summaryrefslogtreecommitdiffstats
path: root/llvm/test/CodeGen/X86
diff options
context:
space:
mode:
Diffstat (limited to 'llvm/test/CodeGen/X86')
-rw-r--r--llvm/test/CodeGen/X86/avx512-ext.ll (renamed from llvm/test/CodeGen/X86/avx512-trunc-ext.ll)26
-rw-r--r--llvm/test/CodeGen/X86/avx512-intrinsics.ll390
-rw-r--r--llvm/test/CodeGen/X86/avx512-trunc.ll364
-rw-r--r--llvm/test/CodeGen/X86/avx512bw-intrinsics.ll78
-rw-r--r--llvm/test/CodeGen/X86/avx512bwvl-intrinsics.ll156
-rw-r--r--llvm/test/CodeGen/X86/avx512vl-intrinsics.ll780
-rw-r--r--llvm/test/CodeGen/X86/masked_memop.ll11
7 files changed, 1776 insertions, 29 deletions
diff --git a/llvm/test/CodeGen/X86/avx512-trunc-ext.ll b/llvm/test/CodeGen/X86/avx512-ext.ll
index f25458972e4..aa1dd4928c3 100644
--- a/llvm/test/CodeGen/X86/avx512-trunc-ext.ll
+++ b/llvm/test/CodeGen/X86/avx512-ext.ll
@@ -1,24 +1,7 @@
; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=knl | FileCheck %s --check-prefix=KNL
; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=skx | FileCheck %s --check-prefix=SKX
-
-; KNL-LABEL: trunc_16x32_to_16x8
-; KNL: vpmovdb
-; KNL: ret
-define <16 x i8> @trunc_16x32_to_16x8(<16 x i32> %i) nounwind readnone {
- %x = trunc <16 x i32> %i to <16 x i8>
- ret <16 x i8> %x
-}
-
-; KNL-LABEL: trunc_8x64_to_8x16
-; KNL: vpmovqw
-; KNL: ret
-define <8 x i16> @trunc_8x64_to_8x16(<8 x i64> %i) nounwind readnone {
- %x = trunc <8 x i64> %i to <8 x i16>
- ret <8 x i16> %x
-}
-
-;SKX-LABEL: zext_8x8mem_to_8x16:
+ ;SKX-LABEL: zext_8x8mem_to_8x16:
;SKX: ## BB#0:
;SKX-NEXT: vpmovw2m %xmm0, %k1
;SKX-NEXT: vpmovzxbw (%rdi), %xmm0 {%k1} {z}
@@ -895,13 +878,6 @@ define <8 x i32> @sext_8i1_8i32(<8 x i32> %a1, <8 x i32> %a2) nounwind {
ret <8 x i32> %y
}
-; KNL-LABEL: trunc_v16i32_to_v16i16
-; KNL: vpmovdw
-; KNL: ret
-define <16 x i16> @trunc_v16i32_to_v16i16(<16 x i32> %x) {
- %1 = trunc <16 x i32> %x to <16 x i16>
- ret <16 x i16> %1
-}
; KNL-LABEL: trunc_i32_to_i1
; KNL: movw $-4, %ax
diff --git a/llvm/test/CodeGen/X86/avx512-intrinsics.ll b/llvm/test/CodeGen/X86/avx512-intrinsics.ll
index 6e50fda7467..7c30063ce28 100644
--- a/llvm/test/CodeGen/X86/avx512-intrinsics.ll
+++ b/llvm/test/CodeGen/X86/avx512-intrinsics.ll
@@ -3119,6 +3119,396 @@ define <16 x float>@test_int_x86_avx512_mask_scalef_ps_512(<16 x float> %x0, <16
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) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmov_qb_512:
+; CHECK: vpmovqb %zmm0, %xmm1 {%k1}
+; CHECK-NEXT: vpmovqb %zmm0, %xmm2 {%k1} {z}
+; CHECK-NEXT: vpmovqb %zmm0, %xmm0
+ %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) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmov_qb_mem_512:
+; CHECK: vpmovqb %zmm0, (%rdi)
+; CHECK: vpmovqb %zmm0, (%rdi) {%k1}
+ 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) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovs_qb_512:
+; CHECK: vpmovsqb %zmm0, %xmm1 {%k1}
+; CHECK-NEXT: vpmovsqb %zmm0, %xmm2 {%k1} {z}
+; CHECK-NEXT: vpmovsqb %zmm0, %xmm0
+ %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) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovs_qb_mem_512:
+; CHECK: vpmovsqb %zmm0, (%rdi)
+; CHECK: vpmovsqb %zmm0, (%rdi) {%k1}
+ 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) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovus_qb_512:
+; CHECK: vpmovusqb %zmm0, %xmm1 {%k1}
+; CHECK-NEXT: vpmovusqb %zmm0, %xmm2 {%k1} {z}
+; CHECK-NEXT: vpmovusqb %zmm0, %xmm0
+ %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) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovus_qb_mem_512:
+; CHECK: vpmovusqb %zmm0, (%rdi)
+; CHECK: vpmovusqb %zmm0, (%rdi) {%k1}
+ 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) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmov_qw_512:
+; CHECK: vpmovqw %zmm0, %xmm1 {%k1}
+; CHECK-NEXT: vpmovqw %zmm0, %xmm2 {%k1} {z}
+; CHECK-NEXT: vpmovqw %zmm0, %xmm0
+ %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) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmov_qw_mem_512:
+; CHECK: vpmovqw %zmm0, (%rdi)
+; CHECK: vpmovqw %zmm0, (%rdi) {%k1}
+ 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) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovs_qw_512:
+; CHECK: vpmovsqw %zmm0, %xmm1 {%k1}
+; CHECK-NEXT: vpmovsqw %zmm0, %xmm2 {%k1} {z}
+; CHECK-NEXT: vpmovsqw %zmm0, %xmm0
+ %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) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovs_qw_mem_512:
+; CHECK: vpmovsqw %zmm0, (%rdi)
+; CHECK: vpmovsqw %zmm0, (%rdi) {%k1}
+ 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) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovus_qw_512:
+; CHECK: vpmovusqw %zmm0, %xmm1 {%k1}
+; CHECK-NEXT: vpmovusqw %zmm0, %xmm2 {%k1} {z}
+; CHECK-NEXT: vpmovusqw %zmm0, %xmm0
+ %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) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovus_qw_mem_512:
+; CHECK: vpmovusqw %zmm0, (%rdi)
+; CHECK: vpmovusqw %zmm0, (%rdi) {%k1}
+ 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
+}
+
+declare <8 x i32> @llvm.x86.avx512.mask.pmov.qd.512(<8 x i64>, <8 x i32>, i8)
+
+define <8 x i32>@test_int_x86_avx512_mask_pmov_qd_512(<8 x i64> %x0, <8 x i32> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmov_qd_512:
+; CHECK: vpmovqd %zmm0, %ymm1 {%k1}
+; CHECK-NEXT: vpmovqd %zmm0, %ymm2 {%k1} {z}
+; CHECK-NEXT: vpmovqd %zmm0, %ymm0
+ %res0 = call <8 x i32> @llvm.x86.avx512.mask.pmov.qd.512(<8 x i64> %x0, <8 x i32> %x1, i8 -1)
+ %res1 = call <8 x i32> @llvm.x86.avx512.mask.pmov.qd.512(<8 x i64> %x0, <8 x i32> %x1, i8 %x2)
+ %res2 = call <8 x i32> @llvm.x86.avx512.mask.pmov.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.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) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmov_qd_mem_512:
+; CHECK: vpmovqd %zmm0, (%rdi)
+; CHECK: vpmovqd %zmm0, (%rdi) {%k1}
+ 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) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovs_qd_512:
+; CHECK: vpmovsqd %zmm0, %ymm1 {%k1}
+; CHECK-NEXT: vpmovsqd %zmm0, %ymm2 {%k1} {z}
+; CHECK-NEXT: vpmovsqd %zmm0, %ymm0
+ %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) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovs_qd_mem_512:
+; CHECK: vpmovsqd %zmm0, (%rdi)
+; CHECK: vpmovsqd %zmm0, (%rdi) {%k1}
+ 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) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovus_qd_512:
+; CHECK: vpmovusqd %zmm0, %ymm1 {%k1}
+; CHECK-NEXT: vpmovusqd %zmm0, %ymm2 {%k1} {z}
+; CHECK-NEXT: vpmovusqd %zmm0, %ymm0
+ %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) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovus_qd_mem_512:
+; CHECK: vpmovusqd %zmm0, (%rdi)
+; CHECK: vpmovusqd %zmm0, (%rdi) {%k1}
+ 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) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmov_db_512:
+; CHECK: vpmovdb %zmm0, %xmm1 {%k1}
+; CHECK-NEXT: vpmovdb %zmm0, %xmm2 {%k1} {z}
+; CHECK-NEXT: vpmovdb %zmm0, %xmm0
+ %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) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmov_db_mem_512:
+; CHECK: vpmovdb %zmm0, (%rdi)
+; CHECK: vpmovdb %zmm0, (%rdi) {%k1}
+ 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) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovs_db_512:
+; CHECK: vpmovsdb %zmm0, %xmm1 {%k1}
+; CHECK-NEXT: vpmovsdb %zmm0, %xmm2 {%k1} {z}
+; CHECK-NEXT: vpmovsdb %zmm0, %xmm0
+ %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) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovs_db_mem_512:
+; CHECK: vpmovsdb %zmm0, (%rdi)
+; CHECK: vpmovsdb %zmm0, (%rdi) {%k1}
+ 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) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovus_db_512:
+; CHECK: vpmovusdb %zmm0, %xmm1 {%k1}
+; CHECK-NEXT: vpmovusdb %zmm0, %xmm2 {%k1} {z}
+; CHECK-NEXT: vpmovusdb %zmm0, %xmm0
+ %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) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovus_db_mem_512:
+; CHECK: vpmovusdb %zmm0, (%rdi)
+; CHECK: vpmovusdb %zmm0, (%rdi) {%k1}
+ 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) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmov_dw_512:
+; CHECK: vpmovdw %zmm0, %ymm1 {%k1}
+; CHECK-NEXT: vpmovdw %zmm0, %ymm2 {%k1} {z}
+; CHECK-NEXT: vpmovdw %zmm0, %ymm0
+ %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) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmov_dw_mem_512:
+; CHECK: vpmovdw %zmm0, (%rdi)
+; CHECK: vpmovdw %zmm0, (%rdi) {%k1}
+ 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) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovs_dw_512:
+; CHECK: vpmovsdw %zmm0, %ymm1 {%k1}
+; CHECK-NEXT: vpmovsdw %zmm0, %ymm2 {%k1} {z}
+; CHECK-NEXT: vpmovsdw %zmm0, %ymm0
+ %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) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovs_dw_mem_512:
+; CHECK: vpmovsdw %zmm0, (%rdi)
+; CHECK: vpmovsdw %zmm0, (%rdi) {%k1}
+ 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) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovus_dw_512:
+; CHECK: vpmovusdw %zmm0, %ymm1 {%k1}
+; CHECK-NEXT: vpmovusdw %zmm0, %ymm2 {%k1} {z}
+; CHECK-NEXT: vpmovusdw %zmm0, %ymm0
+ %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) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovus_dw_mem_512:
+; CHECK: vpmovusdw %zmm0, (%rdi)
+; CHECK: vpmovusdw %zmm0, (%rdi) {%k1}
+ 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 <8 x double> @llvm.x86.avx512.mask.cvtdq2pd.512(<8 x i32>, <8 x double>, i8)
define <8 x double>@test_int_x86_avx512_mask_cvt_dq2pd_512(<8 x i32> %x0, <8 x double> %x1, i8 %x2) {
diff --git a/llvm/test/CodeGen/X86/avx512-trunc.ll b/llvm/test/CodeGen/X86/avx512-trunc.ll
new file mode 100644
index 00000000000..9205feda7eb
--- /dev/null
+++ b/llvm/test/CodeGen/X86/avx512-trunc.ll
@@ -0,0 +1,364 @@
+; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=knl | FileCheck %s --check-prefix=KNL
+; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=skx | FileCheck %s --check-prefix=SKX
+
+ attributes #0 = { nounwind }
+
+; KNL-LABEL: trunc_16x32_to_16x8
+; KNL: vpmovdb
+; KNL: ret
+define <16 x i8> @trunc_16x32_to_16x8(<16 x i32> %i) #0 {
+ %x = trunc <16 x i32> %i to <16 x i8>
+ ret <16 x i8> %x
+}
+
+; KNL-LABEL: trunc_8x64_to_8x16
+; KNL: vpmovqw
+; KNL: ret
+define <8 x i16> @trunc_8x64_to_8x16(<8 x i64> %i) #0 {
+ %x = trunc <8 x i64> %i to <8 x i16>
+ ret <8 x i16> %x
+}
+
+; KNL-LABEL: trunc_v16i32_to_v16i16
+; KNL: vpmovdw
+; KNL: ret
+define <16 x i16> @trunc_v16i32_to_v16i16(<16 x i32> %x) #0 {
+ %1 = trunc <16 x i32> %x to <16 x i16>
+ ret <16 x i16> %1
+}
+
+define <8 x i8> @trunc_qb_512(<8 x i64> %i) #0 {
+; SKX-LABEL: trunc_qb_512:
+; SKX: ## BB#0:
+; SKX-NEXT: vpmovqw %zmm0, %xmm0
+; SKX-NEXT: retq
+ %x = trunc <8 x i64> %i to <8 x i8>
+ ret <8 x i8> %x
+}
+
+define void @trunc_qb_512_mem(<8 x i64> %i, <8 x i8>* %res) #0 {
+; SKX-LABEL: trunc_qb_512_mem:
+; SKX: ## BB#0:
+; SKX-NEXT: vpmovqb %zmm0, (%rdi)
+; SKX-NEXT: retq
+ %x = trunc <8 x i64> %i to <8 x i8>
+ store <8 x i8> %x, <8 x i8>* %res
+ ret void
+}
+
+define <4 x i8> @trunc_qb_256(<4 x i64> %i) #0 {
+; SKX-LABEL: trunc_qb_256:
+; SKX: ## BB#0:
+; SKX-NEXT: vpmovqd %ymm0, %xmm0
+; SKX-NEXT: retq
+ %x = trunc <4 x i64> %i to <4 x i8>
+ ret <4 x i8> %x
+}
+
+define void @trunc_qb_256_mem(<4 x i64> %i, <4 x i8>* %res) #0 {
+; SKX-LABEL: trunc_qb_256_mem:
+; SKX: ## BB#0:
+; SKX-NEXT: vpmovqb %ymm0, (%rdi)
+; SKX-NEXT: retq
+ %x = trunc <4 x i64> %i to <4 x i8>
+ store <4 x i8> %x, <4 x i8>* %res
+ ret void
+}
+
+define <2 x i8> @trunc_qb_128(<2 x i64> %i) #0 {
+; SKX-LABEL: trunc_qb_128:
+; SKX: ## BB#0:
+; SKX-NEXT: retq
+ %x = trunc <2 x i64> %i to <2 x i8>
+ ret <2 x i8> %x
+}
+
+define void @trunc_qb_128_mem(<2 x i64> %i, <2 x i8>* %res) #0 {
+; SKX-LABEL: trunc_qb_128_mem:
+; SKX: ## BB#0:
+; SKX-NEXT: vpmovqb %xmm0, (%rdi)
+; SKX-NEXT: retq
+ %x = trunc <2 x i64> %i to <2 x i8>
+ store <2 x i8> %x, <2 x i8>* %res
+ ret void
+}
+
+define <8 x i16> @trunc_qw_512(<8 x i64> %i) #0 {
+; SKX-LABEL: trunc_qw_512:
+; SKX: ## BB#0:
+; SKX-NEXT: vpmovqw %zmm0, %xmm0
+; SKX-NEXT: retq
+ %x = trunc <8 x i64> %i to <8 x i16>
+ ret <8 x i16> %x
+}
+
+define void @trunc_qw_512_mem(<8 x i64> %i, <8 x i16>* %res) #0 {
+; SKX-LABEL: trunc_qw_512_mem:
+; SKX: ## BB#0:
+; SKX-NEXT: vpmovqw %zmm0, (%rdi)
+; SKX-NEXT: retq
+ %x = trunc <8 x i64> %i to <8 x i16>
+ store <8 x i16> %x, <8 x i16>* %res
+ ret void
+}
+
+define <4 x i16> @trunc_qw_256(<4 x i64> %i) #0 {
+; SKX-LABEL: trunc_qw_256:
+; SKX: ## BB#0:
+; SKX-NEXT: vpmovqd %ymm0, %xmm0
+; SKX-NEXT: retq
+ %x = trunc <4 x i64> %i to <4 x i16>
+ ret <4 x i16> %x
+}
+
+define void @trunc_qw_256_mem(<4 x i64> %i, <4 x i16>* %res) #0 {
+; SKX-LABEL: trunc_qw_256_mem:
+; SKX: ## BB#0:
+; SKX-NEXT: vpmovqw %ymm0, (%rdi)
+; SKX-NEXT: retq
+ %x = trunc <4 x i64> %i to <4 x i16>
+ store <4 x i16> %x, <4 x i16>* %res
+ ret void
+}
+
+define <2 x i16> @trunc_qw_128(<2 x i64> %i) #0 {
+; SKX-LABEL: trunc_qw_128:
+; SKX: ## BB#0:
+; SKX-NEXT: retq
+ %x = trunc <2 x i64> %i to <2 x i16>
+ ret <2 x i16> %x
+}
+
+define void @trunc_qw_128_mem(<2 x i64> %i, <2 x i16>* %res) #0 {
+; SKX-LABEL: trunc_qw_128_mem:
+; SKX: ## BB#0:
+; SKX-NEXT: vpmovqw %xmm0, (%rdi)
+; SKX-NEXT: retq
+ %x = trunc <2 x i64> %i to <2 x i16>
+ store <2 x i16> %x, <2 x i16>* %res
+ ret void
+}
+
+define <8 x i32> @trunc_qd_512(<8 x i64> %i) #0 {
+; SKX-LABEL: trunc_qd_512:
+; SKX: ## BB#0:
+; SKX-NEXT: vpmovqd %zmm0, %ymm0
+; SKX-NEXT: retq
+ %x = trunc <8 x i64> %i to <8 x i32>
+ ret <8 x i32> %x
+}
+
+define void @trunc_qd_512_mem(<8 x i64> %i, <8 x i32>* %res) #0 {
+; SKX-LABEL: trunc_qd_512_mem:
+; SKX: ## BB#0:
+; SKX-NEXT: vpmovqd %zmm0, (%rdi)
+; SKX-NEXT: retq
+ %x = trunc <8 x i64> %i to <8 x i32>
+ store <8 x i32> %x, <8 x i32>* %res
+ ret void
+}
+
+define <4 x i32> @trunc_qd_256(<4 x i64> %i) #0 {
+; SKX-LABEL: trunc_qd_256:
+; SKX: ## BB#0:
+; SKX-NEXT: vpmovqd %ymm0, %xmm0
+; SKX-NEXT: retq
+ %x = trunc <4 x i64> %i to <4 x i32>
+ ret <4 x i32> %x
+}
+
+define void @trunc_qd_256_mem(<4 x i64> %i, <4 x i32>* %res) #0 {
+; SKX-LABEL: trunc_qd_256_mem:
+; SKX: ## BB#0:
+; SKX-NEXT: vpmovqd %ymm0, (%rdi)
+; SKX-NEXT: retq
+ %x = trunc <4 x i64> %i to <4 x i32>
+ store <4 x i32> %x, <4 x i32>* %res
+ ret void
+}
+
+define <2 x i32> @trunc_qd_128(<2 x i64> %i) #0 {
+; SKX-LABEL: trunc_qd_128:
+; SKX: ## BB#0:
+; SKX-NEXT: retq
+ %x = trunc <2 x i64> %i to <2 x i32>
+ ret <2 x i32> %x
+}
+
+define void @trunc_qd_128_mem(<2 x i64> %i, <2 x i32>* %res) #0 {
+; SKX-LABEL: trunc_qd_128_mem:
+; SKX: ## BB#0:
+; SKX-NEXT: vpmovqd %xmm0, (%rdi)
+; SKX-NEXT: retq
+ %x = trunc <2 x i64> %i to <2 x i32>
+ store <2 x i32> %x, <2 x i32>* %res
+ ret void
+}
+
+define <16 x i8> @trunc_db_512(<16 x i32> %i) #0 {
+; SKX-LABEL: trunc_db_512:
+; SKX: ## BB#0:
+; SKX-NEXT: vpmovdb %zmm0, %xmm0
+; SKX-NEXT: retq
+ %x = trunc <16 x i32> %i to <16 x i8>
+ ret <16 x i8> %x
+}
+
+define void @trunc_db_512_mem(<16 x i32> %i, <16 x i8>* %res) #0 {
+; SKX-LABEL: trunc_db_512_mem:
+; SKX: ## BB#0:
+; SKX-NEXT: vpmovdb %zmm0, (%rdi)
+; SKX-NEXT: retq
+ %x = trunc <16 x i32> %i to <16 x i8>
+ store <16 x i8> %x, <16 x i8>* %res
+ ret void
+}
+
+define <8 x i8> @trunc_db_256(<8 x i32> %i) #0 {
+; SKX-LABEL: trunc_db_256:
+; SKX: ## BB#0:
+; SKX-NEXT: vpmovdw %ymm0, %xmm0
+; SKX-NEXT: retq
+ %x = trunc <8 x i32> %i to <8 x i8>
+ ret <8 x i8> %x
+}
+
+define void @trunc_db_256_mem(<8 x i32> %i, <8 x i8>* %res) #0 {
+; SKX-LABEL: trunc_db_256_mem:
+; SKX: ## BB#0:
+; SKX-NEXT: vpmovdb %ymm0, (%rdi)
+; SKX-NEXT: retq
+ %x = trunc <8 x i32> %i to <8 x i8>
+ store <8 x i8> %x, <8 x i8>* %res
+ ret void
+}
+
+define <4 x i8> @trunc_db_128(<4 x i32> %i) #0 {
+; SKX-LABEL: trunc_db_128:
+; SKX: ## BB#0:
+; SKX-NEXT: retq
+ %x = trunc <4 x i32> %i to <4 x i8>
+ ret <4 x i8> %x
+}
+
+define void @trunc_db_128_mem(<4 x i32> %i, <4 x i8>* %res) #0 {
+; SKX-LABEL: trunc_db_128_mem:
+; SKX: ## BB#0:
+; SKX-NEXT: vpmovdb %xmm0, (%rdi)
+; SKX-NEXT: retq
+ %x = trunc <4 x i32> %i to <4 x i8>
+ store <4 x i8> %x, <4 x i8>* %res
+ ret void
+}
+
+define <16 x i16> @trunc_dw_512(<16 x i32> %i) #0 {
+; SKX-LABEL: trunc_dw_512:
+; SKX: ## BB#0:
+; SKX-NEXT: vpmovdw %zmm0, %ymm0
+; SKX-NEXT: retq
+ %x = trunc <16 x i32> %i to <16 x i16>
+ ret <16 x i16> %x
+}
+
+define void @trunc_dw_512_mem(<16 x i32> %i, <16 x i16>* %res) #0 {
+; SKX-LABEL: trunc_dw_512_mem:
+; SKX: ## BB#0:
+; SKX-NEXT: vpmovdw %zmm0, (%rdi)
+; SKX-NEXT: retq
+ %x = trunc <16 x i32> %i to <16 x i16>
+ store <16 x i16> %x, <16 x i16>* %res
+ ret void
+}
+
+define <8 x i16> @trunc_dw_256(<8 x i32> %i) #0 {
+; SKX-LABEL: trunc_dw_256:
+; SKX: ## BB#0:
+; SKX-NEXT: vpmovdw %ymm0, %xmm0
+; SKX-NEXT: retq
+ %x = trunc <8 x i32> %i to <8 x i16>
+ ret <8 x i16> %x
+}
+
+define void @trunc_dw_256_mem(<8 x i32> %i, <8 x i16>* %res) #0 {
+; SKX-LABEL: trunc_dw_256_mem:
+; SKX: ## BB#0:
+; SKX-NEXT: vpmovdw %ymm0, (%rdi)
+; SKX-NEXT: retq
+ %x = trunc <8 x i32> %i to <8 x i16>
+ store <8 x i16> %x, <8 x i16>* %res
+ ret void
+}
+
+define <4 x i16> @trunc_dw_128(<4 x i32> %i) #0 {
+; SKX-LABEL: trunc_dw_128:
+; SKX: ## BB#0:
+; SKX-NEXT: retq
+ %x = trunc <4 x i32> %i to <4 x i16>
+ ret <4 x i16> %x
+}
+
+define void @trunc_dw_128_mem(<4 x i32> %i, <4 x i16>* %res) #0 {
+; SKX-LABEL: trunc_dw_128_mem:
+; SKX: ## BB#0:
+; SKX-NEXT: vpmovdw %xmm0, (%rdi)
+; SKX-NEXT: retq
+ %x = trunc <4 x i32> %i to <4 x i16>
+ store <4 x i16> %x, <4 x i16>* %res
+ ret void
+}
+
+define <32 x i8> @trunc_wb_512(<32 x i16> %i) #0 {
+; SKX-LABEL: trunc_wb_512:
+; SKX: ## BB#0:
+; SKX-NEXT: vpmovwb %zmm0, %ymm0
+; SKX-NEXT: retq
+ %x = trunc <32 x i16> %i to <32 x i8>
+ ret <32 x i8> %x
+}
+
+define void @trunc_wb_512_mem(<32 x i16> %i, <32 x i8>* %res) #0 {
+; SKX-LABEL: trunc_wb_512_mem:
+; SKX: ## BB#0:
+; SKX-NEXT: vpmovwb %zmm0, (%rdi)
+; SKX-NEXT: retq
+ %x = trunc <32 x i16> %i to <32 x i8>
+ store <32 x i8> %x, <32 x i8>* %res
+ ret void
+}
+
+define <16 x i8> @trunc_wb_256(<16 x i16> %i) #0 {
+; SKX-LABEL: trunc_wb_256:
+; SKX: ## BB#0:
+; SKX-NEXT: vpmovwb %ymm0, %xmm0
+; SKX-NEXT: retq
+ %x = trunc <16 x i16> %i to <16 x i8>
+ ret <16 x i8> %x
+}
+
+define void @trunc_wb_256_mem(<16 x i16> %i, <16 x i8>* %res) #0 {
+; SKX-LABEL: trunc_wb_256_mem:
+; SKX: ## BB#0:
+; SKX-NEXT: vpmovwb %ymm0, (%rdi)
+; SKX-NEXT: retq
+ %x = trunc <16 x i16> %i to <16 x i8>
+ store <16 x i8> %x, <16 x i8>* %res
+ ret void
+}
+
+define <8 x i8> @trunc_wb_128(<8 x i16> %i) #0 {
+; SKX-LABEL: trunc_wb_128:
+; SKX: ## BB#0:
+; SKX-NEXT: retq
+ %x = trunc <8 x i16> %i to <8 x i8>
+ ret <8 x i8> %x
+}
+
+define void @trunc_wb_128_mem(<8 x i16> %i, <8 x i8>* %res) #0 {
+; SKX-LABEL: trunc_wb_128_mem:
+; SKX: ## BB#0:
+; SKX-NEXT: vpmovwb %xmm0, (%rdi)
+; SKX-NEXT: retq
+ %x = trunc <8 x i16> %i to <8 x i8>
+ store <8 x i8> %x, <8 x i8>* %res
+ ret void
+}
diff --git a/llvm/test/CodeGen/X86/avx512bw-intrinsics.ll b/llvm/test/CodeGen/X86/avx512bw-intrinsics.ll
index b2b417df2f1..5ad28ab5ab5 100644
--- a/llvm/test/CodeGen/X86/avx512bw-intrinsics.ll
+++ b/llvm/test/CodeGen/X86/avx512bw-intrinsics.ll
@@ -1008,6 +1008,84 @@ define <32 x i16>@test_int_x86_avx512_mask_pmulhr_sw_512(<32 x i16> %x0, <32 x i
ret <32 x i16> %res2
}
+declare <32 x i8> @llvm.x86.avx512.mask.pmov.wb.512(<32 x i16>, <32 x i8>, i32)
+
+define <32 x i8>@test_int_x86_avx512_mask_pmov_wb_512(<32 x i16> %x0, <32 x i8> %x1, i32 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmov_wb_512:
+; CHECK: vpmovwb %zmm0, %ymm1 {%k1}
+; CHECK-NEXT: vpmovwb %zmm0, %ymm2 {%k1} {z}
+; CHECK-NEXT: vpmovwb %zmm0, %ymm0
+ %res0 = call <32 x i8> @llvm.x86.avx512.mask.pmov.wb.512(<32 x i16> %x0, <32 x i8> %x1, i32 -1)
+ %res1 = call <32 x i8> @llvm.x86.avx512.mask.pmov.wb.512(<32 x i16> %x0, <32 x i8> %x1, i32 %x2)
+ %res2 = call <32 x i8> @llvm.x86.avx512.mask.pmov.wb.512(<32 x i16> %x0, <32 x i8> zeroinitializer, i32 %x2)
+ %res3 = add <32 x i8> %res0, %res1
+ %res4 = add <32 x i8> %res3, %res2
+ ret <32 x i8> %res4
+}
+
+declare void @llvm.x86.avx512.mask.pmov.wb.mem.512(i8* %ptr, <32 x i16>, i32)
+
+define void @test_int_x86_avx512_mask_pmov_wb_mem_512(i8* %ptr, <32 x i16> %x1, i32 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmov_wb_mem_512:
+; CHECK: vpmovwb %zmm0, (%rdi)
+; CHECK: vpmovwb %zmm0, (%rdi) {%k1}
+ call void @llvm.x86.avx512.mask.pmov.wb.mem.512(i8* %ptr, <32 x i16> %x1, i32 -1)
+ call void @llvm.x86.avx512.mask.pmov.wb.mem.512(i8* %ptr, <32 x i16> %x1, i32 %x2)
+ ret void
+}
+
+declare <32 x i8> @llvm.x86.avx512.mask.pmovs.wb.512(<32 x i16>, <32 x i8>, i32)
+
+define <32 x i8>@test_int_x86_avx512_mask_pmovs_wb_512(<32 x i16> %x0, <32 x i8> %x1, i32 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovs_wb_512:
+; CHECK: vpmovswb %zmm0, %ymm1 {%k1}
+; CHECK-NEXT: vpmovswb %zmm0, %ymm2 {%k1} {z}
+; CHECK-NEXT: vpmovswb %zmm0, %ymm0
+ %res0 = call <32 x i8> @llvm.x86.avx512.mask.pmovs.wb.512(<32 x i16> %x0, <32 x i8> %x1, i32 -1)
+ %res1 = call <32 x i8> @llvm.x86.avx512.mask.pmovs.wb.512(<32 x i16> %x0, <32 x i8> %x1, i32 %x2)
+ %res2 = call <32 x i8> @llvm.x86.avx512.mask.pmovs.wb.512(<32 x i16> %x0, <32 x i8> zeroinitializer, i32 %x2)
+ %res3 = add <32 x i8> %res0, %res1
+ %res4 = add <32 x i8> %res3, %res2
+ ret <32 x i8> %res4
+}
+
+declare void @llvm.x86.avx512.mask.pmovs.wb.mem.512(i8* %ptr, <32 x i16>, i32)
+
+define void @test_int_x86_avx512_mask_pmovs_wb_mem_512(i8* %ptr, <32 x i16> %x1, i32 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovs_wb_mem_512:
+; CHECK: vpmovswb %zmm0, (%rdi)
+; CHECK: vpmovswb %zmm0, (%rdi) {%k1}
+ call void @llvm.x86.avx512.mask.pmovs.wb.mem.512(i8* %ptr, <32 x i16> %x1, i32 -1)
+ call void @llvm.x86.avx512.mask.pmovs.wb.mem.512(i8* %ptr, <32 x i16> %x1, i32 %x2)
+ ret void
+}
+
+declare <32 x i8> @llvm.x86.avx512.mask.pmovus.wb.512(<32 x i16>, <32 x i8>, i32)
+
+define <32 x i8>@test_int_x86_avx512_mask_pmovus_wb_512(<32 x i16> %x0, <32 x i8> %x1, i32 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovus_wb_512:
+; CHECK: vpmovuswb %zmm0, %ymm1 {%k1}
+; CHECK-NEXT: vpmovuswb %zmm0, %ymm2 {%k1} {z}
+; CHECK-NEXT: vpmovuswb %zmm0, %ymm0
+ %res0 = call <32 x i8> @llvm.x86.avx512.mask.pmovus.wb.512(<32 x i16> %x0, <32 x i8> %x1, i32 -1)
+ %res1 = call <32 x i8> @llvm.x86.avx512.mask.pmovus.wb.512(<32 x i16> %x0, <32 x i8> %x1, i32 %x2)
+ %res2 = call <32 x i8> @llvm.x86.avx512.mask.pmovus.wb.512(<32 x i16> %x0, <32 x i8> zeroinitializer, i32 %x2)
+ %res3 = add <32 x i8> %res0, %res1
+ %res4 = add <32 x i8> %res3, %res2
+ ret <32 x i8> %res4
+}
+
+declare void @llvm.x86.avx512.mask.pmovus.wb.mem.512(i8* %ptr, <32 x i16>, i32)
+
+define void @test_int_x86_avx512_mask_pmovus_wb_mem_512(i8* %ptr, <32 x i16> %x1, i32 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovus_wb_mem_512:
+; CHECK: vpmovuswb %zmm0, (%rdi)
+; CHECK: vpmovuswb %zmm0, (%rdi) {%k1}
+ call void @llvm.x86.avx512.mask.pmovus.wb.mem.512(i8* %ptr, <32 x i16> %x1, i32 -1)
+ call void @llvm.x86.avx512.mask.pmovus.wb.mem.512(i8* %ptr, <32 x i16> %x1, i32 %x2)
+ ret void
+}
+
declare <32 x i16> @llvm.x86.avx512.mask.pmaddubs.w.512(<64 x i8>, <64 x i8>, <32 x i16>, i32)
define <32 x i16>@test_int_x86_avx512_mask_pmaddubs_w_512(<64 x i8> %x0, <64 x i8> %x1, <32 x i16> %x2, i32 %x3) {
diff --git a/llvm/test/CodeGen/X86/avx512bwvl-intrinsics.ll b/llvm/test/CodeGen/X86/avx512bwvl-intrinsics.ll
index 2373dc089ae..ee76ae2a8a3 100644
--- a/llvm/test/CodeGen/X86/avx512bwvl-intrinsics.ll
+++ b/llvm/test/CodeGen/X86/avx512bwvl-intrinsics.ll
@@ -3876,6 +3876,162 @@ define <16 x i16>@test_int_x86_avx512_mask_pmulhr_sw_256(<16 x i16> %x0, <16 x i
ret <16 x i16> %res2
}
+declare <16 x i8> @llvm.x86.avx512.mask.pmov.wb.128(<8 x i16>, <16 x i8>, i8)
+
+define <16 x i8>@test_int_x86_avx512_mask_pmov_wb_128(<8 x i16> %x0, <16 x i8> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmov_wb_128:
+; CHECK: vpmovwb %xmm0, %xmm1 {%k1}
+; CHECK-NEXT: vpmovwb %xmm0, %xmm2 {%k1} {z}
+; CHECK-NEXT: vpmovwb %xmm0, %xmm0
+ %res0 = call <16 x i8> @llvm.x86.avx512.mask.pmov.wb.128(<8 x i16> %x0, <16 x i8> %x1, i8 -1)
+ %res1 = call <16 x i8> @llvm.x86.avx512.mask.pmov.wb.128(<8 x i16> %x0, <16 x i8> %x1, i8 %x2)
+ %res2 = call <16 x i8> @llvm.x86.avx512.mask.pmov.wb.128(<8 x i16> %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.wb.mem.128(i8* %ptr, <8 x i16>, i8)
+
+define void @test_int_x86_avx512_mask_pmov_wb_mem_128(i8* %ptr, <8 x i16> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmov_wb_mem_128:
+; CHECK: vpmovwb %xmm0, (%rdi)
+; CHECK: vpmovwb %xmm0, (%rdi) {%k1}
+ call void @llvm.x86.avx512.mask.pmov.wb.mem.128(i8* %ptr, <8 x i16> %x1, i8 -1)
+ call void @llvm.x86.avx512.mask.pmov.wb.mem.128(i8* %ptr, <8 x i16> %x1, i8 %x2)
+ ret void
+}
+
+declare <16 x i8> @llvm.x86.avx512.mask.pmovs.wb.128(<8 x i16>, <16 x i8>, i8)
+
+define <16 x i8>@test_int_x86_avx512_mask_pmovs_wb_128(<8 x i16> %x0, <16 x i8> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovs_wb_128:
+; CHECK: vpmovswb %xmm0, %xmm1 {%k1}
+; CHECK-NEXT: vpmovswb %xmm0, %xmm2 {%k1} {z}
+; CHECK-NEXT: vpmovswb %xmm0, %xmm0
+ %res0 = call <16 x i8> @llvm.x86.avx512.mask.pmovs.wb.128(<8 x i16> %x0, <16 x i8> %x1, i8 -1)
+ %res1 = call <16 x i8> @llvm.x86.avx512.mask.pmovs.wb.128(<8 x i16> %x0, <16 x i8> %x1, i8 %x2)
+ %res2 = call <16 x i8> @llvm.x86.avx512.mask.pmovs.wb.128(<8 x i16> %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.wb.mem.128(i8* %ptr, <8 x i16>, i8)
+
+define void @test_int_x86_avx512_mask_pmovs_wb_mem_128(i8* %ptr, <8 x i16> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovs_wb_mem_128:
+; CHECK: vpmovswb %xmm0, (%rdi)
+; CHECK: vpmovswb %xmm0, (%rdi) {%k1}
+ call void @llvm.x86.avx512.mask.pmovs.wb.mem.128(i8* %ptr, <8 x i16> %x1, i8 -1)
+ call void @llvm.x86.avx512.mask.pmovs.wb.mem.128(i8* %ptr, <8 x i16> %x1, i8 %x2)
+ ret void
+}
+
+declare <16 x i8> @llvm.x86.avx512.mask.pmovus.wb.128(<8 x i16>, <16 x i8>, i8)
+
+define <16 x i8>@test_int_x86_avx512_mask_pmovus_wb_128(<8 x i16> %x0, <16 x i8> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovus_wb_128:
+; CHECK: vpmovuswb %xmm0, %xmm1 {%k1}
+; CHECK-NEXT: vpmovuswb %xmm0, %xmm2 {%k1} {z}
+; CHECK-NEXT: vpmovuswb %xmm0, %xmm0
+ %res0 = call <16 x i8> @llvm.x86.avx512.mask.pmovus.wb.128(<8 x i16> %x0, <16 x i8> %x1, i8 -1)
+ %res1 = call <16 x i8> @llvm.x86.avx512.mask.pmovus.wb.128(<8 x i16> %x0, <16 x i8> %x1, i8 %x2)
+ %res2 = call <16 x i8> @llvm.x86.avx512.mask.pmovus.wb.128(<8 x i16> %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.wb.mem.128(i8* %ptr, <8 x i16>, i8)
+
+define void @test_int_x86_avx512_mask_pmovus_wb_mem_128(i8* %ptr, <8 x i16> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovus_wb_mem_128:
+; CHECK: vpmovuswb %xmm0, (%rdi)
+; CHECK: vpmovuswb %xmm0, (%rdi) {%k1}
+ call void @llvm.x86.avx512.mask.pmovus.wb.mem.128(i8* %ptr, <8 x i16> %x1, i8 -1)
+ call void @llvm.x86.avx512.mask.pmovus.wb.mem.128(i8* %ptr, <8 x i16> %x1, i8 %x2)
+ ret void
+}
+
+declare <16 x i8> @llvm.x86.avx512.mask.pmov.wb.256(<16 x i16>, <16 x i8>, i16)
+
+define <16 x i8>@test_int_x86_avx512_mask_pmov_wb_256(<16 x i16> %x0, <16 x i8> %x1, i16 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmov_wb_256:
+; CHECK: vpmovwb %ymm0, %xmm1 {%k1}
+; CHECK-NEXT: vpmovwb %ymm0, %xmm2 {%k1} {z}
+; CHECK-NEXT: vpmovwb %ymm0, %xmm0
+ %res0 = call <16 x i8> @llvm.x86.avx512.mask.pmov.wb.256(<16 x i16> %x0, <16 x i8> %x1, i16 -1)
+ %res1 = call <16 x i8> @llvm.x86.avx512.mask.pmov.wb.256(<16 x i16> %x0, <16 x i8> %x1, i16 %x2)
+ %res2 = call <16 x i8> @llvm.x86.avx512.mask.pmov.wb.256(<16 x i16> %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.wb.mem.256(i8* %ptr, <16 x i16>, i16)
+
+define void @test_int_x86_avx512_mask_pmov_wb_mem_256(i8* %ptr, <16 x i16> %x1, i16 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmov_wb_mem_256:
+; CHECK: vpmovwb %ymm0, (%rdi)
+; CHECK: vpmovwb %ymm0, (%rdi) {%k1}
+ call void @llvm.x86.avx512.mask.pmov.wb.mem.256(i8* %ptr, <16 x i16> %x1, i16 -1)
+ call void @llvm.x86.avx512.mask.pmov.wb.mem.256(i8* %ptr, <16 x i16> %x1, i16 %x2)
+ ret void
+}
+
+declare <16 x i8> @llvm.x86.avx512.mask.pmovs.wb.256(<16 x i16>, <16 x i8>, i16)
+
+define <16 x i8>@test_int_x86_avx512_mask_pmovs_wb_256(<16 x i16> %x0, <16 x i8> %x1, i16 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovs_wb_256:
+; CHECK: vpmovswb %ymm0, %xmm1 {%k1}
+; CHECK-NEXT: vpmovswb %ymm0, %xmm2 {%k1} {z}
+; CHECK-NEXT: vpmovswb %ymm0, %xmm0
+ %res0 = call <16 x i8> @llvm.x86.avx512.mask.pmovs.wb.256(<16 x i16> %x0, <16 x i8> %x1, i16 -1)
+ %res1 = call <16 x i8> @llvm.x86.avx512.mask.pmovs.wb.256(<16 x i16> %x0, <16 x i8> %x1, i16 %x2)
+ %res2 = call <16 x i8> @llvm.x86.avx512.mask.pmovs.wb.256(<16 x i16> %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.wb.mem.256(i8* %ptr, <16 x i16>, i16)
+
+define void @test_int_x86_avx512_mask_pmovs_wb_mem_256(i8* %ptr, <16 x i16> %x1, i16 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovs_wb_mem_256:
+; CHECK: vpmovswb %ymm0, (%rdi)
+; CHECK: vpmovswb %ymm0, (%rdi) {%k1}
+ call void @llvm.x86.avx512.mask.pmovs.wb.mem.256(i8* %ptr, <16 x i16> %x1, i16 -1)
+ call void @llvm.x86.avx512.mask.pmovs.wb.mem.256(i8* %ptr, <16 x i16> %x1, i16 %x2)
+ ret void
+}
+
+declare <16 x i8> @llvm.x86.avx512.mask.pmovus.wb.256(<16 x i16>, <16 x i8>, i16)
+
+define <16 x i8>@test_int_x86_avx512_mask_pmovus_wb_256(<16 x i16> %x0, <16 x i8> %x1, i16 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovus_wb_256:
+; CHECK: vpmovuswb %ymm0, %xmm1 {%k1}
+; CHECK-NEXT: vpmovuswb %ymm0, %xmm2 {%k1} {z}
+; CHECK-NEXT: vpmovuswb %ymm0, %xmm0
+ %res0 = call <16 x i8> @llvm.x86.avx512.mask.pmovus.wb.256(<16 x i16> %x0, <16 x i8> %x1, i16 -1)
+ %res1 = call <16 x i8> @llvm.x86.avx512.mask.pmovus.wb.256(<16 x i16> %x0, <16 x i8> %x1, i16 %x2)
+ %res2 = call <16 x i8> @llvm.x86.avx512.mask.pmovus.wb.256(<16 x i16> %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.wb.mem.256(i8* %ptr, <16 x i16>, i16)
+
+define void @test_int_x86_avx512_mask_pmovus_wb_mem_256(i8* %ptr, <16 x i16> %x1, i16 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovus_wb_mem_256:
+; CHECK: vpmovuswb %ymm0, (%rdi)
+; CHECK: vpmovuswb %ymm0, (%rdi) {%k1}
+ call void @llvm.x86.avx512.mask.pmovus.wb.mem.256(i8* %ptr, <16 x i16> %x1, i16 -1)
+ call void @llvm.x86.avx512.mask.pmovus.wb.mem.256(i8* %ptr, <16 x i16> %x1, i16 %x2)
+ ret void
+}
+
declare <4 x i32> @llvm.x86.avx512.mask.pmaddw.d.128(<8 x i16>, <8 x i16>, <4 x i32>, i8)
define <4 x i32>@test_int_x86_avx512_mask_pmaddw_d_128(<8 x i16> %x0, <8 x i16> %x1, <4 x i32> %x2, i8 %x3) {
diff --git a/llvm/test/CodeGen/X86/avx512vl-intrinsics.ll b/llvm/test/CodeGen/X86/avx512vl-intrinsics.ll
index 46ee51f47b6..7812148de1c 100644
--- a/llvm/test/CodeGen/X86/avx512vl-intrinsics.ll
+++ b/llvm/test/CodeGen/X86/avx512vl-intrinsics.ll
@@ -3005,6 +3005,786 @@ define <8 x float>@test_int_x86_avx512_mask_scalef_ps_256(<8 x float> %x0, <8 x
ret <8 x float> %res2
}
+declare <16 x i8> @llvm.x86.avx512.mask.pmov.qb.128(<2 x i64>, <16 x i8>, i8)
+
+define <16 x i8>@test_int_x86_avx512_mask_pmov_qb_128(<2 x i64> %x0, <16 x i8> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmov_qb_128:
+; CHECK: vpmovqb %xmm0, %xmm1 {%k1}
+; CHECK-NEXT: vpmovqb %xmm0, %xmm2 {%k1} {z}
+; CHECK-NEXT: vpmovqb %xmm0, %xmm0
+ %res0 = call <16 x i8> @llvm.x86.avx512.mask.pmov.qb.128(<2 x i64> %x0, <16 x i8> %x1, i8 -1)
+ %res1 = call <16 x i8> @llvm.x86.avx512.mask.pmov.qb.128(<2 x i64> %x0, <16 x i8> %x1, i8 %x2)
+ %res2 = call <16 x i8> @llvm.x86.avx512.mask.pmov.qb.128(<2 x i64> %x0, <16 x i8> zeroinitializer, i8 %x2)
+ %res3 = add <16 x i8> %res0, %res1
+ %res4 = add <16 x i8> %res3, %res2
+ ret <16 x i8> %res4
+}
+
+declare void @llvm.x86.avx512.mask.pmov.qb.mem.128(i8* %ptr, <2 x i64>, i8)
+
+define void @test_int_x86_avx512_mask_pmov_qb_mem_128(i8* %ptr, <2 x i64> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmov_qb_mem_128:
+; CHECK: vpmovqb %xmm0, (%rdi)
+; CHECK: vpmovqb %xmm0, (%rdi) {%k1}
+ call void @llvm.x86.avx512.mask.pmov.qb.mem.128(i8* %ptr, <2 x i64> %x1, i8 -1)
+ call void @llvm.x86.avx512.mask.pmov.qb.mem.128(i8* %ptr, <2 x i64> %x1, i8 %x2)
+ ret void
+}
+
+declare <16 x i8> @llvm.x86.avx512.mask.pmovs.qb.128(<2 x i64>, <16 x i8>, i8)
+
+define <16 x i8>@test_int_x86_avx512_mask_pmovs_qb_128(<2 x i64> %x0, <16 x i8> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovs_qb_128:
+; CHECK: vpmovsqb %xmm0, %xmm1 {%k1}
+; CHECK-NEXT: vpmovsqb %xmm0, %xmm2 {%k1} {z}
+; CHECK-NEXT: vpmovsqb %xmm0, %xmm0
+ %res0 = call <16 x i8> @llvm.x86.avx512.mask.pmovs.qb.128(<2 x i64> %x0, <16 x i8> %x1, i8 -1)
+ %res1 = call <16 x i8> @llvm.x86.avx512.mask.pmovs.qb.128(<2 x i64> %x0, <16 x i8> %x1, i8 %x2)
+ %res2 = call <16 x i8> @llvm.x86.avx512.mask.pmovs.qb.128(<2 x i64> %x0, <16 x i8> zeroinitializer, i8 %x2)
+ %res3 = add <16 x i8> %res0, %res1
+ %res4 = add <16 x i8> %res3, %res2
+ ret <16 x i8> %res4
+}
+
+declare void @llvm.x86.avx512.mask.pmovs.qb.mem.128(i8* %ptr, <2 x i64>, i8)
+
+define void @test_int_x86_avx512_mask_pmovs_qb_mem_128(i8* %ptr, <2 x i64> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovs_qb_mem_128:
+; CHECK: vpmovsqb %xmm0, (%rdi)
+; CHECK: vpmovsqb %xmm0, (%rdi) {%k1}
+ call void @llvm.x86.avx512.mask.pmovs.qb.mem.128(i8* %ptr, <2 x i64> %x1, i8 -1)
+ call void @llvm.x86.avx512.mask.pmovs.qb.mem.128(i8* %ptr, <2 x i64> %x1, i8 %x2)
+ ret void
+}
+
+declare <16 x i8> @llvm.x86.avx512.mask.pmovus.qb.128(<2 x i64>, <16 x i8>, i8)
+
+define <16 x i8>@test_int_x86_avx512_mask_pmovus_qb_128(<2 x i64> %x0, <16 x i8> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovus_qb_128:
+; CHECK: vpmovusqb %xmm0, %xmm1 {%k1}
+; CHECK-NEXT: vpmovusqb %xmm0, %xmm2 {%k1} {z}
+; CHECK-NEXT: vpmovusqb %xmm0, %xmm0
+ %res0 = call <16 x i8> @llvm.x86.avx512.mask.pmovus.qb.128(<2 x i64> %x0, <16 x i8> %x1, i8 -1)
+ %res1 = call <16 x i8> @llvm.x86.avx512.mask.pmovus.qb.128(<2 x i64> %x0, <16 x i8> %x1, i8 %x2)
+ %res2 = call <16 x i8> @llvm.x86.avx512.mask.pmovus.qb.128(<2 x i64> %x0, <16 x i8> zeroinitializer, i8 %x2)
+ %res3 = add <16 x i8> %res0, %res1
+ %res4 = add <16 x i8> %res3, %res2
+ ret <16 x i8> %res4
+}
+
+declare void @llvm.x86.avx512.mask.pmovus.qb.mem.128(i8* %ptr, <2 x i64>, i8)
+
+define void @test_int_x86_avx512_mask_pmovus_qb_mem_128(i8* %ptr, <2 x i64> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovus_qb_mem_128:
+; CHECK: vpmovusqb %xmm0, (%rdi)
+; CHECK: vpmovusqb %xmm0, (%rdi) {%k1}
+ call void @llvm.x86.avx512.mask.pmovus.qb.mem.128(i8* %ptr, <2 x i64> %x1, i8 -1)
+ call void @llvm.x86.avx512.mask.pmovus.qb.mem.128(i8* %ptr, <2 x i64> %x1, i8 %x2)
+ ret void
+}
+
+declare <16 x i8> @llvm.x86.avx512.mask.pmov.qb.256(<4 x i64>, <16 x i8>, i8)
+
+define <16 x i8>@test_int_x86_avx512_mask_pmov_qb_256(<4 x i64> %x0, <16 x i8> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmov_qb_256:
+; CHECK: vpmovqb %ymm0, %xmm1 {%k1}
+; CHECK-NEXT: vpmovqb %ymm0, %xmm2 {%k1} {z}
+; CHECK-NEXT: vpmovqb %ymm0, %xmm0
+ %res0 = call <16 x i8> @llvm.x86.avx512.mask.pmov.qb.256(<4 x i64> %x0, <16 x i8> %x1, i8 -1)
+ %res1 = call <16 x i8> @llvm.x86.avx512.mask.pmov.qb.256(<4 x i64> %x0, <16 x i8> %x1, i8 %x2)
+ %res2 = call <16 x i8> @llvm.x86.avx512.mask.pmov.qb.256(<4 x i64> %x0, <16 x i8> zeroinitializer, i8 %x2)
+ %res3 = add <16 x i8> %res0, %res1
+ %res4 = add <16 x i8> %res3, %res2
+ ret <16 x i8> %res4
+}
+
+declare void @llvm.x86.avx512.mask.pmov.qb.mem.256(i8* %ptr, <4 x i64>, i8)
+
+define void @test_int_x86_avx512_mask_pmov_qb_mem_256(i8* %ptr, <4 x i64> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmov_qb_mem_256:
+; CHECK: vpmovqb %ymm0, (%rdi)
+; CHECK: vpmovqb %ymm0, (%rdi) {%k1}
+ call void @llvm.x86.avx512.mask.pmov.qb.mem.256(i8* %ptr, <4 x i64> %x1, i8 -1)
+ call void @llvm.x86.avx512.mask.pmov.qb.mem.256(i8* %ptr, <4 x i64> %x1, i8 %x2)
+ ret void
+}
+
+declare <16 x i8> @llvm.x86.avx512.mask.pmovs.qb.256(<4 x i64>, <16 x i8>, i8)
+
+define <16 x i8>@test_int_x86_avx512_mask_pmovs_qb_256(<4 x i64> %x0, <16 x i8> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovs_qb_256:
+; CHECK: vpmovsqb %ymm0, %xmm1 {%k1}
+; CHECK-NEXT: vpmovsqb %ymm0, %xmm2 {%k1} {z}
+; CHECK-NEXT: vpmovsqb %ymm0, %xmm0
+ %res0 = call <16 x i8> @llvm.x86.avx512.mask.pmovs.qb.256(<4 x i64> %x0, <16 x i8> %x1, i8 -1)
+ %res1 = call <16 x i8> @llvm.x86.avx512.mask.pmovs.qb.256(<4 x i64> %x0, <16 x i8> %x1, i8 %x2)
+ %res2 = call <16 x i8> @llvm.x86.avx512.mask.pmovs.qb.256(<4 x i64> %x0, <16 x i8> zeroinitializer, i8 %x2)
+ %res3 = add <16 x i8> %res0, %res1
+ %res4 = add <16 x i8> %res3, %res2
+ ret <16 x i8> %res4
+}
+
+declare void @llvm.x86.avx512.mask.pmovs.qb.mem.256(i8* %ptr, <4 x i64>, i8)
+
+define void @test_int_x86_avx512_mask_pmovs_qb_mem_256(i8* %ptr, <4 x i64> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovs_qb_mem_256:
+; CHECK: vpmovsqb %ymm0, (%rdi)
+; CHECK: vpmovsqb %ymm0, (%rdi) {%k1}
+ call void @llvm.x86.avx512.mask.pmovs.qb.mem.256(i8* %ptr, <4 x i64> %x1, i8 -1)
+ call void @llvm.x86.avx512.mask.pmovs.qb.mem.256(i8* %ptr, <4 x i64> %x1, i8 %x2)
+ ret void
+}
+
+declare <16 x i8> @llvm.x86.avx512.mask.pmovus.qb.256(<4 x i64>, <16 x i8>, i8)
+
+define <16 x i8>@test_int_x86_avx512_mask_pmovus_qb_256(<4 x i64> %x0, <16 x i8> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovus_qb_256:
+; CHECK: vpmovusqb %ymm0, %xmm1 {%k1}
+; CHECK-NEXT: vpmovusqb %ymm0, %xmm2 {%k1} {z}
+; CHECK-NEXT: vpmovusqb %ymm0, %xmm0
+ %res0 = call <16 x i8> @llvm.x86.avx512.mask.pmovus.qb.256(<4 x i64> %x0, <16 x i8> %x1, i8 -1)
+ %res1 = call <16 x i8> @llvm.x86.avx512.mask.pmovus.qb.256(<4 x i64> %x0, <16 x i8> %x1, i8 %x2)
+ %res2 = call <16 x i8> @llvm.x86.avx512.mask.pmovus.qb.256(<4 x i64> %x0, <16 x i8> zeroinitializer, i8 %x2)
+ %res3 = add <16 x i8> %res0, %res1
+ %res4 = add <16 x i8> %res3, %res2
+ ret <16 x i8> %res4
+}
+
+declare void @llvm.x86.avx512.mask.pmovus.qb.mem.256(i8* %ptr, <4 x i64>, i8)
+
+define void @test_int_x86_avx512_mask_pmovus_qb_mem_256(i8* %ptr, <4 x i64> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovus_qb_mem_256:
+; CHECK: vpmovusqb %ymm0, (%rdi)
+; CHECK: vpmovusqb %ymm0, (%rdi) {%k1}
+ call void @llvm.x86.avx512.mask.pmovus.qb.mem.256(i8* %ptr, <4 x i64> %x1, i8 -1)
+ call void @llvm.x86.avx512.mask.pmovus.qb.mem.256(i8* %ptr, <4 x i64> %x1, i8 %x2)
+ ret void
+}
+
+declare <8 x i16> @llvm.x86.avx512.mask.pmov.qw.128(<2 x i64>, <8 x i16>, i8)
+
+define <8 x i16>@test_int_x86_avx512_mask_pmov_qw_128(<2 x i64> %x0, <8 x i16> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmov_qw_128:
+; CHECK: vpmovqw %xmm0, %xmm1 {%k1}
+; CHECK-NEXT: vpmovqw %xmm0, %xmm2 {%k1} {z}
+; CHECK-NEXT: vpmovqw %xmm0, %xmm0
+ %res0 = call <8 x i16> @llvm.x86.avx512.mask.pmov.qw.128(<2 x i64> %x0, <8 x i16> %x1, i8 -1)
+ %res1 = call <8 x i16> @llvm.x86.avx512.mask.pmov.qw.128(<2 x i64> %x0, <8 x i16> %x1, i8 %x2)
+ %res2 = call <8 x i16> @llvm.x86.avx512.mask.pmov.qw.128(<2 x i64> %x0, <8 x i16> zeroinitializer, i8 %x2)
+ %res3 = add <8 x i16> %res0, %res1
+ %res4 = add <8 x i16> %res3, %res2
+ ret <8 x i16> %res4
+}
+
+declare void @llvm.x86.avx512.mask.pmov.qw.mem.128(i8* %ptr, <2 x i64>, i8)
+
+define void @test_int_x86_avx512_mask_pmov_qw_mem_128(i8* %ptr, <2 x i64> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmov_qw_mem_128:
+; CHECK: vpmovqw %xmm0, (%rdi)
+; CHECK: vpmovqw %xmm0, (%rdi) {%k1}
+ call void @llvm.x86.avx512.mask.pmov.qw.mem.128(i8* %ptr, <2 x i64> %x1, i8 -1)
+ call void @llvm.x86.avx512.mask.pmov.qw.mem.128(i8* %ptr, <2 x i64> %x1, i8 %x2)
+ ret void
+}
+
+declare <8 x i16> @llvm.x86.avx512.mask.pmovs.qw.128(<2 x i64>, <8 x i16>, i8)
+
+define <8 x i16>@test_int_x86_avx512_mask_pmovs_qw_128(<2 x i64> %x0, <8 x i16> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovs_qw_128:
+; CHECK: vpmovsqw %xmm0, %xmm1 {%k1}
+; CHECK-NEXT: vpmovsqw %xmm0, %xmm2 {%k1} {z}
+; CHECK-NEXT: vpmovsqw %xmm0, %xmm0
+ %res0 = call <8 x i16> @llvm.x86.avx512.mask.pmovs.qw.128(<2 x i64> %x0, <8 x i16> %x1, i8 -1)
+ %res1 = call <8 x i16> @llvm.x86.avx512.mask.pmovs.qw.128(<2 x i64> %x0, <8 x i16> %x1, i8 %x2)
+ %res2 = call <8 x i16> @llvm.x86.avx512.mask.pmovs.qw.128(<2 x i64> %x0, <8 x i16> zeroinitializer, i8 %x2)
+ %res3 = add <8 x i16> %res0, %res1
+ %res4 = add <8 x i16> %res3, %res2
+ ret <8 x i16> %res4
+}
+
+declare void @llvm.x86.avx512.mask.pmovs.qw.mem.128(i8* %ptr, <2 x i64>, i8)
+
+define void @test_int_x86_avx512_mask_pmovs_qw_mem_128(i8* %ptr, <2 x i64> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovs_qw_mem_128:
+; CHECK: vpmovsqw %xmm0, (%rdi)
+; CHECK: vpmovsqw %xmm0, (%rdi) {%k1}
+ call void @llvm.x86.avx512.mask.pmovs.qw.mem.128(i8* %ptr, <2 x i64> %x1, i8 -1)
+ call void @llvm.x86.avx512.mask.pmovs.qw.mem.128(i8* %ptr, <2 x i64> %x1, i8 %x2)
+ ret void
+}
+
+declare <8 x i16> @llvm.x86.avx512.mask.pmovus.qw.128(<2 x i64>, <8 x i16>, i8)
+
+define <8 x i16>@test_int_x86_avx512_mask_pmovus_qw_128(<2 x i64> %x0, <8 x i16> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovus_qw_128:
+; CHECK: vpmovusqw %xmm0, %xmm1 {%k1}
+; CHECK-NEXT: vpmovusqw %xmm0, %xmm2 {%k1} {z}
+; CHECK-NEXT: vpmovusqw %xmm0, %xmm0
+ %res0 = call <8 x i16> @llvm.x86.avx512.mask.pmovus.qw.128(<2 x i64> %x0, <8 x i16> %x1, i8 -1)
+ %res1 = call <8 x i16> @llvm.x86.avx512.mask.pmovus.qw.128(<2 x i64> %x0, <8 x i16> %x1, i8 %x2)
+ %res2 = call <8 x i16> @llvm.x86.avx512.mask.pmovus.qw.128(<2 x i64> %x0, <8 x i16> zeroinitializer, i8 %x2)
+ %res3 = add <8 x i16> %res0, %res1
+ %res4 = add <8 x i16> %res3, %res2
+ ret <8 x i16> %res4
+}
+
+declare void @llvm.x86.avx512.mask.pmovus.qw.mem.128(i8* %ptr, <2 x i64>, i8)
+
+define void @test_int_x86_avx512_mask_pmovus_qw_mem_128(i8* %ptr, <2 x i64> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovus_qw_mem_128:
+; CHECK: vpmovusqw %xmm0, (%rdi)
+; CHECK: vpmovusqw %xmm0, (%rdi) {%k1}
+ call void @llvm.x86.avx512.mask.pmovus.qw.mem.128(i8* %ptr, <2 x i64> %x1, i8 -1)
+ call void @llvm.x86.avx512.mask.pmovus.qw.mem.128(i8* %ptr, <2 x i64> %x1, i8 %x2)
+ ret void
+}
+
+declare <8 x i16> @llvm.x86.avx512.mask.pmov.qw.256(<4 x i64>, <8 x i16>, i8)
+
+define <8 x i16>@test_int_x86_avx512_mask_pmov_qw_256(<4 x i64> %x0, <8 x i16> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmov_qw_256:
+; CHECK: vpmovqw %ymm0, %xmm1 {%k1}
+; CHECK-NEXT: vpmovqw %ymm0, %xmm2 {%k1} {z}
+; CHECK-NEXT: vpmovqw %ymm0, %xmm0
+ %res0 = call <8 x i16> @llvm.x86.avx512.mask.pmov.qw.256(<4 x i64> %x0, <8 x i16> %x1, i8 -1)
+ %res1 = call <8 x i16> @llvm.x86.avx512.mask.pmov.qw.256(<4 x i64> %x0, <8 x i16> %x1, i8 %x2)
+ %res2 = call <8 x i16> @llvm.x86.avx512.mask.pmov.qw.256(<4 x i64> %x0, <8 x i16> zeroinitializer, i8 %x2)
+ %res3 = add <8 x i16> %res0, %res1
+ %res4 = add <8 x i16> %res3, %res2
+ ret <8 x i16> %res4
+}
+
+declare void @llvm.x86.avx512.mask.pmov.qw.mem.256(i8* %ptr, <4 x i64>, i8)
+
+define void @test_int_x86_avx512_mask_pmov_qw_mem_256(i8* %ptr, <4 x i64> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmov_qw_mem_256:
+; CHECK: vpmovqw %ymm0, (%rdi)
+; CHECK: vpmovqw %ymm0, (%rdi) {%k1}
+ call void @llvm.x86.avx512.mask.pmov.qw.mem.256(i8* %ptr, <4 x i64> %x1, i8 -1)
+ call void @llvm.x86.avx512.mask.pmov.qw.mem.256(i8* %ptr, <4 x i64> %x1, i8 %x2)
+ ret void
+}
+
+declare <8 x i16> @llvm.x86.avx512.mask.pmovs.qw.256(<4 x i64>, <8 x i16>, i8)
+
+define <8 x i16>@test_int_x86_avx512_mask_pmovs_qw_256(<4 x i64> %x0, <8 x i16> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovs_qw_256:
+; CHECK: vpmovsqw %ymm0, %xmm1 {%k1}
+; CHECK-NEXT: vpmovsqw %ymm0, %xmm2 {%k1} {z}
+; CHECK-NEXT: vpmovsqw %ymm0, %xmm0
+ %res0 = call <8 x i16> @llvm.x86.avx512.mask.pmovs.qw.256(<4 x i64> %x0, <8 x i16> %x1, i8 -1)
+ %res1 = call <8 x i16> @llvm.x86.avx512.mask.pmovs.qw.256(<4 x i64> %x0, <8 x i16> %x1, i8 %x2)
+ %res2 = call <8 x i16> @llvm.x86.avx512.mask.pmovs.qw.256(<4 x i64> %x0, <8 x i16> zeroinitializer, i8 %x2)
+ %res3 = add <8 x i16> %res0, %res1
+ %res4 = add <8 x i16> %res3, %res2
+ ret <8 x i16> %res4
+}
+
+declare void @llvm.x86.avx512.mask.pmovs.qw.mem.256(i8* %ptr, <4 x i64>, i8)
+
+define void @test_int_x86_avx512_mask_pmovs_qw_mem_256(i8* %ptr, <4 x i64> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovs_qw_mem_256:
+; CHECK: vpmovsqw %ymm0, (%rdi)
+; CHECK: vpmovsqw %ymm0, (%rdi) {%k1}
+ call void @llvm.x86.avx512.mask.pmovs.qw.mem.256(i8* %ptr, <4 x i64> %x1, i8 -1)
+ call void @llvm.x86.avx512.mask.pmovs.qw.mem.256(i8* %ptr, <4 x i64> %x1, i8 %x2)
+ ret void
+}
+
+declare <8 x i16> @llvm.x86.avx512.mask.pmovus.qw.256(<4 x i64>, <8 x i16>, i8)
+
+define <8 x i16>@test_int_x86_avx512_mask_pmovus_qw_256(<4 x i64> %x0, <8 x i16> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovus_qw_256:
+; CHECK: vpmovusqw %ymm0, %xmm1 {%k1}
+; CHECK-NEXT: vpmovusqw %ymm0, %xmm2 {%k1} {z}
+; CHECK-NEXT: vpmovusqw %ymm0, %xmm0
+ %res0 = call <8 x i16> @llvm.x86.avx512.mask.pmovus.qw.256(<4 x i64> %x0, <8 x i16> %x1, i8 -1)
+ %res1 = call <8 x i16> @llvm.x86.avx512.mask.pmovus.qw.256(<4 x i64> %x0, <8 x i16> %x1, i8 %x2)
+ %res2 = call <8 x i16> @llvm.x86.avx512.mask.pmovus.qw.256(<4 x i64> %x0, <8 x i16> zeroinitializer, i8 %x2)
+ %res3 = add <8 x i16> %res0, %res1
+ %res4 = add <8 x i16> %res3, %res2
+ ret <8 x i16> %res4
+}
+
+declare void @llvm.x86.avx512.mask.pmovus.qw.mem.256(i8* %ptr, <4 x i64>, i8)
+
+define void @test_int_x86_avx512_mask_pmovus_qw_mem_256(i8* %ptr, <4 x i64> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovus_qw_mem_256:
+; CHECK: vpmovusqw %ymm0, (%rdi)
+; CHECK: vpmovusqw %ymm0, (%rdi) {%k1}
+ call void @llvm.x86.avx512.mask.pmovus.qw.mem.256(i8* %ptr, <4 x i64> %x1, i8 -1)
+ call void @llvm.x86.avx512.mask.pmovus.qw.mem.256(i8* %ptr, <4 x i64> %x1, i8 %x2)
+ ret void
+}
+
+declare <4 x i32> @llvm.x86.avx512.mask.pmov.qd.128(<2 x i64>, <4 x i32>, i8)
+
+define <4 x i32>@test_int_x86_avx512_mask_pmov_qd_128(<2 x i64> %x0, <4 x i32> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmov_qd_128:
+; CHECK: vpmovqd %xmm0, %xmm1 {%k1}
+; CHECK-NEXT: vpmovqd %xmm0, %xmm2 {%k1} {z}
+; CHECK-NEXT: vpmovqd %xmm0, %xmm0
+ %res0 = call <4 x i32> @llvm.x86.avx512.mask.pmov.qd.128(<2 x i64> %x0, <4 x i32> %x1, i8 -1)
+ %res1 = call <4 x i32> @llvm.x86.avx512.mask.pmov.qd.128(<2 x i64> %x0, <4 x i32> %x1, i8 %x2)
+ %res2 = call <4 x i32> @llvm.x86.avx512.mask.pmov.qd.128(<2 x i64> %x0, <4 x i32> zeroinitializer, i8 %x2)
+ %res3 = add <4 x i32> %res0, %res1
+ %res4 = add <4 x i32> %res3, %res2
+ ret <4 x i32> %res4
+}
+
+declare void @llvm.x86.avx512.mask.pmov.qd.mem.128(i8* %ptr, <2 x i64>, i8)
+
+define void @test_int_x86_avx512_mask_pmov_qd_mem_128(i8* %ptr, <2 x i64> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmov_qd_mem_128:
+; CHECK: vpmovqd %xmm0, (%rdi)
+; CHECK: vpmovqd %xmm0, (%rdi) {%k1}
+ call void @llvm.x86.avx512.mask.pmov.qd.mem.128(i8* %ptr, <2 x i64> %x1, i8 -1)
+ call void @llvm.x86.avx512.mask.pmov.qd.mem.128(i8* %ptr, <2 x i64> %x1, i8 %x2)
+ ret void
+}
+
+declare <4 x i32> @llvm.x86.avx512.mask.pmovs.qd.128(<2 x i64>, <4 x i32>, i8)
+
+define <4 x i32>@test_int_x86_avx512_mask_pmovs_qd_128(<2 x i64> %x0, <4 x i32> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovs_qd_128:
+; CHECK: vpmovsqd %xmm0, %xmm1 {%k1}
+; CHECK-NEXT: vpmovsqd %xmm0, %xmm2 {%k1} {z}
+; CHECK-NEXT: vpmovsqd %xmm0, %xmm0
+ %res0 = call <4 x i32> @llvm.x86.avx512.mask.pmovs.qd.128(<2 x i64> %x0, <4 x i32> %x1, i8 -1)
+ %res1 = call <4 x i32> @llvm.x86.avx512.mask.pmovs.qd.128(<2 x i64> %x0, <4 x i32> %x1, i8 %x2)
+ %res2 = call <4 x i32> @llvm.x86.avx512.mask.pmovs.qd.128(<2 x i64> %x0, <4 x i32> zeroinitializer, i8 %x2)
+ %res3 = add <4 x i32> %res0, %res1
+ %res4 = add <4 x i32> %res3, %res2
+ ret <4 x i32> %res4
+}
+
+declare void @llvm.x86.avx512.mask.pmovs.qd.mem.128(i8* %ptr, <2 x i64>, i8)
+
+define void @test_int_x86_avx512_mask_pmovs_qd_mem_128(i8* %ptr, <2 x i64> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovs_qd_mem_128:
+; CHECK: vpmovsqd %xmm0, (%rdi)
+; CHECK: vpmovsqd %xmm0, (%rdi) {%k1}
+ call void @llvm.x86.avx512.mask.pmovs.qd.mem.128(i8* %ptr, <2 x i64> %x1, i8 -1)
+ call void @llvm.x86.avx512.mask.pmovs.qd.mem.128(i8* %ptr, <2 x i64> %x1, i8 %x2)
+ ret void
+}
+
+declare <4 x i32> @llvm.x86.avx512.mask.pmovus.qd.128(<2 x i64>, <4 x i32>, i8)
+
+define <4 x i32>@test_int_x86_avx512_mask_pmovus_qd_128(<2 x i64> %x0, <4 x i32> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovus_qd_128:
+; CHECK: vpmovusqd %xmm0, %xmm1 {%k1}
+; CHECK-NEXT: vpmovusqd %xmm0, %xmm2 {%k1} {z}
+; CHECK-NEXT: vpmovusqd %xmm0, %xmm0
+ %res0 = call <4 x i32> @llvm.x86.avx512.mask.pmovus.qd.128(<2 x i64> %x0, <4 x i32> %x1, i8 -1)
+ %res1 = call <4 x i32> @llvm.x86.avx512.mask.pmovus.qd.128(<2 x i64> %x0, <4 x i32> %x1, i8 %x2)
+ %res2 = call <4 x i32> @llvm.x86.avx512.mask.pmovus.qd.128(<2 x i64> %x0, <4 x i32> zeroinitializer, i8 %x2)
+ %res3 = add <4 x i32> %res0, %res1
+ %res4 = add <4 x i32> %res3, %res2
+ ret <4 x i32> %res4
+}
+
+declare void @llvm.x86.avx512.mask.pmovus.qd.mem.128(i8* %ptr, <2 x i64>, i8)
+
+define void @test_int_x86_avx512_mask_pmovus_qd_mem_128(i8* %ptr, <2 x i64> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovus_qd_mem_128:
+; CHECK: vpmovusqd %xmm0, (%rdi)
+; CHECK: vpmovusqd %xmm0, (%rdi) {%k1}
+ call void @llvm.x86.avx512.mask.pmovus.qd.mem.128(i8* %ptr, <2 x i64> %x1, i8 -1)
+ call void @llvm.x86.avx512.mask.pmovus.qd.mem.128(i8* %ptr, <2 x i64> %x1, i8 %x2)
+ ret void
+}
+
+declare <4 x i32> @llvm.x86.avx512.mask.pmov.qd.256(<4 x i64>, <4 x i32>, i8)
+
+define <4 x i32>@test_int_x86_avx512_mask_pmov_qd_256(<4 x i64> %x0, <4 x i32> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmov_qd_256:
+; CHECK: vpmovqd %ymm0, %xmm1 {%k1}
+; CHECK-NEXT: vpmovqd %ymm0, %xmm2 {%k1} {z}
+; CHECK-NEXT: vpmovqd %ymm0, %xmm0
+ %res0 = call <4 x i32> @llvm.x86.avx512.mask.pmov.qd.256(<4 x i64> %x0, <4 x i32> %x1, i8 -1)
+ %res1 = call <4 x i32> @llvm.x86.avx512.mask.pmov.qd.256(<4 x i64> %x0, <4 x i32> %x1, i8 %x2)
+ %res2 = call <4 x i32> @llvm.x86.avx512.mask.pmov.qd.256(<4 x i64> %x0, <4 x i32> zeroinitializer, i8 %x2)
+ %res3 = add <4 x i32> %res0, %res1
+ %res4 = add <4 x i32> %res3, %res2
+ ret <4 x i32> %res4
+}
+
+declare void @llvm.x86.avx512.mask.pmov.qd.mem.256(i8* %ptr, <4 x i64>, i8)
+
+define void @test_int_x86_avx512_mask_pmov_qd_mem_256(i8* %ptr, <4 x i64> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmov_qd_mem_256:
+; CHECK: vpmovqd %ymm0, (%rdi)
+; CHECK: vpmovqd %ymm0, (%rdi) {%k1}
+ call void @llvm.x86.avx512.mask.pmov.qd.mem.256(i8* %ptr, <4 x i64> %x1, i8 -1)
+ call void @llvm.x86.avx512.mask.pmov.qd.mem.256(i8* %ptr, <4 x i64> %x1, i8 %x2)
+ ret void
+}
+
+declare <4 x i32> @llvm.x86.avx512.mask.pmovs.qd.256(<4 x i64>, <4 x i32>, i8)
+
+define <4 x i32>@test_int_x86_avx512_mask_pmovs_qd_256(<4 x i64> %x0, <4 x i32> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovs_qd_256:
+; CHECK: vpmovsqd %ymm0, %xmm1 {%k1}
+; CHECK-NEXT: vpmovsqd %ymm0, %xmm2 {%k1} {z}
+; CHECK-NEXT: vpmovsqd %ymm0, %xmm0
+ %res0 = call <4 x i32> @llvm.x86.avx512.mask.pmovs.qd.256(<4 x i64> %x0, <4 x i32> %x1, i8 -1)
+ %res1 = call <4 x i32> @llvm.x86.avx512.mask.pmovs.qd.256(<4 x i64> %x0, <4 x i32> %x1, i8 %x2)
+ %res2 = call <4 x i32> @llvm.x86.avx512.mask.pmovs.qd.256(<4 x i64> %x0, <4 x i32> zeroinitializer, i8 %x2)
+ %res3 = add <4 x i32> %res0, %res1
+ %res4 = add <4 x i32> %res3, %res2
+ ret <4 x i32> %res4
+}
+
+declare void @llvm.x86.avx512.mask.pmovs.qd.mem.256(i8* %ptr, <4 x i64>, i8)
+
+define void @test_int_x86_avx512_mask_pmovs_qd_mem_256(i8* %ptr, <4 x i64> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovs_qd_mem_256:
+; CHECK: vpmovsqd %ymm0, (%rdi)
+; CHECK: vpmovsqd %ymm0, (%rdi) {%k1}
+ call void @llvm.x86.avx512.mask.pmovs.qd.mem.256(i8* %ptr, <4 x i64> %x1, i8 -1)
+ call void @llvm.x86.avx512.mask.pmovs.qd.mem.256(i8* %ptr, <4 x i64> %x1, i8 %x2)
+ ret void
+}
+
+declare <4 x i32> @llvm.x86.avx512.mask.pmovus.qd.256(<4 x i64>, <4 x i32>, i8)
+
+define <4 x i32>@test_int_x86_avx512_mask_pmovus_qd_256(<4 x i64> %x0, <4 x i32> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovus_qd_256:
+; CHECK: vpmovusqd %ymm0, %xmm1 {%k1}
+; CHECK-NEXT: vpmovusqd %ymm0, %xmm2 {%k1} {z}
+; CHECK-NEXT: vpmovusqd %ymm0, %xmm0
+ %res0 = call <4 x i32> @llvm.x86.avx512.mask.pmovus.qd.256(<4 x i64> %x0, <4 x i32> %x1, i8 -1)
+ %res1 = call <4 x i32> @llvm.x86.avx512.mask.pmovus.qd.256(<4 x i64> %x0, <4 x i32> %x1, i8 %x2)
+ %res2 = call <4 x i32> @llvm.x86.avx512.mask.pmovus.qd.256(<4 x i64> %x0, <4 x i32> zeroinitializer, i8 %x2)
+ %res3 = add <4 x i32> %res0, %res1
+ %res4 = add <4 x i32> %res3, %res2
+ ret <4 x i32> %res4
+}
+
+declare void @llvm.x86.avx512.mask.pmovus.qd.mem.256(i8* %ptr, <4 x i64>, i8)
+
+define void @test_int_x86_avx512_mask_pmovus_qd_mem_256(i8* %ptr, <4 x i64> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovus_qd_mem_256:
+; CHECK: vpmovusqd %ymm0, (%rdi)
+; CHECK: vpmovusqd %ymm0, (%rdi) {%k1}
+ call void @llvm.x86.avx512.mask.pmovus.qd.mem.256(i8* %ptr, <4 x i64> %x1, i8 -1)
+ call void @llvm.x86.avx512.mask.pmovus.qd.mem.256(i8* %ptr, <4 x i64> %x1, i8 %x2)
+ ret void
+}
+
+declare <16 x i8> @llvm.x86.avx512.mask.pmov.db.128(<4 x i32>, <16 x i8>, i8)
+
+define <16 x i8>@test_int_x86_avx512_mask_pmov_db_128(<4 x i32> %x0, <16 x i8> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmov_db_128:
+; CHECK: vpmovdb %xmm0, %xmm1 {%k1}
+; CHECK-NEXT: vpmovdb %xmm0, %xmm2 {%k1} {z}
+; CHECK-NEXT: vpmovdb %xmm0, %xmm0
+ %res0 = call <16 x i8> @llvm.x86.avx512.mask.pmov.db.128(<4 x i32> %x0, <16 x i8> %x1, i8 -1)
+ %res1 = call <16 x i8> @llvm.x86.avx512.mask.pmov.db.128(<4 x i32> %x0, <16 x i8> %x1, i8 %x2)
+ %res2 = call <16 x i8> @llvm.x86.avx512.mask.pmov.db.128(<4 x i32> %x0, <16 x i8> zeroinitializer, i8 %x2)
+ %res3 = add <16 x i8> %res0, %res1
+ %res4 = add <16 x i8> %res3, %res2
+ ret <16 x i8> %res4
+}
+
+declare void @llvm.x86.avx512.mask.pmov.db.mem.128(i8* %ptr, <4 x i32>, i8)
+
+define void @test_int_x86_avx512_mask_pmov_db_mem_128(i8* %ptr, <4 x i32> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmov_db_mem_128:
+; CHECK: vpmovdb %xmm0, (%rdi)
+; CHECK: vpmovdb %xmm0, (%rdi) {%k1}
+ call void @llvm.x86.avx512.mask.pmov.db.mem.128(i8* %ptr, <4 x i32> %x1, i8 -1)
+ call void @llvm.x86.avx512.mask.pmov.db.mem.128(i8* %ptr, <4 x i32> %x1, i8 %x2)
+ ret void
+}
+
+declare <16 x i8> @llvm.x86.avx512.mask.pmovs.db.128(<4 x i32>, <16 x i8>, i8)
+
+define <16 x i8>@test_int_x86_avx512_mask_pmovs_db_128(<4 x i32> %x0, <16 x i8> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovs_db_128:
+; CHECK: vpmovsdb %xmm0, %xmm1 {%k1}
+; CHECK-NEXT: vpmovsdb %xmm0, %xmm2 {%k1} {z}
+; CHECK-NEXT: vpmovsdb %xmm0, %xmm0
+ %res0 = call <16 x i8> @llvm.x86.avx512.mask.pmovs.db.128(<4 x i32> %x0, <16 x i8> %x1, i8 -1)
+ %res1 = call <16 x i8> @llvm.x86.avx512.mask.pmovs.db.128(<4 x i32> %x0, <16 x i8> %x1, i8 %x2)
+ %res2 = call <16 x i8> @llvm.x86.avx512.mask.pmovs.db.128(<4 x i32> %x0, <16 x i8> zeroinitializer, i8 %x2)
+ %res3 = add <16 x i8> %res0, %res1
+ %res4 = add <16 x i8> %res3, %res2
+ ret <16 x i8> %res4
+}
+
+declare void @llvm.x86.avx512.mask.pmovs.db.mem.128(i8* %ptr, <4 x i32>, i8)
+
+define void @test_int_x86_avx512_mask_pmovs_db_mem_128(i8* %ptr, <4 x i32> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovs_db_mem_128:
+; CHECK: vpmovsdb %xmm0, (%rdi)
+; CHECK: vpmovsdb %xmm0, (%rdi) {%k1}
+ call void @llvm.x86.avx512.mask.pmovs.db.mem.128(i8* %ptr, <4 x i32> %x1, i8 -1)
+ call void @llvm.x86.avx512.mask.pmovs.db.mem.128(i8* %ptr, <4 x i32> %x1, i8 %x2)
+ ret void
+}
+
+declare <16 x i8> @llvm.x86.avx512.mask.pmovus.db.128(<4 x i32>, <16 x i8>, i8)
+
+define <16 x i8>@test_int_x86_avx512_mask_pmovus_db_128(<4 x i32> %x0, <16 x i8> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovus_db_128:
+; CHECK: vpmovusdb %xmm0, %xmm1 {%k1}
+; CHECK-NEXT: vpmovusdb %xmm0, %xmm2 {%k1} {z}
+; CHECK-NEXT: vpmovusdb %xmm0, %xmm0
+ %res0 = call <16 x i8> @llvm.x86.avx512.mask.pmovus.db.128(<4 x i32> %x0, <16 x i8> %x1, i8 -1)
+ %res1 = call <16 x i8> @llvm.x86.avx512.mask.pmovus.db.128(<4 x i32> %x0, <16 x i8> %x1, i8 %x2)
+ %res2 = call <16 x i8> @llvm.x86.avx512.mask.pmovus.db.128(<4 x i32> %x0, <16 x i8> zeroinitializer, i8 %x2)
+ %res3 = add <16 x i8> %res0, %res1
+ %res4 = add <16 x i8> %res3, %res2
+ ret <16 x i8> %res4
+}
+
+declare void @llvm.x86.avx512.mask.pmovus.db.mem.128(i8* %ptr, <4 x i32>, i8)
+
+define void @test_int_x86_avx512_mask_pmovus_db_mem_128(i8* %ptr, <4 x i32> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovus_db_mem_128:
+; CHECK: vpmovusdb %xmm0, (%rdi)
+; CHECK: vpmovusdb %xmm0, (%rdi) {%k1}
+ call void @llvm.x86.avx512.mask.pmovus.db.mem.128(i8* %ptr, <4 x i32> %x1, i8 -1)
+ call void @llvm.x86.avx512.mask.pmovus.db.mem.128(i8* %ptr, <4 x i32> %x1, i8 %x2)
+ ret void
+}
+
+declare <16 x i8> @llvm.x86.avx512.mask.pmov.db.256(<8 x i32>, <16 x i8>, i8)
+
+define <16 x i8>@test_int_x86_avx512_mask_pmov_db_256(<8 x i32> %x0, <16 x i8> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmov_db_256:
+; CHECK: vpmovdb %ymm0, %xmm1 {%k1}
+; CHECK-NEXT: vpmovdb %ymm0, %xmm2 {%k1} {z}
+; CHECK-NEXT: vpmovdb %ymm0, %xmm0
+ %res0 = call <16 x i8> @llvm.x86.avx512.mask.pmov.db.256(<8 x i32> %x0, <16 x i8> %x1, i8 -1)
+ %res1 = call <16 x i8> @llvm.x86.avx512.mask.pmov.db.256(<8 x i32> %x0, <16 x i8> %x1, i8 %x2)
+ %res2 = call <16 x i8> @llvm.x86.avx512.mask.pmov.db.256(<8 x i32> %x0, <16 x i8> zeroinitializer, i8 %x2)
+ %res3 = add <16 x i8> %res0, %res1
+ %res4 = add <16 x i8> %res3, %res2
+ ret <16 x i8> %res4
+}
+
+declare void @llvm.x86.avx512.mask.pmov.db.mem.256(i8* %ptr, <8 x i32>, i8)
+
+define void @test_int_x86_avx512_mask_pmov_db_mem_256(i8* %ptr, <8 x i32> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmov_db_mem_256:
+; CHECK: vpmovdb %ymm0, (%rdi)
+; CHECK: vpmovdb %ymm0, (%rdi) {%k1}
+ call void @llvm.x86.avx512.mask.pmov.db.mem.256(i8* %ptr, <8 x i32> %x1, i8 -1)
+ call void @llvm.x86.avx512.mask.pmov.db.mem.256(i8* %ptr, <8 x i32> %x1, i8 %x2)
+ ret void
+}
+
+declare <16 x i8> @llvm.x86.avx512.mask.pmovs.db.256(<8 x i32>, <16 x i8>, i8)
+
+define <16 x i8>@test_int_x86_avx512_mask_pmovs_db_256(<8 x i32> %x0, <16 x i8> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovs_db_256:
+; CHECK: vpmovsdb %ymm0, %xmm1 {%k1}
+; CHECK-NEXT: vpmovsdb %ymm0, %xmm2 {%k1} {z}
+; CHECK-NEXT: vpmovsdb %ymm0, %xmm0
+ %res0 = call <16 x i8> @llvm.x86.avx512.mask.pmovs.db.256(<8 x i32> %x0, <16 x i8> %x1, i8 -1)
+ %res1 = call <16 x i8> @llvm.x86.avx512.mask.pmovs.db.256(<8 x i32> %x0, <16 x i8> %x1, i8 %x2)
+ %res2 = call <16 x i8> @llvm.x86.avx512.mask.pmovs.db.256(<8 x i32> %x0, <16 x i8> zeroinitializer, i8 %x2)
+ %res3 = add <16 x i8> %res0, %res1
+ %res4 = add <16 x i8> %res3, %res2
+ ret <16 x i8> %res4
+}
+
+declare void @llvm.x86.avx512.mask.pmovs.db.mem.256(i8* %ptr, <8 x i32>, i8)
+
+define void @test_int_x86_avx512_mask_pmovs_db_mem_256(i8* %ptr, <8 x i32> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovs_db_mem_256:
+; CHECK: vpmovsdb %ymm0, (%rdi)
+; CHECK: vpmovsdb %ymm0, (%rdi) {%k1}
+ call void @llvm.x86.avx512.mask.pmovs.db.mem.256(i8* %ptr, <8 x i32> %x1, i8 -1)
+ call void @llvm.x86.avx512.mask.pmovs.db.mem.256(i8* %ptr, <8 x i32> %x1, i8 %x2)
+ ret void
+}
+
+declare <16 x i8> @llvm.x86.avx512.mask.pmovus.db.256(<8 x i32>, <16 x i8>, i8)
+
+define <16 x i8>@test_int_x86_avx512_mask_pmovus_db_256(<8 x i32> %x0, <16 x i8> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovus_db_256:
+; CHECK: vpmovusdb %ymm0, %xmm1 {%k1}
+; CHECK-NEXT: vpmovusdb %ymm0, %xmm2 {%k1} {z}
+; CHECK-NEXT: vpmovusdb %ymm0, %xmm0
+ %res0 = call <16 x i8> @llvm.x86.avx512.mask.pmovus.db.256(<8 x i32> %x0, <16 x i8> %x1, i8 -1)
+ %res1 = call <16 x i8> @llvm.x86.avx512.mask.pmovus.db.256(<8 x i32> %x0, <16 x i8> %x1, i8 %x2)
+ %res2 = call <16 x i8> @llvm.x86.avx512.mask.pmovus.db.256(<8 x i32> %x0, <16 x i8> zeroinitializer, i8 %x2)
+ %res3 = add <16 x i8> %res0, %res1
+ %res4 = add <16 x i8> %res3, %res2
+ ret <16 x i8> %res4
+}
+
+declare void @llvm.x86.avx512.mask.pmovus.db.mem.256(i8* %ptr, <8 x i32>, i8)
+
+define void @test_int_x86_avx512_mask_pmovus_db_mem_256(i8* %ptr, <8 x i32> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovus_db_mem_256:
+; CHECK: vpmovusdb %ymm0, (%rdi)
+; CHECK: vpmovusdb %ymm0, (%rdi) {%k1}
+ call void @llvm.x86.avx512.mask.pmovus.db.mem.256(i8* %ptr, <8 x i32> %x1, i8 -1)
+ call void @llvm.x86.avx512.mask.pmovus.db.mem.256(i8* %ptr, <8 x i32> %x1, i8 %x2)
+ ret void
+}
+
+declare <8 x i16> @llvm.x86.avx512.mask.pmov.dw.128(<4 x i32>, <8 x i16>, i8)
+
+define <8 x i16>@test_int_x86_avx512_mask_pmov_dw_128(<4 x i32> %x0, <8 x i16> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmov_dw_128:
+; CHECK: vpmovdw %xmm0, %xmm1 {%k1}
+; CHECK-NEXT: vpmovdw %xmm0, %xmm2 {%k1} {z}
+; CHECK-NEXT: vpmovdw %xmm0, %xmm0
+ %res0 = call <8 x i16> @llvm.x86.avx512.mask.pmov.dw.128(<4 x i32> %x0, <8 x i16> %x1, i8 -1)
+ %res1 = call <8 x i16> @llvm.x86.avx512.mask.pmov.dw.128(<4 x i32> %x0, <8 x i16> %x1, i8 %x2)
+ %res2 = call <8 x i16> @llvm.x86.avx512.mask.pmov.dw.128(<4 x i32> %x0, <8 x i16> zeroinitializer, i8 %x2)
+ %res3 = add <8 x i16> %res0, %res1
+ %res4 = add <8 x i16> %res3, %res2
+ ret <8 x i16> %res4
+}
+
+declare void @llvm.x86.avx512.mask.pmov.dw.mem.128(i8* %ptr, <4 x i32>, i8)
+
+define void @test_int_x86_avx512_mask_pmov_dw_mem_128(i8* %ptr, <4 x i32> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmov_dw_mem_128:
+; CHECK: vpmovdw %xmm0, (%rdi)
+; CHECK: vpmovdw %xmm0, (%rdi) {%k1}
+ call void @llvm.x86.avx512.mask.pmov.dw.mem.128(i8* %ptr, <4 x i32> %x1, i8 -1)
+ call void @llvm.x86.avx512.mask.pmov.dw.mem.128(i8* %ptr, <4 x i32> %x1, i8 %x2)
+ ret void
+}
+
+declare <8 x i16> @llvm.x86.avx512.mask.pmovs.dw.128(<4 x i32>, <8 x i16>, i8)
+
+define <8 x i16>@test_int_x86_avx512_mask_pmovs_dw_128(<4 x i32> %x0, <8 x i16> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovs_dw_128:
+; CHECK: vpmovsdw %xmm0, %xmm1 {%k1}
+; CHECK-NEXT: vpmovsdw %xmm0, %xmm2 {%k1} {z}
+; CHECK-NEXT: vpmovsdw %xmm0, %xmm0
+ %res0 = call <8 x i16> @llvm.x86.avx512.mask.pmovs.dw.128(<4 x i32> %x0, <8 x i16> %x1, i8 -1)
+ %res1 = call <8 x i16> @llvm.x86.avx512.mask.pmovs.dw.128(<4 x i32> %x0, <8 x i16> %x1, i8 %x2)
+ %res2 = call <8 x i16> @llvm.x86.avx512.mask.pmovs.dw.128(<4 x i32> %x0, <8 x i16> zeroinitializer, i8 %x2)
+ %res3 = add <8 x i16> %res0, %res1
+ %res4 = add <8 x i16> %res3, %res2
+ ret <8 x i16> %res4
+}
+
+declare void @llvm.x86.avx512.mask.pmovs.dw.mem.128(i8* %ptr, <4 x i32>, i8)
+
+define void @test_int_x86_avx512_mask_pmovs_dw_mem_128(i8* %ptr, <4 x i32> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovs_dw_mem_128:
+; CHECK: vpmovsdw %xmm0, (%rdi)
+; CHECK: vpmovsdw %xmm0, (%rdi) {%k1}
+ call void @llvm.x86.avx512.mask.pmovs.dw.mem.128(i8* %ptr, <4 x i32> %x1, i8 -1)
+ call void @llvm.x86.avx512.mask.pmovs.dw.mem.128(i8* %ptr, <4 x i32> %x1, i8 %x2)
+ ret void
+}
+
+declare <8 x i16> @llvm.x86.avx512.mask.pmovus.dw.128(<4 x i32>, <8 x i16>, i8)
+
+define <8 x i16>@test_int_x86_avx512_mask_pmovus_dw_128(<4 x i32> %x0, <8 x i16> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovus_dw_128:
+; CHECK: vpmovusdw %xmm0, %xmm1 {%k1}
+; CHECK-NEXT: vpmovusdw %xmm0, %xmm2 {%k1} {z}
+; CHECK-NEXT: vpmovusdw %xmm0, %xmm0
+ %res0 = call <8 x i16> @llvm.x86.avx512.mask.pmovus.dw.128(<4 x i32> %x0, <8 x i16> %x1, i8 -1)
+ %res1 = call <8 x i16> @llvm.x86.avx512.mask.pmovus.dw.128(<4 x i32> %x0, <8 x i16> %x1, i8 %x2)
+ %res2 = call <8 x i16> @llvm.x86.avx512.mask.pmovus.dw.128(<4 x i32> %x0, <8 x i16> zeroinitializer, i8 %x2)
+ %res3 = add <8 x i16> %res0, %res1
+ %res4 = add <8 x i16> %res3, %res2
+ ret <8 x i16> %res4
+}
+
+declare void @llvm.x86.avx512.mask.pmovus.dw.mem.128(i8* %ptr, <4 x i32>, i8)
+
+define void @test_int_x86_avx512_mask_pmovus_dw_mem_128(i8* %ptr, <4 x i32> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovus_dw_mem_128:
+; CHECK: vpmovusdw %xmm0, (%rdi)
+; CHECK: vpmovusdw %xmm0, (%rdi) {%k1}
+ call void @llvm.x86.avx512.mask.pmovus.dw.mem.128(i8* %ptr, <4 x i32> %x1, i8 -1)
+ call void @llvm.x86.avx512.mask.pmovus.dw.mem.128(i8* %ptr, <4 x i32> %x1, i8 %x2)
+ ret void
+}
+
+declare <8 x i16> @llvm.x86.avx512.mask.pmov.dw.256(<8 x i32>, <8 x i16>, i8)
+
+define <8 x i16>@test_int_x86_avx512_mask_pmov_dw_256(<8 x i32> %x0, <8 x i16> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmov_dw_256:
+; CHECK: vpmovdw %ymm0, %xmm1 {%k1}
+; CHECK-NEXT: vpmovdw %ymm0, %xmm2 {%k1} {z}
+; CHECK-NEXT: vpmovdw %ymm0, %xmm0
+ %res0 = call <8 x i16> @llvm.x86.avx512.mask.pmov.dw.256(<8 x i32> %x0, <8 x i16> %x1, i8 -1)
+ %res1 = call <8 x i16> @llvm.x86.avx512.mask.pmov.dw.256(<8 x i32> %x0, <8 x i16> %x1, i8 %x2)
+ %res2 = call <8 x i16> @llvm.x86.avx512.mask.pmov.dw.256(<8 x i32> %x0, <8 x i16> zeroinitializer, i8 %x2)
+ %res3 = add <8 x i16> %res0, %res1
+ %res4 = add <8 x i16> %res3, %res2
+ ret <8 x i16> %res4
+}
+
+declare void @llvm.x86.avx512.mask.pmov.dw.mem.256(i8* %ptr, <8 x i32>, i8)
+
+define void @test_int_x86_avx512_mask_pmov_dw_mem_256(i8* %ptr, <8 x i32> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmov_dw_mem_256:
+; CHECK: vpmovdw %ymm0, (%rdi)
+; CHECK: vpmovdw %ymm0, (%rdi) {%k1}
+ call void @llvm.x86.avx512.mask.pmov.dw.mem.256(i8* %ptr, <8 x i32> %x1, i8 -1)
+ call void @llvm.x86.avx512.mask.pmov.dw.mem.256(i8* %ptr, <8 x i32> %x1, i8 %x2)
+ ret void
+}
+
+declare <8 x i16> @llvm.x86.avx512.mask.pmovs.dw.256(<8 x i32>, <8 x i16>, i8)
+
+define <8 x i16>@test_int_x86_avx512_mask_pmovs_dw_256(<8 x i32> %x0, <8 x i16> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovs_dw_256:
+; CHECK: vpmovsdw %ymm0, %xmm1 {%k1}
+; CHECK-NEXT: vpmovsdw %ymm0, %xmm2 {%k1} {z}
+; CHECK-NEXT: vpmovsdw %ymm0, %xmm0
+ %res0 = call <8 x i16> @llvm.x86.avx512.mask.pmovs.dw.256(<8 x i32> %x0, <8 x i16> %x1, i8 -1)
+ %res1 = call <8 x i16> @llvm.x86.avx512.mask.pmovs.dw.256(<8 x i32> %x0, <8 x i16> %x1, i8 %x2)
+ %res2 = call <8 x i16> @llvm.x86.avx512.mask.pmovs.dw.256(<8 x i32> %x0, <8 x i16> zeroinitializer, i8 %x2)
+ %res3 = add <8 x i16> %res0, %res1
+ %res4 = add <8 x i16> %res3, %res2
+ ret <8 x i16> %res4
+}
+
+declare void @llvm.x86.avx512.mask.pmovs.dw.mem.256(i8* %ptr, <8 x i32>, i8)
+
+define void @test_int_x86_avx512_mask_pmovs_dw_mem_256(i8* %ptr, <8 x i32> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovs_dw_mem_256:
+; CHECK: vpmovsdw %ymm0, (%rdi)
+; CHECK: vpmovsdw %ymm0, (%rdi) {%k1}
+ call void @llvm.x86.avx512.mask.pmovs.dw.mem.256(i8* %ptr, <8 x i32> %x1, i8 -1)
+ call void @llvm.x86.avx512.mask.pmovs.dw.mem.256(i8* %ptr, <8 x i32> %x1, i8 %x2)
+ ret void
+}
+
+declare <8 x i16> @llvm.x86.avx512.mask.pmovus.dw.256(<8 x i32>, <8 x i16>, i8)
+
+define <8 x i16>@test_int_x86_avx512_mask_pmovus_dw_256(<8 x i32> %x0, <8 x i16> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovus_dw_256:
+; CHECK: vpmovusdw %ymm0, %xmm1 {%k1}
+; CHECK-NEXT: vpmovusdw %ymm0, %xmm2 {%k1} {z}
+; CHECK-NEXT: vpmovusdw %ymm0, %xmm0
+ %res0 = call <8 x i16> @llvm.x86.avx512.mask.pmovus.dw.256(<8 x i32> %x0, <8 x i16> %x1, i8 -1)
+ %res1 = call <8 x i16> @llvm.x86.avx512.mask.pmovus.dw.256(<8 x i32> %x0, <8 x i16> %x1, i8 %x2)
+ %res2 = call <8 x i16> @llvm.x86.avx512.mask.pmovus.dw.256(<8 x i32> %x0, <8 x i16> zeroinitializer, i8 %x2)
+ %res3 = add <8 x i16> %res0, %res1
+ %res4 = add <8 x i16> %res3, %res2
+ ret <8 x i16> %res4
+}
+
+declare void @llvm.x86.avx512.mask.pmovus.dw.mem.256(i8* %ptr, <8 x i32>, i8)
+
+define void @test_int_x86_avx512_mask_pmovus_dw_mem_256(i8* %ptr, <8 x i32> %x1, i8 %x2) {
+; CHECK-LABEL: test_int_x86_avx512_mask_pmovus_dw_mem_256:
+; CHECK: vpmovusdw %ymm0, (%rdi)
+; CHECK: vpmovusdw %ymm0, (%rdi) {%k1}
+ call void @llvm.x86.avx512.mask.pmovus.dw.mem.256(i8* %ptr, <8 x i32> %x1, i8 -1)
+ call void @llvm.x86.avx512.mask.pmovus.dw.mem.256(i8* %ptr, <8 x i32> %x1, i8 %x2)
+ ret void
+}
+
declare <2 x double> @llvm.x86.avx512.mask.cvtdq2pd.128(<4 x i32>, <2 x double>, i8)
define <2 x double>@test_int_x86_avx512_mask_cvt_dq2pd_128(<4 x i32> %x0, <2 x double> %x1, i8 %x2) {
diff --git a/llvm/test/CodeGen/X86/masked_memop.ll b/llvm/test/CodeGen/X86/masked_memop.ll
index 6c16e634a59..f51d4fa103e 100644
--- a/llvm/test/CodeGen/X86/masked_memop.ll
+++ b/llvm/test/CodeGen/X86/masked_memop.ll
@@ -190,10 +190,13 @@ define void @test14(<2 x i32> %trigger, <2 x float>* %addr, <2 x float> %val) {
; AVX2-LABEL: test15
; AVX2: vpmaskmovd
-; SKX-LABEL: test15
-; SKX: kshiftl
-; SKX: kshiftr
-; SKX: vmovdqu32 {{.*}}{%k1}
+; SKX-LABEL: test15:
+; SKX: ## BB#0:
+; SKX-NEXT: vpandq {{.*}}(%rip), %xmm0, %xmm0
+; SKX-NEXT: vpxor %xmm2, %xmm2, %xmm2
+; SKX-NEXT: vpcmpeqq %xmm2, %xmm0, %k1
+; SKX-NEXT: vpmovqd %xmm1, (%rdi) {%k1}
+; SKX-NEXT: retq
define void @test15(<2 x i32> %trigger, <2 x i32>* %addr, <2 x i32> %val) {
%mask = icmp eq <2 x i32> %trigger, zeroinitializer
call void @llvm.masked.store.v2i32(<2 x i32>%val, <2 x i32>* %addr, i32 4, <2 x i1>%mask)
OpenPOWER on IntegriCloud