; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -disable-peephole -mtriple=x86_64-unknown-unknown -mattr=+avx512f | FileCheck %s --check-prefixes=CHECK,X64
; RUN: llc < %s -disable-peephole -mtriple=i686-unknown-unknown -mattr=+avx512f | FileCheck %s --check-prefixes=CHECK,X86


define <8 x double> @test_mask_compress_pd_512(<8 x double> %data, <8 x double> %passthru, i8 %mask) {
; X64-LABEL: test_mask_compress_pd_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vcompresspd %zmm0, %zmm1 {%k1}
; X64-NEXT:    vmovdqa64 %zmm1, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_mask_compress_pd_512:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vcompresspd %zmm0, %zmm1 {%k1}
; X86-NEXT:    vmovdqa64 %zmm1, %zmm0
; X86-NEXT:    retl
  %1 = bitcast i8 %mask to <8 x i1>
  %2 = call <8 x double> @llvm.x86.avx512.mask.compress.v8f64(<8 x double> %data, <8 x double> %passthru, <8 x i1> %1)
  ret <8 x double> %2
}

define <8 x double> @test_maskz_compress_pd_512(<8 x double> %data, i8 %mask) {
; X64-LABEL: test_maskz_compress_pd_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vcompresspd %zmm0, %zmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_maskz_compress_pd_512:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vcompresspd %zmm0, %zmm0 {%k1} {z}
; X86-NEXT:    retl
  %1 = bitcast i8 %mask to <8 x i1>
  %2 = call <8 x double> @llvm.x86.avx512.mask.compress.v8f64(<8 x double> %data, <8 x double> zeroinitializer, <8 x i1> %1)
  ret <8 x double> %2
}

define <8 x double> @test_compress_pd_512(<8 x double> %data) {
; CHECK-LABEL: test_compress_pd_512:
; CHECK:       # %bb.0:
; CHECK-NEXT:    ret{{[l|q]}}
  %1 = call <8 x double> @llvm.x86.avx512.mask.compress.v8f64(<8 x double> %data, <8 x double> undef, <8 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>)
  ret <8 x double> %1
}

define <16 x float> @test_mask_compress_ps_512(<16 x float> %data, <16 x float> %passthru, i16 %mask) {
; X64-LABEL: test_mask_compress_ps_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vcompressps %zmm0, %zmm1 {%k1}
; X64-NEXT:    vmovdqa64 %zmm1, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_mask_compress_ps_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vcompressps %zmm0, %zmm1 {%k1}
; X86-NEXT:    vmovdqa64 %zmm1, %zmm0
; X86-NEXT:    retl
  %1 = bitcast i16 %mask to <16 x i1>
  %2 = call <16 x float> @llvm.x86.avx512.mask.compress.v16f32(<16 x float> %data, <16 x float> %passthru, <16 x i1> %1)
  ret <16 x float> %2
}

define <16 x float> @test_maskz_compress_ps_512(<16 x float> %data, i16 %mask) {
; X64-LABEL: test_maskz_compress_ps_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vcompressps %zmm0, %zmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_maskz_compress_ps_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vcompressps %zmm0, %zmm0 {%k1} {z}
; X86-NEXT:    retl
  %1 = bitcast i16 %mask to <16 x i1>
  %2 = call <16 x float> @llvm.x86.avx512.mask.compress.v16f32(<16 x float> %data, <16 x float> zeroinitializer, <16 x i1> %1)
  ret <16 x float> %2
}

define <16 x float> @test_compress_ps_512(<16 x float> %data) {
; CHECK-LABEL: test_compress_ps_512:
; CHECK:       # %bb.0:
; CHECK-NEXT:    ret{{[l|q]}}
  %1 = call <16 x float> @llvm.x86.avx512.mask.compress.v16f32(<16 x float> %data, <16 x float> undef, <16 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>)
  ret <16 x float> %1
}

define <8 x i64> @test_mask_compress_q_512(<8 x i64> %data, <8 x i64> %passthru, i8 %mask) {
; X64-LABEL: test_mask_compress_q_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpcompressq %zmm0, %zmm1 {%k1}
; X64-NEXT:    vmovdqa64 %zmm1, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_mask_compress_q_512:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vpcompressq %zmm0, %zmm1 {%k1}
; X86-NEXT:    vmovdqa64 %zmm1, %zmm0
; X86-NEXT:    retl
  %1 = bitcast i8 %mask to <8 x i1>
  %2 = call <8 x i64> @llvm.x86.avx512.mask.compress.v8i64(<8 x i64> %data, <8 x i64> %passthru, <8 x i1> %1)
  ret <8 x i64> %2
}

define <8 x i64> @test_maskz_compress_q_512(<8 x i64> %data, i8 %mask) {
; X64-LABEL: test_maskz_compress_q_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpcompressq %zmm0, %zmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_maskz_compress_q_512:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vpcompressq %zmm0, %zmm0 {%k1} {z}
; X86-NEXT:    retl
  %1 = bitcast i8 %mask to <8 x i1>
  %2 = call <8 x i64> @llvm.x86.avx512.mask.compress.v8i64(<8 x i64> %data, <8 x i64> zeroinitializer, <8 x i1> %1)
  ret <8 x i64> %2
}

define <8 x i64> @test_compress_q_512(<8 x i64> %data) {
; CHECK-LABEL: test_compress_q_512:
; CHECK:       # %bb.0:
; CHECK-NEXT:    ret{{[l|q]}}
  %1 = call <8 x i64> @llvm.x86.avx512.mask.compress.v8i64(<8 x i64> %data, <8 x i64> undef, <8 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>)
  ret <8 x i64> %1
}

define <16 x i32> @test_mask_compress_d_512(<16 x i32> %data, <16 x i32> %passthru, i16 %mask) {
; X64-LABEL: test_mask_compress_d_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpcompressd %zmm0, %zmm1 {%k1}
; X64-NEXT:    vmovdqa64 %zmm1, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_mask_compress_d_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vpcompressd %zmm0, %zmm1 {%k1}
; X86-NEXT:    vmovdqa64 %zmm1, %zmm0
; X86-NEXT:    retl
  %1 = bitcast i16 %mask to <16 x i1>
  %2 = call <16 x i32> @llvm.x86.avx512.mask.compress.v16i32(<16 x i32> %data, <16 x i32> %passthru, <16 x i1> %1)
  ret <16 x i32> %2
}

define <16 x i32> @test_maskz_compress_d_512(<16 x i32> %data, i16 %mask) {
; X64-LABEL: test_maskz_compress_d_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpcompressd %zmm0, %zmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_maskz_compress_d_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vpcompressd %zmm0, %zmm0 {%k1} {z}
; X86-NEXT:    retl
  %1 = bitcast i16 %mask to <16 x i1>
  %2 = call <16 x i32> @llvm.x86.avx512.mask.compress.v16i32(<16 x i32> %data, <16 x i32> zeroinitializer, <16 x i1> %1)
  ret <16 x i32> %2
}

define <16 x i32> @test_compress_d_512(<16 x i32> %data) {
; CHECK-LABEL: test_compress_d_512:
; CHECK:       # %bb.0:
; CHECK-NEXT:    ret{{[l|q]}}
  %1 = call <16 x i32> @llvm.x86.avx512.mask.compress.v16i32(<16 x i32> %data, <16 x i32> undef, <16 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>)
  ret <16 x i32> %1
}

define <8 x double> @test_expand_pd_512(<8 x double> %data) {
; CHECK-LABEL: test_expand_pd_512:
; CHECK:       # %bb.0:
; CHECK-NEXT:    ret{{[l|q]}}
  %1 = call <8 x double> @llvm.x86.avx512.mask.expand.v8f64(<8 x double> %data, <8 x double> undef, <8 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>)
  ret <8 x double> %1
}

define <8 x double> @test_mask_expand_pd_512(<8 x double> %data, <8 x double> %passthru, i8 %mask) {
; X64-LABEL: test_mask_expand_pd_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vexpandpd %zmm0, %zmm1 {%k1}
; X64-NEXT:    vmovdqa64 %zmm1, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_mask_expand_pd_512:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vexpandpd %zmm0, %zmm1 {%k1}
; X86-NEXT:    vmovdqa64 %zmm1, %zmm0
; X86-NEXT:    retl
  %1 = bitcast i8 %mask to <8 x i1>
  %2 = call <8 x double> @llvm.x86.avx512.mask.expand.v8f64(<8 x double> %data, <8 x double> %passthru, <8 x i1> %1)
  ret <8 x double> %2
}

define <8 x double> @test_maskz_expand_pd_512(<8 x double> %data, i8 %mask) {
; X64-LABEL: test_maskz_expand_pd_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vexpandpd %zmm0, %zmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_maskz_expand_pd_512:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vexpandpd %zmm0, %zmm0 {%k1} {z}
; X86-NEXT:    retl
  %1 = bitcast i8 %mask to <8 x i1>
  %2 = call <8 x double> @llvm.x86.avx512.mask.expand.v8f64(<8 x double> %data, <8 x double> zeroinitializer, <8 x i1> %1)
  ret <8 x double> %2
}

define <16 x float> @test_expand_ps_512(<16 x float> %data) {
; CHECK-LABEL: test_expand_ps_512:
; CHECK:       # %bb.0:
; CHECK-NEXT:    ret{{[l|q]}}
  %1 = call <16 x float> @llvm.x86.avx512.mask.expand.v16f32(<16 x float> %data, <16 x float> undef, <16 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>)
  ret <16 x float> %1
}

define <16 x float> @test_mask_expand_ps_512(<16 x float> %data, <16 x float> %passthru, i16 %mask) {
; X64-LABEL: test_mask_expand_ps_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vexpandps %zmm0, %zmm1 {%k1}
; X64-NEXT:    vmovdqa64 %zmm1, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_mask_expand_ps_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vexpandps %zmm0, %zmm1 {%k1}
; X86-NEXT:    vmovdqa64 %zmm1, %zmm0
; X86-NEXT:    retl
  %1 = bitcast i16 %mask to <16 x i1>
  %2 = call <16 x float> @llvm.x86.avx512.mask.expand.v16f32(<16 x float> %data, <16 x float> %passthru, <16 x i1> %1)
  ret <16 x float> %2
}

define <16 x float> @test_maskz_expand_ps_512(<16 x float> %data, i16 %mask) {
; X64-LABEL: test_maskz_expand_ps_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vexpandps %zmm0, %zmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_maskz_expand_ps_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vexpandps %zmm0, %zmm0 {%k1} {z}
; X86-NEXT:    retl
  %1 = bitcast i16 %mask to <16 x i1>
  %2 = call <16 x float> @llvm.x86.avx512.mask.expand.v16f32(<16 x float> %data, <16 x float> zeroinitializer, <16 x i1> %1)
  ret <16 x float> %2
}

define <8 x i64> @test_expand_q_512(<8 x i64> %data) {
; CHECK-LABEL: test_expand_q_512:
; CHECK:       # %bb.0:
; CHECK-NEXT:    ret{{[l|q]}}
  %1 = call <8 x i64> @llvm.x86.avx512.mask.expand.v8i64(<8 x i64> %data, <8 x i64> undef, <8 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>)
  ret <8 x i64> %1
}

define <8 x i64> @test_mask_expand_q_512(<8 x i64> %data, <8 x i64> %passthru, i8 %mask) {
; X64-LABEL: test_mask_expand_q_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpexpandq %zmm0, %zmm1 {%k1}
; X64-NEXT:    vmovdqa64 %zmm1, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_mask_expand_q_512:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vpexpandq %zmm0, %zmm1 {%k1}
; X86-NEXT:    vmovdqa64 %zmm1, %zmm0
; X86-NEXT:    retl
  %1 = bitcast i8 %mask to <8 x i1>
  %2 = call <8 x i64> @llvm.x86.avx512.mask.expand.v8i64(<8 x i64> %data, <8 x i64> %passthru, <8 x i1> %1)
  ret <8 x i64> %2
}

define <8 x i64> @test_maskz_expand_q_512(<8 x i64> %data, i8 %mask) {
; X64-LABEL: test_maskz_expand_q_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpexpandq %zmm0, %zmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_maskz_expand_q_512:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vpexpandq %zmm0, %zmm0 {%k1} {z}
; X86-NEXT:    retl
  %1 = bitcast i8 %mask to <8 x i1>
  %2 = call <8 x i64> @llvm.x86.avx512.mask.expand.v8i64(<8 x i64> %data, <8 x i64> zeroinitializer, <8 x i1> %1)
  ret <8 x i64> %2
}

define <16 x i32> @test_expand_d_512(<16 x i32> %data) {
; CHECK-LABEL: test_expand_d_512:
; CHECK:       # %bb.0:
; CHECK-NEXT:    ret{{[l|q]}}
  %1 = call <16 x i32> @llvm.x86.avx512.mask.expand.v16i32(<16 x i32> %data, <16 x i32> undef, <16 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>)
  ret <16 x i32> %1
}

define <16 x i32> @test_mask_expand_d_512(<16 x i32> %data, <16 x i32> %passthru, i16 %mask) {
; X64-LABEL: test_mask_expand_d_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpexpandd %zmm0, %zmm1 {%k1}
; X64-NEXT:    vmovdqa64 %zmm1, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_mask_expand_d_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vpexpandd %zmm0, %zmm1 {%k1}
; X86-NEXT:    vmovdqa64 %zmm1, %zmm0
; X86-NEXT:    retl
  %1 = bitcast i16 %mask to <16 x i1>
  %2 = call <16 x i32> @llvm.x86.avx512.mask.expand.v16i32(<16 x i32> %data, <16 x i32> %passthru, <16 x i1> %1)
  ret <16 x i32> %2
}

define <16 x i32> @test_maskz_expand_d_512(<16 x i32> %data, i16 %mask) {
; X64-LABEL: test_maskz_expand_d_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpexpandd %zmm0, %zmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_maskz_expand_d_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vpexpandd %zmm0, %zmm0 {%k1} {z}
; X86-NEXT:    retl
  %1 = bitcast i16 %mask to <16 x i1>
  %2 = call <16 x i32> @llvm.x86.avx512.mask.expand.v16i32(<16 x i32> %data, <16 x i32> zeroinitializer, <16 x i1> %1)
  ret <16 x i32> %2
}

define <16 x float> @test_rcp_ps_512(<16 x float> %a0) {
; CHECK-LABEL: test_rcp_ps_512:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vrcp14ps %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %res = call <16 x float> @llvm.x86.avx512.rcp14.ps.512(<16 x float> %a0, <16 x float> zeroinitializer, i16 -1) ; <<16 x float>> [#uses=1]
  ret <16 x float> %res
}
declare <16 x float> @llvm.x86.avx512.rcp14.ps.512(<16 x float>, <16 x float>, i16) nounwind readnone

define <8 x double> @test_rcp_pd_512(<8 x double> %a0) {
; CHECK-LABEL: test_rcp_pd_512:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vrcp14pd %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %res = call <8 x double> @llvm.x86.avx512.rcp14.pd.512(<8 x double> %a0, <8 x double> zeroinitializer, i8 -1) ; <<8 x double>> [#uses=1]
  ret <8 x double> %res
}
declare <8 x double> @llvm.x86.avx512.rcp14.pd.512(<8 x double>, <8 x double>, i8) nounwind readnone

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

define <2 x double> @test_rndscale_sd(<2 x double> %a, <2 x double> %b) {
; CHECK-LABEL: test_rndscale_sd:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vroundsd $11, %xmm1, %xmm0, %xmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %res = call <2 x double> @llvm.x86.avx512.mask.rndscale.sd(<2 x double> %a, <2 x double> %b, <2 x double> undef, i8 -1, i32 11, i32 4)
  ret <2 x double>%res
}

define <2 x double> @test_rndscale_sd_mask(<2 x double> %a, <2 x double> %b, <2 x double> %c, i8 %mask) {
; X64-LABEL: test_rndscale_sd_mask:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vrndscalesd $11, %xmm1, %xmm0, %xmm2 {%k1}
; X64-NEXT:    vmovapd %xmm2, %xmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_rndscale_sd_mask:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vrndscalesd $11, %xmm1, %xmm0, %xmm2 {%k1}
; X86-NEXT:    vmovapd %xmm2, %xmm0
; X86-NEXT:    retl
  %res = call <2 x double> @llvm.x86.avx512.mask.rndscale.sd(<2 x double> %a, <2 x double> %b, <2 x double> %c, i8 %mask, i32 11, i32 4)
  ret <2 x double>%res
}

define <2 x double> @test_rndscale_sd_mask_load(<2 x double> %a, <2 x double>* %bptr, <2 x double> %c, i8 %mask) {
; X64-LABEL: test_rndscale_sd_mask_load:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1
; X64-NEXT:    vrndscalesd $11, (%rdi), %xmm0, %xmm1 {%k1}
; X64-NEXT:    vmovapd %xmm1, %xmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_rndscale_sd_mask_load:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    vrndscalesd $11, (%eax), %xmm0, %xmm1 {%k1}
; X86-NEXT:    vmovapd %xmm1, %xmm0
; X86-NEXT:    retl
  %b = load <2 x double>, <2 x double>* %bptr
  %res = call <2 x double> @llvm.x86.avx512.mask.rndscale.sd(<2 x double> %a, <2 x double> %b, <2 x double> %c, i8 %mask, i32 11, i32 4)
  ret <2 x double>%res
}

define <2 x double> @test_rndscale_sd_maskz(<2 x double> %a, <2 x double> %b, i8 %mask) {
; X64-LABEL: test_rndscale_sd_maskz:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vrndscalesd $11, %xmm1, %xmm0, %xmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_rndscale_sd_maskz:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vrndscalesd $11, %xmm1, %xmm0, %xmm0 {%k1} {z}
; X86-NEXT:    retl
  %res = call <2 x double> @llvm.x86.avx512.mask.rndscale.sd(<2 x double> %a, <2 x double> %b, <2 x double> zeroinitializer, i8 %mask, i32 11, i32 4)
  ret <2 x double>%res
}

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

define <4 x float> @test_rndscale_ss(<4 x float> %a, <4 x float> %b) {
; CHECK-LABEL: test_rndscale_ss:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vroundss $11, %xmm1, %xmm0, %xmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %res = call <4 x float> @llvm.x86.avx512.mask.rndscale.ss(<4 x float> %a, <4 x float> %b, <4 x float> undef, i8 -1, i32 11, i32 4)
  ret <4 x float>%res
}

define <4 x float> @test_rndscale_ss_load(<4 x float> %a, <4 x float>* %bptr) {
; X64-LABEL: test_rndscale_ss_load:
; X64:       # %bb.0:
; X64-NEXT:    vroundss $11, (%rdi), %xmm0, %xmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_rndscale_ss_load:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    vroundss $11, (%eax), %xmm0, %xmm0
; X86-NEXT:    retl
  %b = load <4 x float>, <4 x float>* %bptr
  %res = call <4 x float> @llvm.x86.avx512.mask.rndscale.ss(<4 x float> %a, <4 x float> %b, <4 x float> undef, i8 -1, i32 11, i32 4)
  ret <4 x float>%res
}

define <4 x float> @test_rndscale_ss_mask(<4 x float> %a, <4 x float> %b, <4 x float> %c, i8 %mask) {
; X64-LABEL: test_rndscale_ss_mask:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vrndscaless $11, %xmm1, %xmm0, %xmm2 {%k1}
; X64-NEXT:    vmovaps %xmm2, %xmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_rndscale_ss_mask:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vrndscaless $11, %xmm1, %xmm0, %xmm2 {%k1}
; X86-NEXT:    vmovaps %xmm2, %xmm0
; X86-NEXT:    retl
  %res = call <4 x float> @llvm.x86.avx512.mask.rndscale.ss(<4 x float> %a, <4 x float> %b, <4 x float> %c, i8 %mask, i32 11, i32 4)
  ret <4 x float>%res
}

define <4 x float> @test_rndscale_ss_maskz(<4 x float> %a, <4 x float> %b, i8 %mask) {
; X64-LABEL: test_rndscale_ss_maskz:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vrndscaless $11, %xmm1, %xmm0, %xmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_rndscale_ss_maskz:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vrndscaless $11, %xmm1, %xmm0, %xmm0 {%k1} {z}
; X86-NEXT:    retl
  %res = call <4 x float> @llvm.x86.avx512.mask.rndscale.ss(<4 x float> %a, <4 x float> %b, <4 x float> zeroinitializer, i8 %mask, i32 11, i32 4)
  ret <4 x float>%res
}

declare <8 x double> @llvm.x86.avx512.mask.rndscale.pd.512(<8 x double>, i32, <8 x double>, i8, i32)

define <8 x double> @test7(<8 x double> %a) {
; CHECK-LABEL: test7:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vrndscalepd $11, %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %res = call <8 x double> @llvm.x86.avx512.mask.rndscale.pd.512(<8 x double> %a, i32 11, <8 x double> %a, i8 -1, i32 4)
  ret <8 x double>%res
}

declare <16 x float> @llvm.x86.avx512.mask.rndscale.ps.512(<16 x float>, i32, <16 x float>, i16, i32)

define <16 x float> @test8(<16 x float> %a) {
; CHECK-LABEL: test8:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vrndscaleps $11, %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %res = call <16 x float> @llvm.x86.avx512.mask.rndscale.ps.512(<16 x float> %a, i32 11, <16 x float> %a, i16 -1, i32 4)
  ret <16 x float>%res
}

define <16 x float> @test_rsqrt_ps_512(<16 x float> %a0) {
; CHECK-LABEL: test_rsqrt_ps_512:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vrsqrt14ps %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %res = call <16 x float> @llvm.x86.avx512.rsqrt14.ps.512(<16 x float> %a0, <16 x float> zeroinitializer, i16 -1) ; <<16 x float>> [#uses=1]
  ret <16 x float> %res
}
declare <16 x float> @llvm.x86.avx512.rsqrt14.ps.512(<16 x float>, <16 x float>, i16) nounwind readnone

define <8 x double> @test_sqrt_pd_512(<8 x double> %a0) {
; CHECK-LABEL: test_sqrt_pd_512:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vsqrtpd %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %1 = call <8 x double> @llvm.sqrt.v8f64(<8 x double> %a0)
  ret <8 x double> %1
}

define <8 x double> @test_mask_sqrt_pd_512(<8 x double> %a0, <8 x double> %passthru, i8 %mask) {
; X64-LABEL: test_mask_sqrt_pd_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vsqrtpd %zmm0, %zmm1 {%k1}
; X64-NEXT:    vmovapd %zmm1, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_mask_sqrt_pd_512:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vsqrtpd %zmm0, %zmm1 {%k1}
; X86-NEXT:    vmovapd %zmm1, %zmm0
; X86-NEXT:    retl
  %1 = call <8 x double> @llvm.sqrt.v8f64(<8 x double> %a0)
  %2 = bitcast i8 %mask to <8 x i1>
  %3 = select <8 x i1> %2, <8 x double> %1, <8 x double> %passthru
  ret <8 x double> %3
}

define <8 x double> @test_maskz_sqrt_pd_512(<8 x double> %a0, i8 %mask) {
; X64-LABEL: test_maskz_sqrt_pd_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vsqrtpd %zmm0, %zmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_maskz_sqrt_pd_512:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vsqrtpd %zmm0, %zmm0 {%k1} {z}
; X86-NEXT:    retl
  %1 = call <8 x double> @llvm.sqrt.v8f64(<8 x double> %a0)
  %2 = bitcast i8 %mask to <8 x i1>
  %3 = select <8 x i1> %2, <8 x double> %1, <8 x double> zeroinitializer
  ret <8 x double> %3
}
declare <8 x double> @llvm.sqrt.v8f64(<8 x double>)

define <8 x double> @test_sqrt_round_pd_512(<8 x double> %a0) {
; CHECK-LABEL: test_sqrt_round_pd_512:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vsqrtpd {rz-sae}, %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %1 = call <8 x double> @llvm.x86.avx512.sqrt.pd.512(<8 x double> %a0, i32 11)
  ret <8 x double> %1
}

define <8 x double> @test_mask_sqrt_round_pd_512(<8 x double> %a0, <8 x double> %passthru, i8 %mask) {
; X64-LABEL: test_mask_sqrt_round_pd_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vsqrtpd {rz-sae}, %zmm0, %zmm1 {%k1}
; X64-NEXT:    vmovapd %zmm1, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_mask_sqrt_round_pd_512:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vsqrtpd {rz-sae}, %zmm0, %zmm1 {%k1}
; X86-NEXT:    vmovapd %zmm1, %zmm0
; X86-NEXT:    retl
  %1 = call <8 x double> @llvm.x86.avx512.sqrt.pd.512(<8 x double> %a0, i32 11)
  %2 = bitcast i8 %mask to <8 x i1>
  %3 = select <8 x i1> %2, <8 x double> %1, <8 x double> %passthru
  ret <8 x double> %3
}

define <8 x double> @test_maskz_sqrt_round_pd_512(<8 x double> %a0, i8 %mask) {
; X64-LABEL: test_maskz_sqrt_round_pd_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vsqrtpd {rz-sae}, %zmm0, %zmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_maskz_sqrt_round_pd_512:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vsqrtpd {rz-sae}, %zmm0, %zmm0 {%k1} {z}
; X86-NEXT:    retl
  %1 = call <8 x double> @llvm.x86.avx512.sqrt.pd.512(<8 x double> %a0, i32 11)
  %2 = bitcast i8 %mask to <8 x i1>
  %3 = select <8 x i1> %2, <8 x double> %1, <8 x double> zeroinitializer
  ret <8 x double> %3
}
declare <8 x double> @llvm.x86.avx512.sqrt.pd.512(<8 x double>, i32) nounwind readnone

define <16 x float> @test_sqrt_ps_512(<16 x float> %a0) {
; CHECK-LABEL: test_sqrt_ps_512:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vsqrtps %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %1 = call <16 x float> @llvm.sqrt.v16f32(<16 x float> %a0)
  ret <16 x float> %1
}

define <16 x float> @test_mask_sqrt_ps_512(<16 x float> %a0, <16 x float> %passthru, i16 %mask) {
; X64-LABEL: test_mask_sqrt_ps_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vsqrtps %zmm0, %zmm1 {%k1}
; X64-NEXT:    vmovaps %zmm1, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_mask_sqrt_ps_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vsqrtps %zmm0, %zmm1 {%k1}
; X86-NEXT:    vmovaps %zmm1, %zmm0
; X86-NEXT:    retl
  %1 = call <16 x float> @llvm.sqrt.v16f32(<16 x float> %a0)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x float> %1, <16 x float> %passthru
  ret <16 x float> %3
}

define <16 x float> @test_maskz_sqrt_ps_512(<16 x float> %a0, i16 %mask) {
; X64-LABEL: test_maskz_sqrt_ps_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vsqrtps %zmm0, %zmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_maskz_sqrt_ps_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vsqrtps %zmm0, %zmm0 {%k1} {z}
; X86-NEXT:    retl
  %1 = call <16 x float> @llvm.sqrt.v16f32(<16 x float> %a0)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x float> %1, <16 x float> zeroinitializer
  ret <16 x float> %3
}
declare <16 x float> @llvm.sqrt.v16f32(<16 x float>)

define <16 x float> @test_sqrt_round_ps_512(<16 x float> %a0) {
; CHECK-LABEL: test_sqrt_round_ps_512:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vsqrtps {rz-sae}, %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %1 = call <16 x float> @llvm.x86.avx512.sqrt.ps.512(<16 x float> %a0, i32 11)
  ret <16 x float> %1
}

define <16 x float> @test_mask_sqrt_round_ps_512(<16 x float> %a0, <16 x float> %passthru, i16 %mask) {
; X64-LABEL: test_mask_sqrt_round_ps_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vsqrtps {rz-sae}, %zmm0, %zmm1 {%k1}
; X64-NEXT:    vmovaps %zmm1, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_mask_sqrt_round_ps_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vsqrtps {rz-sae}, %zmm0, %zmm1 {%k1}
; X86-NEXT:    vmovaps %zmm1, %zmm0
; X86-NEXT:    retl
  %1 = call <16 x float> @llvm.x86.avx512.sqrt.ps.512(<16 x float> %a0, i32 11)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x float> %1, <16 x float> %passthru
  ret <16 x float> %3
}

define <16 x float> @test_maskz_sqrt_round_ps_512(<16 x float> %a0, i16 %mask) {
; X64-LABEL: test_maskz_sqrt_round_ps_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vsqrtps {rz-sae}, %zmm0, %zmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_maskz_sqrt_round_ps_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vsqrtps {rz-sae}, %zmm0, %zmm0 {%k1} {z}
; X86-NEXT:    retl
  %1 = call <16 x float> @llvm.x86.avx512.sqrt.ps.512(<16 x float> %a0, i32 11)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x float> %1, <16 x float> zeroinitializer
  ret <16 x float> %3
}
declare <16 x float> @llvm.x86.avx512.sqrt.ps.512(<16 x float>, i32) nounwind readnone

define <8 x double> @test_getexp_pd_512(<8 x double> %a0) {
; CHECK-LABEL: test_getexp_pd_512:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vgetexppd %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %res = call <8 x double> @llvm.x86.avx512.mask.getexp.pd.512(<8 x double> %a0,  <8 x double> zeroinitializer, i8 -1, i32 4)
  ret <8 x double> %res
}
define <8 x double> @test_getexp_round_pd_512(<8 x double> %a0) {
; CHECK-LABEL: test_getexp_round_pd_512:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vgetexppd {sae}, %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %res = call <8 x double> @llvm.x86.avx512.mask.getexp.pd.512(<8 x double> %a0,  <8 x double> zeroinitializer, i8 -1, i32 8)
  ret <8 x double> %res
}
declare <8 x double> @llvm.x86.avx512.mask.getexp.pd.512(<8 x double>, <8 x double>, i8, i32) nounwind readnone

define <16 x float> @test_getexp_ps_512(<16 x float> %a0) {
; CHECK-LABEL: test_getexp_ps_512:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vgetexpps %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %res = call <16 x float> @llvm.x86.avx512.mask.getexp.ps.512(<16 x float> %a0, <16 x float> zeroinitializer, i16 -1, i32 4)
  ret <16 x float> %res
}

define <16 x float> @test_getexp_round_ps_512(<16 x float> %a0) {
; CHECK-LABEL: test_getexp_round_ps_512:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vgetexpps {sae}, %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %res = call <16 x float> @llvm.x86.avx512.mask.getexp.ps.512(<16 x float> %a0, <16 x float> zeroinitializer, i16 -1, i32 8)
  ret <16 x float> %res
}
declare <16 x float> @llvm.x86.avx512.mask.getexp.ps.512(<16 x float>, <16 x float>, i16, i32) nounwind readnone

declare <4 x float> @llvm.x86.avx512.mask.sqrt.ss(<4 x float>, <4 x float>, <4 x float>, i8, i32) nounwind readnone

define <4 x float> @test_sqrt_ss(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2, i8 %mask) {
; X64-LABEL: test_sqrt_ss:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vmovaps %xmm2, %xmm3
; X64-NEXT:    vsqrtss %xmm1, %xmm0, %xmm3 {%k1}
; X64-NEXT:    vsqrtss {rd-sae}, %xmm1, %xmm0, %xmm2 {%k1}
; X64-NEXT:    vaddps %xmm2, %xmm3, %xmm2
; X64-NEXT:    vsqrtss {ru-sae}, %xmm1, %xmm0, %xmm3 {%k1} {z}
; X64-NEXT:    vsqrtss {rz-sae}, %xmm1, %xmm0, %xmm0
; X64-NEXT:    vaddps %xmm0, %xmm3, %xmm0
; X64-NEXT:    vaddps %xmm0, %xmm2, %xmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_sqrt_ss:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vmovaps %xmm2, %xmm3
; X86-NEXT:    vsqrtss %xmm1, %xmm0, %xmm3 {%k1}
; X86-NEXT:    vsqrtss {rd-sae}, %xmm1, %xmm0, %xmm2 {%k1}
; X86-NEXT:    vaddps %xmm2, %xmm3, %xmm2
; X86-NEXT:    vsqrtss {ru-sae}, %xmm1, %xmm0, %xmm3 {%k1} {z}
; X86-NEXT:    vsqrtss {rz-sae}, %xmm1, %xmm0, %xmm0
; X86-NEXT:    vaddps %xmm0, %xmm3, %xmm0
; X86-NEXT:    vaddps %xmm0, %xmm2, %xmm0
; X86-NEXT:    retl
  %res0 = call <4 x float> @llvm.x86.avx512.mask.sqrt.ss(<4 x float>%a0, <4 x float> %a1, <4 x float> %a2, i8 %mask, i32 4)
  %res1 = call <4 x float> @llvm.x86.avx512.mask.sqrt.ss(<4 x float>%a0, <4 x float> %a1, <4 x float> %a2, i8 %mask, i32 1)
  %res2 = call <4 x float> @llvm.x86.avx512.mask.sqrt.ss(<4 x float>%a0, <4 x float> %a1, <4 x float> zeroinitializer, i8 %mask, i32 2)
  %res3 = call <4 x float> @llvm.x86.avx512.mask.sqrt.ss(<4 x float>%a0, <4 x float> %a1, <4 x float> zeroinitializer, i8 -1, i32 3)

  %res.1 = fadd <4 x float> %res0, %res1
  %res.2 = fadd <4 x float> %res2, %res3
  %res   = fadd <4 x float> %res.1, %res.2
  ret <4 x float> %res
}

declare <2 x double> @llvm.x86.avx512.mask.sqrt.sd(<2 x double>, <2 x double>, <2 x double>, i8, i32) nounwind readnone

define <2 x double> @test_sqrt_sd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2, i8 %mask) {
; X64-LABEL: test_sqrt_sd:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vmovapd %xmm2, %xmm3
; X64-NEXT:    vsqrtsd %xmm1, %xmm0, %xmm3 {%k1}
; X64-NEXT:    vsqrtsd {rd-sae}, %xmm1, %xmm0, %xmm2 {%k1}
; X64-NEXT:    vaddpd %xmm2, %xmm3, %xmm2
; X64-NEXT:    vsqrtsd {ru-sae}, %xmm1, %xmm0, %xmm3 {%k1} {z}
; X64-NEXT:    vsqrtsd {rz-sae}, %xmm1, %xmm0, %xmm0
; X64-NEXT:    vaddpd %xmm0, %xmm3, %xmm0
; X64-NEXT:    vaddpd %xmm0, %xmm2, %xmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_sqrt_sd:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vmovapd %xmm2, %xmm3
; X86-NEXT:    vsqrtsd %xmm1, %xmm0, %xmm3 {%k1}
; X86-NEXT:    vsqrtsd {rd-sae}, %xmm1, %xmm0, %xmm2 {%k1}
; X86-NEXT:    vaddpd %xmm2, %xmm3, %xmm2
; X86-NEXT:    vsqrtsd {ru-sae}, %xmm1, %xmm0, %xmm3 {%k1} {z}
; X86-NEXT:    vsqrtsd {rz-sae}, %xmm1, %xmm0, %xmm0
; X86-NEXT:    vaddpd %xmm0, %xmm3, %xmm0
; X86-NEXT:    vaddpd %xmm0, %xmm2, %xmm0
; X86-NEXT:    retl
  %res0 = call <2 x double> @llvm.x86.avx512.mask.sqrt.sd(<2 x double>%a0, <2 x double> %a1, <2 x double> %a2, i8 %mask, i32 4)
  %res1 = call <2 x double> @llvm.x86.avx512.mask.sqrt.sd(<2 x double>%a0, <2 x double> %a1, <2 x double> %a2, i8 %mask, i32 1)
  %res2 = call <2 x double> @llvm.x86.avx512.mask.sqrt.sd(<2 x double>%a0, <2 x double> %a1, <2 x double> zeroinitializer, i8 %mask, i32 2)
  %res3 = call <2 x double> @llvm.x86.avx512.mask.sqrt.sd(<2 x double>%a0, <2 x double> %a1, <2 x double> zeroinitializer, i8 -1, i32 3)

  %res.1 = fadd <2 x double> %res0, %res1
  %res.2 = fadd <2 x double> %res2, %res3
  %res   = fadd <2 x double> %res.1, %res.2
  ret <2 x double> %res
}

define i32 @test_x86_avx512_cvttsd2usi(<2 x double> %a0) {
; CHECK-LABEL: test_x86_avx512_cvttsd2usi:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vcvttsd2usi %xmm0, %ecx
; CHECK-NEXT:    vcvttsd2usi {sae}, %xmm0, %eax
; CHECK-NEXT:    addl %ecx, %eax
; CHECK-NEXT:    ret{{[l|q]}}
  %res0 = call i32 @llvm.x86.avx512.cvttsd2usi(<2 x double> %a0, i32 4) ;
  %res1 = call i32 @llvm.x86.avx512.cvttsd2usi(<2 x double> %a0, i32 8) ;
  %res2 = add i32 %res0, %res1
  ret i32 %res2
}
declare i32 @llvm.x86.avx512.cvttsd2usi(<2 x double>, i32) nounwind readnone

define i32 @test_x86_avx512_cvttsd2si(<2 x double> %a0) {
; CHECK-LABEL: test_x86_avx512_cvttsd2si:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vcvttsd2si %xmm0, %ecx
; CHECK-NEXT:    vcvttsd2si {sae}, %xmm0, %eax
; CHECK-NEXT:    addl %ecx, %eax
; CHECK-NEXT:    ret{{[l|q]}}
  %res0 = call i32 @llvm.x86.avx512.cvttsd2si(<2 x double> %a0, i32 4) ;
  %res1 = call i32 @llvm.x86.avx512.cvttsd2si(<2 x double> %a0, i32 8) ;
  %res2 = add i32 %res0, %res1
  ret i32 %res2
}
declare i32 @llvm.x86.avx512.cvttsd2si(<2 x double>, i32) nounwind readnone

define i32 @test_x86_avx512_cvttss2si(<4 x float> %a0) {
; CHECK-LABEL: test_x86_avx512_cvttss2si:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vcvttss2si {sae}, %xmm0, %ecx
; CHECK-NEXT:    vcvttss2si %xmm0, %eax
; CHECK-NEXT:    addl %ecx, %eax
; CHECK-NEXT:    ret{{[l|q]}}
  %res0 = call i32 @llvm.x86.avx512.cvttss2si(<4 x float> %a0, i32 8) ;
  %res1 = call i32 @llvm.x86.avx512.cvttss2si(<4 x float> %a0, i32 4) ;
  %res2 = add i32 %res0, %res1
  ret i32 %res2
}
declare i32 @llvm.x86.avx512.cvttss2si(<4 x float>, i32) nounwind readnone

define i32 @test_x86_avx512_cvttss2si_load(<4 x float>* %a0) {
; X64-LABEL: test_x86_avx512_cvttss2si_load:
; X64:       # %bb.0:
; X64-NEXT:    vcvttss2si (%rdi), %eax
; X64-NEXT:    retq
;
; X86-LABEL: test_x86_avx512_cvttss2si_load:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    vcvttss2si (%eax), %eax
; X86-NEXT:    retl
  %a1 = load <4 x float>, <4 x float>* %a0
  %res = call i32 @llvm.x86.avx512.cvttss2si(<4 x float> %a1, i32 4) ;
  ret i32 %res
}

define i32 @test_x86_avx512_cvttss2usi(<4 x float> %a0) {
; CHECK-LABEL: test_x86_avx512_cvttss2usi:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vcvttss2usi {sae}, %xmm0, %ecx
; CHECK-NEXT:    vcvttss2usi %xmm0, %eax
; CHECK-NEXT:    addl %ecx, %eax
; CHECK-NEXT:    ret{{[l|q]}}
  %res0 = call i32 @llvm.x86.avx512.cvttss2usi(<4 x float> %a0, i32 8) ;
  %res1 = call i32 @llvm.x86.avx512.cvttss2usi(<4 x float> %a0, i32 4) ;
  %res2 = add i32 %res0, %res1
  ret i32 %res2
}
declare i32 @llvm.x86.avx512.cvttss2usi(<4 x float>, i32) nounwind readnone

define i32 @test_x86_avx512_cvtsd2usi32(<2 x double> %a0) {
; CHECK-LABEL: test_x86_avx512_cvtsd2usi32:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vcvtsd2usi %xmm0, %eax
; CHECK-NEXT:    vcvtsd2usi {rz-sae}, %xmm0, %ecx
; CHECK-NEXT:    addl %eax, %ecx
; CHECK-NEXT:    vcvtsd2usi {rd-sae}, %xmm0, %eax
; CHECK-NEXT:    addl %ecx, %eax
; CHECK-NEXT:    ret{{[l|q]}}

  %res = call i32 @llvm.x86.avx512.vcvtsd2usi32(<2 x double> %a0, i32 4)
  %res1 = call i32 @llvm.x86.avx512.vcvtsd2usi32(<2 x double> %a0, i32 3)
  %res2 = call i32 @llvm.x86.avx512.vcvtsd2usi32(<2 x double> %a0, i32 1)
  %res3 = add i32 %res, %res1
  %res4 = add i32 %res3, %res2
  ret i32 %res4
}
declare i32 @llvm.x86.avx512.vcvtsd2usi32(<2 x double>, i32) nounwind readnone

define i32 @test_x86_avx512_cvtsd2si32(<2 x double> %a0) {
; CHECK-LABEL: test_x86_avx512_cvtsd2si32:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vcvtsd2si %xmm0, %eax
; CHECK-NEXT:    vcvtsd2si {rz-sae}, %xmm0, %ecx
; CHECK-NEXT:    addl %eax, %ecx
; CHECK-NEXT:    vcvtsd2si {rd-sae}, %xmm0, %eax
; CHECK-NEXT:    addl %ecx, %eax
; CHECK-NEXT:    ret{{[l|q]}}

  %res = call i32 @llvm.x86.avx512.vcvtsd2si32(<2 x double> %a0, i32 4)
  %res1 = call i32 @llvm.x86.avx512.vcvtsd2si32(<2 x double> %a0, i32 3)
  %res2 = call i32 @llvm.x86.avx512.vcvtsd2si32(<2 x double> %a0, i32 1)
  %res3 = add i32 %res, %res1
  %res4 = add i32 %res3, %res2
  ret i32 %res4
}
declare i32 @llvm.x86.avx512.vcvtsd2si32(<2 x double>, i32) nounwind readnone

define i32 @test_x86_avx512_cvtss2usi32(<4 x float> %a0) {
; CHECK-LABEL: test_x86_avx512_cvtss2usi32:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vcvtss2usi %xmm0, %eax
; CHECK-NEXT:    vcvtss2usi {rz-sae}, %xmm0, %ecx
; CHECK-NEXT:    addl %eax, %ecx
; CHECK-NEXT:    vcvtss2usi {rd-sae}, %xmm0, %eax
; CHECK-NEXT:    addl %ecx, %eax
; CHECK-NEXT:    ret{{[l|q]}}

  %res = call i32 @llvm.x86.avx512.vcvtss2usi32(<4 x float> %a0, i32 4)
  %res1 = call i32 @llvm.x86.avx512.vcvtss2usi32(<4 x float> %a0, i32 3)
  %res2 = call i32 @llvm.x86.avx512.vcvtss2usi32(<4 x float> %a0, i32 1)
  %res3 = add i32 %res, %res1
  %res4 = add i32 %res3, %res2
  ret i32 %res4
}
declare i32 @llvm.x86.avx512.vcvtss2usi32(<4 x float>, i32) nounwind readnone

define i32 @test_x86_avx512_cvtss2si32(<4 x float> %a0) {
; CHECK-LABEL: test_x86_avx512_cvtss2si32:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vcvtss2si %xmm0, %eax
; CHECK-NEXT:    vcvtss2si {rz-sae}, %xmm0, %ecx
; CHECK-NEXT:    addl %eax, %ecx
; CHECK-NEXT:    vcvtss2si {rd-sae}, %xmm0, %eax
; CHECK-NEXT:    addl %ecx, %eax
; CHECK-NEXT:    ret{{[l|q]}}

  %res = call i32 @llvm.x86.avx512.vcvtss2si32(<4 x float> %a0, i32 4)
  %res1 = call i32 @llvm.x86.avx512.vcvtss2si32(<4 x float> %a0, i32 3)
  %res2 = call i32 @llvm.x86.avx512.vcvtss2si32(<4 x float> %a0, i32 1)
  %res3 = add i32 %res, %res1
  %res4 = add i32 %res3, %res2
  ret i32 %res4
}
declare i32 @llvm.x86.avx512.vcvtss2si32(<4 x float>, i32) nounwind readnone

define <16 x float> @test_x86_vcvtph2ps_512(<16 x i16> %a0) {
; CHECK-LABEL: test_x86_vcvtph2ps_512:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vcvtph2ps %ymm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %res = call <16 x float> @llvm.x86.avx512.mask.vcvtph2ps.512(<16 x i16> %a0, <16 x float> zeroinitializer, i16 -1, i32 4)
  ret <16 x float> %res
}

define <16 x float> @test_x86_vcvtph2ps_512_sae(<16 x i16> %a0) {
; CHECK-LABEL: test_x86_vcvtph2ps_512_sae:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vcvtph2ps {sae}, %ymm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %res = call <16 x float> @llvm.x86.avx512.mask.vcvtph2ps.512(<16 x i16> %a0, <16 x float> zeroinitializer, i16 -1, i32 8)
  ret <16 x float> %res
}

define <16 x float> @test_x86_vcvtph2ps_512_rrk(<16 x i16> %a0,<16 x float> %a1, i16 %mask) {
; X64-LABEL: test_x86_vcvtph2ps_512_rrk:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vcvtph2ps %ymm0, %zmm1 {%k1}
; X64-NEXT:    vmovaps %zmm1, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_x86_vcvtph2ps_512_rrk:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vcvtph2ps %ymm0, %zmm1 {%k1}
; X86-NEXT:    vmovaps %zmm1, %zmm0
; X86-NEXT:    retl
  %res = call <16 x float> @llvm.x86.avx512.mask.vcvtph2ps.512(<16 x i16> %a0, <16 x float> %a1, i16 %mask, i32 4)
  ret <16 x float> %res
}

define <16 x float> @test_x86_vcvtph2ps_512_sae_rrkz(<16 x i16> %a0, i16 %mask) {
; X64-LABEL: test_x86_vcvtph2ps_512_sae_rrkz:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vcvtph2ps {sae}, %ymm0, %zmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_x86_vcvtph2ps_512_sae_rrkz:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vcvtph2ps {sae}, %ymm0, %zmm0 {%k1} {z}
; X86-NEXT:    retl
  %res = call <16 x float> @llvm.x86.avx512.mask.vcvtph2ps.512(<16 x i16> %a0, <16 x float> zeroinitializer, i16 %mask, i32 8)
  ret <16 x float> %res
}

define <16 x float> @test_x86_vcvtph2ps_512_rrkz(<16 x i16> %a0, i16 %mask) {
; X64-LABEL: test_x86_vcvtph2ps_512_rrkz:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vcvtph2ps %ymm0, %zmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_x86_vcvtph2ps_512_rrkz:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vcvtph2ps %ymm0, %zmm0 {%k1} {z}
; X86-NEXT:    retl
  %res = call <16 x float> @llvm.x86.avx512.mask.vcvtph2ps.512(<16 x i16> %a0, <16 x float> zeroinitializer, i16 %mask, i32 4)
  ret <16 x float> %res
}

declare <16 x float> @llvm.x86.avx512.mask.vcvtph2ps.512(<16 x i16>, <16 x float>, i16, i32) nounwind readonly

define <16 x i16> @test_x86_vcvtps2ph_256(<16 x float> %a0, <16 x i16> %src, i16 %mask, <16 x i16> * %dst) {
; X64-LABEL: test_x86_vcvtps2ph_256:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vcvtps2ph $2, %zmm0, %ymm2 {%k1} {z}
; X64-NEXT:    vcvtps2ph $2, %zmm0, %ymm1 {%k1}
; X64-NEXT:    vpaddw %ymm1, %ymm2, %ymm1
; X64-NEXT:    vcvtps2ph $2, %zmm0, (%rsi)
; X64-NEXT:    vmovdqa %ymm1, %ymm0
; X64-NEXT:    retq
;
; X86-LABEL: test_x86_vcvtps2ph_256:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vcvtps2ph $2, %zmm0, %ymm2 {%k1} {z}
; X86-NEXT:    vcvtps2ph $2, %zmm0, %ymm1 {%k1}
; X86-NEXT:    vpaddw %ymm1, %ymm2, %ymm1
; X86-NEXT:    vcvtps2ph $2, %zmm0, (%eax)
; X86-NEXT:    vmovdqa %ymm1, %ymm0
; X86-NEXT:    retl
  %res1 = call <16 x i16> @llvm.x86.avx512.mask.vcvtps2ph.512(<16 x float> %a0, i32 2, <16 x i16> zeroinitializer, i16 -1)
  %res2 = call <16 x i16> @llvm.x86.avx512.mask.vcvtps2ph.512(<16 x float> %a0, i32 2, <16 x i16> zeroinitializer, i16 %mask)
  %res3 = call <16 x i16> @llvm.x86.avx512.mask.vcvtps2ph.512(<16 x float> %a0, i32 2, <16 x i16> %src, i16 %mask)
  store <16 x i16> %res1, <16 x i16> * %dst
  %res  = add <16 x i16> %res2, %res3
  ret <16 x i16> %res
}

declare <16 x i16> @llvm.x86.avx512.mask.vcvtps2ph.512(<16 x float>, i32, <16 x i16>, i16) nounwind readonly

define i16 @test_cmpps(<16 x float> %a, <16 x float> %b) {
; CHECK-LABEL: test_cmpps:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vcmpleps {sae}, %zmm1, %zmm0, %k0
; CHECK-NEXT:    kmovw %k0, %eax
; CHECK-NEXT:    # kill: def $ax killed $ax killed $eax
; CHECK-NEXT:    vzeroupper
; CHECK-NEXT:    ret{{[l|q]}}
  %res = call <16 x i1> @llvm.x86.avx512.cmp.ps.512(<16 x float> %a, <16 x float> %b, i32 2, i32 8)
  %1 = bitcast <16 x i1> %res to i16
  ret i16 %1
}
declare <16 x i1> @llvm.x86.avx512.cmp.ps.512(<16 x float>, <16 x float>, i32, i32)

define i8 @test_cmppd(<8 x double> %a, <8 x double> %b) {
; CHECK-LABEL: test_cmppd:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vcmpneqpd %zmm1, %zmm0, %k0
; CHECK-NEXT:    kmovw %k0, %eax
; CHECK-NEXT:    # kill: def $al killed $al killed $eax
; CHECK-NEXT:    vzeroupper
; CHECK-NEXT:    ret{{[l|q]}}
  %res = call <8 x i1> @llvm.x86.avx512.cmp.pd.512(<8 x double> %a, <8 x double> %b, i32 4, i32 4)
  %1 = bitcast <8 x i1> %res to i8
  ret i8 %1
}
declare <8 x i1> @llvm.x86.avx512.cmp.pd.512(<8 x double>, <8 x double>, i32, i32)

; Function Attrs: nounwind readnone

 ; fp min - max
define <8 x double> @test_vmaxpd(<8 x double> %a0, <8 x double> %a1) {
; CHECK-LABEL: test_vmaxpd:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vmaxpd %zmm1, %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %1 = call <8 x double> @llvm.x86.avx512.max.pd.512(<8 x double> %a0, <8 x double> %a1, i32 4)
  ret <8 x double> %1
}
declare <8 x double> @llvm.x86.avx512.max.pd.512(<8 x double>, <8 x double>, i32)

define <8 x double> @test_vminpd(<8 x double> %a0, <8 x double> %a1) {
; CHECK-LABEL: test_vminpd:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vminpd %zmm1, %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %1 = call <8 x double> @llvm.x86.avx512.min.pd.512(<8 x double> %a0, <8 x double> %a1, i32 4)
  ret <8 x double> %1
}
declare <8 x double> @llvm.x86.avx512.min.pd.512(<8 x double>, <8 x double>, i32)

define void @test_mask_store_ss(i8* %ptr, <4 x float> %data, i8 %mask) {
; X64-LABEL: test_mask_store_ss:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1
; X64-NEXT:    vmovss %xmm0, (%rdi) {%k1}
; X64-NEXT:    retq
;
; X86-LABEL: test_mask_store_ss:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    movb {{[0-9]+}}(%esp), %cl
; X86-NEXT:    kmovw %ecx, %k1
; X86-NEXT:    vmovss %xmm0, (%eax) {%k1}
; X86-NEXT:    retl
  %1 = and i8 %mask, 1
  %2 = bitcast i8* %ptr to <4 x float>*
  %3 = bitcast i8 %1 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>
  call void @llvm.masked.store.v4f32.p0v4f32(<4 x float> %data, <4 x float>* %2, i32 1, <4 x i1> %extract)
  ret void
}
declare void @llvm.masked.store.v4f32.p0v4f32(<4 x float>, <4 x float>*, i32, <4 x i1>) #1


declare <16 x float> @llvm.x86.avx512.sub.ps.512(<16 x float>, <16 x float>, i32)
declare <16 x float> @llvm.x86.avx512.mul.ps.512(<16 x float>, <16 x float>, i32)
declare <8 x double> @llvm.x86.avx512.mul.pd.512(<8 x double>, <8 x double>, i32)

define <16 x float> @test_vsubps_rn(<16 x float> %a0, <16 x float> %a1) {
; CHECK-LABEL: test_vsubps_rn:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vsubps {rn-sae}, %zmm1, %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %1 = call <16 x float> @llvm.x86.avx512.sub.ps.512(<16 x float> %a0, <16 x float> %a1, i32 0)
  ret <16 x float> %1
}

define <16 x float> @test_vsubps_rd(<16 x float> %a0, <16 x float> %a1) {
; CHECK-LABEL: test_vsubps_rd:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vsubps {rd-sae}, %zmm1, %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %1 = call <16 x float> @llvm.x86.avx512.sub.ps.512(<16 x float> %a0, <16 x float> %a1, i32 1)
  ret <16 x float> %1
}

define <16 x float> @test_vsubps_ru(<16 x float> %a0, <16 x float> %a1) {
; CHECK-LABEL: test_vsubps_ru:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vsubps {ru-sae}, %zmm1, %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %1 = call <16 x float> @llvm.x86.avx512.sub.ps.512(<16 x float> %a0, <16 x float> %a1, i32 2)
  ret <16 x float> %1
}

define <16 x float> @test_vsubps_rz(<16 x float> %a0, <16 x float> %a1) {
; CHECK-LABEL: test_vsubps_rz:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vsubps {rz-sae}, %zmm1, %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %1 = call <16 x float> @llvm.x86.avx512.sub.ps.512(<16 x float> %a0, <16 x float> %a1, i32 3)
  ret <16 x float> %1
}

define <16 x float> @test_vmulps_rn(<16 x float> %a0, <16 x float> %a1) {
; CHECK-LABEL: test_vmulps_rn:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vmulps {rn-sae}, %zmm1, %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %1 = call <16 x float> @llvm.x86.avx512.mul.ps.512(<16 x float> %a0, <16 x float> %a1, i32 0)
  ret <16 x float> %1
}

define <16 x float> @test_vmulps_rd(<16 x float> %a0, <16 x float> %a1) {
; CHECK-LABEL: test_vmulps_rd:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vmulps {rd-sae}, %zmm1, %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %1 = call <16 x float> @llvm.x86.avx512.mul.ps.512(<16 x float> %a0, <16 x float> %a1, i32 1)
  ret <16 x float> %1
}

define <16 x float> @test_vmulps_ru(<16 x float> %a0, <16 x float> %a1) {
; CHECK-LABEL: test_vmulps_ru:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vmulps {ru-sae}, %zmm1, %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %1 = call <16 x float> @llvm.x86.avx512.mul.ps.512(<16 x float> %a0, <16 x float> %a1, i32 2)
  ret <16 x float> %1
}

define <16 x float> @test_vmulps_rz(<16 x float> %a0, <16 x float> %a1) {
; CHECK-LABEL: test_vmulps_rz:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vmulps {rz-sae}, %zmm1, %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %1 = call <16 x float> @llvm.x86.avx512.mul.ps.512(<16 x float> %a0, <16 x float> %a1, i32 3)
  ret <16 x float> %1
}

;; mask float
define <16 x float> @test_vmulps_mask_rn(<16 x float> %a0, <16 x float> %a1, i16 %mask) {
; X64-LABEL: test_vmulps_mask_rn:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vmulps {rn-sae}, %zmm1, %zmm0, %zmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_vmulps_mask_rn:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vmulps {rn-sae}, %zmm1, %zmm0, %zmm0 {%k1} {z}
; X86-NEXT:    retl
  %1 = call <16 x float> @llvm.x86.avx512.mul.ps.512(<16 x float> %a0, <16 x float> %a1, i32 0)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x float> %1, <16 x float> zeroinitializer
  ret <16 x float> %3
}

define <16 x float> @test_vmulps_mask_rd(<16 x float> %a0, <16 x float> %a1, i16 %mask) {
; X64-LABEL: test_vmulps_mask_rd:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vmulps {rd-sae}, %zmm1, %zmm0, %zmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_vmulps_mask_rd:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vmulps {rd-sae}, %zmm1, %zmm0, %zmm0 {%k1} {z}
; X86-NEXT:    retl
  %1 = call <16 x float> @llvm.x86.avx512.mul.ps.512(<16 x float> %a0, <16 x float> %a1, i32 1)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x float> %1, <16 x float> zeroinitializer
  ret <16 x float> %3
}

define <16 x float> @test_vmulps_mask_ru(<16 x float> %a0, <16 x float> %a1, i16 %mask) {
; X64-LABEL: test_vmulps_mask_ru:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vmulps {ru-sae}, %zmm1, %zmm0, %zmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_vmulps_mask_ru:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vmulps {ru-sae}, %zmm1, %zmm0, %zmm0 {%k1} {z}
; X86-NEXT:    retl
  %1 = call <16 x float> @llvm.x86.avx512.mul.ps.512(<16 x float> %a0, <16 x float> %a1, i32 2)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x float> %1, <16 x float> zeroinitializer
  ret <16 x float> %3
}

define <16 x float> @test_vmulps_mask_rz(<16 x float> %a0, <16 x float> %a1, i16 %mask) {
; X64-LABEL: test_vmulps_mask_rz:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vmulps {rz-sae}, %zmm1, %zmm0, %zmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_vmulps_mask_rz:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vmulps {rz-sae}, %zmm1, %zmm0, %zmm0 {%k1} {z}
; X86-NEXT:    retl
  %1 = call <16 x float> @llvm.x86.avx512.mul.ps.512(<16 x float> %a0, <16 x float> %a1, i32 3)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x float> %1, <16 x float> zeroinitializer
  ret <16 x float> %3
}

;; With Passthru value
define <16 x float> @test_vmulps_mask_passthru_rn(<16 x float> %a0, <16 x float> %a1, <16 x float> %passthru, i16 %mask) {
; X64-LABEL: test_vmulps_mask_passthru_rn:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vmulps {rn-sae}, %zmm1, %zmm0, %zmm2 {%k1}
; X64-NEXT:    vmovaps %zmm2, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_vmulps_mask_passthru_rn:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vmulps {rn-sae}, %zmm1, %zmm0, %zmm2 {%k1}
; X86-NEXT:    vmovaps %zmm2, %zmm0
; X86-NEXT:    retl
  %1 = call <16 x float> @llvm.x86.avx512.mul.ps.512(<16 x float> %a0, <16 x float> %a1, i32 0)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x float> %1, <16 x float> %passthru
  ret <16 x float> %3
}

define <16 x float> @test_vmulps_mask_passthru_rd(<16 x float> %a0, <16 x float> %a1, <16 x float> %passthru, i16 %mask) {
; X64-LABEL: test_vmulps_mask_passthru_rd:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vmulps {rd-sae}, %zmm1, %zmm0, %zmm2 {%k1}
; X64-NEXT:    vmovaps %zmm2, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_vmulps_mask_passthru_rd:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vmulps {rd-sae}, %zmm1, %zmm0, %zmm2 {%k1}
; X86-NEXT:    vmovaps %zmm2, %zmm0
; X86-NEXT:    retl
  %1 = call <16 x float> @llvm.x86.avx512.mul.ps.512(<16 x float> %a0, <16 x float> %a1, i32 1)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x float> %1, <16 x float> %passthru
  ret <16 x float> %3
}

define <16 x float> @test_vmulps_mask_passthru_ru(<16 x float> %a0, <16 x float> %a1, <16 x float> %passthru, i16 %mask) {
; X64-LABEL: test_vmulps_mask_passthru_ru:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vmulps {ru-sae}, %zmm1, %zmm0, %zmm2 {%k1}
; X64-NEXT:    vmovaps %zmm2, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_vmulps_mask_passthru_ru:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vmulps {ru-sae}, %zmm1, %zmm0, %zmm2 {%k1}
; X86-NEXT:    vmovaps %zmm2, %zmm0
; X86-NEXT:    retl
  %1 = call <16 x float> @llvm.x86.avx512.mul.ps.512(<16 x float> %a0, <16 x float> %a1, i32 2)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x float> %1, <16 x float> %passthru
  ret <16 x float> %3
}

define <16 x float> @test_vmulps_mask_passthru_rz(<16 x float> %a0, <16 x float> %a1, <16 x float> %passthru, i16 %mask) {
; X64-LABEL: test_vmulps_mask_passthru_rz:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vmulps {rz-sae}, %zmm1, %zmm0, %zmm2 {%k1}
; X64-NEXT:    vmovaps %zmm2, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_vmulps_mask_passthru_rz:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vmulps {rz-sae}, %zmm1, %zmm0, %zmm2 {%k1}
; X86-NEXT:    vmovaps %zmm2, %zmm0
; X86-NEXT:    retl
  %1 = call <16 x float> @llvm.x86.avx512.mul.ps.512(<16 x float> %a0, <16 x float> %a1, i32 3)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x float> %1, <16 x float> %passthru
  ret <16 x float> %3
}

;; mask double
define <8 x double> @test_vmulpd_mask_rn(<8 x double> %a0, <8 x double> %a1, i8 %mask) {
; X64-LABEL: test_vmulpd_mask_rn:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vmulpd {rn-sae}, %zmm1, %zmm0, %zmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_vmulpd_mask_rn:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vmulpd {rn-sae}, %zmm1, %zmm0, %zmm0 {%k1} {z}
; X86-NEXT:    retl
  %1 = call <8 x double> @llvm.x86.avx512.mul.pd.512(<8 x double> %a0, <8 x double> %a1, i32 0)
  %2 = bitcast i8 %mask to <8 x i1>
  %3 = select <8 x i1> %2, <8 x double> %1, <8 x double> zeroinitializer
  ret <8 x double> %3
}

define <8 x double> @test_vmulpd_mask_rd(<8 x double> %a0, <8 x double> %a1, i8 %mask) {
; X64-LABEL: test_vmulpd_mask_rd:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vmulpd {rd-sae}, %zmm1, %zmm0, %zmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_vmulpd_mask_rd:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vmulpd {rd-sae}, %zmm1, %zmm0, %zmm0 {%k1} {z}
; X86-NEXT:    retl
  %1 = call <8 x double> @llvm.x86.avx512.mul.pd.512(<8 x double> %a0, <8 x double> %a1, i32 1)
  %2 = bitcast i8 %mask to <8 x i1>
  %3 = select <8 x i1> %2, <8 x double> %1, <8 x double> zeroinitializer
  ret <8 x double> %3
}

define <8 x double> @test_vmulpd_mask_ru(<8 x double> %a0, <8 x double> %a1, i8 %mask) {
; X64-LABEL: test_vmulpd_mask_ru:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vmulpd {ru-sae}, %zmm1, %zmm0, %zmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_vmulpd_mask_ru:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vmulpd {ru-sae}, %zmm1, %zmm0, %zmm0 {%k1} {z}
; X86-NEXT:    retl
  %1 = call <8 x double> @llvm.x86.avx512.mul.pd.512(<8 x double> %a0, <8 x double> %a1, i32 2)
  %2 = bitcast i8 %mask to <8 x i1>
  %3 = select <8 x i1> %2, <8 x double> %1, <8 x double> zeroinitializer
  ret <8 x double> %3
}

define <8 x double> @test_vmulpd_mask_rz(<8 x double> %a0, <8 x double> %a1, i8 %mask) {
; X64-LABEL: test_vmulpd_mask_rz:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vmulpd {rz-sae}, %zmm1, %zmm0, %zmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_vmulpd_mask_rz:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vmulpd {rz-sae}, %zmm1, %zmm0, %zmm0 {%k1} {z}
; X86-NEXT:    retl
  %1 = call <8 x double> @llvm.x86.avx512.mul.pd.512(<8 x double> %a0, <8 x double> %a1, i32 3)
  %2 = bitcast i8 %mask to <8 x i1>
  %3 = select <8 x i1> %2, <8 x double> %1, <8 x double> zeroinitializer
  ret <8 x double> %3
}

define <16 x float> @test_mm512_maskz_add_round_ps_rn_sae(<16 x float> %a0, <16 x float> %a1, i16 %mask) {
; X64-LABEL: test_mm512_maskz_add_round_ps_rn_sae:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vaddps {rn-sae}, %zmm1, %zmm0, %zmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_mm512_maskz_add_round_ps_rn_sae:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vaddps {rn-sae}, %zmm1, %zmm0, %zmm0 {%k1} {z}
; X86-NEXT:    retl
  %1 = call <16 x float> @llvm.x86.avx512.add.ps.512(<16 x float> %a0, <16 x float> %a1, i32 0)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x float> %1, <16 x float> zeroinitializer
  ret <16 x float> %3
}

define <16 x float> @test_mm512_maskz_add_round_ps_rd_sae(<16 x float> %a0, <16 x float> %a1, i16 %mask) {
; X64-LABEL: test_mm512_maskz_add_round_ps_rd_sae:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vaddps {rd-sae}, %zmm1, %zmm0, %zmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_mm512_maskz_add_round_ps_rd_sae:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vaddps {rd-sae}, %zmm1, %zmm0, %zmm0 {%k1} {z}
; X86-NEXT:    retl
  %1 = call <16 x float> @llvm.x86.avx512.add.ps.512(<16 x float> %a0, <16 x float> %a1, i32 1)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x float> %1, <16 x float> zeroinitializer
  ret <16 x float> %3
}

define <16 x float> @test_mm512_maskz_add_round_ps_ru_sae(<16 x float> %a0, <16 x float> %a1, i16 %mask) {
; X64-LABEL: test_mm512_maskz_add_round_ps_ru_sae:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vaddps {ru-sae}, %zmm1, %zmm0, %zmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_mm512_maskz_add_round_ps_ru_sae:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vaddps {ru-sae}, %zmm1, %zmm0, %zmm0 {%k1} {z}
; X86-NEXT:    retl
  %1 = call <16 x float> @llvm.x86.avx512.add.ps.512(<16 x float> %a0, <16 x float> %a1, i32 2)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x float> %1, <16 x float> zeroinitializer
  ret <16 x float> %3
}

define <16 x float> @test_mm512_maskz_add_round_ps_rz_sae(<16 x float> %a0, <16 x float> %a1, i16 %mask) {
; X64-LABEL: test_mm512_maskz_add_round_ps_rz_sae:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vaddps {rz-sae}, %zmm1, %zmm0, %zmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_mm512_maskz_add_round_ps_rz_sae:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vaddps {rz-sae}, %zmm1, %zmm0, %zmm0 {%k1} {z}
; X86-NEXT:    retl
  %1 = call <16 x float> @llvm.x86.avx512.add.ps.512(<16 x float> %a0, <16 x float> %a1, i32 3)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x float> %1, <16 x float> zeroinitializer
  ret <16 x float> %3
}

define <16 x float> @test_mm512_maskz_add_round_ps_current(<16 x float> %a0, <16 x float> %a1, i16 %mask) {
; X64-LABEL: test_mm512_maskz_add_round_ps_current:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vaddps %zmm1, %zmm0, %zmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_mm512_maskz_add_round_ps_current:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vaddps %zmm1, %zmm0, %zmm0 {%k1} {z}
; X86-NEXT:    retl
  %1 = call <16 x float> @llvm.x86.avx512.add.ps.512(<16 x float> %a0, <16 x float> %a1, i32 4)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x float> %1, <16 x float> zeroinitializer
  ret <16 x float> %3
}

define <16 x float> @test_mm512_mask_add_round_ps_rn_sae(<16 x float> %a0, <16 x float> %a1, <16 x float> %src, i16 %mask) {
; X64-LABEL: test_mm512_mask_add_round_ps_rn_sae:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vaddps {rn-sae}, %zmm1, %zmm0, %zmm2 {%k1}
; X64-NEXT:    vmovaps %zmm2, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_mm512_mask_add_round_ps_rn_sae:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vaddps {rn-sae}, %zmm1, %zmm0, %zmm2 {%k1}
; X86-NEXT:    vmovaps %zmm2, %zmm0
; X86-NEXT:    retl
  %1 = call <16 x float> @llvm.x86.avx512.add.ps.512(<16 x float> %a0, <16 x float> %a1, i32 0)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x float> %1, <16 x float> %src
  ret <16 x float> %3
}

define <16 x float> @test_mm512_mask_add_round_ps_rd_sae(<16 x float> %a0, <16 x float> %a1, <16 x float> %src, i16 %mask) {
; X64-LABEL: test_mm512_mask_add_round_ps_rd_sae:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vaddps {rd-sae}, %zmm1, %zmm0, %zmm2 {%k1}
; X64-NEXT:    vmovaps %zmm2, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_mm512_mask_add_round_ps_rd_sae:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vaddps {rd-sae}, %zmm1, %zmm0, %zmm2 {%k1}
; X86-NEXT:    vmovaps %zmm2, %zmm0
; X86-NEXT:    retl
  %1 = call <16 x float> @llvm.x86.avx512.add.ps.512(<16 x float> %a0, <16 x float> %a1, i32 1)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x float> %1, <16 x float> %src
  ret <16 x float> %3
}

define <16 x float> @test_mm512_mask_add_round_ps_ru_sae(<16 x float> %a0, <16 x float> %a1, <16 x float> %src, i16 %mask) {
; X64-LABEL: test_mm512_mask_add_round_ps_ru_sae:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vaddps {ru-sae}, %zmm1, %zmm0, %zmm2 {%k1}
; X64-NEXT:    vmovaps %zmm2, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_mm512_mask_add_round_ps_ru_sae:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vaddps {ru-sae}, %zmm1, %zmm0, %zmm2 {%k1}
; X86-NEXT:    vmovaps %zmm2, %zmm0
; X86-NEXT:    retl
  %1 = call <16 x float> @llvm.x86.avx512.add.ps.512(<16 x float> %a0, <16 x float> %a1, i32 2)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x float> %1, <16 x float> %src
  ret <16 x float> %3
}

define <16 x float> @test_mm512_mask_add_round_ps_rz_sae(<16 x float> %a0, <16 x float> %a1, <16 x float> %src, i16 %mask) {
; X64-LABEL: test_mm512_mask_add_round_ps_rz_sae:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vaddps {rz-sae}, %zmm1, %zmm0, %zmm2 {%k1}
; X64-NEXT:    vmovaps %zmm2, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_mm512_mask_add_round_ps_rz_sae:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vaddps {rz-sae}, %zmm1, %zmm0, %zmm2 {%k1}
; X86-NEXT:    vmovaps %zmm2, %zmm0
; X86-NEXT:    retl
  %1 = call <16 x float> @llvm.x86.avx512.add.ps.512(<16 x float> %a0, <16 x float> %a1, i32 3)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x float> %1, <16 x float> %src
  ret <16 x float> %3
}

define <16 x float> @test_mm512_mask_add_round_ps_current(<16 x float> %a0, <16 x float> %a1, <16 x float> %src, i16 %mask) {
; X64-LABEL: test_mm512_mask_add_round_ps_current:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vaddps %zmm1, %zmm0, %zmm2 {%k1}
; X64-NEXT:    vmovaps %zmm2, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_mm512_mask_add_round_ps_current:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vaddps %zmm1, %zmm0, %zmm2 {%k1}
; X86-NEXT:    vmovaps %zmm2, %zmm0
; X86-NEXT:    retl
  %1 = call <16 x float> @llvm.x86.avx512.add.ps.512(<16 x float> %a0, <16 x float> %a1, i32 4)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x float> %1, <16 x float> %src
  ret <16 x float> %3
}

define <16 x float> @test_mm512_add_round_ps_rn_sae(<16 x float> %a0, <16 x float> %a1, i16 %mask) {
; CHECK-LABEL: test_mm512_add_round_ps_rn_sae:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vaddps {rn-sae}, %zmm1, %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %1 = call <16 x float> @llvm.x86.avx512.add.ps.512(<16 x float> %a0, <16 x float> %a1, i32 0)
  ret <16 x float> %1
}

define <16 x float> @test_mm512_add_round_ps_rd_sae(<16 x float> %a0, <16 x float> %a1, i16 %mask) {
; CHECK-LABEL: test_mm512_add_round_ps_rd_sae:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vaddps {rd-sae}, %zmm1, %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %1 = call <16 x float> @llvm.x86.avx512.add.ps.512(<16 x float> %a0, <16 x float> %a1, i32 1)
  ret <16 x float> %1
}

define <16 x float> @test_mm512_add_round_ps_ru_sae(<16 x float> %a0, <16 x float> %a1, i16 %mask) {
; CHECK-LABEL: test_mm512_add_round_ps_ru_sae:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vaddps {ru-sae}, %zmm1, %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %1 = call <16 x float> @llvm.x86.avx512.add.ps.512(<16 x float> %a0, <16 x float> %a1, i32 2)
  ret <16 x float> %1
}

define <16 x float> @test_mm512_add_round_ps_rz_sae(<16 x float> %a0, <16 x float> %a1, i16 %mask) {
; CHECK-LABEL: test_mm512_add_round_ps_rz_sae:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vaddps {rz-sae}, %zmm1, %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %1 = call <16 x float> @llvm.x86.avx512.add.ps.512(<16 x float> %a0, <16 x float> %a1, i32 3)
  ret <16 x float> %1
}

define <16 x float> @test_mm512_add_round_ps_current(<16 x float> %a0, <16 x float> %a1, i16 %mask) {
; CHECK-LABEL: test_mm512_add_round_ps_current:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vaddps %zmm1, %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %1 = call <16 x float> @llvm.x86.avx512.add.ps.512(<16 x float> %a0, <16 x float> %a1, i32 4)
  ret <16 x float> %1
}
declare <16 x float> @llvm.x86.avx512.add.ps.512(<16 x float>, <16 x float>, i32)

define <16 x float> @test_mm512_mask_sub_round_ps_rn_sae(<16 x float> %a0, <16 x float> %a1, <16 x float> %src, i16 %mask) {
; X64-LABEL: test_mm512_mask_sub_round_ps_rn_sae:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vsubps {rn-sae}, %zmm1, %zmm0, %zmm2 {%k1}
; X64-NEXT:    vmovaps %zmm2, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_mm512_mask_sub_round_ps_rn_sae:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vsubps {rn-sae}, %zmm1, %zmm0, %zmm2 {%k1}
; X86-NEXT:    vmovaps %zmm2, %zmm0
; X86-NEXT:    retl
  %1 = call <16 x float> @llvm.x86.avx512.sub.ps.512(<16 x float> %a0, <16 x float> %a1, i32 0)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x float> %1, <16 x float> %src
  ret <16 x float> %3
}

define <16 x float> @test_mm512_mask_sub_round_ps_rd_sae(<16 x float> %a0, <16 x float> %a1, <16 x float> %src, i16 %mask) {
; X64-LABEL: test_mm512_mask_sub_round_ps_rd_sae:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vsubps {rd-sae}, %zmm1, %zmm0, %zmm2 {%k1}
; X64-NEXT:    vmovaps %zmm2, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_mm512_mask_sub_round_ps_rd_sae:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vsubps {rd-sae}, %zmm1, %zmm0, %zmm2 {%k1}
; X86-NEXT:    vmovaps %zmm2, %zmm0
; X86-NEXT:    retl
  %1 = call <16 x float> @llvm.x86.avx512.sub.ps.512(<16 x float> %a0, <16 x float> %a1, i32 1)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x float> %1, <16 x float> %src
  ret <16 x float> %3
}

define <16 x float> @test_mm512_mask_sub_round_ps_ru_sae(<16 x float> %a0, <16 x float> %a1, <16 x float> %src, i16 %mask) {
; X64-LABEL: test_mm512_mask_sub_round_ps_ru_sae:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vsubps {ru-sae}, %zmm1, %zmm0, %zmm2 {%k1}
; X64-NEXT:    vmovaps %zmm2, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_mm512_mask_sub_round_ps_ru_sae:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vsubps {ru-sae}, %zmm1, %zmm0, %zmm2 {%k1}
; X86-NEXT:    vmovaps %zmm2, %zmm0
; X86-NEXT:    retl
  %1 = call <16 x float> @llvm.x86.avx512.sub.ps.512(<16 x float> %a0, <16 x float> %a1, i32 2)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x float> %1, <16 x float> %src
  ret <16 x float> %3
}

define <16 x float> @test_mm512_mask_sub_round_ps_rz_sae(<16 x float> %a0, <16 x float> %a1, <16 x float> %src, i16 %mask) {
; X64-LABEL: test_mm512_mask_sub_round_ps_rz_sae:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vsubps {rz-sae}, %zmm1, %zmm0, %zmm2 {%k1}
; X64-NEXT:    vmovaps %zmm2, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_mm512_mask_sub_round_ps_rz_sae:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vsubps {rz-sae}, %zmm1, %zmm0, %zmm2 {%k1}
; X86-NEXT:    vmovaps %zmm2, %zmm0
; X86-NEXT:    retl
  %1 = call <16 x float> @llvm.x86.avx512.sub.ps.512(<16 x float> %a0, <16 x float> %a1, i32 3)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x float> %1, <16 x float> %src
  ret <16 x float> %3
}

define <16 x float> @test_mm512_mask_sub_round_ps_current(<16 x float> %a0, <16 x float> %a1, <16 x float> %src, i16 %mask) {
; X64-LABEL: test_mm512_mask_sub_round_ps_current:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vsubps %zmm1, %zmm0, %zmm2 {%k1}
; X64-NEXT:    vmovaps %zmm2, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_mm512_mask_sub_round_ps_current:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vsubps %zmm1, %zmm0, %zmm2 {%k1}
; X86-NEXT:    vmovaps %zmm2, %zmm0
; X86-NEXT:    retl
  %1 = call <16 x float> @llvm.x86.avx512.sub.ps.512(<16 x float> %a0, <16 x float> %a1, i32 4)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x float> %1, <16 x float> %src
  ret <16 x float> %3
}

define <16 x float> @test_mm512_sub_round_ps_rn_sae(<16 x float> %a0, <16 x float> %a1, i16 %mask) {
; CHECK-LABEL: test_mm512_sub_round_ps_rn_sae:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vsubps {rn-sae}, %zmm1, %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %1 = call <16 x float> @llvm.x86.avx512.sub.ps.512(<16 x float> %a0, <16 x float> %a1, i32 0)
  ret <16 x float> %1
}

define <16 x float> @test_mm512_sub_round_ps_rd_sae(<16 x float> %a0, <16 x float> %a1, i16 %mask) {
; CHECK-LABEL: test_mm512_sub_round_ps_rd_sae:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vsubps {rd-sae}, %zmm1, %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %1 = call <16 x float> @llvm.x86.avx512.sub.ps.512(<16 x float> %a0, <16 x float> %a1, i32 1)
  ret <16 x float> %1
}

define <16 x float> @test_mm512_sub_round_ps_ru_sae(<16 x float> %a0, <16 x float> %a1, i16 %mask) {
; CHECK-LABEL: test_mm512_sub_round_ps_ru_sae:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vsubps {ru-sae}, %zmm1, %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %1 = call <16 x float> @llvm.x86.avx512.sub.ps.512(<16 x float> %a0, <16 x float> %a1, i32 2)
  ret <16 x float> %1
}

define <16 x float> @test_mm512_sub_round_ps_rz_sae(<16 x float> %a0, <16 x float> %a1, i16 %mask) {
; CHECK-LABEL: test_mm512_sub_round_ps_rz_sae:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vsubps {rz-sae}, %zmm1, %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %1 = call <16 x float> @llvm.x86.avx512.sub.ps.512(<16 x float> %a0, <16 x float> %a1, i32 3)
  ret <16 x float> %1
}

define <16 x float> @test_mm512_sub_round_ps_current(<16 x float> %a0, <16 x float> %a1, i16 %mask) {
; CHECK-LABEL: test_mm512_sub_round_ps_current:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vsubps %zmm1, %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %1 = call <16 x float> @llvm.x86.avx512.sub.ps.512(<16 x float> %a0, <16 x float> %a1, i32 4)
  ret <16 x float> %1
}

define <16 x float> @test_mm512_maskz_div_round_ps_rn_sae(<16 x float> %a0, <16 x float> %a1, i16 %mask) {
; X64-LABEL: test_mm512_maskz_div_round_ps_rn_sae:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vdivps {rn-sae}, %zmm1, %zmm0, %zmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_mm512_maskz_div_round_ps_rn_sae:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vdivps {rn-sae}, %zmm1, %zmm0, %zmm0 {%k1} {z}
; X86-NEXT:    retl
  %1 = call <16 x float> @llvm.x86.avx512.div.ps.512(<16 x float> %a0, <16 x float> %a1, i32 0)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x float> %1, <16 x float> zeroinitializer
  ret <16 x float> %3
}

define <16 x float> @test_mm512_maskz_div_round_ps_rd_sae(<16 x float> %a0, <16 x float> %a1, i16 %mask) {
; X64-LABEL: test_mm512_maskz_div_round_ps_rd_sae:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vdivps {rd-sae}, %zmm1, %zmm0, %zmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_mm512_maskz_div_round_ps_rd_sae:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vdivps {rd-sae}, %zmm1, %zmm0, %zmm0 {%k1} {z}
; X86-NEXT:    retl
  %1 = call <16 x float> @llvm.x86.avx512.div.ps.512(<16 x float> %a0, <16 x float> %a1, i32 1)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x float> %1, <16 x float> zeroinitializer
  ret <16 x float> %3
}

define <16 x float> @test_mm512_maskz_div_round_ps_ru_sae(<16 x float> %a0, <16 x float> %a1, i16 %mask) {
; X64-LABEL: test_mm512_maskz_div_round_ps_ru_sae:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vdivps {ru-sae}, %zmm1, %zmm0, %zmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_mm512_maskz_div_round_ps_ru_sae:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vdivps {ru-sae}, %zmm1, %zmm0, %zmm0 {%k1} {z}
; X86-NEXT:    retl
  %1 = call <16 x float> @llvm.x86.avx512.div.ps.512(<16 x float> %a0, <16 x float> %a1, i32 2)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x float> %1, <16 x float> zeroinitializer
  ret <16 x float> %3
}

define <16 x float> @test_mm512_maskz_div_round_ps_rz_sae(<16 x float> %a0, <16 x float> %a1, i16 %mask) {
; X64-LABEL: test_mm512_maskz_div_round_ps_rz_sae:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vdivps {rz-sae}, %zmm1, %zmm0, %zmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_mm512_maskz_div_round_ps_rz_sae:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vdivps {rz-sae}, %zmm1, %zmm0, %zmm0 {%k1} {z}
; X86-NEXT:    retl
  %1 = call <16 x float> @llvm.x86.avx512.div.ps.512(<16 x float> %a0, <16 x float> %a1, i32 3)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x float> %1, <16 x float> zeroinitializer
  ret <16 x float> %3
}

define <16 x float> @test_mm512_maskz_div_round_ps_current(<16 x float> %a0, <16 x float> %a1, i16 %mask) {
; X64-LABEL: test_mm512_maskz_div_round_ps_current:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vdivps %zmm1, %zmm0, %zmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_mm512_maskz_div_round_ps_current:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vdivps %zmm1, %zmm0, %zmm0 {%k1} {z}
; X86-NEXT:    retl
  %1 = call <16 x float> @llvm.x86.avx512.div.ps.512(<16 x float> %a0, <16 x float> %a1, i32 4)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x float> %1, <16 x float> zeroinitializer
  ret <16 x float> %3
}

define <16 x float> @test_mm512_mask_div_round_ps_rn_sae(<16 x float> %a0, <16 x float> %a1, <16 x float> %src, i16 %mask) {
; X64-LABEL: test_mm512_mask_div_round_ps_rn_sae:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vdivps {rn-sae}, %zmm1, %zmm0, %zmm2 {%k1}
; X64-NEXT:    vmovaps %zmm2, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_mm512_mask_div_round_ps_rn_sae:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vdivps {rn-sae}, %zmm1, %zmm0, %zmm2 {%k1}
; X86-NEXT:    vmovaps %zmm2, %zmm0
; X86-NEXT:    retl
  %1 = call <16 x float> @llvm.x86.avx512.div.ps.512(<16 x float> %a0, <16 x float> %a1, i32 0)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x float> %1, <16 x float> %src
  ret <16 x float> %3
}

define <16 x float> @test_mm512_mask_div_round_ps_rd_sae(<16 x float> %a0, <16 x float> %a1, <16 x float> %src, i16 %mask) {
; X64-LABEL: test_mm512_mask_div_round_ps_rd_sae:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vdivps {rd-sae}, %zmm1, %zmm0, %zmm2 {%k1}
; X64-NEXT:    vmovaps %zmm2, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_mm512_mask_div_round_ps_rd_sae:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vdivps {rd-sae}, %zmm1, %zmm0, %zmm2 {%k1}
; X86-NEXT:    vmovaps %zmm2, %zmm0
; X86-NEXT:    retl
  %1 = call <16 x float> @llvm.x86.avx512.div.ps.512(<16 x float> %a0, <16 x float> %a1, i32 1)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x float> %1, <16 x float> %src
  ret <16 x float> %3
}

define <16 x float> @test_mm512_mask_div_round_ps_ru_sae(<16 x float> %a0, <16 x float> %a1, <16 x float> %src, i16 %mask) {
; X64-LABEL: test_mm512_mask_div_round_ps_ru_sae:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vdivps {ru-sae}, %zmm1, %zmm0, %zmm2 {%k1}
; X64-NEXT:    vmovaps %zmm2, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_mm512_mask_div_round_ps_ru_sae:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vdivps {ru-sae}, %zmm1, %zmm0, %zmm2 {%k1}
; X86-NEXT:    vmovaps %zmm2, %zmm0
; X86-NEXT:    retl
  %1 = call <16 x float> @llvm.x86.avx512.div.ps.512(<16 x float> %a0, <16 x float> %a1, i32 2)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x float> %1, <16 x float> %src
  ret <16 x float> %3
}

define <16 x float> @test_mm512_mask_div_round_ps_rz_sae(<16 x float> %a0, <16 x float> %a1, <16 x float> %src, i16 %mask) {
; X64-LABEL: test_mm512_mask_div_round_ps_rz_sae:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vdivps {rz-sae}, %zmm1, %zmm0, %zmm2 {%k1}
; X64-NEXT:    vmovaps %zmm2, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_mm512_mask_div_round_ps_rz_sae:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vdivps {rz-sae}, %zmm1, %zmm0, %zmm2 {%k1}
; X86-NEXT:    vmovaps %zmm2, %zmm0
; X86-NEXT:    retl
  %1 = call <16 x float> @llvm.x86.avx512.div.ps.512(<16 x float> %a0, <16 x float> %a1, i32 3)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x float> %1, <16 x float> %src
  ret <16 x float> %3
}

define <16 x float> @test_mm512_mask_div_round_ps_current(<16 x float> %a0, <16 x float> %a1, <16 x float> %src, i16 %mask) {
; X64-LABEL: test_mm512_mask_div_round_ps_current:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vdivps %zmm1, %zmm0, %zmm2 {%k1}
; X64-NEXT:    vmovaps %zmm2, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_mm512_mask_div_round_ps_current:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vdivps %zmm1, %zmm0, %zmm2 {%k1}
; X86-NEXT:    vmovaps %zmm2, %zmm0
; X86-NEXT:    retl
  %1 = call <16 x float> @llvm.x86.avx512.div.ps.512(<16 x float> %a0, <16 x float> %a1, i32 4)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x float> %1, <16 x float> %src
  ret <16 x float> %3
}

define <16 x float> @test_mm512_div_round_ps_rn_sae(<16 x float> %a0, <16 x float> %a1, i16 %mask) {
; CHECK-LABEL: test_mm512_div_round_ps_rn_sae:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vdivps {rn-sae}, %zmm1, %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %1 = call <16 x float> @llvm.x86.avx512.div.ps.512(<16 x float> %a0, <16 x float> %a1, i32 0)
  ret <16 x float> %1
}

define <16 x float> @test_mm512_div_round_ps_rd_sae(<16 x float> %a0, <16 x float> %a1, i16 %mask) {
; CHECK-LABEL: test_mm512_div_round_ps_rd_sae:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vdivps {rd-sae}, %zmm1, %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %1 = call <16 x float> @llvm.x86.avx512.div.ps.512(<16 x float> %a0, <16 x float> %a1, i32 1)
  ret <16 x float> %1
}

define <16 x float> @test_mm512_div_round_ps_ru_sae(<16 x float> %a0, <16 x float> %a1, i16 %mask) {
; CHECK-LABEL: test_mm512_div_round_ps_ru_sae:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vdivps {ru-sae}, %zmm1, %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %1 = call <16 x float> @llvm.x86.avx512.div.ps.512(<16 x float> %a0, <16 x float> %a1, i32 2)
  ret <16 x float> %1
}

define <16 x float> @test_mm512_div_round_ps_rz_sae(<16 x float> %a0, <16 x float> %a1, i16 %mask) {
; CHECK-LABEL: test_mm512_div_round_ps_rz_sae:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vdivps {rz-sae}, %zmm1, %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %1 = call <16 x float> @llvm.x86.avx512.div.ps.512(<16 x float> %a0, <16 x float> %a1, i32 3)
  ret <16 x float> %1
}

define <16 x float> @test_mm512_div_round_ps_current(<16 x float> %a0, <16 x float> %a1, i16 %mask) {
; CHECK-LABEL: test_mm512_div_round_ps_current:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vdivps %zmm1, %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %1 = call <16 x float> @llvm.x86.avx512.div.ps.512(<16 x float> %a0, <16 x float> %a1, i32 4)
  ret <16 x float> %1
}
declare <16 x float> @llvm.x86.avx512.div.ps.512(<16 x float>, <16 x float>, i32)

define <16 x float> @test_mm512_maskz_min_round_ps_sae(<16 x float> %a0, <16 x float> %a1, i16 %mask) {
; X64-LABEL: test_mm512_maskz_min_round_ps_sae:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vminps {sae}, %zmm1, %zmm0, %zmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_mm512_maskz_min_round_ps_sae:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vminps {sae}, %zmm1, %zmm0, %zmm0 {%k1} {z}
; X86-NEXT:    retl
  %1 = call <16 x float> @llvm.x86.avx512.min.ps.512(<16 x float> %a0, <16 x float> %a1, i32 8)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x float> %1, <16 x float> zeroinitializer
  ret <16 x float> %3
}

define <16 x float> @test_mm512_maskz_min_round_ps_current(<16 x float> %a0, <16 x float> %a1, i16 %mask) {
; X64-LABEL: test_mm512_maskz_min_round_ps_current:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vminps %zmm1, %zmm0, %zmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_mm512_maskz_min_round_ps_current:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vminps %zmm1, %zmm0, %zmm0 {%k1} {z}
; X86-NEXT:    retl
  %1 = call <16 x float> @llvm.x86.avx512.min.ps.512(<16 x float> %a0, <16 x float> %a1, i32 4)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x float> %1, <16 x float> zeroinitializer
  ret <16 x float> %3
}

define <16 x float> @test_mm512_mask_min_round_ps_sae(<16 x float> %a0, <16 x float> %a1, <16 x float> %src, i16 %mask) {
; X64-LABEL: test_mm512_mask_min_round_ps_sae:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vminps {sae}, %zmm1, %zmm0, %zmm2 {%k1}
; X64-NEXT:    vmovaps %zmm2, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_mm512_mask_min_round_ps_sae:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vminps {sae}, %zmm1, %zmm0, %zmm2 {%k1}
; X86-NEXT:    vmovaps %zmm2, %zmm0
; X86-NEXT:    retl
  %1 = call <16 x float> @llvm.x86.avx512.min.ps.512(<16 x float> %a0, <16 x float> %a1, i32 8)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x float> %1, <16 x float> %src
  ret <16 x float> %3
}

define <16 x float> @test_mm512_mask_min_round_ps_current(<16 x float> %a0, <16 x float> %a1, <16 x float> %src, i16 %mask) {
; X64-LABEL: test_mm512_mask_min_round_ps_current:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vminps %zmm1, %zmm0, %zmm2 {%k1}
; X64-NEXT:    vmovaps %zmm2, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_mm512_mask_min_round_ps_current:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vminps %zmm1, %zmm0, %zmm2 {%k1}
; X86-NEXT:    vmovaps %zmm2, %zmm0
; X86-NEXT:    retl
  %1 = call <16 x float> @llvm.x86.avx512.min.ps.512(<16 x float> %a0, <16 x float> %a1, i32 4)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x float> %1, <16 x float> %src
  ret <16 x float> %3
}

define <16 x float> @test_mm512_min_round_ps_sae(<16 x float> %a0, <16 x float> %a1, i16 %mask) {
; CHECK-LABEL: test_mm512_min_round_ps_sae:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vminps {sae}, %zmm1, %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %1 = call <16 x float> @llvm.x86.avx512.min.ps.512(<16 x float> %a0, <16 x float> %a1, i32 8)
  ret <16 x float> %1
}

define <16 x float> @test_mm512_min_round_ps_current(<16 x float> %a0, <16 x float> %a1, i16 %mask) {
; CHECK-LABEL: test_mm512_min_round_ps_current:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vminps %zmm1, %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %1 = call <16 x float> @llvm.x86.avx512.min.ps.512(<16 x float> %a0, <16 x float> %a1, i32 4)
  ret <16 x float> %1
}
declare <16 x float> @llvm.x86.avx512.min.ps.512(<16 x float>, <16 x float>, i32)

define <16 x float> @test_mm512_maskz_max_round_ps_sae(<16 x float> %a0, <16 x float> %a1, i16 %mask) {
; X64-LABEL: test_mm512_maskz_max_round_ps_sae:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vmaxps {sae}, %zmm1, %zmm0, %zmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_mm512_maskz_max_round_ps_sae:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vmaxps {sae}, %zmm1, %zmm0, %zmm0 {%k1} {z}
; X86-NEXT:    retl
  %1 = call <16 x float> @llvm.x86.avx512.max.ps.512(<16 x float> %a0, <16 x float> %a1, i32 8)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x float> %1, <16 x float> zeroinitializer
  ret <16 x float> %3
}

define <16 x float> @test_mm512_maskz_max_round_ps_current(<16 x float> %a0, <16 x float> %a1, i16 %mask) {
; X64-LABEL: test_mm512_maskz_max_round_ps_current:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vmaxps %zmm1, %zmm0, %zmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_mm512_maskz_max_round_ps_current:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vmaxps %zmm1, %zmm0, %zmm0 {%k1} {z}
; X86-NEXT:    retl
  %1 = call <16 x float> @llvm.x86.avx512.max.ps.512(<16 x float> %a0, <16 x float> %a1, i32 4)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x float> %1, <16 x float> zeroinitializer
  ret <16 x float> %3
}

define <16 x float> @test_mm512_mask_max_round_ps_sae(<16 x float> %a0, <16 x float> %a1, <16 x float> %src, i16 %mask) {
; X64-LABEL: test_mm512_mask_max_round_ps_sae:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vmaxps {sae}, %zmm1, %zmm0, %zmm2 {%k1}
; X64-NEXT:    vmovaps %zmm2, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_mm512_mask_max_round_ps_sae:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vmaxps {sae}, %zmm1, %zmm0, %zmm2 {%k1}
; X86-NEXT:    vmovaps %zmm2, %zmm0
; X86-NEXT:    retl
  %1 = call <16 x float> @llvm.x86.avx512.max.ps.512(<16 x float> %a0, <16 x float> %a1, i32 8)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x float> %1, <16 x float> %src
  ret <16 x float> %3
}

define <16 x float> @test_mm512_mask_max_round_ps_current(<16 x float> %a0, <16 x float> %a1, <16 x float> %src, i16 %mask) {
; X64-LABEL: test_mm512_mask_max_round_ps_current:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vmaxps %zmm1, %zmm0, %zmm2 {%k1}
; X64-NEXT:    vmovaps %zmm2, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_mm512_mask_max_round_ps_current:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vmaxps %zmm1, %zmm0, %zmm2 {%k1}
; X86-NEXT:    vmovaps %zmm2, %zmm0
; X86-NEXT:    retl
  %1 = call <16 x float> @llvm.x86.avx512.max.ps.512(<16 x float> %a0, <16 x float> %a1, i32 4)
  %2 = bitcast i16 %mask to <16 x i1>
  %3 = select <16 x i1> %2, <16 x float> %1, <16 x float> %src
  ret <16 x float> %3
}

define <16 x float> @test_mm512_max_round_ps_sae(<16 x float> %a0, <16 x float> %a1, i16 %mask) {
; CHECK-LABEL: test_mm512_max_round_ps_sae:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vmaxps {sae}, %zmm1, %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %1 = call <16 x float> @llvm.x86.avx512.max.ps.512(<16 x float> %a0, <16 x float> %a1, i32 8)
  ret <16 x float> %1
}

define <16 x float> @test_mm512_max_round_ps_current(<16 x float> %a0, <16 x float> %a1, i16 %mask) {
; CHECK-LABEL: test_mm512_max_round_ps_current:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vmaxps %zmm1, %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %1 = call <16 x float> @llvm.x86.avx512.max.ps.512(<16 x float> %a0, <16 x float> %a1, i32 4)
  ret <16 x float> %1
}
declare <16 x float> @llvm.x86.avx512.max.ps.512(<16 x float>, <16 x float>, i32)

declare <4 x float> @llvm.x86.avx512.mask.add.ss.round(<4 x float>, <4 x float>, <4 x float>, i8, i32) nounwind readnone

define <4 x float> @test_mask_add_ss_rn(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2, i8 %mask) {
; X64-LABEL: test_mask_add_ss_rn:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vaddss {rn-sae}, %xmm1, %xmm0, %xmm2 {%k1}
; X64-NEXT:    vmovaps %xmm2, %xmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_mask_add_ss_rn:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vaddss {rn-sae}, %xmm1, %xmm0, %xmm2 {%k1}
; X86-NEXT:    vmovaps %xmm2, %xmm0
; X86-NEXT:    retl
  %res = call <4 x float> @llvm.x86.avx512.mask.add.ss.round(<4 x float>%a0, <4 x float> %a1, <4 x float> %a2, i8 %mask, i32 0)
  ret <4 x float> %res
}

define <4 x float> @test_mask_add_ss_rd(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2, i8 %mask) {
; X64-LABEL: test_mask_add_ss_rd:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vaddss {rd-sae}, %xmm1, %xmm0, %xmm2 {%k1}
; X64-NEXT:    vmovaps %xmm2, %xmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_mask_add_ss_rd:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vaddss {rd-sae}, %xmm1, %xmm0, %xmm2 {%k1}
; X86-NEXT:    vmovaps %xmm2, %xmm0
; X86-NEXT:    retl
  %res = call <4 x float> @llvm.x86.avx512.mask.add.ss.round(<4 x float>%a0, <4 x float> %a1, <4 x float> %a2, i8 %mask, i32 1)
  ret <4 x float> %res
}

define <4 x float> @test_mask_add_ss_ru(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2, i8 %mask) {
; X64-LABEL: test_mask_add_ss_ru:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vaddss {ru-sae}, %xmm1, %xmm0, %xmm2 {%k1}
; X64-NEXT:    vmovaps %xmm2, %xmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_mask_add_ss_ru:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vaddss {ru-sae}, %xmm1, %xmm0, %xmm2 {%k1}
; X86-NEXT:    vmovaps %xmm2, %xmm0
; X86-NEXT:    retl
  %res = call <4 x float> @llvm.x86.avx512.mask.add.ss.round(<4 x float>%a0, <4 x float> %a1, <4 x float> %a2, i8 %mask, i32 2)
  ret <4 x float> %res
}

define <4 x float> @test_mask_add_ss_rz(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2, i8 %mask) {
; X64-LABEL: test_mask_add_ss_rz:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vaddss {rz-sae}, %xmm1, %xmm0, %xmm2 {%k1}
; X64-NEXT:    vmovaps %xmm2, %xmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_mask_add_ss_rz:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vaddss {rz-sae}, %xmm1, %xmm0, %xmm2 {%k1}
; X86-NEXT:    vmovaps %xmm2, %xmm0
; X86-NEXT:    retl
  %res = call <4 x float> @llvm.x86.avx512.mask.add.ss.round(<4 x float>%a0, <4 x float> %a1, <4 x float> %a2, i8 %mask, i32 3)
  ret <4 x float> %res
}

define <4 x float> @test_mask_add_ss_current(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2, i8 %mask) {
; X64-LABEL: test_mask_add_ss_current:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vaddss %xmm1, %xmm0, %xmm2 {%k1}
; X64-NEXT:    vmovaps %xmm2, %xmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_mask_add_ss_current:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vaddss %xmm1, %xmm0, %xmm2 {%k1}
; X86-NEXT:    vmovaps %xmm2, %xmm0
; X86-NEXT:    retl
  %res = call <4 x float> @llvm.x86.avx512.mask.add.ss.round(<4 x float>%a0, <4 x float> %a1, <4 x float> %a2, i8 %mask, i32 4)
  ret <4 x float> %res
}

define <4 x float> @test_maskz_add_ss_rn(<4 x float> %a0, <4 x float> %a1, i8 %mask) {
; X64-LABEL: test_maskz_add_ss_rn:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vaddss {rn-sae}, %xmm1, %xmm0, %xmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_maskz_add_ss_rn:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vaddss {rn-sae}, %xmm1, %xmm0, %xmm0 {%k1} {z}
; X86-NEXT:    retl
  %res = call <4 x float> @llvm.x86.avx512.mask.add.ss.round(<4 x float>%a0, <4 x float> %a1, <4 x float> zeroinitializer, i8 %mask, i32 0)
  ret <4 x float> %res
}

define <4 x float> @test_add_ss_rn(<4 x float> %a0, <4 x float> %a1) {
; CHECK-LABEL: test_add_ss_rn:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vaddss {rn-sae}, %xmm1, %xmm0, %xmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %res = call <4 x float> @llvm.x86.avx512.mask.add.ss.round(<4 x float>%a0, <4 x float> %a1, <4 x float> zeroinitializer, i8 -1, i32 0)
  ret <4 x float> %res
}

define <4 x float> @test_mask_add_ss_current_memfold(<4 x float> %a0, float* %a1, <4 x float> %a2, i8 %mask) {
; X64-LABEL: test_mask_add_ss_current_memfold:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1
; X64-NEXT:    vaddss (%rdi), %xmm0, %xmm1 {%k1}
; X64-NEXT:    vmovaps %xmm1, %xmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_mask_add_ss_current_memfold:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    vaddss (%eax), %xmm0, %xmm1 {%k1}
; X86-NEXT:    vmovaps %xmm1, %xmm0
; X86-NEXT:    retl
  %a1.val = load float, float* %a1
  %a1v0 = insertelement <4 x float> undef, float %a1.val, i32 0
  %a1v1 = insertelement <4 x float> %a1v0, float 0.000000e+00, i32 1
  %a1v2 = insertelement <4 x float> %a1v1, float 0.000000e+00, i32 2
  %a1v  = insertelement <4 x float> %a1v2, float 0.000000e+00, i32 3
  %res = call <4 x float> @llvm.x86.avx512.mask.add.ss.round(<4 x float>%a0, <4 x float> %a1v, <4 x float> %a2, i8 %mask, i32 4)
  ret <4 x float> %res
}

define <4 x float> @test_maskz_add_ss_current_memfold(<4 x float> %a0, float* %a1, i8 %mask) {
; X64-LABEL: test_maskz_add_ss_current_memfold:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1
; X64-NEXT:    vaddss (%rdi), %xmm0, %xmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_maskz_add_ss_current_memfold:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    vaddss (%eax), %xmm0, %xmm0 {%k1} {z}
; X86-NEXT:    retl
  %a1.val = load float, float* %a1
  %a1v0 = insertelement <4 x float> undef, float %a1.val, i32 0
  %a1v1 = insertelement <4 x float> %a1v0, float 0.000000e+00, i32 1
  %a1v2 = insertelement <4 x float> %a1v1, float 0.000000e+00, i32 2
  %a1v  = insertelement <4 x float> %a1v2, float 0.000000e+00, i32 3
  %res = call <4 x float> @llvm.x86.avx512.mask.add.ss.round(<4 x float>%a0, <4 x float> %a1v, <4 x float> zeroinitializer, i8 %mask, i32 4)
  ret <4 x float> %res
}

declare <2 x double> @llvm.x86.avx512.mask.add.sd.round(<2 x double>, <2 x double>, <2 x double>, i8, i32) nounwind readnone

define <2 x double> @test_mask_add_sd_rn(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2, i8 %mask) {
; X64-LABEL: test_mask_add_sd_rn:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vaddsd {rn-sae}, %xmm1, %xmm0, %xmm2 {%k1}
; X64-NEXT:    vmovapd %xmm2, %xmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_mask_add_sd_rn:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vaddsd {rn-sae}, %xmm1, %xmm0, %xmm2 {%k1}
; X86-NEXT:    vmovapd %xmm2, %xmm0
; X86-NEXT:    retl
  %res = call <2 x double> @llvm.x86.avx512.mask.add.sd.round(<2 x double>%a0, <2 x double> %a1, <2 x double> %a2, i8 %mask, i32 0)
  ret <2 x double> %res
}

define <2 x double> @test_mask_add_sd_rd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2, i8 %mask) {
; X64-LABEL: test_mask_add_sd_rd:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vaddsd {rd-sae}, %xmm1, %xmm0, %xmm2 {%k1}
; X64-NEXT:    vmovapd %xmm2, %xmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_mask_add_sd_rd:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vaddsd {rd-sae}, %xmm1, %xmm0, %xmm2 {%k1}
; X86-NEXT:    vmovapd %xmm2, %xmm0
; X86-NEXT:    retl
  %res = call <2 x double> @llvm.x86.avx512.mask.add.sd.round(<2 x double>%a0, <2 x double> %a1, <2 x double> %a2, i8 %mask, i32 1)
  ret <2 x double> %res
}

define <2 x double> @test_mask_add_sd_ru(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2, i8 %mask) {
; X64-LABEL: test_mask_add_sd_ru:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vaddsd {ru-sae}, %xmm1, %xmm0, %xmm2 {%k1}
; X64-NEXT:    vmovapd %xmm2, %xmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_mask_add_sd_ru:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vaddsd {ru-sae}, %xmm1, %xmm0, %xmm2 {%k1}
; X86-NEXT:    vmovapd %xmm2, %xmm0
; X86-NEXT:    retl
  %res = call <2 x double> @llvm.x86.avx512.mask.add.sd.round(<2 x double>%a0, <2 x double> %a1, <2 x double> %a2, i8 %mask, i32 2)
  ret <2 x double> %res
}

define <2 x double> @test_mask_add_sd_rz(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2, i8 %mask) {
; X64-LABEL: test_mask_add_sd_rz:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vaddsd {rz-sae}, %xmm1, %xmm0, %xmm2 {%k1}
; X64-NEXT:    vmovapd %xmm2, %xmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_mask_add_sd_rz:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vaddsd {rz-sae}, %xmm1, %xmm0, %xmm2 {%k1}
; X86-NEXT:    vmovapd %xmm2, %xmm0
; X86-NEXT:    retl
  %res = call <2 x double> @llvm.x86.avx512.mask.add.sd.round(<2 x double>%a0, <2 x double> %a1, <2 x double> %a2, i8 %mask, i32 3)
  ret <2 x double> %res
}

define <2 x double> @test_mask_add_sd_current(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2, i8 %mask) {
; X64-LABEL: test_mask_add_sd_current:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vaddsd %xmm1, %xmm0, %xmm2 {%k1}
; X64-NEXT:    vmovapd %xmm2, %xmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_mask_add_sd_current:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vaddsd %xmm1, %xmm0, %xmm2 {%k1}
; X86-NEXT:    vmovapd %xmm2, %xmm0
; X86-NEXT:    retl
  %res = call <2 x double> @llvm.x86.avx512.mask.add.sd.round(<2 x double>%a0, <2 x double> %a1, <2 x double> %a2, i8 %mask, i32 4)
  ret <2 x double> %res
}

define <2 x double> @test_maskz_add_sd_rn(<2 x double> %a0, <2 x double> %a1, i8 %mask) {
; X64-LABEL: test_maskz_add_sd_rn:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vaddsd {rn-sae}, %xmm1, %xmm0, %xmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_maskz_add_sd_rn:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vaddsd {rn-sae}, %xmm1, %xmm0, %xmm0 {%k1} {z}
; X86-NEXT:    retl
  %res = call <2 x double> @llvm.x86.avx512.mask.add.sd.round(<2 x double>%a0, <2 x double> %a1, <2 x double> zeroinitializer, i8 %mask, i32 0)
  ret <2 x double> %res
}

define <2 x double> @test_add_sd_rn(<2 x double> %a0, <2 x double> %a1) {
; CHECK-LABEL: test_add_sd_rn:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vaddsd {rn-sae}, %xmm1, %xmm0, %xmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %res = call <2 x double> @llvm.x86.avx512.mask.add.sd.round(<2 x double>%a0, <2 x double> %a1, <2 x double> zeroinitializer, i8 -1, i32 0)
  ret <2 x double> %res
}

define <2 x double> @test_mask_add_sd_current_memfold(<2 x double> %a0, double* %a1, <2 x double> %a2, i8 %mask) {
; X64-LABEL: test_mask_add_sd_current_memfold:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1
; X64-NEXT:    vaddsd (%rdi), %xmm0, %xmm1 {%k1}
; X64-NEXT:    vmovapd %xmm1, %xmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_mask_add_sd_current_memfold:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    vaddsd (%eax), %xmm0, %xmm1 {%k1}
; X86-NEXT:    vmovapd %xmm1, %xmm0
; X86-NEXT:    retl
  %a1.val = load double, double* %a1
  %a1v0 = insertelement <2 x double> undef, double %a1.val, i32 0
  %a1v = insertelement <2 x double> %a1v0, double 0.000000e+00, i32 1
  %res = call <2 x double> @llvm.x86.avx512.mask.add.sd.round(<2 x double>%a0, <2 x double> %a1v, <2 x double> %a2, i8 %mask, i32 4)
  ret <2 x double> %res
}

define <2 x double> @test_maskz_add_sd_current_memfold(<2 x double> %a0, double* %a1, i8 %mask) {
; X64-LABEL: test_maskz_add_sd_current_memfold:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1
; X64-NEXT:    vaddsd (%rdi), %xmm0, %xmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_maskz_add_sd_current_memfold:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    vaddsd (%eax), %xmm0, %xmm0 {%k1} {z}
; X86-NEXT:    retl
  %a1.val = load double, double* %a1
  %a1v0 = insertelement <2 x double> undef, double %a1.val, i32 0
  %a1v = insertelement <2 x double> %a1v0, double 0.000000e+00, i32 1
  %res = call <2 x double> @llvm.x86.avx512.mask.add.sd.round(<2 x double>%a0, <2 x double> %a1v, <2 x double> zeroinitializer, i8 %mask, i32 4)
  ret <2 x double> %res
}

declare <4 x float> @llvm.x86.avx512.mask.max.ss.round(<4 x float>, <4 x float>, <4 x float>, i8, i32) nounwind readnone

define <4 x float> @test_mask_max_ss_sae(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2, i8 %mask) {
; X64-LABEL: test_mask_max_ss_sae:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vmaxss {sae}, %xmm1, %xmm0, %xmm2 {%k1}
; X64-NEXT:    vmovaps %xmm2, %xmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_mask_max_ss_sae:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vmaxss {sae}, %xmm1, %xmm0, %xmm2 {%k1}
; X86-NEXT:    vmovaps %xmm2, %xmm0
; X86-NEXT:    retl
  %res = call <4 x float> @llvm.x86.avx512.mask.max.ss.round(<4 x float>%a0, <4 x float> %a1, <4 x float> %a2, i8 %mask, i32 8)
  ret <4 x float> %res
}

define <4 x float> @test_maskz_max_ss_sae(<4 x float> %a0, <4 x float> %a1, i8 %mask) {
; X64-LABEL: test_maskz_max_ss_sae:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vmaxss {sae}, %xmm1, %xmm0, %xmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_maskz_max_ss_sae:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vmaxss {sae}, %xmm1, %xmm0, %xmm0 {%k1} {z}
; X86-NEXT:    retl
  %res = call <4 x float> @llvm.x86.avx512.mask.max.ss.round(<4 x float>%a0, <4 x float> %a1, <4 x float> zeroinitializer, i8 %mask, i32 8)
  ret <4 x float> %res
}

define <4 x float> @test_max_ss_sae(<4 x float> %a0, <4 x float> %a1) {
; CHECK-LABEL: test_max_ss_sae:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vmaxss {sae}, %xmm1, %xmm0, %xmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %res = call <4 x float> @llvm.x86.avx512.mask.max.ss.round(<4 x float>%a0, <4 x float> %a1, <4 x float> zeroinitializer, i8 -1, i32 8)
  ret <4 x float> %res
}

define <4 x float> @test_mask_max_ss(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2, i8 %mask) {
; X64-LABEL: test_mask_max_ss:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vmaxss %xmm1, %xmm0, %xmm2 {%k1}
; X64-NEXT:    vmovaps %xmm2, %xmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_mask_max_ss:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vmaxss %xmm1, %xmm0, %xmm2 {%k1}
; X86-NEXT:    vmovaps %xmm2, %xmm0
; X86-NEXT:    retl
  %res = call <4 x float> @llvm.x86.avx512.mask.max.ss.round(<4 x float>%a0, <4 x float> %a1, <4 x float> %a2, i8 %mask, i32 4)
  ret <4 x float> %res
}

define <4 x float> @test_maskz_max_ss(<4 x float> %a0, <4 x float> %a1, i8 %mask) {
; X64-LABEL: test_maskz_max_ss:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vmaxss %xmm1, %xmm0, %xmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_maskz_max_ss:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vmaxss %xmm1, %xmm0, %xmm0 {%k1} {z}
; X86-NEXT:    retl
  %res = call <4 x float> @llvm.x86.avx512.mask.max.ss.round(<4 x float>%a0, <4 x float> %a1, <4 x float> zeroinitializer, i8 %mask, i32 4)
  ret <4 x float> %res
}

define <4 x float> @test_max_ss(<4 x float> %a0, <4 x float> %a1) {
; CHECK-LABEL: test_max_ss:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vmaxss %xmm1, %xmm0, %xmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %res = call <4 x float> @llvm.x86.avx512.mask.max.ss.round(<4 x float>%a0, <4 x float> %a1, <4 x float> zeroinitializer, i8 -1, i32 4)
  ret <4 x float> %res
}

define <4 x float> @test_mask_max_ss_memfold(<4 x float> %a0, float* %a1, <4 x float> %a2, i8 %mask) {
; X64-LABEL: test_mask_max_ss_memfold:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1
; X64-NEXT:    vmaxss (%rdi), %xmm0, %xmm1 {%k1}
; X64-NEXT:    vmovaps %xmm1, %xmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_mask_max_ss_memfold:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    vmaxss (%eax), %xmm0, %xmm1 {%k1}
; X86-NEXT:    vmovaps %xmm1, %xmm0
; X86-NEXT:    retl
  %a1.val = load float, float* %a1
  %a1v0 = insertelement <4 x float> undef, float %a1.val, i32 0
  %a1v1 = insertelement <4 x float> %a1v0, float 0.000000e+00, i32 1
  %a1v2 = insertelement <4 x float> %a1v1, float 0.000000e+00, i32 2
  %a1v  = insertelement <4 x float> %a1v2, float 0.000000e+00, i32 3
  %res = call <4 x float> @llvm.x86.avx512.mask.max.ss.round(<4 x float>%a0, <4 x float> %a1v, <4 x float> %a2, i8 %mask, i32 4)
  ret <4 x float> %res
}

define <4 x float> @test_maskz_max_ss_memfold(<4 x float> %a0, float* %a1, i8 %mask) {
; X64-LABEL: test_maskz_max_ss_memfold:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1
; X64-NEXT:    vmaxss (%rdi), %xmm0, %xmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_maskz_max_ss_memfold:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    vmaxss (%eax), %xmm0, %xmm0 {%k1} {z}
; X86-NEXT:    retl
  %a1.val = load float, float* %a1
  %a1v0 = insertelement <4 x float> undef, float %a1.val, i32 0
  %a1v1 = insertelement <4 x float> %a1v0, float 0.000000e+00, i32 1
  %a1v2 = insertelement <4 x float> %a1v1, float 0.000000e+00, i32 2
  %a1v  = insertelement <4 x float> %a1v2, float 0.000000e+00, i32 3
  %res = call <4 x float> @llvm.x86.avx512.mask.max.ss.round(<4 x float>%a0, <4 x float> %a1v, <4 x float> zeroinitializer, i8 %mask, i32 4)
  ret <4 x float> %res
}
declare <2 x double> @llvm.x86.avx512.mask.max.sd.round(<2 x double>, <2 x double>, <2 x double>, i8, i32) nounwind readnone

define <2 x double> @test_mask_max_sd_sae(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2, i8 %mask) {
; X64-LABEL: test_mask_max_sd_sae:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vmaxsd {sae}, %xmm1, %xmm0, %xmm2 {%k1}
; X64-NEXT:    vmovapd %xmm2, %xmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_mask_max_sd_sae:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vmaxsd {sae}, %xmm1, %xmm0, %xmm2 {%k1}
; X86-NEXT:    vmovapd %xmm2, %xmm0
; X86-NEXT:    retl
  %res = call <2 x double> @llvm.x86.avx512.mask.max.sd.round(<2 x double>%a0, <2 x double> %a1, <2 x double> %a2, i8 %mask, i32 8)
  ret <2 x double> %res
}

define <2 x double> @test_maskz_max_sd_sae(<2 x double> %a0, <2 x double> %a1, i8 %mask) {
; X64-LABEL: test_maskz_max_sd_sae:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vmaxsd {sae}, %xmm1, %xmm0, %xmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_maskz_max_sd_sae:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vmaxsd {sae}, %xmm1, %xmm0, %xmm0 {%k1} {z}
; X86-NEXT:    retl
  %res = call <2 x double> @llvm.x86.avx512.mask.max.sd.round(<2 x double>%a0, <2 x double> %a1, <2 x double> zeroinitializer, i8 %mask, i32 8)
  ret <2 x double> %res
}

define <2 x double> @test_max_sd_sae(<2 x double> %a0, <2 x double> %a1) {
; CHECK-LABEL: test_max_sd_sae:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vmaxsd {sae}, %xmm1, %xmm0, %xmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %res = call <2 x double> @llvm.x86.avx512.mask.max.sd.round(<2 x double>%a0, <2 x double> %a1, <2 x double> zeroinitializer, i8 -1, i32 8)
  ret <2 x double> %res
}

define <2 x double> @test_mask_max_sd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2, i8 %mask) {
; X64-LABEL: test_mask_max_sd:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vmaxsd %xmm1, %xmm0, %xmm2 {%k1}
; X64-NEXT:    vmovapd %xmm2, %xmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_mask_max_sd:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vmaxsd %xmm1, %xmm0, %xmm2 {%k1}
; X86-NEXT:    vmovapd %xmm2, %xmm0
; X86-NEXT:    retl
  %res = call <2 x double> @llvm.x86.avx512.mask.max.sd.round(<2 x double>%a0, <2 x double> %a1, <2 x double> %a2, i8 %mask, i32 4)
  ret <2 x double> %res
}

define <2 x double> @test_maskz_max_sd(<2 x double> %a0, <2 x double> %a1, i8 %mask) {
; X64-LABEL: test_maskz_max_sd:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vmaxsd %xmm1, %xmm0, %xmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_maskz_max_sd:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vmaxsd %xmm1, %xmm0, %xmm0 {%k1} {z}
; X86-NEXT:    retl
  %res = call <2 x double> @llvm.x86.avx512.mask.max.sd.round(<2 x double>%a0, <2 x double> %a1, <2 x double> zeroinitializer, i8 %mask, i32 4)
  ret <2 x double> %res
}

define <2 x double> @test_max_sd(<2 x double> %a0, <2 x double> %a1) {
; CHECK-LABEL: test_max_sd:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vmaxsd %xmm1, %xmm0, %xmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %res = call <2 x double> @llvm.x86.avx512.mask.max.sd.round(<2 x double>%a0, <2 x double> %a1, <2 x double> zeroinitializer, i8 -1, i32 4)
  ret <2 x double> %res
}

define <2 x double> @test_mask_max_sd_memfold(<2 x double> %a0, double* %a1, <2 x double> %a2, i8 %mask) {
; X64-LABEL: test_mask_max_sd_memfold:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1
; X64-NEXT:    vmaxsd (%rdi), %xmm0, %xmm1 {%k1}
; X64-NEXT:    vmovapd %xmm1, %xmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_mask_max_sd_memfold:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    vmaxsd (%eax), %xmm0, %xmm1 {%k1}
; X86-NEXT:    vmovapd %xmm1, %xmm0
; X86-NEXT:    retl
  %a1.val = load double, double* %a1
  %a1v0 = insertelement <2 x double> undef, double %a1.val, i32 0
  %a1v = insertelement <2 x double> %a1v0, double 0.000000e+00, i32 1
  %res = call <2 x double> @llvm.x86.avx512.mask.max.sd.round(<2 x double>%a0, <2 x double> %a1v, <2 x double> %a2, i8 %mask, i32 4)
  ret <2 x double> %res
}

define <2 x double> @test_maskz_max_sd_memfold(<2 x double> %a0, double* %a1, i8 %mask) {
; X64-LABEL: test_maskz_max_sd_memfold:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1
; X64-NEXT:    vmaxsd (%rdi), %xmm0, %xmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_maskz_max_sd_memfold:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    vmaxsd (%eax), %xmm0, %xmm0 {%k1} {z}
; X86-NEXT:    retl
  %a1.val = load double, double* %a1
  %a1v0 = insertelement <2 x double> undef, double %a1.val, i32 0
  %a1v = insertelement <2 x double> %a1v0, double 0.000000e+00, i32 1
  %res = call <2 x double> @llvm.x86.avx512.mask.max.sd.round(<2 x double>%a0, <2 x double> %a1v, <2 x double> zeroinitializer, i8 %mask, i32 4)
  ret <2 x double> %res
}

define <4 x float> @test_x86_avx512_cvtsi2ss32(<4 x float> %a, i32 %b) {
; X64-LABEL: test_x86_avx512_cvtsi2ss32:
; X64:       # %bb.0:
; X64-NEXT:    vcvtsi2ssl %edi, {rz-sae}, %xmm0, %xmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_x86_avx512_cvtsi2ss32:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    vcvtsi2ssl %eax, {rz-sae}, %xmm0, %xmm0
; X86-NEXT:    retl
  %res = call <4 x float> @llvm.x86.avx512.cvtsi2ss32(<4 x float> %a, i32 %b, i32 3) ; <<<4 x float>> [#uses=1]
  ret <4 x float> %res
}
declare <4 x float> @llvm.x86.avx512.cvtsi2ss32(<4 x float>, i32, i32) nounwind readnone

define <4 x float> @test_x86_avx512__mm_cvt_roundu32_ss (<4 x float> %a, i32 %b) {
; X64-LABEL: test_x86_avx512__mm_cvt_roundu32_ss:
; X64:       # %bb.0:
; X64-NEXT:    vcvtusi2ssl %edi, {rd-sae}, %xmm0, %xmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_x86_avx512__mm_cvt_roundu32_ss:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    vcvtusi2ssl %eax, {rd-sae}, %xmm0, %xmm0
; X86-NEXT:    retl
  %res = call <4 x float> @llvm.x86.avx512.cvtusi2ss(<4 x float> %a, i32 %b, i32 1) ; <<<4 x float>> [#uses=1]
  ret <4 x float> %res
}

define <4 x float> @test_x86_avx512__mm_cvt_roundu32_ss_mem(<4 x float> %a, i32* %ptr) {
; X64-LABEL: test_x86_avx512__mm_cvt_roundu32_ss_mem:
; X64:       # %bb.0:
; X64-NEXT:    movl (%rdi), %eax
; X64-NEXT:    vcvtusi2ssl %eax, {rd-sae}, %xmm0, %xmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_x86_avx512__mm_cvt_roundu32_ss_mem:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    movl (%eax), %eax
; X86-NEXT:    vcvtusi2ssl %eax, {rd-sae}, %xmm0, %xmm0
; X86-NEXT:    retl
  %b = load i32, i32* %ptr
  %res = call <4 x float> @llvm.x86.avx512.cvtusi2ss(<4 x float> %a, i32 %b, i32 1) ; <<<4 x float>> [#uses=1]
  ret <4 x float> %res
}

define <4 x float> @test_x86_avx512__mm_cvtu32_ss(<4 x float> %a, i32 %b) {
; X64-LABEL: test_x86_avx512__mm_cvtu32_ss:
; X64:       # %bb.0:
; X64-NEXT:    vcvtusi2ssl %edi, %xmm0, %xmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_x86_avx512__mm_cvtu32_ss:
; X86:       # %bb.0:
; X86-NEXT:    vcvtusi2ssl {{[0-9]+}}(%esp), %xmm0, %xmm0
; X86-NEXT:    retl
  %res = call <4 x float> @llvm.x86.avx512.cvtusi2ss(<4 x float> %a, i32 %b, i32 4) ; <<<4 x float>> [#uses=1]
  ret <4 x float> %res
}

define <4 x float> @test_x86_avx512__mm_cvtu32_ss_mem(<4 x float> %a, i32* %ptr) {
; X64-LABEL: test_x86_avx512__mm_cvtu32_ss_mem:
; X64:       # %bb.0:
; X64-NEXT:    vcvtusi2ssl (%rdi), %xmm0, %xmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_x86_avx512__mm_cvtu32_ss_mem:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    vcvtusi2ssl (%eax), %xmm0, %xmm0
; X86-NEXT:    retl
  %b = load i32, i32* %ptr
  %res = call <4 x float> @llvm.x86.avx512.cvtusi2ss(<4 x float> %a, i32 %b, i32 4) ; <<<4 x float>> [#uses=1]
  ret <4 x float> %res
}
declare <4 x float> @llvm.x86.avx512.cvtusi2ss(<4 x float>, i32, i32) nounwind readnone

declare <16 x i32> @llvm.x86.avx512.vpermi2var.d.512(<16 x i32>, <16 x i32>, <16 x i32>)

define <16 x i32>@test_int_x86_avx512_mask_vpermi2var_d_512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32>* %x2p, <16 x i32> %x4, i16 %x3) {
; X64-LABEL: test_int_x86_avx512_mask_vpermi2var_d_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1
; X64-NEXT:    vmovdqa64 %zmm1, %zmm3
; X64-NEXT:    vpermi2d (%rdi), %zmm0, %zmm3 {%k1}
; X64-NEXT:    vpermt2d %zmm2, %zmm1, %zmm0
; X64-NEXT:    vpaddd %zmm0, %zmm3, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_vpermi2var_d_512:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vmovdqa64 %zmm1, %zmm3
; X86-NEXT:    vpermi2d (%eax), %zmm0, %zmm3 {%k1}
; X86-NEXT:    vpermt2d %zmm2, %zmm1, %zmm0
; X86-NEXT:    vpaddd %zmm0, %zmm3, %zmm0
; X86-NEXT:    retl
  %x2 = load <16 x i32>, <16 x i32>* %x2p
  %1 = call <16 x i32> @llvm.x86.avx512.vpermi2var.d.512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32> %x2)
  %2 = bitcast i16 %x3 to <16 x i1>
  %3 = select <16 x i1> %2, <16 x i32> %1, <16 x i32> %x1
  %4 = call <16 x i32> @llvm.x86.avx512.vpermi2var.d.512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32> %x4)
  %res2 = add <16 x i32> %3, %4
  ret <16 x i32> %res2
}

declare <8 x double> @llvm.x86.avx512.vpermi2var.pd.512(<8 x double>, <8 x i64>, <8 x double>)

define <8 x double>@test_int_x86_avx512_mask_vpermi2var_pd_512(<8 x double> %x0, <8 x i64> %x1, <8 x double> %x2, i8 %x3) {
; X64-LABEL: test_int_x86_avx512_mask_vpermi2var_pd_512:
; X64:       # %bb.0:
; X64-NEXT:    vmovapd %zmm0, %zmm3
; X64-NEXT:    vpermt2pd %zmm2, %zmm1, %zmm3
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpermi2pd %zmm2, %zmm0, %zmm1 {%k1}
; X64-NEXT:    vaddpd %zmm3, %zmm1, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_vpermi2var_pd_512:
; X86:       # %bb.0:
; X86-NEXT:    vmovapd %zmm0, %zmm3
; X86-NEXT:    vpermt2pd %zmm2, %zmm1, %zmm3
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vpermi2pd %zmm2, %zmm0, %zmm1 {%k1}
; X86-NEXT:    vaddpd %zmm3, %zmm1, %zmm0
; X86-NEXT:    retl
  %1 = call <8 x double> @llvm.x86.avx512.vpermi2var.pd.512(<8 x double> %x0, <8 x i64> %x1, <8 x double> %x2)
  %2 = bitcast <8 x i64> %x1 to <8 x double>
  %3 = bitcast i8 %x3 to <8 x i1>
  %4 = select <8 x i1> %3, <8 x double> %1, <8 x double> %2
  %5 = call <8 x double> @llvm.x86.avx512.vpermi2var.pd.512(<8 x double> %x0, <8 x i64> %x1, <8 x double> %x2)
  %6 = bitcast <8 x i64> %x1 to <8 x double>
  %res2 = fadd <8 x double> %4, %5
  ret <8 x double> %res2
}

declare <16 x float> @llvm.x86.avx512.vpermi2var.ps.512(<16 x float>, <16 x i32>, <16 x float>)

define <16 x float>@test_int_x86_avx512_mask_vpermi2var_ps_512(<16 x float> %x0, <16 x i32> %x1, <16 x float> %x2, i16 %x3) {
; X64-LABEL: test_int_x86_avx512_mask_vpermi2var_ps_512:
; X64:       # %bb.0:
; X64-NEXT:    vmovaps %zmm0, %zmm3
; X64-NEXT:    vpermt2ps %zmm2, %zmm1, %zmm3
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpermi2ps %zmm2, %zmm0, %zmm1 {%k1}
; X64-NEXT:    vaddps %zmm3, %zmm1, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_vpermi2var_ps_512:
; X86:       # %bb.0:
; X86-NEXT:    vmovaps %zmm0, %zmm3
; X86-NEXT:    vpermt2ps %zmm2, %zmm1, %zmm3
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vpermi2ps %zmm2, %zmm0, %zmm1 {%k1}
; X86-NEXT:    vaddps %zmm3, %zmm1, %zmm0
; X86-NEXT:    retl
  %1 = call <16 x float> @llvm.x86.avx512.vpermi2var.ps.512(<16 x float> %x0, <16 x i32> %x1, <16 x float> %x2)
  %2 = bitcast <16 x i32> %x1 to <16 x float>
  %3 = bitcast i16 %x3 to <16 x i1>
  %4 = select <16 x i1> %3, <16 x float> %1, <16 x float> %2
  %5 = call <16 x float> @llvm.x86.avx512.vpermi2var.ps.512(<16 x float> %x0, <16 x i32> %x1, <16 x float> %x2)
  %6 = bitcast <16 x i32> %x1 to <16 x float>
  %res2 = fadd <16 x float> %4, %5
  ret <16 x float> %res2
}

declare <8 x i64> @llvm.x86.avx512.vpermi2var.q.512(<8 x i64>, <8 x i64>, <8 x i64>)

define <8 x i64>@test_int_x86_avx512_mask_vpermi2var_q_512(<8 x i64> %x0, <8 x i64> %x1, <8 x i64> %x2, i8 %x3) {
; X64-LABEL: test_int_x86_avx512_mask_vpermi2var_q_512:
; X64:       # %bb.0:
; X64-NEXT:    vmovdqa64 %zmm0, %zmm3
; X64-NEXT:    vpermt2q %zmm2, %zmm1, %zmm3
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpermi2q %zmm2, %zmm0, %zmm1 {%k1}
; X64-NEXT:    vpaddq %zmm3, %zmm1, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_vpermi2var_q_512:
; X86:       # %bb.0:
; X86-NEXT:    vmovdqa64 %zmm0, %zmm3
; X86-NEXT:    vpermt2q %zmm2, %zmm1, %zmm3
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vpermi2q %zmm2, %zmm0, %zmm1 {%k1}
; X86-NEXT:    vpaddq %zmm3, %zmm1, %zmm0
; X86-NEXT:    retl
  %1 = call <8 x i64> @llvm.x86.avx512.vpermi2var.q.512(<8 x i64> %x0, <8 x i64> %x1, <8 x i64> %x2)
  %2 = bitcast i8 %x3 to <8 x i1>
  %3 = select <8 x i1> %2, <8 x i64> %1, <8 x i64> %x1
  %4 = call <8 x i64> @llvm.x86.avx512.vpermi2var.q.512(<8 x i64> %x0, <8 x i64> %x1, <8 x i64> %x2)
  %res2 = add <8 x i64> %3, %4
  ret <8 x i64> %res2
}

define <16 x i32>@test_int_x86_avx512_maskz_vpermt2var_d_512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32>* %x2p, i16 %x3) {
; X64-LABEL: test_int_x86_avx512_maskz_vpermt2var_d_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1
; X64-NEXT:    vmovdqa64 %zmm1, %zmm2
; X64-NEXT:    vpermt2d (%rdi), %zmm0, %zmm2 {%k1} {z}
; X64-NEXT:    vpermt2d %zmm1, %zmm0, %zmm1
; X64-NEXT:    vpaddd %zmm1, %zmm2, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_maskz_vpermt2var_d_512:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vmovdqa64 %zmm1, %zmm2
; X86-NEXT:    vpermt2d (%eax), %zmm0, %zmm2 {%k1} {z}
; X86-NEXT:    vpermt2d %zmm1, %zmm0, %zmm1
; X86-NEXT:    vpaddd %zmm1, %zmm2, %zmm0
; X86-NEXT:    retl
  %x2 = load <16 x i32>, <16 x i32>* %x2p
  %1 = call <16 x i32> @llvm.x86.avx512.vpermi2var.d.512(<16 x i32> %x1, <16 x i32> %x0, <16 x i32> %x2)
  %2 = bitcast i16 %x3 to <16 x i1>
  %3 = select <16 x i1> %2, <16 x i32> %1, <16 x i32> zeroinitializer
  %4 = call <16 x i32> @llvm.x86.avx512.vpermi2var.d.512(<16 x i32> %x1, <16 x i32> %x0, <16 x i32> %x1)
  %res2 = add <16 x i32> %3, %4
  ret <16 x i32> %res2
}

define <8 x double>@test_int_x86_avx512_maskz_vpermt2var_pd_512(<8 x i64> %x0, <8 x double> %x1, double* %x2ptr, i8 %x3) {
; X64-LABEL: test_int_x86_avx512_maskz_vpermt2var_pd_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1
; X64-NEXT:    vmovapd %zmm1, %zmm2
; X64-NEXT:    vpermt2pd (%rdi){1to8}, %zmm0, %zmm2 {%k1} {z}
; X64-NEXT:    vpermt2pd %zmm1, %zmm0, %zmm1
; X64-NEXT:    vaddpd %zmm1, %zmm2, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_maskz_vpermt2var_pd_512:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx
; X86-NEXT:    kmovw %ecx, %k1
; X86-NEXT:    vmovapd %zmm1, %zmm2
; X86-NEXT:    vpermt2pd (%eax){1to8}, %zmm0, %zmm2 {%k1} {z}
; X86-NEXT:    vpermt2pd %zmm1, %zmm0, %zmm1
; X86-NEXT:    vaddpd %zmm1, %zmm2, %zmm0
; X86-NEXT:    retl
  %x2s = load double, double* %x2ptr
  %x2ins = insertelement <8 x double> undef, double %x2s, i32 0
  %x2 = shufflevector <8 x double> %x2ins, <8 x double> undef, <8 x i32> zeroinitializer
  %1 = call <8 x double> @llvm.x86.avx512.vpermi2var.pd.512(<8 x double> %x1, <8 x i64> %x0, <8 x double> %x2)
  %2 = bitcast i8 %x3 to <8 x i1>
  %3 = select <8 x i1> %2, <8 x double> %1, <8 x double> zeroinitializer
  %4 = call <8 x double> @llvm.x86.avx512.vpermi2var.pd.512(<8 x double> %x1, <8 x i64> %x0, <8 x double> %x1)
  %res2 = fadd <8 x double> %3, %4
  ret <8 x double> %res2
}

define <16 x float>@test_int_x86_avx512_maskz_vpermt2var_ps_512(<16 x i32> %x0, <16 x float> %x1, <16 x float> %x2, i16 %x3) {
; X64-LABEL: test_int_x86_avx512_maskz_vpermt2var_ps_512:
; X64:       # %bb.0:
; X64-NEXT:    vmovaps %zmm1, %zmm3
; X64-NEXT:    vpermt2ps %zmm2, %zmm0, %zmm3
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpermt2ps %zmm2, %zmm0, %zmm1 {%k1} {z}
; X64-NEXT:    vaddps %zmm3, %zmm1, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_maskz_vpermt2var_ps_512:
; X86:       # %bb.0:
; X86-NEXT:    vmovaps %zmm1, %zmm3
; X86-NEXT:    vpermt2ps %zmm2, %zmm0, %zmm3
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vpermt2ps %zmm2, %zmm0, %zmm1 {%k1} {z}
; X86-NEXT:    vaddps %zmm3, %zmm1, %zmm0
; X86-NEXT:    retl
  %1 = call <16 x float> @llvm.x86.avx512.vpermi2var.ps.512(<16 x float> %x1, <16 x i32> %x0, <16 x float> %x2)
  %2 = bitcast i16 %x3 to <16 x i1>
  %3 = select <16 x i1> %2, <16 x float> %1, <16 x float> zeroinitializer
  %4 = call <16 x float> @llvm.x86.avx512.vpermi2var.ps.512(<16 x float> %x1, <16 x i32> %x0, <16 x float> %x2)
  %res2 = fadd <16 x float> %3, %4
  ret <16 x float> %res2
}

define <8 x i64>@test_int_x86_avx512_maskz_vpermt2var_q_512(<8 x i64> %x0, <8 x i64> %x1, <8 x i64> %x2, i8 %x3) {
; X64-LABEL: test_int_x86_avx512_maskz_vpermt2var_q_512:
; X64:       # %bb.0:
; X64-NEXT:    vmovdqa64 %zmm1, %zmm3
; X64-NEXT:    vpermt2q %zmm2, %zmm0, %zmm3
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpermt2q %zmm2, %zmm0, %zmm1 {%k1} {z}
; X64-NEXT:    vpaddq %zmm3, %zmm1, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_maskz_vpermt2var_q_512:
; X86:       # %bb.0:
; X86-NEXT:    vmovdqa64 %zmm1, %zmm3
; X86-NEXT:    vpermt2q %zmm2, %zmm0, %zmm3
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vpermt2q %zmm2, %zmm0, %zmm1 {%k1} {z}
; X86-NEXT:    vpaddq %zmm3, %zmm1, %zmm0
; X86-NEXT:    retl
  %1 = call <8 x i64> @llvm.x86.avx512.vpermi2var.q.512(<8 x i64> %x1, <8 x i64> %x0, <8 x i64> %x2)
  %2 = bitcast i8 %x3 to <8 x i1>
  %3 = select <8 x i1> %2, <8 x i64> %1, <8 x i64> zeroinitializer
  %4 = call <8 x i64> @llvm.x86.avx512.vpermi2var.q.512(<8 x i64> %x1, <8 x i64> %x0, <8 x i64> %x2)
  %res2 = add <8 x i64> %3, %4
  ret <8 x i64> %res2
}

define <16 x i32>@test_int_x86_avx512_mask_vpermt2var_d_512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32> %x2, i16 %x3) {
; X64-LABEL: test_int_x86_avx512_mask_vpermt2var_d_512:
; X64:       # %bb.0:
; X64-NEXT:    vmovdqa64 %zmm1, %zmm3
; X64-NEXT:    vpermt2d %zmm2, %zmm0, %zmm3
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpermt2d %zmm2, %zmm0, %zmm1 {%k1}
; X64-NEXT:    vpaddd %zmm3, %zmm1, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_vpermt2var_d_512:
; X86:       # %bb.0:
; X86-NEXT:    vmovdqa64 %zmm1, %zmm3
; X86-NEXT:    vpermt2d %zmm2, %zmm0, %zmm3
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vpermt2d %zmm2, %zmm0, %zmm1 {%k1}
; X86-NEXT:    vpaddd %zmm3, %zmm1, %zmm0
; X86-NEXT:    retl
  %1 = call <16 x i32> @llvm.x86.avx512.vpermi2var.d.512(<16 x i32> %x1, <16 x i32> %x0, <16 x i32> %x2)
  %2 = bitcast i16 %x3 to <16 x i1>
  %3 = select <16 x i1> %2, <16 x i32> %1, <16 x i32> %x1
  %4 = call <16 x i32> @llvm.x86.avx512.vpermi2var.d.512(<16 x i32> %x1, <16 x i32> %x0, <16 x i32> %x2)
  %res2 = add <16 x i32> %3, %4
  ret <16 x i32> %res2
}

declare <8 x double> @llvm.x86.avx512.mask.scalef.pd.512(<8 x double>, <8 x double>, <8 x double>, i8, i32)
define <8 x double>@test_int_x86_avx512_mask_scalef_pd_512(<8 x double> %x0, <8 x double> %x1, <8 x double> %x2, i8 %x3) {
; X64-LABEL: test_int_x86_avx512_mask_scalef_pd_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vscalefpd {rz-sae}, %zmm1, %zmm0, %zmm2 {%k1}
; X64-NEXT:    vscalefpd {rn-sae}, %zmm1, %zmm0, %zmm0
; X64-NEXT:    vaddpd %zmm0, %zmm2, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_scalef_pd_512:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vscalefpd {rz-sae}, %zmm1, %zmm0, %zmm2 {%k1}
; X86-NEXT:    vscalefpd {rn-sae}, %zmm1, %zmm0, %zmm0
; X86-NEXT:    vaddpd %zmm0, %zmm2, %zmm0
; X86-NEXT:    retl
  %res = call <8 x double> @llvm.x86.avx512.mask.scalef.pd.512(<8 x double> %x0, <8 x double> %x1, <8 x double> %x2, i8 %x3, i32 3)
  %res1 = call <8 x double> @llvm.x86.avx512.mask.scalef.pd.512(<8 x double> %x0, <8 x double> %x1, <8 x double> %x2, i8 -1, i32 0)
  %res2 = fadd <8 x double> %res, %res1
  ret <8 x double> %res2
}

declare <16 x float> @llvm.x86.avx512.mask.scalef.ps.512(<16 x float>, <16 x float>, <16 x float>, i16, i32)
define <16 x float>@test_int_x86_avx512_mask_scalef_ps_512(<16 x float> %x0, <16 x float> %x1, <16 x float> %x2, i16 %x3) {
; X64-LABEL: test_int_x86_avx512_mask_scalef_ps_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vscalefps {ru-sae}, %zmm1, %zmm0, %zmm2 {%k1}
; X64-NEXT:    vscalefps {rn-sae}, %zmm1, %zmm0, %zmm0
; X64-NEXT:    vaddps %zmm0, %zmm2, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_scalef_ps_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vscalefps {ru-sae}, %zmm1, %zmm0, %zmm2 {%k1}
; X86-NEXT:    vscalefps {rn-sae}, %zmm1, %zmm0, %zmm0
; X86-NEXT:    vaddps %zmm0, %zmm2, %zmm0
; X86-NEXT:    retl
  %res = call <16 x float> @llvm.x86.avx512.mask.scalef.ps.512(<16 x float> %x0, <16 x float> %x1, <16 x float> %x2, i16 %x3, i32 2)
  %res1 = call <16 x float> @llvm.x86.avx512.mask.scalef.ps.512(<16 x float> %x0, <16 x float> %x1, <16 x float> %x2, i16 -1, i32 0)
  %res2 = fadd <16 x float> %res, %res1
  ret <16 x float> %res2
}

declare <16 x i8> @llvm.x86.avx512.mask.pmov.qb.512(<8 x i64>, <16 x i8>, i8)

define <16 x i8>@test_int_x86_avx512_mask_pmov_qb_512(<8 x i64> %x0, <16 x i8> %x1, i8 %x2) {
; X64-LABEL: test_int_x86_avx512_mask_pmov_qb_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpmovqb %zmm0, %xmm2
; X64-NEXT:    vpmovqb %zmm0, %xmm1 {%k1}
; X64-NEXT:    vpmovqb %zmm0, %xmm0 {%k1} {z}
; X64-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
; X64-NEXT:    vpaddb %xmm0, %xmm2, %xmm0
; X64-NEXT:    vzeroupper
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_pmov_qb_512:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vpmovqb %zmm0, %xmm2
; X86-NEXT:    vpmovqb %zmm0, %xmm1 {%k1}
; X86-NEXT:    vpmovqb %zmm0, %xmm0 {%k1} {z}
; X86-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
; X86-NEXT:    vpaddb %xmm0, %xmm2, %xmm0
; X86-NEXT:    vzeroupper
; X86-NEXT:    retl
    %res0 = call <16 x i8> @llvm.x86.avx512.mask.pmov.qb.512(<8 x i64> %x0, <16 x i8> %x1, i8 -1)
    %res1 = call <16 x i8> @llvm.x86.avx512.mask.pmov.qb.512(<8 x i64> %x0, <16 x i8> %x1, i8 %x2)
    %res2 = call <16 x i8> @llvm.x86.avx512.mask.pmov.qb.512(<8 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.512(i8* %ptr, <8 x i64>, i8)

define void @test_int_x86_avx512_mask_pmov_qb_mem_512(i8* %ptr, <8 x i64> %x1, i8 %x2) {
; X64-LABEL: test_int_x86_avx512_mask_pmov_qb_mem_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1
; X64-NEXT:    vpmovqb %zmm0, (%rdi)
; X64-NEXT:    vpmovqb %zmm0, (%rdi) {%k1}
; X64-NEXT:    vzeroupper
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_pmov_qb_mem_512:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    vpmovqb %zmm0, (%eax)
; X86-NEXT:    vpmovqb %zmm0, (%eax) {%k1}
; X86-NEXT:    vzeroupper
; X86-NEXT:    retl
    call void @llvm.x86.avx512.mask.pmov.qb.mem.512(i8* %ptr, <8 x i64> %x1, i8 -1)
    call void @llvm.x86.avx512.mask.pmov.qb.mem.512(i8* %ptr, <8 x i64> %x1, i8 %x2)
    ret void
}

declare <16 x i8> @llvm.x86.avx512.mask.pmovs.qb.512(<8 x i64>, <16 x i8>, i8)

define <16 x i8>@test_int_x86_avx512_mask_pmovs_qb_512(<8 x i64> %x0, <16 x i8> %x1, i8 %x2) {
; X64-LABEL: test_int_x86_avx512_mask_pmovs_qb_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpmovsqb %zmm0, %xmm2
; X64-NEXT:    vpmovsqb %zmm0, %xmm1 {%k1}
; X64-NEXT:    vpmovsqb %zmm0, %xmm0 {%k1} {z}
; X64-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
; X64-NEXT:    vpaddb %xmm0, %xmm2, %xmm0
; X64-NEXT:    vzeroupper
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_pmovs_qb_512:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vpmovsqb %zmm0, %xmm2
; X86-NEXT:    vpmovsqb %zmm0, %xmm1 {%k1}
; X86-NEXT:    vpmovsqb %zmm0, %xmm0 {%k1} {z}
; X86-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
; X86-NEXT:    vpaddb %xmm0, %xmm2, %xmm0
; X86-NEXT:    vzeroupper
; X86-NEXT:    retl
    %res0 = call <16 x i8> @llvm.x86.avx512.mask.pmovs.qb.512(<8 x i64> %x0, <16 x i8> %x1, i8 -1)
    %res1 = call <16 x i8> @llvm.x86.avx512.mask.pmovs.qb.512(<8 x i64> %x0, <16 x i8> %x1, i8 %x2)
    %res2 = call <16 x i8> @llvm.x86.avx512.mask.pmovs.qb.512(<8 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.512(i8* %ptr, <8 x i64>, i8)

define void @test_int_x86_avx512_mask_pmovs_qb_mem_512(i8* %ptr, <8 x i64> %x1, i8 %x2) {
; X64-LABEL: test_int_x86_avx512_mask_pmovs_qb_mem_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1
; X64-NEXT:    vpmovsqb %zmm0, (%rdi)
; X64-NEXT:    vpmovsqb %zmm0, (%rdi) {%k1}
; X64-NEXT:    vzeroupper
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_pmovs_qb_mem_512:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    vpmovsqb %zmm0, (%eax)
; X86-NEXT:    vpmovsqb %zmm0, (%eax) {%k1}
; X86-NEXT:    vzeroupper
; X86-NEXT:    retl
    call void @llvm.x86.avx512.mask.pmovs.qb.mem.512(i8* %ptr, <8 x i64> %x1, i8 -1)
    call void @llvm.x86.avx512.mask.pmovs.qb.mem.512(i8* %ptr, <8 x i64> %x1, i8 %x2)
    ret void
}

declare <16 x i8> @llvm.x86.avx512.mask.pmovus.qb.512(<8 x i64>, <16 x i8>, i8)

define <16 x i8>@test_int_x86_avx512_mask_pmovus_qb_512(<8 x i64> %x0, <16 x i8> %x1, i8 %x2) {
; X64-LABEL: test_int_x86_avx512_mask_pmovus_qb_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpmovusqb %zmm0, %xmm2
; X64-NEXT:    vpmovusqb %zmm0, %xmm1 {%k1}
; X64-NEXT:    vpmovusqb %zmm0, %xmm0 {%k1} {z}
; X64-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
; X64-NEXT:    vpaddb %xmm0, %xmm2, %xmm0
; X64-NEXT:    vzeroupper
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_pmovus_qb_512:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vpmovusqb %zmm0, %xmm2
; X86-NEXT:    vpmovusqb %zmm0, %xmm1 {%k1}
; X86-NEXT:    vpmovusqb %zmm0, %xmm0 {%k1} {z}
; X86-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
; X86-NEXT:    vpaddb %xmm0, %xmm2, %xmm0
; X86-NEXT:    vzeroupper
; X86-NEXT:    retl
    %res0 = call <16 x i8> @llvm.x86.avx512.mask.pmovus.qb.512(<8 x i64> %x0, <16 x i8> %x1, i8 -1)
    %res1 = call <16 x i8> @llvm.x86.avx512.mask.pmovus.qb.512(<8 x i64> %x0, <16 x i8> %x1, i8 %x2)
    %res2 = call <16 x i8> @llvm.x86.avx512.mask.pmovus.qb.512(<8 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.512(i8* %ptr, <8 x i64>, i8)

define void @test_int_x86_avx512_mask_pmovus_qb_mem_512(i8* %ptr, <8 x i64> %x1, i8 %x2) {
; X64-LABEL: test_int_x86_avx512_mask_pmovus_qb_mem_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1
; X64-NEXT:    vpmovusqb %zmm0, (%rdi)
; X64-NEXT:    vpmovusqb %zmm0, (%rdi) {%k1}
; X64-NEXT:    vzeroupper
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_pmovus_qb_mem_512:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    vpmovusqb %zmm0, (%eax)
; X86-NEXT:    vpmovusqb %zmm0, (%eax) {%k1}
; X86-NEXT:    vzeroupper
; X86-NEXT:    retl
    call void @llvm.x86.avx512.mask.pmovus.qb.mem.512(i8* %ptr, <8 x i64> %x1, i8 -1)
    call void @llvm.x86.avx512.mask.pmovus.qb.mem.512(i8* %ptr, <8 x i64> %x1, i8 %x2)
    ret void
}

declare <8 x i16> @llvm.x86.avx512.mask.pmov.qw.512(<8 x i64>, <8 x i16>, i8)

define <8 x i16>@test_int_x86_avx512_mask_pmov_qw_512(<8 x i64> %x0, <8 x i16> %x1, i8 %x2) {
; X64-LABEL: test_int_x86_avx512_mask_pmov_qw_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpmovqw %zmm0, %xmm2
; X64-NEXT:    vpmovqw %zmm0, %xmm1 {%k1}
; X64-NEXT:    vpmovqw %zmm0, %xmm0 {%k1} {z}
; X64-NEXT:    vpaddw %xmm0, %xmm1, %xmm0
; X64-NEXT:    vpaddw %xmm0, %xmm2, %xmm0
; X64-NEXT:    vzeroupper
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_pmov_qw_512:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vpmovqw %zmm0, %xmm2
; X86-NEXT:    vpmovqw %zmm0, %xmm1 {%k1}
; X86-NEXT:    vpmovqw %zmm0, %xmm0 {%k1} {z}
; X86-NEXT:    vpaddw %xmm0, %xmm1, %xmm0
; X86-NEXT:    vpaddw %xmm0, %xmm2, %xmm0
; X86-NEXT:    vzeroupper
; X86-NEXT:    retl
    %res0 = call <8 x i16> @llvm.x86.avx512.mask.pmov.qw.512(<8 x i64> %x0, <8 x i16> %x1, i8 -1)
    %res1 = call <8 x i16> @llvm.x86.avx512.mask.pmov.qw.512(<8 x i64> %x0, <8 x i16> %x1, i8 %x2)
    %res2 = call <8 x i16> @llvm.x86.avx512.mask.pmov.qw.512(<8 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.512(i8* %ptr, <8 x i64>, i8)

define void @test_int_x86_avx512_mask_pmov_qw_mem_512(i8* %ptr, <8 x i64> %x1, i8 %x2) {
; X64-LABEL: test_int_x86_avx512_mask_pmov_qw_mem_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1
; X64-NEXT:    vpmovqw %zmm0, (%rdi)
; X64-NEXT:    vpmovqw %zmm0, (%rdi) {%k1}
; X64-NEXT:    vzeroupper
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_pmov_qw_mem_512:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    vpmovqw %zmm0, (%eax)
; X86-NEXT:    vpmovqw %zmm0, (%eax) {%k1}
; X86-NEXT:    vzeroupper
; X86-NEXT:    retl
    call void @llvm.x86.avx512.mask.pmov.qw.mem.512(i8* %ptr, <8 x i64> %x1, i8 -1)
    call void @llvm.x86.avx512.mask.pmov.qw.mem.512(i8* %ptr, <8 x i64> %x1, i8 %x2)
    ret void
}

declare <8 x i16> @llvm.x86.avx512.mask.pmovs.qw.512(<8 x i64>, <8 x i16>, i8)

define <8 x i16>@test_int_x86_avx512_mask_pmovs_qw_512(<8 x i64> %x0, <8 x i16> %x1, i8 %x2) {
; X64-LABEL: test_int_x86_avx512_mask_pmovs_qw_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpmovsqw %zmm0, %xmm2
; X64-NEXT:    vpmovsqw %zmm0, %xmm1 {%k1}
; X64-NEXT:    vpmovsqw %zmm0, %xmm0 {%k1} {z}
; X64-NEXT:    vpaddw %xmm0, %xmm1, %xmm0
; X64-NEXT:    vpaddw %xmm0, %xmm2, %xmm0
; X64-NEXT:    vzeroupper
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_pmovs_qw_512:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vpmovsqw %zmm0, %xmm2
; X86-NEXT:    vpmovsqw %zmm0, %xmm1 {%k1}
; X86-NEXT:    vpmovsqw %zmm0, %xmm0 {%k1} {z}
; X86-NEXT:    vpaddw %xmm0, %xmm1, %xmm0
; X86-NEXT:    vpaddw %xmm0, %xmm2, %xmm0
; X86-NEXT:    vzeroupper
; X86-NEXT:    retl
    %res0 = call <8 x i16> @llvm.x86.avx512.mask.pmovs.qw.512(<8 x i64> %x0, <8 x i16> %x1, i8 -1)
    %res1 = call <8 x i16> @llvm.x86.avx512.mask.pmovs.qw.512(<8 x i64> %x0, <8 x i16> %x1, i8 %x2)
    %res2 = call <8 x i16> @llvm.x86.avx512.mask.pmovs.qw.512(<8 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.512(i8* %ptr, <8 x i64>, i8)

define void @test_int_x86_avx512_mask_pmovs_qw_mem_512(i8* %ptr, <8 x i64> %x1, i8 %x2) {
; X64-LABEL: test_int_x86_avx512_mask_pmovs_qw_mem_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1
; X64-NEXT:    vpmovsqw %zmm0, (%rdi)
; X64-NEXT:    vpmovsqw %zmm0, (%rdi) {%k1}
; X64-NEXT:    vzeroupper
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_pmovs_qw_mem_512:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    vpmovsqw %zmm0, (%eax)
; X86-NEXT:    vpmovsqw %zmm0, (%eax) {%k1}
; X86-NEXT:    vzeroupper
; X86-NEXT:    retl
    call void @llvm.x86.avx512.mask.pmovs.qw.mem.512(i8* %ptr, <8 x i64> %x1, i8 -1)
    call void @llvm.x86.avx512.mask.pmovs.qw.mem.512(i8* %ptr, <8 x i64> %x1, i8 %x2)
    ret void
}

declare <8 x i16> @llvm.x86.avx512.mask.pmovus.qw.512(<8 x i64>, <8 x i16>, i8)

define <8 x i16>@test_int_x86_avx512_mask_pmovus_qw_512(<8 x i64> %x0, <8 x i16> %x1, i8 %x2) {
; X64-LABEL: test_int_x86_avx512_mask_pmovus_qw_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpmovusqw %zmm0, %xmm2
; X64-NEXT:    vpmovusqw %zmm0, %xmm1 {%k1}
; X64-NEXT:    vpmovusqw %zmm0, %xmm0 {%k1} {z}
; X64-NEXT:    vpaddw %xmm0, %xmm1, %xmm0
; X64-NEXT:    vpaddw %xmm0, %xmm2, %xmm0
; X64-NEXT:    vzeroupper
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_pmovus_qw_512:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vpmovusqw %zmm0, %xmm2
; X86-NEXT:    vpmovusqw %zmm0, %xmm1 {%k1}
; X86-NEXT:    vpmovusqw %zmm0, %xmm0 {%k1} {z}
; X86-NEXT:    vpaddw %xmm0, %xmm1, %xmm0
; X86-NEXT:    vpaddw %xmm0, %xmm2, %xmm0
; X86-NEXT:    vzeroupper
; X86-NEXT:    retl
    %res0 = call <8 x i16> @llvm.x86.avx512.mask.pmovus.qw.512(<8 x i64> %x0, <8 x i16> %x1, i8 -1)
    %res1 = call <8 x i16> @llvm.x86.avx512.mask.pmovus.qw.512(<8 x i64> %x0, <8 x i16> %x1, i8 %x2)
    %res2 = call <8 x i16> @llvm.x86.avx512.mask.pmovus.qw.512(<8 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.512(i8* %ptr, <8 x i64>, i8)

define void @test_int_x86_avx512_mask_pmovus_qw_mem_512(i8* %ptr, <8 x i64> %x1, i8 %x2) {
; X64-LABEL: test_int_x86_avx512_mask_pmovus_qw_mem_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1
; X64-NEXT:    vpmovusqw %zmm0, (%rdi)
; X64-NEXT:    vpmovusqw %zmm0, (%rdi) {%k1}
; X64-NEXT:    vzeroupper
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_pmovus_qw_mem_512:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    vpmovusqw %zmm0, (%eax)
; X86-NEXT:    vpmovusqw %zmm0, (%eax) {%k1}
; X86-NEXT:    vzeroupper
; X86-NEXT:    retl
    call void @llvm.x86.avx512.mask.pmovus.qw.mem.512(i8* %ptr, <8 x i64> %x1, i8 -1)
    call void @llvm.x86.avx512.mask.pmovus.qw.mem.512(i8* %ptr, <8 x i64> %x1, i8 %x2)
    ret void
}

define <8 x i32>@test_int_x86_avx512_mask_pmov_qd_512(<8 x i64> %x0, <8 x i32> %x1, i8 %x2) {
; X64-LABEL: test_int_x86_avx512_mask_pmov_qd_512:
; X64:       # %bb.0:
; X64-NEXT:    vpmovqd %zmm0, %ymm2
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpmovqd %zmm0, %ymm1 {%k1}
; X64-NEXT:    vpmovqd %zmm0, %ymm0 {%k1} {z}
; X64-NEXT:    vpaddd %ymm0, %ymm1, %ymm0
; X64-NEXT:    vpaddd %ymm0, %ymm2, %ymm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_pmov_qd_512:
; X86:       # %bb.0:
; X86-NEXT:    vpmovqd %zmm0, %ymm2
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vpmovqd %zmm0, %ymm1 {%k1}
; X86-NEXT:    vpmovqd %zmm0, %ymm0 {%k1} {z}
; X86-NEXT:    vpaddd %ymm0, %ymm1, %ymm0
; X86-NEXT:    vpaddd %ymm0, %ymm2, %ymm0
; X86-NEXT:    retl
  %1 = trunc <8 x i64> %x0 to <8 x i32>
  %2 = trunc <8 x i64> %x0 to <8 x i32>
  %3 = bitcast i8 %x2 to <8 x i1>
  %4 = select <8 x i1> %3, <8 x i32> %2, <8 x i32> %x1
  %5 = trunc <8 x i64> %x0 to <8 x i32>
  %6 = bitcast i8 %x2 to <8 x i1>
  %7 = select <8 x i1> %6, <8 x i32> %5, <8 x i32> zeroinitializer
  %res3 = add <8 x i32> %1, %4
  %res4 = add <8 x i32> %res3, %7
  ret <8 x i32> %res4
}

declare void @llvm.x86.avx512.mask.pmov.qd.mem.512(i8* %ptr, <8 x i64>, i8)

define void @test_int_x86_avx512_mask_pmov_qd_mem_512(i8* %ptr, <8 x i64> %x1, i8 %x2) {
; X64-LABEL: test_int_x86_avx512_mask_pmov_qd_mem_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1
; X64-NEXT:    vpmovqd %zmm0, (%rdi)
; X64-NEXT:    vpmovqd %zmm0, (%rdi) {%k1}
; X64-NEXT:    vzeroupper
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_pmov_qd_mem_512:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    vpmovqd %zmm0, (%eax)
; X86-NEXT:    vpmovqd %zmm0, (%eax) {%k1}
; X86-NEXT:    vzeroupper
; X86-NEXT:    retl
    call void @llvm.x86.avx512.mask.pmov.qd.mem.512(i8* %ptr, <8 x i64> %x1, i8 -1)
    call void @llvm.x86.avx512.mask.pmov.qd.mem.512(i8* %ptr, <8 x i64> %x1, i8 %x2)
    ret void
}

declare <8 x i32> @llvm.x86.avx512.mask.pmovs.qd.512(<8 x i64>, <8 x i32>, i8)

define <8 x i32>@test_int_x86_avx512_mask_pmovs_qd_512(<8 x i64> %x0, <8 x i32> %x1, i8 %x2) {
; X64-LABEL: test_int_x86_avx512_mask_pmovs_qd_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpmovsqd %zmm0, %ymm2 {%k1} {z}
; X64-NEXT:    vpmovsqd %zmm0, %ymm1 {%k1}
; X64-NEXT:    vpaddd %ymm2, %ymm1, %ymm1
; X64-NEXT:    vpmovsqd %zmm0, %ymm0
; X64-NEXT:    vpaddd %ymm1, %ymm0, %ymm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_pmovs_qd_512:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vpmovsqd %zmm0, %ymm1 {%k1}
; X86-NEXT:    vpmovsqd %zmm0, %ymm2 {%k1} {z}
; X86-NEXT:    vpaddd %ymm2, %ymm1, %ymm1
; X86-NEXT:    vpmovsqd %zmm0, %ymm0
; X86-NEXT:    vpaddd %ymm1, %ymm0, %ymm0
; X86-NEXT:    retl
    %res0 = call <8 x i32> @llvm.x86.avx512.mask.pmovs.qd.512(<8 x i64> %x0, <8 x i32> %x1, i8 -1)
    %res1 = call <8 x i32> @llvm.x86.avx512.mask.pmovs.qd.512(<8 x i64> %x0, <8 x i32> %x1, i8 %x2)
    %res2 = call <8 x i32> @llvm.x86.avx512.mask.pmovs.qd.512(<8 x i64> %x0, <8 x i32> zeroinitializer, i8 %x2)
    %res3 = add <8 x i32> %res0, %res1
    %res4 = add <8 x i32> %res3, %res2
    ret <8 x i32> %res4
}

declare void @llvm.x86.avx512.mask.pmovs.qd.mem.512(i8* %ptr, <8 x i64>, i8)

define void @test_int_x86_avx512_mask_pmovs_qd_mem_512(i8* %ptr, <8 x i64> %x1, i8 %x2) {
; X64-LABEL: test_int_x86_avx512_mask_pmovs_qd_mem_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1
; X64-NEXT:    vpmovsqd %zmm0, (%rdi)
; X64-NEXT:    vpmovsqd %zmm0, (%rdi) {%k1}
; X64-NEXT:    vzeroupper
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_pmovs_qd_mem_512:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    vpmovsqd %zmm0, (%eax)
; X86-NEXT:    vpmovsqd %zmm0, (%eax) {%k1}
; X86-NEXT:    vzeroupper
; X86-NEXT:    retl
    call void @llvm.x86.avx512.mask.pmovs.qd.mem.512(i8* %ptr, <8 x i64> %x1, i8 -1)
    call void @llvm.x86.avx512.mask.pmovs.qd.mem.512(i8* %ptr, <8 x i64> %x1, i8 %x2)
    ret void
}

declare <8 x i32> @llvm.x86.avx512.mask.pmovus.qd.512(<8 x i64>, <8 x i32>, i8)

define <8 x i32>@test_int_x86_avx512_mask_pmovus_qd_512(<8 x i64> %x0, <8 x i32> %x1, i8 %x2) {
; X64-LABEL: test_int_x86_avx512_mask_pmovus_qd_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpmovusqd %zmm0, %ymm2 {%k1} {z}
; X64-NEXT:    vpmovusqd %zmm0, %ymm1 {%k1}
; X64-NEXT:    vpaddd %ymm2, %ymm1, %ymm1
; X64-NEXT:    vpmovusqd %zmm0, %ymm0
; X64-NEXT:    vpaddd %ymm1, %ymm0, %ymm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_pmovus_qd_512:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vpmovusqd %zmm0, %ymm1 {%k1}
; X86-NEXT:    vpmovusqd %zmm0, %ymm2 {%k1} {z}
; X86-NEXT:    vpaddd %ymm2, %ymm1, %ymm1
; X86-NEXT:    vpmovusqd %zmm0, %ymm0
; X86-NEXT:    vpaddd %ymm1, %ymm0, %ymm0
; X86-NEXT:    retl
    %res0 = call <8 x i32> @llvm.x86.avx512.mask.pmovus.qd.512(<8 x i64> %x0, <8 x i32> %x1, i8 -1)
    %res1 = call <8 x i32> @llvm.x86.avx512.mask.pmovus.qd.512(<8 x i64> %x0, <8 x i32> %x1, i8 %x2)
    %res2 = call <8 x i32> @llvm.x86.avx512.mask.pmovus.qd.512(<8 x i64> %x0, <8 x i32> zeroinitializer, i8 %x2)
    %res3 = add <8 x i32> %res0, %res1
    %res4 = add <8 x i32> %res3, %res2
    ret <8 x i32> %res4
}

declare void @llvm.x86.avx512.mask.pmovus.qd.mem.512(i8* %ptr, <8 x i64>, i8)

define void @test_int_x86_avx512_mask_pmovus_qd_mem_512(i8* %ptr, <8 x i64> %x1, i8 %x2) {
; X64-LABEL: test_int_x86_avx512_mask_pmovus_qd_mem_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1
; X64-NEXT:    vpmovusqd %zmm0, (%rdi)
; X64-NEXT:    vpmovusqd %zmm0, (%rdi) {%k1}
; X64-NEXT:    vzeroupper
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_pmovus_qd_mem_512:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    vpmovusqd %zmm0, (%eax)
; X86-NEXT:    vpmovusqd %zmm0, (%eax) {%k1}
; X86-NEXT:    vzeroupper
; X86-NEXT:    retl
    call void @llvm.x86.avx512.mask.pmovus.qd.mem.512(i8* %ptr, <8 x i64> %x1, i8 -1)
    call void @llvm.x86.avx512.mask.pmovus.qd.mem.512(i8* %ptr, <8 x i64> %x1, i8 %x2)
    ret void
}

declare <16 x i8> @llvm.x86.avx512.mask.pmov.db.512(<16 x i32>, <16 x i8>, i16)

define <16 x i8>@test_int_x86_avx512_mask_pmov_db_512(<16 x i32> %x0, <16 x i8> %x1, i16 %x2) {
; X64-LABEL: test_int_x86_avx512_mask_pmov_db_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpmovdb %zmm0, %xmm2
; X64-NEXT:    vpmovdb %zmm0, %xmm1 {%k1}
; X64-NEXT:    vpmovdb %zmm0, %xmm0 {%k1} {z}
; X64-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
; X64-NEXT:    vpaddb %xmm0, %xmm2, %xmm0
; X64-NEXT:    vzeroupper
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_pmov_db_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vpmovdb %zmm0, %xmm2
; X86-NEXT:    vpmovdb %zmm0, %xmm1 {%k1}
; X86-NEXT:    vpmovdb %zmm0, %xmm0 {%k1} {z}
; X86-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
; X86-NEXT:    vpaddb %xmm0, %xmm2, %xmm0
; X86-NEXT:    vzeroupper
; X86-NEXT:    retl
    %res0 = call <16 x i8> @llvm.x86.avx512.mask.pmov.db.512(<16 x i32> %x0, <16 x i8> %x1, i16 -1)
    %res1 = call <16 x i8> @llvm.x86.avx512.mask.pmov.db.512(<16 x i32> %x0, <16 x i8> %x1, i16 %x2)
    %res2 = call <16 x i8> @llvm.x86.avx512.mask.pmov.db.512(<16 x i32> %x0, <16 x i8> zeroinitializer, i16 %x2)
    %res3 = add <16 x i8> %res0, %res1
    %res4 = add <16 x i8> %res3, %res2
    ret <16 x i8> %res4
}

declare void @llvm.x86.avx512.mask.pmov.db.mem.512(i8* %ptr, <16 x i32>, i16)

define void @test_int_x86_avx512_mask_pmov_db_mem_512(i8* %ptr, <16 x i32> %x1, i16 %x2) {
; X64-LABEL: test_int_x86_avx512_mask_pmov_db_mem_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1
; X64-NEXT:    vpmovdb %zmm0, (%rdi)
; X64-NEXT:    vpmovdb %zmm0, (%rdi) {%k1}
; X64-NEXT:    vzeroupper
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_pmov_db_mem_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    vpmovdb %zmm0, (%eax)
; X86-NEXT:    vpmovdb %zmm0, (%eax) {%k1}
; X86-NEXT:    vzeroupper
; X86-NEXT:    retl
    call void @llvm.x86.avx512.mask.pmov.db.mem.512(i8* %ptr, <16 x i32> %x1, i16 -1)
    call void @llvm.x86.avx512.mask.pmov.db.mem.512(i8* %ptr, <16 x i32> %x1, i16 %x2)
    ret void
}

declare <16 x i8> @llvm.x86.avx512.mask.pmovs.db.512(<16 x i32>, <16 x i8>, i16)

define <16 x i8>@test_int_x86_avx512_mask_pmovs_db_512(<16 x i32> %x0, <16 x i8> %x1, i16 %x2) {
; X64-LABEL: test_int_x86_avx512_mask_pmovs_db_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpmovsdb %zmm0, %xmm2
; X64-NEXT:    vpmovsdb %zmm0, %xmm1 {%k1}
; X64-NEXT:    vpmovsdb %zmm0, %xmm0 {%k1} {z}
; X64-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
; X64-NEXT:    vpaddb %xmm0, %xmm2, %xmm0
; X64-NEXT:    vzeroupper
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_pmovs_db_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vpmovsdb %zmm0, %xmm2
; X86-NEXT:    vpmovsdb %zmm0, %xmm1 {%k1}
; X86-NEXT:    vpmovsdb %zmm0, %xmm0 {%k1} {z}
; X86-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
; X86-NEXT:    vpaddb %xmm0, %xmm2, %xmm0
; X86-NEXT:    vzeroupper
; X86-NEXT:    retl
    %res0 = call <16 x i8> @llvm.x86.avx512.mask.pmovs.db.512(<16 x i32> %x0, <16 x i8> %x1, i16 -1)
    %res1 = call <16 x i8> @llvm.x86.avx512.mask.pmovs.db.512(<16 x i32> %x0, <16 x i8> %x1, i16 %x2)
    %res2 = call <16 x i8> @llvm.x86.avx512.mask.pmovs.db.512(<16 x i32> %x0, <16 x i8> zeroinitializer, i16 %x2)
    %res3 = add <16 x i8> %res0, %res1
    %res4 = add <16 x i8> %res3, %res2
    ret <16 x i8> %res4
}

declare void @llvm.x86.avx512.mask.pmovs.db.mem.512(i8* %ptr, <16 x i32>, i16)

define void @test_int_x86_avx512_mask_pmovs_db_mem_512(i8* %ptr, <16 x i32> %x1, i16 %x2) {
; X64-LABEL: test_int_x86_avx512_mask_pmovs_db_mem_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1
; X64-NEXT:    vpmovsdb %zmm0, (%rdi)
; X64-NEXT:    vpmovsdb %zmm0, (%rdi) {%k1}
; X64-NEXT:    vzeroupper
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_pmovs_db_mem_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    vpmovsdb %zmm0, (%eax)
; X86-NEXT:    vpmovsdb %zmm0, (%eax) {%k1}
; X86-NEXT:    vzeroupper
; X86-NEXT:    retl
    call void @llvm.x86.avx512.mask.pmovs.db.mem.512(i8* %ptr, <16 x i32> %x1, i16 -1)
    call void @llvm.x86.avx512.mask.pmovs.db.mem.512(i8* %ptr, <16 x i32> %x1, i16 %x2)
    ret void
}

declare <16 x i8> @llvm.x86.avx512.mask.pmovus.db.512(<16 x i32>, <16 x i8>, i16)

define <16 x i8>@test_int_x86_avx512_mask_pmovus_db_512(<16 x i32> %x0, <16 x i8> %x1, i16 %x2) {
; X64-LABEL: test_int_x86_avx512_mask_pmovus_db_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpmovusdb %zmm0, %xmm2
; X64-NEXT:    vpmovusdb %zmm0, %xmm1 {%k1}
; X64-NEXT:    vpmovusdb %zmm0, %xmm0 {%k1} {z}
; X64-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
; X64-NEXT:    vpaddb %xmm0, %xmm2, %xmm0
; X64-NEXT:    vzeroupper
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_pmovus_db_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vpmovusdb %zmm0, %xmm2
; X86-NEXT:    vpmovusdb %zmm0, %xmm1 {%k1}
; X86-NEXT:    vpmovusdb %zmm0, %xmm0 {%k1} {z}
; X86-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
; X86-NEXT:    vpaddb %xmm0, %xmm2, %xmm0
; X86-NEXT:    vzeroupper
; X86-NEXT:    retl
    %res0 = call <16 x i8> @llvm.x86.avx512.mask.pmovus.db.512(<16 x i32> %x0, <16 x i8> %x1, i16 -1)
    %res1 = call <16 x i8> @llvm.x86.avx512.mask.pmovus.db.512(<16 x i32> %x0, <16 x i8> %x1, i16 %x2)
    %res2 = call <16 x i8> @llvm.x86.avx512.mask.pmovus.db.512(<16 x i32> %x0, <16 x i8> zeroinitializer, i16 %x2)
    %res3 = add <16 x i8> %res0, %res1
    %res4 = add <16 x i8> %res3, %res2
    ret <16 x i8> %res4
}

declare void @llvm.x86.avx512.mask.pmovus.db.mem.512(i8* %ptr, <16 x i32>, i16)

define void @test_int_x86_avx512_mask_pmovus_db_mem_512(i8* %ptr, <16 x i32> %x1, i16 %x2) {
; X64-LABEL: test_int_x86_avx512_mask_pmovus_db_mem_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1
; X64-NEXT:    vpmovusdb %zmm0, (%rdi)
; X64-NEXT:    vpmovusdb %zmm0, (%rdi) {%k1}
; X64-NEXT:    vzeroupper
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_pmovus_db_mem_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    vpmovusdb %zmm0, (%eax)
; X86-NEXT:    vpmovusdb %zmm0, (%eax) {%k1}
; X86-NEXT:    vzeroupper
; X86-NEXT:    retl
    call void @llvm.x86.avx512.mask.pmovus.db.mem.512(i8* %ptr, <16 x i32> %x1, i16 -1)
    call void @llvm.x86.avx512.mask.pmovus.db.mem.512(i8* %ptr, <16 x i32> %x1, i16 %x2)
    ret void
}

declare <16 x i16> @llvm.x86.avx512.mask.pmov.dw.512(<16 x i32>, <16 x i16>, i16)

define <16 x i16>@test_int_x86_avx512_mask_pmov_dw_512(<16 x i32> %x0, <16 x i16> %x1, i16 %x2) {
; X64-LABEL: test_int_x86_avx512_mask_pmov_dw_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpmovdw %zmm0, %ymm2
; X64-NEXT:    vpmovdw %zmm0, %ymm1 {%k1}
; X64-NEXT:    vpmovdw %zmm0, %ymm0 {%k1} {z}
; X64-NEXT:    vpaddw %ymm0, %ymm1, %ymm0
; X64-NEXT:    vpaddw %ymm0, %ymm2, %ymm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_pmov_dw_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vpmovdw %zmm0, %ymm2
; X86-NEXT:    vpmovdw %zmm0, %ymm1 {%k1}
; X86-NEXT:    vpmovdw %zmm0, %ymm0 {%k1} {z}
; X86-NEXT:    vpaddw %ymm0, %ymm1, %ymm0
; X86-NEXT:    vpaddw %ymm0, %ymm2, %ymm0
; X86-NEXT:    retl
    %res0 = call <16 x i16> @llvm.x86.avx512.mask.pmov.dw.512(<16 x i32> %x0, <16 x i16> %x1, i16 -1)
    %res1 = call <16 x i16> @llvm.x86.avx512.mask.pmov.dw.512(<16 x i32> %x0, <16 x i16> %x1, i16 %x2)
    %res2 = call <16 x i16> @llvm.x86.avx512.mask.pmov.dw.512(<16 x i32> %x0, <16 x i16> zeroinitializer, i16 %x2)
    %res3 = add <16 x i16> %res0, %res1
    %res4 = add <16 x i16> %res3, %res2
    ret <16 x i16> %res4
}

declare void @llvm.x86.avx512.mask.pmov.dw.mem.512(i8* %ptr, <16 x i32>, i16)

define void @test_int_x86_avx512_mask_pmov_dw_mem_512(i8* %ptr, <16 x i32> %x1, i16 %x2) {
; X64-LABEL: test_int_x86_avx512_mask_pmov_dw_mem_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1
; X64-NEXT:    vpmovdw %zmm0, (%rdi)
; X64-NEXT:    vpmovdw %zmm0, (%rdi) {%k1}
; X64-NEXT:    vzeroupper
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_pmov_dw_mem_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    vpmovdw %zmm0, (%eax)
; X86-NEXT:    vpmovdw %zmm0, (%eax) {%k1}
; X86-NEXT:    vzeroupper
; X86-NEXT:    retl
    call void @llvm.x86.avx512.mask.pmov.dw.mem.512(i8* %ptr, <16 x i32> %x1, i16 -1)
    call void @llvm.x86.avx512.mask.pmov.dw.mem.512(i8* %ptr, <16 x i32> %x1, i16 %x2)
    ret void
}

declare <16 x i16> @llvm.x86.avx512.mask.pmovs.dw.512(<16 x i32>, <16 x i16>, i16)

define <16 x i16>@test_int_x86_avx512_mask_pmovs_dw_512(<16 x i32> %x0, <16 x i16> %x1, i16 %x2) {
; X64-LABEL: test_int_x86_avx512_mask_pmovs_dw_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpmovsdw %zmm0, %ymm2
; X64-NEXT:    vpmovsdw %zmm0, %ymm1 {%k1}
; X64-NEXT:    vpmovsdw %zmm0, %ymm0 {%k1} {z}
; X64-NEXT:    vpaddw %ymm0, %ymm1, %ymm0
; X64-NEXT:    vpaddw %ymm0, %ymm2, %ymm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_pmovs_dw_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vpmovsdw %zmm0, %ymm2
; X86-NEXT:    vpmovsdw %zmm0, %ymm1 {%k1}
; X86-NEXT:    vpmovsdw %zmm0, %ymm0 {%k1} {z}
; X86-NEXT:    vpaddw %ymm0, %ymm1, %ymm0
; X86-NEXT:    vpaddw %ymm0, %ymm2, %ymm0
; X86-NEXT:    retl
    %res0 = call <16 x i16> @llvm.x86.avx512.mask.pmovs.dw.512(<16 x i32> %x0, <16 x i16> %x1, i16 -1)
    %res1 = call <16 x i16> @llvm.x86.avx512.mask.pmovs.dw.512(<16 x i32> %x0, <16 x i16> %x1, i16 %x2)
    %res2 = call <16 x i16> @llvm.x86.avx512.mask.pmovs.dw.512(<16 x i32> %x0, <16 x i16> zeroinitializer, i16 %x2)
    %res3 = add <16 x i16> %res0, %res1
    %res4 = add <16 x i16> %res3, %res2
    ret <16 x i16> %res4
}

declare void @llvm.x86.avx512.mask.pmovs.dw.mem.512(i8* %ptr, <16 x i32>, i16)

define void @test_int_x86_avx512_mask_pmovs_dw_mem_512(i8* %ptr, <16 x i32> %x1, i16 %x2) {
; X64-LABEL: test_int_x86_avx512_mask_pmovs_dw_mem_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1
; X64-NEXT:    vpmovsdw %zmm0, (%rdi)
; X64-NEXT:    vpmovsdw %zmm0, (%rdi) {%k1}
; X64-NEXT:    vzeroupper
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_pmovs_dw_mem_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    vpmovsdw %zmm0, (%eax)
; X86-NEXT:    vpmovsdw %zmm0, (%eax) {%k1}
; X86-NEXT:    vzeroupper
; X86-NEXT:    retl
    call void @llvm.x86.avx512.mask.pmovs.dw.mem.512(i8* %ptr, <16 x i32> %x1, i16 -1)
    call void @llvm.x86.avx512.mask.pmovs.dw.mem.512(i8* %ptr, <16 x i32> %x1, i16 %x2)
    ret void
}

declare <16 x i16> @llvm.x86.avx512.mask.pmovus.dw.512(<16 x i32>, <16 x i16>, i16)

define <16 x i16>@test_int_x86_avx512_mask_pmovus_dw_512(<16 x i32> %x0, <16 x i16> %x1, i16 %x2) {
; X64-LABEL: test_int_x86_avx512_mask_pmovus_dw_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpmovusdw %zmm0, %ymm2
; X64-NEXT:    vpmovusdw %zmm0, %ymm1 {%k1}
; X64-NEXT:    vpmovusdw %zmm0, %ymm0 {%k1} {z}
; X64-NEXT:    vpaddw %ymm0, %ymm1, %ymm0
; X64-NEXT:    vpaddw %ymm0, %ymm2, %ymm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_pmovus_dw_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vpmovusdw %zmm0, %ymm2
; X86-NEXT:    vpmovusdw %zmm0, %ymm1 {%k1}
; X86-NEXT:    vpmovusdw %zmm0, %ymm0 {%k1} {z}
; X86-NEXT:    vpaddw %ymm0, %ymm1, %ymm0
; X86-NEXT:    vpaddw %ymm0, %ymm2, %ymm0
; X86-NEXT:    retl
    %res0 = call <16 x i16> @llvm.x86.avx512.mask.pmovus.dw.512(<16 x i32> %x0, <16 x i16> %x1, i16 -1)
    %res1 = call <16 x i16> @llvm.x86.avx512.mask.pmovus.dw.512(<16 x i32> %x0, <16 x i16> %x1, i16 %x2)
    %res2 = call <16 x i16> @llvm.x86.avx512.mask.pmovus.dw.512(<16 x i32> %x0, <16 x i16> zeroinitializer, i16 %x2)
    %res3 = add <16 x i16> %res0, %res1
    %res4 = add <16 x i16> %res3, %res2
    ret <16 x i16> %res4
}

declare void @llvm.x86.avx512.mask.pmovus.dw.mem.512(i8* %ptr, <16 x i32>, i16)

define void @test_int_x86_avx512_mask_pmovus_dw_mem_512(i8* %ptr, <16 x i32> %x1, i16 %x2) {
; X64-LABEL: test_int_x86_avx512_mask_pmovus_dw_mem_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1
; X64-NEXT:    vpmovusdw %zmm0, (%rdi)
; X64-NEXT:    vpmovusdw %zmm0, (%rdi) {%k1}
; X64-NEXT:    vzeroupper
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_pmovus_dw_mem_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    vpmovusdw %zmm0, (%eax)
; X86-NEXT:    vpmovusdw %zmm0, (%eax) {%k1}
; X86-NEXT:    vzeroupper
; X86-NEXT:    retl
    call void @llvm.x86.avx512.mask.pmovus.dw.mem.512(i8* %ptr, <16 x i32> %x1, i16 -1)
    call void @llvm.x86.avx512.mask.pmovus.dw.mem.512(i8* %ptr, <16 x i32> %x1, i16 %x2)
    ret void
}

declare <16 x float> @llvm.x86.avx512.sitofp.round.v16f32.v16i32(<16 x i32>, i32)

define <16 x float>@test_int_x86_avx512_mask_cvt_dq2ps_512(<16 x i32> %x0, <16 x float> %x1, i16 %x2) {
; X64-LABEL: test_int_x86_avx512_mask_cvt_dq2ps_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vcvtdq2ps %zmm0, %zmm1 {%k1}
; X64-NEXT:    vcvtdq2ps {rn-sae}, %zmm0, %zmm0
; X64-NEXT:    vaddps %zmm0, %zmm1, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_cvt_dq2ps_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vcvtdq2ps %zmm0, %zmm1 {%k1}
; X86-NEXT:    vcvtdq2ps {rn-sae}, %zmm0, %zmm0
; X86-NEXT:    vaddps %zmm0, %zmm1, %zmm0
; X86-NEXT:    retl
  %cvt = sitofp <16 x i32> %x0 to <16 x float>
  %1 = bitcast i16 %x2 to <16 x i1>
  %2 = select <16 x i1> %1, <16 x float> %cvt, <16 x float> %x1
  %3 = call <16 x float> @llvm.x86.avx512.sitofp.round.v16f32.v16i32(<16 x i32> %x0, i32 0)
  %res2 = fadd <16 x float> %2, %3
  ret <16 x float> %res2
}

declare <8 x i32> @llvm.x86.avx512.mask.cvtpd2dq.512(<8 x double>, <8 x i32>, i8, i32)

define <8 x i32>@test_int_x86_avx512_mask_cvt_pd2dq_512(<8 x double> %x0, <8 x i32> %x1, i8 %x2) {
; X64-LABEL: test_int_x86_avx512_mask_cvt_pd2dq_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vcvtpd2dq %zmm0, %ymm1 {%k1}
; X64-NEXT:    vcvtpd2dq {rn-sae}, %zmm0, %ymm0
; X64-NEXT:    vpaddd %ymm0, %ymm1, %ymm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_cvt_pd2dq_512:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vcvtpd2dq %zmm0, %ymm1 {%k1}
; X86-NEXT:    vcvtpd2dq {rn-sae}, %zmm0, %ymm0
; X86-NEXT:    vpaddd %ymm0, %ymm1, %ymm0
; X86-NEXT:    retl
  %res = call <8 x i32> @llvm.x86.avx512.mask.cvtpd2dq.512(<8 x double> %x0, <8 x i32> %x1, i8 %x2, i32 4)
  %res1 = call <8 x i32> @llvm.x86.avx512.mask.cvtpd2dq.512(<8 x double> %x0, <8 x i32> %x1, i8 -1, i32 0)
  %res2 = add <8 x i32> %res, %res1
  ret <8 x i32> %res2
}

declare <8 x float> @llvm.x86.avx512.mask.cvtpd2ps.512(<8 x double>, <8 x float>, i8, i32)

define <8 x float>@test_int_x86_avx512_mask_cvt_pd2ps_512(<8 x double> %x0, <8 x float> %x1, i8 %x2) {
; X64-LABEL: test_int_x86_avx512_mask_cvt_pd2ps_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vcvtpd2ps %zmm0, %ymm1 {%k1}
; X64-NEXT:    vcvtpd2ps {ru-sae}, %zmm0, %ymm0
; X64-NEXT:    vaddps %ymm0, %ymm1, %ymm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_cvt_pd2ps_512:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vcvtpd2ps %zmm0, %ymm1 {%k1}
; X86-NEXT:    vcvtpd2ps {ru-sae}, %zmm0, %ymm0
; X86-NEXT:    vaddps %ymm0, %ymm1, %ymm0
; X86-NEXT:    retl
  %res = call <8 x float> @llvm.x86.avx512.mask.cvtpd2ps.512(<8 x double> %x0, <8 x float> %x1, i8 %x2, i32 4)
  %res1 = call <8 x float> @llvm.x86.avx512.mask.cvtpd2ps.512(<8 x double> %x0, <8 x float> %x1, i8 -1, i32 2)
  %res2 = fadd <8 x float> %res, %res1
  ret <8 x float> %res2
}

declare <8 x i32> @llvm.x86.avx512.mask.cvtpd2udq.512(<8 x double>, <8 x i32>, i8, i32)

define <8 x i32>@test_int_x86_avx512_mask_cvt_pd2udq_512(<8 x double> %x0, <8 x i32> %x1, i8 %x2) {
; X64-LABEL: test_int_x86_avx512_mask_cvt_pd2udq_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vcvtpd2udq {ru-sae}, %zmm0, %ymm1 {%k1}
; X64-NEXT:    vcvtpd2udq {rn-sae}, %zmm0, %ymm0
; X64-NEXT:    vpaddd %ymm0, %ymm1, %ymm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_cvt_pd2udq_512:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vcvtpd2udq {ru-sae}, %zmm0, %ymm1 {%k1}
; X86-NEXT:    vcvtpd2udq {rn-sae}, %zmm0, %ymm0
; X86-NEXT:    vpaddd %ymm0, %ymm1, %ymm0
; X86-NEXT:    retl
  %res = call <8 x i32> @llvm.x86.avx512.mask.cvtpd2udq.512(<8 x double> %x0, <8 x i32> %x1, i8 %x2, i32 2)
  %res1 = call <8 x i32> @llvm.x86.avx512.mask.cvtpd2udq.512(<8 x double> %x0, <8 x i32> %x1, i8 -1, i32 0)
  %res2 = add <8 x i32> %res, %res1
  ret <8 x i32> %res2
}

declare <16 x i32> @llvm.x86.avx512.mask.cvtps2dq.512(<16 x float>, <16 x i32>, i16, i32)

define <16 x i32>@test_int_x86_avx512_mask_cvt_ps2dq_512(<16 x float> %x0, <16 x i32> %x1, i16 %x2) {
; X64-LABEL: test_int_x86_avx512_mask_cvt_ps2dq_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vcvtps2dq {ru-sae}, %zmm0, %zmm1 {%k1}
; X64-NEXT:    vcvtps2dq {rn-sae}, %zmm0, %zmm0
; X64-NEXT:    vpaddd %zmm0, %zmm1, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_cvt_ps2dq_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vcvtps2dq {ru-sae}, %zmm0, %zmm1 {%k1}
; X86-NEXT:    vcvtps2dq {rn-sae}, %zmm0, %zmm0
; X86-NEXT:    vpaddd %zmm0, %zmm1, %zmm0
; X86-NEXT:    retl
  %res = call <16 x i32> @llvm.x86.avx512.mask.cvtps2dq.512(<16 x float> %x0, <16 x i32> %x1, i16 %x2, i32 2)
  %res1 = call <16 x i32> @llvm.x86.avx512.mask.cvtps2dq.512(<16 x float> %x0, <16 x i32> %x1, i16 -1, i32 0)
  %res2 = add <16 x i32> %res, %res1
  ret <16 x i32> %res2
}

declare <8 x double> @llvm.x86.avx512.mask.cvtps2pd.512(<8 x float>, <8 x double>, i8, i32)

define <8 x double>@test_int_x86_avx512_mask_cvt_ps2pd_512(<8 x float> %x0, <8 x double> %x1, i8 %x2) {
; X64-LABEL: test_int_x86_avx512_mask_cvt_ps2pd_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vcvtps2pd %ymm0, %zmm1 {%k1}
; X64-NEXT:    vcvtps2pd {sae}, %ymm0, %zmm0
; X64-NEXT:    vaddpd %zmm0, %zmm1, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_cvt_ps2pd_512:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vcvtps2pd %ymm0, %zmm1 {%k1}
; X86-NEXT:    vcvtps2pd {sae}, %ymm0, %zmm0
; X86-NEXT:    vaddpd %zmm0, %zmm1, %zmm0
; X86-NEXT:    retl
  %res = call <8 x double> @llvm.x86.avx512.mask.cvtps2pd.512(<8 x float> %x0, <8 x double> %x1, i8 %x2, i32 4)
  %res1 = call <8 x double> @llvm.x86.avx512.mask.cvtps2pd.512(<8 x float> %x0, <8 x double> %x1, i8 -1, i32 8)
  %res2 = fadd <8 x double> %res, %res1
  ret <8 x double> %res2
}

declare <16 x i32> @llvm.x86.avx512.mask.cvtps2udq.512(<16 x float>, <16 x i32>, i16, i32)

define <16 x i32>@test_int_x86_avx512_mask_cvt_ps2udq_512(<16 x float> %x0, <16 x i32> %x1, i16 %x2) {
; X64-LABEL: test_int_x86_avx512_mask_cvt_ps2udq_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vcvtps2udq {ru-sae}, %zmm0, %zmm1 {%k1}
; X64-NEXT:    vcvtps2udq {rn-sae}, %zmm0, %zmm0
; X64-NEXT:    vpaddd %zmm0, %zmm1, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_cvt_ps2udq_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vcvtps2udq {ru-sae}, %zmm0, %zmm1 {%k1}
; X86-NEXT:    vcvtps2udq {rn-sae}, %zmm0, %zmm0
; X86-NEXT:    vpaddd %zmm0, %zmm1, %zmm0
; X86-NEXT:    retl
  %res = call <16 x i32> @llvm.x86.avx512.mask.cvtps2udq.512(<16 x float> %x0, <16 x i32> %x1, i16 %x2, i32 2)
  %res1 = call <16 x i32> @llvm.x86.avx512.mask.cvtps2udq.512(<16 x float> %x0, <16 x i32> %x1, i16 -1, i32 0)
  %res2 = add <16 x i32> %res, %res1
  ret <16 x i32> %res2
}

declare <8 x i32> @llvm.x86.avx512.mask.cvttpd2dq.512(<8 x double>, <8 x i32>, i8, i32)

define <8 x i32>@test_int_x86_avx512_mask_cvtt_pd2dq_512(<8 x double> %x0, <8 x i32> %x1, i8 %x2) {
; X64-LABEL: test_int_x86_avx512_mask_cvtt_pd2dq_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vcvttpd2dq %zmm0, %ymm1 {%k1}
; X64-NEXT:    vcvttpd2dq {sae}, %zmm0, %ymm0
; X64-NEXT:    vpaddd %ymm0, %ymm1, %ymm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_cvtt_pd2dq_512:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vcvttpd2dq %zmm0, %ymm1 {%k1}
; X86-NEXT:    vcvttpd2dq {sae}, %zmm0, %ymm0
; X86-NEXT:    vpaddd %ymm0, %ymm1, %ymm0
; X86-NEXT:    retl
  %res = call <8 x i32> @llvm.x86.avx512.mask.cvttpd2dq.512(<8 x double> %x0, <8 x i32> %x1, i8 %x2, i32 4)
  %res1 = call <8 x i32> @llvm.x86.avx512.mask.cvttpd2dq.512(<8 x double> %x0, <8 x i32> %x1, i8 -1, i32 8)
  %res2 = add <8 x i32> %res, %res1
  ret <8 x i32> %res2
}

declare <16 x float> @llvm.x86.avx512.uitofp.round.v16f32.v16i32(<16 x i32>, i32)

define <16 x float>@test_int_x86_avx512_mask_cvt_udq2ps_512(<16 x i32> %x0, <16 x float> %x1, i16 %x2) {
; X64-LABEL: test_int_x86_avx512_mask_cvt_udq2ps_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vcvtudq2ps %zmm0, %zmm1 {%k1}
; X64-NEXT:    vcvtudq2ps {rn-sae}, %zmm0, %zmm0
; X64-NEXT:    vaddps %zmm0, %zmm1, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_cvt_udq2ps_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vcvtudq2ps %zmm0, %zmm1 {%k1}
; X86-NEXT:    vcvtudq2ps {rn-sae}, %zmm0, %zmm0
; X86-NEXT:    vaddps %zmm0, %zmm1, %zmm0
; X86-NEXT:    retl
  %cvt = uitofp <16 x i32> %x0 to <16 x float>
  %1 = bitcast i16 %x2 to <16 x i1>
  %2 = select <16 x i1> %1, <16 x float> %cvt, <16 x float> %x1
  %3 = call <16 x float> @llvm.x86.avx512.uitofp.round.v16f32.v16i32(<16 x i32> %x0, i32 0)
  %res2 = fadd <16 x float> %2, %3
  ret <16 x float> %res2
}

declare <8 x i32> @llvm.x86.avx512.mask.cvttpd2udq.512(<8 x double>, <8 x i32>, i8, i32)

define <8 x i32>@test_int_x86_avx512_mask_cvtt_pd2udq_512(<8 x double> %x0, <8 x i32> %x1, i8 %x2) {
; X64-LABEL: test_int_x86_avx512_mask_cvtt_pd2udq_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vcvttpd2udq %zmm0, %ymm1 {%k1}
; X64-NEXT:    vcvttpd2udq {sae}, %zmm0, %ymm0
; X64-NEXT:    vpaddd %ymm0, %ymm1, %ymm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_cvtt_pd2udq_512:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vcvttpd2udq %zmm0, %ymm1 {%k1}
; X86-NEXT:    vcvttpd2udq {sae}, %zmm0, %ymm0
; X86-NEXT:    vpaddd %ymm0, %ymm1, %ymm0
; X86-NEXT:    retl
  %res = call <8 x i32> @llvm.x86.avx512.mask.cvttpd2udq.512(<8 x double> %x0, <8 x i32> %x1, i8 %x2, i32 4)
  %res1 = call <8 x i32> @llvm.x86.avx512.mask.cvttpd2udq.512(<8 x double> %x0, <8 x i32> %x1, i8 -1, i32 8)
  %res2 = add <8 x i32> %res, %res1
  ret <8 x i32> %res2
}

declare <16 x i32> @llvm.x86.avx512.mask.cvttps2dq.512(<16 x float>, <16 x i32>, i16, i32)

define <16 x i32>@test_int_x86_avx512_mask_cvtt_ps2dq_512(<16 x float> %x0, <16 x i32> %x1, i16 %x2) {
; X64-LABEL: test_int_x86_avx512_mask_cvtt_ps2dq_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vcvttps2dq %zmm0, %zmm1 {%k1}
; X64-NEXT:    vcvttps2dq {sae}, %zmm0, %zmm0
; X64-NEXT:    vpaddd %zmm0, %zmm1, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_cvtt_ps2dq_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vcvttps2dq %zmm0, %zmm1 {%k1}
; X86-NEXT:    vcvttps2dq {sae}, %zmm0, %zmm0
; X86-NEXT:    vpaddd %zmm0, %zmm1, %zmm0
; X86-NEXT:    retl
  %res = call <16 x i32> @llvm.x86.avx512.mask.cvttps2dq.512(<16 x float> %x0, <16 x i32> %x1, i16 %x2, i32 4)
  %res1 = call <16 x i32> @llvm.x86.avx512.mask.cvttps2dq.512(<16 x float> %x0, <16 x i32> %x1, i16 -1, i32 8)
  %res2 = add <16 x i32> %res, %res1
  ret <16 x i32> %res2
}

declare <16 x i32> @llvm.x86.avx512.mask.cvttps2udq.512(<16 x float>, <16 x i32>, i16, i32)

define <16 x i32>@test_int_x86_avx512_mask_cvtt_ps2udq_512(<16 x float> %x0, <16 x i32> %x1, i16 %x2) {
; X64-LABEL: test_int_x86_avx512_mask_cvtt_ps2udq_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vcvttps2udq %zmm0, %zmm1 {%k1}
; X64-NEXT:    vcvttps2udq {sae}, %zmm0, %zmm0
; X64-NEXT:    vpaddd %zmm0, %zmm1, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_cvtt_ps2udq_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vcvttps2udq %zmm0, %zmm1 {%k1}
; X86-NEXT:    vcvttps2udq {sae}, %zmm0, %zmm0
; X86-NEXT:    vpaddd %zmm0, %zmm1, %zmm0
; X86-NEXT:    retl
  %res = call <16 x i32> @llvm.x86.avx512.mask.cvttps2udq.512(<16 x float> %x0, <16 x i32> %x1, i16 %x2, i32 4)
  %res1 = call <16 x i32> @llvm.x86.avx512.mask.cvttps2udq.512(<16 x float> %x0, <16 x i32> %x1, i16 -1, i32 8)
  %res2 = add <16 x i32> %res, %res1
  ret <16 x i32> %res2
}

declare <4 x float> @llvm.x86.avx512.mask.getexp.ss(<4 x float>, <4 x float>, <4 x float>, i8, i32) nounwind readnone

define <4 x float> @test_getexp_ss(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2, i8 %mask) {
; X64-LABEL: test_getexp_ss:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vmovaps %xmm2, %xmm3
; X64-NEXT:    vgetexpss %xmm1, %xmm0, %xmm3 {%k1}
; X64-NEXT:    vgetexpss {sae}, %xmm1, %xmm0, %xmm4 {%k1} {z}
; X64-NEXT:    vgetexpss {sae}, %xmm1, %xmm0, %xmm5
; X64-NEXT:    vaddps %xmm5, %xmm4, %xmm4
; X64-NEXT:    vgetexpss {sae}, %xmm1, %xmm0, %xmm2 {%k1}
; X64-NEXT:    vaddps %xmm2, %xmm3, %xmm0
; X64-NEXT:    vaddps %xmm4, %xmm0, %xmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_getexp_ss:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vmovaps %xmm2, %xmm3
; X86-NEXT:    vgetexpss %xmm1, %xmm0, %xmm3 {%k1}
; X86-NEXT:    vgetexpss {sae}, %xmm1, %xmm0, %xmm2 {%k1}
; X86-NEXT:    vaddps %xmm2, %xmm3, %xmm2
; X86-NEXT:    vgetexpss {sae}, %xmm1, %xmm0, %xmm3 {%k1} {z}
; X86-NEXT:    vgetexpss {sae}, %xmm1, %xmm0, %xmm0
; X86-NEXT:    vaddps %xmm0, %xmm3, %xmm0
; X86-NEXT:    vaddps %xmm0, %xmm2, %xmm0
; X86-NEXT:    retl
  %res0 = call <4 x float> @llvm.x86.avx512.mask.getexp.ss(<4 x float>%a0, <4 x float> %a1, <4 x float> %a2, i8 %mask, i32 4)
  %res1 = call <4 x float> @llvm.x86.avx512.mask.getexp.ss(<4 x float>%a0, <4 x float> %a1, <4 x float> %a2, i8 %mask, i32 8)
  %res2 = call <4 x float> @llvm.x86.avx512.mask.getexp.ss(<4 x float>%a0, <4 x float> %a1, <4 x float> zeroinitializer, i8 %mask, i32 8)
  %res3 = call <4 x float> @llvm.x86.avx512.mask.getexp.ss(<4 x float>%a0, <4 x float> %a1, <4 x float> zeroinitializer, i8 -1, i32 8)

  %res.1 = fadd <4 x float> %res0, %res1
  %res.2 = fadd <4 x float> %res2, %res3
  %res   = fadd <4 x float> %res.1, %res.2
  ret <4 x float> %res
}

declare <2 x double> @llvm.x86.avx512.mask.getexp.sd(<2 x double>, <2 x double>, <2 x double>, i8, i32) nounwind readnone

define <2 x double> @test_getexp_sd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2, i8 %mask) {
; X64-LABEL: test_getexp_sd:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vgetexpsd %xmm1, %xmm0, %xmm3
; X64-NEXT:    vmovapd %xmm2, %xmm4
; X64-NEXT:    vgetexpsd %xmm1, %xmm0, %xmm4 {%k1}
; X64-NEXT:    vgetexpsd {sae}, %xmm1, %xmm0, %xmm5 {%k1} {z}
; X64-NEXT:    vaddpd %xmm3, %xmm5, %xmm3
; X64-NEXT:    vgetexpsd {sae}, %xmm1, %xmm0, %xmm2 {%k1}
; X64-NEXT:    vaddpd %xmm2, %xmm4, %xmm0
; X64-NEXT:    vaddpd %xmm3, %xmm0, %xmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_getexp_sd:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vmovapd %xmm2, %xmm3
; X86-NEXT:    vgetexpsd {sae}, %xmm1, %xmm0, %xmm3 {%k1}
; X86-NEXT:    vgetexpsd {sae}, %xmm1, %xmm0, %xmm4 {%k1} {z}
; X86-NEXT:    vgetexpsd %xmm1, %xmm0, %xmm2 {%k1}
; X86-NEXT:    vaddpd %xmm3, %xmm2, %xmm2
; X86-NEXT:    vgetexpsd %xmm1, %xmm0, %xmm0
; X86-NEXT:    vaddpd %xmm0, %xmm4, %xmm0
; X86-NEXT:    vaddpd %xmm0, %xmm2, %xmm0
; X86-NEXT:    retl
  %res0 = call <2 x double> @llvm.x86.avx512.mask.getexp.sd(<2 x double>%a0, <2 x double> %a1, <2 x double> %a2, i8 %mask, i32 4)
  %res1 = call <2 x double> @llvm.x86.avx512.mask.getexp.sd(<2 x double>%a0, <2 x double> %a1, <2 x double> %a2, i8 %mask, i32 8)
  %res2 = call <2 x double> @llvm.x86.avx512.mask.getexp.sd(<2 x double>%a0, <2 x double> %a1, <2 x double> zeroinitializer, i8 %mask, i32 8)
  %res3 = call <2 x double> @llvm.x86.avx512.mask.getexp.sd(<2 x double>%a0, <2 x double> %a1, <2 x double> zeroinitializer, i8 -1, i32 4)

  %res.1 = fadd <2 x double> %res0, %res1
  %res.2 = fadd <2 x double> %res2, %res3
  %res   = fadd <2 x double> %res.1, %res.2
  ret <2 x double> %res
}

declare i8 @llvm.x86.avx512.mask.cmp.sd(<2 x double>, <2 x double>, i32, i8, i32)

define i8@test_int_x86_avx512_mask_cmp_sd(<2 x double> %x0, <2 x double> %x1, i8 %x3, i32 %x4) {
; X64-LABEL: test_int_x86_avx512_mask_cmp_sd:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vcmpnltsd {sae}, %xmm1, %xmm0, %k0 {%k1}
; X64-NEXT:    kmovw %k0, %eax
; X64-NEXT:    # kill: def $al killed $al killed $eax
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_cmp_sd:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vcmpnltsd {sae}, %xmm1, %xmm0, %k0 {%k1}
; X86-NEXT:    kmovw %k0, %eax
; X86-NEXT:    # kill: def $al killed $al killed $eax
; X86-NEXT:    retl

  %res4 = call i8 @llvm.x86.avx512.mask.cmp.sd(<2 x double> %x0, <2 x double> %x1, i32 5, i8 %x3, i32 8)
  ret i8 %res4
}

define i8@test_int_x86_avx512_mask_cmp_sd_all(<2 x double> %x0, <2 x double> %x1, i8 %x3, i32 %x4) {
; X64-LABEL: test_int_x86_avx512_mask_cmp_sd_all:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vcmplesd %xmm1, %xmm0, %k0
; X64-NEXT:    kmovw %k0, %ecx
; X64-NEXT:    vcmpunordsd {sae}, %xmm1, %xmm0, %k0
; X64-NEXT:    kmovw %k0, %edx
; X64-NEXT:    vcmpneqsd %xmm1, %xmm0, %k0 {%k1}
; X64-NEXT:    kmovw %k0, %esi
; X64-NEXT:    vcmpnltsd {sae}, %xmm1, %xmm0, %k0 {%k1}
; X64-NEXT:    kmovw %k0, %eax
; X64-NEXT:    orb %sil, %al
; X64-NEXT:    orb %dl, %al
; X64-NEXT:    orb %cl, %al
; X64-NEXT:    # kill: def $al killed $al killed $eax
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_cmp_sd_all:
; X86:       # %bb.0:
; X86-NEXT:    pushl %ebx
; X86-NEXT:    .cfi_def_cfa_offset 8
; X86-NEXT:    .cfi_offset %ebx, -8
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vcmplesd %xmm1, %xmm0, %k0
; X86-NEXT:    kmovw %k0, %ecx
; X86-NEXT:    vcmpunordsd {sae}, %xmm1, %xmm0, %k0
; X86-NEXT:    kmovw %k0, %edx
; X86-NEXT:    vcmpneqsd %xmm1, %xmm0, %k0 {%k1}
; X86-NEXT:    kmovw %k0, %ebx
; X86-NEXT:    vcmpnltsd {sae}, %xmm1, %xmm0, %k0 {%k1}
; X86-NEXT:    kmovw %k0, %eax
; X86-NEXT:    orb %bl, %al
; X86-NEXT:    orb %dl, %al
; X86-NEXT:    orb %cl, %al
; X86-NEXT:    # kill: def $al killed $al killed $eax
; X86-NEXT:    popl %ebx
; X86-NEXT:    .cfi_def_cfa_offset 4
; X86-NEXT:    retl

  %res1 = call i8 @llvm.x86.avx512.mask.cmp.sd(<2 x double> %x0, <2 x double> %x1, i32 2, i8 -1, i32 4)
  %res2 = call i8 @llvm.x86.avx512.mask.cmp.sd(<2 x double> %x0, <2 x double> %x1, i32 3, i8 -1, i32 8)
  %res3 = call i8 @llvm.x86.avx512.mask.cmp.sd(<2 x double> %x0, <2 x double> %x1, i32 4, i8 %x3, i32 4)
  %res4 = call i8 @llvm.x86.avx512.mask.cmp.sd(<2 x double> %x0, <2 x double> %x1, i32 5, i8 %x3, i32 8)

  %res11 = or i8 %res1, %res2
  %res12 = or i8 %res3, %res4
  %res13 = or i8 %res11, %res12
  ret i8 %res13
}

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

define i8@test_int_x86_avx512_mask_cmp_ss(<4 x float> %x0, <4 x float> %x1, i8 %x3, i32 %x4) {
; X64-LABEL: test_int_x86_avx512_mask_cmp_ss:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vcmpunordss %xmm1, %xmm0, %k0 {%k1}
; X64-NEXT:    kmovw %k0, %eax
; X64-NEXT:    # kill: def $al killed $al killed $eax
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_cmp_ss:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vcmpunordss %xmm1, %xmm0, %k0 {%k1}
; X86-NEXT:    kmovw %k0, %eax
; X86-NEXT:    # kill: def $al killed $al killed $eax
; X86-NEXT:    retl

  %res2 = call i8 @llvm.x86.avx512.mask.cmp.ss(<4 x float> %x0, <4 x float> %x1, i32 3, i8 %x3, i32 4)
  ret i8 %res2
}


define i8@test_int_x86_avx512_mask_cmp_ss_all(<4 x float> %x0, <4 x float> %x1, i8 %x3, i32 %x4) {
; X64-LABEL: test_int_x86_avx512_mask_cmp_ss_all:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vcmpless %xmm1, %xmm0, %k0
; X64-NEXT:    kmovw %k0, %ecx
; X64-NEXT:    vcmpunordss {sae}, %xmm1, %xmm0, %k0
; X64-NEXT:    kmovw %k0, %edx
; X64-NEXT:    vcmpneqss %xmm1, %xmm0, %k0 {%k1}
; X64-NEXT:    kmovw %k0, %esi
; X64-NEXT:    vcmpnltss {sae}, %xmm1, %xmm0, %k0 {%k1}
; X64-NEXT:    kmovw %k0, %eax
; X64-NEXT:    andb %sil, %al
; X64-NEXT:    andb %dl, %al
; X64-NEXT:    andb %cl, %al
; X64-NEXT:    # kill: def $al killed $al killed $eax
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_cmp_ss_all:
; X86:       # %bb.0:
; X86-NEXT:    pushl %ebx
; X86-NEXT:    .cfi_def_cfa_offset 8
; X86-NEXT:    .cfi_offset %ebx, -8
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vcmpless %xmm1, %xmm0, %k0
; X86-NEXT:    kmovw %k0, %ecx
; X86-NEXT:    vcmpunordss {sae}, %xmm1, %xmm0, %k0
; X86-NEXT:    kmovw %k0, %edx
; X86-NEXT:    vcmpneqss %xmm1, %xmm0, %k0 {%k1}
; X86-NEXT:    kmovw %k0, %ebx
; X86-NEXT:    vcmpnltss {sae}, %xmm1, %xmm0, %k0 {%k1}
; X86-NEXT:    kmovw %k0, %eax
; X86-NEXT:    andb %bl, %al
; X86-NEXT:    andb %dl, %al
; X86-NEXT:    andb %cl, %al
; X86-NEXT:    # kill: def $al killed $al killed $eax
; X86-NEXT:    popl %ebx
; X86-NEXT:    .cfi_def_cfa_offset 4
; X86-NEXT:    retl
  %res1 = call i8 @llvm.x86.avx512.mask.cmp.ss(<4 x float> %x0, <4 x float> %x1, i32 2, i8 -1, i32 4)
  %res2 = call i8 @llvm.x86.avx512.mask.cmp.ss(<4 x float> %x0, <4 x float> %x1, i32 3, i8 -1, i32 8)
  %res3 = call i8 @llvm.x86.avx512.mask.cmp.ss(<4 x float> %x0, <4 x float> %x1, i32 4, i8 %x3, i32 4)
  %res4 = call i8 @llvm.x86.avx512.mask.cmp.ss(<4 x float> %x0, <4 x float> %x1, i32 5, i8 %x3, i32 8)

  %res11 = and i8 %res1, %res2
  %res12 = and i8 %res3, %res4
  %res13 = and i8 %res11, %res12
  ret i8 %res13
}

declare <8 x double> @llvm.x86.avx512.mask.getmant.pd.512(<8 x double>, i32, <8 x double>, i8, i32)

define <8 x double>@test_int_x86_avx512_mask_getmant_pd_512(<8 x double> %x0, <8 x double> %x2, i8 %x3) {
; X64-LABEL: test_int_x86_avx512_mask_getmant_pd_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vgetmantpd $11, %zmm0, %zmm1 {%k1}
; X64-NEXT:    vgetmantpd $11, {sae}, %zmm0, %zmm0
; X64-NEXT:    vaddpd %zmm0, %zmm1, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_getmant_pd_512:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vgetmantpd $11, %zmm0, %zmm1 {%k1}
; X86-NEXT:    vgetmantpd $11, {sae}, %zmm0, %zmm0
; X86-NEXT:    vaddpd %zmm0, %zmm1, %zmm0
; X86-NEXT:    retl
  %res = call <8 x double> @llvm.x86.avx512.mask.getmant.pd.512(<8 x double> %x0, i32 11, <8 x double> %x2, i8 %x3, i32 4)
  %res1 = call <8 x double> @llvm.x86.avx512.mask.getmant.pd.512(<8 x double> %x0, i32 11, <8 x double> %x2, i8 -1, i32 8)
  %res2 = fadd <8 x double> %res, %res1
  ret <8 x double> %res2
}

declare <16 x float> @llvm.x86.avx512.mask.getmant.ps.512(<16 x float>, i32, <16 x float>, i16, i32)

define <16 x float>@test_int_x86_avx512_mask_getmant_ps_512(<16 x float> %x0, <16 x float> %x2, i16 %x3) {
; X64-LABEL: test_int_x86_avx512_mask_getmant_ps_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vgetmantps $11, %zmm0, %zmm1 {%k1}
; X64-NEXT:    vgetmantps $11, {sae}, %zmm0, %zmm0
; X64-NEXT:    vaddps %zmm0, %zmm1, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_getmant_ps_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vgetmantps $11, %zmm0, %zmm1 {%k1}
; X86-NEXT:    vgetmantps $11, {sae}, %zmm0, %zmm0
; X86-NEXT:    vaddps %zmm0, %zmm1, %zmm0
; X86-NEXT:    retl
  %res = call <16 x float> @llvm.x86.avx512.mask.getmant.ps.512(<16 x float> %x0, i32 11, <16 x float> %x2, i16 %x3, i32 4)
  %res1 = call <16 x float> @llvm.x86.avx512.mask.getmant.ps.512(<16 x float> %x0, i32 11, <16 x float> %x2, i16 -1, i32 8)
  %res2 = fadd <16 x float> %res, %res1
  ret <16 x float> %res2
}

declare <2 x double> @llvm.x86.avx512.mask.getmant.sd(<2 x double>, <2 x double>, i32, <2 x double>, i8, i32)

define <2 x double>@test_int_x86_avx512_mask_getmant_sd(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 %x3) {
; X64-LABEL: test_int_x86_avx512_mask_getmant_sd:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vgetmantsd $11, %xmm1, %xmm0, %xmm3
; X64-NEXT:    vmovapd %xmm2, %xmm4
; X64-NEXT:    vgetmantsd $11, %xmm1, %xmm0, %xmm4 {%k1}
; X64-NEXT:    vgetmantsd $11, %xmm1, %xmm0, %xmm5 {%k1} {z}
; X64-NEXT:    vaddpd %xmm5, %xmm4, %xmm4
; X64-NEXT:    vgetmantsd $11, {sae}, %xmm1, %xmm0, %xmm2 {%k1}
; X64-NEXT:    vaddpd %xmm3, %xmm2, %xmm0
; X64-NEXT:    vaddpd %xmm0, %xmm4, %xmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_getmant_sd:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vmovapd %xmm2, %xmm3
; X86-NEXT:    vgetmantsd $11, %xmm1, %xmm0, %xmm3 {%k1}
; X86-NEXT:    vgetmantsd $11, %xmm1, %xmm0, %xmm4 {%k1} {z}
; X86-NEXT:    vaddpd %xmm4, %xmm3, %xmm3
; X86-NEXT:    vgetmantsd $11, %xmm1, %xmm0, %xmm4
; X86-NEXT:    vgetmantsd $11, {sae}, %xmm1, %xmm0, %xmm2 {%k1}
; X86-NEXT:    vaddpd %xmm4, %xmm2, %xmm0
; X86-NEXT:    vaddpd %xmm0, %xmm3, %xmm0
; X86-NEXT:    retl
  %res  = call <2 x double> @llvm.x86.avx512.mask.getmant.sd(<2 x double> %x0, <2 x double> %x1, i32 11, <2 x double> %x2, i8 %x3, i32 4)
  %res1 = call <2 x double> @llvm.x86.avx512.mask.getmant.sd(<2 x double> %x0, <2 x double> %x1, i32 11, <2 x double> zeroinitializer, i8 %x3, i32 4)
  %res2 = call <2 x double> @llvm.x86.avx512.mask.getmant.sd(<2 x double> %x0, <2 x double> %x1, i32 11, <2 x double> %x2, i8 %x3, i32 8)
  %res3 = call <2 x double> @llvm.x86.avx512.mask.getmant.sd(<2 x double> %x0, <2 x double> %x1, i32 11, <2 x double> %x2, i8 -1, i32 4)
  %res11 = fadd <2 x double> %res, %res1
  %res12 = fadd <2 x double> %res2, %res3
  %res13 = fadd <2 x double> %res11, %res12
  ret <2 x double> %res13
}

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

define <4 x float>@test_int_x86_avx512_mask_getmant_ss(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 %x3) {
; X64-LABEL: test_int_x86_avx512_mask_getmant_ss:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vgetmantss $11, %xmm1, %xmm0, %xmm3
; X64-NEXT:    vgetmantss $11, %xmm1, %xmm0, %xmm2 {%k1}
; X64-NEXT:    vgetmantss $11, %xmm1, %xmm0, %xmm4 {%k1} {z}
; X64-NEXT:    vaddps %xmm4, %xmm2, %xmm2
; X64-NEXT:    vgetmantss $11, {sae}, %xmm1, %xmm0, %xmm0
; X64-NEXT:    vaddps %xmm3, %xmm0, %xmm0
; X64-NEXT:    vaddps %xmm0, %xmm2, %xmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_getmant_ss:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vgetmantss $11, %xmm1, %xmm0, %xmm2 {%k1}
; X86-NEXT:    vgetmantss $11, %xmm1, %xmm0, %xmm3 {%k1} {z}
; X86-NEXT:    vaddps %xmm3, %xmm2, %xmm2
; X86-NEXT:    vgetmantss $11, %xmm1, %xmm0, %xmm3
; X86-NEXT:    vgetmantss $11, {sae}, %xmm1, %xmm0, %xmm0
; X86-NEXT:    vaddps %xmm3, %xmm0, %xmm0
; X86-NEXT:    vaddps %xmm0, %xmm2, %xmm0
; X86-NEXT:    retl
  %res  = call <4 x float> @llvm.x86.avx512.mask.getmant.ss(<4 x float> %x0, <4 x float> %x1, i32 11, <4 x float> %x2, i8 %x3, i32 4)
  %res1 = call <4 x float> @llvm.x86.avx512.mask.getmant.ss(<4 x float> %x0, <4 x float> %x1, i32 11, <4 x float> zeroinitializer, i8 %x3, i32 4)
  %res2 = call <4 x float> @llvm.x86.avx512.mask.getmant.ss(<4 x float> %x0, <4 x float> %x1, i32 11, <4 x float> %x2, i8 -1, i32 8)
  %res3 = call <4 x float> @llvm.x86.avx512.mask.getmant.ss(<4 x float> %x0, <4 x float> %x1, i32 11, <4 x float> %x2, i8 -1, i32 4)
  %res11 = fadd <4 x float> %res, %res1
  %res12 = fadd <4 x float> %res2, %res3
  %res13 = fadd <4 x float> %res11, %res12
  ret <4 x float> %res13
}

declare <8 x double> @llvm.x86.avx512.vpermilvar.pd.512(<8 x double>, <8 x i64>)

define <8 x double>@test_int_x86_avx512_vpermilvar_pd_512(<8 x double> %x0, <8 x i64> %x1) {
; CHECK-LABEL: test_int_x86_avx512_vpermilvar_pd_512:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vpermilpd %zmm1, %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %res = call <8 x double> @llvm.x86.avx512.vpermilvar.pd.512(<8 x double> %x0, <8 x i64> %x1)
  ret <8 x double> %res
}

define <8 x double>@test_int_x86_avx512_vpermilvar_pd_512_mask(<8 x double> %x0, <8 x i64> %x1, <8 x double> %x2, i8 %mask) {
; X64-LABEL: test_int_x86_avx512_vpermilvar_pd_512_mask:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpermilpd %zmm1, %zmm0, %zmm2 {%k1}
; X64-NEXT:    vmovapd %zmm2, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_vpermilvar_pd_512_mask:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vpermilpd %zmm1, %zmm0, %zmm2 {%k1}
; X86-NEXT:    vmovapd %zmm2, %zmm0
; X86-NEXT:    retl
  %res = call <8 x double> @llvm.x86.avx512.vpermilvar.pd.512(<8 x double> %x0, <8 x i64> %x1)
  %mask.cast = bitcast i8 %mask to <8 x i1>
  %res2 = select <8 x i1> %mask.cast, <8 x double> %res, <8 x double> %x2
  ret <8 x double> %res2
}

define <8 x double>@test_int_x86_avx512_vpermilvar_pd_512_maskz(<8 x double> %x0, <8 x i64> %x1, i8 %mask) {
; X64-LABEL: test_int_x86_avx512_vpermilvar_pd_512_maskz:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpermilpd %zmm1, %zmm0, %zmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_vpermilvar_pd_512_maskz:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vpermilpd %zmm1, %zmm0, %zmm0 {%k1} {z}
; X86-NEXT:    retl
  %res = call <8 x double> @llvm.x86.avx512.vpermilvar.pd.512(<8 x double> %x0, <8 x i64> %x1)
  %mask.cast = bitcast i8 %mask to <8 x i1>
  %res2 = select <8 x i1> %mask.cast, <8 x double> %res, <8 x double> zeroinitializer
  ret <8 x double> %res2
}

declare <16 x float> @llvm.x86.avx512.vpermilvar.ps.512(<16 x float>, <16 x i32>)

define <16 x float>@test_int_x86_avx512_vpermilvar_ps_512(<16 x float> %x0, <16 x i32> %x1) {
; CHECK-LABEL: test_int_x86_avx512_vpermilvar_ps_512:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vpermilps %zmm1, %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %res = call <16 x float> @llvm.x86.avx512.vpermilvar.ps.512(<16 x float> %x0, <16 x i32> %x1)
  ret <16 x float> %res
}

define <16 x float>@test_int_x86_avx512_vpermilvar_ps_512_mask(<16 x float> %x0, <16 x i32> %x1, <16 x float> %x2, i16 %mask) {
; X64-LABEL: test_int_x86_avx512_vpermilvar_ps_512_mask:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpermilps %zmm1, %zmm0, %zmm2 {%k1}
; X64-NEXT:    vmovaps %zmm2, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_vpermilvar_ps_512_mask:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vpermilps %zmm1, %zmm0, %zmm2 {%k1}
; X86-NEXT:    vmovaps %zmm2, %zmm0
; X86-NEXT:    retl
  %res = call <16 x float> @llvm.x86.avx512.vpermilvar.ps.512(<16 x float> %x0, <16 x i32> %x1)
  %mask.cast = bitcast i16 %mask to <16 x i1>
  %res2 = select <16 x i1> %mask.cast, <16 x float> %res, <16 x float> %x2
  ret <16 x float> %res2
}

define <16 x float>@test_int_x86_avx512_vpermilvar_ps_512_maskz(<16 x float> %x0, <16 x i32> %x1, i16 %mask) {
; X64-LABEL: test_int_x86_avx512_vpermilvar_ps_512_maskz:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpermilps %zmm1, %zmm0, %zmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_vpermilvar_ps_512_maskz:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vpermilps %zmm1, %zmm0, %zmm0 {%k1} {z}
; X86-NEXT:    retl
  %res = call <16 x float> @llvm.x86.avx512.vpermilvar.ps.512(<16 x float> %x0, <16 x i32> %x1)
  %mask.cast = bitcast i16 %mask to <16 x i1>
  %res2 = select <16 x i1> %mask.cast, <16 x float> %res, <16 x float> zeroinitializer
  ret <16 x float> %res2
}

; Test case to make sure we can print shuffle decode comments for constant pool loads.
define <16 x float>@test_int_x86_avx512_vpermilvar_ps_512_constant_pool(<16 x float> %x0, <16 x i32> %x1) {
; CHECK-LABEL: test_int_x86_avx512_vpermilvar_ps_512_constant_pool:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vpermilps {{.*#+}} zmm0 = zmm0[1,0,3,2,4,5,6,7,10,11,8,9,14,15,13,12]
; CHECK-NEXT:    ret{{[l|q]}}
  %res = call <16 x float> @llvm.x86.avx512.vpermilvar.ps.512(<16 x float> %x0, <16 x i32> <i32 1, i32 0, i32 3, i32 2, i32 0, i32 1, i32 2, i32 3, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 1, i32 0>)
  ret <16 x float> %res
}

define <16 x float>@test_int_x86_avx512_vpermilvar_ps_512_constant_pool_mask(<16 x float> %x0, <16 x i32> %x1, <16 x float> %x2, i16 %mask) {
; X64-LABEL: test_int_x86_avx512_vpermilvar_ps_512_constant_pool_mask:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpermilps {{.*#+}} zmm2 {%k1} = zmm0[1,0,3,2,4,5,6,7,10,11,8,9,14,15,13,12]
; X64-NEXT:    vmovaps %zmm2, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_vpermilvar_ps_512_constant_pool_mask:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vpermilps {{.*#+}} zmm2 {%k1} = zmm0[1,0,3,2,4,5,6,7,10,11,8,9,14,15,13,12]
; X86-NEXT:    vmovaps %zmm2, %zmm0
; X86-NEXT:    retl
  %res = call <16 x float> @llvm.x86.avx512.vpermilvar.ps.512(<16 x float> %x0, <16 x i32> <i32 1, i32 0, i32 3, i32 2, i32 0, i32 1, i32 2, i32 3, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 1, i32 0>)
  %mask.cast = bitcast i16 %mask to <16 x i1>
  %res2 = select <16 x i1> %mask.cast, <16 x float> %res, <16 x float> %x2
  ret <16 x float> %res2
}

define <16 x float>@test_int_x86_avx512_vpermilvar_ps_512_constant_pool_maskz(<16 x float> %x0, <16 x i32> %x1, i16 %mask) {
; X64-LABEL: test_int_x86_avx512_vpermilvar_ps_512_constant_pool_maskz:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpermilps {{.*#+}} zmm0 {%k1} {z} = zmm0[1,0,3,2,4,5,6,7,10,11,8,9,14,15,13,12]
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_vpermilvar_ps_512_constant_pool_maskz:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vpermilps {{.*#+}} zmm0 {%k1} {z} = zmm0[1,0,3,2,4,5,6,7,10,11,8,9,14,15,13,12]
; X86-NEXT:    retl
  %res = call <16 x float> @llvm.x86.avx512.vpermilvar.ps.512(<16 x float> %x0, <16 x i32> <i32 1, i32 0, i32 3, i32 2, i32 0, i32 1, i32 2, i32 3, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 1, i32 0>)
  %mask.cast = bitcast i16 %mask to <16 x i1>
  %res2 = select <16 x i1> %mask.cast, <16 x float> %res, <16 x float> zeroinitializer
  ret <16 x float> %res2
}

declare <2 x double> @llvm.x86.avx512.mask.cvtss2sd.round(<2 x double>, <4 x float>, <2 x double>, i8, i32)

define <2 x double>@test_int_x86_avx512_mask_cvt_ss2sd_round(<2 x double> %x0,<4 x float> %x1, <2 x double> %x2, i8 %x3) {
; X64-LABEL: test_int_x86_avx512_mask_cvt_ss2sd_round:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vcvtss2sd %xmm1, %xmm0, %xmm2 {%k1}
; X64-NEXT:    vcvtss2sd {sae}, %xmm1, %xmm0, %xmm0
; X64-NEXT:    vaddpd %xmm0, %xmm2, %xmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_cvt_ss2sd_round:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vcvtss2sd %xmm1, %xmm0, %xmm2 {%k1}
; X86-NEXT:    vcvtss2sd {sae}, %xmm1, %xmm0, %xmm0
; X86-NEXT:    vaddpd %xmm0, %xmm2, %xmm0
; X86-NEXT:    retl
  %res = call <2 x double> @llvm.x86.avx512.mask.cvtss2sd.round(<2 x double> %x0, <4 x float> %x1, <2 x double> %x2, i8 %x3, i32 4)
  %res1 = call <2 x double> @llvm.x86.avx512.mask.cvtss2sd.round(<2 x double> %x0, <4 x float> %x1, <2 x double> %x2, i8 -1, i32 8)
  %res2 = fadd <2 x double> %res, %res1
  ret <2 x double> %res2
}

declare <4 x float> @llvm.x86.avx512.mask.cvtsd2ss.round(<4 x float>, <2 x double>, <4 x float>, i8, i32)

define <4 x float>@test_int_x86_avx512_mask_cvt_sd2ss_round(<4 x float> %x0,<2 x double> %x1, <4 x float> %x2, i8 %x3) {
; X64-LABEL: test_int_x86_avx512_mask_cvt_sd2ss_round:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vcvtsd2ss {rz-sae}, %xmm1, %xmm0, %xmm2 {%k1}
; X64-NEXT:    vcvtsd2ss {rn-sae}, %xmm1, %xmm0, %xmm0
; X64-NEXT:    vaddps %xmm0, %xmm2, %xmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_cvt_sd2ss_round:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vcvtsd2ss {rz-sae}, %xmm1, %xmm0, %xmm2 {%k1}
; X86-NEXT:    vcvtsd2ss {rn-sae}, %xmm1, %xmm0, %xmm0
; X86-NEXT:    vaddps %xmm0, %xmm2, %xmm0
; X86-NEXT:    retl
  %res = call <4 x float> @llvm.x86.avx512.mask.cvtsd2ss.round(<4 x float> %x0, <2 x double> %x1, <4 x float> %x2, i8 %x3, i32 3)
  %res1 = call <4 x float> @llvm.x86.avx512.mask.cvtsd2ss.round(<4 x float> %x0, <2 x double> %x1, <4 x float> %x2, i8 -1, i32 8)
  %res2 = fadd <4 x float> %res, %res1
  ret <4 x float> %res2
}

declare <16 x i32> @llvm.x86.avx512.pternlog.d.512(<16 x i32>, <16 x i32>, <16 x i32>, i32)

define <16 x i32>@test_int_x86_avx512_mask_pternlog_d_512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32> %x2, i16 %x4) {
; X64-LABEL: test_int_x86_avx512_mask_pternlog_d_512:
; X64:       # %bb.0:
; X64-NEXT:    vmovdqa64 %zmm0, %zmm3
; X64-NEXT:    vpternlogd $33, %zmm2, %zmm1, %zmm3
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpternlogd $33, %zmm2, %zmm1, %zmm0 {%k1}
; X64-NEXT:    vpaddd %zmm3, %zmm0, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_pternlog_d_512:
; X86:       # %bb.0:
; X86-NEXT:    vmovdqa64 %zmm0, %zmm3
; X86-NEXT:    vpternlogd $33, %zmm2, %zmm1, %zmm3
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vpternlogd $33, %zmm2, %zmm1, %zmm0 {%k1}
; X86-NEXT:    vpaddd %zmm3, %zmm0, %zmm0
; X86-NEXT:    retl
  %1 = call <16 x i32> @llvm.x86.avx512.pternlog.d.512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32> %x2, i32 33)
  %2 = bitcast i16 %x4 to <16 x i1>
  %3 = select <16 x i1> %2, <16 x i32> %1, <16 x i32> %x0
  %4 = call <16 x i32> @llvm.x86.avx512.pternlog.d.512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32> %x2, i32 33)
  %res2 = add <16 x i32> %3, %4
  ret <16 x i32> %res2
}

define <16 x i32>@test_int_x86_avx512_maskz_pternlog_d_512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32> %x2, i16 %x4) {
; X64-LABEL: test_int_x86_avx512_maskz_pternlog_d_512:
; X64:       # %bb.0:
; X64-NEXT:    vmovdqa64 %zmm0, %zmm3
; X64-NEXT:    vpternlogd $33, %zmm2, %zmm1, %zmm3
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpternlogd $33, %zmm2, %zmm1, %zmm0 {%k1} {z}
; X64-NEXT:    vpaddd %zmm3, %zmm0, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_maskz_pternlog_d_512:
; X86:       # %bb.0:
; X86-NEXT:    vmovdqa64 %zmm0, %zmm3
; X86-NEXT:    vpternlogd $33, %zmm2, %zmm1, %zmm3
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vpternlogd $33, %zmm2, %zmm1, %zmm0 {%k1} {z}
; X86-NEXT:    vpaddd %zmm3, %zmm0, %zmm0
; X86-NEXT:    retl
  %1 = call <16 x i32> @llvm.x86.avx512.pternlog.d.512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32> %x2, i32 33)
  %2 = bitcast i16 %x4 to <16 x i1>
  %3 = select <16 x i1> %2, <16 x i32> %1, <16 x i32> zeroinitializer
  %4 = call <16 x i32> @llvm.x86.avx512.pternlog.d.512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32> %x2, i32 33)
  %res2 = add <16 x i32> %3, %4
  ret <16 x i32> %res2
}

declare <8 x i64> @llvm.x86.avx512.pternlog.q.512(<8 x i64>, <8 x i64>, <8 x i64>, i32)

define <8 x i64>@test_int_x86_avx512_mask_pternlog_q_512(<8 x i64> %x0, <8 x i64> %x1, <8 x i64> %x2, i8 %x4) {
; X64-LABEL: test_int_x86_avx512_mask_pternlog_q_512:
; X64:       # %bb.0:
; X64-NEXT:    vmovdqa64 %zmm0, %zmm3
; X64-NEXT:    vpternlogq $33, %zmm2, %zmm1, %zmm3
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpternlogq $33, %zmm2, %zmm1, %zmm0 {%k1}
; X64-NEXT:    vpaddq %zmm3, %zmm0, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_pternlog_q_512:
; X86:       # %bb.0:
; X86-NEXT:    vmovdqa64 %zmm0, %zmm3
; X86-NEXT:    vpternlogq $33, %zmm2, %zmm1, %zmm3
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vpternlogq $33, %zmm2, %zmm1, %zmm0 {%k1}
; X86-NEXT:    vpaddq %zmm3, %zmm0, %zmm0
; X86-NEXT:    retl
  %1 = call <8 x i64> @llvm.x86.avx512.pternlog.q.512(<8 x i64> %x0, <8 x i64> %x1, <8 x i64> %x2, i32 33)
  %2 = bitcast i8 %x4 to <8 x i1>
  %3 = select <8 x i1> %2, <8 x i64> %1, <8 x i64> %x0
  %4 = call <8 x i64> @llvm.x86.avx512.pternlog.q.512(<8 x i64> %x0, <8 x i64> %x1, <8 x i64> %x2, i32 33)
  %res2 = add <8 x i64> %3, %4
  ret <8 x i64> %res2
}

define <8 x i64>@test_int_x86_avx512_maskz_pternlog_q_512(<8 x i64> %x0, <8 x i64> %x1, <8 x i64> %x2, i8 %x4) {
; X64-LABEL: test_int_x86_avx512_maskz_pternlog_q_512:
; X64:       # %bb.0:
; X64-NEXT:    vmovdqa64 %zmm0, %zmm3
; X64-NEXT:    vpternlogq $33, %zmm2, %zmm1, %zmm3
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpternlogq $33, %zmm2, %zmm1, %zmm0 {%k1} {z}
; X64-NEXT:    vpaddq %zmm3, %zmm0, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_maskz_pternlog_q_512:
; X86:       # %bb.0:
; X86-NEXT:    vmovdqa64 %zmm0, %zmm3
; X86-NEXT:    vpternlogq $33, %zmm2, %zmm1, %zmm3
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vpternlogq $33, %zmm2, %zmm1, %zmm0 {%k1} {z}
; X86-NEXT:    vpaddq %zmm3, %zmm0, %zmm0
; X86-NEXT:    retl
  %1 = call <8 x i64> @llvm.x86.avx512.pternlog.q.512(<8 x i64> %x0, <8 x i64> %x1, <8 x i64> %x2, i32 33)
  %2 = bitcast i8 %x4 to <8 x i1>
  %3 = select <8 x i1> %2, <8 x i64> %1, <8 x i64> zeroinitializer
  %4 = call <8 x i64> @llvm.x86.avx512.pternlog.q.512(<8 x i64> %x0, <8 x i64> %x1, <8 x i64> %x2, i32 33)
  %res2 = add <8 x i64> %3, %4
  ret <8 x i64> %res2
}

define i32 @test_x86_avx512_comi_sd_eq_sae(<2 x double> %a0, <2 x double> %a1) {
; CHECK-LABEL: test_x86_avx512_comi_sd_eq_sae:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vcmpeqsd {sae}, %xmm1, %xmm0, %k0
; CHECK-NEXT:    kmovw %k0, %eax
; CHECK-NEXT:    ret{{[l|q]}}
  %res = call i32 @llvm.x86.avx512.vcomi.sd(<2 x double> %a0, <2 x double> %a1, i32 0, i32 8)
  ret i32 %res
}

define i32 @test_x86_avx512_ucomi_sd_eq_sae(<2 x double> %a0, <2 x double> %a1) {
; CHECK-LABEL: test_x86_avx512_ucomi_sd_eq_sae:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vcmpeq_uqsd {sae}, %xmm1, %xmm0, %k0
; CHECK-NEXT:    kmovw %k0, %eax
; CHECK-NEXT:    ret{{[l|q]}}
  %res = call i32 @llvm.x86.avx512.vcomi.sd(<2 x double> %a0, <2 x double> %a1, i32 8, i32 8)
  ret i32 %res
}

define i32 @test_x86_avx512_comi_sd_eq(<2 x double> %a0, <2 x double> %a1) {
; CHECK-LABEL: test_x86_avx512_comi_sd_eq:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vcmpeqsd %xmm1, %xmm0, %k0
; CHECK-NEXT:    kmovw %k0, %eax
; CHECK-NEXT:    ret{{[l|q]}}
  %res = call i32 @llvm.x86.avx512.vcomi.sd(<2 x double> %a0, <2 x double> %a1, i32 0, i32 4)
  ret i32 %res
}

define i32 @test_x86_avx512_ucomi_sd_eq(<2 x double> %a0, <2 x double> %a1) {
; CHECK-LABEL: test_x86_avx512_ucomi_sd_eq:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vcmpeq_uqsd %xmm1, %xmm0, %k0
; CHECK-NEXT:    kmovw %k0, %eax
; CHECK-NEXT:    ret{{[l|q]}}
  %res = call i32 @llvm.x86.avx512.vcomi.sd(<2 x double> %a0, <2 x double> %a1, i32 8, i32 4)
  ret i32 %res
}

define i32 @test_x86_avx512_comi_sd_lt_sae(<2 x double> %a0, <2 x double> %a1) {
; CHECK-LABEL: test_x86_avx512_comi_sd_lt_sae:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vcmpltsd {sae}, %xmm1, %xmm0, %k0
; CHECK-NEXT:    kmovw %k0, %eax
; CHECK-NEXT:    ret{{[l|q]}}
  %res = call i32 @llvm.x86.avx512.vcomi.sd(<2 x double> %a0, <2 x double> %a1, i32 1, i32 8)
  ret i32 %res
}

define i32 @test_x86_avx512_ucomi_sd_lt_sae(<2 x double> %a0, <2 x double> %a1) {
; CHECK-LABEL: test_x86_avx512_ucomi_sd_lt_sae:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vcmpngesd {sae}, %xmm1, %xmm0, %k0
; CHECK-NEXT:    kmovw %k0, %eax
; CHECK-NEXT:    ret{{[l|q]}}
  %res = call i32 @llvm.x86.avx512.vcomi.sd(<2 x double> %a0, <2 x double> %a1, i32 9, i32 8)
  ret i32 %res
}

define i32 @test_x86_avx512_comi_sd_lt(<2 x double> %a0, <2 x double> %a1) {
; CHECK-LABEL: test_x86_avx512_comi_sd_lt:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vcmpltsd %xmm1, %xmm0, %k0
; CHECK-NEXT:    kmovw %k0, %eax
; CHECK-NEXT:    ret{{[l|q]}}
  %res = call i32 @llvm.x86.avx512.vcomi.sd(<2 x double> %a0, <2 x double> %a1, i32 1, i32 4)
  ret i32 %res
}

define i32 @test_x86_avx512_ucomi_sd_lt(<2 x double> %a0, <2 x double> %a1) {
; CHECK-LABEL: test_x86_avx512_ucomi_sd_lt:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vcmpngesd %xmm1, %xmm0, %k0
; CHECK-NEXT:    kmovw %k0, %eax
; CHECK-NEXT:    ret{{[l|q]}}
  %res = call i32 @llvm.x86.avx512.vcomi.sd(<2 x double> %a0, <2 x double> %a1, i32 9, i32 4)
  ret i32 %res
}

declare i32 @llvm.x86.avx512.vcomi.sd(<2 x double>, <2 x double>, i32, i32)

define i32 @test_x86_avx512_ucomi_ss_lt(<4 x float> %a0, <4 x float> %a1) {
; CHECK-LABEL: test_x86_avx512_ucomi_ss_lt:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vcmpngess %xmm1, %xmm0, %k0
; CHECK-NEXT:    kmovw %k0, %eax
; CHECK-NEXT:    ret{{[l|q]}}
  %res = call i32 @llvm.x86.avx512.vcomi.ss(<4 x float> %a0, <4 x float> %a1, i32 9, i32 4)
  ret i32 %res
}

declare i32 @llvm.x86.avx512.vcomi.ss(<4 x float>, <4 x float>, i32, i32)

declare <8 x double> @llvm.x86.avx512.permvar.df.512(<8 x double>, <8 x i64>)

define <8 x double>@test_int_x86_avx512_mask_permvar_df_512(<8 x double> %x0, <8 x i64> %x1, <8 x double> %x2, i8 %x3) {
; X64-LABEL: test_int_x86_avx512_mask_permvar_df_512:
; X64:       # %bb.0:
; X64-NEXT:    vpermpd %zmm0, %zmm1, %zmm3
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpermpd %zmm0, %zmm1, %zmm2 {%k1}
; X64-NEXT:    vpermpd %zmm0, %zmm1, %zmm0 {%k1} {z}
; X64-NEXT:    vaddpd %zmm0, %zmm2, %zmm0
; X64-NEXT:    vaddpd %zmm3, %zmm0, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_permvar_df_512:
; X86:       # %bb.0:
; X86-NEXT:    vpermpd %zmm0, %zmm1, %zmm3
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vpermpd %zmm0, %zmm1, %zmm2 {%k1}
; X86-NEXT:    vpermpd %zmm0, %zmm1, %zmm0 {%k1} {z}
; X86-NEXT:    vaddpd %zmm0, %zmm2, %zmm0
; X86-NEXT:    vaddpd %zmm3, %zmm0, %zmm0
; X86-NEXT:    retl
  %1 = call <8 x double> @llvm.x86.avx512.permvar.df.512(<8 x double> %x0, <8 x i64> %x1)
  %2 = bitcast i8 %x3 to <8 x i1>
  %3 = select <8 x i1> %2, <8 x double> %1, <8 x double> %x2
  %4 = call <8 x double> @llvm.x86.avx512.permvar.df.512(<8 x double> %x0, <8 x i64> %x1)
  %5 = bitcast i8 %x3 to <8 x i1>
  %6 = select <8 x i1> %5, <8 x double> %4, <8 x double> zeroinitializer
  %7 = call <8 x double> @llvm.x86.avx512.permvar.df.512(<8 x double> %x0, <8 x i64> %x1)
  %res3 = fadd <8 x double> %3, %6
  %res4 = fadd <8 x double> %res3, %7
  ret <8 x double> %res4
}

declare <8 x i64> @llvm.x86.avx512.permvar.di.512(<8 x i64>, <8 x i64>)

define <8 x i64>@test_int_x86_avx512_mask_permvar_di_512(<8 x i64> %x0, <8 x i64> %x1, <8 x i64> %x2, i8 %x3) {
; X64-LABEL: test_int_x86_avx512_mask_permvar_di_512:
; X64:       # %bb.0:
; X64-NEXT:    vpermq %zmm0, %zmm1, %zmm3
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpermq %zmm0, %zmm1, %zmm2 {%k1}
; X64-NEXT:    vpermq %zmm0, %zmm1, %zmm0 {%k1} {z}
; X64-NEXT:    vpaddq %zmm3, %zmm0, %zmm0
; X64-NEXT:    vpaddq %zmm0, %zmm2, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_permvar_di_512:
; X86:       # %bb.0:
; X86-NEXT:    vpermq %zmm0, %zmm1, %zmm3
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vpermq %zmm0, %zmm1, %zmm2 {%k1}
; X86-NEXT:    vpermq %zmm0, %zmm1, %zmm0 {%k1} {z}
; X86-NEXT:    vpaddq %zmm3, %zmm0, %zmm0
; X86-NEXT:    vpaddq %zmm0, %zmm2, %zmm0
; X86-NEXT:    retl
  %1 = call <8 x i64> @llvm.x86.avx512.permvar.di.512(<8 x i64> %x0, <8 x i64> %x1)
  %2 = bitcast i8 %x3 to <8 x i1>
  %3 = select <8 x i1> %2, <8 x i64> %1, <8 x i64> %x2
  %4 = call <8 x i64> @llvm.x86.avx512.permvar.di.512(<8 x i64> %x0, <8 x i64> %x1)
  %5 = bitcast i8 %x3 to <8 x i1>
  %6 = select <8 x i1> %5, <8 x i64> %4, <8 x i64> zeroinitializer
  %7 = call <8 x i64> @llvm.x86.avx512.permvar.di.512(<8 x i64> %x0, <8 x i64> %x1)
  %res3 = add <8 x i64> %3, %6
  %res4 = add <8 x i64> %res3, %7
  ret <8 x i64> %res4
}

declare <16 x float> @llvm.x86.avx512.permvar.sf.512(<16 x float>, <16 x i32>)

define <16 x float>@test_int_x86_avx512_mask_permvar_sf_512(<16 x float> %x0, <16 x i32> %x1, <16 x float> %x2, i16 %x3) {
; X64-LABEL: test_int_x86_avx512_mask_permvar_sf_512:
; X64:       # %bb.0:
; X64-NEXT:    vpermps %zmm0, %zmm1, %zmm3
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpermps %zmm0, %zmm1, %zmm2 {%k1}
; X64-NEXT:    vpermps %zmm0, %zmm1, %zmm0 {%k1} {z}
; X64-NEXT:    vaddps %zmm0, %zmm2, %zmm0
; X64-NEXT:    vaddps %zmm3, %zmm0, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_permvar_sf_512:
; X86:       # %bb.0:
; X86-NEXT:    vpermps %zmm0, %zmm1, %zmm3
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vpermps %zmm0, %zmm1, %zmm2 {%k1}
; X86-NEXT:    vpermps %zmm0, %zmm1, %zmm0 {%k1} {z}
; X86-NEXT:    vaddps %zmm0, %zmm2, %zmm0
; X86-NEXT:    vaddps %zmm3, %zmm0, %zmm0
; X86-NEXT:    retl
  %1 = call <16 x float> @llvm.x86.avx512.permvar.sf.512(<16 x float> %x0, <16 x i32> %x1)
  %2 = bitcast i16 %x3 to <16 x i1>
  %3 = select <16 x i1> %2, <16 x float> %1, <16 x float> %x2
  %4 = call <16 x float> @llvm.x86.avx512.permvar.sf.512(<16 x float> %x0, <16 x i32> %x1)
  %5 = bitcast i16 %x3 to <16 x i1>
  %6 = select <16 x i1> %5, <16 x float> %4, <16 x float> zeroinitializer
  %7 = call <16 x float> @llvm.x86.avx512.permvar.sf.512(<16 x float> %x0, <16 x i32> %x1)
  %res3 = fadd <16 x float> %3, %6
  %res4 = fadd <16 x float> %res3, %7
  ret <16 x float> %res4
}

declare <16 x i32> @llvm.x86.avx512.permvar.si.512(<16 x i32>, <16 x i32>)

define <16 x i32>@test_int_x86_avx512_mask_permvar_si_512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32> %x2, i16 %x3) {
; X64-LABEL: test_int_x86_avx512_mask_permvar_si_512:
; X64:       # %bb.0:
; X64-NEXT:    vpermd %zmm0, %zmm1, %zmm3
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpermd %zmm0, %zmm1, %zmm2 {%k1}
; X64-NEXT:    vpermd %zmm0, %zmm1, %zmm0 {%k1} {z}
; X64-NEXT:    vpaddd %zmm3, %zmm0, %zmm0
; X64-NEXT:    vpaddd %zmm0, %zmm2, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_permvar_si_512:
; X86:       # %bb.0:
; X86-NEXT:    vpermd %zmm0, %zmm1, %zmm3
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vpermd %zmm0, %zmm1, %zmm2 {%k1}
; X86-NEXT:    vpermd %zmm0, %zmm1, %zmm0 {%k1} {z}
; X86-NEXT:    vpaddd %zmm3, %zmm0, %zmm0
; X86-NEXT:    vpaddd %zmm0, %zmm2, %zmm0
; X86-NEXT:    retl
  %1 = call <16 x i32> @llvm.x86.avx512.permvar.si.512(<16 x i32> %x0, <16 x i32> %x1)
  %2 = bitcast i16 %x3 to <16 x i1>
  %3 = select <16 x i1> %2, <16 x i32> %1, <16 x i32> %x2
  %4 = call <16 x i32> @llvm.x86.avx512.permvar.si.512(<16 x i32> %x0, <16 x i32> %x1)
  %5 = bitcast i16 %x3 to <16 x i1>
  %6 = select <16 x i1> %5, <16 x i32> %4, <16 x i32> zeroinitializer
  %7 = call <16 x i32> @llvm.x86.avx512.permvar.si.512(<16 x i32> %x0, <16 x i32> %x1)
  %res3 = add <16 x i32> %3, %6
  %res4 = add <16 x i32> %res3, %7
  ret <16 x i32> %res4
}

declare <8 x double> @llvm.x86.avx512.mask.fixupimm.pd.512(<8 x double>, <8 x double>, <8 x i64>, i32, i8, i32)

define <8 x double>@test_int_x86_avx512_mask_fixupimm_pd_512(<8 x double> %x0, <8 x double> %x1, <8 x i64> %x2, i8 %x4) {
; X64-LABEL: test_int_x86_avx512_mask_fixupimm_pd_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vmovapd %zmm0, %zmm3
; X64-NEXT:    vfixupimmpd $4, %zmm2, %zmm1, %zmm3 {%k1}
; X64-NEXT:    vxorpd %xmm4, %xmm4, %xmm4
; X64-NEXT:    vfixupimmpd $5, %zmm2, %zmm1, %zmm4 {%k1} {z}
; X64-NEXT:    vaddpd %zmm4, %zmm3, %zmm3
; X64-NEXT:    vfixupimmpd $3, {sae}, %zmm2, %zmm1, %zmm0
; X64-NEXT:    vaddpd %zmm0, %zmm3, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_fixupimm_pd_512:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vmovapd %zmm0, %zmm3
; X86-NEXT:    vfixupimmpd $4, %zmm2, %zmm1, %zmm3 {%k1}
; X86-NEXT:    vxorpd %xmm4, %xmm4, %xmm4
; X86-NEXT:    vfixupimmpd $5, %zmm2, %zmm1, %zmm4 {%k1} {z}
; X86-NEXT:    vaddpd %zmm4, %zmm3, %zmm3
; X86-NEXT:    vfixupimmpd $3, {sae}, %zmm2, %zmm1, %zmm0
; X86-NEXT:    vaddpd %zmm0, %zmm3, %zmm0
; X86-NEXT:    retl
  %res = call <8 x double> @llvm.x86.avx512.mask.fixupimm.pd.512(<8 x double> %x0, <8 x double> %x1, <8 x i64> %x2, i32 4, i8 %x4, i32 4)
  %res1 = call <8 x double> @llvm.x86.avx512.mask.fixupimm.pd.512(<8 x double> zeroinitializer, <8 x double> %x1, <8 x i64> %x2, i32 5, i8 %x4, i32 4)
  %res2 = call <8 x double> @llvm.x86.avx512.mask.fixupimm.pd.512(<8 x double> %x0, <8 x double> %x1, <8 x i64> %x2, i32 3, i8 -1, i32 8)
  %res3 = fadd <8 x double> %res, %res1
  %res4 = fadd <8 x double> %res3, %res2
  ret <8 x double> %res4
}

define <8 x double>@test_int_x86_avx512_mask_fixupimm_pd_512_load(<8 x double> %x0, <8 x double> %x1, <8 x i64>* %x2ptr) {
; X64-LABEL: test_int_x86_avx512_mask_fixupimm_pd_512_load:
; X64:       # %bb.0:
; X64-NEXT:    vfixupimmpd $3, (%rdi), %zmm1, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_fixupimm_pd_512_load:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    vfixupimmpd $3, (%eax), %zmm1, %zmm0
; X86-NEXT:    retl
  %x2 = load <8 x i64>, <8 x i64>* %x2ptr
  %res = call <8 x double> @llvm.x86.avx512.mask.fixupimm.pd.512(<8 x double> %x0, <8 x double> %x1, <8 x i64> %x2, i32 3, i8 -1, i32 4)
  ret <8 x double> %res
}

declare <8 x double> @llvm.x86.avx512.maskz.fixupimm.pd.512(<8 x double>, <8 x double>, <8 x i64>, i32, i8, i32)

define <8 x double>@test_int_x86_avx512_maskz_fixupimm_pd_512(<8 x double> %x0, <8 x double> %x1, <8 x i64> %x2, i8 %x4) {
; X64-LABEL: test_int_x86_avx512_maskz_fixupimm_pd_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vmovapd %zmm0, %zmm3
; X64-NEXT:    vfixupimmpd $3, %zmm2, %zmm1, %zmm3 {%k1} {z}
; X64-NEXT:    vxorpd %xmm4, %xmm4, %xmm4
; X64-NEXT:    vmovapd %zmm0, %zmm5
; X64-NEXT:    vfixupimmpd $5, %zmm4, %zmm1, %zmm5 {%k1} {z}
; X64-NEXT:    vaddpd %zmm5, %zmm3, %zmm3
; X64-NEXT:    vfixupimmpd $2, {sae}, %zmm2, %zmm1, %zmm0
; X64-NEXT:    vaddpd %zmm0, %zmm3, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_maskz_fixupimm_pd_512:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vmovapd %zmm0, %zmm3
; X86-NEXT:    vfixupimmpd $3, %zmm2, %zmm1, %zmm3 {%k1} {z}
; X86-NEXT:    vxorpd %xmm4, %xmm4, %xmm4
; X86-NEXT:    vmovapd %zmm0, %zmm5
; X86-NEXT:    vfixupimmpd $5, %zmm4, %zmm1, %zmm5 {%k1} {z}
; X86-NEXT:    vaddpd %zmm5, %zmm3, %zmm3
; X86-NEXT:    vfixupimmpd $2, {sae}, %zmm2, %zmm1, %zmm0
; X86-NEXT:    vaddpd %zmm0, %zmm3, %zmm0
; X86-NEXT:    retl
  %res = call <8 x double> @llvm.x86.avx512.maskz.fixupimm.pd.512(<8 x double> %x0, <8 x double> %x1, <8 x i64> %x2, i32 3, i8 %x4, i32 4)
  %res1 = call <8 x double> @llvm.x86.avx512.maskz.fixupimm.pd.512(<8 x double> %x0, <8 x double> %x1, <8 x i64> zeroinitializer, i32 5, i8 %x4, i32 4)
  %res2 = call <8 x double> @llvm.x86.avx512.maskz.fixupimm.pd.512(<8 x double> %x0, <8 x double> %x1, <8 x i64> %x2, i32 2, i8 -1, i32 8)
  %res3 = fadd <8 x double> %res, %res1
  %res4 = fadd <8 x double> %res3, %res2
  ret <8 x double> %res4
}

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

define <4 x float>@test_int_x86_avx512_mask_fixupimm_ss(<4 x float> %x0, <4 x float> %x1, <4 x i32> %x2, i8 %x4) {
; X64-LABEL: test_int_x86_avx512_mask_fixupimm_ss:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vmovaps %xmm0, %xmm3
; X64-NEXT:    vfixupimmss $5, %xmm2, %xmm1, %xmm3 {%k1}
; X64-NEXT:    vxorps %xmm4, %xmm4, %xmm4
; X64-NEXT:    vmovaps %xmm0, %xmm5
; X64-NEXT:    vfixupimmss $5, %xmm4, %xmm1, %xmm5 {%k1}
; X64-NEXT:    vaddps %xmm5, %xmm3, %xmm3
; X64-NEXT:    vfixupimmss $5, {sae}, %xmm2, %xmm1, %xmm0
; X64-NEXT:    vaddps %xmm0, %xmm3, %xmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_fixupimm_ss:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vmovaps %xmm0, %xmm3
; X86-NEXT:    vfixupimmss $5, %xmm2, %xmm1, %xmm3 {%k1}
; X86-NEXT:    vxorps %xmm4, %xmm4, %xmm4
; X86-NEXT:    vmovaps %xmm0, %xmm5
; X86-NEXT:    vfixupimmss $5, %xmm4, %xmm1, %xmm5 {%k1}
; X86-NEXT:    vaddps %xmm5, %xmm3, %xmm3
; X86-NEXT:    vfixupimmss $5, {sae}, %xmm2, %xmm1, %xmm0
; X86-NEXT:    vaddps %xmm0, %xmm3, %xmm0
; X86-NEXT:    retl
  %res = call <4 x float> @llvm.x86.avx512.mask.fixupimm.ss(<4 x float> %x0, <4 x float> %x1, <4 x i32> %x2, i32 5, i8 %x4, i32 4)
  %res1 = call <4 x float> @llvm.x86.avx512.mask.fixupimm.ss(<4 x float> %x0, <4 x float> %x1, <4 x i32> zeroinitializer, i32 5, i8 %x4, i32 4)
  %res2 = call <4 x float> @llvm.x86.avx512.mask.fixupimm.ss(<4 x float> %x0, <4 x float> %x1, <4 x i32> %x2, i32 5, i8 -1, i32 8)
  %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.ss(<4 x float>, <4 x float>, <4 x i32>, i32, i8, i32)

define <4 x float>@test_int_x86_avx512_maskz_fixupimm_ss(<4 x float> %x0, <4 x float> %x1, <4 x i32> %x2, i8 %x4) {
; X64-LABEL: test_int_x86_avx512_maskz_fixupimm_ss:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vmovaps %xmm0, %xmm3
; X64-NEXT:    vfixupimmss $5, %xmm2, %xmm1, %xmm3
; X64-NEXT:    vmovaps %xmm0, %xmm4
; X64-NEXT:    vfixupimmss $5, %xmm2, %xmm1, %xmm4 {%k1} {z}
; X64-NEXT:    vxorps %xmm2, %xmm2, %xmm2
; X64-NEXT:    vfixupimmss $5, {sae}, %xmm2, %xmm1, %xmm0 {%k1} {z}
; X64-NEXT:    vaddps %xmm0, %xmm4, %xmm0
; X64-NEXT:    vaddps %xmm3, %xmm0, %xmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_maskz_fixupimm_ss:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vmovaps %xmm0, %xmm3
; X86-NEXT:    vfixupimmss $5, %xmm2, %xmm1, %xmm3 {%k1} {z}
; X86-NEXT:    vmovaps %xmm0, %xmm4
; X86-NEXT:    vfixupimmss $5, %xmm2, %xmm1, %xmm4
; X86-NEXT:    vxorps %xmm2, %xmm2, %xmm2
; X86-NEXT:    vfixupimmss $5, {sae}, %xmm2, %xmm1, %xmm0 {%k1} {z}
; X86-NEXT:    vaddps %xmm0, %xmm3, %xmm0
; X86-NEXT:    vaddps %xmm4, %xmm0, %xmm0
; X86-NEXT:    retl
  %res = call <4 x float> @llvm.x86.avx512.maskz.fixupimm.ss(<4 x float> %x0, <4 x float> %x1, <4 x i32> %x2, i32 5, i8 %x4, i32 4)
  %res1 = call <4 x float> @llvm.x86.avx512.maskz.fixupimm.ss(<4 x float> %x0, <4 x float> %x1, <4 x i32> zeroinitializer, i32 5, i8 %x4, i32 8)
  %res2 = call <4 x float> @llvm.x86.avx512.maskz.fixupimm.ss(<4 x float> %x0, <4 x float> %x1, <4 x i32> %x2, i32 5, i8 -1, i32 4)
  %res3 = fadd <4 x float> %res, %res1
  %res4 = fadd <4 x float> %res3, %res2
  ret <4 x float> %res4
}

declare <16 x float> @llvm.x86.avx512.mask.fixupimm.ps.512(<16 x float>, <16 x float>, <16 x i32>, i32, i16, i32)

define <16 x float>@test_int_x86_avx512_mask_fixupimm_ps_512(<16 x float> %x0, <16 x float> %x1, <16 x i32> %x2, i16 %x4) {
; X64-LABEL: test_int_x86_avx512_mask_fixupimm_ps_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vmovaps %zmm0, %zmm3
; X64-NEXT:    vfixupimmps $5, %zmm2, %zmm1, %zmm3 {%k1}
; X64-NEXT:    vxorps %xmm4, %xmm4, %xmm4
; X64-NEXT:    vmovaps %zmm0, %zmm5
; X64-NEXT:    vfixupimmps $5, %zmm4, %zmm1, %zmm5 {%k1}
; X64-NEXT:    vaddps %zmm5, %zmm3, %zmm3
; X64-NEXT:    vfixupimmps $5, {sae}, %zmm2, %zmm1, %zmm0
; X64-NEXT:    vaddps %zmm0, %zmm3, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_fixupimm_ps_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vmovaps %zmm0, %zmm3
; X86-NEXT:    vfixupimmps $5, %zmm2, %zmm1, %zmm3 {%k1}
; X86-NEXT:    vxorps %xmm4, %xmm4, %xmm4
; X86-NEXT:    vmovaps %zmm0, %zmm5
; X86-NEXT:    vfixupimmps $5, %zmm4, %zmm1, %zmm5 {%k1}
; X86-NEXT:    vaddps %zmm5, %zmm3, %zmm3
; X86-NEXT:    vfixupimmps $5, {sae}, %zmm2, %zmm1, %zmm0
; X86-NEXT:    vaddps %zmm0, %zmm3, %zmm0
; X86-NEXT:    retl
  %res = call <16 x float> @llvm.x86.avx512.mask.fixupimm.ps.512(<16 x float> %x0, <16 x float> %x1, <16 x i32> %x2, i32 5, i16 %x4, i32 4)
  %res1 = call <16 x float> @llvm.x86.avx512.mask.fixupimm.ps.512(<16 x float> %x0, <16 x float> %x1, <16 x i32> zeroinitializer, i32 5, i16 %x4, i32 4)
  %res2 = call <16 x float> @llvm.x86.avx512.mask.fixupimm.ps.512(<16 x float> %x0, <16 x float> %x1, <16 x i32> %x2, i32 5, i16 -1, i32 8)
  %res3 = fadd <16 x float> %res, %res1
  %res4 = fadd <16 x float> %res3, %res2
  ret <16 x float> %res4
}

define <16 x float>@test_int_x86_avx512_mask_fixupimm_ps_512_load(<16 x float> %x0, <16 x float> %x1, <16 x i32>* %x2ptr) {
; X64-LABEL: test_int_x86_avx512_mask_fixupimm_ps_512_load:
; X64:       # %bb.0:
; X64-NEXT:    vfixupimmps $5, (%rdi), %zmm1, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_fixupimm_ps_512_load:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    vfixupimmps $5, (%eax), %zmm1, %zmm0
; X86-NEXT:    retl
  %x2 = load <16 x i32>, <16 x i32>* %x2ptr
  %res = call <16 x float> @llvm.x86.avx512.mask.fixupimm.ps.512(<16 x float> %x0, <16 x float> %x1, <16 x i32> %x2, i32 5, i16 -1, i32 4)
  ret <16 x float> %res
}

declare <16 x float> @llvm.x86.avx512.maskz.fixupimm.ps.512(<16 x float>, <16 x float>, <16 x i32>, i32, i16, i32)

define <16 x float>@test_int_x86_avx512_maskz_fixupimm_ps_512(<16 x float> %x0, <16 x float> %x1, <16 x i32> %x2, i16 %x4) {
; X64-LABEL: test_int_x86_avx512_maskz_fixupimm_ps_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vmovaps %zmm0, %zmm3
; X64-NEXT:    vfixupimmps $5, %zmm2, %zmm1, %zmm3
; X64-NEXT:    vmovaps %zmm0, %zmm4
; X64-NEXT:    vfixupimmps $5, %zmm2, %zmm1, %zmm4 {%k1} {z}
; X64-NEXT:    vxorps %xmm2, %xmm2, %xmm2
; X64-NEXT:    vfixupimmps $5, {sae}, %zmm2, %zmm1, %zmm0 {%k1} {z}
; X64-NEXT:    vaddps %zmm0, %zmm4, %zmm0
; X64-NEXT:    vaddps %zmm3, %zmm0, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_maskz_fixupimm_ps_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vmovaps %zmm0, %zmm3
; X86-NEXT:    vfixupimmps $5, %zmm2, %zmm1, %zmm3 {%k1} {z}
; X86-NEXT:    vmovaps %zmm0, %zmm4
; X86-NEXT:    vfixupimmps $5, %zmm2, %zmm1, %zmm4
; X86-NEXT:    vxorps %xmm2, %xmm2, %xmm2
; X86-NEXT:    vfixupimmps $5, {sae}, %zmm2, %zmm1, %zmm0 {%k1} {z}
; X86-NEXT:    vaddps %zmm0, %zmm3, %zmm0
; X86-NEXT:    vaddps %zmm4, %zmm0, %zmm0
; X86-NEXT:    retl
  %res = call <16 x float> @llvm.x86.avx512.maskz.fixupimm.ps.512(<16 x float> %x0, <16 x float> %x1, <16 x i32> %x2, i32 5, i16 %x4, i32 4)
  %res1 = call <16 x float> @llvm.x86.avx512.maskz.fixupimm.ps.512(<16 x float> %x0, <16 x float> %x1, <16 x i32> zeroinitializer, i32 5, i16 %x4, i32 8)
  %res2 = call <16 x float> @llvm.x86.avx512.maskz.fixupimm.ps.512(<16 x float> %x0, <16 x float> %x1, <16 x i32> %x2, i32 5, i16 -1, i32 4)
  %res3 = fadd <16 x float> %res, %res1
  %res4 = fadd <16 x float> %res3, %res2
  ret <16 x float> %res4
}

declare <2 x double> @llvm.x86.avx512.mask.fixupimm.sd(<2 x double>, <2 x double>, <2 x i64>, i32, i8, i32)

define <2 x double>@test_int_x86_avx512_mask_fixupimm_sd(<2 x double> %x0, <2 x double> %x1, <2 x i64> %x2, i8 %x4) {
; X64-LABEL: test_int_x86_avx512_mask_fixupimm_sd:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vmovapd %xmm0, %xmm3
; X64-NEXT:    vfixupimmsd $5, %xmm2, %xmm1, %xmm3
; X64-NEXT:    vmovapd %xmm0, %xmm4
; X64-NEXT:    vfixupimmsd $5, %xmm2, %xmm1, %xmm4 {%k1}
; X64-NEXT:    vxorpd %xmm2, %xmm2, %xmm2
; X64-NEXT:    vfixupimmsd $5, {sae}, %xmm2, %xmm1, %xmm0 {%k1}
; X64-NEXT:    vaddpd %xmm0, %xmm4, %xmm0
; X64-NEXT:    vaddpd %xmm3, %xmm0, %xmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_fixupimm_sd:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vmovapd %xmm0, %xmm3
; X86-NEXT:    vfixupimmsd $5, %xmm2, %xmm1, %xmm3 {%k1}
; X86-NEXT:    vmovapd %xmm0, %xmm4
; X86-NEXT:    vfixupimmsd $5, %xmm2, %xmm1, %xmm4
; X86-NEXT:    vxorpd %xmm2, %xmm2, %xmm2
; X86-NEXT:    vfixupimmsd $5, {sae}, %xmm2, %xmm1, %xmm0 {%k1}
; X86-NEXT:    vaddpd %xmm0, %xmm3, %xmm0
; X86-NEXT:    vaddpd %xmm4, %xmm0, %xmm0
; X86-NEXT:    retl
  %res = call <2 x double> @llvm.x86.avx512.mask.fixupimm.sd(<2 x double> %x0, <2 x double> %x1, <2 x i64> %x2, i32 5, i8 %x4, i32 4)
  %res1 = call <2 x double> @llvm.x86.avx512.mask.fixupimm.sd(<2 x double> %x0, <2 x double> %x1, <2 x i64> zeroinitializer, i32 5, i8 %x4, i32 8)
  %res2 = call <2 x double> @llvm.x86.avx512.mask.fixupimm.sd(<2 x double> %x0, <2 x double> %x1, <2 x i64> %x2, i32 5, i8 -1, i32 4)
  %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.sd(<2 x double>, <2 x double>, <2 x i64>, i32, i8, i32)

define <2 x double>@test_int_x86_avx512_maskz_fixupimm_sd(<2 x double> %x0, <2 x double> %x1, <2 x i64> %x2, i8 %x4) {
; X64-LABEL: test_int_x86_avx512_maskz_fixupimm_sd:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vmovapd %xmm0, %xmm3
; X64-NEXT:    vfixupimmsd $5, %xmm2, %xmm1, %xmm3 {%k1} {z}
; X64-NEXT:    vxorpd %xmm4, %xmm4, %xmm4
; X64-NEXT:    vmovapd %xmm0, %xmm5
; X64-NEXT:    vfixupimmsd $5, {sae}, %xmm4, %xmm1, %xmm5 {%k1} {z}
; X64-NEXT:    vaddpd %xmm5, %xmm3, %xmm3
; X64-NEXT:    vfixupimmsd $5, {sae}, %xmm2, %xmm1, %xmm0 {%k1} {z}
; X64-NEXT:    vaddpd %xmm0, %xmm3, %xmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_maskz_fixupimm_sd:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vmovapd %xmm0, %xmm3
; X86-NEXT:    vfixupimmsd $5, %xmm2, %xmm1, %xmm3 {%k1} {z}
; X86-NEXT:    vxorpd %xmm4, %xmm4, %xmm4
; X86-NEXT:    vmovapd %xmm0, %xmm5
; X86-NEXT:    vfixupimmsd $5, {sae}, %xmm4, %xmm1, %xmm5 {%k1} {z}
; X86-NEXT:    vaddpd %xmm5, %xmm3, %xmm3
; X86-NEXT:    vfixupimmsd $5, {sae}, %xmm2, %xmm1, %xmm0 {%k1} {z}
; X86-NEXT:    vaddpd %xmm0, %xmm3, %xmm0
; X86-NEXT:    retl
  %res = call <2 x double> @llvm.x86.avx512.maskz.fixupimm.sd(<2 x double> %x0, <2 x double> %x1, <2 x i64> %x2, i32 5, i8 %x4, i32 4)
  %res1 = call <2 x double> @llvm.x86.avx512.maskz.fixupimm.sd(<2 x double> %x0, <2 x double> %x1, <2 x i64> zeroinitializer, i32 5, i8 %x4, i32 8)
  %res2 = call <2 x double> @llvm.x86.avx512.maskz.fixupimm.sd(<2 x double> %x0, <2 x double> %x1, <2 x i64> %x2, i32 5, i8 %x4, i32 8)
  %res3 = fadd <2 x double> %res, %res1
  %res4 = fadd <2 x double> %res3, %res2
  ret <2 x double> %res4
}

declare double @llvm.fma.f64(double, double, double) #1
declare double @llvm.x86.avx512.vfmadd.f64(double, double, double, i32) #0

define <2 x double>@test_int_x86_avx512_mask_vfmadd_sd(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 %x3,i32 %x4 ){
; X64-LABEL: test_int_x86_avx512_mask_vfmadd_sd:
; X64:       # %bb.0:
; X64-NEXT:    vmovapd %xmm0, %xmm3
; X64-NEXT:    vfmadd213sd {{.*#+}} xmm3 = (xmm1 * xmm3) + xmm2
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vmovapd %xmm0, %xmm4
; X64-NEXT:    vfmadd213sd {{.*#+}} xmm4 = (xmm1 * xmm4) + xmm2
; X64-NEXT:    vaddpd %xmm4, %xmm3, %xmm3
; X64-NEXT:    vmovapd %xmm0, %xmm4
; X64-NEXT:    vfmadd213sd {rz-sae}, %xmm2, %xmm1, %xmm4
; X64-NEXT:    vfmadd213sd {rz-sae}, %xmm2, %xmm1, %xmm0 {%k1}
; X64-NEXT:    vaddpd %xmm0, %xmm4, %xmm0
; X64-NEXT:    vaddpd %xmm0, %xmm3, %xmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_vfmadd_sd:
; X86:       # %bb.0:
; X86-NEXT:    movb {{[0-9]+}}(%esp), %al
; X86-NEXT:    vmovapd %xmm0, %xmm3
; X86-NEXT:    vfmadd213sd {{.*#+}} xmm3 = (xmm1 * xmm3) + xmm2
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vmovapd %xmm0, %xmm4
; X86-NEXT:    vfmadd213sd {{.*#+}} xmm4 = (xmm1 * xmm4) + xmm2
; X86-NEXT:    vaddpd %xmm4, %xmm3, %xmm3
; X86-NEXT:    vmovapd %xmm0, %xmm4
; X86-NEXT:    vfmadd213sd {rz-sae}, %xmm2, %xmm1, %xmm4
; X86-NEXT:    vfmadd213sd {rz-sae}, %xmm2, %xmm1, %xmm0 {%k1}
; X86-NEXT:    vaddpd %xmm0, %xmm4, %xmm0
; X86-NEXT:    vaddpd %xmm0, %xmm3, %xmm0
; X86-NEXT:    retl
  %1 = extractelement <2 x double> %x0, i64 0
  %2 = extractelement <2 x double> %x1, i64 0
  %3 = extractelement <2 x double> %x2, i64 0
  %4 = call double @llvm.fma.f64(double %1, double %2, double %3)
  %5 = insertelement <2 x double> %x0, double %4, i64 0
  %6 = extractelement <2 x double> %x0, i64 0
  %7 = extractelement <2 x double> %x1, i64 0
  %8 = extractelement <2 x double> %x2, i64 0
  %9 = call double @llvm.fma.f64(double %6, double %7, double %8)
  %10 = bitcast i8 %x3 to <8 x i1>
  %11 = extractelement <8 x i1> %10, i64 0
  %12 = select i1 %11, double %9, double %6
  %13 = insertelement <2 x double> %x0, double %12, i64 0
  %14 = extractelement <2 x double> %x0, i64 0
  %15 = extractelement <2 x double> %x1, i64 0
  %16 = extractelement <2 x double> %x2, i64 0
  %17 = call double @llvm.x86.avx512.vfmadd.f64(double %14, double %15, double %16, i32 3)
  %18 = insertelement <2 x double> %x0, double %17, i64 0
  %19 = extractelement <2 x double> %x0, i64 0
  %20 = extractelement <2 x double> %x1, i64 0
  %21 = extractelement <2 x double> %x2, i64 0
  %22 = call double @llvm.x86.avx512.vfmadd.f64(double %19, double %20, double %21, i32 3)
  %23 = bitcast i8 %x3 to <8 x i1>
  %24 = extractelement <8 x i1> %23, i64 0
  %25 = select i1 %24, double %22, double %19
  %26 = insertelement <2 x double> %x0, double %25, i64 0
  %res4 = fadd <2 x double> %5, %13
  %res5 = fadd <2 x double> %18, %26
  %res6 = fadd <2 x double> %res4, %res5
  ret <2 x double> %res6
}

define <4 x float>@test_int_x86_avx512_mask_vfmadd_ss(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 %x3,i32 %x4 ){
; X64-LABEL: test_int_x86_avx512_mask_vfmadd_ss:
; X64:       # %bb.0:
; X64-NEXT:    vmovaps %xmm0, %xmm3
; X64-NEXT:    vfmadd213ss {{.*#+}} xmm3 = (xmm1 * xmm3) + xmm2
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vmovaps %xmm0, %xmm4
; X64-NEXT:    vfmadd213ss {{.*#+}} xmm4 = (xmm1 * xmm4) + xmm2
; X64-NEXT:    vaddps %xmm4, %xmm3, %xmm3
; X64-NEXT:    vmovaps %xmm0, %xmm4
; X64-NEXT:    vfmadd213ss {rz-sae}, %xmm2, %xmm1, %xmm4
; X64-NEXT:    vfmadd213ss {rz-sae}, %xmm2, %xmm1, %xmm0 {%k1}
; X64-NEXT:    vaddps %xmm0, %xmm4, %xmm0
; X64-NEXT:    vaddps %xmm0, %xmm3, %xmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_vfmadd_ss:
; X86:       # %bb.0:
; X86-NEXT:    movb {{[0-9]+}}(%esp), %al
; X86-NEXT:    vmovaps %xmm0, %xmm3
; X86-NEXT:    vfmadd213ss {{.*#+}} xmm3 = (xmm1 * xmm3) + xmm2
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vmovaps %xmm0, %xmm4
; X86-NEXT:    vfmadd213ss {{.*#+}} xmm4 = (xmm1 * xmm4) + xmm2
; X86-NEXT:    vaddps %xmm4, %xmm3, %xmm3
; X86-NEXT:    vmovaps %xmm0, %xmm4
; X86-NEXT:    vfmadd213ss {rz-sae}, %xmm2, %xmm1, %xmm4
; X86-NEXT:    vfmadd213ss {rz-sae}, %xmm2, %xmm1, %xmm0 {%k1}
; X86-NEXT:    vaddps %xmm0, %xmm4, %xmm0
; X86-NEXT:    vaddps %xmm0, %xmm3, %xmm0
; X86-NEXT:    retl
  %1 = extractelement <4 x float> %x0, i64 0
  %2 = extractelement <4 x float> %x1, i64 0
  %3 = extractelement <4 x float> %x2, i64 0
  %4 = call float @llvm.fma.f32(float %1, float %2, float %3)
  %5 = insertelement <4 x float> %x0, float %4, i64 0
  %6 = extractelement <4 x float> %x0, i64 0
  %7 = extractelement <4 x float> %x1, i64 0
  %8 = extractelement <4 x float> %x2, i64 0
  %9 = call float @llvm.fma.f32(float %6, float %7, float %8)
  %10 = bitcast i8 %x3 to <8 x i1>
  %11 = extractelement <8 x i1> %10, i64 0
  %12 = select i1 %11, float %9, float %6
  %13 = insertelement <4 x float> %x0, float %12, i64 0
  %14 = extractelement <4 x float> %x0, i64 0
  %15 = extractelement <4 x float> %x1, i64 0
  %16 = extractelement <4 x float> %x2, i64 0
  %17 = call float @llvm.x86.avx512.vfmadd.f32(float %14, float %15, float %16, i32 3)
  %18 = insertelement <4 x float> %x0, float %17, i64 0
  %19 = extractelement <4 x float> %x0, i64 0
  %20 = extractelement <4 x float> %x1, i64 0
  %21 = extractelement <4 x float> %x2, i64 0
  %22 = call float @llvm.x86.avx512.vfmadd.f32(float %19, float %20, float %21, i32 3)
  %23 = bitcast i8 %x3 to <8 x i1>
  %24 = extractelement <8 x i1> %23, i64 0
  %25 = select i1 %24, float %22, float %19
  %26 = insertelement <4 x float> %x0, float %25, i64 0
  %res4 = fadd <4 x float> %5, %13
  %res5 = fadd <4 x float> %18, %26
  %res6 = fadd <4 x float> %res4, %res5
  ret <4 x float> %res6
}

define <2 x double>@test_int_x86_avx512_maskz_vfmadd_sd(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 %x3,i32 %x4 ){
; X64-LABEL: test_int_x86_avx512_maskz_vfmadd_sd:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vmovapd %xmm0, %xmm3
; X64-NEXT:    vfmadd213sd {{.*#+}} xmm3 = (xmm1 * xmm3) + xmm2
; X64-NEXT:    vfmadd213sd {rz-sae}, %xmm2, %xmm1, %xmm0 {%k1} {z}
; X64-NEXT:    vaddpd %xmm0, %xmm3, %xmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_maskz_vfmadd_sd:
; X86:       # %bb.0:
; X86-NEXT:    movb {{[0-9]+}}(%esp), %al
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vmovapd %xmm0, %xmm3
; X86-NEXT:    vfmadd213sd {{.*#+}} xmm3 = (xmm1 * xmm3) + xmm2
; X86-NEXT:    vfmadd213sd {rz-sae}, %xmm2, %xmm1, %xmm0 {%k1} {z}
; X86-NEXT:    vaddpd %xmm0, %xmm3, %xmm0
; X86-NEXT:    retl
  %1 = extractelement <2 x double> %x0, i64 0
  %2 = extractelement <2 x double> %x1, i64 0
  %3 = extractelement <2 x double> %x2, i64 0
  %4 = call double @llvm.fma.f64(double %1, double %2, double %3)
  %5 = bitcast i8 %x3 to <8 x i1>
  %6 = extractelement <8 x i1> %5, i64 0
  %7 = select i1 %6, double %4, double 0.000000e+00
  %8 = insertelement <2 x double> %x0, double %7, i64 0
  %9 = extractelement <2 x double> %x0, i64 0
  %10 = extractelement <2 x double> %x1, i64 0
  %11 = extractelement <2 x double> %x2, i64 0
  %12 = call double @llvm.x86.avx512.vfmadd.f64(double %9, double %10, double %11, i32 3)
  %13 = bitcast i8 %x3 to <8 x i1>
  %14 = extractelement <8 x i1> %13, i64 0
  %15 = select i1 %14, double %12, double 0.000000e+00
  %16 = insertelement <2 x double> %x0, double %15, i64 0
  %res2 = fadd <2 x double> %8, %16
  ret <2 x double> %res2
}

declare float @llvm.fma.f32(float, float, float) #1
declare float @llvm.x86.avx512.vfmadd.f32(float, float, float, i32) #0

define <4 x float>@test_int_x86_avx512_maskz_vfmadd_ss(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 %x3,i32 %x4 ){
; X64-LABEL: test_int_x86_avx512_maskz_vfmadd_ss:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vfmadd213ss {{.*#+}} xmm0 = (xmm1 * xmm0) + xmm2
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_maskz_vfmadd_ss:
; X86:       # %bb.0:
; X86-NEXT:    movb {{[0-9]+}}(%esp), %al
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vfmadd213ss {{.*#+}} xmm0 = (xmm1 * xmm0) + xmm2
; X86-NEXT:    retl
  %1 = extractelement <4 x float> %x0, i64 0
  %2 = extractelement <4 x float> %x1, i64 0
  %3 = extractelement <4 x float> %x2, i64 0
  %4 = call float @llvm.fma.f32(float %1, float %2, float %3)
  %5 = bitcast i8 %x3 to <8 x i1>
  %6 = extractelement <8 x i1> %5, i64 0
  %7 = select i1 %6, float %4, float 0.000000e+00
  %8 = insertelement <4 x float> %x0, float %7, i64 0
  %9 = extractelement <4 x float> %x0, i64 0
  %10 = extractelement <4 x float> %x1, i64 0
  %11 = extractelement <4 x float> %x2, i64 0
  %12 = call float @llvm.x86.avx512.vfmadd.f32(float %9, float %10, float %11, i32 3)
  %13 = bitcast i8 %x3 to <8 x i1>
  %14 = extractelement <8 x i1> %13, i64 0
  %15 = select i1 %14, float %12, float 0.000000e+00
  %16 = insertelement <4 x float> %x0, float %15, i64 0
  %res2 = fadd <4 x float> %8, %16
  ret <4 x float> %8
}

define <2 x double>@test_int_x86_avx512_mask3_vfmadd_sd(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 %x3,i32 %x4 ){
; X64-LABEL: test_int_x86_avx512_mask3_vfmadd_sd:
; X64:       # %bb.0:
; X64-NEXT:    vmovapd %xmm2, %xmm3
; X64-NEXT:    vfmadd231sd {{.*#+}} xmm3 = (xmm0 * xmm1) + xmm3
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vmovapd %xmm2, %xmm4
; X64-NEXT:    vfmadd231sd {{.*#+}} xmm4 = (xmm0 * xmm1) + xmm4
; X64-NEXT:    vaddpd %xmm4, %xmm3, %xmm3
; X64-NEXT:    vmovapd %xmm2, %xmm4
; X64-NEXT:    vfmadd231sd {rz-sae}, %xmm1, %xmm0, %xmm4
; X64-NEXT:    vfmadd231sd {rz-sae}, %xmm1, %xmm0, %xmm2 {%k1}
; X64-NEXT:    vaddpd %xmm2, %xmm4, %xmm0
; X64-NEXT:    vaddpd %xmm0, %xmm3, %xmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask3_vfmadd_sd:
; X86:       # %bb.0:
; X86-NEXT:    movb {{[0-9]+}}(%esp), %al
; X86-NEXT:    vmovapd %xmm2, %xmm3
; X86-NEXT:    vfmadd231sd {{.*#+}} xmm3 = (xmm0 * xmm1) + xmm3
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vmovapd %xmm2, %xmm4
; X86-NEXT:    vfmadd231sd {{.*#+}} xmm4 = (xmm0 * xmm1) + xmm4
; X86-NEXT:    vaddpd %xmm4, %xmm3, %xmm3
; X86-NEXT:    vmovapd %xmm2, %xmm4
; X86-NEXT:    vfmadd231sd {rz-sae}, %xmm1, %xmm0, %xmm4
; X86-NEXT:    vfmadd231sd {rz-sae}, %xmm1, %xmm0, %xmm2 {%k1}
; X86-NEXT:    vaddpd %xmm2, %xmm4, %xmm0
; X86-NEXT:    vaddpd %xmm0, %xmm3, %xmm0
; X86-NEXT:    retl
  %1 = extractelement <2 x double> %x0, i64 0
  %2 = extractelement <2 x double> %x1, i64 0
  %3 = extractelement <2 x double> %x2, i64 0
  %4 = call double @llvm.fma.f64(double %1, double %2, double %3)
  %5 = insertelement <2 x double> %x2, double %4, i64 0
  %6 = extractelement <2 x double> %x0, i64 0
  %7 = extractelement <2 x double> %x1, i64 0
  %8 = extractelement <2 x double> %x2, i64 0
  %9 = call double @llvm.fma.f64(double %6, double %7, double %8)
  %10 = bitcast i8 %x3 to <8 x i1>
  %11 = extractelement <8 x i1> %10, i64 0
  %12 = select i1 %11, double %9, double %8
  %13 = insertelement <2 x double> %x2, double %12, i64 0
  %14 = extractelement <2 x double> %x0, i64 0
  %15 = extractelement <2 x double> %x1, i64 0
  %16 = extractelement <2 x double> %x2, i64 0
  %17 = call double @llvm.x86.avx512.vfmadd.f64(double %14, double %15, double %16, i32 3)
  %18 = insertelement <2 x double> %x2, double %17, i64 0
  %19 = extractelement <2 x double> %x0, i64 0
  %20 = extractelement <2 x double> %x1, i64 0
  %21 = extractelement <2 x double> %x2, i64 0
  %22 = call double @llvm.x86.avx512.vfmadd.f64(double %19, double %20, double %21, i32 3)
  %23 = bitcast i8 %x3 to <8 x i1>
  %24 = extractelement <8 x i1> %23, i64 0
  %25 = select i1 %24, double %22, double %21
  %26 = insertelement <2 x double> %x2, double %25, i64 0
  %res4 = fadd <2 x double> %5, %13
  %res5 = fadd <2 x double> %18, %26
  %res6 = fadd <2 x double> %res4, %res5
  ret <2 x double> %res6
}

define <4 x float>@test_int_x86_avx512_mask3_vfmadd_ss(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 %x3,i32 %x4 ){
; X64-LABEL: test_int_x86_avx512_mask3_vfmadd_ss:
; X64:       # %bb.0:
; X64-NEXT:    vmovaps %xmm2, %xmm3
; X64-NEXT:    vfmadd231ss {{.*#+}} xmm3 = (xmm0 * xmm1) + xmm3
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vmovaps %xmm2, %xmm4
; X64-NEXT:    vfmadd231ss {{.*#+}} xmm4 = (xmm0 * xmm1) + xmm4
; X64-NEXT:    vaddps %xmm4, %xmm3, %xmm3
; X64-NEXT:    vmovaps %xmm2, %xmm4
; X64-NEXT:    vfmadd231ss {rz-sae}, %xmm1, %xmm0, %xmm4
; X64-NEXT:    vfmadd231ss {rz-sae}, %xmm1, %xmm0, %xmm2 {%k1}
; X64-NEXT:    vaddps %xmm2, %xmm4, %xmm0
; X64-NEXT:    vaddps %xmm0, %xmm3, %xmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask3_vfmadd_ss:
; X86:       # %bb.0:
; X86-NEXT:    movb {{[0-9]+}}(%esp), %al
; X86-NEXT:    vmovaps %xmm2, %xmm3
; X86-NEXT:    vfmadd231ss {{.*#+}} xmm3 = (xmm0 * xmm1) + xmm3
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vmovaps %xmm2, %xmm4
; X86-NEXT:    vfmadd231ss {{.*#+}} xmm4 = (xmm0 * xmm1) + xmm4
; X86-NEXT:    vaddps %xmm4, %xmm3, %xmm3
; X86-NEXT:    vmovaps %xmm2, %xmm4
; X86-NEXT:    vfmadd231ss {rz-sae}, %xmm1, %xmm0, %xmm4
; X86-NEXT:    vfmadd231ss {rz-sae}, %xmm1, %xmm0, %xmm2 {%k1}
; X86-NEXT:    vaddps %xmm2, %xmm4, %xmm0
; X86-NEXT:    vaddps %xmm0, %xmm3, %xmm0
; X86-NEXT:    retl
  %1 = extractelement <4 x float> %x0, i64 0
  %2 = extractelement <4 x float> %x1, i64 0
  %3 = extractelement <4 x float> %x2, i64 0
  %4 = call float @llvm.fma.f32(float %1, float %2, float %3)
  %5 = insertelement <4 x float> %x2, float %4, i64 0
  %6 = extractelement <4 x float> %x0, i64 0
  %7 = extractelement <4 x float> %x1, i64 0
  %8 = extractelement <4 x float> %x2, i64 0
  %9 = call float @llvm.fma.f32(float %6, float %7, float %8)
  %10 = bitcast i8 %x3 to <8 x i1>
  %11 = extractelement <8 x i1> %10, i64 0
  %12 = select i1 %11, float %9, float %8
  %13 = insertelement <4 x float> %x2, float %12, i64 0
  %14 = extractelement <4 x float> %x0, i64 0
  %15 = extractelement <4 x float> %x1, i64 0
  %16 = extractelement <4 x float> %x2, i64 0
  %17 = call float @llvm.x86.avx512.vfmadd.f32(float %14, float %15, float %16, i32 3)
  %18 = insertelement <4 x float> %x2, float %17, i64 0
  %19 = extractelement <4 x float> %x0, i64 0
  %20 = extractelement <4 x float> %x1, i64 0
  %21 = extractelement <4 x float> %x2, i64 0
  %22 = call float @llvm.x86.avx512.vfmadd.f32(float %19, float %20, float %21, i32 3)
  %23 = bitcast i8 %x3 to <8 x i1>
  %24 = extractelement <8 x i1> %23, i64 0
  %25 = select i1 %24, float %22, float %21
  %26 = insertelement <4 x float> %x2, float %25, i64 0
  %res4 = fadd <4 x float> %5, %13
  %res5 = fadd <4 x float> %18, %26
  %res6 = fadd <4 x float> %res4, %res5
  ret <4 x float> %res6
}

define void @fmadd_ss_mask_memfold(float* %a, float* %b, i8 %c) {
; X64-LABEL: fmadd_ss_mask_memfold:
; X64:       # %bb.0:
; X64-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; X64-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; X64-NEXT:    vfmadd213ss {{.*#+}} xmm1 = (xmm0 * xmm1) + xmm0
; X64-NEXT:    kmovw %edx, %k1
; X64-NEXT:    vmovss %xmm1, %xmm0, %xmm0 {%k1}
; X64-NEXT:    vmovss %xmm0, (%rdi)
; X64-NEXT:    retq
;
; X86-LABEL: fmadd_ss_mask_memfold:
; X86:       # %bb.0:
; X86-NEXT:    movb {{[0-9]+}}(%esp), %al
; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; X86-NEXT:    movl {{[0-9]+}}(%esp), %edx
; X86-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; X86-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; X86-NEXT:    vfmadd213ss {{.*#+}} xmm1 = (xmm0 * xmm1) + xmm0
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vmovss %xmm1, %xmm0, %xmm0 {%k1}
; X86-NEXT:    vmovss %xmm0, (%edx)
; X86-NEXT:    retl
  %a.val = load float, float* %a
  %av0 = insertelement <4 x float> undef, float %a.val, i32 0
  %av1 = insertelement <4 x float> %av0, float 0.000000e+00, i32 1
  %av2 = insertelement <4 x float> %av1, float 0.000000e+00, i32 2
  %av  = insertelement <4 x float> %av2, float 0.000000e+00, i32 3

  %b.val = load float, float* %b
  %bv0 = insertelement <4 x float> undef, float %b.val, i32 0
  %bv1 = insertelement <4 x float> %bv0, float 0.000000e+00, i32 1
  %bv2 = insertelement <4 x float> %bv1, float 0.000000e+00, i32 2
  %bv =  insertelement <4 x float> %bv2, float 0.000000e+00, i32 3
  %1 = extractelement <4 x float> %av, i64 0
  %2 = extractelement <4 x float> %bv, i64 0
  %3 = extractelement <4 x float> %av, i64 0
  %4 = call float @llvm.fma.f32(float %1, float %2, float %3)
  %5 = bitcast i8 %c to <8 x i1>
  %6 = extractelement <8 x i1> %5, i64 0
  %7 = select i1 %6, float %4, float %1
  %8 = insertelement <4 x float> %av, float %7, i64 0
  %sr = extractelement <4 x float> %8, i32 0
  store float %sr, float* %a
  ret void
}

define void @fmadd_ss_maskz_memfold(float* %a, float* %b, i8 %c) {
; X64-LABEL: fmadd_ss_maskz_memfold:
; X64:       # %bb.0:
; X64-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; X64-NEXT:    vfmadd231ss {{.*#+}} xmm0 = (xmm0 * mem) + xmm0
; X64-NEXT:    kmovw %edx, %k1
; X64-NEXT:    vmovss %xmm0, %xmm0, %xmm0 {%k1} {z}
; X64-NEXT:    vmovss %xmm0, (%rdi)
; X64-NEXT:    retq
;
; X86-LABEL: fmadd_ss_maskz_memfold:
; X86:       # %bb.0:
; X86-NEXT:    movb {{[0-9]+}}(%esp), %al
; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; X86-NEXT:    movl {{[0-9]+}}(%esp), %edx
; X86-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; X86-NEXT:    vfmadd231ss {{.*#+}} xmm0 = (xmm0 * mem) + xmm0
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vmovss %xmm0, %xmm0, %xmm0 {%k1} {z}
; X86-NEXT:    vmovss %xmm0, (%edx)
; X86-NEXT:    retl
  %a.val = load float, float* %a
  %av0 = insertelement <4 x float> undef, float %a.val, i32 0
  %av1 = insertelement <4 x float> %av0, float 0.000000e+00, i32 1
  %av2 = insertelement <4 x float> %av1, float 0.000000e+00, i32 2
  %av  = insertelement <4 x float> %av2, float 0.000000e+00, i32 3

  %b.val = load float, float* %b
  %bv0 = insertelement <4 x float> undef, float %b.val, i32 0
  %bv1 = insertelement <4 x float> %bv0, float 0.000000e+00, i32 1
  %bv2 = insertelement <4 x float> %bv1, float 0.000000e+00, i32 2
  %bv  = insertelement <4 x float> %bv2, float 0.000000e+00, i32 3
  %1 = extractelement <4 x float> %av, i64 0
  %2 = extractelement <4 x float> %bv, i64 0
  %3 = extractelement <4 x float> %av, i64 0
  %4 = call float @llvm.fma.f32(float %1, float %2, float %3)
  %5 = bitcast i8 %c to <8 x i1>
  %6 = extractelement <8 x i1> %5, i64 0
  %7 = select i1 %6, float %4, float 0.000000e+00
  %8 = insertelement <4 x float> %av, float %7, i64 0
  %sr = extractelement <4 x float> %8, i32 0
  store float %sr, float* %a
  ret void
}

define void @fmadd_sd_mask_memfold(double* %a, double* %b, i8 %c) {
; X64-LABEL: fmadd_sd_mask_memfold:
; X64:       # %bb.0:
; X64-NEXT:    vmovsd {{.*#+}} xmm0 = mem[0],zero
; X64-NEXT:    vmovsd {{.*#+}} xmm1 = mem[0],zero
; X64-NEXT:    vfmadd213sd {{.*#+}} xmm1 = (xmm0 * xmm1) + xmm0
; X64-NEXT:    kmovw %edx, %k1
; X64-NEXT:    vmovsd %xmm1, %xmm0, %xmm0 {%k1}
; X64-NEXT:    vmovsd %xmm0, (%rdi)
; X64-NEXT:    retq
;
; X86-LABEL: fmadd_sd_mask_memfold:
; X86:       # %bb.0:
; X86-NEXT:    movb {{[0-9]+}}(%esp), %al
; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; X86-NEXT:    movl {{[0-9]+}}(%esp), %edx
; X86-NEXT:    vmovsd {{.*#+}} xmm0 = mem[0],zero
; X86-NEXT:    vmovsd {{.*#+}} xmm1 = mem[0],zero
; X86-NEXT:    vfmadd213sd {{.*#+}} xmm1 = (xmm0 * xmm1) + xmm0
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vmovsd %xmm1, %xmm0, %xmm0 {%k1}
; X86-NEXT:    vmovsd %xmm0, (%edx)
; X86-NEXT:    retl
  %a.val = load double, double* %a
  %av0 = insertelement <2 x double> undef, double %a.val, i32 0
  %av = insertelement <2 x double> %av0, double 0.000000e+00, i32 1

  %b.val = load double, double* %b
  %bv0 = insertelement <2 x double> undef, double %b.val, i32 0
  %bv = insertelement <2 x double> %bv0, double 0.000000e+00, i32 1
  %1 = extractelement <2 x double> %av, i64 0
  %2 = extractelement <2 x double> %bv, i64 0
  %3 = extractelement <2 x double> %av, i64 0
  %4 = call double @llvm.fma.f64(double %1, double %2, double %3)
  %5 = bitcast i8 %c to <8 x i1>
  %6 = extractelement <8 x i1> %5, i64 0
  %7 = select i1 %6, double %4, double %1
  %8 = insertelement <2 x double> %av, double %7, i64 0
  %sr = extractelement <2 x double> %8, i32 0
  store double %sr, double* %a
  ret void
}

define void @fmadd_sd_maskz_memfold(double* %a, double* %b, i8 %c) {
; X64-LABEL: fmadd_sd_maskz_memfold:
; X64:       # %bb.0:
; X64-NEXT:    vmovsd {{.*#+}} xmm0 = mem[0],zero
; X64-NEXT:    vfmadd231sd {{.*#+}} xmm0 = (xmm0 * mem) + xmm0
; X64-NEXT:    kmovw %edx, %k1
; X64-NEXT:    vmovsd %xmm0, %xmm0, %xmm0 {%k1} {z}
; X64-NEXT:    vmovsd %xmm0, (%rdi)
; X64-NEXT:    retq
;
; X86-LABEL: fmadd_sd_maskz_memfold:
; X86:       # %bb.0:
; X86-NEXT:    movb {{[0-9]+}}(%esp), %al
; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
; X86-NEXT:    movl {{[0-9]+}}(%esp), %edx
; X86-NEXT:    vmovsd {{.*#+}} xmm0 = mem[0],zero
; X86-NEXT:    vfmadd231sd {{.*#+}} xmm0 = (xmm0 * mem) + xmm0
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vmovsd %xmm0, %xmm0, %xmm0 {%k1} {z}
; X86-NEXT:    vmovsd %xmm0, (%edx)
; X86-NEXT:    retl
  %a.val = load double, double* %a
  %av0 = insertelement <2 x double> undef, double %a.val, i32 0
  %av = insertelement <2 x double> %av0, double 0.000000e+00, i32 1

  %b.val = load double, double* %b
  %bv0 = insertelement <2 x double> undef, double %b.val, i32 0
  %bv = insertelement <2 x double> %bv0, double 0.000000e+00, i32 1
  %1 = extractelement <2 x double> %av, i64 0
  %2 = extractelement <2 x double> %bv, i64 0
  %3 = extractelement <2 x double> %av, i64 0
  %4 = call double @llvm.fma.f64(double %1, double %2, double %3)
  %5 = bitcast i8 %c to <8 x i1>
  %6 = extractelement <8 x i1> %5, i64 0
  %7 = select i1 %6, double %4, double 0.000000e+00
  %8 = insertelement <2 x double> %av, double %7, i64 0
  %sr = extractelement <2 x double> %8, i32 0
  store double %sr, double* %a
  ret void
}

define <2 x double>@test_int_x86_avx512_mask3_vfmsub_sd(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 %x3,i32 %x4 ){
; X64-LABEL: test_int_x86_avx512_mask3_vfmsub_sd:
; X64:       # %bb.0:
; X64-NEXT:    vmovapd %xmm2, %xmm3
; X64-NEXT:    vfmsub231sd {{.*#+}} xmm3 = (xmm0 * xmm1) - xmm3
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vmovapd %xmm2, %xmm4
; X64-NEXT:    vfmsub231sd {{.*#+}} xmm4 = (xmm0 * xmm1) - xmm4
; X64-NEXT:    vaddpd %xmm4, %xmm3, %xmm3
; X64-NEXT:    vmovapd %xmm2, %xmm4
; X64-NEXT:    vfmsub231sd {rz-sae}, %xmm1, %xmm0, %xmm4
; X64-NEXT:    vfmsub231sd {rz-sae}, %xmm1, %xmm0, %xmm2 {%k1}
; X64-NEXT:    vaddpd %xmm2, %xmm4, %xmm0
; X64-NEXT:    vaddpd %xmm0, %xmm3, %xmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask3_vfmsub_sd:
; X86:       # %bb.0:
; X86-NEXT:    movb {{[0-9]+}}(%esp), %al
; X86-NEXT:    vmovapd %xmm2, %xmm3
; X86-NEXT:    vfmsub231sd {{.*#+}} xmm3 = (xmm0 * xmm1) - xmm3
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vmovapd %xmm2, %xmm4
; X86-NEXT:    vfmsub231sd {{.*#+}} xmm4 = (xmm0 * xmm1) - xmm4
; X86-NEXT:    vaddpd %xmm4, %xmm3, %xmm3
; X86-NEXT:    vmovapd %xmm2, %xmm4
; X86-NEXT:    vfmsub231sd {rz-sae}, %xmm1, %xmm0, %xmm4
; X86-NEXT:    vfmsub231sd {rz-sae}, %xmm1, %xmm0, %xmm2 {%k1}
; X86-NEXT:    vaddpd %xmm2, %xmm4, %xmm0
; X86-NEXT:    vaddpd %xmm0, %xmm3, %xmm0
; X86-NEXT:    retl
  %1 = fsub <2 x double> <double -0.000000e+00, double -0.000000e+00>, %x2
  %2 = extractelement <2 x double> %x0, i64 0
  %3 = extractelement <2 x double> %x1, i64 0
  %4 = extractelement <2 x double> %1, i64 0
  %5 = call double @llvm.fma.f64(double %2, double %3, double %4)
  %6 = extractelement <2 x double> %x2, i64 0
  %7 = insertelement <2 x double> %x2, double %5, i64 0
  %8 = fsub <2 x double> <double -0.000000e+00, double -0.000000e+00>, %x2
  %9 = extractelement <2 x double> %x0, i64 0
  %10 = extractelement <2 x double> %x1, i64 0
  %11 = extractelement <2 x double> %8, i64 0
  %12 = call double @llvm.fma.f64(double %9, double %10, double %11)
  %13 = extractelement <2 x double> %x2, i64 0
  %14 = bitcast i8 %x3 to <8 x i1>
  %15 = extractelement <8 x i1> %14, i64 0
  %16 = select i1 %15, double %12, double %13
  %17 = insertelement <2 x double> %x2, double %16, i64 0
  %18 = fsub <2 x double> <double -0.000000e+00, double -0.000000e+00>, %x2
  %19 = extractelement <2 x double> %x0, i64 0
  %20 = extractelement <2 x double> %x1, i64 0
  %21 = extractelement <2 x double> %18, i64 0
  %22 = call double @llvm.x86.avx512.vfmadd.f64(double %19, double %20, double %21, i32 3)
  %23 = extractelement <2 x double> %x2, i64 0
  %24 = insertelement <2 x double> %x2, double %22, i64 0
  %25 = fsub <2 x double> <double -0.000000e+00, double -0.000000e+00>, %x2
  %26 = extractelement <2 x double> %x0, i64 0
  %27 = extractelement <2 x double> %x1, i64 0
  %28 = extractelement <2 x double> %25, i64 0
  %29 = call double @llvm.x86.avx512.vfmadd.f64(double %26, double %27, double %28, i32 3)
  %30 = extractelement <2 x double> %x2, i64 0
  %31 = bitcast i8 %x3 to <8 x i1>
  %32 = extractelement <8 x i1> %31, i64 0
  %33 = select i1 %32, double %29, double %30
  %34 = insertelement <2 x double> %x2, double %33, i64 0
  %res4 = fadd <2 x double> %7, %17
  %res5 = fadd <2 x double> %24, %34
  %res6 = fadd <2 x double> %res4, %res5
  ret <2 x double> %res6
}

define <4 x float>@test_int_x86_avx512_mask3_vfmsub_ss(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 %x3,i32 %x4 ){
; X64-LABEL: test_int_x86_avx512_mask3_vfmsub_ss:
; X64:       # %bb.0:
; X64-NEXT:    vmovaps %xmm2, %xmm3
; X64-NEXT:    vfmsub231ss {{.*#+}} xmm3 = (xmm0 * xmm1) - xmm3
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vmovaps %xmm2, %xmm4
; X64-NEXT:    vfmsub231ss {{.*#+}} xmm4 = (xmm0 * xmm1) - xmm4
; X64-NEXT:    vaddps %xmm4, %xmm3, %xmm3
; X64-NEXT:    vmovaps %xmm2, %xmm4
; X64-NEXT:    vfmsub231ss {rz-sae}, %xmm1, %xmm0, %xmm4
; X64-NEXT:    vfmsub231ss {rz-sae}, %xmm1, %xmm0, %xmm2 {%k1}
; X64-NEXT:    vaddps %xmm2, %xmm4, %xmm0
; X64-NEXT:    vaddps %xmm0, %xmm3, %xmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask3_vfmsub_ss:
; X86:       # %bb.0:
; X86-NEXT:    movb {{[0-9]+}}(%esp), %al
; X86-NEXT:    vmovaps %xmm2, %xmm3
; X86-NEXT:    vfmsub231ss {{.*#+}} xmm3 = (xmm0 * xmm1) - xmm3
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vmovaps %xmm2, %xmm4
; X86-NEXT:    vfmsub231ss {{.*#+}} xmm4 = (xmm0 * xmm1) - xmm4
; X86-NEXT:    vaddps %xmm4, %xmm3, %xmm3
; X86-NEXT:    vmovaps %xmm2, %xmm4
; X86-NEXT:    vfmsub231ss {rz-sae}, %xmm1, %xmm0, %xmm4
; X86-NEXT:    vfmsub231ss {rz-sae}, %xmm1, %xmm0, %xmm2 {%k1}
; X86-NEXT:    vaddps %xmm2, %xmm4, %xmm0
; X86-NEXT:    vaddps %xmm0, %xmm3, %xmm0
; X86-NEXT:    retl
  %1 = fsub <4 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, %x2
  %2 = extractelement <4 x float> %x0, i64 0
  %3 = extractelement <4 x float> %x1, i64 0
  %4 = extractelement <4 x float> %1, i64 0
  %5 = call float @llvm.fma.f32(float %2, float %3, float %4)
  %6 = extractelement <4 x float> %x2, i64 0
  %7 = insertelement <4 x float> %x2, float %5, i64 0
  %8 = fsub <4 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, %x2
  %9 = extractelement <4 x float> %x0, i64 0
  %10 = extractelement <4 x float> %x1, i64 0
  %11 = extractelement <4 x float> %8, i64 0
  %12 = call float @llvm.fma.f32(float %9, float %10, float %11)
  %13 = extractelement <4 x float> %x2, i64 0
  %14 = bitcast i8 %x3 to <8 x i1>
  %15 = extractelement <8 x i1> %14, i64 0
  %16 = select i1 %15, float %12, float %13
  %17 = insertelement <4 x float> %x2, float %16, i64 0
  %18 = fsub <4 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, %x2
  %19 = extractelement <4 x float> %x0, i64 0
  %20 = extractelement <4 x float> %x1, i64 0
  %21 = extractelement <4 x float> %18, i64 0
  %22 = call float @llvm.x86.avx512.vfmadd.f32(float %19, float %20, float %21, i32 3)
  %23 = extractelement <4 x float> %x2, i64 0
  %24 = insertelement <4 x float> %x2, float %22, i64 0
  %25 = fsub <4 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, %x2
  %26 = extractelement <4 x float> %x0, i64 0
  %27 = extractelement <4 x float> %x1, i64 0
  %28 = extractelement <4 x float> %25, i64 0
  %29 = call float @llvm.x86.avx512.vfmadd.f32(float %26, float %27, float %28, i32 3)
  %30 = extractelement <4 x float> %x2, i64 0
  %31 = bitcast i8 %x3 to <8 x i1>
  %32 = extractelement <8 x i1> %31, i64 0
  %33 = select i1 %32, float %29, float %30
  %34 = insertelement <4 x float> %x2, float %33, i64 0
  %res4 = fadd <4 x float> %7, %17
  %res5 = fadd <4 x float> %24, %34
  %res6 = fadd <4 x float> %res4, %res5
  ret <4 x float> %res6
}

define <2 x double>@test_int_x86_avx512_mask3_vfnmsub_sd(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, i8 %x3,i32 %x4 ){
; X64-LABEL: test_int_x86_avx512_mask3_vfnmsub_sd:
; X64:       # %bb.0:
; X64-NEXT:    vmovapd %xmm2, %xmm3
; X64-NEXT:    vfnmsub231sd {{.*#+}} xmm3 = -(xmm0 * xmm1) - xmm3
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vmovapd %xmm2, %xmm4
; X64-NEXT:    vfnmsub231sd {{.*#+}} xmm4 = -(xmm0 * xmm1) - xmm4
; X64-NEXT:    vaddpd %xmm4, %xmm3, %xmm3
; X64-NEXT:    vmovapd %xmm2, %xmm4
; X64-NEXT:    vfnmsub231sd {rz-sae}, %xmm1, %xmm0, %xmm4
; X64-NEXT:    vfnmsub231sd {rz-sae}, %xmm1, %xmm0, %xmm2 {%k1}
; X64-NEXT:    vaddpd %xmm2, %xmm4, %xmm0
; X64-NEXT:    vaddpd %xmm0, %xmm3, %xmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask3_vfnmsub_sd:
; X86:       # %bb.0:
; X86-NEXT:    movb {{[0-9]+}}(%esp), %al
; X86-NEXT:    vmovapd %xmm2, %xmm3
; X86-NEXT:    vfnmsub231sd {{.*#+}} xmm3 = -(xmm0 * xmm1) - xmm3
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vmovapd %xmm2, %xmm4
; X86-NEXT:    vfnmsub231sd {{.*#+}} xmm4 = -(xmm0 * xmm1) - xmm4
; X86-NEXT:    vaddpd %xmm4, %xmm3, %xmm3
; X86-NEXT:    vmovapd %xmm2, %xmm4
; X86-NEXT:    vfnmsub231sd {rz-sae}, %xmm1, %xmm0, %xmm4
; X86-NEXT:    vfnmsub231sd {rz-sae}, %xmm1, %xmm0, %xmm2 {%k1}
; X86-NEXT:    vaddpd %xmm2, %xmm4, %xmm0
; X86-NEXT:    vaddpd %xmm0, %xmm3, %xmm0
; X86-NEXT:    retl
  %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 = extractelement <2 x double> %1, i64 0
  %4 = extractelement <2 x double> %x1, i64 0
  %5 = extractelement <2 x double> %2, i64 0
  %6 = call double @llvm.fma.f64(double %3, double %4, double %5)
  %7 = extractelement <2 x double> %x2, i64 0
  %8 = insertelement <2 x double> %x2, double %6, i64 0
  %9 = fsub <2 x double> <double -0.000000e+00, double -0.000000e+00>, %x0
  %10 = fsub <2 x double> <double -0.000000e+00, double -0.000000e+00>, %x2
  %11 = extractelement <2 x double> %9, i64 0
  %12 = extractelement <2 x double> %x1, i64 0
  %13 = extractelement <2 x double> %10, i64 0
  %14 = call double @llvm.fma.f64(double %11, double %12, double %13)
  %15 = extractelement <2 x double> %x2, i64 0
  %16 = bitcast i8 %x3 to <8 x i1>
  %17 = extractelement <8 x i1> %16, i64 0
  %18 = select i1 %17, double %14, double %15
  %19 = insertelement <2 x double> %x2, double %18, i64 0
  %20 = fsub <2 x double> <double -0.000000e+00, double -0.000000e+00>, %x0
  %21 = fsub <2 x double> <double -0.000000e+00, double -0.000000e+00>, %x2
  %22 = extractelement <2 x double> %20, i64 0
  %23 = extractelement <2 x double> %x1, i64 0
  %24 = extractelement <2 x double> %21, i64 0
  %25 = call double @llvm.x86.avx512.vfmadd.f64(double %22, double %23, double %24, i32 3)
  %26 = extractelement <2 x double> %x2, i64 0
  %27 = insertelement <2 x double> %x2, double %25, i64 0
  %28 = fsub <2 x double> <double -0.000000e+00, double -0.000000e+00>, %x0
  %29 = fsub <2 x double> <double -0.000000e+00, double -0.000000e+00>, %x2
  %30 = extractelement <2 x double> %28, i64 0
  %31 = extractelement <2 x double> %x1, i64 0
  %32 = extractelement <2 x double> %29, i64 0
  %33 = call double @llvm.x86.avx512.vfmadd.f64(double %30, double %31, double %32, i32 3)
  %34 = extractelement <2 x double> %x2, i64 0
  %35 = bitcast i8 %x3 to <8 x i1>
  %36 = extractelement <8 x i1> %35, i64 0
  %37 = select i1 %36, double %33, double %34
  %38 = insertelement <2 x double> %x2, double %37, i64 0
  %res4 = fadd <2 x double> %8, %19
  %res5 = fadd <2 x double> %27, %38
  %res6 = fadd <2 x double> %res4, %res5
  ret <2 x double> %res6
}

define <4 x float>@test_int_x86_avx512_mask3_vfnmsub_ss(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, i8 %x3,i32 %x4 ){
; X64-LABEL: test_int_x86_avx512_mask3_vfnmsub_ss:
; X64:       # %bb.0:
; X64-NEXT:    vmovaps %xmm2, %xmm3
; X64-NEXT:    vfnmsub231ss {{.*#+}} xmm3 = -(xmm0 * xmm1) - xmm3
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vmovaps %xmm2, %xmm4
; X64-NEXT:    vfnmsub231ss {{.*#+}} xmm4 = -(xmm0 * xmm1) - xmm4
; X64-NEXT:    vaddps %xmm4, %xmm3, %xmm3
; X64-NEXT:    vmovaps %xmm2, %xmm4
; X64-NEXT:    vfnmsub231ss {rz-sae}, %xmm1, %xmm0, %xmm4
; X64-NEXT:    vfnmsub231ss {rz-sae}, %xmm1, %xmm0, %xmm2 {%k1}
; X64-NEXT:    vaddps %xmm2, %xmm4, %xmm0
; X64-NEXT:    vaddps %xmm0, %xmm3, %xmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask3_vfnmsub_ss:
; X86:       # %bb.0:
; X86-NEXT:    movb {{[0-9]+}}(%esp), %al
; X86-NEXT:    vmovaps %xmm2, %xmm3
; X86-NEXT:    vfnmsub231ss {{.*#+}} xmm3 = -(xmm0 * xmm1) - xmm3
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vmovaps %xmm2, %xmm4
; X86-NEXT:    vfnmsub231ss {{.*#+}} xmm4 = -(xmm0 * xmm1) - xmm4
; X86-NEXT:    vaddps %xmm4, %xmm3, %xmm3
; X86-NEXT:    vmovaps %xmm2, %xmm4
; X86-NEXT:    vfnmsub231ss {rz-sae}, %xmm1, %xmm0, %xmm4
; X86-NEXT:    vfnmsub231ss {rz-sae}, %xmm1, %xmm0, %xmm2 {%k1}
; X86-NEXT:    vaddps %xmm2, %xmm4, %xmm0
; X86-NEXT:    vaddps %xmm0, %xmm3, %xmm0
; X86-NEXT:    retl
  %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 = extractelement <4 x float> %1, i64 0
  %4 = extractelement <4 x float> %x1, i64 0
  %5 = extractelement <4 x float> %2, i64 0
  %6 = call float @llvm.fma.f32(float %3, float %4, float %5)
  %7 = extractelement <4 x float> %x2, i64 0
  %8 = insertelement <4 x float> %x2, float %6, i64 0
  %9 = fsub <4 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, %x0
  %10 = fsub <4 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, %x2
  %11 = extractelement <4 x float> %9, i64 0
  %12 = extractelement <4 x float> %x1, i64 0
  %13 = extractelement <4 x float> %10, i64 0
  %14 = call float @llvm.fma.f32(float %11, float %12, float %13)
  %15 = extractelement <4 x float> %x2, i64 0
  %16 = bitcast i8 %x3 to <8 x i1>
  %17 = extractelement <8 x i1> %16, i64 0
  %18 = select i1 %17, float %14, float %15
  %19 = insertelement <4 x float> %x2, float %18, i64 0
  %20 = fsub <4 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, %x0
  %21 = fsub <4 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, %x2
  %22 = extractelement <4 x float> %20, i64 0
  %23 = extractelement <4 x float> %x1, i64 0
  %24 = extractelement <4 x float> %21, i64 0
  %25 = call float @llvm.x86.avx512.vfmadd.f32(float %22, float %23, float %24, i32 3)
  %26 = extractelement <4 x float> %x2, i64 0
  %27 = insertelement <4 x float> %x2, float %25, i64 0
  %28 = fsub <4 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, %x0
  %29 = fsub <4 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, %x2
  %30 = extractelement <4 x float> %28, i64 0
  %31 = extractelement <4 x float> %x1, i64 0
  %32 = extractelement <4 x float> %29, i64 0
  %33 = call float @llvm.x86.avx512.vfmadd.f32(float %30, float %31, float %32, i32 3)
  %34 = extractelement <4 x float> %x2, i64 0
  %35 = bitcast i8 %x3 to <8 x i1>
  %36 = extractelement <8 x i1> %35, i64 0
  %37 = select i1 %36, float %33, float %34
  %38 = insertelement <4 x float> %x2, float %37, i64 0
  %res4 = fadd <4 x float> %8, %19
  %res5 = fadd <4 x float> %27, %38
  %res6 = fadd <4 x float> %res4, %res5
  ret <4 x float> %res6
}

define <4 x float>@test_int_x86_avx512_mask3_vfmadd_ss_rm(<4 x float> %x0, <4 x float> %x1, float *%ptr_b ,i8 %x3,i32 %x4) {
; X64-LABEL: test_int_x86_avx512_mask3_vfmadd_ss_rm:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1
; X64-NEXT:    vfmadd231ss {{.*#+}} xmm1 = (xmm0 * mem) + xmm1
; X64-NEXT:    vmovaps %xmm1, %xmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask3_vfmadd_ss_rm:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    movb {{[0-9]+}}(%esp), %cl
; X86-NEXT:    kmovw %ecx, %k1
; X86-NEXT:    vfmadd231ss {{.*#+}} xmm1 = (xmm0 * mem) + xmm1
; X86-NEXT:    vmovaps %xmm1, %xmm0
; X86-NEXT:    retl
  %q = load float, float* %ptr_b
  %vecinit.i = insertelement <4 x float> undef, float %q, i32 0
  %1 = extractelement <4 x float> %x0, i64 0
  %2 = extractelement <4 x float> %vecinit.i, i64 0
  %3 = extractelement <4 x float> %x1, i64 0
  %4 = call float @llvm.fma.f32(float %1, float %2, float %3)
  %5 = bitcast i8 %x3 to <8 x i1>
  %6 = extractelement <8 x i1> %5, i64 0
  %7 = select i1 %6, float %4, float %3
  %8 = insertelement <4 x float> %x1, float %7, i64 0
  ret <4 x float> %8
}

define <4 x float>@test_int_x86_avx512_mask_vfmadd_ss_rm(<4 x float> %x0, <4 x float> %x1,float *%ptr_b ,i8 %x3,i32 %x4) {
; X64-LABEL: test_int_x86_avx512_mask_vfmadd_ss_rm:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %esi, %k1
; X64-NEXT:    vfmadd132ss {{.*#+}} xmm0 = (xmm0 * mem) + xmm1
; X64-NEXT:    retq
;
; X86-LABEL: test_int_x86_avx512_mask_vfmadd_ss_rm:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    movb {{[0-9]+}}(%esp), %cl
; X86-NEXT:    kmovw %ecx, %k1
; X86-NEXT:    vfmadd132ss {{.*#+}} xmm0 = (xmm0 * mem) + xmm1
; X86-NEXT:    retl
  %q = load float, float* %ptr_b
  %vecinit.i = insertelement <4 x float> undef, float %q, i32 0
  %1 = extractelement <4 x float> %x0, i64 0
  %2 = extractelement <4 x float> %vecinit.i, i64 0
  %3 = extractelement <4 x float> %x1, i64 0
  %4 = call float @llvm.fma.f32(float %1, float %2, float %3)
  %5 = bitcast i8 %x3 to <8 x i1>
  %6 = extractelement <8 x i1> %5, i64 0
  %7 = select i1 %6, float %4, float %1
  %8 = insertelement <4 x float> %x0, float %7, i64 0
  ret <4 x float> %8
}


define <4 x float>@test_int_x86_avx512_maskz_vfmadd_ss_rm(<4 x float> %x0, <4 x float> %x1,float *%ptr_b ,i8 %x3,i32 %x4) {
; CHECK-LABEL: test_int_x86_avx512_maskz_vfmadd_ss_rm:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vxorps %xmm1, %xmm1, %xmm1
; CHECK-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1,2,3]
; CHECK-NEXT:    ret{{[l|q]}}
  %q = load float, float* %ptr_b
  %vecinit.i = insertelement <4 x float> undef, float %q, i32 0
  %1 = extractelement <4 x float> %x0, i64 0
  %2 = extractelement <4 x float> %x1, i64 0
  %3 = extractelement <4 x float> %vecinit.i, i64 0
  %4 = call float @llvm.fma.f32(float %1, float %2, float %3)
  %5 = select i1 false, float %4, float 0.000000e+00
  %6 = insertelement <4 x float> %x0, float %5, i64 0
  ret <4 x float> %6
}

define <16 x i32> @test_x86_avx512_psll_d_512(<16 x i32> %a0, <4 x i32> %a1) {
; CHECK-LABEL: test_x86_avx512_psll_d_512:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vpslld %xmm1, %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %res = call <16 x i32> @llvm.x86.avx512.psll.d.512(<16 x i32> %a0, <4 x i32> %a1) ; <<16 x i32>> [#uses=1]
  ret <16 x i32> %res
}
define <16 x i32> @test_x86_avx512_mask_psll_d_512(<16 x i32> %a0, <4 x i32> %a1, <16 x i32> %passthru, i16 %mask) {
; X64-LABEL: test_x86_avx512_mask_psll_d_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpslld %xmm1, %zmm0, %zmm2 {%k1}
; X64-NEXT:    vmovdqa64 %zmm2, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_x86_avx512_mask_psll_d_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vpslld %xmm1, %zmm0, %zmm2 {%k1}
; X86-NEXT:    vmovdqa64 %zmm2, %zmm0
; X86-NEXT:    retl
  %res = call <16 x i32> @llvm.x86.avx512.psll.d.512(<16 x i32> %a0, <4 x i32> %a1) ; <<16 x i32>> [#uses=1]
  %mask.cast = bitcast i16 %mask to <16 x i1>
  %res2 = select <16 x i1> %mask.cast, <16 x i32> %res, <16 x i32> %passthru
  ret <16 x i32> %res2
}
define <16 x i32> @test_x86_avx512_maskz_psll_d_512(<16 x i32> %a0, <4 x i32> %a1, i16 %mask) {
; X64-LABEL: test_x86_avx512_maskz_psll_d_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpslld %xmm1, %zmm0, %zmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_x86_avx512_maskz_psll_d_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vpslld %xmm1, %zmm0, %zmm0 {%k1} {z}
; X86-NEXT:    retl
  %res = call <16 x i32> @llvm.x86.avx512.psll.d.512(<16 x i32> %a0, <4 x i32> %a1) ; <<16 x i32>> [#uses=1]
  %mask.cast = bitcast i16 %mask to <16 x i1>
  %res2 = select <16 x i1> %mask.cast, <16 x i32> %res, <16 x i32> zeroinitializer
  ret <16 x i32> %res2
}
declare <16 x i32> @llvm.x86.avx512.psll.d.512(<16 x i32>, <4 x i32>) nounwind readnone


define <8 x i64> @test_x86_avx512_psll_q_512(<8 x i64> %a0, <2 x i64> %a1) {
; CHECK-LABEL: test_x86_avx512_psll_q_512:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vpsllq %xmm1, %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %res = call <8 x i64> @llvm.x86.avx512.psll.q.512(<8 x i64> %a0, <2 x i64> %a1) ; <<8 x i64>> [#uses=1]
  ret <8 x i64> %res
}
define <8 x i64> @test_x86_avx512_mask_psll_q_512(<8 x i64> %a0, <2 x i64> %a1, <8 x i64> %passthru, i8 %mask) {
; X64-LABEL: test_x86_avx512_mask_psll_q_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpsllq %xmm1, %zmm0, %zmm2 {%k1}
; X64-NEXT:    vmovdqa64 %zmm2, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_x86_avx512_mask_psll_q_512:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vpsllq %xmm1, %zmm0, %zmm2 {%k1}
; X86-NEXT:    vmovdqa64 %zmm2, %zmm0
; X86-NEXT:    retl
  %res = call <8 x i64> @llvm.x86.avx512.psll.q.512(<8 x i64> %a0, <2 x i64> %a1) ; <<8 x i64>> [#uses=1]
  %mask.cast = bitcast i8 %mask to <8 x i1>
  %res2 = select <8 x i1> %mask.cast, <8 x i64> %res, <8 x i64> %passthru
  ret <8 x i64> %res2
}
define <8 x i64> @test_x86_avx512_maskz_psll_q_512(<8 x i64> %a0, <2 x i64> %a1, i8 %mask) {
; X64-LABEL: test_x86_avx512_maskz_psll_q_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpsllq %xmm1, %zmm0, %zmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_x86_avx512_maskz_psll_q_512:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vpsllq %xmm1, %zmm0, %zmm0 {%k1} {z}
; X86-NEXT:    retl
  %res = call <8 x i64> @llvm.x86.avx512.psll.q.512(<8 x i64> %a0, <2 x i64> %a1) ; <<8 x i64>> [#uses=1]
  %mask.cast = bitcast i8 %mask to <8 x i1>
  %res2 = select <8 x i1> %mask.cast, <8 x i64> %res, <8 x i64> zeroinitializer
  ret <8 x i64> %res2
}
declare <8 x i64> @llvm.x86.avx512.psll.q.512(<8 x i64>, <2 x i64>) nounwind readnone


define <16 x i32> @test_x86_avx512_pslli_d_512(<16 x i32> %a0) {
; CHECK-LABEL: test_x86_avx512_pslli_d_512:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vpslld $7, %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %res = call <16 x i32> @llvm.x86.avx512.pslli.d.512(<16 x i32> %a0, i32 7) ; <<16 x i32>> [#uses=1]
  ret <16 x i32> %res
}
define <16 x i32> @test_x86_avx512_mask_pslli_d_512(<16 x i32> %a0, <16 x i32> %passthru, i16 %mask) {
; X64-LABEL: test_x86_avx512_mask_pslli_d_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpslld $7, %zmm0, %zmm1 {%k1}
; X64-NEXT:    vmovdqa64 %zmm1, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_x86_avx512_mask_pslli_d_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vpslld $7, %zmm0, %zmm1 {%k1}
; X86-NEXT:    vmovdqa64 %zmm1, %zmm0
; X86-NEXT:    retl
  %res = call <16 x i32> @llvm.x86.avx512.pslli.d.512(<16 x i32> %a0, i32 7) ; <<16 x i32>> [#uses=1]
  %mask.cast = bitcast i16 %mask to <16 x i1>
  %res2 = select <16 x i1> %mask.cast, <16 x i32> %res, <16 x i32> %passthru
  ret <16 x i32> %res2
}
define <16 x i32> @test_x86_avx512_maskz_pslli_d_512(<16 x i32> %a0, i16 %mask) {
; X64-LABEL: test_x86_avx512_maskz_pslli_d_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpslld $7, %zmm0, %zmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_x86_avx512_maskz_pslli_d_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vpslld $7, %zmm0, %zmm0 {%k1} {z}
; X86-NEXT:    retl
  %res = call <16 x i32> @llvm.x86.avx512.pslli.d.512(<16 x i32> %a0, i32 7) ; <<16 x i32>> [#uses=1]
  %mask.cast = bitcast i16 %mask to <16 x i1>
  %res2 = select <16 x i1> %mask.cast, <16 x i32> %res, <16 x i32> zeroinitializer
  ret <16 x i32> %res2
}
declare <16 x i32> @llvm.x86.avx512.pslli.d.512(<16 x i32>, i32) nounwind readnone


define <8 x i64> @test_x86_avx512_pslli_q_512(<8 x i64> %a0) {
; CHECK-LABEL: test_x86_avx512_pslli_q_512:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vpsllq $7, %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %res = call <8 x i64> @llvm.x86.avx512.pslli.q.512(<8 x i64> %a0, i32 7) ; <<8 x i64>> [#uses=1]
  ret <8 x i64> %res
}
define <8 x i64> @test_x86_avx512_mask_pslli_q_512(<8 x i64> %a0, <8 x i64> %passthru, i8 %mask) {
; X64-LABEL: test_x86_avx512_mask_pslli_q_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpsllq $7, %zmm0, %zmm1 {%k1}
; X64-NEXT:    vmovdqa64 %zmm1, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_x86_avx512_mask_pslli_q_512:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vpsllq $7, %zmm0, %zmm1 {%k1}
; X86-NEXT:    vmovdqa64 %zmm1, %zmm0
; X86-NEXT:    retl
  %res = call <8 x i64> @llvm.x86.avx512.pslli.q.512(<8 x i64> %a0, i32 7) ; <<8 x i64>> [#uses=1]
  %mask.cast = bitcast i8 %mask to <8 x i1>
  %res2 = select <8 x i1> %mask.cast, <8 x i64> %res, <8 x i64> %passthru
  ret <8 x i64> %res2
}
define <8 x i64> @test_x86_avx512_maskz_pslli_q_512(<8 x i64> %a0, <8 x i64> %passthru, i8 %mask) {
; X64-LABEL: test_x86_avx512_maskz_pslli_q_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpsllq $7, %zmm0, %zmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_x86_avx512_maskz_pslli_q_512:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vpsllq $7, %zmm0, %zmm0 {%k1} {z}
; X86-NEXT:    retl
  %res = call <8 x i64> @llvm.x86.avx512.pslli.q.512(<8 x i64> %a0, i32 7) ; <<8 x i64>> [#uses=1]
  %mask.cast = bitcast i8 %mask to <8 x i1>
  %res2 = select <8 x i1> %mask.cast, <8 x i64> %res, <8 x i64> zeroinitializer
  ret <8 x i64> %res2
}
declare <8 x i64> @llvm.x86.avx512.pslli.q.512(<8 x i64>, i32) nounwind readnone


define <8 x i64> @test_x86_avx512_psra_q_512(<8 x i64> %a0, <2 x i64> %a1) {
; CHECK-LABEL: test_x86_avx512_psra_q_512:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vpsraq %xmm1, %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %res = call <8 x i64> @llvm.x86.avx512.psra.q.512(<8 x i64> %a0, <2 x i64> %a1) ; <<8 x i64>> [#uses=1]
  ret <8 x i64> %res
}
define <8 x i64> @test_x86_avx512_mask_psra_q_512(<8 x i64> %a0, <2 x i64> %a1, <8 x i64> %passthru, i8 %mask) {
; X64-LABEL: test_x86_avx512_mask_psra_q_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpsraq %xmm1, %zmm0, %zmm2 {%k1}
; X64-NEXT:    vmovdqa64 %zmm2, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_x86_avx512_mask_psra_q_512:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vpsraq %xmm1, %zmm0, %zmm2 {%k1}
; X86-NEXT:    vmovdqa64 %zmm2, %zmm0
; X86-NEXT:    retl
  %res = call <8 x i64> @llvm.x86.avx512.psra.q.512(<8 x i64> %a0, <2 x i64> %a1) ; <<8 x i64>> [#uses=1]
  %mask.cast = bitcast i8 %mask to <8 x i1>
  %res2 = select <8 x i1> %mask.cast, <8 x i64> %res, <8 x i64> %passthru
  ret <8 x i64> %res2
}
define <8 x i64> @test_x86_avx512_maskz_psra_q_512(<8 x i64> %a0, <2 x i64> %a1, i8 %mask) {
; X64-LABEL: test_x86_avx512_maskz_psra_q_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpsraq %xmm1, %zmm0, %zmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_x86_avx512_maskz_psra_q_512:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vpsraq %xmm1, %zmm0, %zmm0 {%k1} {z}
; X86-NEXT:    retl
  %res = call <8 x i64> @llvm.x86.avx512.psra.q.512(<8 x i64> %a0, <2 x i64> %a1) ; <<8 x i64>> [#uses=1]
  %mask.cast = bitcast i8 %mask to <8 x i1>
  %res2 = select <8 x i1> %mask.cast, <8 x i64> %res, <8 x i64> zeroinitializer
  ret <8 x i64> %res2
}
declare <8 x i64> @llvm.x86.avx512.psra.q.512(<8 x i64>, <2 x i64>) nounwind readnone


define <16 x i32> @test_x86_avx512_psra_d_512(<16 x i32> %a0, <4 x i32> %a1) {
; CHECK-LABEL: test_x86_avx512_psra_d_512:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vpsrad %xmm1, %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %res = call <16 x i32> @llvm.x86.avx512.psra.d.512(<16 x i32> %a0, <4 x i32> %a1) ; <<16 x i32>> [#uses=1]
  ret <16 x i32> %res
}
define <16 x i32> @test_x86_avx512_mask_psra_d_512(<16 x i32> %a0, <4 x i32> %a1, <16 x i32> %passthru, i16 %mask) {
; X64-LABEL: test_x86_avx512_mask_psra_d_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpsrad %xmm1, %zmm0, %zmm2 {%k1}
; X64-NEXT:    vmovdqa64 %zmm2, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_x86_avx512_mask_psra_d_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vpsrad %xmm1, %zmm0, %zmm2 {%k1}
; X86-NEXT:    vmovdqa64 %zmm2, %zmm0
; X86-NEXT:    retl
  %res = call <16 x i32> @llvm.x86.avx512.psra.d.512(<16 x i32> %a0, <4 x i32> %a1) ; <<16 x i32>> [#uses=1]
  %mask.cast = bitcast i16 %mask to <16 x i1>
  %res2 = select <16 x i1> %mask.cast, <16 x i32> %res, <16 x i32> %passthru
  ret <16 x i32> %res2
}
define <16 x i32> @test_x86_avx512_maskz_psra_d_512(<16 x i32> %a0, <4 x i32> %a1, i16 %mask) {
; X64-LABEL: test_x86_avx512_maskz_psra_d_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpsrad %xmm1, %zmm0, %zmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_x86_avx512_maskz_psra_d_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vpsrad %xmm1, %zmm0, %zmm0 {%k1} {z}
; X86-NEXT:    retl
  %res = call <16 x i32> @llvm.x86.avx512.psra.d.512(<16 x i32> %a0, <4 x i32> %a1) ; <<16 x i32>> [#uses=1]
  %mask.cast = bitcast i16 %mask to <16 x i1>
  %res2 = select <16 x i1> %mask.cast, <16 x i32> %res, <16 x i32> zeroinitializer
  ret <16 x i32> %res2
}
declare <16 x i32> @llvm.x86.avx512.psra.d.512(<16 x i32>, <4 x i32>) nounwind readnone


define <8 x i64> @test_x86_avx512_psrai_q_512(<8 x i64> %a0) {
; CHECK-LABEL: test_x86_avx512_psrai_q_512:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vpsraq $7, %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %res = call <8 x i64> @llvm.x86.avx512.psrai.q.512(<8 x i64> %a0, i32 7) ; <<8 x i64>> [#uses=1]
  ret <8 x i64> %res
}
define <8 x i64> @test_x86_avx512_mask_psrai_q_512(<8 x i64> %a0, <8 x i64> %passthru, i8 %mask) {
; X64-LABEL: test_x86_avx512_mask_psrai_q_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpsraq $7, %zmm0, %zmm1 {%k1}
; X64-NEXT:    vmovdqa64 %zmm1, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_x86_avx512_mask_psrai_q_512:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vpsraq $7, %zmm0, %zmm1 {%k1}
; X86-NEXT:    vmovdqa64 %zmm1, %zmm0
; X86-NEXT:    retl
  %res = call <8 x i64> @llvm.x86.avx512.psrai.q.512(<8 x i64> %a0, i32 7) ; <<8 x i64>> [#uses=1]
  %mask.cast = bitcast i8 %mask to <8 x i1>
  %res2 = select <8 x i1> %mask.cast, <8 x i64> %res, <8 x i64> %passthru
  ret <8 x i64> %res2
}
define <8 x i64> @test_x86_avx512_maskz_psrai_q_512(<8 x i64> %a0, i8 %mask) {
; X64-LABEL: test_x86_avx512_maskz_psrai_q_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpsraq $7, %zmm0, %zmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_x86_avx512_maskz_psrai_q_512:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vpsraq $7, %zmm0, %zmm0 {%k1} {z}
; X86-NEXT:    retl
  %res = call <8 x i64> @llvm.x86.avx512.psrai.q.512(<8 x i64> %a0, i32 7) ; <<8 x i64>> [#uses=1]
  %mask.cast = bitcast i8 %mask to <8 x i1>
  %res2 = select <8 x i1> %mask.cast, <8 x i64> %res, <8 x i64> zeroinitializer
  ret <8 x i64> %res2
}
declare <8 x i64> @llvm.x86.avx512.psrai.q.512(<8 x i64>, i32) nounwind readnone


define <16 x i32> @test_x86_avx512_psrai_d_512(<16 x i32> %a0) {
; CHECK-LABEL: test_x86_avx512_psrai_d_512:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vpsrad $7, %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %res = call <16 x i32> @llvm.x86.avx512.psrai.d.512(<16 x i32> %a0, i32 7) ; <<16 x i32>> [#uses=1]
  ret <16 x i32> %res
}
define <16 x i32> @test_x86_avx512_mask_psrai_d_512(<16 x i32> %a0, <16 x i32> %passthru, i16 %mask) {
; X64-LABEL: test_x86_avx512_mask_psrai_d_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpsrad $7, %zmm0, %zmm1 {%k1}
; X64-NEXT:    vmovdqa64 %zmm1, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_x86_avx512_mask_psrai_d_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vpsrad $7, %zmm0, %zmm1 {%k1}
; X86-NEXT:    vmovdqa64 %zmm1, %zmm0
; X86-NEXT:    retl
  %res = call <16 x i32> @llvm.x86.avx512.psrai.d.512(<16 x i32> %a0, i32 7) ; <<16 x i32>> [#uses=1]
  %mask.cast = bitcast i16 %mask to <16 x i1>
  %res2 = select <16 x i1> %mask.cast, <16 x i32> %res, <16 x i32> %passthru
  ret <16 x i32> %res2
}
define <16 x i32> @test_x86_avx512_maskz_psrai_d_512(<16 x i32> %a0, i16 %mask) {
; X64-LABEL: test_x86_avx512_maskz_psrai_d_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpsrad $7, %zmm0, %zmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_x86_avx512_maskz_psrai_d_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vpsrad $7, %zmm0, %zmm0 {%k1} {z}
; X86-NEXT:    retl
  %res = call <16 x i32> @llvm.x86.avx512.psrai.d.512(<16 x i32> %a0, i32 7) ; <<16 x i32>> [#uses=1]
  %mask.cast = bitcast i16 %mask to <16 x i1>
  %res2 = select <16 x i1> %mask.cast, <16 x i32> %res, <16 x i32> zeroinitializer
  ret <16 x i32> %res2
}
declare <16 x i32> @llvm.x86.avx512.psrai.d.512(<16 x i32>, i32) nounwind readnone


define <16 x i32> @test_x86_avx512_psrl_d_512(<16 x i32> %a0, <4 x i32> %a1) {
; CHECK-LABEL: test_x86_avx512_psrl_d_512:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vpsrld %xmm1, %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %res = call <16 x i32> @llvm.x86.avx512.psrl.d.512(<16 x i32> %a0, <4 x i32> %a1) ; <<16 x i32>> [#uses=1]
  ret <16 x i32> %res
}
define <16 x i32> @test_x86_avx512_mask_psrl_d_512(<16 x i32> %a0, <4 x i32> %a1, <16 x i32> %passthru, i16 %mask) {
; X64-LABEL: test_x86_avx512_mask_psrl_d_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpsrld %xmm1, %zmm0, %zmm2 {%k1}
; X64-NEXT:    vmovdqa64 %zmm2, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_x86_avx512_mask_psrl_d_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vpsrld %xmm1, %zmm0, %zmm2 {%k1}
; X86-NEXT:    vmovdqa64 %zmm2, %zmm0
; X86-NEXT:    retl
  %res = call <16 x i32> @llvm.x86.avx512.psrl.d.512(<16 x i32> %a0, <4 x i32> %a1) ; <<16 x i32>> [#uses=1]
  %mask.cast = bitcast i16 %mask to <16 x i1>
  %res2 = select <16 x i1> %mask.cast, <16 x i32> %res, <16 x i32> %passthru
  ret <16 x i32> %res2
}
define <16 x i32> @test_x86_avx512_maskz_psrl_d_512(<16 x i32> %a0, <4 x i32> %a1, i16 %mask) {
; X64-LABEL: test_x86_avx512_maskz_psrl_d_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpsrld %xmm1, %zmm0, %zmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_x86_avx512_maskz_psrl_d_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vpsrld %xmm1, %zmm0, %zmm0 {%k1} {z}
; X86-NEXT:    retl
  %res = call <16 x i32> @llvm.x86.avx512.psrl.d.512(<16 x i32> %a0, <4 x i32> %a1) ; <<16 x i32>> [#uses=1]
  %mask.cast = bitcast i16 %mask to <16 x i1>
  %res2 = select <16 x i1> %mask.cast, <16 x i32> %res, <16 x i32> zeroinitializer
  ret <16 x i32> %res2
}
declare <16 x i32> @llvm.x86.avx512.psrl.d.512(<16 x i32>, <4 x i32>) nounwind readnone


define <8 x i64> @test_x86_avx512_psrl_q_512(<8 x i64> %a0, <2 x i64> %a1) {
; CHECK-LABEL: test_x86_avx512_psrl_q_512:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vpsrlq %xmm1, %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %res = call <8 x i64> @llvm.x86.avx512.psrl.q.512(<8 x i64> %a0, <2 x i64> %a1) ; <<8 x i64>> [#uses=1]
  ret <8 x i64> %res
}
define <8 x i64> @test_x86_avx512_mask_psrl_q_512(<8 x i64> %a0, <2 x i64> %a1, <8 x i64> %passthru, i8 %mask) {
; X64-LABEL: test_x86_avx512_mask_psrl_q_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpsrlq %xmm1, %zmm0, %zmm2 {%k1}
; X64-NEXT:    vmovdqa64 %zmm2, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_x86_avx512_mask_psrl_q_512:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vpsrlq %xmm1, %zmm0, %zmm2 {%k1}
; X86-NEXT:    vmovdqa64 %zmm2, %zmm0
; X86-NEXT:    retl
  %res = call <8 x i64> @llvm.x86.avx512.psrl.q.512(<8 x i64> %a0, <2 x i64> %a1) ; <<8 x i64>> [#uses=1]
  %mask.cast = bitcast i8 %mask to <8 x i1>
  %res2 = select <8 x i1> %mask.cast, <8 x i64> %res, <8 x i64> %passthru
  ret <8 x i64> %res2
}
define <8 x i64> @test_x86_avx512_maskz_psrl_q_512(<8 x i64> %a0, <2 x i64> %a1, i8 %mask) {
; X64-LABEL: test_x86_avx512_maskz_psrl_q_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpsrlq %xmm1, %zmm0, %zmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_x86_avx512_maskz_psrl_q_512:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vpsrlq %xmm1, %zmm0, %zmm0 {%k1} {z}
; X86-NEXT:    retl
  %res = call <8 x i64> @llvm.x86.avx512.psrl.q.512(<8 x i64> %a0, <2 x i64> %a1) ; <<8 x i64>> [#uses=1]
  %mask.cast = bitcast i8 %mask to <8 x i1>
  %res2 = select <8 x i1> %mask.cast, <8 x i64> %res, <8 x i64> zeroinitializer
  ret <8 x i64> %res2
}
declare <8 x i64> @llvm.x86.avx512.psrl.q.512(<8 x i64>, <2 x i64>) nounwind readnone


define <16 x i32> @test_x86_avx512_psrli_d_512(<16 x i32> %a0) {
; CHECK-LABEL: test_x86_avx512_psrli_d_512:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vpsrld $7, %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %res = call <16 x i32> @llvm.x86.avx512.psrli.d.512(<16 x i32> %a0, i32 7) ; <<16 x i32>> [#uses=1]
  ret <16 x i32> %res
}
define <16 x i32> @test_x86_avx512_mask_psrli_d_512(<16 x i32> %a0, <16 x i32> %passthru, i16 %mask) {
; X64-LABEL: test_x86_avx512_mask_psrli_d_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpsrld $7, %zmm0, %zmm1 {%k1}
; X64-NEXT:    vmovdqa64 %zmm1, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_x86_avx512_mask_psrli_d_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vpsrld $7, %zmm0, %zmm1 {%k1}
; X86-NEXT:    vmovdqa64 %zmm1, %zmm0
; X86-NEXT:    retl
  %res = call <16 x i32> @llvm.x86.avx512.psrli.d.512(<16 x i32> %a0, i32 7) ; <<16 x i32>> [#uses=1]
  %mask.cast = bitcast i16 %mask to <16 x i1>
  %res2 = select <16 x i1> %mask.cast, <16 x i32> %res, <16 x i32> %passthru
  ret <16 x i32> %res2
}
define <16 x i32> @test_x86_avx512_maskz_psrli_d_512(<16 x i32> %a0, i16 %mask) {
; X64-LABEL: test_x86_avx512_maskz_psrli_d_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpsrld $7, %zmm0, %zmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_x86_avx512_maskz_psrli_d_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vpsrld $7, %zmm0, %zmm0 {%k1} {z}
; X86-NEXT:    retl
  %res = call <16 x i32> @llvm.x86.avx512.psrli.d.512(<16 x i32> %a0, i32 7) ; <<16 x i32>> [#uses=1]
  %mask.cast = bitcast i16 %mask to <16 x i1>
  %res2 = select <16 x i1> %mask.cast, <16 x i32> %res, <16 x i32> zeroinitializer
  ret <16 x i32> %res2
}
declare <16 x i32> @llvm.x86.avx512.psrli.d.512(<16 x i32>, i32) nounwind readnone


define <8 x i64> @test_x86_avx512_psrli_q_512(<8 x i64> %a0) {
; CHECK-LABEL: test_x86_avx512_psrli_q_512:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vpsrlq $7, %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %res = call <8 x i64> @llvm.x86.avx512.psrli.q.512(<8 x i64> %a0, i32 7) ; <<8 x i64>> [#uses=1]
  ret <8 x i64> %res
}
define <8 x i64> @test_x86_avx512_mask_psrli_q_512(<8 x i64> %a0, <8 x i64> %passthru, i8 %mask) {
; X64-LABEL: test_x86_avx512_mask_psrli_q_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpsrlq $7, %zmm0, %zmm1 {%k1}
; X64-NEXT:    vmovdqa64 %zmm1, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_x86_avx512_mask_psrli_q_512:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vpsrlq $7, %zmm0, %zmm1 {%k1}
; X86-NEXT:    vmovdqa64 %zmm1, %zmm0
; X86-NEXT:    retl
  %res = call <8 x i64> @llvm.x86.avx512.psrli.q.512(<8 x i64> %a0, i32 7) ; <<8 x i64>> [#uses=1]
  %mask.cast = bitcast i8 %mask to <8 x i1>
  %res2 = select <8 x i1> %mask.cast, <8 x i64> %res, <8 x i64> %passthru
  ret <8 x i64> %res2
}
define <8 x i64> @test_x86_avx512_maskz_psrli_q_512(<8 x i64> %a0, i8 %mask) {
; X64-LABEL: test_x86_avx512_maskz_psrli_q_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpsrlq $7, %zmm0, %zmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_x86_avx512_maskz_psrli_q_512:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vpsrlq $7, %zmm0, %zmm0 {%k1} {z}
; X86-NEXT:    retl
  %res = call <8 x i64> @llvm.x86.avx512.psrli.q.512(<8 x i64> %a0, i32 7) ; <<8 x i64>> [#uses=1]
  %mask.cast = bitcast i8 %mask to <8 x i1>
  %res2 = select <8 x i1> %mask.cast, <8 x i64> %res, <8 x i64> zeroinitializer
  ret <8 x i64> %res2
}
declare <8 x i64> @llvm.x86.avx512.psrli.q.512(<8 x i64>, i32) nounwind readnone

define <16 x i32> @test_x86_avx512_psllv_d_512(<16 x i32> %a0, <16 x i32> %a1) {
; CHECK-LABEL: test_x86_avx512_psllv_d_512:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vpsllvd %zmm1, %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %res = call <16 x i32> @llvm.x86.avx512.psllv.d.512(<16 x i32> %a0, <16 x i32> %a1)
  ret <16 x i32> %res
}

define <16 x i32> @test_x86_avx512_psllv_d_512_const() {
; X64-LABEL: test_x86_avx512_psllv_d_512_const:
; X64:       # %bb.0:
; X64-NEXT:    vmovdqa64 {{.*#+}} zmm0 = [2,9,0,4294967295,3,7,4294967295,0,4,5,4294967294,0,5,3,4294967293,0]
; X64-NEXT:    vpsllvd {{.*}}(%rip), %zmm0, %zmm0
; X64-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4294967295]
; X64-NEXT:    vpsllvd {{.*}}(%rip), %zmm1, %zmm1
; X64-NEXT:    vpaddd %zmm1, %zmm0, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_x86_avx512_psllv_d_512_const:
; X86:       # %bb.0:
; X86-NEXT:    vmovdqa64 {{.*#+}} zmm0 = [2,9,0,4294967295,3,7,4294967295,0,4,5,4294967294,0,5,3,4294967293,0]
; X86-NEXT:    vpsllvd {{\.LCPI.*}}, %zmm0, %zmm0
; X86-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4294967295]
; X86-NEXT:    vpsllvd {{\.LCPI.*}}, %zmm1, %zmm1
; X86-NEXT:    vpaddd %zmm1, %zmm0, %zmm0
; X86-NEXT:    retl
  %res0 = call <16 x i32> @llvm.x86.avx512.psllv.d.512(<16 x i32> <i32 2, i32 9, i32 0, i32 -1, i32 3, i32 7, i32 -1, i32 0, i32 4, i32 5, i32 -2, i32 0, i32 5, i32 3, i32 -3, i32 0>, <16 x i32> <i32 1, i32 0, i32 33, i32 -1,i32 2, i32 0, i32 34, i32 -2, i32 3, i32 0, i32 35, i32 -1, i32 4, i32 0, i32 36, i32 -3>)
  %res1 = call <16 x i32> @llvm.x86.avx512.psllv.d.512(<16 x i32> <i32 4, i32 4, i32 4, i32 4, i32 4, i32 4, i32 4, i32 4, i32 4, i32 4, i32 4, i32 4, i32 4, i32 4, i32 4, i32 -1>, <16 x i32> <i32 1, i32 1, i32 1,  i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1,  i32 1, i32 1, i32 1, i32 1, i32 -1>)
  %res2 = add <16 x i32> %res0, %res1
  ret <16 x i32> %res2
}

define <16 x i32> @test_x86_avx512_mask_psllv_d_512(<16 x i32> %a0, <16 x i32> %a1, <16 x i32> %a2, i16 %mask) {
; X64-LABEL: test_x86_avx512_mask_psllv_d_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpsllvd %zmm1, %zmm0, %zmm2 {%k1}
; X64-NEXT:    vmovdqa64 %zmm2, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_x86_avx512_mask_psllv_d_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vpsllvd %zmm1, %zmm0, %zmm2 {%k1}
; X86-NEXT:    vmovdqa64 %zmm2, %zmm0
; X86-NEXT:    retl
  %res = call <16 x i32> @llvm.x86.avx512.psllv.d.512(<16 x i32> %a0, <16 x i32> %a1)
  %mask.cast = bitcast i16 %mask to <16 x i1>
  %res2 = select <16 x i1> %mask.cast, <16 x i32> %res, <16 x i32> %a2
  ret <16 x i32> %res2
}

define <16 x i32> @test_x86_avx512_maskz_psllv_d_512(<16 x i32> %a0, <16 x i32> %a1, i16 %mask) {
; X64-LABEL: test_x86_avx512_maskz_psllv_d_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpsllvd %zmm1, %zmm0, %zmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_x86_avx512_maskz_psllv_d_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vpsllvd %zmm1, %zmm0, %zmm0 {%k1} {z}
; X86-NEXT:    retl
  %res = call <16 x i32> @llvm.x86.avx512.psllv.d.512(<16 x i32> %a0, <16 x i32> %a1)
  %mask.cast = bitcast i16 %mask to <16 x i1>
  %res2 = select <16 x i1> %mask.cast, <16 x i32> %res, <16 x i32> zeroinitializer
  ret <16 x i32> %res2
}

declare <16 x i32> @llvm.x86.avx512.psllv.d.512(<16 x i32>, <16 x i32>) nounwind readnone

define <8 x i64> @test_x86_avx512_psllv_q_512(<8 x i64> %a0, <8 x i64> %a1) {
; CHECK-LABEL: test_x86_avx512_psllv_q_512:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vpsllvq %zmm1, %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %res = call <8 x i64> @llvm.x86.avx512.psllv.q.512(<8 x i64> %a0, <8 x i64> %a1)
  ret <8 x i64> %res
}

define <8 x i64> @test_x86_avx512_psllv_q_512_const() {
; X64-LABEL: test_x86_avx512_psllv_q_512_const:
; X64:       # %bb.0:
; X64-NEXT:    vmovdqa64 {{.*#+}} zmm0 = [2,9,0,18446744073709551615,3,7,18446744073709551615,0]
; X64-NEXT:    vpsllvq {{.*}}(%rip), %zmm0, %zmm0
; X64-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [4,4,4,4,4,4,4,18446744073709551615]
; X64-NEXT:    vpsllvq {{.*}}(%rip), %zmm1, %zmm1
; X64-NEXT:    vpaddq %zmm1, %zmm0, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_x86_avx512_psllv_q_512_const:
; X86:       # %bb.0:
; X86-NEXT:    vmovdqa64 {{.*#+}} zmm0 = [2,0,9,0,0,0,4294967295,4294967295,3,0,7,0,4294967295,4294967295,0,0]
; X86-NEXT:    vpsllvq {{\.LCPI.*}}, %zmm0, %zmm0
; X86-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [4,0,4,0,4,0,4,0,4,0,4,0,4,0,4294967295,4294967295]
; X86-NEXT:    vpsllvq {{\.LCPI.*}}, %zmm1, %zmm1
; X86-NEXT:    vpaddq %zmm1, %zmm0, %zmm0
; X86-NEXT:    retl
  %res0 = call <8 x i64> @llvm.x86.avx512.psllv.q.512(<8 x i64> <i64 2, i64 9, i64 0, i64 -1, i64 3, i64 7, i64 -1, i64 0>, <8 x i64> <i64 1, i64 0, i64 33, i64 -1,i64 2, i64 0, i64 34, i64 -2>)
  %res1 = call <8 x i64> @llvm.x86.avx512.psllv.q.512(<8 x i64> <i64 4, i64 4, i64 4, i64 4, i64 4, i64 4, i64 4, i64 -1>, <8 x i64> <i64 1, i64 1, i64 1, i64 1,  i64 1, i64 1, i64 1, i64 -1>)
  %res2 = add <8 x i64> %res0, %res1
  ret <8 x i64> %res2
}

define <8 x i64> @test_x86_avx512_mask_psllv_q_512(<8 x i64> %a0, <8 x i64> %a1, <8 x i64> %a2, i8 %mask) {
; X64-LABEL: test_x86_avx512_mask_psllv_q_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpsllvq %zmm1, %zmm0, %zmm2 {%k1}
; X64-NEXT:    vmovdqa64 %zmm2, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_x86_avx512_mask_psllv_q_512:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vpsllvq %zmm1, %zmm0, %zmm2 {%k1}
; X86-NEXT:    vmovdqa64 %zmm2, %zmm0
; X86-NEXT:    retl
  %res = call <8 x i64> @llvm.x86.avx512.psllv.q.512(<8 x i64> %a0, <8 x i64> %a1)
  %mask.cast = bitcast i8 %mask to <8 x i1>
  %res2 = select <8 x i1> %mask.cast, <8 x i64> %res, <8 x i64> %a2
  ret <8 x i64> %res2
}

define <8 x i64> @test_x86_avx512_maskz_psllv_q_512(<8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
; X64-LABEL: test_x86_avx512_maskz_psllv_q_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpsllvq %zmm1, %zmm0, %zmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_x86_avx512_maskz_psllv_q_512:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vpsllvq %zmm1, %zmm0, %zmm0 {%k1} {z}
; X86-NEXT:    retl
  %res = call <8 x i64> @llvm.x86.avx512.psllv.q.512(<8 x i64> %a0, <8 x i64> %a1)
  %mask.cast = bitcast i8 %mask to <8 x i1>
  %res2 = select <8 x i1> %mask.cast, <8 x i64> %res, <8 x i64> zeroinitializer
  ret <8 x i64> %res2
}

declare <8 x i64> @llvm.x86.avx512.psllv.q.512(<8 x i64>, <8 x i64>) nounwind readnone

define <16 x i32> @test_x86_avx512_psrav_d_512(<16 x i32> %a0, <16 x i32> %a1) {
; CHECK-LABEL: test_x86_avx512_psrav_d_512:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vpsravd %zmm1, %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %res = call <16 x i32> @llvm.x86.avx512.psrav.d.512(<16 x i32> %a0, <16 x i32> %a1)
  ret <16 x i32> %res
}

define <16 x i32> @test_x86_avx512_mask_psrav_d_512(<16 x i32> %a0, <16 x i32> %a1, <16 x i32> %a2, i16 %mask) {
; X64-LABEL: test_x86_avx512_mask_psrav_d_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpsravd %zmm1, %zmm0, %zmm2 {%k1}
; X64-NEXT:    vmovdqa64 %zmm2, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_x86_avx512_mask_psrav_d_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vpsravd %zmm1, %zmm0, %zmm2 {%k1}
; X86-NEXT:    vmovdqa64 %zmm2, %zmm0
; X86-NEXT:    retl
  %res = call <16 x i32> @llvm.x86.avx512.psrav.d.512(<16 x i32> %a0, <16 x i32> %a1)
  %mask.cast = bitcast i16 %mask to <16 x i1>
  %res2 = select <16 x i1> %mask.cast, <16 x i32> %res, <16 x i32> %a2
  ret <16 x i32> %res2
}

define <16 x i32> @test_x86_avx512_maskz_psrav_d_512(<16 x i32> %a0, <16 x i32> %a1, i16 %mask) {
; X64-LABEL: test_x86_avx512_maskz_psrav_d_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpsravd %zmm1, %zmm0, %zmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_x86_avx512_maskz_psrav_d_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vpsravd %zmm1, %zmm0, %zmm0 {%k1} {z}
; X86-NEXT:    retl
  %res = call <16 x i32> @llvm.x86.avx512.psrav.d.512(<16 x i32> %a0, <16 x i32> %a1)
  %mask.cast = bitcast i16 %mask to <16 x i1>
  %res2 = select <16 x i1> %mask.cast, <16 x i32> %res, <16 x i32> zeroinitializer
  ret <16 x i32> %res2
}

declare <16 x i32> @llvm.x86.avx512.psrav.d.512(<16 x i32>, <16 x i32>) nounwind readnone

define <8 x i64> @test_x86_avx512_psrav_q_512(<8 x i64> %a0, <8 x i64> %a1) {
; CHECK-LABEL: test_x86_avx512_psrav_q_512:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vpsravq %zmm1, %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %res = call <8 x i64> @llvm.x86.avx512.psrav.q.512(<8 x i64> %a0, <8 x i64> %a1)
  ret <8 x i64> %res
}

define <8 x i64> @test_x86_avx512_mask_psrav_q_512(<8 x i64> %a0, <8 x i64> %a1, <8 x i64> %a2, i8 %mask) {
; X64-LABEL: test_x86_avx512_mask_psrav_q_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpsravq %zmm1, %zmm0, %zmm2 {%k1}
; X64-NEXT:    vmovdqa64 %zmm2, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_x86_avx512_mask_psrav_q_512:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vpsravq %zmm1, %zmm0, %zmm2 {%k1}
; X86-NEXT:    vmovdqa64 %zmm2, %zmm0
; X86-NEXT:    retl
  %res = call <8 x i64> @llvm.x86.avx512.psrav.q.512(<8 x i64> %a0, <8 x i64> %a1)
  %mask.cast = bitcast i8 %mask to <8 x i1>
  %res2 = select <8 x i1> %mask.cast, <8 x i64> %res, <8 x i64> %a2
  ret <8 x i64> %res2
}

define <8 x i64> @test_x86_avx512_maskz_psrav_q_512(<8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
; X64-LABEL: test_x86_avx512_maskz_psrav_q_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpsravq %zmm1, %zmm0, %zmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_x86_avx512_maskz_psrav_q_512:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vpsravq %zmm1, %zmm0, %zmm0 {%k1} {z}
; X86-NEXT:    retl
  %res = call <8 x i64> @llvm.x86.avx512.psrav.q.512(<8 x i64> %a0, <8 x i64> %a1)
  %mask.cast = bitcast i8 %mask to <8 x i1>
  %res2 = select <8 x i1> %mask.cast, <8 x i64> %res, <8 x i64> zeroinitializer
  ret <8 x i64> %res2
}

declare <8 x i64> @llvm.x86.avx512.psrav.q.512(<8 x i64>, <8 x i64>) nounwind readnone

define <16 x i32> @test_x86_avx512_psrlv_d_512(<16 x i32> %a0, <16 x i32> %a1) {
; CHECK-LABEL: test_x86_avx512_psrlv_d_512:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vpsrlvd %zmm1, %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %res = call <16 x i32> @llvm.x86.avx512.psrlv.d.512(<16 x i32> %a0, <16 x i32> %a1)
  ret <16 x i32> %res
}

define <16 x i32> @test_x86_avx512_psrlv_d_512_const() {
; X64-LABEL: test_x86_avx512_psrlv_d_512_const:
; X64:       # %bb.0:
; X64-NEXT:    vmovdqa64 {{.*#+}} zmm0 = [2,9,0,4294967295,3,7,4294967295,0,4,5,4294967294,0,5,3,4294967293,0]
; X64-NEXT:    vpsrlvd {{.*}}(%rip), %zmm0, %zmm0
; X64-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4294967295]
; X64-NEXT:    vpsrlvd {{.*}}(%rip), %zmm1, %zmm1
; X64-NEXT:    vpaddd %zmm1, %zmm0, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_x86_avx512_psrlv_d_512_const:
; X86:       # %bb.0:
; X86-NEXT:    vmovdqa64 {{.*#+}} zmm0 = [2,9,0,4294967295,3,7,4294967295,0,4,5,4294967294,0,5,3,4294967293,0]
; X86-NEXT:    vpsrlvd {{\.LCPI.*}}, %zmm0, %zmm0
; X86-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4294967295]
; X86-NEXT:    vpsrlvd {{\.LCPI.*}}, %zmm1, %zmm1
; X86-NEXT:    vpaddd %zmm1, %zmm0, %zmm0
; X86-NEXT:    retl
  %res0 = call <16 x i32> @llvm.x86.avx512.psrlv.d.512(<16 x i32> <i32 2, i32 9, i32 0, i32 -1, i32 3, i32 7, i32 -1, i32 0, i32 4, i32 5, i32 -2, i32 0, i32 5, i32 3, i32 -3, i32 0>, <16 x i32> <i32 1, i32 0, i32 33, i32 -1,i32 2, i32 0, i32 34, i32 -2, i32 3, i32 0, i32 35, i32 -1, i32 4, i32 0, i32 36, i32 -3>)
  %res1 = call <16 x i32> @llvm.x86.avx512.psrlv.d.512(<16 x i32> <i32 4, i32 4, i32 4, i32 4, i32 4, i32 4, i32 4, i32 4, i32 4, i32 4, i32 4, i32 4, i32 4, i32 4, i32 4, i32 -1>, <16 x i32> <i32 1, i32 1, i32 1,  i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1,  i32 1, i32 1, i32 1, i32 1, i32 -1  >)
  %res2 = add <16 x i32> %res0, %res1
  ret <16 x i32> %res2
}

define <16 x i32> @test_x86_avx512_mask_psrlv_d_512(<16 x i32> %a0, <16 x i32> %a1, <16 x i32> %a2, i16 %mask) {
; X64-LABEL: test_x86_avx512_mask_psrlv_d_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpsrlvd %zmm1, %zmm0, %zmm2 {%k1}
; X64-NEXT:    vmovdqa64 %zmm2, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_x86_avx512_mask_psrlv_d_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vpsrlvd %zmm1, %zmm0, %zmm2 {%k1}
; X86-NEXT:    vmovdqa64 %zmm2, %zmm0
; X86-NEXT:    retl
  %res = call <16 x i32> @llvm.x86.avx512.psrlv.d.512(<16 x i32> %a0, <16 x i32> %a1)
  %mask.cast = bitcast i16 %mask to <16 x i1>
  %res2 = select <16 x i1> %mask.cast, <16 x i32> %res, <16 x i32> %a2
  ret <16 x i32> %res2
}

define <16 x i32> @test_x86_avx512_maskz_psrlv_d_512(<16 x i32> %a0, <16 x i32> %a1, i16 %mask) {
; X64-LABEL: test_x86_avx512_maskz_psrlv_d_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpsrlvd %zmm1, %zmm0, %zmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_x86_avx512_maskz_psrlv_d_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1
; X86-NEXT:    vpsrlvd %zmm1, %zmm0, %zmm0 {%k1} {z}
; X86-NEXT:    retl
  %res = call <16 x i32> @llvm.x86.avx512.psrlv.d.512(<16 x i32> %a0, <16 x i32> %a1)
  %mask.cast = bitcast i16 %mask to <16 x i1>
  %res2 = select <16 x i1> %mask.cast, <16 x i32> %res, <16 x i32> zeroinitializer
  ret <16 x i32> %res2
}

declare <16 x i32> @llvm.x86.avx512.psrlv.d.512(<16 x i32>, <16 x i32>) nounwind readnone

define <8 x i64> @test_x86_avx512_psrlv_q_512(<8 x i64> %a0, <8 x i64> %a1) {
; CHECK-LABEL: test_x86_avx512_psrlv_q_512:
; CHECK:       # %bb.0:
; CHECK-NEXT:    vpsrlvq %zmm1, %zmm0, %zmm0
; CHECK-NEXT:    ret{{[l|q]}}
  %res = call <8 x i64> @llvm.x86.avx512.psrlv.q.512(<8 x i64> %a0, <8 x i64> %a1)
  ret <8 x i64> %res
}

define <8 x i64> @test_x86_avx512_psrlv_q_512_const() {
; X64-LABEL: test_x86_avx512_psrlv_q_512_const:
; X64:       # %bb.0:
; X64-NEXT:    vmovdqa64 {{.*#+}} zmm0 = [2,9,0,18446744073709551615,3,7,18446744073709551615,0]
; X64-NEXT:    vpsrlvq {{.*}}(%rip), %zmm0, %zmm0
; X64-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [4,4,4,4,4,4,4,18446744073709551615]
; X64-NEXT:    vpsrlvq {{.*}}(%rip), %zmm1, %zmm1
; X64-NEXT:    vpaddq %zmm1, %zmm0, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_x86_avx512_psrlv_q_512_const:
; X86:       # %bb.0:
; X86-NEXT:    vmovdqa64 {{.*#+}} zmm0 = [2,0,9,0,0,0,4294967295,4294967295,3,0,7,0,4294967295,4294967295,0,0]
; X86-NEXT:    vpsrlvq {{\.LCPI.*}}, %zmm0, %zmm0
; X86-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [4,0,4,0,4,0,4,0,4,0,4,0,4,0,4294967295,4294967295]
; X86-NEXT:    vpsrlvq {{\.LCPI.*}}, %zmm1, %zmm1
; X86-NEXT:    vpaddq %zmm1, %zmm0, %zmm0
; X86-NEXT:    retl
  %res0 = call <8 x i64> @llvm.x86.avx512.psrlv.q.512(<8 x i64> <i64 2, i64 9, i64 0, i64 -1, i64 3, i64 7, i64 -1, i64 0>, <8 x i64> <i64 1, i64 0, i64 33, i64 -1,i64 2, i64 0, i64 34, i64 -2>)
  %res1 = call <8 x i64> @llvm.x86.avx512.psrlv.q.512(<8 x i64> <i64 4, i64 4, i64 4, i64 4, i64 4, i64 4, i64 4, i64 -1>, <8 x i64> <i64 1, i64 1, i64 1, i64 1,  i64 1, i64 1, i64 1, i64 -1>)
  %res2 = add <8 x i64> %res0, %res1
  ret <8 x i64> %res2
}

define <8 x i64> @test_x86_avx512_mask_psrlv_q_512(<8 x i64> %a0, <8 x i64> %a1, <8 x i64> %a2, i8 %mask) {
; X64-LABEL: test_x86_avx512_mask_psrlv_q_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpsrlvq %zmm1, %zmm0, %zmm2 {%k1}
; X64-NEXT:    vmovdqa64 %zmm2, %zmm0
; X64-NEXT:    retq
;
; X86-LABEL: test_x86_avx512_mask_psrlv_q_512:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vpsrlvq %zmm1, %zmm0, %zmm2 {%k1}
; X86-NEXT:    vmovdqa64 %zmm2, %zmm0
; X86-NEXT:    retl
  %res = call <8 x i64> @llvm.x86.avx512.psrlv.q.512(<8 x i64> %a0, <8 x i64> %a1)
  %mask.cast = bitcast i8 %mask to <8 x i1>
  %res2 = select <8 x i1> %mask.cast, <8 x i64> %res, <8 x i64> %a2
  ret <8 x i64> %res2
}

define <8 x i64> @test_x86_avx512_maskz_psrlv_q_512(<8 x i64> %a0, <8 x i64> %a1, i8 %mask) {
; X64-LABEL: test_x86_avx512_maskz_psrlv_q_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovw %edi, %k1
; X64-NEXT:    vpsrlvq %zmm1, %zmm0, %zmm0 {%k1} {z}
; X64-NEXT:    retq
;
; X86-LABEL: test_x86_avx512_maskz_psrlv_q_512:
; X86:       # %bb.0:
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vpsrlvq %zmm1, %zmm0, %zmm0 {%k1} {z}
; X86-NEXT:    retl
  %res = call <8 x i64> @llvm.x86.avx512.psrlv.q.512(<8 x i64> %a0, <8 x i64> %a1)
  %mask.cast = bitcast i8 %mask to <8 x i1>
  %res2 = select <8 x i1> %mask.cast, <8 x i64> %res, <8 x i64> zeroinitializer
  ret <8 x i64> %res2
}

declare <8 x i64> @llvm.x86.avx512.psrlv.q.512(<8 x i64>, <8 x i64>) nounwind readnone

define <16 x float> @bad_mask_transition(<8 x double> %a, <8 x double> %b, <8 x double> %c, <8 x double> %d, <16 x float> %e, <16 x float> %f) {
; X64-LABEL: bad_mask_transition:
; X64:       # %bb.0: # %entry
; X64-NEXT:    vcmplt_oqpd %zmm1, %zmm0, %k0
; X64-NEXT:    kmovw %k0, %eax
; X64-NEXT:    vcmplt_oqpd %zmm3, %zmm2, %k0
; X64-NEXT:    kmovw %k0, %ecx
; X64-NEXT:    movzbl %al, %eax
; X64-NEXT:    movzbl %cl, %ecx
; X64-NEXT:    kmovw %eax, %k0
; X64-NEXT:    kmovw %ecx, %k1
; X64-NEXT:    kunpckbw %k0, %k1, %k1
; X64-NEXT:    vblendmps %zmm5, %zmm4, %zmm0 {%k1}
; X64-NEXT:    retq
;
; X86-LABEL: bad_mask_transition:
; X86:       # %bb.0: # %entry
; X86-NEXT:    pushl %ebp
; X86-NEXT:    .cfi_def_cfa_offset 8
; X86-NEXT:    .cfi_offset %ebp, -8
; X86-NEXT:    movl %esp, %ebp
; X86-NEXT:    .cfi_def_cfa_register %ebp
; X86-NEXT:    andl $-64, %esp
; X86-NEXT:    subl $64, %esp
; X86-NEXT:    vmovaps 72(%ebp), %zmm3
; X86-NEXT:    vcmplt_oqpd %zmm1, %zmm0, %k0
; X86-NEXT:    kmovw %k0, %eax
; X86-NEXT:    vcmplt_oqpd 8(%ebp), %zmm2, %k0
; X86-NEXT:    kmovw %k0, %ecx
; X86-NEXT:    movzbl %al, %eax
; X86-NEXT:    movzbl %cl, %ecx
; X86-NEXT:    kmovw %eax, %k0
; X86-NEXT:    kmovw %ecx, %k1
; X86-NEXT:    kunpckbw %k0, %k1, %k1
; X86-NEXT:    vmovaps 136(%ebp), %zmm3 {%k1}
; X86-NEXT:    vmovaps %zmm3, %zmm0
; X86-NEXT:    movl %ebp, %esp
; X86-NEXT:    popl %ebp
; X86-NEXT:    .cfi_def_cfa %esp, 4
; X86-NEXT:    retl
entry:
  %0 = call <8 x i1> @llvm.x86.avx512.cmp.pd.512(<8 x double> %a, <8 x double> %b, i32 17, i32 4)
  %1 = bitcast <8 x i1> %0 to i8
  %2 = call <8 x i1> @llvm.x86.avx512.cmp.pd.512(<8 x double> %c, <8 x double> %d, i32 17, i32 4)
  %3 = bitcast <8 x i1> %2 to i8
  %conv = zext i8 %1 to i16
  %conv2 = zext i8 %3 to i16
  %4 = bitcast i16 %conv to <16 x i1>
  %5 = bitcast i16 %conv2 to <16 x i1>
  %6 = shufflevector <16 x i1> %4, <16 x i1> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
  %7 = shufflevector <16 x i1> %5, <16 x i1> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
  %8 = shufflevector <8 x i1> %6, <8 x i1> %7, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
  %9 = select <16 x i1> %8, <16 x float> %f, <16 x float> %e
  ret <16 x float> %9
}

define <16 x float> @bad_mask_transition_2(<8 x double> %a, <8 x double> %b, <8 x double> %c, <8 x double> %d, <16 x float> %e, <16 x float> %f) {
; X64-LABEL: bad_mask_transition_2:
; X64:       # %bb.0: # %entry
; X64-NEXT:    vcmplt_oqpd %zmm1, %zmm0, %k0
; X64-NEXT:    kmovw %k0, %eax
; X64-NEXT:    movzbl %al, %eax
; X64-NEXT:    kmovw %eax, %k1
; X64-NEXT:    vblendmps %zmm5, %zmm4, %zmm0 {%k1}
; X64-NEXT:    retq
;
; X86-LABEL: bad_mask_transition_2:
; X86:       # %bb.0: # %entry
; X86-NEXT:    pushl %ebp
; X86-NEXT:    .cfi_def_cfa_offset 8
; X86-NEXT:    .cfi_offset %ebp, -8
; X86-NEXT:    movl %esp, %ebp
; X86-NEXT:    .cfi_def_cfa_register %ebp
; X86-NEXT:    andl $-64, %esp
; X86-NEXT:    subl $64, %esp
; X86-NEXT:    vmovaps 72(%ebp), %zmm2
; X86-NEXT:    vcmplt_oqpd %zmm1, %zmm0, %k0
; X86-NEXT:    kmovw %k0, %eax
; X86-NEXT:    movzbl %al, %eax
; X86-NEXT:    kmovw %eax, %k1
; X86-NEXT:    vmovaps 136(%ebp), %zmm2 {%k1}
; X86-NEXT:    vmovaps %zmm2, %zmm0
; X86-NEXT:    movl %ebp, %esp
; X86-NEXT:    popl %ebp
; X86-NEXT:    .cfi_def_cfa %esp, 4
; X86-NEXT:    retl
entry:
  %0 = call <8 x i1> @llvm.x86.avx512.cmp.pd.512(<8 x double> %a, <8 x double> %b, i32 17, i32 4)
  %1 = bitcast <8 x i1> %0 to i8
  %conv = zext i8 %1 to i16
  %2 = bitcast i16 %conv to <16 x i1>
  %3 = select <16 x i1> %2, <16 x float> %f, <16 x float> %e
  ret <16 x float> %3
}

declare <8 x double> @llvm.x86.avx512.mask.compress.v8f64(<8 x double>, <8 x double>, <8 x i1>)
declare <16 x float> @llvm.x86.avx512.mask.compress.v16f32(<16 x float>, <16 x float>, <16 x i1>)
declare <8 x i64> @llvm.x86.avx512.mask.compress.v8i64(<8 x i64>, <8 x i64>, <8 x i1>)
declare <16 x i32> @llvm.x86.avx512.mask.compress.v16i32(<16 x i32>, <16 x i32>, <16 x i1>)
declare <8 x double> @llvm.x86.avx512.mask.expand.v8f64(<8 x double>, <8 x double>, <8 x i1>)
declare <16 x float> @llvm.x86.avx512.mask.expand.v16f32(<16 x float>, <16 x float>, <16 x i1>)
declare <8 x i64> @llvm.x86.avx512.mask.expand.v8i64(<8 x i64>, <8 x i64>, <8 x i1>)
declare <16 x i32> @llvm.x86.avx512.mask.expand.v16i32(<16 x i32>, <16 x i32>, <16 x i1>)