; 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

define <2 x double> @test_mask_compress_pd_128(<2 x double> %data, <2 x double> %passthru, i8 %mask) {
; X86-LABEL: test_mask_compress_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:    vcompresspd %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x8a,0xc1]
; X86-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_compress_pd_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vcompresspd %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x8a,0xc1]
; X64-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = bitcast i8 %mask to <8 x i1>
  %extract = shufflevector <8 x i1> %1, <8 x i1> %1, <2 x i32> <i32 0, i32 1>
  %2 = call <2 x double> @llvm.x86.avx512.mask.compress.v2f64(<2 x double> %data, <2 x double> %passthru, <2 x i1> %extract)
  ret <2 x double> %2
}

define <2 x double> @test_maskz_compress_pd_128(<2 x double> %data, i8 %mask) {
; X86-LABEL: test_maskz_compress_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:    vcompresspd %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x8a,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_maskz_compress_pd_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vcompresspd %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x8a,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = bitcast i8 %mask to <8 x i1>
  %extract = shufflevector <8 x i1> %1, <8 x i1> %1, <2 x i32> <i32 0, i32 1>
  %2 = call <2 x double> @llvm.x86.avx512.mask.compress.v2f64(<2 x double> %data, <2 x double> zeroinitializer, <2 x i1> %extract)
  ret <2 x double> %2
}

define <2 x double> @test_compress_pd_128(<2 x double> %data) {
; CHECK-LABEL: test_compress_pd_128:
; CHECK:       # %bb.0:
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %1 = call <2 x double> @llvm.x86.avx512.mask.compress.v2f64(<2 x double> %data, <2 x double> undef, <2 x i1> <i1 true, i1 true>)
  ret <2 x double> %1
}

define <4 x float> @test_mask_compress_ps_128(<4 x float> %data, <4 x float> %passthru, i8 %mask) {
; X86-LABEL: test_mask_compress_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:    vcompressps %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x8a,0xc1]
; X86-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_compress_ps_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vcompressps %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x8a,0xc1]
; X64-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = bitcast i8 %mask to <8 x i1>
  %extract = shufflevector <8 x i1> %1, <8 x i1> %1, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %2 = call <4 x float> @llvm.x86.avx512.mask.compress.v4f32(<4 x float> %data, <4 x float> %passthru, <4 x i1> %extract)
  ret <4 x float> %2
}

define <4 x float> @test_maskz_compress_ps_128(<4 x float> %data, i8 %mask) {
; X86-LABEL: test_maskz_compress_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:    vcompressps %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x8a,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_maskz_compress_ps_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vcompressps %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x8a,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = bitcast i8 %mask to <8 x i1>
  %extract = shufflevector <8 x i1> %1, <8 x i1> %1, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %2 = call <4 x float> @llvm.x86.avx512.mask.compress.v4f32(<4 x float> %data, <4 x float> zeroinitializer, <4 x i1> %extract)
  ret <4 x float> %2
}

define <4 x float> @test_compress_ps_128(<4 x float> %data) {
; CHECK-LABEL: test_compress_ps_128:
; CHECK:       # %bb.0:
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %1 = call <4 x float> @llvm.x86.avx512.mask.compress.v4f32(<4 x float> %data, <4 x float> undef, <4 x i1> <i1 true, i1 true, i1 true, i1 true>)
  ret <4 x float> %1
}

define <2 x i64> @test_mask_compress_q_128(<2 x i64> %data, <2 x i64> %passthru, i8 %mask) {
; X86-LABEL: test_mask_compress_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:    vpcompressq %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x8b,0xc1]
; X86-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_compress_q_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpcompressq %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x8b,0xc1]
; X64-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = bitcast i8 %mask to <8 x i1>
  %extract = shufflevector <8 x i1> %1, <8 x i1> %1, <2 x i32> <i32 0, i32 1>
  %2 = call <2 x i64> @llvm.x86.avx512.mask.compress.v2i64(<2 x i64> %data, <2 x i64> %passthru, <2 x i1> %extract)
  ret <2 x i64> %2
}

define <2 x i64> @test_maskz_compress_q_128(<2 x i64> %data, i8 %mask) {
; X86-LABEL: test_maskz_compress_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:    vpcompressq %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x8b,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_maskz_compress_q_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpcompressq %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x8b,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = bitcast i8 %mask to <8 x i1>
  %extract = shufflevector <8 x i1> %1, <8 x i1> %1, <2 x i32> <i32 0, i32 1>
  %2 = call <2 x i64> @llvm.x86.avx512.mask.compress.v2i64(<2 x i64> %data, <2 x i64> zeroinitializer, <2 x i1> %extract)
  ret <2 x i64> %2
}

define <2 x i64> @test_compress_q_128(<2 x i64> %data) {
; CHECK-LABEL: test_compress_q_128:
; CHECK:       # %bb.0:
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %1 = call <2 x i64> @llvm.x86.avx512.mask.compress.v2i64(<2 x i64> %data, <2 x i64> undef, <2 x i1> <i1 true, i1 true>)
  ret <2 x i64> %1
}

define <4 x i32> @test_mask_compress_d_128(<4 x i32> %data, <4 x i32> %passthru, i8 %mask) {
; X86-LABEL: test_mask_compress_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:    vpcompressd %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x8b,0xc1]
; X86-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_compress_d_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpcompressd %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x8b,0xc1]
; X64-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = bitcast i8 %mask to <8 x i1>
  %extract = shufflevector <8 x i1> %1, <8 x i1> %1, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %2 = call <4 x i32> @llvm.x86.avx512.mask.compress.v4i32(<4 x i32> %data, <4 x i32> %passthru, <4 x i1> %extract)
  ret <4 x i32> %2
}

define <4 x i32> @test_maskz_compress_d_128(<4 x i32> %data, i8 %mask) {
; X86-LABEL: test_maskz_compress_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:    vpcompressd %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x8b,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_maskz_compress_d_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpcompressd %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x8b,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = bitcast i8 %mask to <8 x i1>
  %extract = shufflevector <8 x i1> %1, <8 x i1> %1, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %2 = call <4 x i32> @llvm.x86.avx512.mask.compress.v4i32(<4 x i32> %data, <4 x i32> zeroinitializer, <4 x i1> %extract)
  ret <4 x i32> %2
}

define <4 x i32> @test_compress_d_128(<4 x i32> %data) {
; CHECK-LABEL: test_compress_d_128:
; CHECK:       # %bb.0:
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %1 = call <4 x i32> @llvm.x86.avx512.mask.compress.v4i32(<4 x i32> %data, <4 x i32> undef, <4 x i1> <i1 true, i1 true, i1 true, i1 true>)
  ret <4 x i32> %1
}

define <2 x double> @test_expand_pd_128(<2 x double> %data) {
; CHECK-LABEL: test_expand_pd_128:
; CHECK:       # %bb.0:
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %1 = call <2 x double> @llvm.x86.avx512.mask.expand.v2f64(<2 x double> %data, <2 x double> undef, <2 x i1> <i1 true, i1 true>)
  ret <2 x double> %1
}

define <2 x double> @test_mask_expand_pd_128(<2 x double> %data, <2 x double> %passthru, i8 %mask) {
; X86-LABEL: test_mask_expand_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:    vexpandpd %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x88,0xc8]
; X86-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_expand_pd_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vexpandpd %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x88,0xc8]
; X64-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = bitcast i8 %mask to <8 x i1>
  %extract = shufflevector <8 x i1> %1, <8 x i1> %1, <2 x i32> <i32 0, i32 1>
  %2 = call <2 x double> @llvm.x86.avx512.mask.expand.v2f64(<2 x double> %data, <2 x double> %passthru, <2 x i1> %extract)
  ret <2 x double> %2
}

define <2 x double> @test_maskz_expand_pd_128(<2 x double> %data, i8 %mask) {
; X86-LABEL: test_maskz_expand_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:    vexpandpd %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x88,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_maskz_expand_pd_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vexpandpd %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x88,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = bitcast i8 %mask to <8 x i1>
  %extract = shufflevector <8 x i1> %1, <8 x i1> %1, <2 x i32> <i32 0, i32 1>
  %2 = call <2 x double> @llvm.x86.avx512.mask.expand.v2f64(<2 x double> %data, <2 x double> zeroinitializer, <2 x i1> %extract)
  ret <2 x double> %2
}

define <4 x float> @test_expand_ps_128(<4 x float> %data) {
; CHECK-LABEL: test_expand_ps_128:
; CHECK:       # %bb.0:
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %1 = call <4 x float> @llvm.x86.avx512.mask.expand.v4f32(<4 x float> %data, <4 x float> undef, <4 x i1> <i1 true, i1 true, i1 true, i1 true>)
  ret <4 x float> %1
}

define <4 x float> @test_mask_expand_ps_128(<4 x float> %data, <4 x float> %passthru, i8 %mask) {
; X86-LABEL: test_mask_expand_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:    vexpandps %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x88,0xc8]
; X86-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_expand_ps_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vexpandps %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x88,0xc8]
; X64-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = bitcast i8 %mask to <8 x i1>
  %extract = shufflevector <8 x i1> %1, <8 x i1> %1, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %2 = call <4 x float> @llvm.x86.avx512.mask.expand.v4f32(<4 x float> %data, <4 x float> %passthru, <4 x i1> %extract)
  ret <4 x float> %2
}

define <4 x float> @test_maskz_expand_ps_128(<4 x float> %data, i8 %mask) {
; X86-LABEL: test_maskz_expand_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:    vexpandps %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x88,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_maskz_expand_ps_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vexpandps %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x88,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = bitcast i8 %mask to <8 x i1>
  %extract = shufflevector <8 x i1> %1, <8 x i1> %1, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %2 = call <4 x float> @llvm.x86.avx512.mask.expand.v4f32(<4 x float> %data, <4 x float> zeroinitializer, <4 x i1> %extract)
  ret <4 x float> %2
}

define <2 x i64> @test_expand_q_128(<2 x i64> %data) {
; CHECK-LABEL: test_expand_q_128:
; CHECK:       # %bb.0:
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %1 = call <2 x i64> @llvm.x86.avx512.mask.expand.v2i64(<2 x i64> %data, <2 x i64> undef, <2 x i1> <i1 true, i1 true>)
  ret <2 x i64> %1
}

define <2 x i64> @test_mask_expand_q_128(<2 x i64> %data, <2 x i64> %passthru, i8 %mask) {
; X86-LABEL: test_mask_expand_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:    vpexpandq %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x89,0xc8]
; X86-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_expand_q_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpexpandq %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x89,0xc8]
; X64-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = bitcast i8 %mask to <8 x i1>
  %extract = shufflevector <8 x i1> %1, <8 x i1> %1, <2 x i32> <i32 0, i32 1>
  %2 = call <2 x i64> @llvm.x86.avx512.mask.expand.v2i64(<2 x i64> %data, <2 x i64> %passthru, <2 x i1> %extract)
  ret <2 x i64> %2
}

define <2 x i64> @test_maskz_expand_q_128(<2 x i64> %data, i8 %mask) {
; X86-LABEL: test_maskz_expand_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:    vpexpandq %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x89,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_maskz_expand_q_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpexpandq %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x89,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = bitcast i8 %mask to <8 x i1>
  %extract = shufflevector <8 x i1> %1, <8 x i1> %1, <2 x i32> <i32 0, i32 1>
  %2 = call <2 x i64> @llvm.x86.avx512.mask.expand.v2i64(<2 x i64> %data, <2 x i64> zeroinitializer, <2 x i1> %extract)
  ret <2 x i64> %2
}

define <4 x i32> @test_expand_d_128(<4 x i32> %data) {
; CHECK-LABEL: test_expand_d_128:
; CHECK:       # %bb.0:
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %1 = call <4 x i32> @llvm.x86.avx512.mask.expand.v4i32(<4 x i32> %data, <4 x i32> undef, <4 x i1> <i1 true, i1 true, i1 true, i1 true>)
  ret <4 x i32> %1
}

define <4 x i32> @test_mask_expand_d_128(<4 x i32> %data, <4 x i32> %passthru, i8 %mask) {
; X86-LABEL: test_mask_expand_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:    vpexpandd %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x89,0xc8]
; X86-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_expand_d_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpexpandd %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x89,0xc8]
; X64-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = bitcast i8 %mask to <8 x i1>
  %extract = shufflevector <8 x i1> %1, <8 x i1> %1, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %2 = call <4 x i32> @llvm.x86.avx512.mask.expand.v4i32(<4 x i32> %data, <4 x i32> %passthru, <4 x i1> %extract)
  ret <4 x i32> %2
}

define <4 x i32> @test_maskz_expand_d_128(<4 x i32> %data, i8 %mask) {
; X86-LABEL: test_maskz_expand_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:    vpexpandd %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x89,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_maskz_expand_d_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpexpandd %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x89,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = bitcast i8 %mask to <8 x i1>
  %extract = shufflevector <8 x i1> %1, <8 x i1> %1, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %2 = call <4 x i32> @llvm.x86.avx512.mask.expand.v4i32(<4 x i32> %data, <4 x i32> zeroinitializer, <4 x i1> %extract)
  ret <4 x i32> %2
}

define <4 x double> @test_mask_compress_pd_256(<4 x double> %data, <4 x double> %passthru, i8 %mask) {
; X86-LABEL: test_mask_compress_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:    vcompresspd %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x8a,0xc1]
; X86-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_compress_pd_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vcompresspd %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x8a,0xc1]
; X64-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = bitcast i8 %mask to <8 x i1>
  %extract = shufflevector <8 x i1> %1, <8 x i1> %1, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %2 = call <4 x double> @llvm.x86.avx512.mask.compress.v4f64(<4 x double> %data, <4 x double> %passthru, <4 x i1> %extract)
  ret <4 x double> %2
}

define <4 x double> @test_maskz_compress_pd_256(<4 x double> %data, i8 %mask) {
; X86-LABEL: test_maskz_compress_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:    vcompresspd %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x8a,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_maskz_compress_pd_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vcompresspd %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x8a,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = bitcast i8 %mask to <8 x i1>
  %extract = shufflevector <8 x i1> %1, <8 x i1> %1, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %2 = call <4 x double> @llvm.x86.avx512.mask.compress.v4f64(<4 x double> %data, <4 x double> zeroinitializer, <4 x i1> %extract)
  ret <4 x double> %2
}

define <4 x double> @test_compress_pd_256(<4 x double> %data) {
; CHECK-LABEL: test_compress_pd_256:
; CHECK:       # %bb.0:
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %1 = call <4 x double> @llvm.x86.avx512.mask.compress.v4f64(<4 x double> %data, <4 x double> undef, <4 x i1> <i1 true, i1 true, i1 true, i1 true>)
  ret <4 x double> %1
}

define <8 x float> @test_mask_compress_ps_256(<8 x float> %data, <8 x float> %passthru, i8 %mask) {
; X86-LABEL: test_mask_compress_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:    vcompressps %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x8a,0xc1]
; X86-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_compress_ps_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vcompressps %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x8a,0xc1]
; X64-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = bitcast i8 %mask to <8 x i1>
  %2 = call <8 x float> @llvm.x86.avx512.mask.compress.v8f32(<8 x float> %data, <8 x float> %passthru, <8 x i1> %1)
  ret <8 x float> %2
}

define <8 x float> @test_maskz_compress_ps_256(<8 x float> %data, i8 %mask) {
; X86-LABEL: test_maskz_compress_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:    vcompressps %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x8a,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_maskz_compress_ps_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vcompressps %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x8a,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = bitcast i8 %mask to <8 x i1>
  %2 = call <8 x float> @llvm.x86.avx512.mask.compress.v8f32(<8 x float> %data, <8 x float> zeroinitializer, <8 x i1> %1)
  ret <8 x float> %2
}

define <8 x float> @test_compress_ps_256(<8 x float> %data) {
; CHECK-LABEL: test_compress_ps_256:
; CHECK:       # %bb.0:
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %1 = call <8 x float> @llvm.x86.avx512.mask.compress.v8f32(<8 x float> %data, <8 x float> undef, <8 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>)
  ret <8 x float> %1
}

define <4 x i64> @test_mask_compress_q_256(<4 x i64> %data, <4 x i64> %passthru, i8 %mask) {
; X86-LABEL: test_mask_compress_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:    vpcompressq %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x8b,0xc1]
; X86-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_compress_q_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpcompressq %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x8b,0xc1]
; X64-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = bitcast i8 %mask to <8 x i1>
  %extract = shufflevector <8 x i1> %1, <8 x i1> %1, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %2 = call <4 x i64> @llvm.x86.avx512.mask.compress.v4i64(<4 x i64> %data, <4 x i64> %passthru, <4 x i1> %extract)
  ret <4 x i64> %2
}

define <4 x i64> @test_maskz_compress_q_256(<4 x i64> %data, i8 %mask) {
; X86-LABEL: test_maskz_compress_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:    vpcompressq %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x8b,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_maskz_compress_q_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpcompressq %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x8b,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = bitcast i8 %mask to <8 x i1>
  %extract = shufflevector <8 x i1> %1, <8 x i1> %1, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %2 = call <4 x i64> @llvm.x86.avx512.mask.compress.v4i64(<4 x i64> %data, <4 x i64> zeroinitializer, <4 x i1> %extract)
  ret <4 x i64> %2
}

define <4 x i64> @test_compress_q_256(<4 x i64> %data) {
; CHECK-LABEL: test_compress_q_256:
; CHECK:       # %bb.0:
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %1 = call <4 x i64> @llvm.x86.avx512.mask.compress.v4i64(<4 x i64> %data, <4 x i64> undef, <4 x i1> <i1 true, i1 true, i1 true, i1 true>)
  ret <4 x i64> %1
}

define <8 x i32> @test_mask_compress_d_256(<8 x i32> %data, <8 x i32> %passthru, i8 %mask) {
; X86-LABEL: test_mask_compress_d_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:    vpcompressd %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x8b,0xc1]
; X86-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_compress_d_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpcompressd %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x8b,0xc1]
; X64-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = bitcast i8 %mask to <8 x i1>
  %2 = call <8 x i32> @llvm.x86.avx512.mask.compress.v8i32(<8 x i32> %data, <8 x i32> %passthru, <8 x i1> %1)
  ret <8 x i32> %2
}

define <8 x i32> @test_maskz_compress_d_256(<8 x i32> %data, i8 %mask) {
; X86-LABEL: test_maskz_compress_d_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:    vpcompressd %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x8b,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_maskz_compress_d_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpcompressd %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x8b,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = bitcast i8 %mask to <8 x i1>
  %2 = call <8 x i32> @llvm.x86.avx512.mask.compress.v8i32(<8 x i32> %data, <8 x i32> zeroinitializer, <8 x i1> %1)
  ret <8 x i32> %2
}

define <8 x i32> @test_compress_d_256(<8 x i32> %data) {
; CHECK-LABEL: test_compress_d_256:
; CHECK:       # %bb.0:
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %1 = call <8 x i32> @llvm.x86.avx512.mask.compress.v8i32(<8 x i32> %data, <8 x i32> undef, <8 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>)
  ret <8 x i32> %1
}

define <4 x double> @test_expand_pd_256(<4 x double> %data) {
; CHECK-LABEL: test_expand_pd_256:
; CHECK:       # %bb.0:
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %1 = call <4 x double> @llvm.x86.avx512.mask.expand.v4f64(<4 x double> %data, <4 x double> undef, <4 x i1> <i1 true, i1 true, i1 true, i1 true>)
  ret <4 x double> %1
}

define <4 x double> @test_mask_expand_pd_256(<4 x double> %data, <4 x double> %passthru, i8 %mask) {
; X86-LABEL: test_mask_expand_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:    vexpandpd %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x88,0xc8]
; X86-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_expand_pd_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vexpandpd %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x88,0xc8]
; X64-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = bitcast i8 %mask to <8 x i1>
  %extract = shufflevector <8 x i1> %1, <8 x i1> %1, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %2 = call <4 x double> @llvm.x86.avx512.mask.expand.v4f64(<4 x double> %data, <4 x double> %passthru, <4 x i1> %extract)
  ret <4 x double> %2
}

define <4 x double> @test_maskz_expand_pd_256(<4 x double> %data, i8 %mask) {
; X86-LABEL: test_maskz_expand_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:    vexpandpd %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x88,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_maskz_expand_pd_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vexpandpd %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x88,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = bitcast i8 %mask to <8 x i1>
  %extract = shufflevector <8 x i1> %1, <8 x i1> %1, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %2 = call <4 x double> @llvm.x86.avx512.mask.expand.v4f64(<4 x double> %data, <4 x double> zeroinitializer, <4 x i1> %extract)
  ret <4 x double> %2
}

define <8 x float> @test_expand_ps_256(<8 x float> %data) {
; CHECK-LABEL: test_expand_ps_256:
; CHECK:       # %bb.0:
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %1 = call <8 x float> @llvm.x86.avx512.mask.expand.v8f32(<8 x float> %data, <8 x float> undef, <8 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>)
  ret <8 x float> %1
}

define <8 x float> @test_mask_expand_ps_256(<8 x float> %data, <8 x float> %passthru, i8 %mask) {
; X86-LABEL: test_mask_expand_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:    vexpandps %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x88,0xc8]
; X86-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_expand_ps_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vexpandps %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x88,0xc8]
; X64-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = bitcast i8 %mask to <8 x i1>
  %2 = call <8 x float> @llvm.x86.avx512.mask.expand.v8f32(<8 x float> %data, <8 x float> %passthru, <8 x i1> %1)
  ret <8 x float> %2
}

define <8 x float> @test_maskz_expand_ps_256(<8 x float> %data, i8 %mask) {
; X86-LABEL: test_maskz_expand_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:    vexpandps %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x88,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_maskz_expand_ps_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vexpandps %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x88,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = bitcast i8 %mask to <8 x i1>
  %2 = call <8 x float> @llvm.x86.avx512.mask.expand.v8f32(<8 x float> %data, <8 x float> zeroinitializer, <8 x i1> %1)
  ret <8 x float> %2
}

define <4 x i64> @test_expand_q_256(<4 x i64> %data) {
; CHECK-LABEL: test_expand_q_256:
; CHECK:       # %bb.0:
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %1 = call <4 x i64> @llvm.x86.avx512.mask.expand.v4i64(<4 x i64> %data, <4 x i64> undef, <4 x i1> <i1 true, i1 true, i1 true, i1 true>)
  ret <4 x i64> %1
}

define <4 x i64> @test_mask_expand_q_256(<4 x i64> %data, <4 x i64> %passthru, i8 %mask) {
; X86-LABEL: test_mask_expand_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:    vpexpandq %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x89,0xc8]
; X86-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_expand_q_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpexpandq %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x89,0xc8]
; X64-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = bitcast i8 %mask to <8 x i1>
  %extract = shufflevector <8 x i1> %1, <8 x i1> %1, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %2 = call <4 x i64> @llvm.x86.avx512.mask.expand.v4i64(<4 x i64> %data, <4 x i64> %passthru, <4 x i1> %extract)
  ret <4 x i64> %2
}

define <4 x i64> @test_maskz_expand_q_256(<4 x i64> %data, i8 %mask) {
; X86-LABEL: test_maskz_expand_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:    vpexpandq %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x89,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_maskz_expand_q_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpexpandq %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x89,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = bitcast i8 %mask to <8 x i1>
  %extract = shufflevector <8 x i1> %1, <8 x i1> %1, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %2 = call <4 x i64> @llvm.x86.avx512.mask.expand.v4i64(<4 x i64> %data, <4 x i64> zeroinitializer, <4 x i1> %extract)
  ret <4 x i64> %2
}

define <8 x i32> @test_expand_d_256(<8 x i32> %data) {
; CHECK-LABEL: test_expand_d_256:
; CHECK:       # %bb.0:
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %1 = call <8 x i32> @llvm.x86.avx512.mask.expand.v8i32(<8 x i32> %data, <8 x i32> undef, <8 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>)
  ret <8 x i32> %1
}

define <8 x i32> @test_mask_expand_d_256(<8 x i32> %data, <8 x i32> %passthru, i8 %mask) {
; X86-LABEL: test_mask_expand_d_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:    vpexpandd %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x89,0xc8]
; X86-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_expand_d_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpexpandd %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x89,0xc8]
; X64-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = bitcast i8 %mask to <8 x i1>
  %2 = call <8 x i32> @llvm.x86.avx512.mask.expand.v8i32(<8 x i32> %data, <8 x i32> %passthru, <8 x i1> %1)
  ret <8 x i32> %2
}

define <8 x i32> @test_maskz_expand_d_256(<8 x i32> %data, i8 %mask) {
; X86-LABEL: test_maskz_expand_d_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:    vpexpandd %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x89,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_maskz_expand_d_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpexpandd %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x89,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = bitcast i8 %mask to <8 x i1>
  %2 = call <8 x i32> @llvm.x86.avx512.mask.expand.v8i32(<8 x i32> %data, <8 x i32> zeroinitializer, <8 x i1> %1)
  ret <8 x i32> %2
}

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 <8 x i1> @llvm.x86.avx512.cmp.ps.256(<8 x float> %a, <8 x float> %b, i32 2)
  %1 = bitcast <8 x i1> %res to i8
  ret i8 %1
}
declare <8 x i1> @llvm.x86.avx512.cmp.ps.256(<8 x float>, <8 x float>, i32)

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 <4 x i1> @llvm.x86.avx512.cmp.ps.128(<4 x float> %a, <4 x float> %b, i32 2)
  %1 = shufflevector <4 x i1> %res, <4 x i1> zeroinitializer, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
  %2 = bitcast <8 x i1> %1 to i8
  ret i8 %2
}
declare <4 x i1> @llvm.x86.avx512.cmp.ps.128(<4 x float>, <4 x float>, i32)

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 <4 x i1> @llvm.x86.avx512.cmp.pd.256(<4 x double> %a, <4 x double> %b, i32 2)
  %1 = shufflevector <4 x i1> %res, <4 x i1> zeroinitializer, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
  %2 = bitcast <8 x i1> %1 to i8
  ret i8 %2
}
declare <4 x i1> @llvm.x86.avx512.cmp.pd.256(<4 x double>, <4 x double>, i32)

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 <2 x i1> @llvm.x86.avx512.cmp.pd.128(<2 x double> %a, <2 x double> %b, i32 2)
  %1 = shufflevector <2 x i1> %res, <2 x i1> zeroinitializer, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 2, i32 3, i32 2, i32 3>
  %2 = bitcast <8 x i1> %1 to i8
  ret i8 %2
}
declare <2 x i1> @llvm.x86.avx512.cmp.pd.128(<2 x double>, <2 x double>, i32)

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]
  %1 = call <8 x float> @llvm.x86.avx.max.ps.256(<8 x float> %a0, <8 x float> %a1)
  %2 = bitcast i8 %mask to <8 x i1>
  %3 = select <8 x i1> %2, <8 x float> %1, <8 x float> zeroinitializer
  ret <8 x float> %3
}

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]
  %1 = call <8 x float> @llvm.x86.avx.max.ps.256(<8 x float> %a0, <8 x float> %a1)
  %2 = bitcast i8 %mask to <8 x i1>
  %3 = select <8 x i1> %2, <8 x float> %1, <8 x float> %src
  ret <8 x float> %3
}

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]
  %1 = call <8 x float> @llvm.x86.avx.max.ps.256(<8 x float> %a0, <8 x float> %a1)
  ret <8 x float> %1
}
declare <8 x float> @llvm.x86.avx.max.ps.256(<8 x float>, <8 x float>)

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]
  %1 = call <4 x float> @llvm.x86.sse.max.ps(<4 x float> %a0, <4 x float> %a1)
  %2 = bitcast i8 %mask to <8 x i1>
  %extract = shufflevector <8 x i1> %2, <8 x i1> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %3 = select <4 x i1> %extract, <4 x float> %1, <4 x float> zeroinitializer
  ret <4 x float> %3
}

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]
  %1 = call <4 x float> @llvm.x86.sse.max.ps(<4 x float> %a0, <4 x float> %a1)
  %2 = bitcast i8 %mask to <8 x i1>
  %extract = shufflevector <8 x i1> %2, <8 x i1> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %3 = select <4 x i1> %extract, <4 x float> %1, <4 x float> %src
  ret <4 x float> %3
}

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]
  %1 = call <4 x float> @llvm.x86.sse.max.ps(<4 x float> %a0, <4 x float> %a1)
  ret <4 x float> %1
}
declare <4 x float> @llvm.x86.sse.max.ps(<4 x float>, <4 x float>)

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]
  %1 = call <8 x float> @llvm.x86.avx.min.ps.256(<8 x float> %a0, <8 x float> %a1)
  %2 = bitcast i8 %mask to <8 x i1>
  %3 = select <8 x i1> %2, <8 x float> %1, <8 x float> zeroinitializer
  ret <8 x float> %3
}

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]
  %1 = call <8 x float> @llvm.x86.avx.min.ps.256(<8 x float> %a0, <8 x float> %a1)
  %2 = bitcast i8 %mask to <8 x i1>
  %3 = select <8 x i1> %2, <8 x float> %1, <8 x float> %src
  ret <8 x float> %3
}

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]
  %1 = call <8 x float> @llvm.x86.avx.min.ps.256(<8 x float> %a0, <8 x float> %a1)
  ret <8 x float> %1
}
declare <8 x float> @llvm.x86.avx.min.ps.256(<8 x float>, <8 x float>)

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]
  %1 = call <4 x float> @llvm.x86.sse.min.ps(<4 x float> %a0, <4 x float> %a1)
  %2 = bitcast i8 %mask to <8 x i1>
  %extract = shufflevector <8 x i1> %2, <8 x i1> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %3 = select <4 x i1> %extract, <4 x float> %1, <4 x float> zeroinitializer
  ret <4 x float> %3
}

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]
  %1 = call <4 x float> @llvm.x86.sse.min.ps(<4 x float> %a0, <4 x float> %a1)
  %2 = bitcast i8 %mask to <8 x i1>
  %extract = shufflevector <8 x i1> %2, <8 x i1> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %3 = select <4 x i1> %extract, <4 x float> %1, <4 x float> %src
  ret <4 x float> %3
}

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]
  %1 = call <4 x float> @llvm.x86.sse.min.ps(<4 x float> %a0, <4 x float> %a1)
  ret <4 x float> %1
}
declare <4 x float> @llvm.x86.sse.min.ps(<4 x float>, <4 x float>)

define <4 x double> @test_getexp_pd_256(<4 x double> %a0) {
; CHECK-LABEL: test_getexp_pd_256:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vgetexppd %ymm0, %ymm0 # encoding: [0x62,0xf2,0xfd,0x28,0x42,0xc0]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <4 x double> @llvm.x86.avx512.mask.getexp.pd.256(<4 x double> %a0,  <4 x double> zeroinitializer, i8 -1)
  ret <4 x double> %res
}

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

define <8 x float> @test_getexp_ps_256(<8 x float> %a0) {
; CHECK-LABEL: test_getexp_ps_256:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vgetexpps %ymm0, %ymm0 # encoding: [0x62,0xf2,0x7d,0x28,0x42,0xc0]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask.getexp.ps.256(<8 x float> %a0, <8 x float> zeroinitializer, i8 -1)
  ret <8 x float> %res
}
declare <8 x float> @llvm.x86.avx512.mask.getexp.ps.256(<8 x float>, <8 x float>, i8) nounwind readnone

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

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]
  %1 = call <4 x i32> @llvm.x86.avx512.vpermi2var.d.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2)
  %2 = bitcast i8 %x3 to <8 x i1>
  %extract = 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> %extract, <4 x i32> %1, <4 x i32> %x1
  %4 = call <4 x i32> @llvm.x86.avx512.vpermi2var.d.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2)
  %res2 = add <4 x i32> %3, %4
  ret <4 x i32> %res2
}

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]
  %1 = call <4 x i32> @llvm.x86.avx512.vpermi2var.d.128(<4 x i32> %x1, <4 x i32> %x0, <4 x i32> %x2)
  %2 = bitcast i8 %x3 to <8 x i1>
  %extract = 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> %extract, <4 x i32> %1, <4 x i32> %x1
  %4 = call <4 x i32> @llvm.x86.avx512.vpermi2var.d.128(<4 x i32> %x1, <4 x i32> %x0, <4 x i32> %x2)
  %res2 = add <4 x i32> %3, %4
  ret <4 x i32> %res2
}

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]
  %1 = call <4 x i32> @llvm.x86.avx512.vpermi2var.d.128(<4 x i32> %x1, <4 x i32> %x0, <4 x i32> %x2)
  %2 = bitcast i8 %x3 to <8 x i1>
  %extract = 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> %extract, <4 x i32> %1, <4 x i32> zeroinitializer
  %4 = call <4 x i32> @llvm.x86.avx512.vpermi2var.d.128(<4 x i32> %x1, <4 x i32> %x0, <4 x i32> %x2)
  %res2 = add <4 x i32> %3, %4
  ret <4 x i32> %res2
}

declare <8 x i32> @llvm.x86.avx512.vpermi2var.d.256(<8 x i32>, <8 x i32>, <8 x i32>)

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]
  %1 = call <8 x i32> @llvm.x86.avx512.vpermi2var.d.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2)
  %2 = bitcast i8 %x3 to <8 x i1>
  %3 = select <8 x i1> %2, <8 x i32> %1, <8 x i32> %x1
  %4 = call <8 x i32> @llvm.x86.avx512.vpermi2var.d.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2)
  %res2 = add <8 x i32> %3, %4
  ret <8 x i32> %res2
}

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]
  %1 = call <8 x i32> @llvm.x86.avx512.vpermi2var.d.256(<8 x i32> %x1, <8 x i32> %x0, <8 x i32> %x2)
  %2 = bitcast i8 %x3 to <8 x i1>
  %3 = select <8 x i1> %2, <8 x i32> %1, <8 x i32> %x1
  %4 = call <8 x i32> @llvm.x86.avx512.vpermi2var.d.256(<8 x i32> %x1, <8 x i32> %x0, <8 x i32> %x2)
  %res2 = add <8 x i32> %3, %4
  ret <8 x i32> %res2
}

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]
  %1 = call <8 x i32> @llvm.x86.avx512.vpermi2var.d.256(<8 x i32> %x1, <8 x i32> %x0, <8 x i32> %x2)
  %2 = bitcast i8 %x3 to <8 x i1>
  %3 = select <8 x i1> %2, <8 x i32> %1, <8 x i32> zeroinitializer
  %4 = call <8 x i32> @llvm.x86.avx512.vpermi2var.d.256(<8 x i32> %x1, <8 x i32> %x0, <8 x i32> %x2)
  %res2 = add <8 x i32> %3, %4
  ret <8 x i32> %res2
}

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

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]
  %1 = call <2 x double> @llvm.x86.avx512.vpermi2var.pd.128(<2 x double> %x0, <2 x i64> %x1, <2 x double> %x2)
  %2 = bitcast <2 x i64> %x1 to <2 x double>
  %3 = bitcast i8 %x3 to <8 x i1>
  %extract = shufflevector <8 x i1> %3, <8 x i1> %3, <2 x i32> <i32 0, i32 1>
  %4 = select <2 x i1> %extract, <2 x double> %1, <2 x double> %2
  %5 = call <2 x double> @llvm.x86.avx512.vpermi2var.pd.128(<2 x double> %x0, <2 x i64> %x1, <2 x double> %x2)
  %6 = bitcast <2 x i64> %x1 to <2 x double>
  %res2 = fadd <2 x double> %4, %5
  ret <2 x double> %res2
}

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

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]
  %1 = call <4 x double> @llvm.x86.avx512.vpermi2var.pd.256(<4 x double> %x0, <4 x i64> %x1, <4 x double> %x2)
  %2 = bitcast <4 x i64> %x1 to <4 x double>
  %3 = bitcast i8 %x3 to <8 x i1>
  %extract = shufflevector <8 x i1> %3, <8 x i1> %3, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %4 = select <4 x i1> %extract, <4 x double> %1, <4 x double> %2
  %5 = call <4 x double> @llvm.x86.avx512.vpermi2var.pd.256(<4 x double> %x0, <4 x i64> %x1, <4 x double> %x2)
  %6 = bitcast <4 x i64> %x1 to <4 x double>
  %res2 = fadd <4 x double> %4, %5
  ret <4 x double> %res2
}

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

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]
  %1 = call <4 x float> @llvm.x86.avx512.vpermi2var.ps.128(<4 x float> %x0, <4 x i32> %x1, <4 x float> %x2)
  %2 = bitcast <4 x i32> %x1 to <4 x float>
  %3 = bitcast i8 %x3 to <8 x i1>
  %extract = shufflevector <8 x i1> %3, <8 x i1> %3, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %4 = select <4 x i1> %extract, <4 x float> %1, <4 x float> %2
  %5 = call <4 x float> @llvm.x86.avx512.vpermi2var.ps.128(<4 x float> %x0, <4 x i32> %x1, <4 x float> %x2)
  %6 = bitcast <4 x i32> %x1 to <4 x float>
  %res2 = fadd <4 x float> %4, %5
  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>
  %1 = call <4 x float> @llvm.x86.avx512.vpermi2var.ps.128(<4 x float> %x0, <4 x i32> %x1cast, <4 x float> %x2)
  %2 = bitcast <4 x i32> %x1cast to <4 x float>
  %3 = bitcast i8 %x3 to <8 x i1>
  %extract = shufflevector <8 x i1> %3, <8 x i1> %3, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %4 = select <4 x i1> %extract, <4 x float> %1, <4 x float> %2
  ret <4 x float> %4
}

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

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]
  %1 = call <8 x float> @llvm.x86.avx512.vpermi2var.ps.256(<8 x float> %x0, <8 x i32> %x1, <8 x float> %x2)
  %2 = bitcast <8 x i32> %x1 to <8 x float>
  %3 = bitcast i8 %x3 to <8 x i1>
  %4 = select <8 x i1> %3, <8 x float> %1, <8 x float> %2
  %5 = call <8 x float> @llvm.x86.avx512.vpermi2var.ps.256(<8 x float> %x0, <8 x i32> %x1, <8 x float> %x2)
  %6 = bitcast <8 x i32> %x1 to <8 x float>
  %res2 = fadd <8 x float> %4, %5
  ret <8 x float> %res2
}

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

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]
  %1 = call <2 x i64> @llvm.x86.avx512.vpermi2var.q.128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2)
  %2 = bitcast i8 %x3 to <8 x i1>
  %extract = shufflevector <8 x i1> %2, <8 x i1> %2, <2 x i32> <i32 0, i32 1>
  %3 = select <2 x i1> %extract, <2 x i64> %1, <2 x i64> %x1
  %4 = call <2 x i64> @llvm.x86.avx512.vpermi2var.q.128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2)
  %res2 = add <2 x i64> %3, %4
  ret <2 x i64> %res2
}

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]
  %1 = call <2 x i64> @llvm.x86.avx512.vpermi2var.q.128(<2 x i64> %x1, <2 x i64> %x0, <2 x i64> %x2)
  %2 = bitcast i8 %x3 to <8 x i1>
  %extract = shufflevector <8 x i1> %2, <8 x i1> %2, <2 x i32> <i32 0, i32 1>
  %3 = select <2 x i1> %extract, <2 x i64> %1, <2 x i64> %x1
  %4 = call <2 x i64> @llvm.x86.avx512.vpermi2var.q.128(<2 x i64> %x1, <2 x i64> %x0, <2 x i64> %x2)
  %res2 = add <2 x i64> %3, %4
  ret <2 x i64> %res2
}

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]
  %1 = call <2 x i64> @llvm.x86.avx512.vpermi2var.q.128(<2 x i64> %x1, <2 x i64> %x0, <2 x i64> %x2)
  %2 = bitcast i8 %x3 to <8 x i1>
  %extract = shufflevector <8 x i1> %2, <8 x i1> %2, <2 x i32> <i32 0, i32 1>
  %3 = select <2 x i1> %extract, <2 x i64> %1, <2 x i64> zeroinitializer
  %4 = call <2 x i64> @llvm.x86.avx512.vpermi2var.q.128(<2 x i64> %x1, <2 x i64> %x0, <2 x i64> %x2)
  %res2 = add <2 x i64> %3, %4
  ret <2 x i64> %res2
}

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

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]
  %1 = call <4 x i64> @llvm.x86.avx512.vpermi2var.q.256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2)
  %2 = bitcast i8 %x3 to <8 x i1>
  %extract = 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> %extract, <4 x i64> %1, <4 x i64> %x1
  %4 = call <4 x i64> @llvm.x86.avx512.vpermi2var.q.256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2)
  %res2 = add <4 x i64> %3, %4
  ret <4 x i64> %res2
}

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]
  %1 = call <4 x i64> @llvm.x86.avx512.vpermi2var.q.256(<4 x i64> %x1, <4 x i64> %x0, <4 x i64> %x2)
  %2 = bitcast i8 %x3 to <8 x i1>
  %extract = 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> %extract, <4 x i64> %1, <4 x i64> %x1
  %4 = call <4 x i64> @llvm.x86.avx512.vpermi2var.q.256(<4 x i64> %x1, <4 x i64> %x0, <4 x i64> %x2)
  %res2 = add <4 x i64> %3, %4
  ret <4 x i64> %res2
}

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]
  %1 = call <4 x i64> @llvm.x86.avx512.vpermi2var.q.256(<4 x i64> %x1, <4 x i64> %x0, <4 x i64> %x2)
  %2 = bitcast i8 %x3 to <8 x i1>
  %extract = 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> %extract, <4 x i64> %1, <4 x i64> zeroinitializer
  %4 = call <4 x i64> @llvm.x86.avx512.vpermi2var.q.256(<4 x i64> %x1, <4 x i64> %x0, <4 x i64> %x2)
  %res2 = add <4 x i64> %3, %4
  ret <4 x i64> %res2
}

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

define <2 x double>@test_int_x86_avx512_mask_scalef_pd_128(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_scalef_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:    vscalefpd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x2c,0xd1]
; X86-NEXT:    vscalefpd %xmm1, %xmm0, %xmm0 # encoding: [0x62,0xf2,0xfd,0x08,0x2c,0xc1]
; X86-NEXT:    vaddpd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x58,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_scalef_pd_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vscalefpd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x2c,0xd1]
; X64-NEXT:    vscalefpd %xmm1, %xmm0, %xmm0 # encoding: [0x62,0xf2,0xfd,0x08,0x2c,0xc1]
; X64-NEXT:    vaddpd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x58,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x double> @llvm.x86.avx512.mask.scalef.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.scalef.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.scalef.pd.256(<4 x double>, <4 x double>, <4 x double>, i8)

define <4 x double>@test_int_x86_avx512_mask_scalef_pd_256(<4 x double> %x0, <4 x double> %x1, <4 x double> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_scalef_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:    vscalefpd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x2c,0xd1]
; X86-NEXT:    vscalefpd %ymm1, %ymm0, %ymm0 # encoding: [0x62,0xf2,0xfd,0x28,0x2c,0xc1]
; 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_scalef_pd_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vscalefpd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x2c,0xd1]
; X64-NEXT:    vscalefpd %ymm1, %ymm0, %ymm0 # encoding: [0x62,0xf2,0xfd,0x28,0x2c,0xc1]
; 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.scalef.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.scalef.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.scalef.ps.128(<4 x float>, <4 x float>, <4 x float>, i8)

define <4 x float>@test_int_x86_avx512_mask_scalef_ps_128(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_scalef_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:    vscalefps %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x2c,0xd1]
; X86-NEXT:    vscalefps %xmm1, %xmm0, %xmm0 # encoding: [0x62,0xf2,0x7d,0x08,0x2c,0xc1]
; 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_scalef_ps_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vscalefps %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x2c,0xd1]
; X64-NEXT:    vscalefps %xmm1, %xmm0, %xmm0 # encoding: [0x62,0xf2,0x7d,0x08,0x2c,0xc1]
; 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.scalef.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.scalef.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.scalef.ps.256(<8 x float>, <8 x float>, <8 x float>, i8)

define <8 x float>@test_int_x86_avx512_mask_scalef_ps_256(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_scalef_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:    vscalefps %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x2c,0xd1]
; X86-NEXT:    vscalefps %ymm1, %ymm0, %ymm0 # encoding: [0x62,0xf2,0x7d,0x28,0x2c,0xc1]
; X86-NEXT:    vaddps %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xec,0x58,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_scalef_ps_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vscalefps %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x2c,0xd1]
; X64-NEXT:    vscalefps %ymm1, %ymm0, %ymm0 # encoding: [0x62,0xf2,0x7d,0x28,0x2c,0xc1]
; X64-NEXT:    vaddps %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xec,0x58,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask.scalef.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.scalef.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 <16 x i8> @llvm.x86.avx512.mask.pmov.qb.128(<2 x i64>, <16 x i8>, i8)

define <16 x i8>@test_int_x86_avx512_mask_pmov_qb_128(<2 x i64> %x0, <16 x i8> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmov_qb_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:    vpmovqb %xmm0, %xmm2 # encoding: [0x62,0xf2,0x7e,0x08,0x32,0xc2]
; X86-NEXT:    vpmovqb %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x32,0xc1]
; X86-NEXT:    vpmovqb %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0x89,0x32,0xc0]
; X86-NEXT:    vpaddb %xmm0, %xmm1, %xmm0 # encoding: [0xc5,0xf1,0xfc,0xc0]
; X86-NEXT:    vpaddb %xmm0, %xmm2, %xmm0 # encoding: [0xc5,0xe9,0xfc,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmov_qb_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmovqb %xmm0, %xmm2 # encoding: [0x62,0xf2,0x7e,0x08,0x32,0xc2]
; X64-NEXT:    vpmovqb %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x32,0xc1]
; X64-NEXT:    vpmovqb %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0x89,0x32,0xc0]
; X64-NEXT:    vpaddb %xmm0, %xmm1, %xmm0 # encoding: [0xc5,0xf1,0xfc,0xc0]
; X64-NEXT:    vpaddb %xmm0, %xmm2, %xmm0 # encoding: [0xc5,0xe9,0xfc,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
    %res0 = call <16 x i8> @llvm.x86.avx512.mask.pmov.qb.128(<2 x i64> %x0, <16 x i8> %x1, i8 -1)
    %res1 = call <16 x i8> @llvm.x86.avx512.mask.pmov.qb.128(<2 x i64> %x0, <16 x i8> %x1, i8 %x2)
    %res2 = call <16 x i8> @llvm.x86.avx512.mask.pmov.qb.128(<2 x i64> %x0, <16 x i8> zeroinitializer, i8 %x2)
    %res3 = add <16 x i8> %res0, %res1
    %res4 = add <16 x i8> %res3, %res2
    ret <16 x i8> %res4
}

declare void @llvm.x86.avx512.mask.pmov.qb.mem.128(i8* %ptr, <2 x i64>, i8)

define void @test_int_x86_avx512_mask_pmov_qb_mem_128(i8* %ptr, <2 x i64> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmov_qb_mem_128:
; X86:       # %bb.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:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpmovqb %xmm0, (%eax) # encoding: [0x62,0xf2,0x7e,0x08,0x32,0x00]
; X86-NEXT:    vpmovqb %xmm0, (%eax) {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x32,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmov_qb_mem_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpmovqb %xmm0, (%rdi) # encoding: [0x62,0xf2,0x7e,0x08,0x32,0x07]
; X64-NEXT:    vpmovqb %xmm0, (%rdi) {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x32,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
    call void @llvm.x86.avx512.mask.pmov.qb.mem.128(i8* %ptr, <2 x i64> %x1, i8 -1)
    call void @llvm.x86.avx512.mask.pmov.qb.mem.128(i8* %ptr, <2 x i64> %x1, i8 %x2)
    ret void
}

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

define <16 x i8>@test_int_x86_avx512_mask_pmovs_qb_128(<2 x i64> %x0, <16 x i8> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmovs_qb_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:    vpmovsqb %xmm0, %xmm2 # encoding: [0x62,0xf2,0x7e,0x08,0x22,0xc2]
; X86-NEXT:    vpmovsqb %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x22,0xc1]
; X86-NEXT:    vpmovsqb %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0x89,0x22,0xc0]
; X86-NEXT:    vpaddb %xmm0, %xmm1, %xmm0 # encoding: [0xc5,0xf1,0xfc,0xc0]
; X86-NEXT:    vpaddb %xmm0, %xmm2, %xmm0 # encoding: [0xc5,0xe9,0xfc,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmovs_qb_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmovsqb %xmm0, %xmm2 # encoding: [0x62,0xf2,0x7e,0x08,0x22,0xc2]
; X64-NEXT:    vpmovsqb %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x22,0xc1]
; X64-NEXT:    vpmovsqb %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0x89,0x22,0xc0]
; X64-NEXT:    vpaddb %xmm0, %xmm1, %xmm0 # encoding: [0xc5,0xf1,0xfc,0xc0]
; X64-NEXT:    vpaddb %xmm0, %xmm2, %xmm0 # encoding: [0xc5,0xe9,0xfc,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
    %res0 = call <16 x i8> @llvm.x86.avx512.mask.pmovs.qb.128(<2 x i64> %x0, <16 x i8> %x1, i8 -1)
    %res1 = call <16 x i8> @llvm.x86.avx512.mask.pmovs.qb.128(<2 x i64> %x0, <16 x i8> %x1, i8 %x2)
    %res2 = call <16 x i8> @llvm.x86.avx512.mask.pmovs.qb.128(<2 x i64> %x0, <16 x i8> zeroinitializer, i8 %x2)
    %res3 = add <16 x i8> %res0, %res1
    %res4 = add <16 x i8> %res3, %res2
    ret <16 x i8> %res4
}

declare void @llvm.x86.avx512.mask.pmovs.qb.mem.128(i8* %ptr, <2 x i64>, i8)

define void @test_int_x86_avx512_mask_pmovs_qb_mem_128(i8* %ptr, <2 x i64> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmovs_qb_mem_128:
; X86:       # %bb.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:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpmovsqb %xmm0, (%eax) # encoding: [0x62,0xf2,0x7e,0x08,0x22,0x00]
; X86-NEXT:    vpmovsqb %xmm0, (%eax) {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x22,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmovs_qb_mem_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpmovsqb %xmm0, (%rdi) # encoding: [0x62,0xf2,0x7e,0x08,0x22,0x07]
; X64-NEXT:    vpmovsqb %xmm0, (%rdi) {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x22,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
    call void @llvm.x86.avx512.mask.pmovs.qb.mem.128(i8* %ptr, <2 x i64> %x1, i8 -1)
    call void @llvm.x86.avx512.mask.pmovs.qb.mem.128(i8* %ptr, <2 x i64> %x1, i8 %x2)
    ret void
}

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

define <16 x i8>@test_int_x86_avx512_mask_pmovus_qb_128(<2 x i64> %x0, <16 x i8> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmovus_qb_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:    vpmovusqb %xmm0, %xmm2 # encoding: [0x62,0xf2,0x7e,0x08,0x12,0xc2]
; X86-NEXT:    vpmovusqb %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x12,0xc1]
; X86-NEXT:    vpmovusqb %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0x89,0x12,0xc0]
; X86-NEXT:    vpaddb %xmm0, %xmm1, %xmm0 # encoding: [0xc5,0xf1,0xfc,0xc0]
; X86-NEXT:    vpaddb %xmm0, %xmm2, %xmm0 # encoding: [0xc5,0xe9,0xfc,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmovus_qb_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmovusqb %xmm0, %xmm2 # encoding: [0x62,0xf2,0x7e,0x08,0x12,0xc2]
; X64-NEXT:    vpmovusqb %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x12,0xc1]
; X64-NEXT:    vpmovusqb %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0x89,0x12,0xc0]
; X64-NEXT:    vpaddb %xmm0, %xmm1, %xmm0 # encoding: [0xc5,0xf1,0xfc,0xc0]
; X64-NEXT:    vpaddb %xmm0, %xmm2, %xmm0 # encoding: [0xc5,0xe9,0xfc,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
    %res0 = call <16 x i8> @llvm.x86.avx512.mask.pmovus.qb.128(<2 x i64> %x0, <16 x i8> %x1, i8 -1)
    %res1 = call <16 x i8> @llvm.x86.avx512.mask.pmovus.qb.128(<2 x i64> %x0, <16 x i8> %x1, i8 %x2)
    %res2 = call <16 x i8> @llvm.x86.avx512.mask.pmovus.qb.128(<2 x i64> %x0, <16 x i8> zeroinitializer, i8 %x2)
    %res3 = add <16 x i8> %res0, %res1
    %res4 = add <16 x i8> %res3, %res2
    ret <16 x i8> %res4
}

declare void @llvm.x86.avx512.mask.pmovus.qb.mem.128(i8* %ptr, <2 x i64>, i8)

define void @test_int_x86_avx512_mask_pmovus_qb_mem_128(i8* %ptr, <2 x i64> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmovus_qb_mem_128:
; X86:       # %bb.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:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpmovusqb %xmm0, (%eax) # encoding: [0x62,0xf2,0x7e,0x08,0x12,0x00]
; X86-NEXT:    vpmovusqb %xmm0, (%eax) {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x12,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmovus_qb_mem_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpmovusqb %xmm0, (%rdi) # encoding: [0x62,0xf2,0x7e,0x08,0x12,0x07]
; X64-NEXT:    vpmovusqb %xmm0, (%rdi) {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x12,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
    call void @llvm.x86.avx512.mask.pmovus.qb.mem.128(i8* %ptr, <2 x i64> %x1, i8 -1)
    call void @llvm.x86.avx512.mask.pmovus.qb.mem.128(i8* %ptr, <2 x i64> %x1, i8 %x2)
    ret void
}

declare <16 x i8> @llvm.x86.avx512.mask.pmov.qb.256(<4 x i64>, <16 x i8>, i8)

define <16 x i8>@test_int_x86_avx512_mask_pmov_qb_256(<4 x i64> %x0, <16 x i8> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmov_qb_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:    vpmovqb %ymm0, %xmm2 # encoding: [0x62,0xf2,0x7e,0x28,0x32,0xc2]
; X86-NEXT:    vpmovqb %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x32,0xc1]
; X86-NEXT:    vpmovqb %ymm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0xa9,0x32,0xc0]
; X86-NEXT:    vpaddb %xmm0, %xmm1, %xmm0 # encoding: [0xc5,0xf1,0xfc,0xc0]
; X86-NEXT:    vpaddb %xmm0, %xmm2, %xmm0 # encoding: [0xc5,0xe9,0xfc,0xc0]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmov_qb_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmovqb %ymm0, %xmm2 # encoding: [0x62,0xf2,0x7e,0x28,0x32,0xc2]
; X64-NEXT:    vpmovqb %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x32,0xc1]
; X64-NEXT:    vpmovqb %ymm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0xa9,0x32,0xc0]
; X64-NEXT:    vpaddb %xmm0, %xmm1, %xmm0 # encoding: [0xc5,0xf1,0xfc,0xc0]
; X64-NEXT:    vpaddb %xmm0, %xmm2, %xmm0 # encoding: [0xc5,0xe9,0xfc,0xc0]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
    %res0 = call <16 x i8> @llvm.x86.avx512.mask.pmov.qb.256(<4 x i64> %x0, <16 x i8> %x1, i8 -1)
    %res1 = call <16 x i8> @llvm.x86.avx512.mask.pmov.qb.256(<4 x i64> %x0, <16 x i8> %x1, i8 %x2)
    %res2 = call <16 x i8> @llvm.x86.avx512.mask.pmov.qb.256(<4 x i64> %x0, <16 x i8> zeroinitializer, i8 %x2)
    %res3 = add <16 x i8> %res0, %res1
    %res4 = add <16 x i8> %res3, %res2
    ret <16 x i8> %res4
}

declare void @llvm.x86.avx512.mask.pmov.qb.mem.256(i8* %ptr, <4 x i64>, i8)

define void @test_int_x86_avx512_mask_pmov_qb_mem_256(i8* %ptr, <4 x i64> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmov_qb_mem_256:
; X86:       # %bb.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:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpmovqb %ymm0, (%eax) # encoding: [0x62,0xf2,0x7e,0x28,0x32,0x00]
; X86-NEXT:    vpmovqb %ymm0, (%eax) {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x32,0x00]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmov_qb_mem_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpmovqb %ymm0, (%rdi) # encoding: [0x62,0xf2,0x7e,0x28,0x32,0x07]
; X64-NEXT:    vpmovqb %ymm0, (%rdi) {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x32,0x07]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
    call void @llvm.x86.avx512.mask.pmov.qb.mem.256(i8* %ptr, <4 x i64> %x1, i8 -1)
    call void @llvm.x86.avx512.mask.pmov.qb.mem.256(i8* %ptr, <4 x i64> %x1, i8 %x2)
    ret void
}

declare <16 x i8> @llvm.x86.avx512.mask.pmovs.qb.256(<4 x i64>, <16 x i8>, i8)

define <16 x i8>@test_int_x86_avx512_mask_pmovs_qb_256(<4 x i64> %x0, <16 x i8> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmovs_qb_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:    vpmovsqb %ymm0, %xmm2 # encoding: [0x62,0xf2,0x7e,0x28,0x22,0xc2]
; X86-NEXT:    vpmovsqb %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x22,0xc1]
; X86-NEXT:    vpmovsqb %ymm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0xa9,0x22,0xc0]
; X86-NEXT:    vpaddb %xmm0, %xmm1, %xmm0 # encoding: [0xc5,0xf1,0xfc,0xc0]
; X86-NEXT:    vpaddb %xmm0, %xmm2, %xmm0 # encoding: [0xc5,0xe9,0xfc,0xc0]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmovs_qb_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmovsqb %ymm0, %xmm2 # encoding: [0x62,0xf2,0x7e,0x28,0x22,0xc2]
; X64-NEXT:    vpmovsqb %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x22,0xc1]
; X64-NEXT:    vpmovsqb %ymm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0xa9,0x22,0xc0]
; X64-NEXT:    vpaddb %xmm0, %xmm1, %xmm0 # encoding: [0xc5,0xf1,0xfc,0xc0]
; X64-NEXT:    vpaddb %xmm0, %xmm2, %xmm0 # encoding: [0xc5,0xe9,0xfc,0xc0]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
    %res0 = call <16 x i8> @llvm.x86.avx512.mask.pmovs.qb.256(<4 x i64> %x0, <16 x i8> %x1, i8 -1)
    %res1 = call <16 x i8> @llvm.x86.avx512.mask.pmovs.qb.256(<4 x i64> %x0, <16 x i8> %x1, i8 %x2)
    %res2 = call <16 x i8> @llvm.x86.avx512.mask.pmovs.qb.256(<4 x i64> %x0, <16 x i8> zeroinitializer, i8 %x2)
    %res3 = add <16 x i8> %res0, %res1
    %res4 = add <16 x i8> %res3, %res2
    ret <16 x i8> %res4
}

declare void @llvm.x86.avx512.mask.pmovs.qb.mem.256(i8* %ptr, <4 x i64>, i8)

define void @test_int_x86_avx512_mask_pmovs_qb_mem_256(i8* %ptr, <4 x i64> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmovs_qb_mem_256:
; X86:       # %bb.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:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpmovsqb %ymm0, (%eax) # encoding: [0x62,0xf2,0x7e,0x28,0x22,0x00]
; X86-NEXT:    vpmovsqb %ymm0, (%eax) {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x22,0x00]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmovs_qb_mem_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpmovsqb %ymm0, (%rdi) # encoding: [0x62,0xf2,0x7e,0x28,0x22,0x07]
; X64-NEXT:    vpmovsqb %ymm0, (%rdi) {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x22,0x07]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
    call void @llvm.x86.avx512.mask.pmovs.qb.mem.256(i8* %ptr, <4 x i64> %x1, i8 -1)
    call void @llvm.x86.avx512.mask.pmovs.qb.mem.256(i8* %ptr, <4 x i64> %x1, i8 %x2)
    ret void
}

declare <16 x i8> @llvm.x86.avx512.mask.pmovus.qb.256(<4 x i64>, <16 x i8>, i8)

define <16 x i8>@test_int_x86_avx512_mask_pmovus_qb_256(<4 x i64> %x0, <16 x i8> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmovus_qb_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:    vpmovusqb %ymm0, %xmm2 # encoding: [0x62,0xf2,0x7e,0x28,0x12,0xc2]
; X86-NEXT:    vpmovusqb %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x12,0xc1]
; X86-NEXT:    vpmovusqb %ymm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0xa9,0x12,0xc0]
; X86-NEXT:    vpaddb %xmm0, %xmm1, %xmm0 # encoding: [0xc5,0xf1,0xfc,0xc0]
; X86-NEXT:    vpaddb %xmm0, %xmm2, %xmm0 # encoding: [0xc5,0xe9,0xfc,0xc0]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmovus_qb_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmovusqb %ymm0, %xmm2 # encoding: [0x62,0xf2,0x7e,0x28,0x12,0xc2]
; X64-NEXT:    vpmovusqb %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x12,0xc1]
; X64-NEXT:    vpmovusqb %ymm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0xa9,0x12,0xc0]
; X64-NEXT:    vpaddb %xmm0, %xmm1, %xmm0 # encoding: [0xc5,0xf1,0xfc,0xc0]
; X64-NEXT:    vpaddb %xmm0, %xmm2, %xmm0 # encoding: [0xc5,0xe9,0xfc,0xc0]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
    %res0 = call <16 x i8> @llvm.x86.avx512.mask.pmovus.qb.256(<4 x i64> %x0, <16 x i8> %x1, i8 -1)
    %res1 = call <16 x i8> @llvm.x86.avx512.mask.pmovus.qb.256(<4 x i64> %x0, <16 x i8> %x1, i8 %x2)
    %res2 = call <16 x i8> @llvm.x86.avx512.mask.pmovus.qb.256(<4 x i64> %x0, <16 x i8> zeroinitializer, i8 %x2)
    %res3 = add <16 x i8> %res0, %res1
    %res4 = add <16 x i8> %res3, %res2
    ret <16 x i8> %res4
}

declare void @llvm.x86.avx512.mask.pmovus.qb.mem.256(i8* %ptr, <4 x i64>, i8)

define void @test_int_x86_avx512_mask_pmovus_qb_mem_256(i8* %ptr, <4 x i64> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmovus_qb_mem_256:
; X86:       # %bb.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:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpmovusqb %ymm0, (%eax) # encoding: [0x62,0xf2,0x7e,0x28,0x12,0x00]
; X86-NEXT:    vpmovusqb %ymm0, (%eax) {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x12,0x00]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmovus_qb_mem_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpmovusqb %ymm0, (%rdi) # encoding: [0x62,0xf2,0x7e,0x28,0x12,0x07]
; X64-NEXT:    vpmovusqb %ymm0, (%rdi) {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x12,0x07]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
    call void @llvm.x86.avx512.mask.pmovus.qb.mem.256(i8* %ptr, <4 x i64> %x1, i8 -1)
    call void @llvm.x86.avx512.mask.pmovus.qb.mem.256(i8* %ptr, <4 x i64> %x1, i8 %x2)
    ret void
}

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

define <8 x i16>@test_int_x86_avx512_mask_pmov_qw_128(<2 x i64> %x0, <8 x i16> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmov_qw_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:    vpmovqw %xmm0, %xmm2 # encoding: [0x62,0xf2,0x7e,0x08,0x34,0xc2]
; X86-NEXT:    vpmovqw %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x34,0xc1]
; X86-NEXT:    vpmovqw %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0x89,0x34,0xc0]
; X86-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 # encoding: [0xc5,0xf1,0xfd,0xc0]
; X86-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 # encoding: [0xc5,0xe9,0xfd,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmov_qw_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmovqw %xmm0, %xmm2 # encoding: [0x62,0xf2,0x7e,0x08,0x34,0xc2]
; X64-NEXT:    vpmovqw %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x34,0xc1]
; X64-NEXT:    vpmovqw %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0x89,0x34,0xc0]
; X64-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 # encoding: [0xc5,0xf1,0xfd,0xc0]
; X64-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 # encoding: [0xc5,0xe9,0xfd,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
    %res0 = call <8 x i16> @llvm.x86.avx512.mask.pmov.qw.128(<2 x i64> %x0, <8 x i16> %x1, i8 -1)
    %res1 = call <8 x i16> @llvm.x86.avx512.mask.pmov.qw.128(<2 x i64> %x0, <8 x i16> %x1, i8 %x2)
    %res2 = call <8 x i16> @llvm.x86.avx512.mask.pmov.qw.128(<2 x i64> %x0, <8 x i16> zeroinitializer, i8 %x2)
    %res3 = add <8 x i16> %res0, %res1
    %res4 = add <8 x i16> %res3, %res2
    ret <8 x i16> %res4
}

declare void @llvm.x86.avx512.mask.pmov.qw.mem.128(i8* %ptr, <2 x i64>, i8)

define void @test_int_x86_avx512_mask_pmov_qw_mem_128(i8* %ptr, <2 x i64> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmov_qw_mem_128:
; X86:       # %bb.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:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpmovqw %xmm0, (%eax) # encoding: [0x62,0xf2,0x7e,0x08,0x34,0x00]
; X86-NEXT:    vpmovqw %xmm0, (%eax) {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x34,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmov_qw_mem_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpmovqw %xmm0, (%rdi) # encoding: [0x62,0xf2,0x7e,0x08,0x34,0x07]
; X64-NEXT:    vpmovqw %xmm0, (%rdi) {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x34,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
    call void @llvm.x86.avx512.mask.pmov.qw.mem.128(i8* %ptr, <2 x i64> %x1, i8 -1)
    call void @llvm.x86.avx512.mask.pmov.qw.mem.128(i8* %ptr, <2 x i64> %x1, i8 %x2)
    ret void
}

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

define <8 x i16>@test_int_x86_avx512_mask_pmovs_qw_128(<2 x i64> %x0, <8 x i16> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmovs_qw_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:    vpmovsqw %xmm0, %xmm2 # encoding: [0x62,0xf2,0x7e,0x08,0x24,0xc2]
; X86-NEXT:    vpmovsqw %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x24,0xc1]
; X86-NEXT:    vpmovsqw %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0x89,0x24,0xc0]
; X86-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 # encoding: [0xc5,0xf1,0xfd,0xc0]
; X86-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 # encoding: [0xc5,0xe9,0xfd,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmovs_qw_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmovsqw %xmm0, %xmm2 # encoding: [0x62,0xf2,0x7e,0x08,0x24,0xc2]
; X64-NEXT:    vpmovsqw %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x24,0xc1]
; X64-NEXT:    vpmovsqw %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0x89,0x24,0xc0]
; X64-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 # encoding: [0xc5,0xf1,0xfd,0xc0]
; X64-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 # encoding: [0xc5,0xe9,0xfd,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
    %res0 = call <8 x i16> @llvm.x86.avx512.mask.pmovs.qw.128(<2 x i64> %x0, <8 x i16> %x1, i8 -1)
    %res1 = call <8 x i16> @llvm.x86.avx512.mask.pmovs.qw.128(<2 x i64> %x0, <8 x i16> %x1, i8 %x2)
    %res2 = call <8 x i16> @llvm.x86.avx512.mask.pmovs.qw.128(<2 x i64> %x0, <8 x i16> zeroinitializer, i8 %x2)
    %res3 = add <8 x i16> %res0, %res1
    %res4 = add <8 x i16> %res3, %res2
    ret <8 x i16> %res4
}

declare void @llvm.x86.avx512.mask.pmovs.qw.mem.128(i8* %ptr, <2 x i64>, i8)

define void @test_int_x86_avx512_mask_pmovs_qw_mem_128(i8* %ptr, <2 x i64> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmovs_qw_mem_128:
; X86:       # %bb.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:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpmovsqw %xmm0, (%eax) # encoding: [0x62,0xf2,0x7e,0x08,0x24,0x00]
; X86-NEXT:    vpmovsqw %xmm0, (%eax) {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x24,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmovs_qw_mem_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpmovsqw %xmm0, (%rdi) # encoding: [0x62,0xf2,0x7e,0x08,0x24,0x07]
; X64-NEXT:    vpmovsqw %xmm0, (%rdi) {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x24,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
    call void @llvm.x86.avx512.mask.pmovs.qw.mem.128(i8* %ptr, <2 x i64> %x1, i8 -1)
    call void @llvm.x86.avx512.mask.pmovs.qw.mem.128(i8* %ptr, <2 x i64> %x1, i8 %x2)
    ret void
}

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

define <8 x i16>@test_int_x86_avx512_mask_pmovus_qw_128(<2 x i64> %x0, <8 x i16> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmovus_qw_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:    vpmovusqw %xmm0, %xmm2 # encoding: [0x62,0xf2,0x7e,0x08,0x14,0xc2]
; X86-NEXT:    vpmovusqw %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x14,0xc1]
; X86-NEXT:    vpmovusqw %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0x89,0x14,0xc0]
; X86-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 # encoding: [0xc5,0xf1,0xfd,0xc0]
; X86-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 # encoding: [0xc5,0xe9,0xfd,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmovus_qw_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmovusqw %xmm0, %xmm2 # encoding: [0x62,0xf2,0x7e,0x08,0x14,0xc2]
; X64-NEXT:    vpmovusqw %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x14,0xc1]
; X64-NEXT:    vpmovusqw %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0x89,0x14,0xc0]
; X64-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 # encoding: [0xc5,0xf1,0xfd,0xc0]
; X64-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 # encoding: [0xc5,0xe9,0xfd,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
    %res0 = call <8 x i16> @llvm.x86.avx512.mask.pmovus.qw.128(<2 x i64> %x0, <8 x i16> %x1, i8 -1)
    %res1 = call <8 x i16> @llvm.x86.avx512.mask.pmovus.qw.128(<2 x i64> %x0, <8 x i16> %x1, i8 %x2)
    %res2 = call <8 x i16> @llvm.x86.avx512.mask.pmovus.qw.128(<2 x i64> %x0, <8 x i16> zeroinitializer, i8 %x2)
    %res3 = add <8 x i16> %res0, %res1
    %res4 = add <8 x i16> %res3, %res2
    ret <8 x i16> %res4
}

declare void @llvm.x86.avx512.mask.pmovus.qw.mem.128(i8* %ptr, <2 x i64>, i8)

define void @test_int_x86_avx512_mask_pmovus_qw_mem_128(i8* %ptr, <2 x i64> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmovus_qw_mem_128:
; X86:       # %bb.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:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpmovusqw %xmm0, (%eax) # encoding: [0x62,0xf2,0x7e,0x08,0x14,0x00]
; X86-NEXT:    vpmovusqw %xmm0, (%eax) {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x14,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmovus_qw_mem_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpmovusqw %xmm0, (%rdi) # encoding: [0x62,0xf2,0x7e,0x08,0x14,0x07]
; X64-NEXT:    vpmovusqw %xmm0, (%rdi) {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x14,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
    call void @llvm.x86.avx512.mask.pmovus.qw.mem.128(i8* %ptr, <2 x i64> %x1, i8 -1)
    call void @llvm.x86.avx512.mask.pmovus.qw.mem.128(i8* %ptr, <2 x i64> %x1, i8 %x2)
    ret void
}

declare <8 x i16> @llvm.x86.avx512.mask.pmov.qw.256(<4 x i64>, <8 x i16>, i8)

define <8 x i16>@test_int_x86_avx512_mask_pmov_qw_256(<4 x i64> %x0, <8 x i16> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmov_qw_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:    vpmovqw %ymm0, %xmm2 # encoding: [0x62,0xf2,0x7e,0x28,0x34,0xc2]
; X86-NEXT:    vpmovqw %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x34,0xc1]
; X86-NEXT:    vpmovqw %ymm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0xa9,0x34,0xc0]
; X86-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 # encoding: [0xc5,0xf1,0xfd,0xc0]
; X86-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 # encoding: [0xc5,0xe9,0xfd,0xc0]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmov_qw_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmovqw %ymm0, %xmm2 # encoding: [0x62,0xf2,0x7e,0x28,0x34,0xc2]
; X64-NEXT:    vpmovqw %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x34,0xc1]
; X64-NEXT:    vpmovqw %ymm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0xa9,0x34,0xc0]
; X64-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 # encoding: [0xc5,0xf1,0xfd,0xc0]
; X64-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 # encoding: [0xc5,0xe9,0xfd,0xc0]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
    %res0 = call <8 x i16> @llvm.x86.avx512.mask.pmov.qw.256(<4 x i64> %x0, <8 x i16> %x1, i8 -1)
    %res1 = call <8 x i16> @llvm.x86.avx512.mask.pmov.qw.256(<4 x i64> %x0, <8 x i16> %x1, i8 %x2)
    %res2 = call <8 x i16> @llvm.x86.avx512.mask.pmov.qw.256(<4 x i64> %x0, <8 x i16> zeroinitializer, i8 %x2)
    %res3 = add <8 x i16> %res0, %res1
    %res4 = add <8 x i16> %res3, %res2
    ret <8 x i16> %res4
}

declare void @llvm.x86.avx512.mask.pmov.qw.mem.256(i8* %ptr, <4 x i64>, i8)

define void @test_int_x86_avx512_mask_pmov_qw_mem_256(i8* %ptr, <4 x i64> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmov_qw_mem_256:
; X86:       # %bb.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:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpmovqw %ymm0, (%eax) # encoding: [0x62,0xf2,0x7e,0x28,0x34,0x00]
; X86-NEXT:    vpmovqw %ymm0, (%eax) {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x34,0x00]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmov_qw_mem_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpmovqw %ymm0, (%rdi) # encoding: [0x62,0xf2,0x7e,0x28,0x34,0x07]
; X64-NEXT:    vpmovqw %ymm0, (%rdi) {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x34,0x07]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
    call void @llvm.x86.avx512.mask.pmov.qw.mem.256(i8* %ptr, <4 x i64> %x1, i8 -1)
    call void @llvm.x86.avx512.mask.pmov.qw.mem.256(i8* %ptr, <4 x i64> %x1, i8 %x2)
    ret void
}

declare <8 x i16> @llvm.x86.avx512.mask.pmovs.qw.256(<4 x i64>, <8 x i16>, i8)

define <8 x i16>@test_int_x86_avx512_mask_pmovs_qw_256(<4 x i64> %x0, <8 x i16> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmovs_qw_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:    vpmovsqw %ymm0, %xmm2 # encoding: [0x62,0xf2,0x7e,0x28,0x24,0xc2]
; X86-NEXT:    vpmovsqw %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x24,0xc1]
; X86-NEXT:    vpmovsqw %ymm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0xa9,0x24,0xc0]
; X86-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 # encoding: [0xc5,0xf1,0xfd,0xc0]
; X86-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 # encoding: [0xc5,0xe9,0xfd,0xc0]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmovs_qw_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmovsqw %ymm0, %xmm2 # encoding: [0x62,0xf2,0x7e,0x28,0x24,0xc2]
; X64-NEXT:    vpmovsqw %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x24,0xc1]
; X64-NEXT:    vpmovsqw %ymm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0xa9,0x24,0xc0]
; X64-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 # encoding: [0xc5,0xf1,0xfd,0xc0]
; X64-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 # encoding: [0xc5,0xe9,0xfd,0xc0]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
    %res0 = call <8 x i16> @llvm.x86.avx512.mask.pmovs.qw.256(<4 x i64> %x0, <8 x i16> %x1, i8 -1)
    %res1 = call <8 x i16> @llvm.x86.avx512.mask.pmovs.qw.256(<4 x i64> %x0, <8 x i16> %x1, i8 %x2)
    %res2 = call <8 x i16> @llvm.x86.avx512.mask.pmovs.qw.256(<4 x i64> %x0, <8 x i16> zeroinitializer, i8 %x2)
    %res3 = add <8 x i16> %res0, %res1
    %res4 = add <8 x i16> %res3, %res2
    ret <8 x i16> %res4
}

declare void @llvm.x86.avx512.mask.pmovs.qw.mem.256(i8* %ptr, <4 x i64>, i8)

define void @test_int_x86_avx512_mask_pmovs_qw_mem_256(i8* %ptr, <4 x i64> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmovs_qw_mem_256:
; X86:       # %bb.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:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpmovsqw %ymm0, (%eax) # encoding: [0x62,0xf2,0x7e,0x28,0x24,0x00]
; X86-NEXT:    vpmovsqw %ymm0, (%eax) {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x24,0x00]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmovs_qw_mem_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpmovsqw %ymm0, (%rdi) # encoding: [0x62,0xf2,0x7e,0x28,0x24,0x07]
; X64-NEXT:    vpmovsqw %ymm0, (%rdi) {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x24,0x07]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
    call void @llvm.x86.avx512.mask.pmovs.qw.mem.256(i8* %ptr, <4 x i64> %x1, i8 -1)
    call void @llvm.x86.avx512.mask.pmovs.qw.mem.256(i8* %ptr, <4 x i64> %x1, i8 %x2)
    ret void
}

declare <8 x i16> @llvm.x86.avx512.mask.pmovus.qw.256(<4 x i64>, <8 x i16>, i8)

define <8 x i16>@test_int_x86_avx512_mask_pmovus_qw_256(<4 x i64> %x0, <8 x i16> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmovus_qw_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:    vpmovusqw %ymm0, %xmm2 # encoding: [0x62,0xf2,0x7e,0x28,0x14,0xc2]
; X86-NEXT:    vpmovusqw %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x14,0xc1]
; X86-NEXT:    vpmovusqw %ymm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0xa9,0x14,0xc0]
; X86-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 # encoding: [0xc5,0xf1,0xfd,0xc0]
; X86-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 # encoding: [0xc5,0xe9,0xfd,0xc0]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmovus_qw_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmovusqw %ymm0, %xmm2 # encoding: [0x62,0xf2,0x7e,0x28,0x14,0xc2]
; X64-NEXT:    vpmovusqw %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x14,0xc1]
; X64-NEXT:    vpmovusqw %ymm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0xa9,0x14,0xc0]
; X64-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 # encoding: [0xc5,0xf1,0xfd,0xc0]
; X64-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 # encoding: [0xc5,0xe9,0xfd,0xc0]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
    %res0 = call <8 x i16> @llvm.x86.avx512.mask.pmovus.qw.256(<4 x i64> %x0, <8 x i16> %x1, i8 -1)
    %res1 = call <8 x i16> @llvm.x86.avx512.mask.pmovus.qw.256(<4 x i64> %x0, <8 x i16> %x1, i8 %x2)
    %res2 = call <8 x i16> @llvm.x86.avx512.mask.pmovus.qw.256(<4 x i64> %x0, <8 x i16> zeroinitializer, i8 %x2)
    %res3 = add <8 x i16> %res0, %res1
    %res4 = add <8 x i16> %res3, %res2
    ret <8 x i16> %res4
}

declare void @llvm.x86.avx512.mask.pmovus.qw.mem.256(i8* %ptr, <4 x i64>, i8)

define void @test_int_x86_avx512_mask_pmovus_qw_mem_256(i8* %ptr, <4 x i64> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmovus_qw_mem_256:
; X86:       # %bb.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:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpmovusqw %ymm0, (%eax) # encoding: [0x62,0xf2,0x7e,0x28,0x14,0x00]
; X86-NEXT:    vpmovusqw %ymm0, (%eax) {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x14,0x00]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmovus_qw_mem_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpmovusqw %ymm0, (%rdi) # encoding: [0x62,0xf2,0x7e,0x28,0x14,0x07]
; X64-NEXT:    vpmovusqw %ymm0, (%rdi) {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x14,0x07]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
    call void @llvm.x86.avx512.mask.pmovus.qw.mem.256(i8* %ptr, <4 x i64> %x1, i8 -1)
    call void @llvm.x86.avx512.mask.pmovus.qw.mem.256(i8* %ptr, <4 x i64> %x1, i8 %x2)
    ret void
}

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

define <4 x i32>@test_int_x86_avx512_mask_pmov_qd_128(<2 x i64> %x0, <4 x i32> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmov_qd_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:    vpmovqd %xmm0, %xmm2 # encoding: [0x62,0xf2,0x7e,0x08,0x35,0xc2]
; X86-NEXT:    vpmovqd %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x35,0xc1]
; X86-NEXT:    vpmovqd %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0x89,0x35,0xc0]
; 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_mask_pmov_qd_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmovqd %xmm0, %xmm2 # encoding: [0x62,0xf2,0x7e,0x08,0x35,0xc2]
; X64-NEXT:    vpmovqd %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x35,0xc1]
; X64-NEXT:    vpmovqd %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0x89,0x35,0xc0]
; 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]
    %res0 = call <4 x i32> @llvm.x86.avx512.mask.pmov.qd.128(<2 x i64> %x0, <4 x i32> %x1, i8 -1)
    %res1 = call <4 x i32> @llvm.x86.avx512.mask.pmov.qd.128(<2 x i64> %x0, <4 x i32> %x1, i8 %x2)
    %res2 = call <4 x i32> @llvm.x86.avx512.mask.pmov.qd.128(<2 x i64> %x0, <4 x i32> zeroinitializer, i8 %x2)
    %res3 = add <4 x i32> %res0, %res1
    %res4 = add <4 x i32> %res3, %res2
    ret <4 x i32> %res4
}

declare void @llvm.x86.avx512.mask.pmov.qd.mem.128(i8* %ptr, <2 x i64>, i8)

define void @test_int_x86_avx512_mask_pmov_qd_mem_128(i8* %ptr, <2 x i64> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmov_qd_mem_128:
; X86:       # %bb.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:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpmovqd %xmm0, (%eax) # encoding: [0x62,0xf2,0x7e,0x08,0x35,0x00]
; X86-NEXT:    vpmovqd %xmm0, (%eax) {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x35,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmov_qd_mem_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpmovqd %xmm0, (%rdi) # encoding: [0x62,0xf2,0x7e,0x08,0x35,0x07]
; X64-NEXT:    vpmovqd %xmm0, (%rdi) {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x35,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
    call void @llvm.x86.avx512.mask.pmov.qd.mem.128(i8* %ptr, <2 x i64> %x1, i8 -1)
    call void @llvm.x86.avx512.mask.pmov.qd.mem.128(i8* %ptr, <2 x i64> %x1, i8 %x2)
    ret void
}

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

define <4 x i32>@test_int_x86_avx512_mask_pmovs_qd_128(<2 x i64> %x0, <4 x i32> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmovs_qd_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:    vpmovsqd %xmm0, %xmm0 # encoding: [0x62,0xf2,0x7e,0x08,0x25,0xc0]
; X86-NEXT:    vmovdqa32 %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x6f,0xc8]
; X86-NEXT:    vmovdqa32 %xmm0, %xmm2 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x6f,0xd0]
; X86-NEXT:    vpaddd %xmm2, %xmm1, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xca]
; X86-NEXT:    vpaddd %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfe,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmovs_qd_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmovsqd %xmm0, %xmm0 # encoding: [0x62,0xf2,0x7e,0x08,0x25,0xc0]
; X64-NEXT:    vmovdqa32 %xmm0, %xmm2 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x6f,0xd0]
; X64-NEXT:    vmovdqa32 %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x6f,0xc8]
; X64-NEXT:    vpaddd %xmm2, %xmm1, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xca]
; X64-NEXT:    vpaddd %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfe,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
    %res0 = call <4 x i32> @llvm.x86.avx512.mask.pmovs.qd.128(<2 x i64> %x0, <4 x i32> %x1, i8 -1)
    %res1 = call <4 x i32> @llvm.x86.avx512.mask.pmovs.qd.128(<2 x i64> %x0, <4 x i32> %x1, i8 %x2)
    %res2 = call <4 x i32> @llvm.x86.avx512.mask.pmovs.qd.128(<2 x i64> %x0, <4 x i32> zeroinitializer, i8 %x2)
    %res3 = add <4 x i32> %res0, %res1
    %res4 = add <4 x i32> %res3, %res2
    ret <4 x i32> %res4
}

declare void @llvm.x86.avx512.mask.pmovs.qd.mem.128(i8* %ptr, <2 x i64>, i8)

define void @test_int_x86_avx512_mask_pmovs_qd_mem_128(i8* %ptr, <2 x i64> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmovs_qd_mem_128:
; X86:       # %bb.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:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpmovsqd %xmm0, (%eax) # encoding: [0x62,0xf2,0x7e,0x08,0x25,0x00]
; X86-NEXT:    vpmovsqd %xmm0, (%eax) {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x25,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmovs_qd_mem_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpmovsqd %xmm0, (%rdi) # encoding: [0x62,0xf2,0x7e,0x08,0x25,0x07]
; X64-NEXT:    vpmovsqd %xmm0, (%rdi) {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x25,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
    call void @llvm.x86.avx512.mask.pmovs.qd.mem.128(i8* %ptr, <2 x i64> %x1, i8 -1)
    call void @llvm.x86.avx512.mask.pmovs.qd.mem.128(i8* %ptr, <2 x i64> %x1, i8 %x2)
    ret void
}

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

define <4 x i32>@test_int_x86_avx512_mask_pmovus_qd_128(<2 x i64> %x0, <4 x i32> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmovus_qd_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:    vpmovusqd %xmm0, %xmm2 # encoding: [0x62,0xf2,0x7e,0x08,0x15,0xc2]
; X86-NEXT:    vpmovusqd %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x15,0xc1]
; X86-NEXT:    vpmovusqd %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0x89,0x15,0xc0]
; 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_mask_pmovus_qd_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmovusqd %xmm0, %xmm2 # encoding: [0x62,0xf2,0x7e,0x08,0x15,0xc2]
; X64-NEXT:    vpmovusqd %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x15,0xc1]
; X64-NEXT:    vpmovusqd %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0x89,0x15,0xc0]
; 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]
    %res0 = call <4 x i32> @llvm.x86.avx512.mask.pmovus.qd.128(<2 x i64> %x0, <4 x i32> %x1, i8 -1)
    %res1 = call <4 x i32> @llvm.x86.avx512.mask.pmovus.qd.128(<2 x i64> %x0, <4 x i32> %x1, i8 %x2)
    %res2 = call <4 x i32> @llvm.x86.avx512.mask.pmovus.qd.128(<2 x i64> %x0, <4 x i32> zeroinitializer, i8 %x2)
    %res3 = add <4 x i32> %res0, %res1
    %res4 = add <4 x i32> %res3, %res2
    ret <4 x i32> %res4
}

declare void @llvm.x86.avx512.mask.pmovus.qd.mem.128(i8* %ptr, <2 x i64>, i8)

define void @test_int_x86_avx512_mask_pmovus_qd_mem_128(i8* %ptr, <2 x i64> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmovus_qd_mem_128:
; X86:       # %bb.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:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpmovusqd %xmm0, (%eax) # encoding: [0x62,0xf2,0x7e,0x08,0x15,0x00]
; X86-NEXT:    vpmovusqd %xmm0, (%eax) {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x15,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmovus_qd_mem_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpmovusqd %xmm0, (%rdi) # encoding: [0x62,0xf2,0x7e,0x08,0x15,0x07]
; X64-NEXT:    vpmovusqd %xmm0, (%rdi) {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x15,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
    call void @llvm.x86.avx512.mask.pmovus.qd.mem.128(i8* %ptr, <2 x i64> %x1, i8 -1)
    call void @llvm.x86.avx512.mask.pmovus.qd.mem.128(i8* %ptr, <2 x i64> %x1, i8 %x2)
    ret void
}

define <4 x i32>@test_int_x86_avx512_mask_pmov_qd_256(<4 x i64> %x0, <4 x i32> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmov_qd_256:
; X86:       # %bb.0:
; X86-NEXT:    vpmovqd %ymm0, %xmm2 # encoding: [0x62,0xf2,0x7e,0x28,0x35,0xc2]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
; X86-NEXT:    vpmovqd %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x35,0xc1]
; X86-NEXT:    vpmovqd %ymm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0xa9,0x35,0xc0]
; 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:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmov_qd_256:
; X64:       # %bb.0:
; X64-NEXT:    vpmovqd %ymm0, %xmm2 # encoding: [0x62,0xf2,0x7e,0x28,0x35,0xc2]
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmovqd %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x35,0xc1]
; X64-NEXT:    vpmovqd %ymm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0xa9,0x35,0xc0]
; 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:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = trunc <4 x i64> %x0 to <4 x i32>
  %2 = trunc <4 x i64> %x0 to <4 x i32>
  %3 = bitcast i8 %x2 to <8 x i1>
  %extract1 = shufflevector <8 x i1> %3, <8 x i1> %3, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %4 = select <4 x i1> %extract1, <4 x i32> %2, <4 x i32> %x1
  %5 = trunc <4 x i64> %x0 to <4 x i32>
  %6 = bitcast i8 %x2 to <8 x i1>
  %extract = shufflevector <8 x i1> %6, <8 x i1> %6, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %7 = select <4 x i1> %extract, <4 x i32> %5, <4 x i32> zeroinitializer
  %res3 = add <4 x i32> %1, %4
  %res4 = add <4 x i32> %res3, %7
  ret <4 x i32> %res4
}

declare void @llvm.x86.avx512.mask.pmov.qd.mem.256(i8* %ptr, <4 x i64>, i8)

define void @test_int_x86_avx512_mask_pmov_qd_mem_256(i8* %ptr, <4 x i64> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmov_qd_mem_256:
; X86:       # %bb.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:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpmovqd %ymm0, (%eax) # encoding: [0x62,0xf2,0x7e,0x28,0x35,0x00]
; X86-NEXT:    vpmovqd %ymm0, (%eax) {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x35,0x00]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmov_qd_mem_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpmovqd %ymm0, (%rdi) # encoding: [0x62,0xf2,0x7e,0x28,0x35,0x07]
; X64-NEXT:    vpmovqd %ymm0, (%rdi) {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x35,0x07]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
    call void @llvm.x86.avx512.mask.pmov.qd.mem.256(i8* %ptr, <4 x i64> %x1, i8 -1)
    call void @llvm.x86.avx512.mask.pmov.qd.mem.256(i8* %ptr, <4 x i64> %x1, i8 %x2)
    ret void
}

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

define <4 x i32>@test_int_x86_avx512_mask_pmovs_qd_256(<4 x i64> %x0, <4 x i32> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmovs_qd_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:    vpmovsqd %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x25,0xc1]
; X86-NEXT:    vpmovsqd %ymm0, %xmm2 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0xa9,0x25,0xc2]
; X86-NEXT:    vpaddd %xmm2, %xmm1, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xca]
; X86-NEXT:    vpmovsqd %ymm0, %xmm0 # encoding: [0x62,0xf2,0x7e,0x28,0x25,0xc0]
; X86-NEXT:    vpaddd %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfe,0xc1]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmovs_qd_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmovsqd %ymm0, %xmm2 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0xa9,0x25,0xc2]
; X64-NEXT:    vpmovsqd %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x25,0xc1]
; X64-NEXT:    vpaddd %xmm2, %xmm1, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xca]
; X64-NEXT:    vpmovsqd %ymm0, %xmm0 # encoding: [0x62,0xf2,0x7e,0x28,0x25,0xc0]
; X64-NEXT:    vpaddd %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfe,0xc1]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
    %res0 = call <4 x i32> @llvm.x86.avx512.mask.pmovs.qd.256(<4 x i64> %x0, <4 x i32> %x1, i8 -1)
    %res1 = call <4 x i32> @llvm.x86.avx512.mask.pmovs.qd.256(<4 x i64> %x0, <4 x i32> %x1, i8 %x2)
    %res2 = call <4 x i32> @llvm.x86.avx512.mask.pmovs.qd.256(<4 x i64> %x0, <4 x i32> zeroinitializer, i8 %x2)
    %res3 = add <4 x i32> %res0, %res1
    %res4 = add <4 x i32> %res3, %res2
    ret <4 x i32> %res4
}

declare void @llvm.x86.avx512.mask.pmovs.qd.mem.256(i8* %ptr, <4 x i64>, i8)

define void @test_int_x86_avx512_mask_pmovs_qd_mem_256(i8* %ptr, <4 x i64> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmovs_qd_mem_256:
; X86:       # %bb.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:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpmovsqd %ymm0, (%eax) # encoding: [0x62,0xf2,0x7e,0x28,0x25,0x00]
; X86-NEXT:    vpmovsqd %ymm0, (%eax) {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x25,0x00]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmovs_qd_mem_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpmovsqd %ymm0, (%rdi) # encoding: [0x62,0xf2,0x7e,0x28,0x25,0x07]
; X64-NEXT:    vpmovsqd %ymm0, (%rdi) {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x25,0x07]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
    call void @llvm.x86.avx512.mask.pmovs.qd.mem.256(i8* %ptr, <4 x i64> %x1, i8 -1)
    call void @llvm.x86.avx512.mask.pmovs.qd.mem.256(i8* %ptr, <4 x i64> %x1, i8 %x2)
    ret void
}

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

define <4 x i32>@test_int_x86_avx512_mask_pmovus_qd_256(<4 x i64> %x0, <4 x i32> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmovus_qd_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:    vpmovusqd %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x15,0xc1]
; X86-NEXT:    vpmovusqd %ymm0, %xmm2 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0xa9,0x15,0xc2]
; X86-NEXT:    vpaddd %xmm2, %xmm1, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xca]
; X86-NEXT:    vpmovusqd %ymm0, %xmm0 # encoding: [0x62,0xf2,0x7e,0x28,0x15,0xc0]
; X86-NEXT:    vpaddd %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfe,0xc1]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmovus_qd_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmovusqd %ymm0, %xmm2 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0xa9,0x15,0xc2]
; X64-NEXT:    vpmovusqd %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x15,0xc1]
; X64-NEXT:    vpaddd %xmm2, %xmm1, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xca]
; X64-NEXT:    vpmovusqd %ymm0, %xmm0 # encoding: [0x62,0xf2,0x7e,0x28,0x15,0xc0]
; X64-NEXT:    vpaddd %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfe,0xc1]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
    %res0 = call <4 x i32> @llvm.x86.avx512.mask.pmovus.qd.256(<4 x i64> %x0, <4 x i32> %x1, i8 -1)
    %res1 = call <4 x i32> @llvm.x86.avx512.mask.pmovus.qd.256(<4 x i64> %x0, <4 x i32> %x1, i8 %x2)
    %res2 = call <4 x i32> @llvm.x86.avx512.mask.pmovus.qd.256(<4 x i64> %x0, <4 x i32> zeroinitializer, i8 %x2)
    %res3 = add <4 x i32> %res0, %res1
    %res4 = add <4 x i32> %res3, %res2
    ret <4 x i32> %res4
}

declare void @llvm.x86.avx512.mask.pmovus.qd.mem.256(i8* %ptr, <4 x i64>, i8)

define void @test_int_x86_avx512_mask_pmovus_qd_mem_256(i8* %ptr, <4 x i64> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmovus_qd_mem_256:
; X86:       # %bb.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:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpmovusqd %ymm0, (%eax) # encoding: [0x62,0xf2,0x7e,0x28,0x15,0x00]
; X86-NEXT:    vpmovusqd %ymm0, (%eax) {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x15,0x00]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmovus_qd_mem_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpmovusqd %ymm0, (%rdi) # encoding: [0x62,0xf2,0x7e,0x28,0x15,0x07]
; X64-NEXT:    vpmovusqd %ymm0, (%rdi) {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x15,0x07]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
    call void @llvm.x86.avx512.mask.pmovus.qd.mem.256(i8* %ptr, <4 x i64> %x1, i8 -1)
    call void @llvm.x86.avx512.mask.pmovus.qd.mem.256(i8* %ptr, <4 x i64> %x1, i8 %x2)
    ret void
}

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

define <16 x i8>@test_int_x86_avx512_mask_pmov_db_128(<4 x i32> %x0, <16 x i8> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmov_db_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:    vpmovdb %xmm0, %xmm2 # encoding: [0x62,0xf2,0x7e,0x08,0x31,0xc2]
; X86-NEXT:    vpmovdb %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x31,0xc1]
; X86-NEXT:    vpmovdb %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0x89,0x31,0xc0]
; X86-NEXT:    vpaddb %xmm0, %xmm1, %xmm0 # encoding: [0xc5,0xf1,0xfc,0xc0]
; X86-NEXT:    vpaddb %xmm0, %xmm2, %xmm0 # encoding: [0xc5,0xe9,0xfc,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmov_db_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmovdb %xmm0, %xmm2 # encoding: [0x62,0xf2,0x7e,0x08,0x31,0xc2]
; X64-NEXT:    vpmovdb %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x31,0xc1]
; X64-NEXT:    vpmovdb %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0x89,0x31,0xc0]
; X64-NEXT:    vpaddb %xmm0, %xmm1, %xmm0 # encoding: [0xc5,0xf1,0xfc,0xc0]
; X64-NEXT:    vpaddb %xmm0, %xmm2, %xmm0 # encoding: [0xc5,0xe9,0xfc,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
    %res0 = call <16 x i8> @llvm.x86.avx512.mask.pmov.db.128(<4 x i32> %x0, <16 x i8> %x1, i8 -1)
    %res1 = call <16 x i8> @llvm.x86.avx512.mask.pmov.db.128(<4 x i32> %x0, <16 x i8> %x1, i8 %x2)
    %res2 = call <16 x i8> @llvm.x86.avx512.mask.pmov.db.128(<4 x i32> %x0, <16 x i8> zeroinitializer, i8 %x2)
    %res3 = add <16 x i8> %res0, %res1
    %res4 = add <16 x i8> %res3, %res2
    ret <16 x i8> %res4
}

declare void @llvm.x86.avx512.mask.pmov.db.mem.128(i8* %ptr, <4 x i32>, i8)

define void @test_int_x86_avx512_mask_pmov_db_mem_128(i8* %ptr, <4 x i32> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmov_db_mem_128:
; X86:       # %bb.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:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpmovdb %xmm0, (%eax) # encoding: [0x62,0xf2,0x7e,0x08,0x31,0x00]
; X86-NEXT:    vpmovdb %xmm0, (%eax) {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x31,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmov_db_mem_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpmovdb %xmm0, (%rdi) # encoding: [0x62,0xf2,0x7e,0x08,0x31,0x07]
; X64-NEXT:    vpmovdb %xmm0, (%rdi) {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x31,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
    call void @llvm.x86.avx512.mask.pmov.db.mem.128(i8* %ptr, <4 x i32> %x1, i8 -1)
    call void @llvm.x86.avx512.mask.pmov.db.mem.128(i8* %ptr, <4 x i32> %x1, i8 %x2)
    ret void
}

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

define <16 x i8>@test_int_x86_avx512_mask_pmovs_db_128(<4 x i32> %x0, <16 x i8> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmovs_db_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:    vpmovsdb %xmm0, %xmm2 # encoding: [0x62,0xf2,0x7e,0x08,0x21,0xc2]
; X86-NEXT:    vpmovsdb %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x21,0xc1]
; X86-NEXT:    vpmovsdb %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0x89,0x21,0xc0]
; X86-NEXT:    vpaddb %xmm0, %xmm1, %xmm0 # encoding: [0xc5,0xf1,0xfc,0xc0]
; X86-NEXT:    vpaddb %xmm0, %xmm2, %xmm0 # encoding: [0xc5,0xe9,0xfc,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmovs_db_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmovsdb %xmm0, %xmm2 # encoding: [0x62,0xf2,0x7e,0x08,0x21,0xc2]
; X64-NEXT:    vpmovsdb %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x21,0xc1]
; X64-NEXT:    vpmovsdb %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0x89,0x21,0xc0]
; X64-NEXT:    vpaddb %xmm0, %xmm1, %xmm0 # encoding: [0xc5,0xf1,0xfc,0xc0]
; X64-NEXT:    vpaddb %xmm0, %xmm2, %xmm0 # encoding: [0xc5,0xe9,0xfc,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
    %res0 = call <16 x i8> @llvm.x86.avx512.mask.pmovs.db.128(<4 x i32> %x0, <16 x i8> %x1, i8 -1)
    %res1 = call <16 x i8> @llvm.x86.avx512.mask.pmovs.db.128(<4 x i32> %x0, <16 x i8> %x1, i8 %x2)
    %res2 = call <16 x i8> @llvm.x86.avx512.mask.pmovs.db.128(<4 x i32> %x0, <16 x i8> zeroinitializer, i8 %x2)
    %res3 = add <16 x i8> %res0, %res1
    %res4 = add <16 x i8> %res3, %res2
    ret <16 x i8> %res4
}

declare void @llvm.x86.avx512.mask.pmovs.db.mem.128(i8* %ptr, <4 x i32>, i8)

define void @test_int_x86_avx512_mask_pmovs_db_mem_128(i8* %ptr, <4 x i32> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmovs_db_mem_128:
; X86:       # %bb.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:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpmovsdb %xmm0, (%eax) # encoding: [0x62,0xf2,0x7e,0x08,0x21,0x00]
; X86-NEXT:    vpmovsdb %xmm0, (%eax) {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x21,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmovs_db_mem_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpmovsdb %xmm0, (%rdi) # encoding: [0x62,0xf2,0x7e,0x08,0x21,0x07]
; X64-NEXT:    vpmovsdb %xmm0, (%rdi) {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x21,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
    call void @llvm.x86.avx512.mask.pmovs.db.mem.128(i8* %ptr, <4 x i32> %x1, i8 -1)
    call void @llvm.x86.avx512.mask.pmovs.db.mem.128(i8* %ptr, <4 x i32> %x1, i8 %x2)
    ret void
}

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

define <16 x i8>@test_int_x86_avx512_mask_pmovus_db_128(<4 x i32> %x0, <16 x i8> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmovus_db_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:    vpmovusdb %xmm0, %xmm2 # encoding: [0x62,0xf2,0x7e,0x08,0x11,0xc2]
; X86-NEXT:    vpmovusdb %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x11,0xc1]
; X86-NEXT:    vpmovusdb %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0x89,0x11,0xc0]
; X86-NEXT:    vpaddb %xmm0, %xmm1, %xmm0 # encoding: [0xc5,0xf1,0xfc,0xc0]
; X86-NEXT:    vpaddb %xmm0, %xmm2, %xmm0 # encoding: [0xc5,0xe9,0xfc,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmovus_db_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmovusdb %xmm0, %xmm2 # encoding: [0x62,0xf2,0x7e,0x08,0x11,0xc2]
; X64-NEXT:    vpmovusdb %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x11,0xc1]
; X64-NEXT:    vpmovusdb %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0x89,0x11,0xc0]
; X64-NEXT:    vpaddb %xmm0, %xmm1, %xmm0 # encoding: [0xc5,0xf1,0xfc,0xc0]
; X64-NEXT:    vpaddb %xmm0, %xmm2, %xmm0 # encoding: [0xc5,0xe9,0xfc,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
    %res0 = call <16 x i8> @llvm.x86.avx512.mask.pmovus.db.128(<4 x i32> %x0, <16 x i8> %x1, i8 -1)
    %res1 = call <16 x i8> @llvm.x86.avx512.mask.pmovus.db.128(<4 x i32> %x0, <16 x i8> %x1, i8 %x2)
    %res2 = call <16 x i8> @llvm.x86.avx512.mask.pmovus.db.128(<4 x i32> %x0, <16 x i8> zeroinitializer, i8 %x2)
    %res3 = add <16 x i8> %res0, %res1
    %res4 = add <16 x i8> %res3, %res2
    ret <16 x i8> %res4
}

declare void @llvm.x86.avx512.mask.pmovus.db.mem.128(i8* %ptr, <4 x i32>, i8)

define void @test_int_x86_avx512_mask_pmovus_db_mem_128(i8* %ptr, <4 x i32> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmovus_db_mem_128:
; X86:       # %bb.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:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpmovusdb %xmm0, (%eax) # encoding: [0x62,0xf2,0x7e,0x08,0x11,0x00]
; X86-NEXT:    vpmovusdb %xmm0, (%eax) {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x11,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmovus_db_mem_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpmovusdb %xmm0, (%rdi) # encoding: [0x62,0xf2,0x7e,0x08,0x11,0x07]
; X64-NEXT:    vpmovusdb %xmm0, (%rdi) {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x11,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
    call void @llvm.x86.avx512.mask.pmovus.db.mem.128(i8* %ptr, <4 x i32> %x1, i8 -1)
    call void @llvm.x86.avx512.mask.pmovus.db.mem.128(i8* %ptr, <4 x i32> %x1, i8 %x2)
    ret void
}

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

define <16 x i8>@test_int_x86_avx512_mask_pmov_db_256(<8 x i32> %x0, <16 x i8> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmov_db_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:    vpmovdb %ymm0, %xmm2 # encoding: [0x62,0xf2,0x7e,0x28,0x31,0xc2]
; X86-NEXT:    vpmovdb %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x31,0xc1]
; X86-NEXT:    vpmovdb %ymm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0xa9,0x31,0xc0]
; X86-NEXT:    vpaddb %xmm0, %xmm1, %xmm0 # encoding: [0xc5,0xf1,0xfc,0xc0]
; X86-NEXT:    vpaddb %xmm0, %xmm2, %xmm0 # encoding: [0xc5,0xe9,0xfc,0xc0]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmov_db_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmovdb %ymm0, %xmm2 # encoding: [0x62,0xf2,0x7e,0x28,0x31,0xc2]
; X64-NEXT:    vpmovdb %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x31,0xc1]
; X64-NEXT:    vpmovdb %ymm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0xa9,0x31,0xc0]
; X64-NEXT:    vpaddb %xmm0, %xmm1, %xmm0 # encoding: [0xc5,0xf1,0xfc,0xc0]
; X64-NEXT:    vpaddb %xmm0, %xmm2, %xmm0 # encoding: [0xc5,0xe9,0xfc,0xc0]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
    %res0 = call <16 x i8> @llvm.x86.avx512.mask.pmov.db.256(<8 x i32> %x0, <16 x i8> %x1, i8 -1)
    %res1 = call <16 x i8> @llvm.x86.avx512.mask.pmov.db.256(<8 x i32> %x0, <16 x i8> %x1, i8 %x2)
    %res2 = call <16 x i8> @llvm.x86.avx512.mask.pmov.db.256(<8 x i32> %x0, <16 x i8> zeroinitializer, i8 %x2)
    %res3 = add <16 x i8> %res0, %res1
    %res4 = add <16 x i8> %res3, %res2
    ret <16 x i8> %res4
}

declare void @llvm.x86.avx512.mask.pmov.db.mem.256(i8* %ptr, <8 x i32>, i8)

define void @test_int_x86_avx512_mask_pmov_db_mem_256(i8* %ptr, <8 x i32> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmov_db_mem_256:
; X86:       # %bb.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:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpmovdb %ymm0, (%eax) # encoding: [0x62,0xf2,0x7e,0x28,0x31,0x00]
; X86-NEXT:    vpmovdb %ymm0, (%eax) {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x31,0x00]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmov_db_mem_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpmovdb %ymm0, (%rdi) # encoding: [0x62,0xf2,0x7e,0x28,0x31,0x07]
; X64-NEXT:    vpmovdb %ymm0, (%rdi) {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x31,0x07]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
    call void @llvm.x86.avx512.mask.pmov.db.mem.256(i8* %ptr, <8 x i32> %x1, i8 -1)
    call void @llvm.x86.avx512.mask.pmov.db.mem.256(i8* %ptr, <8 x i32> %x1, i8 %x2)
    ret void
}

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

define <16 x i8>@test_int_x86_avx512_mask_pmovs_db_256(<8 x i32> %x0, <16 x i8> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmovs_db_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:    vpmovsdb %ymm0, %xmm2 # encoding: [0x62,0xf2,0x7e,0x28,0x21,0xc2]
; X86-NEXT:    vpmovsdb %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x21,0xc1]
; X86-NEXT:    vpmovsdb %ymm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0xa9,0x21,0xc0]
; X86-NEXT:    vpaddb %xmm0, %xmm1, %xmm0 # encoding: [0xc5,0xf1,0xfc,0xc0]
; X86-NEXT:    vpaddb %xmm0, %xmm2, %xmm0 # encoding: [0xc5,0xe9,0xfc,0xc0]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmovs_db_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmovsdb %ymm0, %xmm2 # encoding: [0x62,0xf2,0x7e,0x28,0x21,0xc2]
; X64-NEXT:    vpmovsdb %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x21,0xc1]
; X64-NEXT:    vpmovsdb %ymm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0xa9,0x21,0xc0]
; X64-NEXT:    vpaddb %xmm0, %xmm1, %xmm0 # encoding: [0xc5,0xf1,0xfc,0xc0]
; X64-NEXT:    vpaddb %xmm0, %xmm2, %xmm0 # encoding: [0xc5,0xe9,0xfc,0xc0]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
    %res0 = call <16 x i8> @llvm.x86.avx512.mask.pmovs.db.256(<8 x i32> %x0, <16 x i8> %x1, i8 -1)
    %res1 = call <16 x i8> @llvm.x86.avx512.mask.pmovs.db.256(<8 x i32> %x0, <16 x i8> %x1, i8 %x2)
    %res2 = call <16 x i8> @llvm.x86.avx512.mask.pmovs.db.256(<8 x i32> %x0, <16 x i8> zeroinitializer, i8 %x2)
    %res3 = add <16 x i8> %res0, %res1
    %res4 = add <16 x i8> %res3, %res2
    ret <16 x i8> %res4
}

declare void @llvm.x86.avx512.mask.pmovs.db.mem.256(i8* %ptr, <8 x i32>, i8)

define void @test_int_x86_avx512_mask_pmovs_db_mem_256(i8* %ptr, <8 x i32> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmovs_db_mem_256:
; X86:       # %bb.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:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpmovsdb %ymm0, (%eax) # encoding: [0x62,0xf2,0x7e,0x28,0x21,0x00]
; X86-NEXT:    vpmovsdb %ymm0, (%eax) {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x21,0x00]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmovs_db_mem_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpmovsdb %ymm0, (%rdi) # encoding: [0x62,0xf2,0x7e,0x28,0x21,0x07]
; X64-NEXT:    vpmovsdb %ymm0, (%rdi) {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x21,0x07]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
    call void @llvm.x86.avx512.mask.pmovs.db.mem.256(i8* %ptr, <8 x i32> %x1, i8 -1)
    call void @llvm.x86.avx512.mask.pmovs.db.mem.256(i8* %ptr, <8 x i32> %x1, i8 %x2)
    ret void
}

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

define <16 x i8>@test_int_x86_avx512_mask_pmovus_db_256(<8 x i32> %x0, <16 x i8> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmovus_db_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:    vpmovusdb %ymm0, %xmm2 # encoding: [0x62,0xf2,0x7e,0x28,0x11,0xc2]
; X86-NEXT:    vpmovusdb %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x11,0xc1]
; X86-NEXT:    vpmovusdb %ymm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0xa9,0x11,0xc0]
; X86-NEXT:    vpaddb %xmm0, %xmm1, %xmm0 # encoding: [0xc5,0xf1,0xfc,0xc0]
; X86-NEXT:    vpaddb %xmm0, %xmm2, %xmm0 # encoding: [0xc5,0xe9,0xfc,0xc0]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmovus_db_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmovusdb %ymm0, %xmm2 # encoding: [0x62,0xf2,0x7e,0x28,0x11,0xc2]
; X64-NEXT:    vpmovusdb %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x11,0xc1]
; X64-NEXT:    vpmovusdb %ymm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0xa9,0x11,0xc0]
; X64-NEXT:    vpaddb %xmm0, %xmm1, %xmm0 # encoding: [0xc5,0xf1,0xfc,0xc0]
; X64-NEXT:    vpaddb %xmm0, %xmm2, %xmm0 # encoding: [0xc5,0xe9,0xfc,0xc0]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
    %res0 = call <16 x i8> @llvm.x86.avx512.mask.pmovus.db.256(<8 x i32> %x0, <16 x i8> %x1, i8 -1)
    %res1 = call <16 x i8> @llvm.x86.avx512.mask.pmovus.db.256(<8 x i32> %x0, <16 x i8> %x1, i8 %x2)
    %res2 = call <16 x i8> @llvm.x86.avx512.mask.pmovus.db.256(<8 x i32> %x0, <16 x i8> zeroinitializer, i8 %x2)
    %res3 = add <16 x i8> %res0, %res1
    %res4 = add <16 x i8> %res3, %res2
    ret <16 x i8> %res4
}

declare void @llvm.x86.avx512.mask.pmovus.db.mem.256(i8* %ptr, <8 x i32>, i8)

define void @test_int_x86_avx512_mask_pmovus_db_mem_256(i8* %ptr, <8 x i32> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmovus_db_mem_256:
; X86:       # %bb.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:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpmovusdb %ymm0, (%eax) # encoding: [0x62,0xf2,0x7e,0x28,0x11,0x00]
; X86-NEXT:    vpmovusdb %ymm0, (%eax) {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x11,0x00]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmovus_db_mem_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpmovusdb %ymm0, (%rdi) # encoding: [0x62,0xf2,0x7e,0x28,0x11,0x07]
; X64-NEXT:    vpmovusdb %ymm0, (%rdi) {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x11,0x07]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
    call void @llvm.x86.avx512.mask.pmovus.db.mem.256(i8* %ptr, <8 x i32> %x1, i8 -1)
    call void @llvm.x86.avx512.mask.pmovus.db.mem.256(i8* %ptr, <8 x i32> %x1, i8 %x2)
    ret void
}

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

define <8 x i16>@test_int_x86_avx512_mask_pmov_dw_128(<4 x i32> %x0, <8 x i16> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmov_dw_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:    vpmovdw %xmm0, %xmm2 # encoding: [0x62,0xf2,0x7e,0x08,0x33,0xc2]
; X86-NEXT:    vpmovdw %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x33,0xc1]
; X86-NEXT:    vpmovdw %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0x89,0x33,0xc0]
; X86-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 # encoding: [0xc5,0xf1,0xfd,0xc0]
; X86-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 # encoding: [0xc5,0xe9,0xfd,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmov_dw_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmovdw %xmm0, %xmm2 # encoding: [0x62,0xf2,0x7e,0x08,0x33,0xc2]
; X64-NEXT:    vpmovdw %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x33,0xc1]
; X64-NEXT:    vpmovdw %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0x89,0x33,0xc0]
; X64-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 # encoding: [0xc5,0xf1,0xfd,0xc0]
; X64-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 # encoding: [0xc5,0xe9,0xfd,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
    %res0 = call <8 x i16> @llvm.x86.avx512.mask.pmov.dw.128(<4 x i32> %x0, <8 x i16> %x1, i8 -1)
    %res1 = call <8 x i16> @llvm.x86.avx512.mask.pmov.dw.128(<4 x i32> %x0, <8 x i16> %x1, i8 %x2)
    %res2 = call <8 x i16> @llvm.x86.avx512.mask.pmov.dw.128(<4 x i32> %x0, <8 x i16> zeroinitializer, i8 %x2)
    %res3 = add <8 x i16> %res0, %res1
    %res4 = add <8 x i16> %res3, %res2
    ret <8 x i16> %res4
}

declare void @llvm.x86.avx512.mask.pmov.dw.mem.128(i8* %ptr, <4 x i32>, i8)

define void @test_int_x86_avx512_mask_pmov_dw_mem_128(i8* %ptr, <4 x i32> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmov_dw_mem_128:
; X86:       # %bb.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:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpmovdw %xmm0, (%eax) # encoding: [0x62,0xf2,0x7e,0x08,0x33,0x00]
; X86-NEXT:    vpmovdw %xmm0, (%eax) {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x33,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmov_dw_mem_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpmovdw %xmm0, (%rdi) # encoding: [0x62,0xf2,0x7e,0x08,0x33,0x07]
; X64-NEXT:    vpmovdw %xmm0, (%rdi) {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x33,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
    call void @llvm.x86.avx512.mask.pmov.dw.mem.128(i8* %ptr, <4 x i32> %x1, i8 -1)
    call void @llvm.x86.avx512.mask.pmov.dw.mem.128(i8* %ptr, <4 x i32> %x1, i8 %x2)
    ret void
}

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

define <8 x i16>@test_int_x86_avx512_mask_pmovs_dw_128(<4 x i32> %x0, <8 x i16> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmovs_dw_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:    vpmovsdw %xmm0, %xmm2 # encoding: [0x62,0xf2,0x7e,0x08,0x23,0xc2]
; X86-NEXT:    vpmovsdw %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x23,0xc1]
; X86-NEXT:    vpmovsdw %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0x89,0x23,0xc0]
; X86-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 # encoding: [0xc5,0xf1,0xfd,0xc0]
; X86-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 # encoding: [0xc5,0xe9,0xfd,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmovs_dw_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmovsdw %xmm0, %xmm2 # encoding: [0x62,0xf2,0x7e,0x08,0x23,0xc2]
; X64-NEXT:    vpmovsdw %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x23,0xc1]
; X64-NEXT:    vpmovsdw %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0x89,0x23,0xc0]
; X64-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 # encoding: [0xc5,0xf1,0xfd,0xc0]
; X64-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 # encoding: [0xc5,0xe9,0xfd,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
    %res0 = call <8 x i16> @llvm.x86.avx512.mask.pmovs.dw.128(<4 x i32> %x0, <8 x i16> %x1, i8 -1)
    %res1 = call <8 x i16> @llvm.x86.avx512.mask.pmovs.dw.128(<4 x i32> %x0, <8 x i16> %x1, i8 %x2)
    %res2 = call <8 x i16> @llvm.x86.avx512.mask.pmovs.dw.128(<4 x i32> %x0, <8 x i16> zeroinitializer, i8 %x2)
    %res3 = add <8 x i16> %res0, %res1
    %res4 = add <8 x i16> %res3, %res2
    ret <8 x i16> %res4
}

declare void @llvm.x86.avx512.mask.pmovs.dw.mem.128(i8* %ptr, <4 x i32>, i8)

define void @test_int_x86_avx512_mask_pmovs_dw_mem_128(i8* %ptr, <4 x i32> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmovs_dw_mem_128:
; X86:       # %bb.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:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpmovsdw %xmm0, (%eax) # encoding: [0x62,0xf2,0x7e,0x08,0x23,0x00]
; X86-NEXT:    vpmovsdw %xmm0, (%eax) {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x23,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmovs_dw_mem_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpmovsdw %xmm0, (%rdi) # encoding: [0x62,0xf2,0x7e,0x08,0x23,0x07]
; X64-NEXT:    vpmovsdw %xmm0, (%rdi) {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x23,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
    call void @llvm.x86.avx512.mask.pmovs.dw.mem.128(i8* %ptr, <4 x i32> %x1, i8 -1)
    call void @llvm.x86.avx512.mask.pmovs.dw.mem.128(i8* %ptr, <4 x i32> %x1, i8 %x2)
    ret void
}

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

define <8 x i16>@test_int_x86_avx512_mask_pmovus_dw_128(<4 x i32> %x0, <8 x i16> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmovus_dw_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:    vpmovusdw %xmm0, %xmm2 # encoding: [0x62,0xf2,0x7e,0x08,0x13,0xc2]
; X86-NEXT:    vpmovusdw %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x13,0xc1]
; X86-NEXT:    vpmovusdw %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0x89,0x13,0xc0]
; X86-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 # encoding: [0xc5,0xf1,0xfd,0xc0]
; X86-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 # encoding: [0xc5,0xe9,0xfd,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmovus_dw_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmovusdw %xmm0, %xmm2 # encoding: [0x62,0xf2,0x7e,0x08,0x13,0xc2]
; X64-NEXT:    vpmovusdw %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x13,0xc1]
; X64-NEXT:    vpmovusdw %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0x89,0x13,0xc0]
; X64-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 # encoding: [0xc5,0xf1,0xfd,0xc0]
; X64-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 # encoding: [0xc5,0xe9,0xfd,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
    %res0 = call <8 x i16> @llvm.x86.avx512.mask.pmovus.dw.128(<4 x i32> %x0, <8 x i16> %x1, i8 -1)
    %res1 = call <8 x i16> @llvm.x86.avx512.mask.pmovus.dw.128(<4 x i32> %x0, <8 x i16> %x1, i8 %x2)
    %res2 = call <8 x i16> @llvm.x86.avx512.mask.pmovus.dw.128(<4 x i32> %x0, <8 x i16> zeroinitializer, i8 %x2)
    %res3 = add <8 x i16> %res0, %res1
    %res4 = add <8 x i16> %res3, %res2
    ret <8 x i16> %res4
}

declare void @llvm.x86.avx512.mask.pmovus.dw.mem.128(i8* %ptr, <4 x i32>, i8)

define void @test_int_x86_avx512_mask_pmovus_dw_mem_128(i8* %ptr, <4 x i32> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmovus_dw_mem_128:
; X86:       # %bb.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:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpmovusdw %xmm0, (%eax) # encoding: [0x62,0xf2,0x7e,0x08,0x13,0x00]
; X86-NEXT:    vpmovusdw %xmm0, (%eax) {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x13,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmovus_dw_mem_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpmovusdw %xmm0, (%rdi) # encoding: [0x62,0xf2,0x7e,0x08,0x13,0x07]
; X64-NEXT:    vpmovusdw %xmm0, (%rdi) {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x13,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
    call void @llvm.x86.avx512.mask.pmovus.dw.mem.128(i8* %ptr, <4 x i32> %x1, i8 -1)
    call void @llvm.x86.avx512.mask.pmovus.dw.mem.128(i8* %ptr, <4 x i32> %x1, i8 %x2)
    ret void
}

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

define <8 x i16>@test_int_x86_avx512_mask_pmov_dw_256(<8 x i32> %x0, <8 x i16> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmov_dw_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:    vpmovdw %ymm0, %xmm2 # encoding: [0x62,0xf2,0x7e,0x28,0x33,0xc2]
; X86-NEXT:    vpmovdw %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x33,0xc1]
; X86-NEXT:    vpmovdw %ymm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0xa9,0x33,0xc0]
; X86-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 # encoding: [0xc5,0xf1,0xfd,0xc0]
; X86-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 # encoding: [0xc5,0xe9,0xfd,0xc0]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmov_dw_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmovdw %ymm0, %xmm2 # encoding: [0x62,0xf2,0x7e,0x28,0x33,0xc2]
; X64-NEXT:    vpmovdw %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x33,0xc1]
; X64-NEXT:    vpmovdw %ymm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0xa9,0x33,0xc0]
; X64-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 # encoding: [0xc5,0xf1,0xfd,0xc0]
; X64-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 # encoding: [0xc5,0xe9,0xfd,0xc0]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
    %res0 = call <8 x i16> @llvm.x86.avx512.mask.pmov.dw.256(<8 x i32> %x0, <8 x i16> %x1, i8 -1)
    %res1 = call <8 x i16> @llvm.x86.avx512.mask.pmov.dw.256(<8 x i32> %x0, <8 x i16> %x1, i8 %x2)
    %res2 = call <8 x i16> @llvm.x86.avx512.mask.pmov.dw.256(<8 x i32> %x0, <8 x i16> zeroinitializer, i8 %x2)
    %res3 = add <8 x i16> %res0, %res1
    %res4 = add <8 x i16> %res3, %res2
    ret <8 x i16> %res4
}

declare void @llvm.x86.avx512.mask.pmov.dw.mem.256(i8* %ptr, <8 x i32>, i8)

define void @test_int_x86_avx512_mask_pmov_dw_mem_256(i8* %ptr, <8 x i32> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmov_dw_mem_256:
; X86:       # %bb.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:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpmovdw %ymm0, (%eax) # encoding: [0x62,0xf2,0x7e,0x28,0x33,0x00]
; X86-NEXT:    vpmovdw %ymm0, (%eax) {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x33,0x00]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmov_dw_mem_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpmovdw %ymm0, (%rdi) # encoding: [0x62,0xf2,0x7e,0x28,0x33,0x07]
; X64-NEXT:    vpmovdw %ymm0, (%rdi) {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x33,0x07]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
    call void @llvm.x86.avx512.mask.pmov.dw.mem.256(i8* %ptr, <8 x i32> %x1, i8 -1)
    call void @llvm.x86.avx512.mask.pmov.dw.mem.256(i8* %ptr, <8 x i32> %x1, i8 %x2)
    ret void
}

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

define <8 x i16>@test_int_x86_avx512_mask_pmovs_dw_256(<8 x i32> %x0, <8 x i16> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmovs_dw_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:    vpmovsdw %ymm0, %xmm2 # encoding: [0x62,0xf2,0x7e,0x28,0x23,0xc2]
; X86-NEXT:    vpmovsdw %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x23,0xc1]
; X86-NEXT:    vpmovsdw %ymm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0xa9,0x23,0xc0]
; X86-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 # encoding: [0xc5,0xf1,0xfd,0xc0]
; X86-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 # encoding: [0xc5,0xe9,0xfd,0xc0]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmovs_dw_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmovsdw %ymm0, %xmm2 # encoding: [0x62,0xf2,0x7e,0x28,0x23,0xc2]
; X64-NEXT:    vpmovsdw %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x23,0xc1]
; X64-NEXT:    vpmovsdw %ymm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0xa9,0x23,0xc0]
; X64-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 # encoding: [0xc5,0xf1,0xfd,0xc0]
; X64-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 # encoding: [0xc5,0xe9,0xfd,0xc0]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
    %res0 = call <8 x i16> @llvm.x86.avx512.mask.pmovs.dw.256(<8 x i32> %x0, <8 x i16> %x1, i8 -1)
    %res1 = call <8 x i16> @llvm.x86.avx512.mask.pmovs.dw.256(<8 x i32> %x0, <8 x i16> %x1, i8 %x2)
    %res2 = call <8 x i16> @llvm.x86.avx512.mask.pmovs.dw.256(<8 x i32> %x0, <8 x i16> zeroinitializer, i8 %x2)
    %res3 = add <8 x i16> %res0, %res1
    %res4 = add <8 x i16> %res3, %res2
    ret <8 x i16> %res4
}

declare void @llvm.x86.avx512.mask.pmovs.dw.mem.256(i8* %ptr, <8 x i32>, i8)

define void @test_int_x86_avx512_mask_pmovs_dw_mem_256(i8* %ptr, <8 x i32> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmovs_dw_mem_256:
; X86:       # %bb.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:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpmovsdw %ymm0, (%eax) # encoding: [0x62,0xf2,0x7e,0x28,0x23,0x00]
; X86-NEXT:    vpmovsdw %ymm0, (%eax) {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x23,0x00]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmovs_dw_mem_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpmovsdw %ymm0, (%rdi) # encoding: [0x62,0xf2,0x7e,0x28,0x23,0x07]
; X64-NEXT:    vpmovsdw %ymm0, (%rdi) {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x23,0x07]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
    call void @llvm.x86.avx512.mask.pmovs.dw.mem.256(i8* %ptr, <8 x i32> %x1, i8 -1)
    call void @llvm.x86.avx512.mask.pmovs.dw.mem.256(i8* %ptr, <8 x i32> %x1, i8 %x2)
    ret void
}

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

define <8 x i16>@test_int_x86_avx512_mask_pmovus_dw_256(<8 x i32> %x0, <8 x i16> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmovus_dw_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:    vpmovusdw %ymm0, %xmm2 # encoding: [0x62,0xf2,0x7e,0x28,0x13,0xc2]
; X86-NEXT:    vpmovusdw %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x13,0xc1]
; X86-NEXT:    vpmovusdw %ymm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0xa9,0x13,0xc0]
; X86-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 # encoding: [0xc5,0xf1,0xfd,0xc0]
; X86-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 # encoding: [0xc5,0xe9,0xfd,0xc0]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmovus_dw_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpmovusdw %ymm0, %xmm2 # encoding: [0x62,0xf2,0x7e,0x28,0x13,0xc2]
; X64-NEXT:    vpmovusdw %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x13,0xc1]
; X64-NEXT:    vpmovusdw %ymm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0xa9,0x13,0xc0]
; X64-NEXT:    vpaddw %xmm0, %xmm1, %xmm0 # encoding: [0xc5,0xf1,0xfd,0xc0]
; X64-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 # encoding: [0xc5,0xe9,0xfd,0xc0]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
    %res0 = call <8 x i16> @llvm.x86.avx512.mask.pmovus.dw.256(<8 x i32> %x0, <8 x i16> %x1, i8 -1)
    %res1 = call <8 x i16> @llvm.x86.avx512.mask.pmovus.dw.256(<8 x i32> %x0, <8 x i16> %x1, i8 %x2)
    %res2 = call <8 x i16> @llvm.x86.avx512.mask.pmovus.dw.256(<8 x i32> %x0, <8 x i16> zeroinitializer, i8 %x2)
    %res3 = add <8 x i16> %res0, %res1
    %res4 = add <8 x i16> %res3, %res2
    ret <8 x i16> %res4
}

declare void @llvm.x86.avx512.mask.pmovus.dw.mem.256(i8* %ptr, <8 x i32>, i8)

define void @test_int_x86_avx512_mask_pmovus_dw_mem_256(i8* %ptr, <8 x i32> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_pmovus_dw_mem_256:
; X86:       # %bb.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:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    vpmovusdw %ymm0, (%eax) # encoding: [0x62,0xf2,0x7e,0x28,0x13,0x00]
; X86-NEXT:    vpmovusdw %ymm0, (%eax) {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x13,0x00]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_pmovus_dw_mem_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
; X64-NEXT:    vpmovusdw %ymm0, (%rdi) # encoding: [0x62,0xf2,0x7e,0x28,0x13,0x07]
; X64-NEXT:    vpmovusdw %ymm0, (%rdi) {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x13,0x07]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
    call void @llvm.x86.avx512.mask.pmovus.dw.mem.256(i8* %ptr, <8 x i32> %x1, i8 -1)
    call void @llvm.x86.avx512.mask.pmovus.dw.mem.256(i8* %ptr, <8 x i32> %x1, i8 %x2)
    ret void
}

declare <4 x i32> @llvm.x86.avx512.mask.cvtpd2dq.128(<2 x double>, <4 x i32>, i8)

define <4 x i32>@test_int_x86_avx512_mask_cvt_pd2dq_128(<2 x double> %x0, <4 x i32> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_cvt_pd2dq_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:    vcvtpd2dq %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xff,0x09,0xe6,0xc8]
; X86-NEXT:    vcvtpd2dq %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xfb,0xe6,0xc0]
; 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_cvt_pd2dq_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vcvtpd2dq %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xff,0x09,0xe6,0xc8]
; X64-NEXT:    vcvtpd2dq %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xfb,0xe6,0xc0]
; 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.cvtpd2dq.128(<2 x double> %x0, <4 x i32> %x1, i8 %x2)
  %res1 = call <4 x i32> @llvm.x86.avx512.mask.cvtpd2dq.128(<2 x double> %x0, <4 x i32> %x1, i8 -1)
  %res2 = add <4 x i32> %res, %res1
  ret <4 x i32> %res2
}

define <4 x i32>@test_int_x86_avx512_mask_cvt_pd2dq_128_zext(<2 x double> %x0, <4 x i32> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_cvt_pd2dq_128_zext:
; 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:    vcvtpd2dq %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xff,0x09,0xe6,0xc8]
; X86-NEXT:    vcvtpd2dq %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xfb,0xe6,0xc0]
; 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_cvt_pd2dq_128_zext:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vcvtpd2dq %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xff,0x09,0xe6,0xc8]
; X64-NEXT:    vcvtpd2dq %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xfb,0xe6,0xc0]
; 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.cvtpd2dq.128(<2 x double> %x0, <4 x i32> %x1, i8 %x2)
  %res1 = shufflevector <4 x i32> %res, <4 x i32> zeroinitializer, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
  %res2 = call <4 x i32> @llvm.x86.avx512.mask.cvtpd2dq.128(<2 x double> %x0, <4 x i32> %x1, i8 -1)
  %res3 = shufflevector <4 x i32> %res2, <4 x i32> zeroinitializer, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
  %res4 = add <4 x i32> %res1, %res3
  ret <4 x i32> %res4
}

declare <4 x float> @llvm.x86.avx512.mask.cvtpd2ps(<2 x double>, <4 x float>, i8)

define <4 x float>@test_int_x86_avx512_mask_cvt_pd2ps(<2 x double> %x0, <4 x float> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_cvt_pd2ps:
; 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:    vcvtpd2ps %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xfd,0x09,0x5a,0xc8]
; X86-NEXT:    vcvtpd2ps %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x5a,0xc0]
; X86-NEXT:    vaddps %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf0,0x58,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_cvt_pd2ps:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vcvtpd2ps %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xfd,0x09,0x5a,0xc8]
; X64-NEXT:    vcvtpd2ps %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x5a,0xc0]
; 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.cvtpd2ps(<2 x double> %x0, <4 x float> %x1, i8 %x2)
  %res1 = call <4 x float> @llvm.x86.avx512.mask.cvtpd2ps(<2 x double> %x0, <4 x float> %x1, i8 -1)
  %res2 = fadd <4 x float> %res, %res1
  ret <4 x float> %res2
}

define <4 x float>@test_int_x86_avx512_mask_cvt_pd2ps_zext(<2 x double> %x0, <4 x float> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_cvt_pd2ps_zext:
; 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:    vcvtpd2ps %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xfd,0x09,0x5a,0xc8]
; X86-NEXT:    vcvtpd2ps %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x5a,0xc0]
; X86-NEXT:    vaddps %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf0,0x58,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_cvt_pd2ps_zext:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vcvtpd2ps %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xfd,0x09,0x5a,0xc8]
; X64-NEXT:    vcvtpd2ps %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x5a,0xc0]
; 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.cvtpd2ps(<2 x double> %x0, <4 x float> %x1, i8 %x2)
  %res1 = shufflevector <4 x float> %res, <4 x float> zeroinitializer, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
  %res2 = call <4 x float> @llvm.x86.avx512.mask.cvtpd2ps(<2 x double> %x0, <4 x float> %x1, i8 -1)
  %res3 = shufflevector <4 x float> %res2, <4 x float> zeroinitializer, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
  %res4 = fadd <4 x float> %res1, %res3
  ret <4 x float> %res4
}

declare <4 x i32> @llvm.x86.avx512.mask.cvtpd2udq.128(<2 x double>, <4 x i32>, i8)

define <4 x i32>@test_int_x86_avx512_mask_cvt_pd2udq_128(<2 x double> %x0, <4 x i32> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_cvt_pd2udq_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:    vcvtpd2udq %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xfc,0x09,0x79,0xc8]
; X86-NEXT:    vcvtpd2udq %xmm0, %xmm0 # encoding: [0x62,0xf1,0xfc,0x08,0x79,0xc0]
; 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_cvt_pd2udq_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vcvtpd2udq %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xfc,0x09,0x79,0xc8]
; X64-NEXT:    vcvtpd2udq %xmm0, %xmm0 # encoding: [0x62,0xf1,0xfc,0x08,0x79,0xc0]
; 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.cvtpd2udq.128(<2 x double> %x0, <4 x i32> %x1, i8 %x2)
  %res1 = call <4 x i32> @llvm.x86.avx512.mask.cvtpd2udq.128(<2 x double> %x0, <4 x i32> %x1, i8 -1)
  %res2 = add <4 x i32> %res, %res1
  ret <4 x i32> %res2
}

define <4 x i32>@test_int_x86_avx512_mask_cvt_pd2udq_128_zext(<2 x double> %x0, <4 x i32> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_cvt_pd2udq_128_zext:
; 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:    vcvtpd2udq %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xfc,0x09,0x79,0xc8]
; X86-NEXT:    vcvtpd2udq %xmm0, %xmm0 # encoding: [0x62,0xf1,0xfc,0x08,0x79,0xc0]
; 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_cvt_pd2udq_128_zext:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vcvtpd2udq %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xfc,0x09,0x79,0xc8]
; X64-NEXT:    vcvtpd2udq %xmm0, %xmm0 # encoding: [0x62,0xf1,0xfc,0x08,0x79,0xc0]
; 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.cvtpd2udq.128(<2 x double> %x0, <4 x i32> %x1, i8 %x2)
  %res1 = shufflevector <4 x i32> %res, <4 x i32> zeroinitializer, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
  %res2 = call <4 x i32> @llvm.x86.avx512.mask.cvtpd2udq.128(<2 x double> %x0, <4 x i32> %x1, i8 -1)
  %res3 = shufflevector <4 x i32> %res2, <4 x i32> zeroinitializer, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
  %res4 = add <4 x i32> %res1, %res3
  ret <4 x i32> %res4
}

declare <4 x i32> @llvm.x86.avx512.mask.cvtpd2udq.256(<4 x double>, <4 x i32>, i8)

define <4 x i32>@test_int_x86_avx512_mask_cvt_pd2udq_256(<4 x double> %x0, <4 x i32> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_cvt_pd2udq_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:    vcvtpd2udq %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xfc,0x29,0x79,0xc8]
; X86-NEXT:    vcvtpd2udq %ymm0, %xmm0 # encoding: [0x62,0xf1,0xfc,0x28,0x79,0xc0]
; X86-NEXT:    vpaddd %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc0]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_cvt_pd2udq_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vcvtpd2udq %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xfc,0x29,0x79,0xc8]
; X64-NEXT:    vcvtpd2udq %ymm0, %xmm0 # encoding: [0x62,0xf1,0xfc,0x28,0x79,0xc0]
; X64-NEXT:    vpaddd %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc0]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i32> @llvm.x86.avx512.mask.cvtpd2udq.256(<4 x double> %x0, <4 x i32> %x1, i8 %x2)
  %res1 = call <4 x i32> @llvm.x86.avx512.mask.cvtpd2udq.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.cvtps2dq.128(<4 x float>, <4 x i32>, i8)

define <4 x i32>@test_int_x86_avx512_mask_cvt_ps2dq_128(<4 x float> %x0, <4 x i32> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_cvt_ps2dq_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:    vcvtps2dq %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x5b,0xc8]
; X86-NEXT:    vcvtps2dq %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x5b,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_cvt_ps2dq_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vcvtps2dq %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x5b,0xc8]
; X64-NEXT:    vcvtps2dq %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x5b,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.cvtps2dq.128(<4 x float> %x0, <4 x i32> %x1, i8 %x2)
  %res1 = call <4 x i32> @llvm.x86.avx512.mask.cvtps2dq.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.cvtps2dq.256(<8 x float>, <8 x i32>, i8)

define <8 x i32>@test_int_x86_avx512_mask_cvt_ps2dq_256(<8 x float> %x0, <8 x i32> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_cvt_ps2dq_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:    vcvtps2dq %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x5b,0xc8]
; X86-NEXT:    vcvtps2dq %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x5b,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_cvt_ps2dq_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vcvtps2dq %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x5b,0xc8]
; X64-NEXT:    vcvtps2dq %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x5b,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.cvtps2dq.256(<8 x float> %x0, <8 x i32> %x1, i8 %x2)
  %res1 = call <8 x i32> @llvm.x86.avx512.mask.cvtps2dq.256(<8 x float> %x0, <8 x i32> %x1, i8 -1)
  %res2 = add <8 x i32> %res, %res1
  ret <8 x i32> %res2
}

declare <4 x i32> @llvm.x86.avx512.mask.cvtps2udq.128(<4 x float>, <4 x i32>, i8)

define <4 x i32>@test_int_x86_avx512_mask_cvt_ps2udq_128(<4 x float> %x0, <4 x i32> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_cvt_ps2udq_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:    vcvtps2udq %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7c,0x09,0x79,0xc8]
; X86-NEXT:    vcvtps2udq %xmm0, %xmm0 # encoding: [0x62,0xf1,0x7c,0x08,0x79,0xc0]
; 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_cvt_ps2udq_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vcvtps2udq %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7c,0x09,0x79,0xc8]
; X64-NEXT:    vcvtps2udq %xmm0, %xmm0 # encoding: [0x62,0xf1,0x7c,0x08,0x79,0xc0]
; 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.cvtps2udq.128(<4 x float> %x0, <4 x i32> %x1, i8 %x2)
  %res1 = call <4 x i32> @llvm.x86.avx512.mask.cvtps2udq.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.cvtps2udq.256(<8 x float>, <8 x i32>, i8)

define <8 x i32>@test_int_x86_avx512_mask_cvt_ps2udq_256(<8 x float> %x0, <8 x i32> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_cvt_ps2udq_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:    vcvtps2udq %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7c,0x29,0x79,0xc8]
; X86-NEXT:    vcvtps2udq %ymm0, %ymm0 # encoding: [0x62,0xf1,0x7c,0x28,0x79,0xc0]
; 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_cvt_ps2udq_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vcvtps2udq %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7c,0x29,0x79,0xc8]
; X64-NEXT:    vcvtps2udq %ymm0, %ymm0 # encoding: [0x62,0xf1,0x7c,0x28,0x79,0xc0]
; 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.cvtps2udq.256(<8 x float> %x0, <8 x i32> %x1, i8 %x2)
  %res1 = call <8 x i32> @llvm.x86.avx512.mask.cvtps2udq.256(<8 x float> %x0, <8 x i32> %x1, i8 -1)
  %res2 = add <8 x i32> %res, %res1
  ret <8 x i32> %res2
}

declare <4 x i32> @llvm.x86.avx512.mask.cvttpd2dq.128(<2 x double>, <4 x i32>, i8)

define <4 x i32>@test_int_x86_avx512_mask_cvtt_pd2dq_128(<2 x double> %x0, <4 x i32> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_cvtt_pd2dq_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:    vcvttpd2dq %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xfd,0x09,0xe6,0xc8]
; X86-NEXT:    vcvttpd2dq %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xe6,0xc0]
; 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_cvtt_pd2dq_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vcvttpd2dq %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xfd,0x09,0xe6,0xc8]
; X64-NEXT:    vcvttpd2dq %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xe6,0xc0]
; 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.cvttpd2dq.128(<2 x double> %x0, <4 x i32> %x1, i8 %x2)
  %res1 = call <4 x i32> @llvm.x86.avx512.mask.cvttpd2dq.128(<2 x double> %x0, <4 x i32> %x1, i8 -1)
  %res2 = add <4 x i32> %res, %res1
  ret <4 x i32> %res2
}

define <4 x i32>@test_int_x86_avx512_mask_cvtt_pd2dq_128_zext(<2 x double> %x0, <4 x i32> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_cvtt_pd2dq_128_zext:
; 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:    vcvttpd2dq %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xfd,0x09,0xe6,0xc8]
; X86-NEXT:    vcvttpd2dq %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xe6,0xc0]
; 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_cvtt_pd2dq_128_zext:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vcvttpd2dq %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xfd,0x09,0xe6,0xc8]
; X64-NEXT:    vcvttpd2dq %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xe6,0xc0]
; 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.cvttpd2dq.128(<2 x double> %x0, <4 x i32> %x1, i8 %x2)
  %res1 = shufflevector <4 x i32> %res, <4 x i32> zeroinitializer, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
  %res2 = call <4 x i32> @llvm.x86.avx512.mask.cvttpd2dq.128(<2 x double> %x0, <4 x i32> %x1, i8 -1)
  %res3 = shufflevector <4 x i32> %res2, <4 x i32> zeroinitializer, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
  %res4 = add <4 x i32> %res1, %res3
  ret <4 x i32> %res4
}

declare <4 x i32> @llvm.x86.avx512.mask.cvttpd2udq.128(<2 x double>, <4 x i32>, i8)

define <4 x i32>@test_int_x86_avx512_mask_cvtt_pd2udq_128(<2 x double> %x0, <4 x i32> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_cvtt_pd2udq_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:    vcvttpd2udq %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xfc,0x09,0x78,0xc8]
; X86-NEXT:    vcvttpd2udq %xmm0, %xmm0 # encoding: [0x62,0xf1,0xfc,0x08,0x78,0xc0]
; 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_cvtt_pd2udq_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vcvttpd2udq %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xfc,0x09,0x78,0xc8]
; X64-NEXT:    vcvttpd2udq %xmm0, %xmm0 # encoding: [0x62,0xf1,0xfc,0x08,0x78,0xc0]
; 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.cvttpd2udq.128(<2 x double> %x0, <4 x i32> %x1, i8 %x2)
  %res1 = call <4 x i32> @llvm.x86.avx512.mask.cvttpd2udq.128(<2 x double> %x0, <4 x i32> %x1, i8 -1)
  %res2 = add <4 x i32> %res, %res1
  ret <4 x i32> %res2
}

define <4 x i32>@test_int_x86_avx512_mask_cvtt_pd2udq_128_zext(<2 x double> %x0, <4 x i32> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_cvtt_pd2udq_128_zext:
; 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:    vcvttpd2udq %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xfc,0x09,0x78,0xc8]
; X86-NEXT:    vcvttpd2udq %xmm0, %xmm0 # encoding: [0x62,0xf1,0xfc,0x08,0x78,0xc0]
; 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_cvtt_pd2udq_128_zext:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vcvttpd2udq %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xfc,0x09,0x78,0xc8]
; X64-NEXT:    vcvttpd2udq %xmm0, %xmm0 # encoding: [0x62,0xf1,0xfc,0x08,0x78,0xc0]
; 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.cvttpd2udq.128(<2 x double> %x0, <4 x i32> %x1, i8 %x2)
  %res1 = shufflevector <4 x i32> %res, <4 x i32> zeroinitializer, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
  %res2 = call <4 x i32> @llvm.x86.avx512.mask.cvttpd2udq.128(<2 x double> %x0, <4 x i32> %x1, i8 -1)
  %res3 = shufflevector <4 x i32> %res2, <4 x i32> zeroinitializer, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
  %res4 = add <4 x i32> %res1, %res3
  ret <4 x i32> %res4
}

declare <4 x i32> @llvm.x86.avx512.mask.cvttpd2udq.256(<4 x double>, <4 x i32>, i8)

define <4 x i32>@test_int_x86_avx512_mask_cvtt_pd2udq_256(<4 x double> %x0, <4 x i32> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_cvtt_pd2udq_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:    vcvttpd2udq %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xfc,0x29,0x78,0xc8]
; X86-NEXT:    vcvttpd2udq %ymm0, %xmm0 # encoding: [0x62,0xf1,0xfc,0x28,0x78,0xc0]
; X86-NEXT:    vpaddd %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc0]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_cvtt_pd2udq_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vcvttpd2udq %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xfc,0x29,0x78,0xc8]
; X64-NEXT:    vcvttpd2udq %ymm0, %xmm0 # encoding: [0x62,0xf1,0xfc,0x28,0x78,0xc0]
; X64-NEXT:    vpaddd %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfe,0xc0]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i32> @llvm.x86.avx512.mask.cvttpd2udq.256(<4 x double> %x0, <4 x i32> %x1, i8 %x2)
  %res1 = call <4 x i32> @llvm.x86.avx512.mask.cvttpd2udq.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.cvttps2udq.128(<4 x float>, <4 x i32>, i8)

define <4 x i32>@test_int_x86_avx512_mask_cvtt_ps2udq_128(<4 x float> %x0, <4 x i32> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_cvtt_ps2udq_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:    vcvttps2udq %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7c,0x09,0x78,0xc8]
; X86-NEXT:    vcvttps2udq %xmm0, %xmm0 # encoding: [0x62,0xf1,0x7c,0x08,0x78,0xc0]
; 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_cvtt_ps2udq_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vcvttps2udq %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7c,0x09,0x78,0xc8]
; X64-NEXT:    vcvttps2udq %xmm0, %xmm0 # encoding: [0x62,0xf1,0x7c,0x08,0x78,0xc0]
; 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.cvttps2udq.128(<4 x float> %x0, <4 x i32> %x1, i8 %x2)
  %res1 = call <4 x i32> @llvm.x86.avx512.mask.cvttps2udq.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.cvttps2udq.256(<8 x float>, <8 x i32>, i8)

define <8 x i32>@test_int_x86_avx512_mask_cvtt_ps2udq_256(<8 x float> %x0, <8 x i32> %x1, i8 %x2) {
; X86-LABEL: test_int_x86_avx512_mask_cvtt_ps2udq_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:    vcvttps2udq %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7c,0x29,0x78,0xc8]
; X86-NEXT:    vcvttps2udq %ymm0, %ymm0 # encoding: [0x62,0xf1,0x7c,0x28,0x78,0xc0]
; 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_cvtt_ps2udq_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vcvttps2udq %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7c,0x29,0x78,0xc8]
; X64-NEXT:    vcvttps2udq %ymm0, %ymm0 # encoding: [0x62,0xf1,0x7c,0x28,0x78,0xc0]
; 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.cvttps2udq.256(<8 x float> %x0, <8 x i32> %x1, i8 %x2)
  %res1 = call <8 x i32> @llvm.x86.avx512.mask.cvttps2udq.256(<8 x float> %x0, <8 x i32> %x1, i8 -1)
  %res2 = add <8 x i32> %res, %res1
  ret <8 x i32> %res2
}

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

define <2 x double>@test_int_x86_avx512_mask_rndscale_pd_128(<2 x double> %x0, <2 x double> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_rndscale_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:    vrndscalepd $4, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x09,0xc8,0x04]
; X86-NEXT:    vrndscalepd $88, %xmm0, %xmm0 # encoding: [0x62,0xf3,0xfd,0x08,0x09,0xc0,0x58]
; X86-NEXT:    vaddpd %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0x58,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_rndscale_pd_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vrndscalepd $4, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x09,0xc8,0x04]
; X64-NEXT:    vrndscalepd $88, %xmm0, %xmm0 # encoding: [0x62,0xf3,0xfd,0x08,0x09,0xc0,0x58]
; 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.rndscale.pd.128(<2 x double> %x0, i32 4, <2 x double> %x2, i8 %x3)
  %res1 = call <2 x double> @llvm.x86.avx512.mask.rndscale.pd.128(<2 x double> %x0, i32 88, <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.rndscale.pd.256(<4 x double>, i32, <4 x double>, i8)

define <4 x double>@test_int_x86_avx512_mask_rndscale_pd_256(<4 x double> %x0, <4 x double> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_rndscale_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:    vrndscalepd $4, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x09,0xc8,0x04]
; X86-NEXT:    vrndscalepd $88, %ymm0, %ymm0 # encoding: [0x62,0xf3,0xfd,0x28,0x09,0xc0,0x58]
; 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_rndscale_pd_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vrndscalepd $4, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x09,0xc8,0x04]
; X64-NEXT:    vrndscalepd $88, %ymm0, %ymm0 # encoding: [0x62,0xf3,0xfd,0x28,0x09,0xc0,0x58]
; 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.rndscale.pd.256(<4 x double> %x0, i32 4, <4 x double> %x2, i8 %x3)
  %res1 = call <4 x double> @llvm.x86.avx512.mask.rndscale.pd.256(<4 x double> %x0, i32 88, <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.rndscale.ps.128(<4 x float>, i32, <4 x float>, i8)

define <4 x float>@test_int_x86_avx512_mask_rndscale_ps_128(<4 x float> %x0, <4 x float> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_rndscale_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:    vrndscaleps $88, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x08,0xc8,0x58]
; X86-NEXT:    vroundps $4, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x08,0xc0,0x04]
; X86-NEXT:    vaddps %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf0,0x58,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_rndscale_ps_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vrndscaleps $88, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x08,0xc8,0x58]
; X64-NEXT:    vroundps $4, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x08,0xc0,0x04]
; 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.rndscale.ps.128(<4 x float> %x0, i32 88, <4 x float> %x2, i8 %x3)
  %res1 = call <4 x float> @llvm.x86.avx512.mask.rndscale.ps.128(<4 x float> %x0, i32 4, <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.rndscale.ps.256(<8 x float>, i32, <8 x float>, i8)

define <8 x float>@test_int_x86_avx512_mask_rndscale_ps_256(<8 x float> %x0, <8 x float> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_rndscale_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:    vrndscaleps $5, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x08,0xc8,0x05]
; X86-NEXT:    vrndscaleps $66, %ymm0, %ymm0 # encoding: [0x62,0xf3,0x7d,0x28,0x08,0xc0,0x42]
; 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_rndscale_ps_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vrndscaleps $5, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x08,0xc8,0x05]
; X64-NEXT:    vrndscaleps $66, %ymm0, %ymm0 # encoding: [0x62,0xf3,0x7d,0x28,0x08,0xc0,0x42]
; 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.rndscale.ps.256(<8 x float> %x0, i32 5, <8 x float> %x2, i8 %x3)
  %res1 = call <8 x float> @llvm.x86.avx512.mask.rndscale.ps.256(<8 x float> %x0, i32 66, <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.getmant.pd.128(<2 x double>, i32, <2 x double>, i8)

define <2 x double>@test_int_x86_avx512_mask_getmant_pd_128(<2 x double> %x0, <2 x double> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_getmant_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:    vgetmantpd $11, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x26,0xc8,0x0b]
; X86-NEXT:    vgetmantpd $11, %xmm0, %xmm2 {%k1} {z} # encoding: [0x62,0xf3,0xfd,0x89,0x26,0xd0,0x0b]
; X86-NEXT:    vgetmantpd $11, %xmm0, %xmm0 # encoding: [0x62,0xf3,0xfd,0x08,0x26,0xc0,0x0b]
; X86-NEXT:    vaddpd %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0x58,0xc0]
; X86-NEXT:    vaddpd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x58,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_getmant_pd_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vgetmantpd $11, %xmm0, %xmm2 {%k1} {z} # encoding: [0x62,0xf3,0xfd,0x89,0x26,0xd0,0x0b]
; X64-NEXT:    vgetmantpd $11, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x26,0xc8,0x0b]
; X64-NEXT:    vgetmantpd $11, %xmm0, %xmm0 # encoding: [0x62,0xf3,0xfd,0x08,0x26,0xc0,0x0b]
; X64-NEXT:    vaddpd %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0x58,0xc0]
; X64-NEXT:    vaddpd %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x58,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x double> @llvm.x86.avx512.mask.getmant.pd.128(<2 x double> %x0, i32 11, <2 x double> %x2, i8 %x3)
  %res2 = call <2 x double> @llvm.x86.avx512.mask.getmant.pd.128(<2 x double> %x0, i32 11, <2 x double> zeroinitializer, i8 %x3)
  %res1 = call <2 x double> @llvm.x86.avx512.mask.getmant.pd.128(<2 x double> %x0, i32 11, <2 x double> %x2, i8 -1)
  %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.getmant.pd.256(<4 x double>, i32, <4 x double>, i8)

define <4 x double>@test_int_x86_avx512_mask_getmant_pd_256(<4 x double> %x0, <4 x double> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_getmant_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:    vgetmantpd $11, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x26,0xc8,0x0b]
; X86-NEXT:    vgetmantpd $11, %ymm0, %ymm0 # encoding: [0x62,0xf3,0xfd,0x28,0x26,0xc0,0x0b]
; 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_getmant_pd_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vgetmantpd $11, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x26,0xc8,0x0b]
; X64-NEXT:    vgetmantpd $11, %ymm0, %ymm0 # encoding: [0x62,0xf3,0xfd,0x28,0x26,0xc0,0x0b]
; 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.getmant.pd.256(<4 x double> %x0, i32 11, <4 x double> %x2, i8 %x3)
  %res1 = call <4 x double> @llvm.x86.avx512.mask.getmant.pd.256(<4 x double> %x0, i32 11, <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.getmant.ps.128(<4 x float>, i32, <4 x float>, i8)

define <4 x float>@test_int_x86_avx512_mask_getmant_ps_128(<4 x float> %x0, <4 x float> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_getmant_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:    vgetmantps $11, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x26,0xc8,0x0b]
; X86-NEXT:    vgetmantps $11, %xmm0, %xmm0 # encoding: [0x62,0xf3,0x7d,0x08,0x26,0xc0,0x0b]
; X86-NEXT:    vaddps %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf0,0x58,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_getmant_ps_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vgetmantps $11, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x26,0xc8,0x0b]
; X64-NEXT:    vgetmantps $11, %xmm0, %xmm0 # encoding: [0x62,0xf3,0x7d,0x08,0x26,0xc0,0x0b]
; 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.getmant.ps.128(<4 x float> %x0, i32 11, <4 x float> %x2, i8 %x3)
  %res1 = call <4 x float> @llvm.x86.avx512.mask.getmant.ps.128(<4 x float> %x0, i32 11, <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.getmant.ps.256(<8 x float>, i32, <8 x float>, i8)

define <8 x float>@test_int_x86_avx512_mask_getmant_ps_256(<8 x float> %x0, <8 x float> %x2, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_getmant_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:    vgetmantps $11, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x26,0xc8,0x0b]
; X86-NEXT:    vgetmantps $11, %ymm0, %ymm0 # encoding: [0x62,0xf3,0x7d,0x28,0x26,0xc0,0x0b]
; 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_getmant_ps_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vgetmantps $11, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x26,0xc8,0x0b]
; X64-NEXT:    vgetmantps $11, %ymm0, %ymm0 # encoding: [0x62,0xf3,0x7d,0x28,0x26,0xc0,0x0b]
; 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.getmant.ps.256(<8 x float> %x0, i32 11, <8 x float> %x2, i8 %x3)
  %res1 = call <8 x float> @llvm.x86.avx512.mask.getmant.ps.256(<8 x float> %x0, i32 11, <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.pternlog.d.128(<4 x i32>, <4 x i32>, <4 x i32>, i32)

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]
  %1 = call <4 x i32> @llvm.x86.avx512.pternlog.d.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i32 33)
  %2 = bitcast i8 %x4 to <8 x i1>
  %extract = 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> %extract, <4 x i32> %1, <4 x i32> %x0
  %4 = call <4 x i32> @llvm.x86.avx512.pternlog.d.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i32 33)
  %res2 = add <4 x i32> %3, %4
  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]
  %1 = call <4 x i32> @llvm.x86.avx512.pternlog.d.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i32 33)
  %2 = bitcast i8 %x4 to <8 x i1>
  %extract = 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> %extract, <4 x i32> %1, <4 x i32> zeroinitializer
  %4 = call <4 x i32> @llvm.x86.avx512.pternlog.d.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i32 33)
  %res2 = add <4 x i32> %3, %4
  ret <4 x i32> %res2
}

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

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]
  %1 = call <8 x i32> @llvm.x86.avx512.pternlog.d.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i32 33)
  %2 = bitcast i8 %x4 to <8 x i1>
  %3 = select <8 x i1> %2, <8 x i32> %1, <8 x i32> %x0
  %4 = call <8 x i32> @llvm.x86.avx512.pternlog.d.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i32 33)
  %res2 = add <8 x i32> %3, %4
  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]
  %1 = call <8 x i32> @llvm.x86.avx512.pternlog.d.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i32 33)
  %2 = bitcast i8 %x4 to <8 x i1>
  %3 = select <8 x i1> %2, <8 x i32> %1, <8 x i32> zeroinitializer
  %4 = call <8 x i32> @llvm.x86.avx512.pternlog.d.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i32 33)
  %res2 = add <8 x i32> %3, %4
  ret <8 x i32> %res2
}

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

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]
  %1 = call <2 x i64> @llvm.x86.avx512.pternlog.q.128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i32 33)
  %2 = bitcast i8 %x4 to <8 x i1>
  %extract = shufflevector <8 x i1> %2, <8 x i1> %2, <2 x i32> <i32 0, i32 1>
  %3 = select <2 x i1> %extract, <2 x i64> %1, <2 x i64> %x0
  %4 = call <2 x i64> @llvm.x86.avx512.pternlog.q.128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i32 33)
  %res2 = add <2 x i64> %3, %4
  ret <2 x i64> %res2
}

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]
  %1 = call <2 x i64> @llvm.x86.avx512.pternlog.q.128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i32 33)
  %2 = bitcast i8 %x4 to <8 x i1>
  %extract = shufflevector <8 x i1> %2, <8 x i1> %2, <2 x i32> <i32 0, i32 1>
  %3 = select <2 x i1> %extract, <2 x i64> %1, <2 x i64> zeroinitializer
  %4 = call <2 x i64> @llvm.x86.avx512.pternlog.q.128(<2 x i64> %x0, <2 x i64> %x1, <2 x i64> %x2, i32 33)
  %res2 = add <2 x i64> %3, %4
  ret <2 x i64> %res2
}

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

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]
  %1 = call <4 x i64> @llvm.x86.avx512.pternlog.q.256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2, i32 33)
  %2 = bitcast i8 %x4 to <8 x i1>
  %extract = 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> %extract, <4 x i64> %1, <4 x i64> %x0
  %4 = call <4 x i64> @llvm.x86.avx512.pternlog.q.256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2, i32 33)
  %res2 = add <4 x i64> %3, %4
  ret <4 x i64> %res2
}

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]
  %1 = call <4 x i64> @llvm.x86.avx512.pternlog.q.256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2, i32 33)
  %2 = bitcast i8 %x4 to <8 x i1>
  %extract = 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> %extract, <4 x i64> %1, <4 x i64> zeroinitializer
  %4 = call <4 x i64> @llvm.x86.avx512.pternlog.q.256(<4 x i64> %x0, <4 x i64> %x1, <4 x i64> %x2, i32 33)
  %res2 = add <4 x i64> %3, %4
  ret <4 x i64> %res2
}

define <4 x float> @test_x86_vcvtph2ps_128(<8 x i16> %a0) {
; CHECK-LABEL: test_x86_vcvtph2ps_128:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vcvtph2ps %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x13,0xc0]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask.vcvtph2ps.128(<8 x i16> %a0, <4 x float> zeroinitializer, i8 -1)
  ret <4 x float> %res
}

define <4 x float> @test_x86_vcvtph2ps_128_rrk(<8 x i16> %a0,<4 x float> %a1, i8 %mask) {
; X86-LABEL: test_x86_vcvtph2ps_128_rrk:
; 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:    vcvtph2ps %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x13,0xc8]
; X86-NEXT:    vmovaps %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x28,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_x86_vcvtph2ps_128_rrk:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vcvtph2ps %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x13,0xc8]
; X64-NEXT:    vmovaps %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x28,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask.vcvtph2ps.128(<8 x i16> %a0, <4 x float> %a1, i8 %mask)
  ret <4 x float> %res
}


define <4 x float> @test_x86_vcvtph2ps_128_rrkz(<8 x i16> %a0, i8 %mask) {
; X86-LABEL: test_x86_vcvtph2ps_128_rrkz:
; 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:    vcvtph2ps %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x13,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_x86_vcvtph2ps_128_rrkz:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vcvtph2ps %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x13,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask.vcvtph2ps.128(<8 x i16> %a0, <4 x float> zeroinitializer, i8 %mask)
  ret <4 x float> %res
}

declare <4 x float> @llvm.x86.avx512.mask.vcvtph2ps.128(<8 x i16>, <4 x float>, i8) nounwind readonly

define <8 x float> @test_x86_vcvtph2ps_256(<8 x i16> %a0) {
; CHECK-LABEL: test_x86_vcvtph2ps_256:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vcvtph2ps %xmm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x13,0xc0]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask.vcvtph2ps.256(<8 x i16> %a0, <8 x float> zeroinitializer, i8 -1)
  ret <8 x float> %res
}

define <8 x float> @test_x86_vcvtph2ps_256_rrk(<8 x i16> %a0,<8 x float> %a1, i8 %mask) {
; X86-LABEL: test_x86_vcvtph2ps_256_rrk:
; 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:    vcvtph2ps %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x13,0xc8]
; X86-NEXT:    vmovaps %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x28,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_x86_vcvtph2ps_256_rrk:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vcvtph2ps %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x13,0xc8]
; X64-NEXT:    vmovaps %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x28,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask.vcvtph2ps.256(<8 x i16> %a0, <8 x float> %a1, i8 %mask)
  ret <8 x float> %res
}

define <8 x float> @test_x86_vcvtph2ps_256_rrkz(<8 x i16> %a0, i8 %mask) {
; X86-LABEL: test_x86_vcvtph2ps_256_rrkz:
; 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:    vcvtph2ps %xmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x13,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_x86_vcvtph2ps_256_rrkz:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vcvtph2ps %xmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x13,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.mask.vcvtph2ps.256(<8 x i16> %a0, <8 x float> zeroinitializer, i8 %mask)
  ret <8 x float> %res
}

declare <8 x float> @llvm.x86.avx512.mask.vcvtph2ps.256(<8 x i16>, <8 x float>, i8) nounwind readonly

define <8 x i16> @test_x86_vcvtps2ph_128(<4 x float> %a0, i8 %mask, <8 x i16> %src) {
; X86-LABEL: test_x86_vcvtps2ph_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:    vcvtps2ph $2, %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x1d,0xc2,0x02]
; X86-NEXT:    vcvtps2ph $2, %xmm0, %xmm3 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0x89,0x1d,0xc3,0x02]
; X86-NEXT:    vcvtps2ph $2, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x1d,0xc1,0x02]
; X86-NEXT:    vpaddw %xmm1, %xmm3, %xmm0 # encoding: [0xc5,0xe1,0xfd,0xc1]
; X86-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 # encoding: [0xc5,0xe9,0xfd,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_x86_vcvtps2ph_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vcvtps2ph $2, %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x1d,0xc2,0x02]
; X64-NEXT:    vcvtps2ph $2, %xmm0, %xmm3 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0x89,0x1d,0xc3,0x02]
; X64-NEXT:    vcvtps2ph $2, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x1d,0xc1,0x02]
; X64-NEXT:    vpaddw %xmm1, %xmm3, %xmm0 # encoding: [0xc5,0xe1,0xfd,0xc1]
; X64-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 # encoding: [0xc5,0xe9,0xfd,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res1 = call <8 x i16> @llvm.x86.avx512.mask.vcvtps2ph.128(<4 x float> %a0, i32 2, <8 x i16> zeroinitializer, i8 -1)
  %res2 = call <8 x i16> @llvm.x86.avx512.mask.vcvtps2ph.128(<4 x float> %a0, i32 2, <8 x i16> zeroinitializer, i8 %mask)
  %res3 = call <8 x i16> @llvm.x86.avx512.mask.vcvtps2ph.128(<4 x float> %a0, i32 2, <8 x i16> %src, i8 %mask)
  %res0 = add <8 x i16> %res1, %res2
  %res = add <8 x i16> %res3, %res0
  ret <8 x i16> %res
}

declare <8 x i16> @llvm.x86.avx512.mask.vcvtps2ph.128(<4 x float>, i32, <8 x i16>, i8) nounwind readonly

define <8 x i16> @test_x86_vcvtps2ph_256(<8 x float> %a0, i8 %mask, <8 x i16> %src) {
; X86-LABEL: test_x86_vcvtps2ph_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:    vcvtps2ph $2, %ymm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x7d,0x1d,0xc2,0x02]
; X86-NEXT:    vcvtps2ph $2, %ymm0, %xmm3 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0xa9,0x1d,0xc3,0x02]
; X86-NEXT:    vcvtps2ph $2, %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x1d,0xc1,0x02]
; X86-NEXT:    vpaddw %xmm1, %xmm3, %xmm0 # encoding: [0xc5,0xe1,0xfd,0xc1]
; X86-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 # encoding: [0xc5,0xe9,0xfd,0xc0]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_x86_vcvtps2ph_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vcvtps2ph $2, %ymm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x7d,0x1d,0xc2,0x02]
; X64-NEXT:    vcvtps2ph $2, %ymm0, %xmm3 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0xa9,0x1d,0xc3,0x02]
; X64-NEXT:    vcvtps2ph $2, %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x1d,0xc1,0x02]
; X64-NEXT:    vpaddw %xmm1, %xmm3, %xmm0 # encoding: [0xc5,0xe1,0xfd,0xc1]
; X64-NEXT:    vpaddw %xmm0, %xmm2, %xmm0 # encoding: [0xc5,0xe9,0xfd,0xc0]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
  %res1 = call <8 x i16> @llvm.x86.avx512.mask.vcvtps2ph.256(<8 x float> %a0, i32 2, <8 x i16> zeroinitializer, i8 -1)
  %res2 = call <8 x i16> @llvm.x86.avx512.mask.vcvtps2ph.256(<8 x float> %a0, i32 2, <8 x i16> zeroinitializer, i8 %mask)
  %res3 = call <8 x i16> @llvm.x86.avx512.mask.vcvtps2ph.256(<8 x float> %a0, i32 2, <8 x i16> %src, i8 %mask)
  %res0 = add <8 x i16> %res1, %res2
  %res = add <8 x i16> %res3, %res0
  ret <8 x i16> %res
}

declare <8 x i16> @llvm.x86.avx512.mask.vcvtps2ph.256(<8 x float>, i32, <8 x i16>, i8) nounwind readonly

define <8 x float> @test_rsqrt_ps_256_rr(<8 x float> %a0) {
; CHECK-LABEL: test_rsqrt_ps_256_rr:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vrsqrt14ps %ymm0, %ymm0 # encoding: [0x62,0xf2,0x7d,0x28,0x4e,0xc0]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.rsqrt14.ps.256(<8 x float> %a0, <8 x float> zeroinitializer, i8 -1)
  ret <8 x float> %res
}

define <8 x float> @test_rsqrt_ps_256_rrkz(<8 x float> %a0, i8 %mask) {
; X86-LABEL: test_rsqrt_ps_256_rrkz:
; 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:    vrsqrt14ps %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x4e,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_rsqrt_ps_256_rrkz:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vrsqrt14ps %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x4e,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.rsqrt14.ps.256(<8 x float> %a0, <8 x float> zeroinitializer, i8 %mask)
  ret <8 x float> %res
}

define <8 x float> @test_rsqrt_ps_256_rrk(<8 x float> %a0, <8 x float> %a1, i8 %mask) {
; X86-LABEL: test_rsqrt_ps_256_rrk:
; 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:    vrsqrt14ps %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x4e,0xc8]
; X86-NEXT:    vmovaps %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x28,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_rsqrt_ps_256_rrk:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vrsqrt14ps %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x4e,0xc8]
; X64-NEXT:    vmovaps %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x28,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.rsqrt14.ps.256(<8 x float> %a0, <8 x float> %a1, i8 %mask)
  ret <8 x float> %res
}

define <4 x float> @test_rsqrt_ps_128_rr(<4 x float> %a0) {
; CHECK-LABEL: test_rsqrt_ps_128_rr:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vrsqrt14ps %xmm0, %xmm0 # encoding: [0x62,0xf2,0x7d,0x08,0x4e,0xc0]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.rsqrt14.ps.128(<4 x float> %a0, <4 x float> zeroinitializer, i8 -1)
  ret <4 x float> %res
}

define <4 x float> @test_rsqrt_ps_128_rrkz(<4 x float> %a0, i8 %mask) {
; X86-LABEL: test_rsqrt_ps_128_rrkz:
; 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:    vrsqrt14ps %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x4e,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_rsqrt_ps_128_rrkz:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vrsqrt14ps %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x4e,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.rsqrt14.ps.128(<4 x float> %a0, <4 x float> zeroinitializer, i8 %mask)
  ret <4 x float> %res
}

define <4 x float> @test_rsqrt_ps_128_rrk(<4 x float> %a0, <4 x float> %a1, i8 %mask) {
; X86-LABEL: test_rsqrt_ps_128_rrk:
; 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:    vrsqrt14ps %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x4e,0xc8]
; X86-NEXT:    vmovaps %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x28,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_rsqrt_ps_128_rrk:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vrsqrt14ps %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x4e,0xc8]
; X64-NEXT:    vmovaps %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x28,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.rsqrt14.ps.128(<4 x float> %a0, <4 x float> %a1, i8 %mask)
  ret <4 x float> %res
}

declare <8 x float> @llvm.x86.avx512.rsqrt14.ps.256(<8 x float>, <8 x float>, i8) nounwind readnone
declare <4 x float> @llvm.x86.avx512.rsqrt14.ps.128(<4 x float>, <4 x float>, i8) nounwind readnone

define <8 x float> @test_rcp_ps_256_rr(<8 x float> %a0) {
; CHECK-LABEL: test_rcp_ps_256_rr:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vrcp14ps %ymm0, %ymm0 # encoding: [0x62,0xf2,0x7d,0x28,0x4c,0xc0]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.rcp14.ps.256(<8 x float> %a0, <8 x float> zeroinitializer, i8 -1)
  ret <8 x float> %res
}

define <8 x float> @test_rcp_ps_256_rrkz(<8 x float> %a0, i8 %mask) {
; X86-LABEL: test_rcp_ps_256_rrkz:
; 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:    vrcp14ps %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x4c,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_rcp_ps_256_rrkz:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vrcp14ps %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x4c,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.rcp14.ps.256(<8 x float> %a0, <8 x float> zeroinitializer, i8 %mask)
  ret <8 x float> %res
}

define <8 x float> @test_rcp_ps_256_rrk(<8 x float> %a0, <8 x float> %a1, i8 %mask) {
; X86-LABEL: test_rcp_ps_256_rrk:
; 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:    vrcp14ps %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x4c,0xc8]
; X86-NEXT:    vmovaps %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x28,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_rcp_ps_256_rrk:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vrcp14ps %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x4c,0xc8]
; X64-NEXT:    vmovaps %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x28,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x float> @llvm.x86.avx512.rcp14.ps.256(<8 x float> %a0, <8 x float> %a1, i8 %mask)
  ret <8 x float> %res
}

define <4 x float> @test_rcp_ps_128_rr(<4 x float> %a0) {
; CHECK-LABEL: test_rcp_ps_128_rr:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vrcp14ps %xmm0, %xmm0 # encoding: [0x62,0xf2,0x7d,0x08,0x4c,0xc0]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.rcp14.ps.128(<4 x float> %a0, <4 x float> zeroinitializer, i8 -1)
  ret <4 x float> %res
}

define <4 x float> @test_rcp_ps_128_rrkz(<4 x float> %a0, i8 %mask) {
; X86-LABEL: test_rcp_ps_128_rrkz:
; 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:    vrcp14ps %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x4c,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_rcp_ps_128_rrkz:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vrcp14ps %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x4c,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.rcp14.ps.128(<4 x float> %a0, <4 x float> zeroinitializer, i8 %mask)
  ret <4 x float> %res
}

define <4 x float> @test_rcp_ps_128_rrk(<4 x float> %a0, <4 x float> %a1, i8 %mask) {
; X86-LABEL: test_rcp_ps_128_rrk:
; 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:    vrcp14ps %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x4c,0xc8]
; X86-NEXT:    vmovaps %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x28,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_rcp_ps_128_rrk:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vrcp14ps %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x4c,0xc8]
; X64-NEXT:    vmovaps %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x28,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.rcp14.ps.128(<4 x float> %a0, <4 x float> %a1, i8 %mask)
  ret <4 x float> %res
}

declare <8 x float> @llvm.x86.avx512.rcp14.ps.256(<8 x float>, <8 x float>, i8) nounwind readnone
declare <4 x float> @llvm.x86.avx512.rcp14.ps.128(<4 x float>, <4 x float>, i8) nounwind readnone

define <4 x double> @test_rsqrt_pd_256_rr(<4 x double> %a0) {
; CHECK-LABEL: test_rsqrt_pd_256_rr:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vrsqrt14pd %ymm0, %ymm0 # encoding: [0x62,0xf2,0xfd,0x28,0x4e,0xc0]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <4 x double> @llvm.x86.avx512.rsqrt14.pd.256(<4 x double> %a0, <4 x double> zeroinitializer, i8 -1)
  ret <4 x double> %res
}

define <4 x double> @test_rsqrt_pd_256_rrkz(<4 x double> %a0, i8 %mask) {
; X86-LABEL: test_rsqrt_pd_256_rrkz:
; 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:    vrsqrt14pd %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x4e,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_rsqrt_pd_256_rrkz:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vrsqrt14pd %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x4e,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x double> @llvm.x86.avx512.rsqrt14.pd.256(<4 x double> %a0, <4 x double> zeroinitializer, i8 %mask)
  ret <4 x double> %res
}

define <4 x double> @test_rsqrt_pd_256_rrk(<4 x double> %a0, <4 x double> %a1, i8 %mask) {
; X86-LABEL: test_rsqrt_pd_256_rrk:
; 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:    vrsqrt14pd %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x4e,0xc8]
; X86-NEXT:    vmovapd %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x28,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_rsqrt_pd_256_rrk:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vrsqrt14pd %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x4e,0xc8]
; X64-NEXT:    vmovapd %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x28,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x double> @llvm.x86.avx512.rsqrt14.pd.256(<4 x double> %a0, <4 x double> %a1, i8 %mask)
  ret <4 x double> %res
}

define <2 x double> @test_rsqrt_pd_128_rr(<2 x double> %a0) {
; CHECK-LABEL: test_rsqrt_pd_128_rr:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vrsqrt14pd %xmm0, %xmm0 # encoding: [0x62,0xf2,0xfd,0x08,0x4e,0xc0]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <2 x double> @llvm.x86.avx512.rsqrt14.pd.128(<2 x double> %a0, <2 x double> zeroinitializer, i8 -1)
  ret <2 x double> %res
}

define <2 x double> @test_rsqrt_pd_128_rrkz(<2 x double> %a0, i8 %mask) {
; X86-LABEL: test_rsqrt_pd_128_rrkz:
; 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:    vrsqrt14pd %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x4e,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_rsqrt_pd_128_rrkz:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vrsqrt14pd %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x4e,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x double> @llvm.x86.avx512.rsqrt14.pd.128(<2 x double> %a0, <2 x double> zeroinitializer, i8 %mask)
  ret <2 x double> %res
}

define <2 x double> @test_rsqrt_pd_128_rrk(<2 x double> %a0, <2 x double> %a1, i8 %mask) {
; X86-LABEL: test_rsqrt_pd_128_rrk:
; 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:    vrsqrt14pd %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x4e,0xc8]
; X86-NEXT:    vmovapd %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x28,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_rsqrt_pd_128_rrk:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vrsqrt14pd %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x4e,0xc8]
; X64-NEXT:    vmovapd %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x28,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x double> @llvm.x86.avx512.rsqrt14.pd.128(<2 x double> %a0, <2 x double> %a1, i8 %mask)
  ret <2 x double> %res
}

declare <4 x double> @llvm.x86.avx512.rsqrt14.pd.256(<4 x double>, <4 x double>, i8) nounwind readnone
declare <2 x double> @llvm.x86.avx512.rsqrt14.pd.128(<2 x double>, <2 x double>, i8) nounwind readnone

define <4 x double> @test_rcp_pd_256_rr(<4 x double> %a0) {
; CHECK-LABEL: test_rcp_pd_256_rr:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vrcp14pd %ymm0, %ymm0 # encoding: [0x62,0xf2,0xfd,0x28,0x4c,0xc0]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <4 x double> @llvm.x86.avx512.rcp14.pd.256(<4 x double> %a0, <4 x double> zeroinitializer, i8 -1)
  ret <4 x double> %res
}

define <4 x double> @test_rcp_pd_256_rrkz(<4 x double> %a0, i8 %mask) {
; X86-LABEL: test_rcp_pd_256_rrkz:
; 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:    vrcp14pd %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x4c,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_rcp_pd_256_rrkz:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vrcp14pd %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x4c,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x double> @llvm.x86.avx512.rcp14.pd.256(<4 x double> %a0, <4 x double> zeroinitializer, i8 %mask)
  ret <4 x double> %res
}

define <4 x double> @test_rcp_pd_256_rrk(<4 x double> %a0, <4 x double> %a1, i8 %mask) {
; X86-LABEL: test_rcp_pd_256_rrk:
; 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:    vrcp14pd %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x4c,0xc8]
; X86-NEXT:    vmovapd %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x28,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_rcp_pd_256_rrk:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vrcp14pd %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x4c,0xc8]
; X64-NEXT:    vmovapd %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x28,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x double> @llvm.x86.avx512.rcp14.pd.256(<4 x double> %a0, <4 x double> %a1, i8 %mask)
  ret <4 x double> %res
}

define <2 x double> @test_rcp_pd_128_rr(<2 x double> %a0) {
; CHECK-LABEL: test_rcp_pd_128_rr:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vrcp14pd %xmm0, %xmm0 # encoding: [0x62,0xf2,0xfd,0x08,0x4c,0xc0]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <2 x double> @llvm.x86.avx512.rcp14.pd.128(<2 x double> %a0, <2 x double> zeroinitializer, i8 -1)
  ret <2 x double> %res
}

define <2 x double> @test_rcp_pd_128_rrkz(<2 x double> %a0, i8 %mask) {
; X86-LABEL: test_rcp_pd_128_rrkz:
; 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:    vrcp14pd %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x4c,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_rcp_pd_128_rrkz:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vrcp14pd %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x4c,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x double> @llvm.x86.avx512.rcp14.pd.128(<2 x double> %a0, <2 x double> zeroinitializer, i8 %mask)
  ret <2 x double> %res
}

define <2 x double> @test_rcp_pd_128_rrk(<2 x double> %a0, <2 x double> %a1, i8 %mask) {
; X86-LABEL: test_rcp_pd_128_rrk:
; 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:    vrcp14pd %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x4c,0xc8]
; X86-NEXT:    vmovapd %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x28,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_rcp_pd_128_rrk:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vrcp14pd %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x4c,0xc8]
; X64-NEXT:    vmovapd %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x28,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x double> @llvm.x86.avx512.rcp14.pd.128(<2 x double> %a0, <2 x double> %a1, i8 %mask)
  ret <2 x double> %res
}

declare <4 x double> @llvm.x86.avx512.rcp14.pd.256(<4 x double>, <4 x double>, i8) nounwind readnone
declare <2 x double> @llvm.x86.avx512.rcp14.pd.128(<2 x double>, <2 x double>, i8) nounwind readnone

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

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]
  %1 = call <4 x double> @llvm.x86.avx512.permvar.df.256(<4 x double> %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 double> %1, <4 x double> %x2
  %4 = call <4 x double> @llvm.x86.avx512.permvar.df.256(<4 x double> %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 double> %4, <4 x double> zeroinitializer
  %7 = call <4 x double> @llvm.x86.avx512.permvar.df.256(<4 x double> %x0, <4 x i64> %x1)
  %res3 = fadd <4 x double> %3, %6
  %res4 = fadd <4 x double> %res3, %7
  ret <4 x double> %res4
}

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

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]
  %1 = call <4 x i64> @llvm.x86.avx512.permvar.di.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.permvar.di.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.permvar.di.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 <2 x double> @llvm.x86.avx512.mask.fixupimm.pd.128(<2 x double>, <2 x double>, <2 x i64>, i32, i8)

define <2 x double>@test_int_x86_avx512_mask_fixupimm_pd_128(<2 x double> %x0, <2 x double> %x1, <2 x i64> %x2, i8 %x4) {
; X86-LABEL: test_int_x86_avx512_mask_fixupimm_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:    vmovapd %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x28,0xd8]
; X86-NEXT:    vfixupimmpd $5, %xmm2, %xmm1, %xmm3 {%k1} # encoding: [0x62,0xf3,0xf5,0x09,0x54,0xda,0x05]
; X86-NEXT:    vxorpd %xmm4, %xmm4, %xmm4 # encoding: [0xc5,0xd9,0x57,0xe4]
; X86-NEXT:    vfixupimmpd $4, %xmm2, %xmm1, %xmm4 {%k1} {z} # encoding: [0x62,0xf3,0xf5,0x89,0x54,0xe2,0x04]
; X86-NEXT:    vaddpd %xmm4, %xmm3, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0x58,0xdc]
; X86-NEXT:    vfixupimmpd $3, %xmm2, %xmm1, %xmm0 # encoding: [0x62,0xf3,0xf5,0x08,0x54,0xc2,0x03]
; X86-NEXT:    vaddpd %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0x58,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_fixupimm_pd_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vmovapd %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x28,0xd8]
; X64-NEXT:    vfixupimmpd $5, %xmm2, %xmm1, %xmm3 {%k1} # encoding: [0x62,0xf3,0xf5,0x09,0x54,0xda,0x05]
; X64-NEXT:    vxorpd %xmm4, %xmm4, %xmm4 # encoding: [0xc5,0xd9,0x57,0xe4]
; X64-NEXT:    vfixupimmpd $4, %xmm2, %xmm1, %xmm4 {%k1} {z} # encoding: [0x62,0xf3,0xf5,0x89,0x54,0xe2,0x04]
; X64-NEXT:    vaddpd %xmm4, %xmm3, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0x58,0xdc]
; X64-NEXT:    vfixupimmpd $3, %xmm2, %xmm1, %xmm0 # encoding: [0x62,0xf3,0xf5,0x08,0x54,0xc2,0x03]
; X64-NEXT:    vaddpd %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0x58,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x double> @llvm.x86.avx512.mask.fixupimm.pd.128(<2 x double> %x0, <2 x double> %x1,<2 x i64> %x2, i32 5, i8 %x4)
  %res1 = call <2 x double> @llvm.x86.avx512.mask.fixupimm.pd.128(<2 x double> zeroinitializer, <2 x double> %x1, <2 x i64> %x2, i32 4, i8 %x4)
  %res2 = call <2 x double> @llvm.x86.avx512.mask.fixupimm.pd.128(<2 x double> %x0, <2 x double> %x1, <2 x i64> %x2, i32 3, i8 -1)
  %res3 = fadd <2 x double> %res, %res1
  %res4 = fadd <2 x double> %res3, %res2
  ret <2 x double> %res4
}

declare <2 x double> @llvm.x86.avx512.maskz.fixupimm.pd.128(<2 x double>, <2 x double>, <2 x i64>, i32, i8)

define <2 x double>@test_int_x86_avx512_maskz_fixupimm_pd_128(<2 x double> %x0, <2 x double> %x1, <2 x i64> %x2, i8 %x4) {
; X86-LABEL: test_int_x86_avx512_maskz_fixupimm_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:    vmovapd %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x28,0xd8]
; X86-NEXT:    vfixupimmpd $5, %xmm2, %xmm1, %xmm3 {%k1} {z} # encoding: [0x62,0xf3,0xf5,0x89,0x54,0xda,0x05]
; X86-NEXT:    vxorpd %xmm2, %xmm2, %xmm2 # encoding: [0xc5,0xe9,0x57,0xd2]
; X86-NEXT:    vfixupimmpd $3, %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf3,0xf5,0x89,0x54,0xc2,0x03]
; X86-NEXT:    vaddpd %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0x58,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_maskz_fixupimm_pd_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vmovapd %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x28,0xd8]
; X64-NEXT:    vfixupimmpd $5, %xmm2, %xmm1, %xmm3 {%k1} {z} # encoding: [0x62,0xf3,0xf5,0x89,0x54,0xda,0x05]
; X64-NEXT:    vxorpd %xmm2, %xmm2, %xmm2 # encoding: [0xc5,0xe9,0x57,0xd2]
; X64-NEXT:    vfixupimmpd $3, %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf3,0xf5,0x89,0x54,0xc2,0x03]
; X64-NEXT:    vaddpd %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0x58,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x double> @llvm.x86.avx512.maskz.fixupimm.pd.128(<2 x double> %x0, <2 x double> %x1, <2 x i64> %x2, i32 5, i8 %x4)
  %res1 = call <2 x double> @llvm.x86.avx512.maskz.fixupimm.pd.128(<2 x double> %x0, <2 x double> %x1, <2 x i64> zeroinitializer, i32 3, i8 %x4)
  ;%res2 = call <2 x double> @llvm.x86.avx512.maskz.fixupimm.pd.128(<2 x double> %x0, <2 x double> %x1, <2 x i64> %x2, i32 4, i8 -1)
  %res3 = fadd <2 x double> %res, %res1
  ;%res4 = fadd <2 x double> %res3, %res2
  ret <2 x double> %res3
}

declare <4 x double> @llvm.x86.avx512.mask.fixupimm.pd.256(<4 x double>, <4 x double>, <4 x i64>, i32, i8)

define <4 x double>@test_int_x86_avx512_mask_fixupimm_pd_256(<4 x double> %x0, <4 x double> %x1, <4 x i64> %x2, i8 %x4) {
; X86-LABEL: test_int_x86_avx512_mask_fixupimm_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:    vmovapd %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x28,0xd8]
; X86-NEXT:    vfixupimmpd $4, %ymm2, %ymm1, %ymm3 {%k1} # encoding: [0x62,0xf3,0xf5,0x29,0x54,0xda,0x04]
; X86-NEXT:    vxorpd %xmm4, %xmm4, %xmm4 # encoding: [0xc5,0xd9,0x57,0xe4]
; X86-NEXT:    vfixupimmpd $5, %ymm2, %ymm1, %ymm4 {%k1} {z} # encoding: [0x62,0xf3,0xf5,0xa9,0x54,0xe2,0x05]
; X86-NEXT:    vaddpd %ymm4, %ymm3, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0x58,0xdc]
; X86-NEXT:    vfixupimmpd $3, %ymm2, %ymm1, %ymm0 # encoding: [0x62,0xf3,0xf5,0x28,0x54,0xc2,0x03]
; 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_fixupimm_pd_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vmovapd %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x28,0xd8]
; X64-NEXT:    vfixupimmpd $4, %ymm2, %ymm1, %ymm3 {%k1} # encoding: [0x62,0xf3,0xf5,0x29,0x54,0xda,0x04]
; X64-NEXT:    vxorpd %xmm4, %xmm4, %xmm4 # encoding: [0xc5,0xd9,0x57,0xe4]
; X64-NEXT:    vfixupimmpd $5, %ymm2, %ymm1, %ymm4 {%k1} {z} # encoding: [0x62,0xf3,0xf5,0xa9,0x54,0xe2,0x05]
; X64-NEXT:    vaddpd %ymm4, %ymm3, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0x58,0xdc]
; X64-NEXT:    vfixupimmpd $3, %ymm2, %ymm1, %ymm0 # encoding: [0x62,0xf3,0xf5,0x28,0x54,0xc2,0x03]
; 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.fixupimm.pd.256(<4 x double> %x0, <4 x double> %x1, <4 x i64> %x2, i32 4, i8 %x4)
  %res1 = call <4 x double> @llvm.x86.avx512.mask.fixupimm.pd.256(<4 x double> zeroinitializer, <4 x double> %x1, <4 x i64> %x2 , i32 5, i8 %x4)
  %res2 = call <4 x double> @llvm.x86.avx512.mask.fixupimm.pd.256(<4 x double> %x0, <4 x double> %x1, <4 x i64> %x2, i32 3, i8 -1)
  %res3 = fadd <4 x double> %res, %res1
  %res4 = fadd <4 x double> %res3, %res2
  ret <4 x double> %res4
}

declare <4 x double> @llvm.x86.avx512.maskz.fixupimm.pd.256(<4 x double>, <4 x double>, <4 x i64>, i32, i8)

define <4 x double>@test_int_x86_avx512_maskz_fixupimm_pd_256(<4 x double> %x0, <4 x double> %x1, <4 x i64> %x2, i8 %x4) {
; X86-LABEL: test_int_x86_avx512_maskz_fixupimm_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:    vmovapd %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x28,0xd8]
; X86-NEXT:    vfixupimmpd $5, %ymm2, %ymm1, %ymm3 {%k1} {z} # encoding: [0x62,0xf3,0xf5,0xa9,0x54,0xda,0x05]
; X86-NEXT:    vxorpd %xmm4, %xmm4, %xmm4 # encoding: [0xc5,0xd9,0x57,0xe4]
; X86-NEXT:    vmovapd %ymm0, %ymm5 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x28,0xe8]
; X86-NEXT:    vfixupimmpd $4, %ymm4, %ymm1, %ymm5 {%k1} {z} # encoding: [0x62,0xf3,0xf5,0xa9,0x54,0xec,0x04]
; X86-NEXT:    vaddpd %ymm5, %ymm3, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0x58,0xdd]
; X86-NEXT:    vfixupimmpd $3, %ymm2, %ymm1, %ymm0 # encoding: [0x62,0xf3,0xf5,0x28,0x54,0xc2,0x03]
; 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_maskz_fixupimm_pd_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vmovapd %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x28,0xd8]
; X64-NEXT:    vfixupimmpd $5, %ymm2, %ymm1, %ymm3 {%k1} {z} # encoding: [0x62,0xf3,0xf5,0xa9,0x54,0xda,0x05]
; X64-NEXT:    vxorpd %xmm4, %xmm4, %xmm4 # encoding: [0xc5,0xd9,0x57,0xe4]
; X64-NEXT:    vmovapd %ymm0, %ymm5 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x28,0xe8]
; X64-NEXT:    vfixupimmpd $4, %ymm4, %ymm1, %ymm5 {%k1} {z} # encoding: [0x62,0xf3,0xf5,0xa9,0x54,0xec,0x04]
; X64-NEXT:    vaddpd %ymm5, %ymm3, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0x58,0xdd]
; X64-NEXT:    vfixupimmpd $3, %ymm2, %ymm1, %ymm0 # encoding: [0x62,0xf3,0xf5,0x28,0x54,0xc2,0x03]
; 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.maskz.fixupimm.pd.256(<4 x double> %x0, <4 x double> %x1, <4 x i64> %x2, i32 5, i8 %x4)
  %res1 = call <4 x double> @llvm.x86.avx512.maskz.fixupimm.pd.256(<4 x double> %x0, <4 x double> %x1, <4 x i64> zeroinitializer, i32 4, i8 %x4)
  %res2 = call <4 x double> @llvm.x86.avx512.maskz.fixupimm.pd.256(<4 x double> %x0, <4 x double> %x1, <4 x i64> %x2, i32 3, i8 -1)
  %res3 = fadd <4 x double> %res, %res1
  %res4 = fadd <4 x double> %res3, %res2
  ret <4 x double> %res4
}

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

define <4 x float>@test_int_x86_avx512_mask_fixupimm_ps_128(<4 x float> %x0, <4 x float> %x1, <4 x i32> %x2, i8 %x4) {
; X86-LABEL: test_int_x86_avx512_mask_fixupimm_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:    vmovaps %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x28,0xd8]
; X86-NEXT:    vfixupimmps $5, %xmm2, %xmm1, %xmm3 {%k1} # encoding: [0x62,0xf3,0x75,0x09,0x54,0xda,0x05]
; X86-NEXT:    vmovaps %xmm0, %xmm4 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x28,0xe0]
; X86-NEXT:    vfixupimmps $5, %xmm2, %xmm1, %xmm4 # encoding: [0x62,0xf3,0x75,0x08,0x54,0xe2,0x05]
; X86-NEXT:    vxorps %xmm2, %xmm2, %xmm2 # encoding: [0xc5,0xe8,0x57,0xd2]
; X86-NEXT:    vfixupimmps $5, %xmm2, %xmm1, %xmm0 {%k1} # encoding: [0x62,0xf3,0x75,0x09,0x54,0xc2,0x05]
; X86-NEXT:    vaddps %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe0,0x58,0xc0]
; X86-NEXT:    vaddps %xmm4, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x58,0xc4]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_fixupimm_ps_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vmovaps %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x28,0xd8]
; X64-NEXT:    vfixupimmps $5, %xmm2, %xmm1, %xmm3 # encoding: [0x62,0xf3,0x75,0x08,0x54,0xda,0x05]
; X64-NEXT:    vmovaps %xmm0, %xmm4 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x28,0xe0]
; X64-NEXT:    vfixupimmps $5, %xmm2, %xmm1, %xmm4 {%k1} # encoding: [0x62,0xf3,0x75,0x09,0x54,0xe2,0x05]
; X64-NEXT:    vxorps %xmm2, %xmm2, %xmm2 # encoding: [0xc5,0xe8,0x57,0xd2]
; X64-NEXT:    vfixupimmps $5, %xmm2, %xmm1, %xmm0 {%k1} # encoding: [0x62,0xf3,0x75,0x09,0x54,0xc2,0x05]
; X64-NEXT:    vaddps %xmm0, %xmm4, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xd8,0x58,0xc0]
; X64-NEXT:    vaddps %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x58,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.mask.fixupimm.ps.128(<4 x float> %x0, <4 x float> %x1, <4 x i32> %x2, i32 5, i8 %x4)
  %res1 = call <4 x float> @llvm.x86.avx512.mask.fixupimm.ps.128(<4 x float> %x0, <4 x float> %x1, <4 x i32> zeroinitializer, i32 5, i8 %x4)
  %res2 = call <4 x float> @llvm.x86.avx512.mask.fixupimm.ps.128(<4 x float> %x0, <4 x float> %x1, <4 x i32> %x2, i32 5, i8 -1)
  %res3 = fadd <4 x float> %res, %res1
  %res4 = fadd <4 x float> %res3, %res2
  ret <4 x float> %res4
}

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

define <4 x float>@test_int_x86_avx512_maskz_fixupimm_ps_128(<4 x float> %x0, <4 x float> %x1, <4 x i32> %x2, i8 %x4) {
; X86-LABEL: test_int_x86_avx512_maskz_fixupimm_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:    vmovaps %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x28,0xd8]
; X86-NEXT:    vfixupimmps $5, %xmm2, %xmm1, %xmm3 {%k1} {z} # encoding: [0x62,0xf3,0x75,0x89,0x54,0xda,0x05]
; X86-NEXT:    vmovaps %xmm0, %xmm4 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x28,0xe0]
; X86-NEXT:    vfixupimmps $5, %xmm2, %xmm1, %xmm4 # encoding: [0x62,0xf3,0x75,0x08,0x54,0xe2,0x05]
; X86-NEXT:    vxorps %xmm2, %xmm2, %xmm2 # encoding: [0xc5,0xe8,0x57,0xd2]
; X86-NEXT:    vfixupimmps $5, %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf3,0x75,0x89,0x54,0xc2,0x05]
; X86-NEXT:    vaddps %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe0,0x58,0xc0]
; X86-NEXT:    vaddps %xmm4, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x58,0xc4]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_maskz_fixupimm_ps_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vmovaps %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x28,0xd8]
; X64-NEXT:    vfixupimmps $5, %xmm2, %xmm1, %xmm3 # encoding: [0x62,0xf3,0x75,0x08,0x54,0xda,0x05]
; X64-NEXT:    vmovaps %xmm0, %xmm4 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x28,0xe0]
; X64-NEXT:    vfixupimmps $5, %xmm2, %xmm1, %xmm4 {%k1} {z} # encoding: [0x62,0xf3,0x75,0x89,0x54,0xe2,0x05]
; X64-NEXT:    vxorps %xmm2, %xmm2, %xmm2 # encoding: [0xc5,0xe8,0x57,0xd2]
; X64-NEXT:    vfixupimmps $5, %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf3,0x75,0x89,0x54,0xc2,0x05]
; X64-NEXT:    vaddps %xmm0, %xmm4, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xd8,0x58,0xc0]
; X64-NEXT:    vaddps %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf8,0x58,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x float> @llvm.x86.avx512.maskz.fixupimm.ps.128(<4 x float> %x0, <4 x float> %x1, <4 x i32> %x2, i32 5, i8 %x4)
  %res1 = call <4 x float> @llvm.x86.avx512.maskz.fixupimm.ps.128(<4 x float> %x0, <4 x float> %x1, <4 x i32> zeroinitializer, i32 5, i8 %x4)
  %res2 = call <4 x float> @llvm.x86.avx512.maskz.fixupimm.ps.128(<4 x float> %x0, <4 x float> %x1, <4 x i32> %x2, i32 5, i8 -1)
  %res3 = fadd <4 x float> %res, %res1
  %res4 = fadd <4 x float> %res3, %res2
  ret <4 x float> %res4
}

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

define <8 x float>@test_int_x86_avx512_mask_fixupimm_ps_256(<8 x float> %x0, <8 x float> %x1, <8 x i32> %x2, i8 %x4) {
; X86-LABEL: test_int_x86_avx512_mask_fixupimm_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:    vmovaps %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x28,0xd8]
; X86-NEXT:    vfixupimmps $5, %ymm2, %ymm1, %ymm3 {%k1} # encoding: [0x62,0xf3,0x75,0x29,0x54,0xda,0x05]
; X86-NEXT:    vmovaps %ymm0, %ymm4 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x28,0xe0]
; X86-NEXT:    vfixupimmps $5, %ymm2, %ymm1, %ymm4 # encoding: [0x62,0xf3,0x75,0x28,0x54,0xe2,0x05]
; X86-NEXT:    vxorps %xmm2, %xmm2, %xmm2 # encoding: [0xc5,0xe8,0x57,0xd2]
; X86-NEXT:    vfixupimmps $5, %ymm2, %ymm1, %ymm0 {%k1} # encoding: [0x62,0xf3,0x75,0x29,0x54,0xc2,0x05]
; X86-NEXT:    vaddps %ymm0, %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xe4,0x58,0xc0]
; X86-NEXT:    vaddps %ymm4, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x58,0xc4]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_fixupimm_ps_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vmovaps %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x28,0xd8]
; X64-NEXT:    vfixupimmps $5, %ymm2, %ymm1, %ymm3 # encoding: [0x62,0xf3,0x75,0x28,0x54,0xda,0x05]
; X64-NEXT:    vmovaps %ymm0, %ymm4 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x28,0xe0]
; X64-NEXT:    vfixupimmps $5, %ymm2, %ymm1, %ymm4 {%k1} # encoding: [0x62,0xf3,0x75,0x29,0x54,0xe2,0x05]
; X64-NEXT:    vxorps %xmm2, %xmm2, %xmm2 # encoding: [0xc5,0xe8,0x57,0xd2]
; X64-NEXT:    vfixupimmps $5, %ymm2, %ymm1, %ymm0 {%k1} # encoding: [0x62,0xf3,0x75,0x29,0x54,0xc2,0x05]
; X64-NEXT:    vaddps %ymm0, %ymm4, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xdc,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.fixupimm.ps.256(<8 x float> %x0, <8 x float> %x1, <8 x i32> %x2, i32 5, i8 %x4)
  %res1 = call <8 x float> @llvm.x86.avx512.mask.fixupimm.ps.256(<8 x float> %x0, <8 x float> %x1, <8 x i32> zeroinitializer, i32 5, i8 %x4)
  %res2 = call <8 x float> @llvm.x86.avx512.mask.fixupimm.ps.256(<8 x float> %x0, <8 x float> %x1, <8 x i32> %x2, i32 5, i8 -1)
  %res3 = fadd <8 x float> %res, %res1
  %res4 = fadd <8 x float> %res3, %res2
  ret <8 x float> %res4
}

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

define <8 x float>@test_int_x86_avx512_maskz_fixupimm_ps_256(<8 x float> %x0, <8 x float> %x1, <8 x i32> %x2, i8 %x4) {
; X86-LABEL: test_int_x86_avx512_maskz_fixupimm_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:    vmovaps %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x28,0xd8]
; X86-NEXT:    vfixupimmps $5, %ymm2, %ymm1, %ymm3 {%k1} {z} # encoding: [0x62,0xf3,0x75,0xa9,0x54,0xda,0x05]
; X86-NEXT:    vmovaps %ymm0, %ymm4 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x28,0xe0]
; X86-NEXT:    vfixupimmps $5, %ymm2, %ymm1, %ymm4 # encoding: [0x62,0xf3,0x75,0x28,0x54,0xe2,0x05]
; X86-NEXT:    vxorps %xmm2, %xmm2, %xmm2 # encoding: [0xc5,0xe8,0x57,0xd2]
; X86-NEXT:    vfixupimmps $5, %ymm2, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf3,0x75,0xa9,0x54,0xc2,0x05]
; X86-NEXT:    vaddps %ymm0, %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xe4,0x58,0xc0]
; X86-NEXT:    vaddps %ymm4, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x58,0xc4]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_maskz_fixupimm_ps_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vmovaps %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x28,0xd8]
; X64-NEXT:    vfixupimmps $5, %ymm2, %ymm1, %ymm3 # encoding: [0x62,0xf3,0x75,0x28,0x54,0xda,0x05]
; X64-NEXT:    vmovaps %ymm0, %ymm4 # EVEX TO VEX Compression encoding: [0xc5,0xfc,0x28,0xe0]
; X64-NEXT:    vfixupimmps $5, %ymm2, %ymm1, %ymm4 {%k1} {z} # encoding: [0x62,0xf3,0x75,0xa9,0x54,0xe2,0x05]
; X64-NEXT:    vxorps %xmm2, %xmm2, %xmm2 # encoding: [0xc5,0xe8,0x57,0xd2]
; X64-NEXT:    vfixupimmps $5, %ymm2, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf3,0x75,0xa9,0x54,0xc2,0x05]
; X64-NEXT:    vaddps %ymm0, %ymm4, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xdc,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.maskz.fixupimm.ps.256(<8 x float> %x0, <8 x float> %x1, <8 x i32> %x2, i32 5, i8 %x4)
  %res1 = call <8 x float> @llvm.x86.avx512.maskz.fixupimm.ps.256(<8 x float> %x0, <8 x float> %x1, <8 x i32> zeroinitializer, i32 5, i8 %x4)
  %res2 = call <8 x float> @llvm.x86.avx512.maskz.fixupimm.ps.256(<8 x float> %x0, <8 x float> %x1, <8 x i32> %x2, i32 5, i8 -1)
  %res3 = fadd <8 x float> %res, %res1
  %res4 = fadd <8 x float> %res3, %res2
  ret <8 x float> %res4
}

define <2 x i64> @test_x86_avx512_psra_q_128(<2 x i64> %a0, <2 x i64> %a1) {
; CHECK-LABEL: test_x86_avx512_psra_q_128:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vpsraq %xmm1, %xmm0, %xmm0 # encoding: [0x62,0xf1,0xfd,0x08,0xe2,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <2 x i64> @llvm.x86.avx512.psra.q.128(<2 x i64> %a0, <2 x i64> %a1) ; <<2 x i64>> [#uses=1]
  ret <2 x i64> %res
}
define <2 x i64> @test_x86_avx512_mask_psra_q_128(<2 x i64> %a0, <2 x i64> %a1, <2 x i64> %passthru, i8 %mask) {
; X86-LABEL: test_x86_avx512_mask_psra_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:    vpsraq %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0xfd,0x09,0xe2,0xd1]
; X86-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_x86_avx512_mask_psra_q_128:
; X64:       # %bb.0:
; 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:    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.psra.q.128(<2 x i64> %a0, <2 x i64> %a1) ; <<2 x i64>> [#uses=1]
  %mask.cast = bitcast i8 %mask to <8 x i1>
  %mask.extract = shufflevector <8 x i1> %mask.cast, <8 x i1> undef, <2 x i32> <i32 0, i32 1>
  %res2 = select <2 x i1> %mask.extract, <2 x i64> %res, <2 x i64> %passthru
  ret <2 x i64> %res2
}
define <2 x i64> @test_x86_avx512_maskz_psra_q_128(<2 x i64> %a0, <2 x i64> %a1, i8 %mask) {
; X86-LABEL: test_x86_avx512_maskz_psra_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:    vpsraq %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0x89,0xe2,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_x86_avx512_maskz_psra_q_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpsraq %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0x89,0xe2,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x i64> @llvm.x86.avx512.psra.q.128(<2 x i64> %a0, <2 x i64> %a1) ; <<2 x i64>> [#uses=1]
  %mask.cast = bitcast i8 %mask to <8 x i1>
  %mask.extract = shufflevector <8 x i1> %mask.cast, <8 x i1> undef, <2 x i32> <i32 0, i32 1>
  %res2 = select <2 x i1> %mask.extract, <2 x i64> %res, <2 x i64> zeroinitializer
  ret <2 x i64> %res2
}
declare <2 x i64> @llvm.x86.avx512.psra.q.128(<2 x i64>, <2 x i64>) nounwind readnone


define <4 x i64> @test_x86_avx512_psra_q_256(<4 x i64> %a0, <2 x i64> %a1) {
; CHECK-LABEL: test_x86_avx512_psra_q_256:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vpsraq %xmm1, %ymm0, %ymm0 # encoding: [0x62,0xf1,0xfd,0x28,0xe2,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <4 x i64> @llvm.x86.avx512.psra.q.256(<4 x i64> %a0, <2 x i64> %a1) ; <<4 x i64>> [#uses=1]
  ret <4 x i64> %res
}
define <4 x i64> @test_x86_avx512_mask_psra_q_256(<4 x i64> %a0, <2 x i64> %a1, <4 x i64> %passthru, i8 %mask) {
; X86-LABEL: test_x86_avx512_mask_psra_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:    vpsraq %xmm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0xfd,0x29,0xe2,0xd1]
; X86-NEXT:    vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_x86_avx512_mask_psra_q_256:
; X64:       # %bb.0:
; 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:    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.psra.q.256(<4 x i64> %a0, <2 x i64> %a1) ; <<4 x i64>> [#uses=1]
  %mask.cast = bitcast i8 %mask to <8 x i1>
  %mask.extract = shufflevector <8 x i1> %mask.cast, <8 x i1> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %res2 = select <4 x i1> %mask.extract, <4 x i64> %res, <4 x i64> %passthru
  ret <4 x i64> %res2
}
define <4 x i64> @test_x86_avx512_maskz_psra_q_256(<4 x i64> %a0, <2 x i64> %a1, <4 x i64> %passthru, i8 %mask) {
; X86-LABEL: test_x86_avx512_maskz_psra_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:    vpsraq %xmm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0xa9,0xe2,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_x86_avx512_maskz_psra_q_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpsraq %xmm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0xa9,0xe2,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i64> @llvm.x86.avx512.psra.q.256(<4 x i64> %a0, <2 x i64> %a1) ; <<4 x i64>> [#uses=1]
  %mask.cast = bitcast i8 %mask to <8 x i1>
  %mask.extract = shufflevector <8 x i1> %mask.cast, <8 x i1> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %res2 = select <4 x i1> %mask.extract, <4 x i64> %res, <4 x i64> zeroinitializer
  ret <4 x i64> %res2
}
declare <4 x i64> @llvm.x86.avx512.psra.q.256(<4 x i64>, <2 x i64>) nounwind readnone


define <2 x i64> @test_x86_avx512_psrai_q_128(<2 x i64> %a0) {
; CHECK-LABEL: test_x86_avx512_psrai_q_128:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vpsraq $7, %xmm0, %xmm0 # encoding: [0x62,0xf1,0xfd,0x08,0x72,0xe0,0x07]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <2 x i64> @llvm.x86.avx512.psrai.q.128(<2 x i64> %a0, i32 7) ; <<2 x i64>> [#uses=1]
  ret <2 x i64> %res
}
define <2 x i64> @test_x86_avx512_mask_psrai_q_128(<2 x i64> %a0, <2 x i64> %passthru, i8 %mask) {
; X86-LABEL: test_x86_avx512_mask_psrai_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:    vpsraq $7, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x09,0x72,0xe0,0x07]
; X86-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_x86_avx512_mask_psrai_q_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpsraq $7, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x09,0x72,0xe0,0x07]
; X64-NEXT:    vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x i64> @llvm.x86.avx512.psrai.q.128(<2 x i64> %a0, i32 7) ; <<2 x i64>> [#uses=1]
  %mask.cast = bitcast i8 %mask to <8 x i1>
  %mask.extract = shufflevector <8 x i1> %mask.cast, <8 x i1> undef, <2 x i32> <i32 0, i32 1>
  %res2 = select <2 x i1> %mask.extract, <2 x i64> %res, <2 x i64> %passthru
  ret <2 x i64> %res2
}
define <2 x i64> @test_x86_avx512_maskz_psrai_q_128(<2 x i64> %a0, i8 %mask) {
; X86-LABEL: test_x86_avx512_maskz_psrai_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:    vpsraq $7, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0x89,0x72,0xe0,0x07]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_x86_avx512_maskz_psrai_q_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpsraq $7, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0x89,0x72,0xe0,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x i64> @llvm.x86.avx512.psrai.q.128(<2 x i64> %a0, i32 7) ; <<2 x i64>> [#uses=1]
  %mask.cast = bitcast i8 %mask to <8 x i1>
  %mask.extract = shufflevector <8 x i1> %mask.cast, <8 x i1> undef, <2 x i32> <i32 0, i32 1>
  %res2 = select <2 x i1> %mask.extract, <2 x i64> %res, <2 x i64> zeroinitializer
  ret <2 x i64> %res2
}
declare <2 x i64> @llvm.x86.avx512.psrai.q.128(<2 x i64>, i32) nounwind readnone


define <4 x i64> @test_x86_avx512_psrai_q_256(<4 x i64> %a0) {
; CHECK-LABEL: test_x86_avx512_psrai_q_256:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vpsraq $7, %ymm0, %ymm0 # encoding: [0x62,0xf1,0xfd,0x28,0x72,0xe0,0x07]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <4 x i64> @llvm.x86.avx512.psrai.q.256(<4 x i64> %a0, i32 7) ; <<4 x i64>> [#uses=1]
  ret <4 x i64> %res
}
define <4 x i64> @test_x86_avx512_mask_psrai_q_256(<4 x i64> %a0, <4 x i64> %passthru, i8 %mask) {
; X86-LABEL: test_x86_avx512_mask_psrai_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:    vpsraq $7, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x29,0x72,0xe0,0x07]
; X86-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_x86_avx512_mask_psrai_q_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpsraq $7, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0xf5,0x29,0x72,0xe0,0x07]
; X64-NEXT:    vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i64> @llvm.x86.avx512.psrai.q.256(<4 x i64> %a0, i32 7) ; <<4 x i64>> [#uses=1]
  %mask.cast = bitcast i8 %mask to <8 x i1>
  %mask.extract = shufflevector <8 x i1> %mask.cast, <8 x i1> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %res2 = select <4 x i1> %mask.extract, <4 x i64> %res, <4 x i64> %passthru
  ret <4 x i64> %res2
}
define <4 x i64> @test_x86_avx512_maskz_psrai_q_256(<4 x i64> %a0, i8 %mask) {
; X86-LABEL: test_x86_avx512_maskz_psrai_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:    vpsraq $7, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0xa9,0x72,0xe0,0x07]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_x86_avx512_maskz_psrai_q_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpsraq $7, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0xa9,0x72,0xe0,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i64> @llvm.x86.avx512.psrai.q.256(<4 x i64> %a0, i32 7) ; <<4 x i64>> [#uses=1]
  %mask.cast = bitcast i8 %mask to <8 x i1>
  %mask.extract = shufflevector <8 x i1> %mask.cast, <8 x i1> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %res2 = select <4 x i1> %mask.extract, <4 x i64> %res, <4 x i64> zeroinitializer
  ret <4 x i64> %res2
}
declare <4 x i64> @llvm.x86.avx512.psrai.q.256(<4 x i64>, i32) nounwind readnone

define <2 x i64> @test_x86_avx512_psrav_q_128(<2 x i64> %a0, <2 x i64> %a1) {
; CHECK-LABEL: test_x86_avx512_psrav_q_128:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vpsravq %xmm1, %xmm0, %xmm0 # encoding: [0x62,0xf2,0xfd,0x08,0x46,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <2 x i64> @llvm.x86.avx512.psrav.q.128(<2 x i64> %a0, <2 x i64> %a1)
  ret <2 x i64> %res
}

define <2 x i64> @test_x86_avx512_mask_psrav_q_128(<2 x i64> %a0, <2 x i64> %a1, <2 x i64> %a2, i8 %mask) {
; X86-LABEL: test_x86_avx512_mask_psrav_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:    vpsravq %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x46,0xd1]
; X86-NEXT:    vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_x86_avx512_mask_psrav_q_128:
; X64:       # %bb.0:
; 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:    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.psrav.q.128(<2 x i64> %a0, <2 x i64> %a1)
  %mask.cast = bitcast i8 %mask to <8 x i1>
  %mask.extract = shufflevector <8 x i1> %mask.cast, <8 x i1> undef, <2 x i32> <i32 0, i32 1>
  %res2 = select <2 x i1> %mask.extract, <2 x i64> %res, <2 x i64> %a2
  ret <2 x i64> %res2
}

define <2 x i64> @test_x86_avx512_maskz_psrav_q_128(<2 x i64> %a0, <2 x i64> %a1, i8 %mask) {
; X86-LABEL: test_x86_avx512_maskz_psrav_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:    vpsravq %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x46,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_x86_avx512_maskz_psrav_q_128:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpsravq %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x46,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <2 x i64> @llvm.x86.avx512.psrav.q.128(<2 x i64> %a0, <2 x i64> %a1)
  %mask.cast = bitcast i8 %mask to <8 x i1>
  %mask.extract = shufflevector <8 x i1> %mask.cast, <8 x i1> undef, <2 x i32> <i32 0, i32 1>
  %res2 = select <2 x i1> %mask.extract, <2 x i64> %res, <2 x i64> zeroinitializer
  ret <2 x i64> %res2
}

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

define <4 x i64> @test_x86_avx512_psrav_q_256(<4 x i64> %a0, <4 x i64> %a1) {
; CHECK-LABEL: test_x86_avx512_psrav_q_256:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vpsravq %ymm1, %ymm0, %ymm0 # encoding: [0x62,0xf2,0xfd,0x28,0x46,0xc1]
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <4 x i64> @llvm.x86.avx512.psrav.q.256(<4 x i64> %a0, <4 x i64> %a1)
  ret <4 x i64> %res
}

define <4 x i64> @test_x86_avx512_mask_psrav_q_256(<4 x i64> %a0, <4 x i64> %a1, <4 x i64> %a2, i8 %mask) {
; X86-LABEL: test_x86_avx512_mask_psrav_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:    vpsravq %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x46,0xd1]
; X86-NEXT:    vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_x86_avx512_mask_psrav_q_256:
; X64:       # %bb.0:
; 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:    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.psrav.q.256(<4 x i64> %a0, <4 x i64> %a1)
  %mask.cast = bitcast i8 %mask to <8 x i1>
  %mask.extract = shufflevector <8 x i1> %mask.cast, <8 x i1> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %res2 = select <4 x i1> %mask.extract, <4 x i64> %res, <4 x i64> %a2
  ret <4 x i64> %res2
}

define <4 x i64> @test_x86_avx512_maskz_psrav_q_256(<4 x i64> %a0, <4 x i64> %a1, i8 %mask) {
; X86-LABEL: test_x86_avx512_maskz_psrav_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:    vpsravq %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x46,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_x86_avx512_maskz_psrav_q_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
; X64-NEXT:    vpsravq %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x46,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <4 x i64> @llvm.x86.avx512.psrav.q.256(<4 x i64> %a0, <4 x i64> %a1)
  %mask.cast = bitcast i8 %mask to <8 x i1>
  %mask.extract = shufflevector <8 x i1> %mask.cast, <8 x i1> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %res2 = select <4 x i1> %mask.extract, <4 x i64> %res, <4 x i64> zeroinitializer
  ret <4 x i64> %res2
}

declare <4 x i64> @llvm.x86.avx512.psrav.q.256(<4 x i64>, <4 x i64>) 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]
  %1 = call <8 x float> @llvm.fma.v8f32(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2)
  ret <8 x float> %1
}

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]
  %1 = call <8 x float> @llvm.fma.v8f32(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2)
  %2 = bitcast i8 %mask to <8 x i1>
  %3 = select <8 x i1> %2, <8 x float> %1, <8 x float> %a0
  ret <8 x float> %3
}

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]
  %1 = call <4 x float> @llvm.fma.v4f32(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2)
  ret <4 x float> %1
}

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]
  %1 = call <4 x float> @llvm.fma.v4f32(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2)
  %2 = bitcast i8 %mask to <8 x i1>
  %extract = 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> %extract, <4 x float> %1, <4 x float> %a0
  ret <4 x float> %3
}

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]
  %1 = call <4 x double> @llvm.fma.v4f64(<4 x double> %a, <4 x double> %b, <4 x double> %c)
  ret <4 x double> %1
}

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]
  %1 = call <4 x double> @llvm.fma.v4f64(<4 x double> %a, <4 x double> %b, <4 x double> %c)
  %2 = bitcast i8 %mask to <8 x i1>
  %extract = 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> %extract, <4 x double> %1, <4 x double> %a
  ret <4 x double> %3
}

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]
  %1 = call <2 x double> @llvm.fma.v2f64(<2 x double> %a, <2 x double> %b, <2 x double> %c)
  ret <2 x double> %1
}

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]
  %1 = call <2 x double> @llvm.fma.v2f64(<2 x double> %a, <2 x double> %b, <2 x double> %c)
  %2 = bitcast i8 %mask to <8 x i1>
  %extract = shufflevector <8 x i1> %2, <8 x i1> %2, <2 x i32> <i32 0, i32 1>
  %3 = select <2 x i1> %extract, <2 x double> %1, <2 x double> %a
  ret <2 x double> %3
}

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]
  %1 = call <2 x double> @llvm.fma.v2f64(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2)
  %2 = bitcast i8 %x3 to <8 x i1>
  %extract = shufflevector <8 x i1> %2, <8 x i1> %2, <2 x i32> <i32 0, i32 1>
  %3 = select <2 x i1> %extract, <2 x double> %1, <2 x double> %x2
  ret <2 x double> %3
}

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]
  %1 = call <2 x double> @llvm.fma.v2f64(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2)
  %2 = bitcast i8 %x3 to <8 x i1>
  %extract = shufflevector <8 x i1> %2, <8 x i1> %2, <2 x i32> <i32 0, i32 1>
  %3 = select <2 x i1> %extract, <2 x double> %1, <2 x double> zeroinitializer
  ret <2 x double> %3
}

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]
  %1 = call <4 x double> @llvm.fma.v4f64(<4 x double> %x0, <4 x double> %x1, <4 x double> %x2)
  %2 = bitcast i8 %x3 to <8 x i1>
  %extract = 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> %extract, <4 x double> %1, <4 x double> %x2
  ret <4 x double> %3
}

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]
  %1 = call <4 x double> @llvm.fma.v4f64(<4 x double> %x0, <4 x double> %x1, <4 x double> %x2)
  %2 = bitcast i8 %x3 to <8 x i1>
  %extract = 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> %extract, <4 x double> %1, <4 x double> zeroinitializer
  ret <4 x double> %3
}

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]
  %1 = call <4 x float> @llvm.fma.v4f32(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2)
  %2 = bitcast i8 %x3 to <8 x i1>
  %extract = 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> %extract, <4 x float> %1, <4 x float> %x2
  ret <4 x float> %3
}

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]
  %1 = call <4 x float> @llvm.fma.v4f32(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2)
  %2 = bitcast i8 %x3 to <8 x i1>
  %extract = 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> %extract, <4 x float> %1, <4 x float> zeroinitializer
  ret <4 x float> %3
}

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]
  %1 = call <8 x float> @llvm.fma.v8f32(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2)
  %2 = bitcast i8 %x3 to <8 x i1>
  %3 = select <8 x i1> %2, <8 x float> %1, <8 x float> %x2
  ret <8 x float> %3
}

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]
  %1 = call <8 x float> @llvm.fma.v8f32(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2)
  %2 = bitcast i8 %x3 to <8 x i1>
  %3 = select <8 x i1> %2, <8 x float> %1, <8 x float> zeroinitializer
  ret <8 x float> %3
}

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]
  %1 = fsub <2 x double> <double -0.000000e+00, double -0.000000e+00>, %x2
  %2 = call <2 x double> @llvm.fma.v2f64(<2 x double> %x0, <2 x double> %x1, <2 x double> %1)
  %3 = bitcast i8 %x3 to <8 x i1>
  %extract = shufflevector <8 x i1> %3, <8 x i1> %3, <2 x i32> <i32 0, i32 1>
  %4 = select <2 x i1> %extract, <2 x double> %2, <2 x double> %x2
  ret <2 x double> %4
}

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]
  %1 = fsub <4 x double> <double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00>, %x2
  %2 = call <4 x double> @llvm.fma.v4f64(<4 x double> %x0, <4 x double> %x1, <4 x double> %1)
  %3 = bitcast i8 %x3 to <8 x i1>
  %extract = shufflevector <8 x i1> %3, <8 x i1> %3, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %4 = select <4 x i1> %extract, <4 x double> %2, <4 x double> %x2
  ret <4 x double> %4
}

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]
  %1 = fsub <4 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, %x2
  %2 = call <4 x float> @llvm.fma.v4f32(<4 x float> %x0, <4 x float> %x1, <4 x float> %1)
  %3 = bitcast i8 %x3 to <8 x i1>
  %extract = shufflevector <8 x i1> %3, <8 x i1> %3, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %4 = select <4 x i1> %extract, <4 x float> %2, <4 x float> %x2
  ret <4 x float> %4
}

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]
  %1 = fsub <8 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, %x2
  %2 = call <8 x float> @llvm.fma.v8f32(<8 x float> %x0, <8 x float> %x1, <8 x float> %1)
  %3 = bitcast i8 %x3 to <8 x i1>
  %4 = select <8 x i1> %3, <8 x float> %2, <8 x float> %x2
  ret <8 x float> %4
}

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]
  %1 = fsub <8 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, %a1
  %2 = call <8 x float> @llvm.fma.v8f32(<8 x float> %a0, <8 x float> %1, <8 x float> %a2)
  ret <8 x float> %2
}

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]
  %1 = fsub <8 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, %a1
  %2 = call <8 x float> @llvm.fma.v8f32(<8 x float> %a0, <8 x float> %1, <8 x float> %a2)
  %3 = bitcast i8 %mask to <8 x i1>
  %4 = select <8 x i1> %3, <8 x float> %2, <8 x float> %a0
  ret <8 x float> %4
}

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]
  %1 = fsub <4 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, %a1
  %2 = call <4 x float> @llvm.fma.v4f32(<4 x float> %a0, <4 x float> %1, <4 x float> %a2)
  ret <4 x float> %2
}

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]
  %1 = fsub <4 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, %a1
  %2 = call <4 x float> @llvm.fma.v4f32(<4 x float> %a0, <4 x float> %1, <4 x float> %a2)
  %3 = bitcast i8 %mask to <8 x i1>
  %extract = shufflevector <8 x i1> %3, <8 x i1> %3, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %4 = select <4 x i1> %extract, <4 x float> %2, <4 x float> %a0
  ret <4 x float> %4
}

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]
  %1 = fsub <4 x double> <double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00>, %a1
  %2 = call <4 x double> @llvm.fma.v4f64(<4 x double> %a0, <4 x double> %1, <4 x double> %a2)
  ret <4 x double> %2
}

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]
  %1 = fsub <4 x double> <double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00>, %a1
  %2 = call <4 x double> @llvm.fma.v4f64(<4 x double> %a0, <4 x double> %1, <4 x double> %a2)
  %3 = bitcast i8 %mask to <8 x i1>
  %extract = shufflevector <8 x i1> %3, <8 x i1> %3, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %4 = select <4 x i1> %extract, <4 x double> %2, <4 x double> %a0
  ret <4 x double> %4
}

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]
  %1 = fsub <2 x double> <double -0.000000e+00, double -0.000000e+00>, %a1
  %2 = call <2 x double> @llvm.fma.v2f64(<2 x double> %a0, <2 x double> %1, <2 x double> %a2)
  ret <2 x double> %2
}

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]
  %1 = fsub <2 x double> <double -0.000000e+00, double -0.000000e+00>, %a1
  %2 = call <2 x double> @llvm.fma.v2f64(<2 x double> %a0, <2 x double> %1, <2 x double> %a2)
  %3 = bitcast i8 %mask to <8 x i1>
  %extract = shufflevector <8 x i1> %3, <8 x i1> %3, <2 x i32> <i32 0, i32 1>
  %4 = select <2 x i1> %extract, <2 x double> %2, <2 x double> %a0
  ret <2 x double> %4
}

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]
  %1 = fsub <8 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, %a1
  %2 = fsub <8 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, %a2
  %3 = call <8 x float> @llvm.fma.v8f32(<8 x float> %a0, <8 x float> %1, <8 x float> %2)
  ret <8 x float> %3
}

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]
  %1 = fsub <8 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, %a1
  %2 = fsub <8 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, %a2
  %3 = call <8 x float> @llvm.fma.v8f32(<8 x float> %a0, <8 x float> %1, <8 x float> %2)
  %4 = bitcast i8 %mask to <8 x i1>
  %5 = select <8 x i1> %4, <8 x float> %3, <8 x float> %a0
  ret <8 x float> %5
}

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]
  %1 = fsub <4 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, %a1
  %2 = fsub <4 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, %a2
  %3 = call <4 x float> @llvm.fma.v4f32(<4 x float> %a0, <4 x float> %1, <4 x float> %2)
  ret <4 x float> %3
}

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]
  %1 = fsub <4 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, %a1
  %2 = fsub <4 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, %a2
  %3 = call <4 x float> @llvm.fma.v4f32(<4 x float> %a0, <4 x float> %1, <4 x float> %2)
  %4 = bitcast i8 %mask to <8 x i1>
  %extract = shufflevector <8 x i1> %4, <8 x i1> %4, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %5 = select <4 x i1> %extract, <4 x float> %3, <4 x float> %a0
  ret <4 x float> %5
}

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]
  %1 = fsub <4 x double> <double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00>, %a1
  %2 = fsub <4 x double> <double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00>, %a2
  %3 = call <4 x double> @llvm.fma.v4f64(<4 x double> %a0, <4 x double> %1, <4 x double> %2)
  ret <4 x double> %3
}

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]
  %1 = fsub <4 x double> <double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00>, %a1
  %2 = fsub <4 x double> <double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00>, %a2
  %3 = call <4 x double> @llvm.fma.v4f64(<4 x double> %a0, <4 x double> %1, <4 x double> %2)
  %4 = bitcast i8 %mask to <8 x i1>
  %extract = shufflevector <8 x i1> %4, <8 x i1> %4, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %5 = select <4 x i1> %extract, <4 x double> %3, <4 x double> %a0
  ret <4 x double> %5
}

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]
  %1 = fsub <2 x double> <double -0.000000e+00, double -0.000000e+00>, %a1
  %2 = fsub <2 x double> <double -0.000000e+00, double -0.000000e+00>, %a2
  %3 = call <2 x double> @llvm.fma.v2f64(<2 x double> %a0, <2 x double> %1, <2 x double> %2)
  ret <2 x double> %3
}

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]
  %1 = fsub <2 x double> <double -0.000000e+00, double -0.000000e+00>, %a1
  %2 = fsub <2 x double> <double -0.000000e+00, double -0.000000e+00>, %a2
  %3 = call <2 x double> @llvm.fma.v2f64(<2 x double> %a0, <2 x double> %1, <2 x double> %2)
  %4 = bitcast i8 %mask to <8 x i1>
  %extract = shufflevector <8 x i1> %4, <8 x i1> %4, <2 x i32> <i32 0, i32 1>
  %5 = select <2 x i1> %extract, <2 x double> %3, <2 x double> %a0
  ret <2 x double> %5
}

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]
  %1 = fsub <2 x double> <double -0.000000e+00, double -0.000000e+00>, %x0
  %2 = fsub <2 x double> <double -0.000000e+00, double -0.000000e+00>, %x2
  %3 = call <2 x double> @llvm.fma.v2f64(<2 x double> %1, <2 x double> %x1, <2 x double> %2)
  %4 = bitcast i8 %x3 to <8 x i1>
  %extract = shufflevector <8 x i1> %4, <8 x i1> %4, <2 x i32> <i32 0, i32 1>
  %5 = select <2 x i1> %extract, <2 x double> %3, <2 x double> %x2
  ret <2 x double> %5
}

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]
  %1 = fsub <4 x double> <double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00>, %x0
  %2 = fsub <4 x double> <double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00>, %x2
  %3 = call <4 x double> @llvm.fma.v4f64(<4 x double> %1, <4 x double> %x1, <4 x double> %2)
  %4 = bitcast i8 %x3 to <8 x i1>
  %extract = shufflevector <8 x i1> %4, <8 x i1> %4, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %5 = select <4 x i1> %extract, <4 x double> %3, <4 x double> %x2
  ret <4 x double> %5
}

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]
  %1 = fsub <4 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, %x0
  %2 = fsub <4 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, %x2
  %3 = call <4 x float> @llvm.fma.v4f32(<4 x float> %1, <4 x float> %x1, <4 x float> %2)
  %4 = bitcast i8 %x3 to <8 x i1>
  %extract = shufflevector <8 x i1> %4, <8 x i1> %4, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  %5 = select <4 x i1> %extract, <4 x float> %3, <4 x float> %x2
  ret <4 x float> %5
}

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]
  %1 = fsub <8 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, %x0
  %2 = fsub <8 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, %x2
  %3 = call <8 x float> @llvm.fma.v8f32(<8 x float> %1, <8 x float> %x1, <8 x float> %2)
  %4 = bitcast i8 %x3 to <8 x i1>
  %5 = select <8 x i1> %4, <8 x float> %3, <8 x float> %x2
  ret <8 x float> %5
}

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]
  %1 = call <8 x float> @llvm.fma.v8f32(<8 x float> %a, <8 x float> %b, <8 x float> %c)
  %2 = fsub <8 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, %c
  %3 = call <8 x float> @llvm.fma.v8f32(<8 x float> %a, <8 x float> %b, <8 x float> %2)
  %4 = shufflevector <8 x float> %3, <8 x float> %1, <8 x i32> <i32 0, i32 9, i32 2, i32 11, i32 4, i32 13, i32 6, i32 15>
  ret <8 x float> %4
}

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]
  %1 = call <8 x float> @llvm.fma.v8f32(<8 x float> %a, <8 x float> %b, <8 x float> %c)
  %2 = fsub <8 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, %c
  %3 = call <8 x float> @llvm.fma.v8f32(<8 x float> %a, <8 x float> %b, <8 x float> %2)
  %4 = shufflevector <8 x float> %3, <8 x float> %1, <8 x i32> <i32 0, i32 9, i32 2, i32 11, i32 4, i32 13, i32 6, i32 15>
  %5 = bitcast i8 %mask to <8 x i1>
  %6 = select <8 x i1> %5, <8 x float> %4, <8 x float> %a
  ret <8 x float> %6
}

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]
  %1 = call <4 x float> @llvm.fma.v4f32(<4 x float> %a, <4 x float> %b, <4 x float> %c)
  %2 = fsub <4 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, %c
  %3 = call <4 x float> @llvm.fma.v4f32(<4 x float> %a, <4 x float> %b, <4 x float> %2)
  %4 = shufflevector <4 x float> %3, <4 x float> %1, <4 x i32> <i32 0, i32 5, i32 2, i32 7>
  ret <4 x float> %4
}

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]
  %1 = call <4 x float> @llvm.fma.v4f32(<4 x float> %a, <4 x float> %b, <4 x float> %c)
  %2 = fsub <4 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, %c
  %3 = call <4 x float> @llvm.fma.v4f32(<4 x float> %a, <4 x float> %b, <4 x float> %2)
  %4 = shufflevector <4 x float> %3, <4 x float> %1, <4 x i32> <i32 0, i32 5, i32 2, i32 7>
  %5 = bitcast i8 %mask 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 float> %4, <4 x float> %a
  ret <4 x float> %6
}

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]
  %1 = call <4 x double> @llvm.fma.v4f64(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2)
  %2 = fsub <4 x double> <double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00>, %a2
  %3 = call <4 x double> @llvm.fma.v4f64(<4 x double> %a0, <4 x double> %a1, <4 x double> %2)
  %4 = shufflevector <4 x double> %3, <4 x double> %1, <4 x i32> <i32 0, i32 5, i32 2, i32 7>
  ret <4 x double> %4
}

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]
  %1 = call <4 x double> @llvm.fma.v4f64(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2)
  %2 = fsub <4 x double> <double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00>, %a2
  %3 = call <4 x double> @llvm.fma.v4f64(<4 x double> %a0, <4 x double> %a1, <4 x double> %2)
  %4 = shufflevector <4 x double> %3, <4 x double> %1, <4 x i32> <i32 0, i32 5, i32 2, i32 7>
  %5 = bitcast i8 %mask 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 double> %4, <4 x double> %a0
  ret <4 x double> %6
}

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]
  %1 = call <2 x double> @llvm.fma.v2f64(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2)
  %2 = fsub <2 x double> <double -0.000000e+00, double -0.000000e+00>, %a2
  %3 = call <2 x double> @llvm.fma.v2f64(<2 x double> %a0, <2 x double> %a1, <2 x double> %2)
  %4 = shufflevector <2 x double> %3, <2 x double> %1, <2 x i32> <i32 0, i32 3>
  ret <2 x double> %4
}

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]
  %1 = call <2 x double> @llvm.fma.v2f64(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2)
  %2 = fsub <2 x double> <double -0.000000e+00, double -0.000000e+00>, %a2
  %3 = call <2 x double> @llvm.fma.v2f64(<2 x double> %a0, <2 x double> %a1, <2 x double> %2)
  %4 = shufflevector <2 x double> %3, <2 x double> %1, <2 x i32> <i32 0, i32 3>
  %5 = bitcast i8 %mask 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 double> %4, <2 x double> %a0
  ret <2 x double> %6
}

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]
  %1 = call <2 x double> @llvm.fma.v2f64(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2)
  %2 = fsub <2 x double> <double -0.000000e+00, double -0.000000e+00>, %x2
  %3 = call <2 x double> @llvm.fma.v2f64(<2 x double> %x0, <2 x double> %x1, <2 x double> %2)
  %4 = shufflevector <2 x double> %3, <2 x double> %1, <2 x i32> <i32 0, 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 double> %4, <2 x double> %x2
  ret <2 x double> %6
}

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]
  %1 = call <2 x double> @llvm.fma.v2f64(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2)
  %2 = fsub <2 x double> <double -0.000000e+00, double -0.000000e+00>, %x2
  %3 = call <2 x double> @llvm.fma.v2f64(<2 x double> %x0, <2 x double> %x1, <2 x double> %2)
  %4 = shufflevector <2 x double> %3, <2 x double> %1, <2 x i32> <i32 0, 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 double> %4, <2 x double> zeroinitializer
  ret <2 x double> %6
}

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]
  %1 = call <4 x double> @llvm.fma.v4f64(<4 x double> %x0, <4 x double> %x1, <4 x double> %x2)
  %2 = fsub <4 x double> <double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00>, %x2
  %3 = call <4 x double> @llvm.fma.v4f64(<4 x double> %x0, <4 x double> %x1, <4 x double> %2)
  %4 = shufflevector <4 x double> %3, <4 x double> %1, <4 x i32> <i32 0, i32 5, i32 2, i32 7>
  %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 double> %4, <4 x double> %x2
  ret <4 x double> %6
}

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]
  %1 = call <4 x double> @llvm.fma.v4f64(<4 x double> %x0, <4 x double> %x1, <4 x double> %x2)
  %2 = fsub <4 x double> <double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00>, %x2
  %3 = call <4 x double> @llvm.fma.v4f64(<4 x double> %x0, <4 x double> %x1, <4 x double> %2)
  %4 = shufflevector <4 x double> %3, <4 x double> %1, <4 x i32> <i32 0, i32 5, i32 2, i32 7>
  %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 double> %4, <4 x double> zeroinitializer
  ret <4 x double> %6
}

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]
  %1 = call <4 x float> @llvm.fma.v4f32(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2)
  %2 = fsub <4 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, %x2
  %3 = call <4 x float> @llvm.fma.v4f32(<4 x float> %x0, <4 x float> %x1, <4 x float> %2)
  %4 = shufflevector <4 x float> %3, <4 x float> %1, <4 x i32> <i32 0, i32 5, i32 2, i32 7>
  %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 float> %4, <4 x float> %x2
  ret <4 x float> %6
}

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]
  %1 = call <4 x float> @llvm.fma.v4f32(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2)
  %2 = fsub <4 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, %x2
  %3 = call <4 x float> @llvm.fma.v4f32(<4 x float> %x0, <4 x float> %x1, <4 x float> %2)
  %4 = shufflevector <4 x float> %3, <4 x float> %1, <4 x i32> <i32 0, i32 5, i32 2, i32 7>
  %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 float> %4, <4 x float> zeroinitializer
  ret <4 x float> %6
}

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]
  %1 = call <8 x float> @llvm.fma.v8f32(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2)
  %2 = fsub <8 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, %x2
  %3 = call <8 x float> @llvm.fma.v8f32(<8 x float> %x0, <8 x float> %x1, <8 x float> %2)
  %4 = shufflevector <8 x float> %3, <8 x float> %1, <8 x i32> <i32 0, i32 9, i32 2, i32 11, i32 4, i32 13, i32 6, i32 15>
  %5 = bitcast i8 %x3 to <8 x i1>
  %6 = select <8 x i1> %5, <8 x float> %4, <8 x float> %x2
  ret <8 x float> %6
}

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]
  %1 = call <8 x float> @llvm.fma.v8f32(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2)
  %2 = fsub <8 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, %x2
  %3 = call <8 x float> @llvm.fma.v8f32(<8 x float> %x0, <8 x float> %x1, <8 x float> %2)
  %4 = shufflevector <8 x float> %3, <8 x float> %1, <8 x i32> <i32 0, i32 9, i32 2, i32 11, i32 4, i32 13, i32 6, i32 15>
  %5 = bitcast i8 %x3 to <8 x i1>
  %6 = select <8 x i1> %5, <8 x float> %4, <8 x float> zeroinitializer
  ret <8 x float> %6
}

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]
  %1 = call <2 x double> @llvm.fma.v2f64(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2)
  %2 = fsub <2 x double> <double -0.000000e+00, double -0.000000e+00>, %x2
  %3 = call <2 x double> @llvm.fma.v2f64(<2 x double> %x0, <2 x double> %x1, <2 x double> %2)
  %4 = shufflevector <2 x double> %1, <2 x double> %3, <2 x i32> <i32 0, 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 double> %4, <2 x double> %x2
  ret <2 x double> %6
}

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]
  %1 = call <4 x double> @llvm.fma.v4f64(<4 x double> %x0, <4 x double> %x1, <4 x double> %x2)
  %2 = fsub <4 x double> <double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00>, %x2
  %3 = call <4 x double> @llvm.fma.v4f64(<4 x double> %x0, <4 x double> %x1, <4 x double> %2)
  %4 = shufflevector <4 x double> %1, <4 x double> %3, <4 x i32> <i32 0, i32 5, i32 2, i32 7>
  %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 double> %4, <4 x double> %x2
  ret <4 x double> %6
}

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]
  %1 = call <4 x float> @llvm.fma.v4f32(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2)
  %2 = fsub <4 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, %x2
  %3 = call <4 x float> @llvm.fma.v4f32(<4 x float> %x0, <4 x float> %x1, <4 x float> %2)
  %4 = shufflevector <4 x float> %1, <4 x float> %3, <4 x i32> <i32 0, i32 5, i32 2, i32 7>
  %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 float> %4, <4 x float> %x2
  ret <4 x float> %6
}

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]
  %1 = call <8 x float> @llvm.fma.v8f32(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2)
  %2 = fsub <8 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, %x2
  %3 = call <8 x float> @llvm.fma.v8f32(<8 x float> %x0, <8 x float> %x1, <8 x float> %2)
  %4 = shufflevector <8 x float> %1, <8 x float> %3, <8 x i32> <i32 0, i32 9, i32 2, i32 11, i32 4, i32 13, i32 6, i32 15>
  %5 = bitcast i8 %x3 to <8 x i1>
  %6 = select <8 x i1> %5, <8 x float> %4, <8 x float> %x2
  ret <8 x float> %6
}

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
  %1 = call <4 x float> @llvm.fma.v4f32(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2)
  %2 = bitcast i8 %mask to <8 x i1>
  %extract = 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> %extract, <4 x float> %1, <4 x float> %a0
  ret <4 x float> %3
}

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
  %1 = call <4 x float> @llvm.fma.v4f32(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2)
  %2 = bitcast i8 %mask to <8 x i1>
  %extract = 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> %extract, <4 x float> %1, <4 x float> %a0
  ret <4 x float> %3
}

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
  %1 = call <4 x float> @llvm.fma.v4f32(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2)
  ret <4 x float> %1
}

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
  %1 = call <4 x float> @llvm.fma.v4f32(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2)
  ret <4 x float> %1
}

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
  %1 = call <4 x float> @llvm.fma.v4f32(<4 x float> %a0, <4 x float> %a1, <4 x float> %vecinit6.i)
  %2 = bitcast i8 %mask to <8 x i1>
  %extract = 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> %extract, <4 x float> %1, <4 x float> %a0
  ret <4 x float> %3
}

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
  %1 = call <4 x float> @llvm.fma.v4f32(<4 x float> %a0, <4 x float> %a1, <4 x float> %vecinit6.i)
  %2 = bitcast i8 %mask to <8 x i1>
  %extract = 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> %extract, <4 x float> %1, <4 x float> %a0
  ret <4 x float> %3
}

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
  %1 = call <4 x float> @llvm.fma.v4f32(<4 x float> %a0, <4 x float> %a1, <4 x float> %vecinit6.i)
  ret <4 x float> %1
}

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
  %1 = call <4 x float> @llvm.fma.v4f32(<4 x float> %a0, <4 x float> %a1, <4 x float> %vecinit6.i)
  ret <4 x float> %1
}

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
  %1 = call <2 x double> @llvm.fma.v2f64(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2)
  %2 = bitcast i8 %mask to <8 x i1>
  %extract = shufflevector <8 x i1> %2, <8 x i1> %2, <2 x i32> <i32 0, i32 1>
  %3 = select <2 x i1> %extract, <2 x double> %1, <2 x double> %a0
  ret <2 x double> %3
}

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
  %1 = call <2 x double> @llvm.fma.v2f64(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2)
  ret <2 x double> %1
}

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
  %1 = call <4 x double> @llvm.fma.v4f64(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2)
  %2 = bitcast i8 %mask to <8 x i1>
  %extract = 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> %extract, <4 x double> %1, <4 x double> %a0
  ret <4 x double> %3
}

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
  %1 = call <4 x double> @llvm.fma.v4f64(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2)
  ret <4 x double> %1
}

declare <8 x float> @llvm.fma.v8f32(<8 x float>, <8 x float>, <8 x float>)
declare <4 x float> @llvm.fma.v4f32(<4 x float>, <4 x float>, <4 x float>)
declare <4 x double> @llvm.fma.v4f64(<4 x double>, <4 x double>, <4 x double>)
declare <2 x double> @llvm.fma.v2f64(<2 x double>, <2 x double>, <2 x double>)
declare <2 x double> @llvm.x86.avx512.mask.compress.v2f64(<2 x double>, <2 x double>, <2 x i1>)
declare <4 x float> @llvm.x86.avx512.mask.compress.v4f32(<4 x float>, <4 x float>, <4 x i1>)
declare <2 x i64> @llvm.x86.avx512.mask.compress.v2i64(<2 x i64>, <2 x i64>, <2 x i1>)
declare <4 x i32> @llvm.x86.avx512.mask.compress.v4i32(<4 x i32>, <4 x i32>, <4 x i1>)
declare <2 x double> @llvm.x86.avx512.mask.expand.v2f64(<2 x double>, <2 x double>, <2 x i1>)
declare <4 x float> @llvm.x86.avx512.mask.expand.v4f32(<4 x float>, <4 x float>, <4 x i1>)
declare <2 x i64> @llvm.x86.avx512.mask.expand.v2i64(<2 x i64>, <2 x i64>, <2 x i1>)
declare <4 x i32> @llvm.x86.avx512.mask.expand.v4i32(<4 x i32>, <4 x i32>, <4 x i1>)
declare <4 x double> @llvm.x86.avx512.mask.compress.v4f64(<4 x double>, <4 x double>, <4 x i1>)
declare <8 x float> @llvm.x86.avx512.mask.compress.v8f32(<8 x float>, <8 x float>, <8 x i1>)
declare <4 x i64> @llvm.x86.avx512.mask.compress.v4i64(<4 x i64>, <4 x i64>, <4 x i1>)
declare <8 x i32> @llvm.x86.avx512.mask.compress.v8i32(<8 x i32>, <8 x i32>, <8 x i1>)
declare <4 x double> @llvm.x86.avx512.mask.expand.v4f64(<4 x double>, <4 x double>, <4 x i1>)
declare <8 x float> @llvm.x86.avx512.mask.expand.v8f32(<8 x float>, <8 x float>, <8 x i1>)
declare <4 x i64> @llvm.x86.avx512.mask.expand.v4i64(<4 x i64>, <4 x i64>, <4 x i1>)
declare <8 x i32> @llvm.x86.avx512.mask.expand.v8i32(<8 x i32>, <8 x i32>, <8 x i1>)