; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+avx512vbmi2 --show-mc-encoding | FileCheck %s --check-prefixes=CHECK,X86
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512vbmi2 --show-mc-encoding | FileCheck %s --check-prefixes=CHECK,X64

define <32 x i16> @test_mask_expand_load_w_512(i8* %addr, <32 x i16> %data, i32 %mask) {
; X86-LABEL: test_mask_expand_load_w_512:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpexpandw (%eax), %zmm0 {%k1} # encoding: [0x62,0xf2,0xfd,0x49,0x62,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_expand_load_w_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpexpandw (%rdi), %zmm0 {%k1} # encoding: [0x62,0xf2,0xfd,0x49,0x62,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <32 x i16> @llvm.x86.avx512.mask.expand.load.w.512(i8* %addr, <32 x i16> %data, i32 %mask)
  ret <32 x i16> %res
}

define <32 x i16> @test_maskz_expand_load_w_512(i8* %addr, i32 %mask) {
; X86-LABEL: test_maskz_expand_load_w_512:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpexpandw (%eax), %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xc9,0x62,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_maskz_expand_load_w_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpexpandw (%rdi), %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xc9,0x62,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <32 x i16> @llvm.x86.avx512.mask.expand.load.w.512(i8* %addr, <32 x i16> zeroinitializer, i32 %mask)
  ret <32 x i16> %res
}

declare <32 x i16> @llvm.x86.avx512.mask.expand.load.w.512(i8* %addr, <32 x i16> %data, i32 %mask)

define <32 x i16> @test_expand_load_w_512(i8* %addr, <32 x i16> %data) {
; X86-LABEL: test_expand_load_w_512:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kxnord %k0, %k0, %k1 # encoding: [0xc4,0xe1,0xfd,0x46,0xc8]
; X86-NEXT:    vpexpandw (%eax), %zmm0 {%k1} # encoding: [0x62,0xf2,0xfd,0x49,0x62,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_expand_load_w_512:
; X64:       # %bb.0:
; X64-NEXT:    kxnord %k0, %k0, %k1 # encoding: [0xc4,0xe1,0xfd,0x46,0xc8]
; X64-NEXT:    vpexpandw (%rdi), %zmm0 {%k1} # encoding: [0x62,0xf2,0xfd,0x49,0x62,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <32 x i16> @llvm.x86.avx512.mask.expand.load.w.512(i8* %addr, <32 x i16> %data, i32 -1)
  ret <32 x i16> %res
}

define <64 x i8> @test_mask_expand_load_b_512(i8* %addr, <64 x i8> %data, i64 %mask) {
; X86-LABEL: test_mask_expand_load_b_512:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovq {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf8,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpexpandb (%eax), %zmm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x49,0x62,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_expand_load_b_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovq %rsi, %k1 # encoding: [0xc4,0xe1,0xfb,0x92,0xce]
; X64-NEXT:    vpexpandb (%rdi), %zmm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x49,0x62,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <64 x i8> @llvm.x86.avx512.mask.expand.load.b.512(i8* %addr, <64 x i8> %data, i64 %mask)
  ret <64 x i8> %res
}

define <64 x i8> @test_maskz_expand_load_b_512(i8* %addr, i64 %mask) {
; X86-LABEL: test_maskz_expand_load_b_512:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovq {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf8,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpexpandb (%eax), %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xc9,0x62,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_maskz_expand_load_b_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovq %rsi, %k1 # encoding: [0xc4,0xe1,0xfb,0x92,0xce]
; X64-NEXT:    vpexpandb (%rdi), %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xc9,0x62,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <64 x i8> @llvm.x86.avx512.mask.expand.load.b.512(i8* %addr, <64 x i8> zeroinitializer, i64 %mask)
  ret <64 x i8> %res
}

declare <64 x i8> @llvm.x86.avx512.mask.expand.load.b.512(i8* %addr, <64 x i8> %data, i64 %mask)

define <64 x i8> @test_expand_load_b_512(i8* %addr, <64 x i8> %data) {
; X86-LABEL: test_expand_load_b_512:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kxnorq %k0, %k0, %k1 # encoding: [0xc4,0xe1,0xfc,0x46,0xc8]
; X86-NEXT:    vpexpandb (%eax), %zmm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x49,0x62,0x00]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_expand_load_b_512:
; X64:       # %bb.0:
; X64-NEXT:    kxnorq %k0, %k0, %k1 # encoding: [0xc4,0xe1,0xfc,0x46,0xc8]
; X64-NEXT:    vpexpandb (%rdi), %zmm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x49,0x62,0x07]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <64 x i8> @llvm.x86.avx512.mask.expand.load.b.512(i8* %addr, <64 x i8> %data, i64 -1)
  ret <64 x i8> %res
}

define void @test_mask_compress_store_w_512(i8* %addr, <32 x i16> %data, i32 %mask) {
; X86-LABEL: test_mask_compress_store_w_512:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpcompressw %zmm0, (%eax) {%k1} # encoding: [0x62,0xf2,0xfd,0x49,0x63,0x00]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_compress_store_w_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vpcompressw %zmm0, (%rdi) {%k1} # encoding: [0x62,0xf2,0xfd,0x49,0x63,0x07]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
  call void @llvm.x86.avx512.mask.compress.store.w.512(i8* %addr, <32 x i16> %data, i32 %mask)
  ret void
}

declare void @llvm.x86.avx512.mask.compress.store.w.512(i8* %addr, <32 x i16> %data, i32 %mask)

define <32 x i16> @test_mask_compress_w_512(<32 x i16> %data, <32 x i16> %passthru, i32 %mask) {
; X86-LABEL: test_mask_compress_w_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpcompressw %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x49,0x63,0xc1]
; X86-NEXT:    vmovdqa64 %zmm1, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_compress_w_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpcompressw %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x49,0x63,0xc1]
; X64-NEXT:    vmovdqa64 %zmm1, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <32 x i16> @llvm.x86.avx512.mask.compress.w.512(<32 x i16> %data, <32 x i16> %passthru, i32 %mask)
  ret <32 x i16> %res
}

define <32 x i16> @test_maskz_compress_w_512(<32 x i16> %data, i32 %mask) {
; X86-LABEL: test_maskz_compress_w_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpcompressw %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xc9,0x63,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_maskz_compress_w_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpcompressw %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xc9,0x63,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <32 x i16> @llvm.x86.avx512.mask.compress.w.512(<32 x i16> %data, <32 x i16> zeroinitializer, i32 %mask)
  ret <32 x i16> %res
}

define <32 x i16> @test_compress_w_512(<32 x i16> %data) {
; CHECK-LABEL: test_compress_w_512:
; CHECK:       # %bb.0:
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <32 x i16> @llvm.x86.avx512.mask.compress.w.512(<32 x i16> %data, <32 x i16> undef, i32 -1)
  ret <32 x i16> %res
}

declare <32 x i16> @llvm.x86.avx512.mask.compress.w.512(<32 x i16> %data, <32 x i16> %src0, i32 %mask)

define void @test_compress_store_w_512(i8* %addr, <32 x i16> %data) {
; X86-LABEL: test_compress_store_w_512:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kxnord %k0, %k0, %k1 # encoding: [0xc4,0xe1,0xfd,0x46,0xc8]
; X86-NEXT:    vpcompressw %zmm0, (%eax) {%k1} # encoding: [0x62,0xf2,0xfd,0x49,0x63,0x00]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_compress_store_w_512:
; X64:       # %bb.0:
; X64-NEXT:    kxnord %k0, %k0, %k1 # encoding: [0xc4,0xe1,0xfd,0x46,0xc8]
; X64-NEXT:    vpcompressw %zmm0, (%rdi) {%k1} # encoding: [0x62,0xf2,0xfd,0x49,0x63,0x07]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
  call void @llvm.x86.avx512.mask.compress.store.w.512(i8* %addr, <32 x i16> %data, i32 -1)
  ret void
}

define void @test_mask_compress_store_b_512(i8* %addr, <64 x i8> %data, i64 %mask) {
; X86-LABEL: test_mask_compress_store_b_512:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovq {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf8,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vpcompressb %zmm0, (%eax) {%k1} # encoding: [0x62,0xf2,0x7d,0x49,0x63,0x00]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_compress_store_b_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovq %rsi, %k1 # encoding: [0xc4,0xe1,0xfb,0x92,0xce]
; X64-NEXT:    vpcompressb %zmm0, (%rdi) {%k1} # encoding: [0x62,0xf2,0x7d,0x49,0x63,0x07]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
  call void @llvm.x86.avx512.mask.compress.store.b.512(i8* %addr, <64 x i8> %data, i64 %mask)
  ret void
}

declare void @llvm.x86.avx512.mask.compress.store.b.512(i8* %addr, <64 x i8> %data, i64 %mask)

define <64 x i8> @test_mask_compress_b_512(<64 x i8> %data, <64 x i8> %passthru, i64 %mask) {
; X86-LABEL: test_mask_compress_b_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k0 # encoding: [0xc4,0xe1,0xf9,0x90,0x44,0x24,0x04]
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
; X86-NEXT:    kunpckdq %k0, %k1, %k1 # encoding: [0xc4,0xe1,0xf4,0x4b,0xc8]
; X86-NEXT:    vpcompressb %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x49,0x63,0xc1]
; X86-NEXT:    vmovdqa64 %zmm1, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc1]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_mask_compress_b_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovq %rdi, %k1 # encoding: [0xc4,0xe1,0xfb,0x92,0xcf]
; X64-NEXT:    vpcompressb %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x49,0x63,0xc1]
; X64-NEXT:    vmovdqa64 %zmm1, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc1]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <64 x i8> @llvm.x86.avx512.mask.compress.b.512(<64 x i8> %data, <64 x i8> %passthru, i64 %mask)
  ret <64 x i8> %res
}

define <64 x i8> @test_maskz_compress_b_512(<64 x i8> %data, i64 %mask) {
; X86-LABEL: test_maskz_compress_b_512:
; X86:       # %bb.0:
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k0 # encoding: [0xc4,0xe1,0xf9,0x90,0x44,0x24,0x04]
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
; X86-NEXT:    kunpckdq %k0, %k1, %k1 # encoding: [0xc4,0xe1,0xf4,0x4b,0xc8]
; X86-NEXT:    vpcompressb %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xc9,0x63,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_maskz_compress_b_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovq %rdi, %k1 # encoding: [0xc4,0xe1,0xfb,0x92,0xcf]
; X64-NEXT:    vpcompressb %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xc9,0x63,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <64 x i8> @llvm.x86.avx512.mask.compress.b.512(<64 x i8> %data, <64 x i8> zeroinitializer, i64 %mask)
  ret <64 x i8> %res
}

define <64 x i8> @test_compress_b_512(<64 x i8> %data) {
; CHECK-LABEL: test_compress_b_512:
; CHECK:       # %bb.0:
; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
  %res = call <64 x i8> @llvm.x86.avx512.mask.compress.b.512(<64 x i8> %data, <64 x i8> undef, i64 -1)
  ret <64 x i8> %res
}

declare <64 x i8> @llvm.x86.avx512.mask.compress.b.512(<64 x i8> %data, <64 x i8> %src0, i64 %mask)

define void @test_compress_store_b_512(i8* %addr, <64 x i8> %data) {
; X86-LABEL: test_compress_store_b_512:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kxnorq %k0, %k0, %k1 # encoding: [0xc4,0xe1,0xfc,0x46,0xc8]
; X86-NEXT:    vpcompressb %zmm0, (%eax) {%k1} # encoding: [0x62,0xf2,0x7d,0x49,0x63,0x00]
; X86-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_compress_store_b_512:
; X64:       # %bb.0:
; X64-NEXT:    kxnorq %k0, %k0, %k1 # encoding: [0xc4,0xe1,0xfc,0x46,0xc8]
; X64-NEXT:    vpcompressb %zmm0, (%rdi) {%k1} # encoding: [0x62,0xf2,0x7d,0x49,0x63,0x07]
; X64-NEXT:    vzeroupper # encoding: [0xc5,0xf8,0x77]
; X64-NEXT:    retq # encoding: [0xc3]
  call void @llvm.x86.avx512.mask.compress.store.b.512(i8* %addr, <64 x i8> %data, i64 -1)
  ret void
}

define <16 x i32>@test_int_x86_avx512_mask_vpshld_d_512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32> %x3, i16 %x4) {
; X86-LABEL: test_int_x86_avx512_mask_vpshld_d_512:
; X86:       # %bb.0:
; X86-NEXT:    vpshldd $22, %zmm1, %zmm0, %zmm3 # encoding: [0x62,0xf3,0x7d,0x48,0x71,0xd9,0x16]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpshldd $22, %zmm1, %zmm0, %zmm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x49,0x71,0xd1,0x16]
; X86-NEXT:    vpaddd %zmm3, %zmm2, %zmm0 # encoding: [0x62,0xf1,0x6d,0x48,0xfe,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_vpshld_d_512:
; X64:       # %bb.0:
; X64-NEXT:    vpshldd $22, %zmm1, %zmm0, %zmm3 # encoding: [0x62,0xf3,0x7d,0x48,0x71,0xd9,0x16]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpshldd $22, %zmm1, %zmm0, %zmm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x49,0x71,0xd1,0x16]
; X64-NEXT:    vpaddd %zmm3, %zmm2, %zmm0 # encoding: [0x62,0xf1,0x6d,0x48,0xfe,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <16 x i32> @llvm.x86.avx512.mask.vpshld.d.512(<16 x i32> %x0, <16 x i32> %x1, i32 22, <16 x i32> %x3, i16 %x4)
  %res1 = call <16 x i32> @llvm.x86.avx512.mask.vpshld.d.512(<16 x i32> %x0, <16 x i32> %x1, i32 22, <16 x i32> %x3, i16 -1)
  %res2 = add <16 x i32> %res, %res1
  ret <16 x i32> %res2
}
declare <16 x i32> @llvm.x86.avx512.mask.vpshld.d.512(<16 x i32>, <16 x i32>, i32, <16 x i32>, i16)

define <8 x i64>@test_int_x86_avx512_mask_vpshld_q_512(<8 x i64> %x0, <8 x i64> %x1, <8 x i64> %x3, i8 %x4) {
; X86-LABEL: test_int_x86_avx512_mask_vpshld_q_512:
; X86:       # %bb.0:
; X86-NEXT:    vpshldq $22, %zmm1, %zmm0, %zmm3 # encoding: [0x62,0xf3,0xfd,0x48,0x71,0xd9,0x16]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpshldq $22, %zmm1, %zmm0, %zmm2 {%k1} # encoding: [0x62,0xf3,0xfd,0x49,0x71,0xd1,0x16]
; X86-NEXT:    vpaddq %zmm3, %zmm2, %zmm0 # encoding: [0x62,0xf1,0xed,0x48,0xd4,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_vpshld_q_512:
; X64:       # %bb.0:
; X64-NEXT:    vpshldq $22, %zmm1, %zmm0, %zmm3 # encoding: [0x62,0xf3,0xfd,0x48,0x71,0xd9,0x16]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpshldq $22, %zmm1, %zmm0, %zmm2 {%k1} # encoding: [0x62,0xf3,0xfd,0x49,0x71,0xd1,0x16]
; X64-NEXT:    vpaddq %zmm3, %zmm2, %zmm0 # encoding: [0x62,0xf1,0xed,0x48,0xd4,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x i64> @llvm.x86.avx512.mask.vpshld.q.512(<8 x i64> %x0, <8 x i64> %x1, i32 22, <8 x i64> %x3, i8 %x4)
  %res1 = call <8 x i64> @llvm.x86.avx512.mask.vpshld.q.512(<8 x i64> %x0, <8 x i64> %x1, i32 22, <8 x i64> %x3, i8 -1)
  %res2 = add <8 x i64> %res, %res1
  ret <8 x i64> %res2
}
declare <8 x i64> @llvm.x86.avx512.mask.vpshld.q.512(<8 x i64>, <8 x i64>, i32, <8 x i64>, i8)

define <32 x i16>@test_int_x86_avx512_mask_vpshld_w_512(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x3, i32 %x4) {
; X86-LABEL: test_int_x86_avx512_mask_vpshld_w_512:
; X86:       # %bb.0:
; X86-NEXT:    vpshldw $6, %zmm1, %zmm0, %zmm3 # encoding: [0x62,0xf3,0xfd,0x48,0x70,0xd9,0x06]
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpshldw $6, %zmm1, %zmm0, %zmm2 {%k1} # encoding: [0x62,0xf3,0xfd,0x49,0x70,0xd1,0x06]
; X86-NEXT:    vpaddw %zmm3, %zmm2, %zmm0 # encoding: [0x62,0xf1,0x6d,0x48,0xfd,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_vpshld_w_512:
; X64:       # %bb.0:
; X64-NEXT:    vpshldw $6, %zmm1, %zmm0, %zmm3 # encoding: [0x62,0xf3,0xfd,0x48,0x70,0xd9,0x06]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpshldw $6, %zmm1, %zmm0, %zmm2 {%k1} # encoding: [0x62,0xf3,0xfd,0x49,0x70,0xd1,0x06]
; X64-NEXT:    vpaddw %zmm3, %zmm2, %zmm0 # encoding: [0x62,0xf1,0x6d,0x48,0xfd,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <32 x i16> @llvm.x86.avx512.mask.vpshld.w.512(<32 x i16> %x0, <32 x i16> %x1, i32 6, <32 x i16> %x3, i32 %x4)
  %res1 = call <32 x i16> @llvm.x86.avx512.mask.vpshld.w.512(<32 x i16> %x0, <32 x i16> %x1, i32 6, <32 x i16> %x3, i32 -1)
  %res2 = add <32 x i16> %res, %res1
  ret <32 x i16> %res2
}
declare <32 x i16> @llvm.x86.avx512.mask.vpshld.w.512(<32 x i16>, <32 x i16>, i32, <32 x i16>, i32)

define <16 x i32>@test_int_x86_avx512_mask_vpshrd_d_512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32> %x3, i16 %x4) {
; X86-LABEL: test_int_x86_avx512_mask_vpshrd_d_512:
; X86:       # %bb.0:
; X86-NEXT:    vpshrdd $22, %zmm1, %zmm0, %zmm3 # encoding: [0x62,0xf3,0x7d,0x48,0x73,0xd9,0x16]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpshrdd $22, %zmm1, %zmm0, %zmm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x49,0x73,0xd1,0x16]
; X86-NEXT:    vpaddd %zmm3, %zmm2, %zmm0 # encoding: [0x62,0xf1,0x6d,0x48,0xfe,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_vpshrd_d_512:
; X64:       # %bb.0:
; X64-NEXT:    vpshrdd $22, %zmm1, %zmm0, %zmm3 # encoding: [0x62,0xf3,0x7d,0x48,0x73,0xd9,0x16]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpshrdd $22, %zmm1, %zmm0, %zmm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x49,0x73,0xd1,0x16]
; X64-NEXT:    vpaddd %zmm3, %zmm2, %zmm0 # encoding: [0x62,0xf1,0x6d,0x48,0xfe,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <16 x i32> @llvm.x86.avx512.mask.vpshrd.d.512(<16 x i32> %x0, <16 x i32> %x1, i32 22, <16 x i32> %x3, i16 %x4)
  %res1 = call <16 x i32> @llvm.x86.avx512.mask.vpshrd.d.512(<16 x i32> %x0, <16 x i32> %x1, i32 22, <16 x i32> %x3, i16 -1)
  %res2 = add <16 x i32> %res, %res1
  ret <16 x i32> %res2
}
declare <16 x i32> @llvm.x86.avx512.mask.vpshrd.d.512(<16 x i32>, <16 x i32>, i32, <16 x i32>, i16)

define <8 x i64>@test_int_x86_avx512_mask_vpshrd_q_512(<8 x i64> %x0, <8 x i64> %x1, <8 x i64> %x3, i8 %x4) {
; X86-LABEL: test_int_x86_avx512_mask_vpshrd_q_512:
; X86:       # %bb.0:
; X86-NEXT:    vpshrdq $22, %zmm1, %zmm0, %zmm3 # encoding: [0x62,0xf3,0xfd,0x48,0x73,0xd9,0x16]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpshrdq $22, %zmm1, %zmm0, %zmm2 {%k1} # encoding: [0x62,0xf3,0xfd,0x49,0x73,0xd1,0x16]
; X86-NEXT:    vpaddq %zmm3, %zmm2, %zmm0 # encoding: [0x62,0xf1,0xed,0x48,0xd4,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_vpshrd_q_512:
; X64:       # %bb.0:
; X64-NEXT:    vpshrdq $22, %zmm1, %zmm0, %zmm3 # encoding: [0x62,0xf3,0xfd,0x48,0x73,0xd9,0x16]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpshrdq $22, %zmm1, %zmm0, %zmm2 {%k1} # encoding: [0x62,0xf3,0xfd,0x49,0x73,0xd1,0x16]
; X64-NEXT:    vpaddq %zmm3, %zmm2, %zmm0 # encoding: [0x62,0xf1,0xed,0x48,0xd4,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <8 x i64> @llvm.x86.avx512.mask.vpshrd.q.512(<8 x i64> %x0, <8 x i64> %x1, i32 22, <8 x i64> %x3, i8 %x4)
  %res1 = call <8 x i64> @llvm.x86.avx512.mask.vpshrd.q.512(<8 x i64> %x0, <8 x i64> %x1, i32 22, <8 x i64> %x3, i8 -1)
  %res2 = add <8 x i64> %res, %res1
  ret <8 x i64> %res2
}
declare <8 x i64> @llvm.x86.avx512.mask.vpshrd.q.512(<8 x i64>, <8 x i64>, i32, <8 x i64>, i8)

define <32 x i16>@test_int_x86_avx512_mask_vpshrd_w_512(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x3, i32 %x4) {
; X86-LABEL: test_int_x86_avx512_mask_vpshrd_w_512:
; X86:       # %bb.0:
; X86-NEXT:    vpshrdw $6, %zmm1, %zmm0, %zmm3 # encoding: [0x62,0xf3,0xfd,0x48,0x72,0xd9,0x06]
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpshrdw $6, %zmm1, %zmm0, %zmm2 {%k1} # encoding: [0x62,0xf3,0xfd,0x49,0x72,0xd1,0x06]
; X86-NEXT:    vpaddw %zmm3, %zmm2, %zmm0 # encoding: [0x62,0xf1,0x6d,0x48,0xfd,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_vpshrd_w_512:
; X64:       # %bb.0:
; X64-NEXT:    vpshrdw $6, %zmm1, %zmm0, %zmm3 # encoding: [0x62,0xf3,0xfd,0x48,0x72,0xd9,0x06]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpshrdw $6, %zmm1, %zmm0, %zmm2 {%k1} # encoding: [0x62,0xf3,0xfd,0x49,0x72,0xd1,0x06]
; X64-NEXT:    vpaddw %zmm3, %zmm2, %zmm0 # encoding: [0x62,0xf1,0x6d,0x48,0xfd,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %res = call <32 x i16> @llvm.x86.avx512.mask.vpshrd.w.512(<32 x i16> %x0, <32 x i16> %x1, i32 6, <32 x i16> %x3, i32 %x4)
  %res1 = call <32 x i16> @llvm.x86.avx512.mask.vpshrd.w.512(<32 x i16> %x0, <32 x i16> %x1, i32 6, <32 x i16> %x3, i32 -1)
  %res2 = add <32 x i16> %res, %res1
  ret <32 x i16> %res2
}
declare <32 x i16> @llvm.x86.avx512.mask.vpshrd.w.512(<32 x i16>, <32 x i16>, i32, <32 x i16>, i32)

define <16 x i32>@test_int_x86_avx512_mask_vpshld_d_512_2(<16 x i32> %x0, <16 x i32> %x1, <16 x i32> %x3, i16 %x4) {
; X86-LABEL: test_int_x86_avx512_mask_vpshld_d_512_2:
; X86:       # %bb.0:
; X86-NEXT:    vpshldd $22, %zmm1, %zmm0, %zmm3 # encoding: [0x62,0xf3,0x7d,0x48,0x71,0xd9,0x16]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpshldd $22, %zmm1, %zmm0, %zmm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x49,0x71,0xd1,0x16]
; X86-NEXT:    vpaddd %zmm3, %zmm2, %zmm0 # encoding: [0x62,0xf1,0x6d,0x48,0xfe,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_vpshld_d_512_2:
; X64:       # %bb.0:
; X64-NEXT:    vpshldd $22, %zmm1, %zmm0, %zmm3 # encoding: [0x62,0xf3,0x7d,0x48,0x71,0xd9,0x16]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpshldd $22, %zmm1, %zmm0, %zmm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x49,0x71,0xd1,0x16]
; X64-NEXT:    vpaddd %zmm3, %zmm2, %zmm0 # encoding: [0x62,0xf1,0x6d,0x48,0xfe,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = call <16 x i32> @llvm.x86.avx512.vpshld.d.512(<16 x i32> %x0, <16 x i32> %x1, i32 22)
  %2 = bitcast i16 %x4 to <16 x i1>
  %3 = select <16 x i1> %2, <16 x i32> %1, <16 x i32> %x3
  %4 = call <16 x i32> @llvm.x86.avx512.vpshld.d.512(<16 x i32> %x0, <16 x i32> %x1, i32 22)
  %res2 = add <16 x i32> %3, %4
  ret <16 x i32> %res2
}
declare <16 x i32> @llvm.x86.avx512.vpshld.d.512(<16 x i32>, <16 x i32>, i32)

define <8 x i64>@test_int_x86_avx512_mask_vpshld_q_512_2(<8 x i64> %x0, <8 x i64> %x1, <8 x i64> %x3, i8 %x4) {
; X86-LABEL: test_int_x86_avx512_mask_vpshld_q_512_2:
; X86:       # %bb.0:
; X86-NEXT:    vpshldq $22, %zmm1, %zmm0, %zmm3 # encoding: [0x62,0xf3,0xfd,0x48,0x71,0xd9,0x16]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpshldq $22, %zmm1, %zmm0, %zmm2 {%k1} # encoding: [0x62,0xf3,0xfd,0x49,0x71,0xd1,0x16]
; X86-NEXT:    vpaddq %zmm3, %zmm2, %zmm0 # encoding: [0x62,0xf1,0xed,0x48,0xd4,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_vpshld_q_512_2:
; X64:       # %bb.0:
; X64-NEXT:    vpshldq $22, %zmm1, %zmm0, %zmm3 # encoding: [0x62,0xf3,0xfd,0x48,0x71,0xd9,0x16]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpshldq $22, %zmm1, %zmm0, %zmm2 {%k1} # encoding: [0x62,0xf3,0xfd,0x49,0x71,0xd1,0x16]
; X64-NEXT:    vpaddq %zmm3, %zmm2, %zmm0 # encoding: [0x62,0xf1,0xed,0x48,0xd4,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = call <8 x i64> @llvm.x86.avx512.vpshld.q.512(<8 x i64> %x0, <8 x i64> %x1, i32 22)
  %2 = bitcast i8 %x4 to <8 x i1>
  %3 = select <8 x i1> %2, <8 x i64> %1, <8 x i64> %x3
  %4 = call <8 x i64> @llvm.x86.avx512.vpshld.q.512(<8 x i64> %x0, <8 x i64> %x1, i32 22)
  %res2 = add <8 x i64> %3, %4
  ret <8 x i64> %res2
}
declare <8 x i64> @llvm.x86.avx512.vpshld.q.512(<8 x i64>, <8 x i64>, i32)

define <32 x i16>@test_int_x86_avx512_mask_vpshld_w_512_2(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x3, i32 %x4) {
; X86-LABEL: test_int_x86_avx512_mask_vpshld_w_512_2:
; X86:       # %bb.0:
; X86-NEXT:    vpshldw $6, %zmm1, %zmm0, %zmm3 # encoding: [0x62,0xf3,0xfd,0x48,0x70,0xd9,0x06]
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpshldw $6, %zmm1, %zmm0, %zmm2 {%k1} # encoding: [0x62,0xf3,0xfd,0x49,0x70,0xd1,0x06]
; X86-NEXT:    vpaddw %zmm3, %zmm2, %zmm0 # encoding: [0x62,0xf1,0x6d,0x48,0xfd,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_vpshld_w_512_2:
; X64:       # %bb.0:
; X64-NEXT:    vpshldw $6, %zmm1, %zmm0, %zmm3 # encoding: [0x62,0xf3,0xfd,0x48,0x70,0xd9,0x06]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpshldw $6, %zmm1, %zmm0, %zmm2 {%k1} # encoding: [0x62,0xf3,0xfd,0x49,0x70,0xd1,0x06]
; X64-NEXT:    vpaddw %zmm3, %zmm2, %zmm0 # encoding: [0x62,0xf1,0x6d,0x48,0xfd,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = call <32 x i16> @llvm.x86.avx512.vpshld.w.512(<32 x i16> %x0, <32 x i16> %x1, i32 6)
  %2 = bitcast i32 %x4 to <32 x i1>
  %3 = select <32 x i1> %2, <32 x i16> %1, <32 x i16> %x3
  %4 = call <32 x i16> @llvm.x86.avx512.vpshld.w.512(<32 x i16> %x0, <32 x i16> %x1, i32 6)
  %res2 = add <32 x i16> %3, %4
  ret <32 x i16> %res2
}
declare <32 x i16> @llvm.x86.avx512.vpshld.w.512(<32 x i16>, <32 x i16>, i32)

define <16 x i32>@test_int_x86_avx512_mask_vpshrd_d_512_2(<16 x i32> %x0, <16 x i32> %x1, <16 x i32> %x3, i16 %x4) {
; X86-LABEL: test_int_x86_avx512_mask_vpshrd_d_512_2:
; X86:       # %bb.0:
; X86-NEXT:    vpshrdd $22, %zmm1, %zmm0, %zmm3 # encoding: [0x62,0xf3,0x7d,0x48,0x73,0xd9,0x16]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpshrdd $22, %zmm1, %zmm0, %zmm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x49,0x73,0xd1,0x16]
; X86-NEXT:    vpaddd %zmm3, %zmm2, %zmm0 # encoding: [0x62,0xf1,0x6d,0x48,0xfe,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_vpshrd_d_512_2:
; X64:       # %bb.0:
; X64-NEXT:    vpshrdd $22, %zmm1, %zmm0, %zmm3 # encoding: [0x62,0xf3,0x7d,0x48,0x73,0xd9,0x16]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpshrdd $22, %zmm1, %zmm0, %zmm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x49,0x73,0xd1,0x16]
; X64-NEXT:    vpaddd %zmm3, %zmm2, %zmm0 # encoding: [0x62,0xf1,0x6d,0x48,0xfe,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = call <16 x i32> @llvm.x86.avx512.vpshrd.d.512(<16 x i32> %x0, <16 x i32> %x1, i32 22)
  %2 = bitcast i16 %x4 to <16 x i1>
  %3 = select <16 x i1> %2, <16 x i32> %1, <16 x i32> %x3
  %4 = call <16 x i32> @llvm.x86.avx512.vpshrd.d.512(<16 x i32> %x0, <16 x i32> %x1, i32 22)
  %res2 = add <16 x i32> %3, %4
  ret <16 x i32> %res2
}
declare <16 x i32> @llvm.x86.avx512.vpshrd.d.512(<16 x i32>, <16 x i32>, i32)

define <8 x i64>@test_int_x86_avx512_mask_vpshrd_q_512_2(<8 x i64> %x0, <8 x i64> %x1, <8 x i64> %x3, i8 %x4) {
; X86-LABEL: test_int_x86_avx512_mask_vpshrd_q_512_2:
; X86:       # %bb.0:
; X86-NEXT:    vpshrdq $22, %zmm1, %zmm0, %zmm3 # encoding: [0x62,0xf3,0xfd,0x48,0x73,0xd9,0x16]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT:    kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
; X86-NEXT:    vpshrdq $22, %zmm1, %zmm0, %zmm2 {%k1} # encoding: [0x62,0xf3,0xfd,0x49,0x73,0xd1,0x16]
; X86-NEXT:    vpaddq %zmm3, %zmm2, %zmm0 # encoding: [0x62,0xf1,0xed,0x48,0xd4,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_vpshrd_q_512_2:
; X64:       # %bb.0:
; X64-NEXT:    vpshrdq $22, %zmm1, %zmm0, %zmm3 # encoding: [0x62,0xf3,0xfd,0x48,0x73,0xd9,0x16]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpshrdq $22, %zmm1, %zmm0, %zmm2 {%k1} # encoding: [0x62,0xf3,0xfd,0x49,0x73,0xd1,0x16]
; X64-NEXT:    vpaddq %zmm3, %zmm2, %zmm0 # encoding: [0x62,0xf1,0xed,0x48,0xd4,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = call <8 x i64> @llvm.x86.avx512.vpshrd.q.512(<8 x i64> %x0, <8 x i64> %x1, i32 22)
  %2 = bitcast i8 %x4 to <8 x i1>
  %3 = select <8 x i1> %2, <8 x i64> %1, <8 x i64> %x3
  %4 = call <8 x i64> @llvm.x86.avx512.vpshrd.q.512(<8 x i64> %x0, <8 x i64> %x1, i32 22)
  %res2 = add <8 x i64> %3, %4
  ret <8 x i64> %res2
}
declare <8 x i64> @llvm.x86.avx512.vpshrd.q.512(<8 x i64>, <8 x i64>, i32)

define <32 x i16>@test_int_x86_avx512_mask_vpshrd_w_512_2(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x3, i32 %x4) {
; X86-LABEL: test_int_x86_avx512_mask_vpshrd_w_512_2:
; X86:       # %bb.0:
; X86-NEXT:    vpshrdw $6, %zmm1, %zmm0, %zmm3 # encoding: [0x62,0xf3,0xfd,0x48,0x72,0xd9,0x06]
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
; X86-NEXT:    vpshrdw $6, %zmm1, %zmm0, %zmm2 {%k1} # encoding: [0x62,0xf3,0xfd,0x49,0x72,0xd1,0x06]
; X86-NEXT:    vpaddw %zmm3, %zmm2, %zmm0 # encoding: [0x62,0xf1,0x6d,0x48,0xfd,0xc3]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_vpshrd_w_512_2:
; X64:       # %bb.0:
; X64-NEXT:    vpshrdw $6, %zmm1, %zmm0, %zmm3 # encoding: [0x62,0xf3,0xfd,0x48,0x72,0xd9,0x06]
; X64-NEXT:    kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
; X64-NEXT:    vpshrdw $6, %zmm1, %zmm0, %zmm2 {%k1} # encoding: [0x62,0xf3,0xfd,0x49,0x72,0xd1,0x06]
; X64-NEXT:    vpaddw %zmm3, %zmm2, %zmm0 # encoding: [0x62,0xf1,0x6d,0x48,0xfd,0xc3]
; X64-NEXT:    retq # encoding: [0xc3]
  %1 = call <32 x i16> @llvm.x86.avx512.vpshrd.w.512(<32 x i16> %x0, <32 x i16> %x1, i32 6)
  %2 = bitcast i32 %x4 to <32 x i1>
  %3 = select <32 x i1> %2, <32 x i16> %1, <32 x i16> %x3
  %4 = call <32 x i16> @llvm.x86.avx512.vpshrd.w.512(<32 x i16> %x0, <32 x i16> %x1, i32 6)
  %res2 = add <32 x i16> %3, %4
  ret <32 x i16> %res2
}
declare <32 x i16> @llvm.x86.avx512.vpshrd.w.512(<32 x i16>, <32 x i16>, i32)

declare <16 x i32> @llvm.x86.avx512.mask.vpshrdv.d.512(<16 x i32>, <16 x i32>, <16 x i32>, i16)
declare <16 x i32> @llvm.x86.avx512.maskz.vpshrdv.d.512(<16 x i32>, <16 x i32>, <16 x i32>, i16)

define <16 x i32>@test_int_x86_avx512_mask_vpshrdv_d_512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32>* %x2p, <16 x i32> %x4, i16 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_vpshrdv_d_512:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vmovdqa64 %zmm0, %zmm3 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xd8]
; X86-NEXT:    vpshrdvd (%eax), %zmm1, %zmm3 {%k1} # encoding: [0x62,0xf2,0x75,0x49,0x73,0x18]
; X86-NEXT:    vmovdqa64 %zmm0, %zmm4 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xe0]
; X86-NEXT:    vpshrdvd %zmm2, %zmm1, %zmm4 # encoding: [0x62,0xf2,0x75,0x48,0x73,0xe2]
; X86-NEXT:    vpshrdvd %zmm2, %zmm1, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xc9,0x73,0xc2]
; X86-NEXT:    vpaddd %zmm0, %zmm4, %zmm0 # encoding: [0x62,0xf1,0x5d,0x48,0xfe,0xc0]
; X86-NEXT:    vpaddd %zmm0, %zmm3, %zmm0 # encoding: [0x62,0xf1,0x65,0x48,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_vpshrdv_d_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vmovdqa64 %zmm0, %zmm3 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xd8]
; X64-NEXT:    vpshrdvd (%rdi), %zmm1, %zmm3 {%k1} # encoding: [0x62,0xf2,0x75,0x49,0x73,0x1f]
; X64-NEXT:    vmovdqa64 %zmm0, %zmm4 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xe0]
; X64-NEXT:    vpshrdvd %zmm2, %zmm1, %zmm4 # encoding: [0x62,0xf2,0x75,0x48,0x73,0xe2]
; X64-NEXT:    vpshrdvd %zmm2, %zmm1, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xc9,0x73,0xc2]
; X64-NEXT:    vpaddd %zmm0, %zmm4, %zmm0 # encoding: [0x62,0xf1,0x5d,0x48,0xfe,0xc0]
; X64-NEXT:    vpaddd %zmm0, %zmm3, %zmm0 # encoding: [0x62,0xf1,0x65,0x48,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %x2 = load <16 x i32>, <16 x i32>* %x2p
  %res = call <16 x i32> @llvm.x86.avx512.mask.vpshrdv.d.512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32> %x2, i16 %x3)
  %res1 = call <16 x i32> @llvm.x86.avx512.mask.vpshrdv.d.512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32> %x4, i16 -1)
  %res2 = call <16 x i32> @llvm.x86.avx512.maskz.vpshrdv.d.512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32> %x4, i16  %x3)
  %res3 = add <16 x i32> %res, %res1
  %res4 = add <16 x i32> %res2, %res3
  ret <16 x i32> %res4
}

declare <8 x i64> @llvm.x86.avx512.mask.vpshrdv.q.512(<8 x i64>, <8 x i64>, <8 x i64>, i8)
declare <8 x i64> @llvm.x86.avx512.maskz.vpshrdv.q.512(<8 x i64>, <8 x i64>, <8 x i64>, i8)

define <8 x i64>@test_int_x86_avx512_mask_vpshrdv_q_512(<8 x i64> %x0, <8 x i64> %x1, <8 x i64>* %x2p, <8 x i64> %x4, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_vpshrdv_q_512:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
; X86-NEXT:    kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
; X86-NEXT:    vmovdqa64 %zmm0, %zmm3 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xd8]
; X86-NEXT:    vpshrdvq (%eax), %zmm1, %zmm3 {%k1} # encoding: [0x62,0xf2,0xf5,0x49,0x73,0x18]
; X86-NEXT:    vmovdqa64 %zmm0, %zmm4 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xe0]
; X86-NEXT:    vpshrdvq %zmm2, %zmm1, %zmm4 # encoding: [0x62,0xf2,0xf5,0x48,0x73,0xe2]
; X86-NEXT:    vpshrdvq %zmm2, %zmm1, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0xc9,0x73,0xc2]
; X86-NEXT:    vpaddq %zmm0, %zmm4, %zmm0 # encoding: [0x62,0xf1,0xdd,0x48,0xd4,0xc0]
; X86-NEXT:    vpaddq %zmm0, %zmm3, %zmm0 # encoding: [0x62,0xf1,0xe5,0x48,0xd4,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_vpshrdv_q_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vmovdqa64 %zmm0, %zmm3 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xd8]
; X64-NEXT:    vpshrdvq (%rdi), %zmm1, %zmm3 {%k1} # encoding: [0x62,0xf2,0xf5,0x49,0x73,0x1f]
; X64-NEXT:    vmovdqa64 %zmm0, %zmm4 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xe0]
; X64-NEXT:    vpshrdvq %zmm2, %zmm1, %zmm4 # encoding: [0x62,0xf2,0xf5,0x48,0x73,0xe2]
; X64-NEXT:    vpshrdvq %zmm2, %zmm1, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0xc9,0x73,0xc2]
; X64-NEXT:    vpaddq %zmm0, %zmm4, %zmm0 # encoding: [0x62,0xf1,0xdd,0x48,0xd4,0xc0]
; X64-NEXT:    vpaddq %zmm0, %zmm3, %zmm0 # encoding: [0x62,0xf1,0xe5,0x48,0xd4,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %x2 = load <8 x i64>, <8 x i64>* %x2p
  %res = call <8 x i64> @llvm.x86.avx512.mask.vpshrdv.q.512(<8 x i64> %x0, <8 x i64> %x1, <8 x i64> %x2, i8 %x3)
  %res1 = call <8 x i64> @llvm.x86.avx512.mask.vpshrdv.q.512(<8 x i64> %x0, <8 x i64> %x1, <8 x i64> %x4, i8 -1)
  %res2 = call <8 x i64> @llvm.x86.avx512.maskz.vpshrdv.q.512(<8 x i64> %x0, <8 x i64> %x1, <8 x i64> %x4, i8  %x3)
  %res3 = add <8 x i64> %res, %res1
  %res4 = add <8 x i64> %res2, %res3
  ret <8 x i64> %res4
}

declare <32 x i16> @llvm.x86.avx512.mask.vpshrdv.w.512(<32 x i16>, <32 x i16>, <32 x i16>, i32)
declare <32 x i16> @llvm.x86.avx512.maskz.vpshrdv.w.512(<32 x i16>, <32 x i16>, <32 x i16>, i32)

define <32 x i16>@test_int_x86_avx512_mask_vpshrdv_w_512(<32 x i16> %x0, <32 x i16> %x1, <32 x i16>* %x2p, <32 x i16> %x4, i32 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_vpshrdv_w_512:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vmovdqa64 %zmm0, %zmm3 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xd8]
; X86-NEXT:    vpshrdvw (%eax), %zmm1, %zmm3 {%k1} # encoding: [0x62,0xf2,0xf5,0x49,0x72,0x18]
; X86-NEXT:    vmovdqa64 %zmm0, %zmm4 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xe0]
; X86-NEXT:    vpshrdvw %zmm2, %zmm1, %zmm4 # encoding: [0x62,0xf2,0xf5,0x48,0x72,0xe2]
; X86-NEXT:    vpshrdvw %zmm2, %zmm1, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0xc9,0x72,0xc2]
; X86-NEXT:    vpaddw %zmm0, %zmm4, %zmm0 # encoding: [0x62,0xf1,0x5d,0x48,0xfd,0xc0]
; X86-NEXT:    vpaddw %zmm0, %zmm3, %zmm0 # encoding: [0x62,0xf1,0x65,0x48,0xfd,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_vpshrdv_w_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vmovdqa64 %zmm0, %zmm3 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xd8]
; X64-NEXT:    vpshrdvw (%rdi), %zmm1, %zmm3 {%k1} # encoding: [0x62,0xf2,0xf5,0x49,0x72,0x1f]
; X64-NEXT:    vmovdqa64 %zmm0, %zmm4 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xe0]
; X64-NEXT:    vpshrdvw %zmm2, %zmm1, %zmm4 # encoding: [0x62,0xf2,0xf5,0x48,0x72,0xe2]
; X64-NEXT:    vpshrdvw %zmm2, %zmm1, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0xc9,0x72,0xc2]
; X64-NEXT:    vpaddw %zmm0, %zmm4, %zmm0 # encoding: [0x62,0xf1,0x5d,0x48,0xfd,0xc0]
; X64-NEXT:    vpaddw %zmm0, %zmm3, %zmm0 # encoding: [0x62,0xf1,0x65,0x48,0xfd,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %x2 = load <32 x i16>, <32 x i16>* %x2p
  %res = call <32 x i16> @llvm.x86.avx512.mask.vpshrdv.w.512(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x2, i32 %x3)
  %res1 = call <32 x i16> @llvm.x86.avx512.mask.vpshrdv.w.512(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x4, i32 -1)
  %res2 = call <32 x i16> @llvm.x86.avx512.maskz.vpshrdv.w.512(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x4, i32  %x3)
  %res3 = add <32 x i16> %res, %res1
  %res4 = add <32 x i16> %res2, %res3
  ret <32 x i16> %res4
}

declare <16 x i32> @llvm.x86.avx512.mask.vpshldv.d.512(<16 x i32>, <16 x i32>, <16 x i32>, i16)
declare <16 x i32> @llvm.x86.avx512.maskz.vpshldv.d.512(<16 x i32>, <16 x i32>, <16 x i32>, i16)

define <16 x i32>@test_int_x86_avx512_mask_vpshldv_d_512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32>* %x2p, <16 x i32> %x4, i16 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_vpshldv_d_512:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vmovdqa64 %zmm0, %zmm3 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xd8]
; X86-NEXT:    vpshldvd (%eax), %zmm1, %zmm3 {%k1} # encoding: [0x62,0xf2,0x75,0x49,0x71,0x18]
; X86-NEXT:    vmovdqa64 %zmm0, %zmm4 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xe0]
; X86-NEXT:    vpshldvd %zmm2, %zmm1, %zmm4 # encoding: [0x62,0xf2,0x75,0x48,0x71,0xe2]
; X86-NEXT:    vpshldvd %zmm2, %zmm1, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xc9,0x71,0xc2]
; X86-NEXT:    vpaddd %zmm0, %zmm4, %zmm0 # encoding: [0x62,0xf1,0x5d,0x48,0xfe,0xc0]
; X86-NEXT:    vpaddd %zmm0, %zmm3, %zmm0 # encoding: [0x62,0xf1,0x65,0x48,0xfe,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_vpshldv_d_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vmovdqa64 %zmm0, %zmm3 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xd8]
; X64-NEXT:    vpshldvd (%rdi), %zmm1, %zmm3 {%k1} # encoding: [0x62,0xf2,0x75,0x49,0x71,0x1f]
; X64-NEXT:    vmovdqa64 %zmm0, %zmm4 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xe0]
; X64-NEXT:    vpshldvd %zmm2, %zmm1, %zmm4 # encoding: [0x62,0xf2,0x75,0x48,0x71,0xe2]
; X64-NEXT:    vpshldvd %zmm2, %zmm1, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xc9,0x71,0xc2]
; X64-NEXT:    vpaddd %zmm0, %zmm4, %zmm0 # encoding: [0x62,0xf1,0x5d,0x48,0xfe,0xc0]
; X64-NEXT:    vpaddd %zmm0, %zmm3, %zmm0 # encoding: [0x62,0xf1,0x65,0x48,0xfe,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %x2 = load <16 x i32>, <16 x i32>* %x2p
  %res = call <16 x i32> @llvm.x86.avx512.mask.vpshldv.d.512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32> %x2, i16 %x3)
  %res1 = call <16 x i32> @llvm.x86.avx512.mask.vpshldv.d.512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32> %x4, i16 -1)
  %res2 = call <16 x i32> @llvm.x86.avx512.maskz.vpshldv.d.512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32> %x4, i16  %x3)
  %res3 = add <16 x i32> %res, %res1
  %res4 = add <16 x i32> %res2, %res3
  ret <16 x i32> %res4
}

declare <8 x i64> @llvm.x86.avx512.mask.vpshldv.q.512(<8 x i64>, <8 x i64>, <8 x i64>, i8)
declare <8 x i64> @llvm.x86.avx512.maskz.vpshldv.q.512(<8 x i64>, <8 x i64>, <8 x i64>, i8)

define <8 x i64>@test_int_x86_avx512_mask_vpshldv_q_512(<8 x i64> %x0, <8 x i64> %x1, <8 x i64>* %x2p, <8 x i64> %x4, i8 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_vpshldv_q_512:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
; X86-NEXT:    kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
; X86-NEXT:    vmovdqa64 %zmm0, %zmm3 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xd8]
; X86-NEXT:    vpshldvq (%eax), %zmm1, %zmm3 {%k1} # encoding: [0x62,0xf2,0xf5,0x49,0x71,0x18]
; X86-NEXT:    vmovdqa64 %zmm0, %zmm4 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xe0]
; X86-NEXT:    vpshldvq %zmm2, %zmm1, %zmm4 # encoding: [0x62,0xf2,0xf5,0x48,0x71,0xe2]
; X86-NEXT:    vpshldvq %zmm2, %zmm1, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0xc9,0x71,0xc2]
; X86-NEXT:    vpaddq %zmm0, %zmm4, %zmm0 # encoding: [0x62,0xf1,0xdd,0x48,0xd4,0xc0]
; X86-NEXT:    vpaddq %zmm0, %zmm3, %zmm0 # encoding: [0x62,0xf1,0xe5,0x48,0xd4,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_vpshldv_q_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vmovdqa64 %zmm0, %zmm3 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xd8]
; X64-NEXT:    vpshldvq (%rdi), %zmm1, %zmm3 {%k1} # encoding: [0x62,0xf2,0xf5,0x49,0x71,0x1f]
; X64-NEXT:    vmovdqa64 %zmm0, %zmm4 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xe0]
; X64-NEXT:    vpshldvq %zmm2, %zmm1, %zmm4 # encoding: [0x62,0xf2,0xf5,0x48,0x71,0xe2]
; X64-NEXT:    vpshldvq %zmm2, %zmm1, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0xc9,0x71,0xc2]
; X64-NEXT:    vpaddq %zmm0, %zmm4, %zmm0 # encoding: [0x62,0xf1,0xdd,0x48,0xd4,0xc0]
; X64-NEXT:    vpaddq %zmm0, %zmm3, %zmm0 # encoding: [0x62,0xf1,0xe5,0x48,0xd4,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %x2 = load <8 x i64>, <8 x i64>* %x2p
  %res = call <8 x i64> @llvm.x86.avx512.mask.vpshldv.q.512(<8 x i64> %x0, <8 x i64> %x1, <8 x i64> %x2, i8 %x3)
  %res1 = call <8 x i64> @llvm.x86.avx512.mask.vpshldv.q.512(<8 x i64> %x0, <8 x i64> %x1, <8 x i64> %x4, i8 -1)
  %res2 = call <8 x i64> @llvm.x86.avx512.maskz.vpshldv.q.512(<8 x i64> %x0, <8 x i64> %x1, <8 x i64> %x4, i8  %x3)
  %res3 = add <8 x i64> %res, %res1
  %res4 = add <8 x i64> %res2, %res3
  ret <8 x i64> %res4
}

declare <32 x i16> @llvm.x86.avx512.mask.vpshldv.w.512(<32 x i16>, <32 x i16>, <32 x i16>, i32)
declare <32 x i16> @llvm.x86.avx512.maskz.vpshldv.w.512(<32 x i16>, <32 x i16>, <32 x i16>, i32)

define <32 x i16>@test_int_x86_avx512_mask_vpshldv_w_512(<32 x i16> %x0, <32 x i16> %x1, <32 x i16>* %x2p, <32 x i16> %x4, i32 %x3) {
; X86-LABEL: test_int_x86_avx512_mask_vpshldv_w_512:
; X86:       # %bb.0:
; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
; X86-NEXT:    kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
; X86-NEXT:    vmovdqa64 %zmm0, %zmm3 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xd8]
; X86-NEXT:    vpshldvw (%eax), %zmm1, %zmm3 {%k1} # encoding: [0x62,0xf2,0xf5,0x49,0x70,0x18]
; X86-NEXT:    vmovdqa64 %zmm0, %zmm4 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xe0]
; X86-NEXT:    vpshldvw %zmm2, %zmm1, %zmm4 # encoding: [0x62,0xf2,0xf5,0x48,0x70,0xe2]
; X86-NEXT:    vpshldvw %zmm2, %zmm1, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0xc9,0x70,0xc2]
; X86-NEXT:    vpaddw %zmm0, %zmm4, %zmm0 # encoding: [0x62,0xf1,0x5d,0x48,0xfd,0xc0]
; X86-NEXT:    vpaddw %zmm0, %zmm3, %zmm0 # encoding: [0x62,0xf1,0x65,0x48,0xfd,0xc0]
; X86-NEXT:    retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_mask_vpshldv_w_512:
; X64:       # %bb.0:
; X64-NEXT:    kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
; X64-NEXT:    vmovdqa64 %zmm0, %zmm3 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xd8]
; X64-NEXT:    vpshldvw (%rdi), %zmm1, %zmm3 {%k1} # encoding: [0x62,0xf2,0xf5,0x49,0x70,0x1f]
; X64-NEXT:    vmovdqa64 %zmm0, %zmm4 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xe0]
; X64-NEXT:    vpshldvw %zmm2, %zmm1, %zmm4 # encoding: [0x62,0xf2,0xf5,0x48,0x70,0xe2]
; X64-NEXT:    vpshldvw %zmm2, %zmm1, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0xc9,0x70,0xc2]
; X64-NEXT:    vpaddw %zmm0, %zmm4, %zmm0 # encoding: [0x62,0xf1,0x5d,0x48,0xfd,0xc0]
; X64-NEXT:    vpaddw %zmm0, %zmm3, %zmm0 # encoding: [0x62,0xf1,0x65,0x48,0xfd,0xc0]
; X64-NEXT:    retq # encoding: [0xc3]
  %x2 = load <32 x i16>, <32 x i16>* %x2p
  %res = call <32 x i16> @llvm.x86.avx512.mask.vpshldv.w.512(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x2, i32 %x3)
  %res1 = call <32 x i16> @llvm.x86.avx512.mask.vpshldv.w.512(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x4, i32 -1)
  %res2 = call <32 x i16> @llvm.x86.avx512.maskz.vpshldv.w.512(<32 x i16> %x0, <32 x i16> %x1, <32 x i16> %x4, i32  %x3)
  %res3 = add <32 x i16> %res, %res1
  %res4 = add <32 x i16> %res2, %res3
  ret <32 x i16> %res4
}