; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+avx512ifma,+avx512vl --show-mc-encoding | FileCheck %s --check-prefixes=CHECK,X86
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512ifma,+avx512vl --show-mc-encoding | FileCheck %s --check-prefixes=CHECK,X64

declare <2 x i64> @llvm.x86.avx512.vpmadd52h.uq.128(<2 x i64>, <2 x i64>, <2 x i64>)

define <2 x i64>@test_int_x86_avx512_vpmadd52h_uq_128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2) {
; CHECK-LABEL: test_int_x86_avx512_vpmadd52h_uq_128:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vpmadd52huq %xmm2, %xmm1, %xmm0 # encoding: [0x62,0xf2,0xf5,0x08,0xb5,0xc2]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]

  %1 = call <2 x i64> @llvm.x86.avx512.vpmadd52h.uq.128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2)
  ret <2 x i64> %1
}

define <2 x i64>@test_int_x86_avx512_mask_vpmadd52h_uq_128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_vpmadd52h_uq_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:    vpmadd52huq %xmm2, %xmm1, %xmm0 {%k1} # encoding: [0x62,0xf2,0xf5,0x09,0xb5,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_vpmadd52h_uq_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmadd52huq %xmm2, %xmm1, %xmm0 {%k1} # encoding: [0x62,0xf2,0xf5,0x09,0xb5,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]

  %1 = call <2 x i64> @llvm.x86.avx512.vpmadd52h.uq.128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2)
  %2 = bitcast i8 %x3 to <8 x i1>
  %extract2 = shufflevector <8 x i1> %2, <8 x i1> %2, <2 x i32> <i32 0, i32 1>
  %3 = select <2 x i1> %extract2, <2 x i64> %1, <2 x i64> %x0
  ret <2 x i64> %3
}

declare <4 x i64> @llvm.x86.avx512.vpmadd52h.uq.256(<4 x i64>, <4 x i64>, <4 x i64>)

define <4 x i64>@test_int_x86_avx512_vpmadd52h_uq_256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2) {
; CHECK-LABEL: test_int_x86_avx512_vpmadd52h_uq_256:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vpmadd52huq %ymm2, %ymm1, %ymm0 # encoding: [0x62,0xf2,0xf5,0x28,0xb5,0xc2]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]

  %1 = call <4 x i64> @llvm.x86.avx512.vpmadd52h.uq.256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2)
  ret <4 x i64> %1
}

define <4 x i64>@test_int_x86_avx512_mask_vpmadd52h_uq_256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_vpmadd52h_uq_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:    vpmadd52huq %ymm2, %ymm1, %ymm0 {%k1} # encoding: [0x62,0xf2,0xf5,0x29,0xb5,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_vpmadd52h_uq_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmadd52huq %ymm2, %ymm1, %ymm0 {%k1} # encoding: [0x62,0xf2,0xf5,0x29,0xb5,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]

  %1 = call <4 x i64> @llvm.x86.avx512.vpmadd52h.uq.256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2)
  %2 = bitcast i8 %x3 to <8 x i1>
  %extract2 = 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> %extract2, <4 x i64> %1, <4 x i64> %x0
  ret <4 x i64> %3
}

define <2 x i64>@test_int_x86_avx512_maskz_vpmadd52h_uq_128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_maskz_vpmadd52h_uq_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:    vpmadd52huq %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0x89,0xb5,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_maskz_vpmadd52h_uq_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmadd52huq %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0x89,0xb5,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]

  %1 = call <2 x i64> @llvm.x86.avx512.vpmadd52h.uq.128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2)
  %2 = bitcast i8 %x3 to <8 x i1>
  %extract2 = shufflevector <8 x i1> %2, <8 x i1> %2, <2 x i32> <i32 0, i32 1>
  %3 = select <2 x i1> %extract2, <2 x i64> %1, <2 x i64> zeroinitializer
  ret <2 x i64> %3
}

define <4 x i64>@test_int_x86_avx512_maskz_vpmadd52h_uq_256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_maskz_vpmadd52h_uq_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:    vpmadd52huq %ymm2, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0xa9,0xb5,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_maskz_vpmadd52h_uq_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmadd52huq %ymm2, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0xa9,0xb5,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]

  %1 = call <4 x i64> @llvm.x86.avx512.vpmadd52h.uq.256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2)
  %2 = bitcast i8 %x3 to <8 x i1>
  %extract2 = 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> %extract2, <4 x i64> %1, <4 x i64> zeroinitializer
  ret <4 x i64> %3
}

declare <2 x i64> @llvm.x86.avx512.vpmadd52l.uq.128(<2 x i64>, <2 x i64>, <2 x i64>)

define <2 x i64>@test_int_x86_avx512_vpmadd52l_uq_128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2) {
; CHECK-LABEL: test_int_x86_avx512_vpmadd52l_uq_128:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vpmadd52luq %xmm2, %xmm1, %xmm0 # encoding: [0x62,0xf2,0xf5,0x08,0xb4,0xc2]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]

  %1 = call <2 x i64> @llvm.x86.avx512.vpmadd52l.uq.128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2)
  ret <2 x i64> %1
}

define <2 x i64>@test_int_x86_avx512_mask_vpmadd52l_uq_128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_vpmadd52l_uq_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:    vpmadd52luq %xmm2, %xmm1, %xmm0 {%k1} # encoding: [0x62,0xf2,0xf5,0x09,0xb4,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_vpmadd52l_uq_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmadd52luq %xmm2, %xmm1, %xmm0 {%k1} # encoding: [0x62,0xf2,0xf5,0x09,0xb4,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]

  %1 = call <2 x i64> @llvm.x86.avx512.vpmadd52l.uq.128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2)
  %2 = bitcast i8 %x3 to <8 x i1>
  %extract2 = shufflevector <8 x i1> %2, <8 x i1> %2, <2 x i32> <i32 0, i32 1>
  %3 = select <2 x i1> %extract2, <2 x i64> %1, <2 x i64> %x0
  ret <2 x i64> %3
}

declare <4 x i64> @llvm.x86.avx512.vpmadd52l.uq.256(<4 x i64>, <4 x i64>, <4 x i64>)

define <4 x i64>@test_int_x86_avx512_vpmadd52l_uq_256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2) {
; CHECK-LABEL: test_int_x86_avx512_vpmadd52l_uq_256:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vpmadd52luq %ymm2, %ymm1, %ymm0 # encoding: [0x62,0xf2,0xf5,0x28,0xb4,0xc2]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]

  %1 = call <4 x i64> @llvm.x86.avx512.vpmadd52l.uq.256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2)
  ret <4 x i64> %1
}

define <4 x i64>@test_int_x86_avx512_mask_vpmadd52l_uq_256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_vpmadd52l_uq_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:    vpmadd52luq %ymm2, %ymm1, %ymm0 {%k1} # encoding: [0x62,0xf2,0xf5,0x29,0xb4,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_vpmadd52l_uq_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmadd52luq %ymm2, %ymm1, %ymm0 {%k1} # encoding: [0x62,0xf2,0xf5,0x29,0xb4,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]

  %1 = call <4 x i64> @llvm.x86.avx512.vpmadd52l.uq.256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2)
  %2 = bitcast i8 %x3 to <8 x i1>
  %extract2 = 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> %extract2, <4 x i64> %1, <4 x i64> %x0
  ret <4 x i64> %3
}

define <2 x i64>@test_int_x86_avx512_maskz_vpmadd52l_uq_128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_maskz_vpmadd52l_uq_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:    vpmadd52luq %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0x89,0xb4,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_maskz_vpmadd52l_uq_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmadd52luq %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0x89,0xb4,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]

  %1 = call <2 x i64> @llvm.x86.avx512.vpmadd52l.uq.128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2)
  %2 = bitcast i8 %x3 to <8 x i1>
  %extract2 = shufflevector <8 x i1> %2, <8 x i1> %2, <2 x i32> <i32 0, i32 1>
  %3 = select <2 x i1> %extract2, <2 x i64> %1, <2 x i64> zeroinitializer
  ret <2 x i64> %3
}

define <4 x i64>@test_int_x86_avx512_maskz_vpmadd52l_uq_256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_maskz_vpmadd52l_uq_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:    vpmadd52luq %ymm2, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0xa9,0xb4,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_maskz_vpmadd52l_uq_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmadd52luq %ymm2, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0xa9,0xb4,0xc2]
; X64-NEXT:    retq # encoding: [0xc3]

  %1 = call <4 x i64> @llvm.x86.avx512.vpmadd52l.uq.256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2)
  %2 = bitcast i8 %x3 to <8 x i1>
  %extract2 = 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> %extract2, <4 x i64> %1, <4 x i64> zeroinitializer
  ret <4 x i64> %3
}