ymm2, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x76,0xca]
; X86-NEXT:    vpaddq %ymm3, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xd4,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_vpermi2var_q_256:
; X64:       # %bb.0:
; X64-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
; X64-NEXT:    vpermt2q %ymm2, %ymm1, %ymm3 # encoding: [0x62,0xf2,0xf5,0x28,0x7e,0xda]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpermi2q %ymm2, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x76,0xca]
; X64-NEXT:    vpaddq %ymm3, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xd4,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i64> @llvm.x86.avx512.mask.vpermi2var.q.256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2, i8 %x3)
  %res1 = call <4 x i64> @llvm.x86.avx512.mask.vpermi2var.q.256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2, i8 -1)
  %res2 = add <4 x i64> %res, %res1
  ret <4 x i64> %res2
}

declare <4 x i64> @llvm.x86.avx512.mask.vpermt2var.q.256(<4 x i64>, <4 x i64>, <4 x i64>, i8)

define <4 x i64>@test_int_x86_avx512_mask_vpermt2var_q_256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_vpermt2var_q_256:
; X86:       # %bb.0:
; X86-NEXT:    vmovdqa %ymm1, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd9]
; X86-NEXT:    vpermt2q %ymm2, %ymm0, %ymm3 # encoding: [0x62,0xf2,0xfd,0x28,0x7e,0xda]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpermt2q %ymm2, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x7e,0xca]
; X86-NEXT:    vpaddq %ymm3, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xd4,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_vpermt2var_q_256:
; X64:       # %bb.0:
; X64-NEXT:    vmovdqa %ymm1, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd9]
; X64-NEXT:    vpermt2q %ymm2, %ymm0, %ymm3 # encoding: [0x62,0xf2,0xfd,0x28,0x7e,0xda]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpermt2q %ymm2, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x7e,0xca]
; X64-NEXT:    vpaddq %ymm3, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xd4,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i64> @llvm.x86.avx512.mask.vpermt2var.q.256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2, i8 %x3)
  %res1 = call <4 x i64> @llvm.x86.avx512.mask.vpermt2var.q.256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2, i8 -1)
  %res2 = add <4 x i64> %res, %res1
  ret <4 x i64> %res2
}

declare <4 x i64> @llvm.x86.avx512.maskz.vpermt2var.q.256(<4 x i64>, <4 x i64>, <4 x i64>, i8)

define <4 x i64>@test_int_x86_avx512_maskz_vpermt2var_q_256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_maskz_vpermt2var_q_256:
; X86:       # %bb.0:
; X86-NEXT:    vmovdqa %ymm1, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd9]
; X86-NEXT:    vpermt2q %ymm2, %ymm0, %ymm3 # encoding: [0x62,0xf2,0xfd,0x28,0x7e,0xda]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpermt2q %ymm2, %ymm0, %ymm1 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x7e,0xca]
; X86-NEXT:    vpaddq %ymm3, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xd4,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_maskz_vpermt2var_q_256:
; X64:       # %bb.0:
; X64-NEXT:    vmovdqa %ymm1, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd9]
; X64-NEXT:    vpermt2q %ymm2, %ymm0, %ymm3 # encoding: [0x62,0xf2,0xfd,0x28,0x7e,0xda]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpermt2q %ymm2, %ymm0, %ymm1 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x7e,0xca]
; X64-NEXT:    vpaddq %ymm3, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xd4,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i64> @llvm.x86.avx512.maskz.vpermt2var.q.256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2, i8 %x3)
  %res1 = call <4 x i64> @llvm.x86.avx512.maskz.vpermt2var.q.256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2, i8 -1)
  %res2 = add <4 x i64> %res, %res1
  ret <4 x i64> %res2
}

define void @test_mask_compress_store_pd_128(i8* %addr, <2 x double> %data, i8 %mask) {
; X86-LABEL: test_mask_compress_store_pd_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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vcompresspd %xmm0, (%eax) {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x8a,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_compress_store_pd_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vcompresspd %xmm0, (%rdi) {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x8a,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  call void @llvm.x86.avx512.mask.compress.store.pd.128(i8* %addr, <2 x double> %data, i8 %mask)
  ret void
}

declare void @llvm.x86.avx512.mask.compress.store.pd.128(i8* %addr, <2 x double> %data, i8 %mask)

define void @test_compress_store_pd_128(i8* %addr, <2 x double> %data) {
; X86-LABEL: test_compress_store_pd_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kxnorw %k0, %k0, %k1 # encoding: [0xc5,0xfc,0x46,0xc8]
; X86-NEXT:    vcompresspd %xmm0, (%eax) {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x8a,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_compress_store_pd_128:
; X64:       # %bb.0:
; X64-NEXT:    kxnorw %k0, %k0, %k1 # encoding: [0xc5,0xfc,0x46,0xc8]
; X64-NEXT:    vcompresspd %xmm0, (%rdi) {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x8a,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  call void @llvm.x86.avx512.mask.compress.store.pd.128(i8* %addr, <2 x double> %data, i8 -1)
  ret void
}

define void @test_mask_compress_store_ps_128(i8* %addr, <4 x float> %data, i8 %mask) {
; X86-LABEL: test_mask_compress_store_ps_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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vcompressps %xmm0, (%eax) {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x8a,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_compress_store_ps_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vcompressps %xmm0, (%rdi) {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x8a,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  call void @llvm.x86.avx512.mask.compress.store.ps.128(i8* %addr, <4 x float> %data, i8 %mask)
  ret void
}

declare void @llvm.x86.avx512.mask.compress.store.ps.128(i8* %addr, <4 x float> %data, i8 %mask)

define void @test_compress_store_ps_128(i8* %addr, <4 x float> %data) {
; X86-LABEL: test_compress_store_ps_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kxnorw %k0, %k0, %k1 # encoding: [0xc5,0xfc,0x46,0xc8]
; X86-NEXT:    vcompressps %xmm0, (%eax) {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x8a,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_compress_store_ps_128:
; X64:       # %bb.0:
; X64-NEXT:    kxnorw %k0, %k0, %k1 # encoding: [0xc5,0xfc,0x46,0xc8]
; X64-NEXT:    vcompressps %xmm0, (%rdi) {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x8a,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  call void @llvm.x86.avx512.mask.compress.store.ps.128(i8* %addr, <4 x float> %data, i8 -1)
  ret void
}

define void @test_mask_compress_store_q_128(i8* %addr, <2 x i64> %data, i8 %mask) {
; X86-LABEL: test_mask_compress_store_q_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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpcompressq %xmm0, (%eax) {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x8b,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_compress_store_q_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpcompressq %xmm0, (%rdi) {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x8b,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  call void @llvm.x86.avx512.mask.compress.store.q.128(i8* %addr, <2 x i64> %data, i8 %mask)
  ret void
}

declare void @llvm.x86.avx512.mask.compress.store.q.128(i8* %addr, <2 x i64> %data, i8 %mask)

define void @test_compress_store_q_128(i8* %addr, <2 x i64> %data) {
; X86-LABEL: test_compress_store_q_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kxnorw %k0, %k0, %k1 # encoding: [0xc5,0xfc,0x46,0xc8]
; X86-NEXT:    vpcompressq %xmm0, (%eax) {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x8b,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_compress_store_q_128:
; X64:       # %bb.0:
; X64-NEXT:    kxnorw %k0, %k0, %k1 # encoding: [0xc5,0xfc,0x46,0xc8]
; X64-NEXT:    vpcompressq %xmm0, (%rdi) {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x8b,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  call void @llvm.x86.avx512.mask.compress.store.q.128(i8* %addr, <2 x i64> %data, i8 -1)
  ret void
}

define void @test_mask_compress_store_d_128(i8* %addr, <4 x i32> %data, i8 %mask) {
; X86-LABEL: test_mask_compress_store_d_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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpcompressd %xmm0, (%eax) {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x8b,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_compress_store_d_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpcompressd %xmm0, (%rdi) {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x8b,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  call void @llvm.x86.avx512.mask.compress.store.d.128(i8* %addr, <4 x i32> %data, i8 %mask)
  ret void
}

declare void @llvm.x86.avx512.mask.compress.store.d.128(i8* %addr, <4 x i32> %data, i8 %mask)

define void @test_compress_store_d_128(i8* %addr, <4 x i32> %data) {
; X86-LABEL: test_compress_store_d_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kxnorw %k0, %k0, %k1 # encoding: [0xc5,0xfc,0x46,0xc8]
; X86-NEXT:    vpcompressd %xmm0, (%eax) {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x8b,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_compress_store_d_128:
; X64:       # %bb.0:
; X64-NEXT:    kxnorw %k0, %k0, %k1 # encoding: [0xc5,0xfc,0x46,0xc8]
; X64-NEXT:    vpcompressd %xmm0, (%rdi) {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x8b,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  call void @llvm.x86.avx512.mask.compress.store.d.128(i8* %addr, <4 x i32> %data, i8 -1)
  ret void
}

define <2 x double> @test_mask_expand_load_pd_128(i8* %addr, <2 x double> %data, i8 %mask) {
; X86-LABEL: test_mask_expand_load_pd_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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vexpandpd (%eax), %xmm0 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x88,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_expand_load_pd_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vexpandpd (%rdi), %xmm0 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x88,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x double> @llvm.x86.avx512.mask.expand.load.pd.128(i8* %addr, <2 x double> %data, i8 %mask)
  ret <2 x double> %res
}

define <2 x double> @test_maskz_expand_load_pd_128(i8* %addr, i8 %mask) {
; X86-LABEL: test_maskz_expand_load_pd_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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vexpandpd (%eax), %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x88,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_maskz_expand_load_pd_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vexpandpd (%rdi), %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x88,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x double> @llvm.x86.avx512.mask.expand.load.pd.128(i8* %addr, <2 x double> zeroinitializer, i8 %mask)
  ret <2 x double> %res
}

declare <2 x double> @llvm.x86.avx512.mask.expand.load.pd.128(i8* %addr, <2 x double> %data, i8 %mask)

define <2 x double> @test_expand_load_pd_128(i8* %addr, <2 x double> %data) {
; X86-LABEL: test_expand_load_pd_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kxnorw %k0, %k0, %k1 # encoding: [0xc5,0xfc,0x46,0xc8]
; X86-NEXT:    vexpandpd (%eax), %xmm0 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x88,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_expand_load_pd_128:
; X64:       # %bb.0:
; X64-NEXT:    kxnorw %k0, %k0, %k1 # encoding: [0xc5,0xfc,0x46,0xc8]
; X64-NEXT:    vexpandpd (%rdi), %xmm0 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x88,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x double> @llvm.x86.avx512.mask.expand.load.pd.128(i8* %addr, <2 x double> %data, i8 -1)
  ret <2 x double> %res
}

define <4 x float> @test_mask_expand_load_ps_128(i8* %addr, <4 x float> %data, i8 %mask) {
; X86-LABEL: test_mask_expand_load_ps_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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vexpandps (%eax), %xmm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x88,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_expand_load_ps_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vexpandps (%rdi), %xmm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x88,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask.expand.load.ps.128(i8* %addr, <4 x float> %data, i8 %mask)
  ret <4 x float> %res
}

define <4 x float> @test_maskz_expand_load_ps_128(i8* %addr, i8 %mask) {
; X86-LABEL: test_maskz_expand_load_ps_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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vexpandps (%eax), %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x88,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_maskz_expand_load_ps_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vexpandps (%rdi), %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x88,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask.expand.load.ps.128(i8* %addr, <4 x float> zeroinitializer, i8 %mask)
  ret <4 x float> %res
}

declare <4 x float> @llvm.x86.avx512.mask.expand.load.ps.128(i8* %addr, <4 x float> %data, i8 %mask)

define <4 x float> @test_expand_load_ps_128(i8* %addr, <4 x float> %data) {
; X86-LABEL: test_expand_load_ps_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kxnorw %k0, %k0, %k1 # encoding: [0xc5,0xfc,0x46,0xc8]
; X86-NEXT:    vexpandps (%eax), %xmm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x88,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_expand_load_ps_128:
; X64:       # %bb.0:
; X64-NEXT:    kxnorw %k0, %k0, %k1 # encoding: [0xc5,0xfc,0x46,0xc8]
; X64-NEXT:    vexpandps (%rdi), %xmm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x88,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask.expand.load.ps.128(i8* %addr, <4 x float> %data, i8 -1)
  ret <4 x float> %res
}

define <2 x i64> @test_mask_expand_load_q_128(i8* %addr, <2 x i64> %data, i8 %mask) {
; X86-LABEL: test_mask_expand_load_q_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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpexpandq (%eax), %xmm0 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x89,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_expand_load_q_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpexpandq (%rdi), %xmm0 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x89,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x i64> @llvm.x86.avx512.mask.expand.load.q.128(i8* %addr, <2 x i64> %data, i8 %mask)
  ret <2 x i64> %res
}

define <2 x i64> @test_maskz_expand_load_q_128(i8* %addr, i8 %mask) {
; X86-LABEL: test_maskz_expand_load_q_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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpexpandq (%eax), %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x89,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_maskz_expand_load_q_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpexpandq (%rdi), %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x89,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x i64> @llvm.x86.avx512.mask.expand.load.q.128(i8* %addr, <2 x i64> zeroinitializer, i8 %mask)
  ret <2 x i64> %res
}

declare <2 x i64> @llvm.x86.avx512.mask.expand.load.q.128(i8* %addr, <2 x i64> %data, i8 %mask)

define <2 x i64> @test_expand_load_q_128(i8* %addr, <2 x i64> %data) {
; X86-LABEL: test_expand_load_q_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kxnorw %k0, %k0, %k1 # encoding: [0xc5,0xfc,0x46,0xc8]
; X86-NEXT:    vpexpandq (%eax), %xmm0 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x89,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_expand_load_q_128:
; X64:       # %bb.0:
; X64-NEXT:    kxnorw %k0, %k0, %k1 # encoding: [0xc5,0xfc,0x46,0xc8]
; X64-NEXT:    vpexpandq (%rdi), %xmm0 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x89,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x i64> @llvm.x86.avx512.mask.expand.load.q.128(i8* %addr, <2 x i64> %data, i8 -1)
  ret <2 x i64> %res
}

define <4 x i32> @test_mask_expand_load_d_128(i8* %addr, <4 x i32> %data, i8 %mask) {
; X86-LABEL: test_mask_expand_load_d_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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpexpandd (%eax), %xmm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x89,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_expand_load_d_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpexpandd (%rdi), %xmm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x89,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i32> @llvm.x86.avx512.mask.expand.load.d.128(i8* %addr, <4 x i32> %data, i8 %mask)
  ret <4 x i32> %res
}

define <4 x i32> @test_maskz_expand_load_d_128(i8* %addr, i8 %mask) {
; X86-LABEL: test_maskz_expand_load_d_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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpexpandd (%eax), %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x89,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_maskz_expand_load_d_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpexpandd (%rdi), %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x89,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i32> @llvm.x86.avx512.mask.expand.load.d.128(i8* %addr, <4 x i32> zeroinitializer, i8 %mask)
  ret <4 x i32> %res
}

declare <4 x i32> @llvm.x86.avx512.mask.expand.load.d.128(i8* %addr, <4 x i32> %data, i8 %mask)

define <4 x i32> @test_expand_load_d_128(i8* %addr, <4 x i32> %data) {
; X86-LABEL: test_expand_load_d_128:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kxnorw %k0, %k0, %k1 # encoding: [0xc5,0xfc,0x46,0xc8]
; X86-NEXT:    vpexpandd (%eax), %xmm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x89,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_expand_load_d_128:
; X64:       # %bb.0:
; X64-NEXT:    kxnorw %k0, %k0, %k1 # encoding: [0xc5,0xfc,0x46,0xc8]
; X64-NEXT:    vpexpandd (%rdi), %xmm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x89,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i32> @llvm.x86.avx512.mask.expand.load.d.128(i8* %addr, <4 x i32> %data, i8 -1)
  ret <4 x i32> %res
}

define void @test_mask_compress_store_pd_256(i8* %addr, <4 x double> %data, i8 %mask) {
; X86-LABEL: test_mask_compress_store_pd_256:
; 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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vcompresspd %ymm0, (%eax) {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x8a,0x00]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_compress_store_pd_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vcompresspd %ymm0, (%rdi) {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x8a,0x07]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
  call void @llvm.x86.avx512.mask.compress.store.pd.256(i8* %addr, <4 x double> %data, i8 %mask)
  ret void
}

declare void @llvm.x86.avx512.mask.compress.store.pd.256(i8* %addr, <4 x double> %data, i8 %mask)

define void @test_compress_store_pd_256(i8* %addr, <4 x double> %data) {
; X86-LABEL: test_compress_store_pd_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kxnorw %k0, %k0, %k1 # encoding: [0xc5,0xfc,0x46,0xc8]
; X86-NEXT:    vcompresspd %ymm0, (%eax) {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x8a,0x00]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_compress_store_pd_256:
; X64:       # %bb.0:
; X64-NEXT:    kxnorw %k0, %k0, %k1 # encoding: [0xc5,0xfc,0x46,0xc8]
; X64-NEXT:    vcompresspd %ymm0, (%rdi) {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x8a,0x07]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
  call void @llvm.x86.avx512.mask.compress.store.pd.256(i8* %addr, <4 x double> %data, i8 -1)
  ret void
}

define void @test_mask_compress_store_ps_256(i8* %addr, <8 x float> %data, i8 %mask) {
; X86-LABEL: test_mask_compress_store_ps_256:
; 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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vcompressps %ymm0, (%eax) {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x8a,0x00]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_compress_store_ps_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vcompressps %ymm0, (%rdi) {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x8a,0x07]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
  call void @llvm.x86.avx512.mask.compress.store.ps.256(i8* %addr, <8 x float> %data, i8 %mask)
  ret void
}

declare void @llvm.x86.avx512.mask.compress.store.ps.256(i8* %addr, <8 x float> %data, i8 %mask)

define void @test_compress_store_ps_256(i8* %addr, <8 x float> %data) {
; X86-LABEL: test_compress_store_ps_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kxnorw %k0, %k0, %k1 # encoding: [0xc5,0xfc,0x46,0xc8]
; X86-NEXT:    vcompressps %ymm0, (%eax) {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x8a,0x00]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_compress_store_ps_256:
; X64:       # %bb.0:
; X64-NEXT:    kxnorw %k0, %k0, %k1 # encoding: [0xc5,0xfc,0x46,0xc8]
; X64-NEXT:    vcompressps %ymm0, (%rdi) {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x8a,0x07]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
  call void @llvm.x86.avx512.mask.compress.store.ps.256(i8* %addr, <8 x float> %data, i8 -1)
  ret void
}

define void @test_mask_compress_store_q_256(i8* %addr, <4 x i64> %data, i8 %mask) {
; X86-LABEL: test_mask_compress_store_q_256:
; 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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpcompressq %ymm0, (%eax) {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x8b,0x00]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_compress_store_q_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpcompressq %ymm0, (%rdi) {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x8b,0x07]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
  call void @llvm.x86.avx512.mask.compress.store.q.256(i8* %addr, <4 x i64> %data, i8 %mask)
  ret void
}

declare void @llvm.x86.avx512.mask.compress.store.q.256(i8* %addr, <4 x i64> %data, i8 %mask)

define void @test_compress_store_q_256(i8* %addr, <4 x i64> %data) {
; X86-LABEL: test_compress_store_q_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kxnorw %k0, %k0, %k1 # encoding: [0xc5,0xfc,0x46,0xc8]
; X86-NEXT:    vpcompressq %ymm0, (%eax) {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x8b,0x00]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_compress_store_q_256:
; X64:       # %bb.0:
; X64-NEXT:    kxnorw %k0, %k0, %k1 # encoding: [0xc5,0xfc,0x46,0xc8]
; X64-NEXT:    vpcompressq %ymm0, (%rdi) {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x8b,0x07]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
  call void @llvm.x86.avx512.mask.compress.store.q.256(i8* %addr, <4 x i64> %data, i8 -1)
  ret void
}

define void @test_mask_compress_store_d_256(i8* %addr, <8 x i32> %data, i8 %mask) {
; X86-LABEL: test_mask_compress_store_d_256:
; 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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpcompressd %ymm0, (%eax) {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x8b,0x00]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_compress_store_d_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpcompressd %ymm0, (%rdi) {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x8b,0x07]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
  call void @llvm.x86.avx512.mask.compress.store.d.256(i8* %addr, <8 x i32> %data, i8 %mask)
  ret void
}

declare void @llvm.x86.avx512.mask.compress.store.d.256(i8* %addr, <8 x i32> %data, i8 %mask)

define void @test_compress_store_d_256(i8* %addr, <8 x i32> %data) {
; X86-LABEL: test_compress_store_d_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kxnorw %k0, %k0, %k1 # encoding: [0xc5,0xfc,0x46,0xc8]
; X86-NEXT:    vpcompressd %ymm0, (%eax) {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x8b,0x00]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_compress_store_d_256:
; X64:       # %bb.0:
; X64-NEXT:    kxnorw %k0, %k0, %k1 # encoding: [0xc5,0xfc,0x46,0xc8]
; X64-NEXT:    vpcompressd %ymm0, (%rdi) {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x8b,0x07]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
  call void @llvm.x86.avx512.mask.compress.store.d.256(i8* %addr, <8 x i32> %data, i8 -1)
  ret void
}

define <4 x double> @test_mask_expand_load_pd_256(i8* %addr, <4 x double> %data, i8 %mask) {
; X86-LABEL: test_mask_expand_load_pd_256:
; 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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vexpandpd (%eax), %ymm0 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x88,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_expand_load_pd_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vexpandpd (%rdi), %ymm0 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x88,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x double> @llvm.x86.avx512.mask.expand.load.pd.256(i8* %addr, <4 x double> %data, i8 %mask)
  ret <4 x double> %res
}

define <4 x double> @test_maskz_expand_load_pd_256(i8* %addr, i8 %mask) {
; X86-LABEL: test_maskz_expand_load_pd_256:
; 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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vexpandpd (%eax), %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x88,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_maskz_expand_load_pd_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vexpandpd (%rdi), %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x88,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x double> @llvm.x86.avx512.mask.expand.load.pd.256(i8* %addr, <4 x double> zeroinitializer, i8 %mask)
  ret <4 x double> %res
}

declare <4 x double> @llvm.x86.avx512.mask.expand.load.pd.256(i8* %addr, <4 x double> %data, i8 %mask)

define <4 x double> @test_expand_load_pd_256(i8* %addr, <4 x double> %data) {
; X86-LABEL: test_expand_load_pd_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kxnorw %k0, %k0, %k1 # encoding: [0xc5,0xfc,0x46,0xc8]
; X86-NEXT:    vexpandpd (%eax), %ymm0 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x88,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_expand_load_pd_256:
; X64:       # %bb.0:
; X64-NEXT:    kxnorw %k0, %k0, %k1 # encoding: [0xc5,0xfc,0x46,0xc8]
; X64-NEXT:    vexpandpd (%rdi), %ymm0 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x88,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x double> @llvm.x86.avx512.mask.expand.load.pd.256(i8* %addr, <4 x double> %data, i8 -1)
  ret <4 x double> %res
}

define <8 x float> @test_mask_expand_load_ps_256(i8* %addr, <8 x float> %data, i8 %mask) {
; X86-LABEL: test_mask_expand_load_ps_256:
; 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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vexpandps (%eax), %ymm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x88,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_expand_load_ps_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vexpandps (%rdi), %ymm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x88,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask.expand.load.ps.256(i8* %addr, <8 x float> %data, i8 %mask)
  ret <8 x float> %res
}

define <8 x float> @test_maskz_expand_load_ps_256(i8* %addr, i8 %mask) {
; X86-LABEL: test_maskz_expand_load_ps_256:
; 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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vexpandps (%eax), %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x88,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_maskz_expand_load_ps_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vexpandps (%rdi), %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x88,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask.expand.load.ps.256(i8* %addr, <8 x float> zeroinitializer, i8 %mask)
  ret <8 x float> %res
}

declare <8 x float> @llvm.x86.avx512.mask.expand.load.ps.256(i8* %addr, <8 x float> %data, i8 %mask)

define <8 x float> @test_expand_load_ps_256(i8* %addr, <8 x float> %data) {
; X86-LABEL: test_expand_load_ps_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kxnorw %k0, %k0, %k1 # encoding: [0xc5,0xfc,0x46,0xc8]
; X86-NEXT:    vexpandps (%eax), %ymm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x88,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_expand_load_ps_256:
; X64:       # %bb.0:
; X64-NEXT:    kxnorw %k0, %k0, %k1 # encoding: [0xc5,0xfc,0x46,0xc8]
; X64-NEXT:    vexpandps (%rdi), %ymm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x88,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask.expand.load.ps.256(i8* %addr, <8 x float> %data, i8 -1)
  ret <8 x float> %res
}

define <4 x i64> @test_mask_expand_load_q_256(i8* %addr, <4 x i64> %data, i8 %mask) {
; X86-LABEL: test_mask_expand_load_q_256:
; 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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpexpandq (%eax), %ymm0 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x89,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_expand_load_q_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpexpandq (%rdi), %ymm0 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x89,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i64> @llvm.x86.avx512.mask.expand.load.q.256(i8* %addr, <4 x i64> %data, i8 %mask)
  ret <4 x i64> %res
}

define <4 x i64> @test_maskz_expand_load_q_256(i8* %addr, i8 %mask) {
; X86-LABEL: test_maskz_expand_load_q_256:
; 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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpexpandq (%eax), %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x89,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_maskz_expand_load_q_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpexpandq (%rdi), %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x89,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i64> @llvm.x86.avx512.mask.expand.load.q.256(i8* %addr, <4 x i64> zeroinitializer, i8 %mask)
  ret <4 x i64> %res
}

declare <4 x i64> @llvm.x86.avx512.mask.expand.load.q.256(i8* %addr, <4 x i64> %data, i8 %mask)

define <4 x i64> @test_expand_load_q_256(i8* %addr, <4 x i64> %data) {
; X86-LABEL: test_expand_load_q_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kxnorw %k0, %k0, %k1 # encoding: [0xc5,0xfc,0x46,0xc8]
; X86-NEXT:    vpexpandq (%eax), %ymm0 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x89,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_expand_load_q_256:
; X64:       # %bb.0:
; X64-NEXT:    kxnorw %k0, %k0, %k1 # encoding: [0xc5,0xfc,0x46,0xc8]
; X64-NEXT:    vpexpandq (%rdi), %ymm0 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x89,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i64> @llvm.x86.avx512.mask.expand.load.q.256(i8* %addr, <4 x i64> %data, i8 -1)
  ret <4 x i64> %res
}

define <8 x i32> @test_mask_expand_load_d_256(i8* %addr, <8 x i32> %data, i8 %mask) {
; X86-LABEL: test_mask_expand_load_d_256:
; 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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpexpandd (%eax), %ymm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x89,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_expand_load_d_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpexpandd (%rdi), %ymm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x89,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x i32> @llvm.x86.avx512.mask.expand.load.d.256(i8* %addr, <8 x i32> %data, i8 %mask)
  ret <8 x i32> %res
}

define <8 x i32> @test_maskz_expand_load_d_256(i8* %addr, i8 %mask) {
; X86-LABEL: test_maskz_expand_load_d_256:
; 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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vpexpandd (%eax), %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x89,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_maskz_expand_load_d_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpexpandd (%rdi), %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x89,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x i32> @llvm.x86.avx512.mask.expand.load.d.256(i8* %addr, <8 x i32> zeroinitializer, i8 %mask)
  ret <8 x i32> %res
}

declare <8 x i32> @llvm.x86.avx512.mask.expand.load.d.256(i8* %addr, <8 x i32> %data, i8 %mask)

define <8 x i32> @test_expand_load_d_256(i8* %addr, <8 x i32> %data) {
; X86-LABEL: test_expand_load_d_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kxnorw %k0, %k0, %k1 # encoding: [0xc5,0xfc,0x46,0xc8]
; X86-NEXT:    vpexpandd (%eax), %ymm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x89,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_expand_load_d_256:
; X64:       # %bb.0:
; X64-NEXT:    kxnorw %k0, %k0, %k1 # encoding: [0xc5,0xfc,0x46,0xc8]
; X64-NEXT:    vpexpandd (%rdi), %ymm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x89,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x i32> @llvm.x86.avx512.mask.expand.load.d.256(i8* %addr, <8 x i32> %data, i8 -1)
  ret <8 x i32> %res
}

define <4 x double> @test_sqrt_pd_256(<4 x double> %a0, i8 %mask) {
; X86-LABEL: test_sqrt_pd_256:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vsqrtpd %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0xa9,0x51,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_sqrt_pd_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vsqrtpd %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0xa9,0x51,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x double> @llvm.x86.avx512.mask.sqrt.pd.256(<4 x double> %a0,  <4 x double> zeroinitializer, i8 %mask)
  ret <4 x double> %res
}
declare <4 x double> @llvm.x86.avx512.mask.sqrt.pd.256(<4 x double>, <4 x double>, i8) nounwind readnone

define <8 x float> @test_sqrt_ps_256(<8 x float> %a0, i8 %mask) {
; X86-LABEL: test_sqrt_ps_256:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vsqrtps %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7c,0xa9,0x51,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_sqrt_ps_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vsqrtps %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7c,0xa9,0x51,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask.sqrt.ps.256(<8 x float> %a0, <8 x float> zeroinitializer, i8 %mask)
  ret <8 x float> %res
}

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

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

define <4 x i32>@test_int_x86_avx512_mask_prorv_d_128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_prorv_d_128:
; X86:       # %bb.0:
; X86-NEXT:    vprorvd %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf2,0x7d,0x08,0x14,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vprorvd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x14,0xd1]
; X86-NEXT:    vprorvd %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x14,0xc1]
; X86-NEXT:    vpaddd %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfe,0xc3]
; X86-NEXT:    vpaddd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_prorv_d_128:
; X64:       # %bb.0:
; X64-NEXT:    vprorvd %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf2,0x7d,0x08,0x14,0xd9]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vprorvd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x14,0xd1]
; X64-NEXT:    vprorvd %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x14,0xc1]
; X64-NEXT:    vpaddd %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfe,0xc3]
; X64-NEXT:    vpaddd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i32> @llvm.x86.avx512.mask.prorv.d.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 %x3)
  %res1 = call <4 x i32> @llvm.x86.avx512.mask.prorv.d.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> zeroinitializer, i8 %x3)
  %res2 = call <4 x i32> @llvm.x86.avx512.mask.prorv.d.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 -1)
  %res3 = add <4 x i32> %res, %res1
  %res4 = add <4 x i32> %res3, %res2
  ret <4 x i32> %res4
}

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

define <8 x i32>@test_int_x86_avx512_mask_prorv_d_256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_prorv_d_256:
; X86:       # %bb.0:
; X86-NEXT:    vprorvd %ymm1, %ymm0, %ymm3 # encoding: [0x62,0xf2,0x7d,0x28,0x14,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vprorvd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x14,0xd1]
; X86-NEXT:    vprorvd %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x14,0xc1]
; X86-NEXT:    vpaddd %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfe,0xc3]
; X86-NEXT:    vpaddd %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_prorv_d_256:
; X64:       # %bb.0:
; X64-NEXT:    vprorvd %ymm1, %ymm0, %ymm3 # encoding: [0x62,0xf2,0x7d,0x28,0x14,0xd9]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vprorvd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x14,0xd1]
; X64-NEXT:    vprorvd %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x14,0xc1]
; X64-NEXT:    vpaddd %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfe,0xc3]
; X64-NEXT:    vpaddd %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x i32> @llvm.x86.avx512.mask.prorv.d.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 %x3)
  %res1 = call <8 x i32> @llvm.x86.avx512.mask.prorv.d.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> zeroinitializer, i8 %x3)
  %res2 = call <8 x i32> @llvm.x86.avx512.mask.prorv.d.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 -1)
  %res3 = add <8 x i32> %res, %res1
  %res4 = add <8 x i32> %res3, %res2
  ret <8 x i32> %res4
}

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

define <2 x i64>@test_int_x86_avx512_mask_prorv_q_128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_prorv_q_128:
; X86:       # %bb.0:
; X86-NEXT:    vprorvq %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf2,0xfd,0x08,0x14,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vprorvq %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x14,0xd1]
; X86-NEXT:    vprorvq %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x14,0xc1]
; X86-NEXT:    vpaddq %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd4,0xc3]
; X86-NEXT:    vpaddq %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xd4,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_prorv_q_128:
; X64:       # %bb.0:
; X64-NEXT:    vprorvq %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf2,0xfd,0x08,0x14,0xd9]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vprorvq %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x14,0xd1]
; X64-NEXT:    vprorvq %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x14,0xc1]
; X64-NEXT:    vpaddq %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd4,0xc3]
; X64-NEXT:    vpaddq %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xd4,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x i64> @llvm.x86.avx512.mask.prorv.q.128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i8 %x3)
  %res1 = call <2 x i64> @llvm.x86.avx512.mask.prorv.q.128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> zeroinitializer, i8 %x3)
  %res2 = call <2 x i64> @llvm.x86.avx512.mask.prorv.q.128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, 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.prorv.q.256(<4 x i64>, <4 x i64>, <4 x i64>, i8)

define <4 x i64>@test_int_x86_avx512_mask_prorv_q_256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_prorv_q_256:
; X86:       # %bb.0:
; X86-NEXT:    vprorvq %ymm1, %ymm0, %ymm3 # encoding: [0x62,0xf2,0xfd,0x28,0x14,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vprorvq %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x14,0xd1]
; X86-NEXT:    vprorvq %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x14,0xc1]
; X86-NEXT:    vpaddq %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd4,0xc3]
; X86-NEXT:    vpaddq %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xd4,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_prorv_q_256:
; X64:       # %bb.0:
; X64-NEXT:    vprorvq %ymm1, %ymm0, %ymm3 # encoding: [0x62,0xf2,0xfd,0x28,0x14,0xd9]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vprorvq %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x14,0xd1]
; X64-NEXT:    vprorvq %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x14,0xc1]
; X64-NEXT:    vpaddq %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd4,0xc3]
; X64-NEXT:    vpaddq %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xd4,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i64> @llvm.x86.avx512.mask.prorv.q.256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2, i8 %x3)
  %res1 = call <4 x i64> @llvm.x86.avx512.mask.prorv.q.256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> zeroinitializer, i8 %x3)
  %res2 = call <4 x i64> @llvm.x86.avx512.mask.prorv.q.256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2, i8 -1)
  %res3 = add <4 x i64> %res, %res1
  %res4 = add <4 x i64> %res3, %res2
  ret <4 x i64> %res4
}

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

define <4 x i32>@test_int_x86_avx512_mask_prol_d_128(<4 x i32> %x0, i32 %x1, <4 x i32> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_prol_d_128:
; X86:       # %bb.0:
; X86-NEXT:    vprold $3, %xmm0, %xmm2 # encoding: [0x62,0xf1,0x6d,0x08,0x72,0xc8,0x03]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vprold $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x72,0xc8,0x03]
; X86-NEXT:    vprold $3, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x72,0xc8,0x03]
; X86-NEXT:    vpaddd %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfe,0xc2]
; X86-NEXT:    vpaddd %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_prol_d_128:
; X64:       # %bb.0:
; X64-NEXT:    vprold $3, %xmm0, %xmm2 # encoding: [0x62,0xf1,0x6d,0x08,0x72,0xc8,0x03]
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vprold $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x72,0xc8,0x03]
; X64-NEXT:    vprold $3, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x72,0xc8,0x03]
; X64-NEXT:    vpaddd %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfe,0xc2]
; X64-NEXT:    vpaddd %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i32> @llvm.x86.avx512.mask.prol.d.128(<4 x i32> %x0, i32 3, <4 x i32> %x2, i8 %x3)
  %res1 = call <4 x i32> @llvm.x86.avx512.mask.prol.d.128(<4 x i32> %x0, i32 3, <4 x i32> zeroinitializer, i8 %x3)
  %res2 = call <4 x i32> @llvm.x86.avx512.mask.prol.d.128(<4 x i32> %x0, i32 3, <4 x i32> %x2, i8 -1)
  %res3 = add <4 x i32> %res, %res1
  %res4 = add <4 x i32> %res3, %res2
  ret <4 x i32> %res4
}

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

define <8 x i32>@test_int_x86_avx512_mask_prol_d_256(<8 x i32> %x0, i32 %x1, <8 x i32> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_prol_d_256:
; X86:       # %bb.0:
; X86-NEXT:    vprold $3, %ymm0, %ymm2 # encoding: [0x62,0xf1,0x6d,0x28,0x72,0xc8,0x03]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vprold $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x72,0xc8,0x03]
; X86-NEXT:    vprold $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x72,0xc8,0x03]
; X86-NEXT:    vpaddd %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfe,0xc2]
; X86-NEXT:    vpaddd %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_prol_d_256:
; X64:       # %bb.0:
; X64-NEXT:    vprold $3, %ymm0, %ymm2 # encoding: [0x62,0xf1,0x6d,0x28,0x72,0xc8,0x03]
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vprold $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x72,0xc8,0x03]
; X64-NEXT:    vprold $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x72,0xc8,0x03]
; X64-NEXT:    vpaddd %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfe,0xc2]
; X64-NEXT:    vpaddd %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x i32> @llvm.x86.avx512.mask.prol.d.256(<8 x i32> %x0, i32 3, <8 x i32> %x2, i8 %x3)
  %res1 = call <8 x i32> @llvm.x86.avx512.mask.prol.d.256(<8 x i32> %x0, i32 3, <8 x i32> zeroinitializer, i8 %x3)
  %res2 = call <8 x i32> @llvm.x86.avx512.mask.prol.d.256(<8 x i32> %x0, i32 3, <8 x i32> %x2, i8 -1)
  %res3 = add <8 x i32> %res, %res1
  %res4 = add <8 x i32> %res3, %res2
  ret <8 x i32> %res4
}

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

define <2 x i64>@test_int_x86_avx512_mask_prol_q_128(<2 x i64> %x0, i32 %x1, <2 x i64> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_prol_q_128:
; X86:       # %bb.0:
; X86-NEXT:    vprolq $3, %xmm0, %xmm2 # encoding: [0x62,0xf1,0xed,0x08,0x72,0xc8,0x03]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vprolq $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x09,0x72,0xc8,0x03]
; X86-NEXT:    vprolq $3, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0x89,0x72,0xc8,0x03]
; 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_prol_q_128:
; X64:       # %bb.0:
; X64-NEXT:    vprolq $3, %xmm0, %xmm2 # encoding: [0x62,0xf1,0xed,0x08,0x72,0xc8,0x03]
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vprolq $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x09,0x72,0xc8,0x03]
; X64-NEXT:    vprolq $3, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0x89,0x72,0xc8,0x03]
; 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.prol.q.128(<2 x i64> %x0, i32 3, <2 x i64> %x2, i8 %x3)
  %res1 = call <2 x i64> @llvm.x86.avx512.mask.prol.q.128(<2 x i64> %x0, i32 3, <2 x i64> zeroinitializer, i8 %x3)
  %res2 = call <2 x i64> @llvm.x86.avx512.mask.prol.q.128(<2 x i64> %x0, i32 3, <2 x i64> %x2, 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.prol.q.256(<4 x i64>, i32, <4 x i64>, i8)

define <4 x i64>@test_int_x86_avx512_mask_prol_q_256(<4 x i64> %x0, i32 %x1, <4 x i64> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_prol_q_256:
; X86:       # %bb.0:
; X86-NEXT:    vprolq $3, %ymm0, %ymm2 # encoding: [0x62,0xf1,0xed,0x28,0x72,0xc8,0x03]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vprolq $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x29,0x72,0xc8,0x03]
; X86-NEXT:    vprolq $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0xa9,0x72,0xc8,0x03]
; 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_prol_q_256:
; X64:       # %bb.0:
; X64-NEXT:    vprolq $3, %ymm0, %ymm2 # encoding: [0x62,0xf1,0xed,0x28,0x72,0xc8,0x03]
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vprolq $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x29,0x72,0xc8,0x03]
; X64-NEXT:    vprolq $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0xa9,0x72,0xc8,0x03]
; 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.prol.q.256(<4 x i64> %x0, i32 3, <4 x i64> %x2, i8 %x3)
  %res1 = call <4 x i64> @llvm.x86.avx512.mask.prol.q.256(<4 x i64> %x0, i32 3, <4 x i64> zeroinitializer, i8 %x3)
  %res2 = call <4 x i64> @llvm.x86.avx512.mask.prol.q.256(<4 x i64> %x0, i32 3, <4 x i64> %x2, i8 -1)
  %res3 = add <4 x i64> %res, %res1
  %res4 = add <4 x i64> %res3, %res2
  ret <4 x i64> %res4
}

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

define <4 x i32>@test_int_x86_avx512_mask_prolv_d_128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_prolv_d_128:
; X86:       # %bb.0:
; X86-NEXT:    vprolvd %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf2,0x7d,0x08,0x15,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vprolvd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x15,0xd1]
; X86-NEXT:    vprolvd %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x15,0xc1]
; X86-NEXT:    vpaddd %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfe,0xc3]
; X86-NEXT:    vpaddd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_prolv_d_128:
; X64:       # %bb.0:
; X64-NEXT:    vprolvd %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf2,0x7d,0x08,0x15,0xd9]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vprolvd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x15,0xd1]
; X64-NEXT:    vprolvd %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x15,0xc1]
; X64-NEXT:    vpaddd %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfe,0xc3]
; X64-NEXT:    vpaddd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i32> @llvm.x86.avx512.mask.prolv.d.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 %x3)
  %res1 = call <4 x i32> @llvm.x86.avx512.mask.prolv.d.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> zeroinitializer, i8 %x3)
  %res2 = call <4 x i32> @llvm.x86.avx512.mask.prolv.d.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 -1)
  %res3 = add <4 x i32> %res, %res1
  %res4 = add <4 x i32> %res3, %res2
  ret <4 x i32> %res4
}

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

define <8 x i32>@test_int_x86_avx512_mask_prolv_d_256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_prolv_d_256:
; X86:       # %bb.0:
; X86-NEXT:    vprolvd %ymm1, %ymm0, %ymm3 # encoding: [0x62,0xf2,0x7d,0x28,0x15,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vprolvd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x15,0xd1]
; X86-NEXT:    vprolvd %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x15,0xc1]
; X86-NEXT:    vpaddd %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfe,0xc3]
; X86-NEXT:    vpaddd %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_prolv_d_256:
; X64:       # %bb.0:
; X64-NEXT:    vprolvd %ymm1, %ymm0, %ymm3 # encoding: [0x62,0xf2,0x7d,0x28,0x15,0xd9]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vprolvd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x15,0xd1]
; X64-NEXT:    vprolvd %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x15,0xc1]
; X64-NEXT:    vpaddd %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfe,0xc3]
; X64-NEXT:    vpaddd %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x i32> @llvm.x86.avx512.mask.prolv.d.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 %x3)
  %res1 = call <8 x i32> @llvm.x86.avx512.mask.prolv.d.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> zeroinitializer, i8 %x3)
  %res2 = call <8 x i32> @llvm.x86.avx512.mask.prolv.d.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 -1)
  %res3 = add <8 x i32> %res, %res1
  %res4 = add <8 x i32> %res3, %res2
  ret <8 x i32> %res4
}

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

define <2 x i64>@test_int_x86_avx512_mask_prolv_q_128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_prolv_q_128:
; X86:       # %bb.0:
; X86-NEXT:    vprolvq %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf2,0xfd,0x08,0x15,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vprolvq %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x15,0xd1]
; X86-NEXT:    vprolvq %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x15,0xc1]
; X86-NEXT:    vpaddq %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd4,0xc3]
; X86-NEXT:    vpaddq %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xd4,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_prolv_q_128:
; X64:       # %bb.0:
; X64-NEXT:    vprolvq %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf2,0xfd,0x08,0x15,0xd9]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vprolvq %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x15,0xd1]
; X64-NEXT:    vprolvq %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x15,0xc1]
; X64-NEXT:    vpaddq %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd4,0xc3]
; X64-NEXT:    vpaddq %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xd4,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x i64> @llvm.x86.avx512.mask.prolv.q.128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i8 %x3)
  %res1 = call <2 x i64> @llvm.x86.avx512.mask.prolv.q.128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> zeroinitializer, i8 %x3)
  %res2 = call <2 x i64> @llvm.x86.avx512.mask.prolv.q.128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, 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.prolv.q.256(<4 x i64>, <4 x i64>, <4 x i64>, i8)

define <4 x i64>@test_int_x86_avx512_mask_prolv_q_256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_prolv_q_256:
; X86:       # %bb.0:
; X86-NEXT:    vprolvq %ymm1, %ymm0, %ymm3 # encoding: [0x62,0xf2,0xfd,0x28,0x15,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vprolvq %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x15,0xd1]
; X86-NEXT:    vprolvq %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x15,0xc1]
; X86-NEXT:    vpaddq %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd4,0xc3]
; X86-NEXT:    vpaddq %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xd4,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_prolv_q_256:
; X64:       # %bb.0:
; X64-NEXT:    vprolvq %ymm1, %ymm0, %ymm3 # encoding: [0x62,0xf2,0xfd,0x28,0x15,0xd9]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vprolvq %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x15,0xd1]
; X64-NEXT:    vprolvq %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x15,0xc1]
; X64-NEXT:    vpaddq %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd4,0xc3]
; X64-NEXT:    vpaddq %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xd4,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i64> @llvm.x86.avx512.mask.prolv.q.256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2, i8 %x3)
  %res1 = call <4 x i64> @llvm.x86.avx512.mask.prolv.q.256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> zeroinitializer, i8 %x3)
  %res2 = call <4 x i64> @llvm.x86.avx512.mask.prolv.q.256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2, i8 -1)
  %res3 = add <4 x i64> %res, %res1
  %res4 = add <4 x i64> %res3, %res2
  ret <4 x i64> %res4
}

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

define <4 x i32>@test_int_x86_avx512_mask_pror_d_128(<4 x i32> %x0, i32 %x1, <4 x i32> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_pror_d_128:
; X86:       # %bb.0:
; X86-NEXT:    vprord $3, %xmm0, %xmm2 # encoding: [0x62,0xf1,0x6d,0x08,0x72,0xc0,0x03]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vprord $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x72,0xc0,0x03]
; X86-NEXT:    vprord $3, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x72,0xc0,0x03]
; X86-NEXT:    vpaddd %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfe,0xc2]
; X86-NEXT:    vpaddd %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pror_d_128:
; X64:       # %bb.0:
; X64-NEXT:    vprord $3, %xmm0, %xmm2 # encoding: [0x62,0xf1,0x6d,0x08,0x72,0xc0,0x03]
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vprord $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x72,0xc0,0x03]
; X64-NEXT:    vprord $3, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x72,0xc0,0x03]
; X64-NEXT:    vpaddd %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfe,0xc2]
; X64-NEXT:    vpaddd %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i32> @llvm.x86.avx512.mask.pror.d.128(<4 x i32> %x0, i32 3, <4 x i32> %x2, i8 %x3)
  %res1 = call <4 x i32> @llvm.x86.avx512.mask.pror.d.128(<4 x i32> %x0, i32 3, <4 x i32> zeroinitializer, i8 %x3)
  %res2 = call <4 x i32> @llvm.x86.avx512.mask.pror.d.128(<4 x i32> %x0, i32 3, <4 x i32> %x2, i8 -1)
  %res3 = add <4 x i32> %res, %res1
  %res4 = add <4 x i32> %res3, %res2
  ret <4 x i32> %res4
}

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

define <8 x i32>@test_int_x86_avx512_mask_pror_d_256(<8 x i32> %x0, i32 %x1, <8 x i32> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_pror_d_256:
; X86:       # %bb.0:
; X86-NEXT:    vprord $3, %ymm0, %ymm2 # encoding: [0x62,0xf1,0x6d,0x28,0x72,0xc0,0x03]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vprord $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x72,0xc0,0x03]
; X86-NEXT:    vprord $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x72,0xc0,0x03]
; X86-NEXT:    vpaddd %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfe,0xc2]
; X86-NEXT:    vpaddd %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pror_d_256:
; X64:       # %bb.0:
; X64-NEXT:    vprord $3, %ymm0, %ymm2 # encoding: [0x62,0xf1,0x6d,0x28,0x72,0xc0,0x03]
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vprord $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x72,0xc0,0x03]
; X64-NEXT:    vprord $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x72,0xc0,0x03]
; X64-NEXT:    vpaddd %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfe,0xc2]
; X64-NEXT:    vpaddd %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x i32> @llvm.x86.avx512.mask.pror.d.256(<8 x i32> %x0, i32 3, <8 x i32> %x2, i8 %x3)
  %res1 = call <8 x i32> @llvm.x86.avx512.mask.pror.d.256(<8 x i32> %x0, i32 3, <8 x i32> zeroinitializer, i8 %x3)
  %res2 = call <8 x i32> @llvm.x86.avx512.mask.pror.d.256(<8 x i32> %x0, i32 3, <8 x i32> %x2, i8 -1)
  %res3 = add <8 x i32> %res, %res1
  %res4 = add <8 x i32> %res3, %res2
  ret <8 x i32> %res4
}

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

define <2 x i64>@test_int_x86_avx512_mask_pror_q_128(<2 x i64> %x0, i32 %x1, <2 x i64> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_pror_q_128:
; X86:       # %bb.0:
; X86-NEXT:    vprorq $3, %xmm0, %xmm2 # encoding: [0x62,0xf1,0xed,0x08,0x72,0xc0,0x03]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vprorq $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x09,0x72,0xc0,0x03]
; X86-NEXT:    vprorq $3, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0x89,0x72,0xc0,0x03]
; 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_pror_q_128:
; X64:       # %bb.0:
; X64-NEXT:    vprorq $3, %xmm0, %xmm2 # encoding: [0x62,0xf1,0xed,0x08,0x72,0xc0,0x03]
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vprorq $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x09,0x72,0xc0,0x03]
; X64-NEXT:    vprorq $3, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0x89,0x72,0xc0,0x03]
; 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.pror.q.128(<2 x i64> %x0, i32 3, <2 x i64> %x2, i8 %x3)
  %res1 = call <2 x i64> @llvm.x86.avx512.mask.pror.q.128(<2 x i64> %x0, i32 3, <2 x i64> zeroinitializer, i8 %x3)
  %res2 = call <2 x i64> @llvm.x86.avx512.mask.pror.q.128(<2 x i64> %x0, i32 3, <2 x i64> %x2, 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.pror.q.256(<4 x i64>, i32, <4 x i64>, i8)

define <4 x i64>@test_int_x86_avx512_mask_pror_q_256(<4 x i64> %x0, i32 %x1, <4 x i64> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_pror_q_256:
; X86:       # %bb.0:
; X86-NEXT:    vprorq $3, %ymm0, %ymm2 # encoding: [0x62,0xf1,0xed,0x28,0x72,0xc0,0x03]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vprorq $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x29,0x72,0xc0,0x03]
; X86-NEXT:    vprorq $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0xa9,0x72,0xc0,0x03]
; 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_pror_q_256:
; X64:       # %bb.0:
; X64-NEXT:    vprorq $3, %ymm0, %ymm2 # encoding: [0x62,0xf1,0xed,0x28,0x72,0xc0,0x03]
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vprorq $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x29,0x72,0xc0,0x03]
; X64-NEXT:    vprorq $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0xa9,0x72,0xc0,0x03]
; 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.pror.q.256(<4 x i64> %x0, i32 3, <4 x i64> %x2, i8 %x3)
  %res1 = call <4 x i64> @llvm.x86.avx512.mask.pror.q.256(<4 x i64> %x0, i32 3, <4 x i64> zeroinitializer, i8 %x3)
  %res2 = call <4 x i64> @llvm.x86.avx512.mask.pror.q.256(<4 x i64> %x0, i32 3, <4 x i64> %x2, i8 -1)
  %res3 = add <4 x i64> %res, %res1
  %res4 = add <4 x i64> %res3, %res2
  ret <4 x i64> %res4
}

declare <4 x i32> @llvm.x86.avx512.prorv.d.128(<4 x i32>, <4 x i32>)

define <4 x i32>@test_int_x86_avx512_prorv_d_128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_prorv_d_128:
; X86:       # %bb.0:
; X86-NEXT:    vprorvd %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf2,0x7d,0x08,0x14,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vprorvd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x14,0xd1]
; X86-NEXT:    vprorvd %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x14,0xc1]
; X86-NEXT:    vpaddd %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfe,0xc3]
; X86-NEXT:    vpaddd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_prorv_d_128:
; X64:       # %bb.0:
; X64-NEXT:    vprorvd %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf2,0x7d,0x08,0x14,0xd9]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vprorvd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x14,0xd1]
; X64-NEXT:    vprorvd %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x14,0xc1]
; X64-NEXT:    vpaddd %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfe,0xc3]
; X64-NEXT:    vpaddd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = call <4 x i32> @llvm.x86.avx512.prorv.d.128(<4 x i32> %x0, <4 x i32> %x1)
  %2 = bitcast i8 %x3 to <8 x i1>
  %extract1 = shufflevector <8 x i1> %2, <8 x i1> %2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %3 = select <4 x i1> %extract1, <4 x i32> %1, <4 x i32> %x2
  %4 = call <4 x i32> @llvm.x86.avx512.prorv.d.128(<4 x i32> %x0, <4 x i32> %x1)
  %5 = bitcast i8 %x3 to <8 x i1>
  %extract = shufflevector <8 x i1> %5, <8 x i1> %5, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %6 = select <4 x i1> %extract, <4 x i32> %4, <4 x i32> zeroinitializer
  %7 = call <4 x i32> @llvm.x86.avx512.prorv.d.128(<4 x i32> %x0, <4 x i32> %x1)
  %res3 = add <4 x i32> %3, %6
  %res4 = add <4 x i32> %res3, %7
  ret <4 x i32> %res4
}

declare <8 x i32> @llvm.x86.avx512.prorv.d.256(<8 x i32>, <8 x i32>)

define <8 x i32>@test_int_x86_avx512_prorv_d_256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_prorv_d_256:
; X86:       # %bb.0:
; X86-NEXT:    vprorvd %ymm1, %ymm0, %ymm3 # encoding: [0x62,0xf2,0x7d,0x28,0x14,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vprorvd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x14,0xd1]
; X86-NEXT:    vprorvd %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x14,0xc1]
; X86-NEXT:    vpaddd %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfe,0xc3]
; X86-NEXT:    vpaddd %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_prorv_d_256:
; X64:       # %bb.0:
; X64-NEXT:    vprorvd %ymm1, %ymm0, %ymm3 # encoding: [0x62,0xf2,0x7d,0x28,0x14,0xd9]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vprorvd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x14,0xd1]
; X64-NEXT:    vprorvd %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x14,0xc1]
; X64-NEXT:    vpaddd %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfe,0xc3]
; X64-NEXT:    vpaddd %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = call <8 x i32> @llvm.x86.avx512.prorv.d.256(<8 x i32> %x0, <8 x i32> %x1)
  %2 = bitcast i8 %x3 to <8 x i1>
  %3 = select <8 x i1> %2, <8 x i32> %1, <8 x i32> %x2
  %4 = call <8 x i32> @llvm.x86.avx512.prorv.d.256(<8 x i32> %x0, <8 x i32> %x1)
  %5 = bitcast i8 %x3 to <8 x i1>
  %6 = select <8 x i1> %5, <8 x i32> %4, <8 x i32> zeroinitializer
  %7 = call <8 x i32> @llvm.x86.avx512.prorv.d.256(<8 x i32> %x0, <8 x i32> %x1)
  %res3 = add <8 x i32> %3, %6
  %res4 = add <8 x i32> %res3, %7
  ret <8 x i32> %res4
}

declare <2 x i64> @llvm.x86.avx512.prorv.q.128(<2 x i64>, <2 x i64>)

define <2 x i64>@test_int_x86_avx512_prorv_q_128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_prorv_q_128:
; X86:       # %bb.0:
; X86-NEXT:    vprorvq %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf2,0xfd,0x08,0x14,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vprorvq %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x14,0xd1]
; X86-NEXT:    vprorvq %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x14,0xc1]
; X86-NEXT:    vpaddq %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd4,0xc3]
; X86-NEXT:    vpaddq %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xd4,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_prorv_q_128:
; X64:       # %bb.0:
; X64-NEXT:    vprorvq %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf2,0xfd,0x08,0x14,0xd9]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vprorvq %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x14,0xd1]
; X64-NEXT:    vprorvq %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x14,0xc1]
; X64-NEXT:    vpaddq %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd4,0xc3]
; X64-NEXT:    vpaddq %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xd4,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = call <2 x i64> @llvm.x86.avx512.prorv.q.128(<2 x i64> %x0, <2 x i64> %x1)
  %2 = bitcast i8 %x3 to <8 x i1>
  %extract1 = shufflevector <8 x i1> %2, <8 x i1> %2, <2 x i32> <i32 0, i32 1>
  %3 = select <2 x i1> %extract1, <2 x i64> %1, <2 x i64> %x2
  %4 = call <2 x i64> @llvm.x86.avx512.prorv.q.128(<2 x i64> %x0, <2 x i64> %x1)
  %5 = bitcast i8 %x3 to <8 x i1>
  %extract = shufflevector <8 x i1> %5, <8 x i1> %5, <2 x i32> <i32 0, i32 1>
  %6 = select <2 x i1> %extract, <2 x i64> %4, <2 x i64> zeroinitializer
  %7 = call <2 x i64> @llvm.x86.avx512.prorv.q.128(<2 x i64> %x0, <2 x i64> %x1)
  %res3 = add <2 x i64> %3, %6
  %res4 = add <2 x i64> %res3, %7
  ret <2 x i64> %res4
}

declare <4 x i64> @llvm.x86.avx512.prorv.q.256(<4 x i64>, <4 x i64>)

define <4 x i64>@test_int_x86_avx512_prorv_q_256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_prorv_q_256:
; X86:       # %bb.0:
; X86-NEXT:    vprorvq %ymm1, %ymm0, %ymm3 # encoding: [0x62,0xf2,0xfd,0x28,0x14,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vprorvq %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x14,0xd1]
; X86-NEXT:    vprorvq %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x14,0xc1]
; X86-NEXT:    vpaddq %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd4,0xc3]
; X86-NEXT:    vpaddq %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xd4,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_prorv_q_256:
; X64:       # %bb.0:
; X64-NEXT:    vprorvq %ymm1, %ymm0, %ymm3 # encoding: [0x62,0xf2,0xfd,0x28,0x14,0xd9]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vprorvq %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x14,0xd1]
; X64-NEXT:    vprorvq %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x14,0xc1]
; X64-NEXT:    vpaddq %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd4,0xc3]
; X64-NEXT:    vpaddq %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xd4,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = call <4 x i64> @llvm.x86.avx512.prorv.q.256(<4 x i64> %x0, <4 x i64> %x1)
  %2 = bitcast i8 %x3 to <8 x i1>
  %extract1 = shufflevector <8 x i1> %2, <8 x i1> %2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %3 = select <4 x i1> %extract1, <4 x i64> %1, <4 x i64> %x2
  %4 = call <4 x i64> @llvm.x86.avx512.prorv.q.256(<4 x i64> %x0, <4 x i64> %x1)
  %5 = bitcast i8 %x3 to <8 x i1>
  %extract = shufflevector <8 x i1> %5, <8 x i1> %5, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %6 = select <4 x i1> %extract, <4 x i64> %4, <4 x i64> zeroinitializer
  %7 = call <4 x i64> @llvm.x86.avx512.prorv.q.256(<4 x i64> %x0, <4 x i64> %x1)
  %res3 = add <4 x i64> %3, %6
  %res4 = add <4 x i64> %res3, %7
  ret <4 x i64> %res4
}

declare <4 x i32> @llvm.x86.avx512.prol.d.128(<4 x i32>, i32)

define <4 x i32>@test_int_x86_avx512_prol_d_128(<4 x i32> %x0, i32 %x1, <4 x i32> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_prol_d_128:
; X86:       # %bb.0:
; X86-NEXT:    vprold $3, %xmm0, %xmm2 # encoding: [0x62,0xf1,0x6d,0x08,0x72,0xc8,0x03]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vprold $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x72,0xc8,0x03]
; X86-NEXT:    vprold $3, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x72,0xc8,0x03]
; X86-NEXT:    vpaddd %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfe,0xc2]
; X86-NEXT:    vpaddd %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_prol_d_128:
; X64:       # %bb.0:
; X64-NEXT:    vprold $3, %xmm0, %xmm2 # encoding: [0x62,0xf1,0x6d,0x08,0x72,0xc8,0x03]
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vprold $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x72,0xc8,0x03]
; X64-NEXT:    vprold $3, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x72,0xc8,0x03]
; X64-NEXT:    vpaddd %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfe,0xc2]
; X64-NEXT:    vpaddd %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = call <4 x i32> @llvm.x86.avx512.prol.d.128(<4 x i32> %x0, i32 3)
  %2 = bitcast i8 %x3 to <8 x i1>
  %extract1 = shufflevector <8 x i1> %2, <8 x i1> %2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %3 = select <4 x i1> %extract1, <4 x i32> %1, <4 x i32> %x2
  %4 = call <4 x i32> @llvm.x86.avx512.prol.d.128(<4 x i32> %x0, i32 3)
  %5 = bitcast i8 %x3 to <8 x i1>
  %extract = shufflevector <8 x i1> %5, <8 x i1> %5, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %6 = select <4 x i1> %extract, <4 x i32> %4, <4 x i32> zeroinitializer
  %7 = call <4 x i32> @llvm.x86.avx512.prol.d.128(<4 x i32> %x0, i32 3)
  %res3 = add <4 x i32> %3, %6
  %res4 = add <4 x i32> %res3, %7
  ret <4 x i32> %res4
}

declare <8 x i32> @llvm.x86.avx512.prol.d.256(<8 x i32>, i32)

define <8 x i32>@test_int_x86_avx512_prol_d_256(<8 x i32> %x0, i32 %x1, <8 x i32> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_prol_d_256:
; X86:       # %bb.0:
; X86-NEXT:    vprold $3, %ymm0, %ymm2 # encoding: [0x62,0xf1,0x6d,0x28,0x72,0xc8,0x03]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vprold $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x72,0xc8,0x03]
; X86-NEXT:    vprold $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x72,0xc8,0x03]
; X86-NEXT:    vpaddd %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfe,0xc2]
; X86-NEXT:    vpaddd %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_prol_d_256:
; X64:       # %bb.0:
; X64-NEXT:    vprold $3, %ymm0, %ymm2 # encoding: [0x62,0xf1,0x6d,0x28,0x72,0xc8,0x03]
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vprold $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x72,0xc8,0x03]
; X64-NEXT:    vprold $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x72,0xc8,0x03]
; X64-NEXT:    vpaddd %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfe,0xc2]
; X64-NEXT:    vpaddd %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = call <8 x i32> @llvm.x86.avx512.prol.d.256(<8 x i32> %x0, i32 3)
  %2 = bitcast i8 %x3 to <8 x i1>
  %3 = select <8 x i1> %2, <8 x i32> %1, <8 x i32> %x2
  %4 = call <8 x i32> @llvm.x86.avx512.prol.d.256(<8 x i32> %x0, i32 3)
  %5 = bitcast i8 %x3 to <8 x i1>
  %6 = select <8 x i1> %5, <8 x i32> %4, <8 x i32> zeroinitializer
  %7 = call <8 x i32> @llvm.x86.avx512.prol.d.256(<8 x i32> %x0, i32 3)
  %res3 = add <8 x i32> %3, %6
  %res4 = add <8 x i32> %res3, %7
  ret <8 x i32> %res4
}

declare <2 x i64> @llvm.x86.avx512.prol.q.128(<2 x i64>, i32)

define <2 x i64>@test_int_x86_avx512_prol_q_128(<2 x i64> %x0, i32 %x1, <2 x i64> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_prol_q_128:
; X86:       # %bb.0:
; X86-NEXT:    vprolq $3, %xmm0, %xmm2 # encoding: [0x62,0xf1,0xed,0x08,0x72,0xc8,0x03]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vprolq $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x09,0x72,0xc8,0x03]
; X86-NEXT:    vprolq $3, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0x89,0x72,0xc8,0x03]
; 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_prol_q_128:
; X64:       # %bb.0:
; X64-NEXT:    vprolq $3, %xmm0, %xmm2 # encoding: [0x62,0xf1,0xed,0x08,0x72,0xc8,0x03]
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vprolq $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x09,0x72,0xc8,0x03]
; X64-NEXT:    vprolq $3, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0x89,0x72,0xc8,0x03]
; 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]
  %1 = call <2 x i64> @llvm.x86.avx512.prol.q.128(<2 x i64> %x0, i32 3)
  %2 = bitcast i8 %x3 to <8 x i1>
  %extract1 = shufflevector <8 x i1> %2, <8 x i1> %2, <2 x i32> <i32 0, i32 1>
  %3 = select <2 x i1> %extract1, <2 x i64> %1, <2 x i64> %x2
  %4 = call <2 x i64> @llvm.x86.avx512.prol.q.128(<2 x i64> %x0, i32 3)
  %5 = bitcast i8 %x3 to <8 x i1>
  %extract = shufflevector <8 x i1> %5, <8 x i1> %5, <2 x i32> <i32 0, i32 1>
  %6 = select <2 x i1> %extract, <2 x i64> %4, <2 x i64> zeroinitializer
  %7 = call <2 x i64> @llvm.x86.avx512.prol.q.128(<2 x i64> %x0, i32 3)
  %res3 = add <2 x i64> %3, %6
  %res4 = add <2 x i64> %res3, %7
  ret <2 x i64> %res4
}

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

define <4 x i64>@test_int_x86_avx512_prol_q_256(<4 x i64> %x0, i32 %x1, <4 x i64> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_prol_q_256:
; X86:       # %bb.0:
; X86-NEXT:    vprolq $3, %ymm0, %ymm2 # encoding: [0x62,0xf1,0xed,0x28,0x72,0xc8,0x03]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vprolq $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x29,0x72,0xc8,0x03]
; X86-NEXT:    vprolq $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0xa9,0x72,0xc8,0x03]
; 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_prol_q_256:
; X64:       # %bb.0:
; X64-NEXT:    vprolq $3, %ymm0, %ymm2 # encoding: [0x62,0xf1,0xed,0x28,0x72,0xc8,0x03]
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vprolq $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x29,0x72,0xc8,0x03]
; X64-NEXT:    vprolq $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0xa9,0x72,0xc8,0x03]
; 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]
  %1 = call <4 x i64> @llvm.x86.avx512.prol.q.256(<4 x i64> %x0, i32 3)
  %2 = bitcast i8 %x3 to <8 x i1>
  %extract1 = shufflevector <8 x i1> %2, <8 x i1> %2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %3 = select <4 x i1> %extract1, <4 x i64> %1, <4 x i64> %x2
  %4 = call <4 x i64> @llvm.x86.avx512.prol.q.256(<4 x i64> %x0, i32 3)
  %5 = bitcast i8 %x3 to <8 x i1>
  %extract = shufflevector <8 x i1> %5, <8 x i1> %5, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %6 = select <4 x i1> %extract, <4 x i64> %4, <4 x i64> zeroinitializer
  %7 = call <4 x i64> @llvm.x86.avx512.prol.q.256(<4 x i64> %x0, i32 3)
  %res3 = add <4 x i64> %3, %6
  %res4 = add <4 x i64> %res3, %7
  ret <4 x i64> %res4
}

declare <4 x i32> @llvm.x86.avx512.prolv.d.128(<4 x i32>, <4 x i32>)

define <4 x i32>@test_int_x86_avx512_prolv_d_128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_prolv_d_128:
; X86:       # %bb.0:
; X86-NEXT:    vprolvd %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf2,0x7d,0x08,0x15,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vprolvd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x15,0xd1]
; X86-NEXT:    vprolvd %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x15,0xc1]
; X86-NEXT:    vpaddd %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfe,0xc3]
; X86-NEXT:    vpaddd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_prolv_d_128:
; X64:       # %bb.0:
; X64-NEXT:    vprolvd %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf2,0x7d,0x08,0x15,0xd9]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vprolvd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x15,0xd1]
; X64-NEXT:    vprolvd %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x15,0xc1]
; X64-NEXT:    vpaddd %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfe,0xc3]
; X64-NEXT:    vpaddd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = call <4 x i32> @llvm.x86.avx512.prolv.d.128(<4 x i32> %x0, <4 x i32> %x1)
  %2 = bitcast i8 %x3 to <8 x i1>
  %extract1 = shufflevector <8 x i1> %2, <8 x i1> %2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %3 = select <4 x i1> %extract1, <4 x i32> %1, <4 x i32> %x2
  %4 = call <4 x i32> @llvm.x86.avx512.prolv.d.128(<4 x i32> %x0, <4 x i32> %x1)
  %5 = bitcast i8 %x3 to <8 x i1>
  %extract = shufflevector <8 x i1> %5, <8 x i1> %5, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %6 = select <4 x i1> %extract, <4 x i32> %4, <4 x i32> zeroinitializer
  %7 = call <4 x i32> @llvm.x86.avx512.prolv.d.128(<4 x i32> %x0, <4 x i32> %x1)
  %res3 = add <4 x i32> %3, %6
  %res4 = add <4 x i32> %res3, %7
  ret <4 x i32> %res4
}

declare <8 x i32> @llvm.x86.avx512.prolv.d.256(<8 x i32>, <8 x i32>)

define <8 x i32>@test_int_x86_avx512_prolv_d_256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_prolv_d_256:
; X86:       # %bb.0:
; X86-NEXT:    vprolvd %ymm1, %ymm0, %ymm3 # encoding: [0x62,0xf2,0x7d,0x28,0x15,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vprolvd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x15,0xd1]
; X86-NEXT:    vprolvd %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x15,0xc1]
; X86-NEXT:    vpaddd %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfe,0xc3]
; X86-NEXT:    vpaddd %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_prolv_d_256:
; X64:       # %bb.0:
; X64-NEXT:    vprolvd %ymm1, %ymm0, %ymm3 # encoding: [0x62,0xf2,0x7d,0x28,0x15,0xd9]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vprolvd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x15,0xd1]
; X64-NEXT:    vprolvd %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x15,0xc1]
; X64-NEXT:    vpaddd %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfe,0xc3]
; X64-NEXT:    vpaddd %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = call <8 x i32> @llvm.x86.avx512.prolv.d.256(<8 x i32> %x0, <8 x i32> %x1)
  %2 = bitcast i8 %x3 to <8 x i1>
  %3 = select <8 x i1> %2, <8 x i32> %1, <8 x i32> %x2
  %4 = call <8 x i32> @llvm.x86.avx512.prolv.d.256(<8 x i32> %x0, <8 x i32> %x1)
  %5 = bitcast i8 %x3 to <8 x i1>
  %6 = select <8 x i1> %5, <8 x i32> %4, <8 x i32> zeroinitializer
  %7 = call <8 x i32> @llvm.x86.avx512.prolv.d.256(<8 x i32> %x0, <8 x i32> %x1)
  %res3 = add <8 x i32> %3, %6
  %res4 = add <8 x i32> %res3, %7
  ret <8 x i32> %res4
}

declare <2 x i64> @llvm.x86.avx512.prolv.q.128(<2 x i64>, <2 x i64>)

define <2 x i64>@test_int_x86_avx512_prolv_q_128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_prolv_q_128:
; X86:       # %bb.0:
; X86-NEXT:    vprolvq %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf2,0xfd,0x08,0x15,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vprolvq %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x15,0xd1]
; X86-NEXT:    vprolvq %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x15,0xc1]
; X86-NEXT:    vpaddq %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd4,0xc3]
; X86-NEXT:    vpaddq %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xd4,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_prolv_q_128:
; X64:       # %bb.0:
; X64-NEXT:    vprolvq %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf2,0xfd,0x08,0x15,0xd9]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vprolvq %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x15,0xd1]
; X64-NEXT:    vprolvq %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x15,0xc1]
; X64-NEXT:    vpaddq %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd4,0xc3]
; X64-NEXT:    vpaddq %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xd4,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = call <2 x i64> @llvm.x86.avx512.prolv.q.128(<2 x i64> %x0, <2 x i64> %x1)
  %2 = bitcast i8 %x3 to <8 x i1>
  %extract1 = shufflevector <8 x i1> %2, <8 x i1> %2, <2 x i32> <i32 0, i32 1>
  %3 = select <2 x i1> %extract1, <2 x i64> %1, <2 x i64> %x2
  %4 = call <2 x i64> @llvm.x86.avx512.prolv.q.128(<2 x i64> %x0, <2 x i64> %x1)
  %5 = bitcast i8 %x3 to <8 x i1>
  %extract = shufflevector <8 x i1> %5, <8 x i1> %5, <2 x i32> <i32 0, i32 1>
  %6 = select <2 x i1> %extract, <2 x i64> %4, <2 x i64> zeroinitializer
  %7 = call <2 x i64> @llvm.x86.avx512.prolv.q.128(<2 x i64> %x0, <2 x i64> %x1)
  %res3 = add <2 x i64> %3, %6
  %res4 = add <2 x i64> %res3, %7
  ret <2 x i64> %res4
}

declare <4 x i64> @llvm.x86.avx512.prolv.q.256(<4 x i64>, <4 x i64>)

define <4 x i64>@test_int_x86_avx512_prolv_q_256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_prolv_q_256:
; X86:       # %bb.0:
; X86-NEXT:    vprolvq %ymm1, %ymm0, %ymm3 # encoding: [0x62,0xf2,0xfd,0x28,0x15,0xd9]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vprolvq %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x15,0xd1]
; X86-NEXT:    vprolvq %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x15,0xc1]
; X86-NEXT:    vpaddq %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd4,0xc3]
; X86-NEXT:    vpaddq %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xd4,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_prolv_q_256:
; X64:       # %bb.0:
; X64-NEXT:    vprolvq %ymm1, %ymm0, %ymm3 # encoding: [0x62,0xf2,0xfd,0x28,0x15,0xd9]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vprolvq %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x15,0xd1]
; X64-NEXT:    vprolvq %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x15,0xc1]
; X64-NEXT:    vpaddq %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd4,0xc3]
; X64-NEXT:    vpaddq %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xd4,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = call <4 x i64> @llvm.x86.avx512.prolv.q.256(<4 x i64> %x0, <4 x i64> %x1)
  %2 = bitcast i8 %x3 to <8 x i1>
  %extract1 = shufflevector <8 x i1> %2, <8 x i1> %2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %3 = select <4 x i1> %extract1, <4 x i64> %1, <4 x i64> %x2
  %4 = call <4 x i64> @llvm.x86.avx512.prolv.q.256(<4 x i64> %x0, <4 x i64> %x1)
  %5 = bitcast i8 %x3 to <8 x i1>
  %extract = shufflevector <8 x i1> %5, <8 x i1> %5, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %6 = select <4 x i1> %extract, <4 x i64> %4, <4 x i64> zeroinitializer
  %7 = call <4 x i64> @llvm.x86.avx512.prolv.q.256(<4 x i64> %x0, <4 x i64> %x1)
  %res3 = add <4 x i64> %3, %6
  %res4 = add <4 x i64> %res3, %7
  ret <4 x i64> %res4
}

declare <4 x i32> @llvm.x86.avx512.pror.d.128(<4 x i32>, i32)

define <4 x i32>@test_int_x86_avx512_pror_d_128(<4 x i32> %x0, i32 %x1, <4 x i32> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_pror_d_128:
; X86:       # %bb.0:
; X86-NEXT:    vprord $3, %xmm0, %xmm2 # encoding: [0x62,0xf1,0x6d,0x08,0x72,0xc0,0x03]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vprord $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x72,0xc0,0x03]
; X86-NEXT:    vprord $3, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x72,0xc0,0x03]
; X86-NEXT:    vpaddd %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfe,0xc2]
; X86-NEXT:    vpaddd %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_pror_d_128:
; X64:       # %bb.0:
; X64-NEXT:    vprord $3, %xmm0, %xmm2 # encoding: [0x62,0xf1,0x6d,0x08,0x72,0xc0,0x03]
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vprord $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x72,0xc0,0x03]
; X64-NEXT:    vprord $3, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x72,0xc0,0x03]
; X64-NEXT:    vpaddd %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfe,0xc2]
; X64-NEXT:    vpaddd %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = call <4 x i32> @llvm.x86.avx512.pror.d.128(<4 x i32> %x0, i32 3)
  %2 = bitcast i8 %x3 to <8 x i1>
  %extract1 = shufflevector <8 x i1> %2, <8 x i1> %2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %3 = select <4 x i1> %extract1, <4 x i32> %1, <4 x i32> %x2
  %4 = call <4 x i32> @llvm.x86.avx512.pror.d.128(<4 x i32> %x0, i32 3)
  %5 = bitcast i8 %x3 to <8 x i1>
  %extract = shufflevector <8 x i1> %5, <8 x i1> %5, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %6 = select <4 x i1> %extract, <4 x i32> %4, <4 x i32> zeroinitializer
  %7 = call <4 x i32> @llvm.x86.avx512.pror.d.128(<4 x i32> %x0, i32 3)
  %res3 = add <4 x i32> %3, %6
  %res4 = add <4 x i32> %res3, %7
  ret <4 x i32> %res4
}

declare <8 x i32> @llvm.x86.avx512.pror.d.256(<8 x i32>, i32)

define <8 x i32>@test_int_x86_avx512_pror_d_256(<8 x i32> %x0, i32 %x1, <8 x i32> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_pror_d_256:
; X86:       # %bb.0:
; X86-NEXT:    vprord $3, %ymm0, %ymm2 # encoding: [0x62,0xf1,0x6d,0x28,0x72,0xc0,0x03]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vprord $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x72,0xc0,0x03]
; X86-NEXT:    vprord $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x72,0xc0,0x03]
; X86-NEXT:    vpaddd %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfe,0xc2]
; X86-NEXT:    vpaddd %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_pror_d_256:
; X64:       # %bb.0:
; X64-NEXT:    vprord $3, %ymm0, %ymm2 # encoding: [0x62,0xf1,0x6d,0x28,0x72,0xc0,0x03]
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vprord $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x72,0xc0,0x03]
; X64-NEXT:    vprord $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x72,0xc0,0x03]
; X64-NEXT:    vpaddd %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfe,0xc2]
; X64-NEXT:    vpaddd %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = call <8 x i32> @llvm.x86.avx512.pror.d.256(<8 x i32> %x0, i32 3)
  %2 = bitcast i8 %x3 to <8 x i1>
  %3 = select <8 x i1> %2, <8 x i32> %1, <8 x i32> %x2
  %4 = call <8 x i32> @llvm.x86.avx512.pror.d.256(<8 x i32> %x0, i32 3)
  %5 = bitcast i8 %x3 to <8 x i1>
  %6 = select <8 x i1> %5, <8 x i32> %4, <8 x i32> zeroinitializer
  %7 = call <8 x i32> @llvm.x86.avx512.pror.d.256(<8 x i32> %x0, i32 3)
  %res3 = add <8 x i32> %3, %6
  %res4 = add <8 x i32> %res3, %7
  ret <8 x i32> %res4
}

declare <2 x i64> @llvm.x86.avx512.pror.q.128(<2 x i64>, i32)

define <2 x i64>@test_int_x86_avx512_pror_q_128(<2 x i64> %x0, i32 %x1, <2 x i64> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_pror_q_128:
; X86:       # %bb.0:
; X86-NEXT:    vprorq $3, %xmm0, %xmm2 # encoding: [0x62,0xf1,0xed,0x08,0x72,0xc0,0x03]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vprorq $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x09,0x72,0xc0,0x03]
; X86-NEXT:    vprorq $3, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0x89,0x72,0xc0,0x03]
; 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_pror_q_128:
; X64:       # %bb.0:
; X64-NEXT:    vprorq $3, %xmm0, %xmm2 # encoding: [0x62,0xf1,0xed,0x08,0x72,0xc0,0x03]
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vprorq $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x09,0x72,0xc0,0x03]
; X64-NEXT:    vprorq $3, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0x89,0x72,0xc0,0x03]
; 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]
  %1 = call <2 x i64> @llvm.x86.avx512.pror.q.128(<2 x i64> %x0, i32 3)
  %2 = bitcast i8 %x3 to <8 x i1>
  %extract1 = shufflevector <8 x i1> %2, <8 x i1> %2, <2 x i32> <i32 0, i32 1>
  %3 = select <2 x i1> %extract1, <2 x i64> %1, <2 x i64> %x2
  %4 = call <2 x i64> @llvm.x86.avx512.pror.q.128(<2 x i64> %x0, i32 3)
  %5 = bitcast i8 %x3 to <8 x i1>
  %extract = shufflevector <8 x i1> %5, <8 x i1> %5, <2 x i32> <i32 0, i32 1>
  %6 = select <2 x i1> %extract, <2 x i64> %4, <2 x i64> zeroinitializer
  %7 = call <2 x i64> @llvm.x86.avx512.pror.q.128(<2 x i64> %x0, i32 3)
  %res3 = add <2 x i64> %3, %6
  %res4 = add <2 x i64> %res3, %7
  ret <2 x i64> %res4
}

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

define <4 x i64>@test_int_x86_avx512_pror_q_256(<4 x i64> %x0, i32 %x1, <4 x i64> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_pror_q_256:
; X86:       # %bb.0:
; X86-NEXT:    vprorq $3, %ymm0, %ymm2 # encoding: [0x62,0xf1,0xed,0x28,0x72,0xc0,0x03]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vprorq $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x29,0x72,0xc0,0x03]
; X86-NEXT:    vprorq $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0xa9,0x72,0xc0,0x03]
; 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_pror_q_256:
; X64:       # %bb.0:
; X64-NEXT:    vprorq $3, %ymm0, %ymm2 # encoding: [0x62,0xf1,0xed,0x28,0x72,0xc0,0x03]
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vprorq $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x29,0x72,0xc0,0x03]
; X64-NEXT:    vprorq $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0xa9,0x72,0xc0,0x03]
; 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]
  %1 = call <4 x i64> @llvm.x86.avx512.pror.q.256(<4 x i64> %x0, i32 3)
  %2 = bitcast i8 %x3 to <8 x i1>
  %extract1 = shufflevector <8 x i1> %2, <8 x i1> %2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %3 = select <4 x i1> %extract1, <4 x i64> %1, <4 x i64> %x2
  %4 = call <4 x i64> @llvm.x86.avx512.pror.q.256(<4 x i64> %x0, i32 3)
  %5 = bitcast i8 %x3 to <8 x i1>
  %extract = shufflevector <8 x i1> %5, <8 x i1> %5, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %6 = select <4 x i1> %extract, <4 x i64> %4, <4 x i64> zeroinitializer
  %7 = call <4 x i64> @llvm.x86.avx512.pror.q.256(<4 x i64> %x0, i32 3)
  %res3 = add <4 x i64> %3, %6
  %res4 = add <4 x i64> %res3, %7
  ret <4 x i64> %res4
}

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

define <8 x float> @test_vfmadd256_ps(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2) {
; CHECK-LABEL: test_vfmadd256_ps:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vfmadd213ps %ymm2, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x75,0xa8,0xc2]
; CHECK-NEXT:    # ymm0 = (ymm1 * ymm0) + ymm2
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask.vfmadd.ps.256(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2, i8 -1) nounwind
  ret <8 x float> %res
}

define <8 x float> @test_mask_vfmadd256_ps(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2, i8 %mask) {
; X86-LABEL: test_mask_vfmadd256_ps:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vfmadd132ps %ymm1, %ymm2, %ymm0 {%k1} # encoding: [0x62,0xf2,0x6d,0x29,0x98,0xc1]
; X86-NEXT:    # ymm0 = (ymm0 * ymm1) + ymm2
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_vfmadd256_ps:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vfmadd132ps %ymm1, %ymm2, %ymm0 {%k1} # encoding: [0x62,0xf2,0x6d,0x29,0x98,0xc1]
; X64-NEXT:    # ymm0 = (ymm0 * ymm1) + ymm2
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask.vfmadd.ps.256(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2, i8 %mask) nounwind
  ret <8 x float> %res
}

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

define <4 x float> @test_vfmadd128_ps(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2) {
; CHECK-LABEL: test_vfmadd128_ps:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vfmadd213ps %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x71,0xa8,0xc2]
; CHECK-NEXT:    # xmm0 = (xmm1 * xmm0) + xmm2
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask.vfmadd.ps.128(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2, i8 -1) nounwind
  ret <4 x float> %res
}

define <4 x float> @test_mask_vfmadd128_ps(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2, i8 %mask) {
; X86-LABEL: test_mask_vfmadd128_ps:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vfmadd132ps %xmm1, %xmm2, %xmm0 {%k1} # encoding: [0x62,0xf2,0x6d,0x09,0x98,0xc1]
; X86-NEXT:    # xmm0 = (xmm0 * xmm1) + xmm2
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_vfmadd128_ps:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vfmadd132ps %xmm1, %xmm2, %xmm0 {%k1} # encoding: [0x62,0xf2,0x6d,0x09,0x98,0xc1]
; X64-NEXT:    # xmm0 = (xmm0 * xmm1) + xmm2
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask.vfmadd.ps.128(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2, i8 %mask) nounwind
  ret <4 x float> %res
}

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

define <4 x double> @test_fmadd256_pd(<4 x double> %a, <4 x double> %b, <4 x double> %c) {
; CHECK-LABEL: test_fmadd256_pd:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vfmadd213pd %ymm2, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0xf5,0xa8,0xc2]
; CHECK-NEXT:    # ymm0 = (ymm1 * ymm0) + ymm2
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <4 x double> @llvm.x86.avx512.mask.vfmadd.pd.256(<4 x double> %a, <4 x double> %b, <4 x double> %c, i8 -1)
  ret <4 x double> %res
}

define <4 x double> @test_mask_fmadd256_pd(<4 x double> %a, <4 x double> %b, <4 x double> %c, i8 %mask) {
; X86-LABEL: test_mask_fmadd256_pd:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vfmadd132pd %ymm1, %ymm2, %ymm0 {%k1} # encoding: [0x62,0xf2,0xed,0x29,0x98,0xc1]
; X86-NEXT:    # ymm0 = (ymm0 * ymm1) + ymm2
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_fmadd256_pd:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vfmadd132pd %ymm1, %ymm2, %ymm0 {%k1} # encoding: [0x62,0xf2,0xed,0x29,0x98,0xc1]
; X64-NEXT:    # ymm0 = (ymm0 * ymm1) + ymm2
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x double> @llvm.x86.avx512.mask.vfmadd.pd.256(<4 x double> %a, <4 x double> %b, <4 x double> %c, i8 %mask)
  ret <4 x double> %res
}

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

define <2 x double> @test_fmadd128_pd(<2 x double> %a, <2 x double> %b, <2 x double> %c) {
; CHECK-LABEL: test_fmadd128_pd:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vfmadd213pd %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0xf1,0xa8,0xc2]
; CHECK-NEXT:    # xmm0 = (xmm1 * xmm0) + xmm2
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <2 x double> @llvm.x86.avx512.mask.vfmadd.pd.128(<2 x double> %a, <2 x double> %b, <2 x double> %c, i8 -1)
  ret <2 x double> %res
}

define <2 x double> @test_mask_fmadd128_pd(<2 x double> %a, <2 x double> %b, <2 x double> %c, i8 %mask) {
; X86-LABEL: test_mask_fmadd128_pd:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vfmadd132pd %xmm1, %xmm2, %xmm0 {%k1} # encoding: [0x62,0xf2,0xed,0x09,0x98,0xc1]
; X86-NEXT:    # xmm0 = (xmm0 * xmm1) + xmm2
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_fmadd128_pd:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vfmadd132pd %xmm1, %xmm2, %xmm0 {%k1} # encoding: [0x62,0xf2,0xed,0x09,0x98,0xc1]
; X64-NEXT:    # xmm0 = (xmm0 * xmm1) + xmm2
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x double> @llvm.x86.avx512.mask.vfmadd.pd.128(<2 x double> %a, <2 x double> %b, <2 x double> %c, i8 %mask)
  ret <2 x double> %res
}

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

define <2 x double>@test_int_x86_avx512_mask3_vfmadd_pd_128(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask3_vfmadd_pd_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vfmadd231pd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0xb8,0xd1]
; X86-NEXT:    # xmm2 = (xmm0 * xmm1) + xmm2
; X86-NEXT:    vmovapd %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x28,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask3_vfmadd_pd_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vfmadd231pd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0xb8,0xd1]
; X64-NEXT:    # xmm2 = (xmm0 * xmm1) + xmm2
; X64-NEXT:    vmovapd %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x28,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x double> @llvm.x86.avx512.mask3.vfmadd.pd.128(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 %x3)
  ret <2 x double> %res
}

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

define <2 x double>@test_int_x86_avx512_maskz_vfmadd_pd_128(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_maskz_vfmadd_pd_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vfmadd213pd %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0x89,0xa8,0xc2]
; X86-NEXT:    # xmm0 = (xmm1 * xmm0) + xmm2
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_maskz_vfmadd_pd_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vfmadd213pd %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0x89,0xa8,0xc2]
; X64-NEXT:    # xmm0 = (xmm1 * xmm0) + xmm2
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x double> @llvm.x86.avx512.maskz.vfmadd.pd.128(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 %x3)
  ret <2 x double> %res
}

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

define <4 x double>@test_int_x86_avx512_mask3_vfmadd_pd_256(<4 x double> %x0, <4 x double> %x1, <4 x double> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask3_vfmadd_pd_256:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vfmadd231pd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0xb8,0xd1]
; X86-NEXT:    # ymm2 = (ymm0 * ymm1) + ymm2
; X86-NEXT:    vmovapd %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x28,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask3_vfmadd_pd_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vfmadd231pd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0xb8,0xd1]
; X64-NEXT:    # ymm2 = (ymm0 * ymm1) + ymm2
; X64-NEXT:    vmovapd %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x28,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x double> @llvm.x86.avx512.mask3.vfmadd.pd.256(<4 x double> %x0, <4 x double> %x1, <4 x double> %x2, i8 %x3)
  ret <4 x double> %res
}

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

define <4 x double>@test_int_x86_avx512_maskz_vfmadd_pd_256(<4 x double> %x0, <4 x double> %x1, <4 x double> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_maskz_vfmadd_pd_256:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vfmadd213pd %ymm2, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0xa9,0xa8,0xc2]
; X86-NEXT:    # ymm0 = (ymm1 * ymm0) + ymm2
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_maskz_vfmadd_pd_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vfmadd213pd %ymm2, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0xa9,0xa8,0xc2]
; X64-NEXT:    # ymm0 = (ymm1 * ymm0) + ymm2
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x double> @llvm.x86.avx512.maskz.vfmadd.pd.256(<4 x double> %x0, <4 x double> %x1, <4 x double> %x2, i8 %x3)
  ret <4 x double> %res
}

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

define <4 x float>@test_int_x86_avx512_mask3_vfmadd_ps_128(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask3_vfmadd_ps_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vfmadd231ps %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0xb8,0xd1]
; X86-NEXT:    # xmm2 = (xmm0 * xmm1) + xmm2
; X86-NEXT:    vmovaps %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x28,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask3_vfmadd_ps_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vfmadd231ps %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0xb8,0xd1]
; X64-NEXT:    # xmm2 = (xmm0 * xmm1) + xmm2
; X64-NEXT:    vmovaps %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x28,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask3.vfmadd.ps.128(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 %x3)
  ret <4 x float> %res
}

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

define <4 x float>@test_int_x86_avx512_maskz_vfmadd_ps_128(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_maskz_vfmadd_ps_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vfmadd213ps %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0x89,0xa8,0xc2]
; X86-NEXT:    # xmm0 = (xmm1 * xmm0) + xmm2
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_maskz_vfmadd_ps_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vfmadd213ps %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0x89,0xa8,0xc2]
; X64-NEXT:    # xmm0 = (xmm1 * xmm0) + xmm2
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.maskz.vfmadd.ps.128(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 %x3)
  ret <4 x float> %res
}

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

define <8 x float>@test_int_x86_avx512_mask3_vfmadd_ps_256(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask3_vfmadd_ps_256:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vfmadd231ps %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0xb8,0xd1]
; X86-NEXT:    # ymm2 = (ymm0 * ymm1) + ymm2
; X86-NEXT:    vmovaps %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x28,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask3_vfmadd_ps_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vfmadd231ps %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0xb8,0xd1]
; X64-NEXT:    # ymm2 = (ymm0 * ymm1) + ymm2
; X64-NEXT:    vmovaps %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x28,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask3.vfmadd.ps.256(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2, i8 %x3)
  ret <8 x float> %res
}

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

define <8 x float>@test_int_x86_avx512_maskz_vfmadd_ps_256(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_maskz_vfmadd_ps_256:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vfmadd213ps %ymm2, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xa9,0xa8,0xc2]
; X86-NEXT:    # ymm0 = (ymm1 * ymm0) + ymm2
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_maskz_vfmadd_ps_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vfmadd213ps %ymm2, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xa9,0xa8,0xc2]
; X64-NEXT:    # ymm0 = (ymm1 * ymm0) + ymm2
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.maskz.vfmadd.ps.256(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2, i8 %x3)
  ret <8 x float> %res
}


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

define <2 x double>@test_int_x86_avx512_mask3_vfmsub_pd_128(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask3_vfmsub_pd_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vfmsub231pd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0xba,0xd1]
; X86-NEXT:    # xmm2 = (xmm0 * xmm1) - xmm2
; X86-NEXT:    vmovapd %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x28,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask3_vfmsub_pd_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vfmsub231pd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0xba,0xd1]
; X64-NEXT:    # xmm2 = (xmm0 * xmm1) - xmm2
; X64-NEXT:    vmovapd %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x28,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x double> @llvm.x86.avx512.mask3.vfmsub.pd.128(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 %x3)
  ret <2 x double> %res
}


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

define <4 x double>@test_int_x86_avx512_mask3_vfmsub_pd_256(<4 x double> %x0, <4 x double> %x1, <4 x double> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask3_vfmsub_pd_256:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vfmsub231pd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0xba,0xd1]
; X86-NEXT:    # ymm2 = (ymm0 * ymm1) - ymm2
; X86-NEXT:    vmovapd %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x28,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask3_vfmsub_pd_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vfmsub231pd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0xba,0xd1]
; X64-NEXT:    # ymm2 = (ymm0 * ymm1) - ymm2
; X64-NEXT:    vmovapd %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x28,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x double> @llvm.x86.avx512.mask3.vfmsub.pd.256(<4 x double> %x0, <4 x double> %x1, <4 x double> %x2, i8 %x3)
  ret <4 x double> %res
}

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

define <4 x float>@test_int_x86_avx512_mask3_vfmsub_ps_128(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask3_vfmsub_ps_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vfmsub231ps %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0xba,0xd1]
; X86-NEXT:    # xmm2 = (xmm0 * xmm1) - xmm2
; X86-NEXT:    vmovaps %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x28,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask3_vfmsub_ps_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vfmsub231ps %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0xba,0xd1]
; X64-NEXT:    # xmm2 = (xmm0 * xmm1) - xmm2
; X64-NEXT:    vmovaps %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x28,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask3.vfmsub.ps.128(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 %x3)
  ret <4 x float> %res
}

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

define <8 x float>@test_int_x86_avx512_mask3_vfmsub_ps_256(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask3_vfmsub_ps_256:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vfmsub231ps %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0xba,0xd1]
; X86-NEXT:    # ymm2 = (ymm0 * ymm1) - ymm2
; X86-NEXT:    vmovaps %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x28,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask3_vfmsub_ps_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vfmsub231ps %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0xba,0xd1]
; X64-NEXT:    # ymm2 = (ymm0 * ymm1) - ymm2
; X64-NEXT:    vmovaps %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x28,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask3.vfmsub.ps.256(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2, i8 %x3)
  ret <8 x float> %res
}

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

define <8 x float> @test_vfnmadd256_ps(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2) {
; CHECK-LABEL: test_vfnmadd256_ps:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vfnmadd213ps %ymm2, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x75,0xac,0xc2]
; CHECK-NEXT:    # ymm0 = -(ymm1 * ymm0) + ymm2
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask.vfnmadd.ps.256(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2, i8 -1) nounwind
  ret <8 x float> %res
}

define <8 x float> @test_mask_vfnmadd256_ps(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2, i8 %mask) {
; X86-LABEL: test_mask_vfnmadd256_ps:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vfnmadd132ps %ymm1, %ymm2, %ymm0 {%k1} # encoding: [0x62,0xf2,0x6d,0x29,0x9c,0xc1]
; X86-NEXT:    # ymm0 = -(ymm0 * ymm1) + ymm2
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_vfnmadd256_ps:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vfnmadd132ps %ymm1, %ymm2, %ymm0 {%k1} # encoding: [0x62,0xf2,0x6d,0x29,0x9c,0xc1]
; X64-NEXT:    # ymm0 = -(ymm0 * ymm1) + ymm2
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask.vfnmadd.ps.256(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2, i8 %mask) nounwind
  ret <8 x float> %res
}

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

define <4 x float> @test_vfnmadd128_ps(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2) {
; CHECK-LABEL: test_vfnmadd128_ps:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vfnmadd213ps %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x71,0xac,0xc2]
; CHECK-NEXT:    # xmm0 = -(xmm1 * xmm0) + xmm2
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask.vfnmadd.ps.128(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2, i8 -1) nounwind
  ret <4 x float> %res
}

define <4 x float> @test_mask_vfnmadd128_ps(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2, i8 %mask) {
; X86-LABEL: test_mask_vfnmadd128_ps:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vfnmadd132ps %xmm1, %xmm2, %xmm0 {%k1} # encoding: [0x62,0xf2,0x6d,0x09,0x9c,0xc1]
; X86-NEXT:    # xmm0 = -(xmm0 * xmm1) + xmm2
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_vfnmadd128_ps:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vfnmadd132ps %xmm1, %xmm2, %xmm0 {%k1} # encoding: [0x62,0xf2,0x6d,0x09,0x9c,0xc1]
; X64-NEXT:    # xmm0 = -(xmm0 * xmm1) + xmm2
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask.vfnmadd.ps.128(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2, i8 %mask) nounwind
  ret <4 x float> %res
}

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

define <4 x double> @test_vfnmadd256_pd(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2) {
; CHECK-LABEL: test_vfnmadd256_pd:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vfnmadd213pd %ymm2, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0xf5,0xac,0xc2]
; CHECK-NEXT:    # ymm0 = -(ymm1 * ymm0) + ymm2
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <4 x double> @llvm.x86.avx512.mask.vfnmadd.pd.256(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2, i8 -1) nounwind
  ret <4 x double> %res
}

define <4 x double> @test_mask_vfnmadd256_pd(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2, i8 %mask) {
; X86-LABEL: test_mask_vfnmadd256_pd:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vfnmadd132pd %ymm1, %ymm2, %ymm0 {%k1} # encoding: [0x62,0xf2,0xed,0x29,0x9c,0xc1]
; X86-NEXT:    # ymm0 = -(ymm0 * ymm1) + ymm2
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_vfnmadd256_pd:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vfnmadd132pd %ymm1, %ymm2, %ymm0 {%k1} # encoding: [0x62,0xf2,0xed,0x29,0x9c,0xc1]
; X64-NEXT:    # ymm0 = -(ymm0 * ymm1) + ymm2
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x double> @llvm.x86.avx512.mask.vfnmadd.pd.256(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2, i8 %mask) nounwind
  ret <4 x double> %res
}

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

define <2 x double> @test_vfnmadd128_pd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2) {
; CHECK-LABEL: test_vfnmadd128_pd:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vfnmadd213pd %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0xf1,0xac,0xc2]
; CHECK-NEXT:    # xmm0 = -(xmm1 * xmm0) + xmm2
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <2 x double> @llvm.x86.avx512.mask.vfnmadd.pd.128(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2, i8 -1) nounwind
  ret <2 x double> %res
}

define <2 x double> @test_mask_vfnmadd128_pd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2, i8 %mask) {
; X86-LABEL: test_mask_vfnmadd128_pd:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vfnmadd132pd %xmm1, %xmm2, %xmm0 {%k1} # encoding: [0x62,0xf2,0xed,0x09,0x9c,0xc1]
; X86-NEXT:    # xmm0 = -(xmm0 * xmm1) + xmm2
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_vfnmadd128_pd:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vfnmadd132pd %xmm1, %xmm2, %xmm0 {%k1} # encoding: [0x62,0xf2,0xed,0x09,0x9c,0xc1]
; X64-NEXT:    # xmm0 = -(xmm0 * xmm1) + xmm2
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x double> @llvm.x86.avx512.mask.vfnmadd.pd.128(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2, i8 %mask) nounwind
  ret <2 x double> %res
}

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

define <8 x float> @test_vfnmsub256_ps(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2) {
; CHECK-LABEL: test_vfnmsub256_ps:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vfnmsub213ps %ymm2, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x75,0xae,0xc2]
; CHECK-NEXT:    # ymm0 = -(ymm1 * ymm0) - ymm2
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask.vfnmsub.ps.256(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2, i8 -1) nounwind
  ret <8 x float> %res
}

define <8 x float> @test_mask_vfnmsub256_ps(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2, i8 %mask) {
; X86-LABEL: test_mask_vfnmsub256_ps:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vfnmsub132ps %ymm1, %ymm2, %ymm0 {%k1} # encoding: [0x62,0xf2,0x6d,0x29,0x9e,0xc1]
; X86-NEXT:    # ymm0 = -(ymm0 * ymm1) - ymm2
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_vfnmsub256_ps:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vfnmsub132ps %ymm1, %ymm2, %ymm0 {%k1} # encoding: [0x62,0xf2,0x6d,0x29,0x9e,0xc1]
; X64-NEXT:    # ymm0 = -(ymm0 * ymm1) - ymm2
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask.vfnmsub.ps.256(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2, i8 %mask) nounwind
  ret <8 x float> %res
}

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

define <4 x float> @test_vfnmsub128_ps(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2) {
; CHECK-LABEL: test_vfnmsub128_ps:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vfnmsub213ps %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x71,0xae,0xc2]
; CHECK-NEXT:    # xmm0 = -(xmm1 * xmm0) - xmm2
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask.vfnmsub.ps.128(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2, i8 -1) nounwind
  ret <4 x float> %res
}

define <4 x float> @test_mask_vfnmsub128_ps(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2, i8 %mask) {
; X86-LABEL: test_mask_vfnmsub128_ps:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vfnmsub132ps %xmm1, %xmm2, %xmm0 {%k1} # encoding: [0x62,0xf2,0x6d,0x09,0x9e,0xc1]
; X86-NEXT:    # xmm0 = -(xmm0 * xmm1) - xmm2
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_vfnmsub128_ps:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vfnmsub132ps %xmm1, %xmm2, %xmm0 {%k1} # encoding: [0x62,0xf2,0x6d,0x09,0x9e,0xc1]
; X64-NEXT:    # xmm0 = -(xmm0 * xmm1) - xmm2
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask.vfnmsub.ps.128(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2, i8 %mask) nounwind
  ret <4 x float> %res
}

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

define <4 x double> @test_vfnmsub256_pd(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2) {
; CHECK-LABEL: test_vfnmsub256_pd:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vfnmsub213pd %ymm2, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0xf5,0xae,0xc2]
; CHECK-NEXT:    # ymm0 = -(ymm1 * ymm0) - ymm2
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <4 x double> @llvm.x86.avx512.mask.vfnmsub.pd.256(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2, i8 -1) nounwind
  ret <4 x double> %res
}

define <4 x double> @test_mask_vfnmsub256_pd(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2, i8 %mask) {
; X86-LABEL: test_mask_vfnmsub256_pd:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vfnmsub132pd %ymm1, %ymm2, %ymm0 {%k1} # encoding: [0x62,0xf2,0xed,0x29,0x9e,0xc1]
; X86-NEXT:    # ymm0 = -(ymm0 * ymm1) - ymm2
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_vfnmsub256_pd:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vfnmsub132pd %ymm1, %ymm2, %ymm0 {%k1} # encoding: [0x62,0xf2,0xed,0x29,0x9e,0xc1]
; X64-NEXT:    # ymm0 = -(ymm0 * ymm1) - ymm2
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x double> @llvm.x86.avx512.mask.vfnmsub.pd.256(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2, i8 %mask) nounwind
  ret <4 x double> %res
}

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

define <2 x double> @test_vfnmsub128_pd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2) {
; CHECK-LABEL: test_vfnmsub128_pd:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vfnmsub213pd %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0xf1,0xae,0xc2]
; CHECK-NEXT:    # xmm0 = -(xmm1 * xmm0) - xmm2
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <2 x double> @llvm.x86.avx512.mask.vfnmsub.pd.128(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2, i8 -1) nounwind
  ret <2 x double> %res
}

define <2 x double> @test_mask_vfnmsub128_pd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2, i8 %mask) {
; X86-LABEL: test_mask_vfnmsub128_pd:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vfnmsub132pd %xmm1, %xmm2, %xmm0 {%k1} # encoding: [0x62,0xf2,0xed,0x09,0x9e,0xc1]
; X86-NEXT:    # xmm0 = -(xmm0 * xmm1) - xmm2
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_vfnmsub128_pd:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vfnmsub132pd %xmm1, %xmm2, %xmm0 {%k1} # encoding: [0x62,0xf2,0xed,0x09,0x9e,0xc1]
; X64-NEXT:    # xmm0 = -(xmm0 * xmm1) - xmm2
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x double> @llvm.x86.avx512.mask.vfnmsub.pd.128(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2, i8 %mask) nounwind
  ret <2 x double> %res
}

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

define <2 x double>@test_int_x86_avx512_mask3_vfnmsub_pd_128(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask3_vfnmsub_pd_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vfnmsub231pd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0xbe,0xd1]
; X86-NEXT:    # xmm2 = -(xmm0 * xmm1) - xmm2
; X86-NEXT:    vmovapd %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x28,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask3_vfnmsub_pd_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vfnmsub231pd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0xbe,0xd1]
; X64-NEXT:    # xmm2 = -(xmm0 * xmm1) - xmm2
; X64-NEXT:    vmovapd %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x28,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x double> @llvm.x86.avx512.mask3.vfnmsub.pd.128(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 %x3)
  ret <2 x double> %res
}

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

define <4 x double>@test_int_x86_avx512_mask3_vfnmsub_pd_256(<4 x double> %x0, <4 x double> %x1, <4 x double> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask3_vfnmsub_pd_256:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vfnmsub231pd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0xbe,0xd1]
; X86-NEXT:    # ymm2 = -(ymm0 * ymm1) - ymm2
; X86-NEXT:    vmovapd %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x28,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask3_vfnmsub_pd_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vfnmsub231pd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0xbe,0xd1]
; X64-NEXT:    # ymm2 = -(ymm0 * ymm1) - ymm2
; X64-NEXT:    vmovapd %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x28,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x double> @llvm.x86.avx512.mask3.vfnmsub.pd.256(<4 x double> %x0, <4 x double> %x1, <4 x double> %x2, i8 %x3)
  ret <4 x double> %res
}

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

define <4 x float>@test_int_x86_avx512_mask3_vfnmsub_ps_128(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask3_vfnmsub_ps_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vfnmsub231ps %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0xbe,0xd1]
; X86-NEXT:    # xmm2 = -(xmm0 * xmm1) - xmm2
; X86-NEXT:    vmovaps %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x28,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask3_vfnmsub_ps_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vfnmsub231ps %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0xbe,0xd1]
; X64-NEXT:    # xmm2 = -(xmm0 * xmm1) - xmm2
; X64-NEXT:    vmovaps %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x28,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask3.vfnmsub.ps.128(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 %x3)
  ret <4 x float> %res
}

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

define <8 x float>@test_int_x86_avx512_mask3_vfnmsub_ps_256(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask3_vfnmsub_ps_256:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vfnmsub231ps %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0xbe,0xd1]
; X86-NEXT:    # ymm2 = -(ymm0 * ymm1) - ymm2
; X86-NEXT:    vmovaps %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x28,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask3_vfnmsub_ps_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vfnmsub231ps %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0xbe,0xd1]
; X64-NEXT:    # ymm2 = -(ymm0 * ymm1) - ymm2
; X64-NEXT:    vmovaps %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x28,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask3.vfnmsub.ps.256(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2, i8 %x3)
  ret <8 x float> %res
}

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

define <8 x float> @test_fmaddsub256_ps(<8 x float> %a, <8 x float> %b, <8 x float> %c) {
; CHECK-LABEL: test_fmaddsub256_ps:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vfmaddsub213ps %ymm2, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x75,0xa6,0xc2]
; CHECK-NEXT:    # ymm0 = (ymm1 * ymm0) +/- ymm2
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask.vfmaddsub.ps.256(<8 x float> %a, <8 x float> %b, <8 x float> %c, i8 -1)
  ret <8 x float> %res
}

define <8 x float> @test_mask_fmaddsub256_ps(<8 x float> %a, <8 x float> %b, <8 x float> %c, i8 %mask) {
; X86-LABEL: test_mask_fmaddsub256_ps:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vfmaddsub132ps %ymm1, %ymm2, %ymm0 {%k1} # encoding: [0x62,0xf2,0x6d,0x29,0x96,0xc1]
; X86-NEXT:    # ymm0 = (ymm0 * ymm1) +/- ymm2
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_fmaddsub256_ps:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vfmaddsub132ps %ymm1, %ymm2, %ymm0 {%k1} # encoding: [0x62,0xf2,0x6d,0x29,0x96,0xc1]
; X64-NEXT:    # ymm0 = (ymm0 * ymm1) +/- ymm2
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask.vfmaddsub.ps.256(<8 x float> %a, <8 x float> %b, <8 x float> %c, i8 %mask)
  ret <8 x float> %res
}

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

define <4 x float> @test_fmaddsub128_ps(<4 x float> %a, <4 x float> %b, <4 x float> %c) {
; CHECK-LABEL: test_fmaddsub128_ps:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vfmaddsub213ps %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x71,0xa6,0xc2]
; CHECK-NEXT:    # xmm0 = (xmm1 * xmm0) +/- xmm2
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask.vfmaddsub.ps.128(<4 x float> %a, <4 x float> %b, <4 x float> %c, i8 -1)
  ret <4 x float> %res
}

define <4 x float> @test_mask_fmaddsub128_ps(<4 x float> %a, <4 x float> %b, <4 x float> %c, i8 %mask) {
; X86-LABEL: test_mask_fmaddsub128_ps:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vfmaddsub132ps %xmm1, %xmm2, %xmm0 {%k1} # encoding: [0x62,0xf2,0x6d,0x09,0x96,0xc1]
; X86-NEXT:    # xmm0 = (xmm0 * xmm1) +/- xmm2
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_fmaddsub128_ps:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vfmaddsub132ps %xmm1, %xmm2, %xmm0 {%k1} # encoding: [0x62,0xf2,0x6d,0x09,0x96,0xc1]
; X64-NEXT:    # xmm0 = (xmm0 * xmm1) +/- xmm2
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask.vfmaddsub.ps.128(<4 x float> %a, <4 x float> %b, <4 x float> %c, i8 %mask)
  ret <4 x float> %res
}

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

define <4 x double> @test_vfmaddsub256_pd(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2) {
; CHECK-LABEL: test_vfmaddsub256_pd:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vfmaddsub213pd %ymm2, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0xf5,0xa6,0xc2]
; CHECK-NEXT:    # ymm0 = (ymm1 * ymm0) +/- ymm2
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <4 x double> @llvm.x86.avx512.mask.vfmaddsub.pd.256(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2, i8 -1) nounwind
  ret <4 x double> %res
}

define <4 x double> @test_mask_vfmaddsub256_pd(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2, i8 %mask) {
; X86-LABEL: test_mask_vfmaddsub256_pd:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vfmaddsub132pd %ymm1, %ymm2, %ymm0 {%k1} # encoding: [0x62,0xf2,0xed,0x29,0x96,0xc1]
; X86-NEXT:    # ymm0 = (ymm0 * ymm1) +/- ymm2
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_vfmaddsub256_pd:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vfmaddsub132pd %ymm1, %ymm2, %ymm0 {%k1} # encoding: [0x62,0xf2,0xed,0x29,0x96,0xc1]
; X64-NEXT:    # ymm0 = (ymm0 * ymm1) +/- ymm2
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x double> @llvm.x86.avx512.mask.vfmaddsub.pd.256(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2, i8 %mask) nounwind
  ret <4 x double> %res
}

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

define <2 x double> @test_vfmaddsub128_pd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2) {
; CHECK-LABEL: test_vfmaddsub128_pd:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vfmaddsub213pd %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0xf1,0xa6,0xc2]
; CHECK-NEXT:    # xmm0 = (xmm1 * xmm0) +/- xmm2
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <2 x double> @llvm.x86.avx512.mask.vfmaddsub.pd.128(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2, i8 -1) nounwind
  ret <2 x double> %res
}

define <2 x double> @test_mask_vfmaddsub128_pd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2, i8 %mask) {
; X86-LABEL: test_mask_vfmaddsub128_pd:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vfmaddsub132pd %xmm1, %xmm2, %xmm0 {%k1} # encoding: [0x62,0xf2,0xed,0x09,0x96,0xc1]
; X86-NEXT:    # xmm0 = (xmm0 * xmm1) +/- xmm2
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_vfmaddsub128_pd:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vfmaddsub132pd %xmm1, %xmm2, %xmm0 {%k1} # encoding: [0x62,0xf2,0xed,0x09,0x96,0xc1]
; X64-NEXT:    # xmm0 = (xmm0 * xmm1) +/- xmm2
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x double> @llvm.x86.avx512.mask.vfmaddsub.pd.128(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2, i8 %mask) nounwind
  ret <2 x double> %res
}

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

define <2 x double>@test_int_x86_avx512_mask3_vfmaddsub_pd_128(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask3_vfmaddsub_pd_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vfmaddsub231pd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0xb6,0xd1]
; X86-NEXT:    # xmm2 = (xmm0 * xmm1) +/- xmm2
; X86-NEXT:    vmovapd %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x28,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask3_vfmaddsub_pd_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vfmaddsub231pd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0xb6,0xd1]
; X64-NEXT:    # xmm2 = (xmm0 * xmm1) +/- xmm2
; X64-NEXT:    vmovapd %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x28,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x double> @llvm.x86.avx512.mask3.vfmaddsub.pd.128(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 %x3)
  ret <2 x double> %res
}

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

define <2 x double>@test_int_x86_avx512_maskz_vfmaddsub_pd_128(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_maskz_vfmaddsub_pd_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vfmaddsub213pd %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0x89,0xa6,0xc2]
; X86-NEXT:    # xmm0 = (xmm1 * xmm0) +/- xmm2
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_maskz_vfmaddsub_pd_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vfmaddsub213pd %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0x89,0xa6,0xc2]
; X64-NEXT:    # xmm0 = (xmm1 * xmm0) +/- xmm2
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x double> @llvm.x86.avx512.maskz.vfmaddsub.pd.128(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 %x3)
  ret <2 x double> %res
}

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

define <4 x double>@test_int_x86_avx512_mask3_vfmaddsub_pd_256(<4 x double> %x0, <4 x double> %x1, <4 x double> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask3_vfmaddsub_pd_256:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vfmaddsub231pd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0xb6,0xd1]
; X86-NEXT:    # ymm2 = (ymm0 * ymm1) +/- ymm2
; X86-NEXT:    vmovapd %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x28,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask3_vfmaddsub_pd_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vfmaddsub231pd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0xb6,0xd1]
; X64-NEXT:    # ymm2 = (ymm0 * ymm1) +/- ymm2
; X64-NEXT:    vmovapd %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x28,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x double> @llvm.x86.avx512.mask3.vfmaddsub.pd.256(<4 x double> %x0, <4 x double> %x1, <4 x double> %x2, i8 %x3)
  ret <4 x double> %res
}

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

define <4 x double>@test_int_x86_avx512_maskz_vfmaddsub_pd_256(<4 x double> %x0, <4 x double> %x1, <4 x double> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_maskz_vfmaddsub_pd_256:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vfmaddsub213pd %ymm2, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0xa9,0xa6,0xc2]
; X86-NEXT:    # ymm0 = (ymm1 * ymm0) +/- ymm2
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_maskz_vfmaddsub_pd_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vfmaddsub213pd %ymm2, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0xa9,0xa6,0xc2]
; X64-NEXT:    # ymm0 = (ymm1 * ymm0) +/- ymm2
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x double> @llvm.x86.avx512.maskz.vfmaddsub.pd.256(<4 x double> %x0, <4 x double> %x1, <4 x double> %x2, i8 %x3)
  ret <4 x double> %res
}

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

define <4 x float>@test_int_x86_avx512_mask3_vfmaddsub_ps_128(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask3_vfmaddsub_ps_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vfmaddsub231ps %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0xb6,0xd1]
; X86-NEXT:    # xmm2 = (xmm0 * xmm1) +/- xmm2
; X86-NEXT:    vmovaps %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x28,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask3_vfmaddsub_ps_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vfmaddsub231ps %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0xb6,0xd1]
; X64-NEXT:    # xmm2 = (xmm0 * xmm1) +/- xmm2
; X64-NEXT:    vmovaps %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x28,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask3.vfmaddsub.ps.128(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 %x3)
  ret <4 x float> %res
}

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

define <4 x float>@test_int_x86_avx512_maskz_vfmaddsub_ps_128(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_maskz_vfmaddsub_ps_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vfmaddsub213ps %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0x89,0xa6,0xc2]
; X86-NEXT:    # xmm0 = (xmm1 * xmm0) +/- xmm2
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_maskz_vfmaddsub_ps_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vfmaddsub213ps %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0x89,0xa6,0xc2]
; X64-NEXT:    # xmm0 = (xmm1 * xmm0) +/- xmm2
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.maskz.vfmaddsub.ps.128(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 %x3)
  ret <4 x float> %res
}

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

define <8 x float>@test_int_x86_avx512_mask3_vfmaddsub_ps_256(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask3_vfmaddsub_ps_256:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vfmaddsub231ps %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0xb6,0xd1]
; X86-NEXT:    # ymm2 = (ymm0 * ymm1) +/- ymm2
; X86-NEXT:    vmovaps %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x28,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask3_vfmaddsub_ps_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vfmaddsub231ps %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0xb6,0xd1]
; X64-NEXT:    # ymm2 = (ymm0 * ymm1) +/- ymm2
; X64-NEXT:    vmovaps %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x28,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask3.vfmaddsub.ps.256(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2, i8 %x3)
  ret <8 x float> %res
}

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

define <8 x float>@test_int_x86_avx512_maskz_vfmaddsub_ps_256(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_maskz_vfmaddsub_ps_256:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vfmaddsub213ps %ymm2, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xa9,0xa6,0xc2]
; X86-NEXT:    # ymm0 = (ymm1 * ymm0) +/- ymm2
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_maskz_vfmaddsub_ps_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vfmaddsub213ps %ymm2, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xa9,0xa6,0xc2]
; X64-NEXT:    # ymm0 = (ymm1 * ymm0) +/- ymm2
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.maskz.vfmaddsub.ps.256(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2, i8 %x3)
  ret <8 x float> %res
}

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

define <2 x double>@test_int_x86_avx512_mask3_vfmsubadd_pd_128(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask3_vfmsubadd_pd_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vfmsubadd231pd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0xb7,0xd1]
; X86-NEXT:    # xmm2 = (xmm0 * xmm1) -/+ xmm2
; X86-NEXT:    vmovapd %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x28,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask3_vfmsubadd_pd_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vfmsubadd231pd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0xb7,0xd1]
; X64-NEXT:    # xmm2 = (xmm0 * xmm1) -/+ xmm2
; X64-NEXT:    vmovapd %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x28,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x double> @llvm.x86.avx512.mask3.vfmsubadd.pd.128(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 %x3)
  ret <2 x double> %res
}

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

define <4 x double>@test_int_x86_avx512_mask3_vfmsubadd_pd_256(<4 x double> %x0, <4 x double> %x1, <4 x double> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask3_vfmsubadd_pd_256:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vfmsubadd231pd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0xb7,0xd1]
; X86-NEXT:    # ymm2 = (ymm0 * ymm1) -/+ ymm2
; X86-NEXT:    vmovapd %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x28,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask3_vfmsubadd_pd_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vfmsubadd231pd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0xb7,0xd1]
; X64-NEXT:    # ymm2 = (ymm0 * ymm1) -/+ ymm2
; X64-NEXT:    vmovapd %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x28,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x double> @llvm.x86.avx512.mask3.vfmsubadd.pd.256(<4 x double> %x0, <4 x double> %x1, <4 x double> %x2, i8 %x3)
  ret <4 x double> %res
}

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

define <4 x float>@test_int_x86_avx512_mask3_vfmsubadd_ps_128(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask3_vfmsubadd_ps_128:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vfmsubadd231ps %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0xb7,0xd1]
; X86-NEXT:    # xmm2 = (xmm0 * xmm1) -/+ xmm2
; X86-NEXT:    vmovaps %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x28,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask3_vfmsubadd_ps_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vfmsubadd231ps %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0xb7,0xd1]
; X64-NEXT:    # xmm2 = (xmm0 * xmm1) -/+ xmm2
; X64-NEXT:    vmovaps %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x28,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask3.vfmsubadd.ps.128(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 %x3)
  ret <4 x float> %res
}

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

define <8 x float>@test_int_x86_avx512_mask3_vfmsubadd_ps_256(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask3_vfmsubadd_ps_256:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vfmsubadd231ps %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0xb7,0xd1]
; X86-NEXT:    # ymm2 = (ymm0 * ymm1) -/+ ymm2
; X86-NEXT:    vmovaps %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x28,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask3_vfmsubadd_ps_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vfmsubadd231ps %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0xb7,0xd1]
; X64-NEXT:    # ymm2 = (ymm0 * ymm1) -/+ ymm2
; X64-NEXT:    vmovaps %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x28,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask3.vfmsubadd.ps.256(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2, i8 %x3)
  ret <8 x float> %res
}


define <4 x float> @test_mask_vfmadd128_ps_rmk(<4 x float> %a0, <4 x float> %a1, <4 x float>* %ptr_a2, i8 %mask) {
; X86-LABEL: test_mask_vfmadd128_ps_rmk:
; 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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vfmadd213ps (%eax), %xmm1, %xmm0 {%k1} # encoding: [0x62,0xf2,0x75,0x09,0xa8,0x00]
; X86-NEXT:    # xmm0 = (xmm1 * xmm0) + mem
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_vfmadd128_ps_rmk:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vfmadd213ps (%rdi), %xmm1, %xmm0 {%k1} # encoding: [0x62,0xf2,0x75,0x09,0xa8,0x07]
; X64-NEXT:    # xmm0 = (xmm1 * xmm0) + mem
; X64-NEXT:    retq # encoding: [0xc3]
  %a2 = load <4 x float>, <4 x float>* %ptr_a2
  %res = call <4 x float> @llvm.x86.avx512.mask.vfmadd.ps.128(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2, i8 %mask) nounwind
  ret <4 x float> %res
}

define <4 x float> @test_mask_vfmadd128_ps_rmka(<4 x float> %a0, <4 x float> %a1, <4 x float>* %ptr_a2, i8 %mask) {
; X86-LABEL: test_mask_vfmadd128_ps_rmka:
; 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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vfmadd213ps (%eax), %xmm1, %xmm0 {%k1} # encoding: [0x62,0xf2,0x75,0x09,0xa8,0x00]
; X86-NEXT:    # xmm0 = (xmm1 * xmm0) + mem
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_vfmadd128_ps_rmka:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vfmadd213ps (%rdi), %xmm1, %xmm0 {%k1} # encoding: [0x62,0xf2,0x75,0x09,0xa8,0x07]
; X64-NEXT:    # xmm0 = (xmm1 * xmm0) + mem
; X64-NEXT:    retq # encoding: [0xc3]
  %a2 = load <4 x float>, <4 x float>* %ptr_a2, align 8
  %res = call <4 x float> @llvm.x86.avx512.mask.vfmadd.ps.128(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2, i8 %mask) nounwind
  ret <4 x float> %res
}

define <4 x float> @test_mask_vfmadd128_ps_rmkz(<4 x float> %a0, <4 x float> %a1, <4 x float>* %ptr_a2) {
; X86-LABEL: test_mask_vfmadd128_ps_rmkz:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vfmadd213ps (%eax), %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x71,0xa8,0x00]
; X86-NEXT:    # xmm0 = (xmm1 * xmm0) + mem
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_vfmadd128_ps_rmkz:
; X64:       # %bb.0:
; X64-NEXT:    vfmadd213ps (%rdi), %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x71,0xa8,0x07]
; X64-NEXT:    # xmm0 = (xmm1 * xmm0) + mem
; X64-NEXT:    retq # encoding: [0xc3]
  %a2 = load <4 x float>, <4 x float>* %ptr_a2
  %res = call <4 x float> @llvm.x86.avx512.mask.vfmadd.ps.128(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2, i8 -1) nounwind
  ret <4 x float> %res
}

define <4 x float> @test_mask_vfmadd128_ps_rmkza(<4 x float> %a0, <4 x float> %a1, <4 x float>* %ptr_a2) {
; X86-LABEL: test_mask_vfmadd128_ps_rmkza:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vfmadd213ps (%eax), %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x71,0xa8,0x00]
; X86-NEXT:    # xmm0 = (xmm1 * xmm0) + mem
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_vfmadd128_ps_rmkza:
; X64:       # %bb.0:
; X64-NEXT:    vfmadd213ps (%rdi), %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x71,0xa8,0x07]
; X64-NEXT:    # xmm0 = (xmm1 * xmm0) + mem
; X64-NEXT:    retq # encoding: [0xc3]
  %a2 = load <4 x float>, <4 x float>* %ptr_a2, align 4
  %res = call <4 x float> @llvm.x86.avx512.mask.vfmadd.ps.128(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2, i8 -1) nounwind
  ret <4 x float> %res
}

define <4 x float> @test_mask_vfmadd128_ps_rmb(<4 x float> %a0, <4 x float> %a1, float* %ptr_a2, i8 %mask) {
; X86-LABEL: test_mask_vfmadd128_ps_rmb:
; 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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vfmadd213ps (%eax){1to4}, %xmm1, %xmm0 {%k1} # encoding: [0x62,0xf2,0x75,0x19,0xa8,0x00]
; X86-NEXT:    # xmm0 = (xmm1 * xmm0) + mem
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_vfmadd128_ps_rmb:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vfmadd213ps (%rdi){1to4}, %xmm1, %xmm0 {%k1} # encoding: [0x62,0xf2,0x75,0x19,0xa8,0x07]
; X64-NEXT:    # xmm0 = (xmm1 * xmm0) + mem
; X64-NEXT:    retq # encoding: [0xc3]
  %q = load float, float* %ptr_a2
  %vecinit.i = insertelement <4 x float> undef, float %q, i32 0
  %vecinit2.i = insertelement <4 x float> %vecinit.i, float %q, i32 1
  %vecinit4.i = insertelement <4 x float> %vecinit2.i, float %q, i32 2
  %vecinit6.i = insertelement <4 x float> %vecinit4.i, float %q, i32 3
  %res = call <4 x float> @llvm.x86.avx512.mask.vfmadd.ps.128(<4 x float> %a0, <4 x float> %a1, <4 x float> %vecinit6.i, i8 %mask) nounwind
  ret <4 x float> %res
}

define <4 x float> @test_mask_vfmadd128_ps_rmba(<4 x float> %a0, <4 x float> %a1, float* %ptr_a2, i8 %mask) {
; X86-LABEL: test_mask_vfmadd128_ps_rmba:
; 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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vfmadd213ps (%eax){1to4}, %xmm1, %xmm0 {%k1} # encoding: [0x62,0xf2,0x75,0x19,0xa8,0x00]
; X86-NEXT:    # xmm0 = (xmm1 * xmm0) + mem
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_vfmadd128_ps_rmba:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vfmadd213ps (%rdi){1to4}, %xmm1, %xmm0 {%k1} # encoding: [0x62,0xf2,0x75,0x19,0xa8,0x07]
; X64-NEXT:    # xmm0 = (xmm1 * xmm0) + mem
; X64-NEXT:    retq # encoding: [0xc3]
  %q = load float, float* %ptr_a2, align 4
  %vecinit.i = insertelement <4 x float> undef, float %q, i32 0
  %vecinit2.i = insertelement <4 x float> %vecinit.i, float %q, i32 1
  %vecinit4.i = insertelement <4 x float> %vecinit2.i, float %q, i32 2
  %vecinit6.i = insertelement <4 x float> %vecinit4.i, float %q, i32 3
  %res = call <4 x float> @llvm.x86.avx512.mask.vfmadd.ps.128(<4 x float> %a0, <4 x float> %a1, <4 x float> %vecinit6.i, i8 %mask) nounwind
  ret <4 x float> %res
}

define <4 x float> @test_mask_vfmadd128_ps_rmbz(<4 x float> %a0, <4 x float> %a1, float* %ptr_a2) {
; X86-LABEL: test_mask_vfmadd128_ps_rmbz:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vfmadd213ps (%eax){1to4}, %xmm1, %xmm0 # encoding: [0x62,0xf2,0x75,0x18,0xa8,0x00]
; X86-NEXT:    # xmm0 = (xmm1 * xmm0) + mem
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_vfmadd128_ps_rmbz:
; X64:       # %bb.0:
; X64-NEXT:    vfmadd213ps (%rdi){1to4}, %xmm1, %xmm0 # encoding: [0x62,0xf2,0x75,0x18,0xa8,0x07]
; X64-NEXT:    # xmm0 = (xmm1 * xmm0) + mem
; X64-NEXT:    retq # encoding: [0xc3]
  %q = load float, float* %ptr_a2
  %vecinit.i = insertelement <4 x float> undef, float %q, i32 0
  %vecinit2.i = insertelement <4 x float> %vecinit.i, float %q, i32 1
  %vecinit4.i = insertelement <4 x float> %vecinit2.i, float %q, i32 2
  %vecinit6.i = insertelement <4 x float> %vecinit4.i, float %q, i32 3
  %res = call <4 x float> @llvm.x86.avx512.mask.vfmadd.ps.128(<4 x float> %a0, <4 x float> %a1, <4 x float> %vecinit6.i, i8 -1) nounwind
  ret <4 x float> %res
}

define <4 x float> @test_mask_vfmadd128_ps_rmbza(<4 x float> %a0, <4 x float> %a1, float* %ptr_a2) {
; X86-LABEL: test_mask_vfmadd128_ps_rmbza:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vfmadd213ps (%eax){1to4}, %xmm1, %xmm0 # encoding: [0x62,0xf2,0x75,0x18,0xa8,0x00]
; X86-NEXT:    # xmm0 = (xmm1 * xmm0) + mem
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_vfmadd128_ps_rmbza:
; X64:       # %bb.0:
; X64-NEXT:    vfmadd213ps (%rdi){1to4}, %xmm1, %xmm0 # encoding: [0x62,0xf2,0x75,0x18,0xa8,0x07]
; X64-NEXT:    # xmm0 = (xmm1 * xmm0) + mem
; X64-NEXT:    retq # encoding: [0xc3]
  %q = load float, float* %ptr_a2, align 4
  %vecinit.i = insertelement <4 x float> undef, float %q, i32 0
  %vecinit2.i = insertelement <4 x float> %vecinit.i, float %q, i32 1
  %vecinit4.i = insertelement <4 x float> %vecinit2.i, float %q, i32 2
  %vecinit6.i = insertelement <4 x float> %vecinit4.i, float %q, i32 3
  %res = call <4 x float> @llvm.x86.avx512.mask.vfmadd.ps.128(<4 x float> %a0, <4 x float> %a1, <4 x float> %vecinit6.i, i8 -1) nounwind
  ret <4 x float> %res
}

define <2 x double> @test_mask_vfmadd128_pd_rmk(<2 x double> %a0, <2 x double> %a1, <2 x double>* %ptr_a2, i8 %mask) {
; X86-LABEL: test_mask_vfmadd128_pd_rmk:
; 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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vfmadd213pd (%eax), %xmm1, %xmm0 {%k1} # encoding: [0x62,0xf2,0xf5,0x09,0xa8,0x00]
; X86-NEXT:    # xmm0 = (xmm1 * xmm0) + mem
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_vfmadd128_pd_rmk:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vfmadd213pd (%rdi), %xmm1, %xmm0 {%k1} # encoding: [0x62,0xf2,0xf5,0x09,0xa8,0x07]
; X64-NEXT:    # xmm0 = (xmm1 * xmm0) + mem
; X64-NEXT:    retq # encoding: [0xc3]
  %a2 = load <2 x double>, <2 x double>* %ptr_a2
  %res = call <2 x double> @llvm.x86.avx512.mask.vfmadd.pd.128(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2, i8 %mask) nounwind
  ret <2 x double> %res
}

define <2 x double> @test_mask_vfmadd128_pd_rmkz(<2 x double> %a0, <2 x double> %a1, <2 x double>* %ptr_a2) {
; X86-LABEL: test_mask_vfmadd128_pd_rmkz:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vfmadd213pd (%eax), %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0xf1,0xa8,0x00]
; X86-NEXT:    # xmm0 = (xmm1 * xmm0) + mem
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_vfmadd128_pd_rmkz:
; X64:       # %bb.0:
; X64-NEXT:    vfmadd213pd (%rdi), %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0xf1,0xa8,0x07]
; X64-NEXT:    # xmm0 = (xmm1 * xmm0) + mem
; X64-NEXT:    retq # encoding: [0xc3]
  %a2 = load <2 x double>, <2 x double>* %ptr_a2
  %res = call <2 x double> @llvm.x86.avx512.mask.vfmadd.pd.128(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2, i8 -1) nounwind
  ret <2 x double> %res
}

define <4 x double> @test_mask_vfmadd256_pd_rmk(<4 x double> %a0, <4 x double> %a1, <4 x double>* %ptr_a2, i8 %mask) {
; X86-LABEL: test_mask_vfmadd256_pd_rmk:
; 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:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
; X86-NEXT:    vfmadd213pd (%eax), %ymm1, %ymm0 {%k1} # encoding: [0x62,0xf2,0xf5,0x29,0xa8,0x00]
; X86-NEXT:    # ymm0 = (ymm1 * ymm0) + mem
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_vfmadd256_pd_rmk:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vfmadd213pd (%rdi), %ymm1, %ymm0 {%k1} # encoding: [0x62,0xf2,0xf5,0x29,0xa8,0x07]
; X64-NEXT:    # ymm0 = (ymm1 * ymm0) + mem
; X64-NEXT:    retq # encoding: [0xc3]
  %a2 = load <4 x double>, <4 x double>* %ptr_a2
  %res = call <4 x double> @llvm.x86.avx512.mask.vfmadd.pd.256(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2, i8 %mask) nounwind
  ret <4 x double> %res
}

define <4 x double> @test_mask_vfmadd256_pd_rmkz(<4 x double> %a0, <4 x double> %a1, <4 x double>* %ptr_a2) {
; X86-LABEL: test_mask_vfmadd256_pd_rmkz:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vfmadd213pd (%eax), %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0xf5,0xa8,0x00]
; X86-NEXT:    # ymm0 = (ymm1 * ymm0) + mem
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_vfmadd256_pd_rmkz:
; X64:       # %bb.0:
; X64-NEXT:    vfmadd213pd (%rdi), %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0xf5,0xa8,0x07]
; X64-NEXT:    # ymm0 = (ymm1 * ymm0) + mem
; X64-NEXT:    retq # encoding: [0xc3]
  %a2 = load <4 x double>, <4 x double>* %ptr_a2
  %res = call <4 x double> @llvm.x86.avx512.mask.vfmadd.pd.256(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2, i8 -1) nounwind
  ret <4 x double> %res
}