maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity
llvm-project/llvm/test/CodeGen/X86/avx512bwvl-intrinsics-upgra...

269 lines
16 KiB
LLVM
Raw Normal View History

[AVX512] Remove masked store intrinsics. Clang now emits generic masked store intrinsics instead. The intrinsics will be autoupgraded to the same generic masked stores. llvm-svn: 271245
2016-05-31 09:50:02 +08:00
; NOTE: Assertions have been autogenerated by update_llc_test_checks.py
; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=knl -mattr=+avx512bw -mattr=+avx512vl --show-mc-encoding| FileCheck %s
declare void @llvm.x86.avx512.mask.storeu.b.128(i8*, <16 x i8>, i16)
define void@test_int_x86_avx512_mask_storeu_b_128(i8* %ptr1, i8* %ptr2, <16 x i8> %x1, i16 %x2) {
; CHECK-LABEL: test_int_x86_avx512_mask_storeu_b_128:
; CHECK: ## BB#0:
; CHECK-NEXT: kmovw %edx, %k1 ## encoding: [0xc5,0xf8,0x92,0xca]
; CHECK-NEXT: vmovdqu8 %xmm0, (%rdi) {%k1} ## encoding: [0x62,0xf1,0x7f,0x09,0x7f,0x07]
; CHECK-NEXT: vmovdqu8 %xmm0, (%rsi) ## encoding: [0x62,0xf1,0x7f,0x08,0x7f,0x06]
; CHECK-NEXT: retq ## encoding: [0xc3]
call void @llvm.x86.avx512.mask.storeu.b.128(i8* %ptr1, <16 x i8> %x1, i16 %x2)
call void @llvm.x86.avx512.mask.storeu.b.128(i8* %ptr2, <16 x i8> %x1, i16 -1)
ret void
}
declare void @llvm.x86.avx512.mask.storeu.b.256(i8*, <32 x i8>, i32)
define void@test_int_x86_avx512_mask_storeu_b_256(i8* %ptr1, i8* %ptr2, <32 x i8> %x1, i32 %x2) {
; CHECK-LABEL: test_int_x86_avx512_mask_storeu_b_256:
; CHECK: ## BB#0:
; CHECK-NEXT: kmovd %edx, %k1 ## encoding: [0xc5,0xfb,0x92,0xca]
; CHECK-NEXT: vmovdqu8 %ymm0, (%rdi) {%k1} ## encoding: [0x62,0xf1,0x7f,0x29,0x7f,0x07]
; CHECK-NEXT: vmovdqu8 %ymm0, (%rsi) ## encoding: [0x62,0xf1,0x7f,0x28,0x7f,0x06]
; CHECK-NEXT: retq ## encoding: [0xc3]
call void @llvm.x86.avx512.mask.storeu.b.256(i8* %ptr1, <32 x i8> %x1, i32 %x2)
call void @llvm.x86.avx512.mask.storeu.b.256(i8* %ptr2, <32 x i8> %x1, i32 -1)
ret void
}
declare void @llvm.x86.avx512.mask.storeu.w.128(i8*, <8 x i16>, i8)
define void@test_int_x86_avx512_mask_storeu_w_128(i8* %ptr1, i8* %ptr2, <8 x i16> %x1, i8 %x2) {
; CHECK-LABEL: test_int_x86_avx512_mask_storeu_w_128:
; CHECK: ## BB#0:
; CHECK-NEXT: kmovw %edx, %k1 ## encoding: [0xc5,0xf8,0x92,0xca]
; CHECK-NEXT: vmovdqu16 %xmm0, (%rdi) {%k1} ## encoding: [0x62,0xf1,0xff,0x09,0x7f,0x07]
; CHECK-NEXT: vmovdqu16 %xmm0, (%rsi) ## encoding: [0x62,0xf1,0xff,0x08,0x7f,0x06]
; CHECK-NEXT: retq ## encoding: [0xc3]
call void @llvm.x86.avx512.mask.storeu.w.128(i8* %ptr1, <8 x i16> %x1, i8 %x2)
call void @llvm.x86.avx512.mask.storeu.w.128(i8* %ptr2, <8 x i16> %x1, i8 -1)
ret void
}
declare void @llvm.x86.avx512.mask.storeu.w.256(i8*, <16 x i16>, i16)
define void@test_int_x86_avx512_mask_storeu_w_256(i8* %ptr1, i8* %ptr2, <16 x i16> %x1, i16 %x2) {
; CHECK-LABEL: test_int_x86_avx512_mask_storeu_w_256:
; CHECK: ## BB#0:
; CHECK-NEXT: kmovw %edx, %k1 ## encoding: [0xc5,0xf8,0x92,0xca]
; CHECK-NEXT: vmovdqu16 %ymm0, (%rdi) {%k1} ## encoding: [0x62,0xf1,0xff,0x29,0x7f,0x07]
; CHECK-NEXT: vmovdqu16 %ymm0, (%rsi) ## encoding: [0x62,0xf1,0xff,0x28,0x7f,0x06]
; CHECK-NEXT: retq ## encoding: [0xc3]
call void @llvm.x86.avx512.mask.storeu.w.256(i8* %ptr1, <16 x i16> %x1, i16 %x2)
call void @llvm.x86.avx512.mask.storeu.w.256(i8* %ptr2, <16 x i16> %x1, i16 -1)
ret void
}
[AVX512] Remove masked load intrinsics. Clang now emits generic masked load intrinsics instead. The intrinsics will be autoupgraded to the same generic masked loads. llvm-svn: 271478
2016-06-02 12:19:36 +08:00
declare <8 x i16> @llvm.x86.avx512.mask.loadu.w.128(i8*, <8 x i16>, i8)
define <8 x i16>@test_int_x86_avx512_mask_loadu_w_128(i8* %ptr, i8* %ptr2, <8 x i16> %x1, i8 %mask) {
; CHECK-LABEL: test_int_x86_avx512_mask_loadu_w_128:
; CHECK: ## BB#0:
; CHECK-NEXT: vmovdqu16 (%rdi), %xmm0 ## encoding: [0x62,0xf1,0xff,0x08,0x6f,0x07]
; CHECK-NEXT: kmovw %edx, %k1 ## encoding: [0xc5,0xf8,0x92,0xca]
; CHECK-NEXT: vmovdqu16 (%rsi), %xmm0 {%k1} ## encoding: [0x62,0xf1,0xff,0x09,0x6f,0x06]
; CHECK-NEXT: vmovdqu16 (%rdi), %xmm1 {%k1} {z} ## encoding: [0x62,0xf1,0xff,0x89,0x6f,0x0f]
; CHECK-NEXT: vpaddw %xmm1, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xfd,0xc1]
; CHECK-NEXT: retq ## encoding: [0xc3]
%res0 = call <8 x i16> @llvm.x86.avx512.mask.loadu.w.128(i8* %ptr, <8 x i16> %x1, i8 -1)
%res = call <8 x i16> @llvm.x86.avx512.mask.loadu.w.128(i8* %ptr2, <8 x i16> %res0, i8 %mask)
%res1 = call <8 x i16> @llvm.x86.avx512.mask.loadu.w.128(i8* %ptr, <8 x i16> zeroinitializer, i8 %mask)
%res2 = add <8 x i16> %res, %res1
ret <8 x i16> %res2
}
declare <16 x i16> @llvm.x86.avx512.mask.loadu.w.256(i8*, <16 x i16>, i16)
define <16 x i16>@test_int_x86_avx512_mask_loadu_w_256(i8* %ptr, i8* %ptr2, <16 x i16> %x1, i16 %mask) {
; CHECK-LABEL: test_int_x86_avx512_mask_loadu_w_256:
; CHECK: ## BB#0:
; CHECK-NEXT: vmovdqu16 (%rdi), %ymm0 ## encoding: [0x62,0xf1,0xff,0x28,0x6f,0x07]
; CHECK-NEXT: kmovw %edx, %k1 ## encoding: [0xc5,0xf8,0x92,0xca]
; CHECK-NEXT: vmovdqu16 (%rsi), %ymm0 {%k1} ## encoding: [0x62,0xf1,0xff,0x29,0x6f,0x06]
; CHECK-NEXT: vmovdqu16 (%rdi), %ymm1 {%k1} {z} ## encoding: [0x62,0xf1,0xff,0xa9,0x6f,0x0f]
; CHECK-NEXT: vpaddw %ymm1, %ymm0, %ymm0 ## encoding: [0x62,0xf1,0x7d,0x28,0xfd,0xc1]
; CHECK-NEXT: retq ## encoding: [0xc3]
%res0 = call <16 x i16> @llvm.x86.avx512.mask.loadu.w.256(i8* %ptr, <16 x i16> %x1, i16 -1)
%res = call <16 x i16> @llvm.x86.avx512.mask.loadu.w.256(i8* %ptr2, <16 x i16> %res0, i16 %mask)
%res1 = call <16 x i16> @llvm.x86.avx512.mask.loadu.w.256(i8* %ptr, <16 x i16> zeroinitializer, i16 %mask)
%res2 = add <16 x i16> %res, %res1
ret <16 x i16> %res2
}
declare <16 x i8> @llvm.x86.avx512.mask.loadu.b.128(i8*, <16 x i8>, i16)
define <16 x i8>@test_int_x86_avx512_mask_loadu_b_128(i8* %ptr, i8* %ptr2, <16 x i8> %x1, i16 %mask) {
; CHECK-LABEL: test_int_x86_avx512_mask_loadu_b_128:
; CHECK: ## BB#0:
; CHECK-NEXT: vmovdqu8 (%rdi), %xmm0 ## encoding: [0x62,0xf1,0x7f,0x08,0x6f,0x07]
; CHECK-NEXT: kmovw %edx, %k1 ## encoding: [0xc5,0xf8,0x92,0xca]
; CHECK-NEXT: vmovdqu8 (%rsi), %xmm0 {%k1} ## encoding: [0x62,0xf1,0x7f,0x09,0x6f,0x06]
; CHECK-NEXT: vmovdqu8 (%rdi), %xmm1 {%k1} {z} ## encoding: [0x62,0xf1,0x7f,0x89,0x6f,0x0f]
; CHECK-NEXT: vpaddb %xmm1, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xfc,0xc1]
; CHECK-NEXT: retq ## encoding: [0xc3]
%res0 = call <16 x i8> @llvm.x86.avx512.mask.loadu.b.128(i8* %ptr, <16 x i8> %x1, i16 -1)
%res = call <16 x i8> @llvm.x86.avx512.mask.loadu.b.128(i8* %ptr2, <16 x i8> %res0, i16 %mask)
%res1 = call <16 x i8> @llvm.x86.avx512.mask.loadu.b.128(i8* %ptr, <16 x i8> zeroinitializer, i16 %mask)
%res2 = add <16 x i8> %res, %res1
ret <16 x i8> %res2
}
declare <32 x i8> @llvm.x86.avx512.mask.loadu.b.256(i8*, <32 x i8>, i32)
define <32 x i8>@test_int_x86_avx512_mask_loadu_b_256(i8* %ptr, i8* %ptr2, <32 x i8> %x1, i32 %mask) {
; CHECK-LABEL: test_int_x86_avx512_mask_loadu_b_256:
; CHECK: ## BB#0:
; CHECK-NEXT: vmovdqu8 (%rdi), %ymm0 ## encoding: [0x62,0xf1,0x7f,0x28,0x6f,0x07]
; CHECK-NEXT: kmovd %edx, %k1 ## encoding: [0xc5,0xfb,0x92,0xca]
; CHECK-NEXT: vmovdqu8 (%rsi), %ymm0 {%k1} ## encoding: [0x62,0xf1,0x7f,0x29,0x6f,0x06]
; CHECK-NEXT: vmovdqu8 (%rdi), %ymm1 {%k1} {z} ## encoding: [0x62,0xf1,0x7f,0xa9,0x6f,0x0f]
; CHECK-NEXT: vpaddb %ymm1, %ymm0, %ymm0 ## encoding: [0x62,0xf1,0x7d,0x28,0xfc,0xc1]
; CHECK-NEXT: retq ## encoding: [0xc3]
%res0 = call <32 x i8> @llvm.x86.avx512.mask.loadu.b.256(i8* %ptr, <32 x i8> %x1, i32 -1)
%res = call <32 x i8> @llvm.x86.avx512.mask.loadu.b.256(i8* %ptr2, <32 x i8> %res0, i32 %mask)
%res1 = call <32 x i8> @llvm.x86.avx512.mask.loadu.b.256(i8* %ptr, <32 x i8> zeroinitializer, i32 %mask)
%res2 = add <32 x i8> %res, %res1
ret <32 x i8> %res2
}
[AVX512] Remove masked palignr intrinsics and auto-upgrade them to native IR of vector shuffle and select. llvm-svn: 271872
2016-06-06 14:12:54 +08:00
declare <16 x i8> @llvm.x86.avx512.mask.palignr.128(<16 x i8>, <16 x i8>, i32, <16 x i8>, i16)
define <16 x i8>@test_int_x86_avx512_mask_palignr_128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %x3, i16 %x4) {
; CHECK-LABEL: test_int_x86_avx512_mask_palignr_128:
; CHECK: ## BB#0:
; CHECK-NEXT: vpalignr $2, %xmm1, %xmm0, %xmm3 ## encoding: [0x62,0xf3,0x7d,0x08,0x0f,0xd9,0x02]
[X86] Fix a test I failed to re-generate in r272249. llvm-svn: 272250
2016-06-09 15:10:34 +08:00
; CHECK-NEXT: ## xmm3 = xmm1[2,3,4,5,6,7,8,9,10,11,12,13,14,15],xmm0[0,1]
[AVX512] Remove masked palignr intrinsics and auto-upgrade them to native IR of vector shuffle and select. llvm-svn: 271872
2016-06-06 14:12:54 +08:00
; CHECK-NEXT: kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
; CHECK-NEXT: vpalignr $2, %xmm1, %xmm0, %xmm2 {%k1} ## encoding: [0x62,0xf3,0x7d,0x09,0x0f,0xd1,0x02]
[AVX512] Fix shuffle comment printing to handle the masked versions of some shuffles. Previously we were printing the mask operands as the register names. llvm-svn: 272367
2016-06-10 12:48:05 +08:00
; CHECK-NEXT: ## xmm2 = xmm1[2,3,4,5,6,7,8,9,10,11,12,13,14,15],xmm0[0,1]
[AVX512] Remove masked palignr intrinsics and auto-upgrade them to native IR of vector shuffle and select. llvm-svn: 271872
2016-06-06 14:12:54 +08:00
; CHECK-NEXT: vpalignr $2, %xmm1, %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf3,0x7d,0x89,0x0f,0xc1,0x02]
[AVX512] Fix shuffle comment printing to handle the masked versions of some shuffles. Previously we were printing the mask operands as the register names. llvm-svn: 272367
2016-06-10 12:48:05 +08:00
; CHECK-NEXT: ## xmm0 = xmm1[2,3,4,5,6,7,8,9,10,11,12,13,14,15],xmm0[0,1]
[AVX512] Remove masked palignr intrinsics and auto-upgrade them to native IR of vector shuffle and select. llvm-svn: 271872
2016-06-06 14:12:54 +08:00
; CHECK-NEXT: vpaddb %xmm0, %xmm2, %xmm0 ## encoding: [0x62,0xf1,0x6d,0x08,0xfc,0xc0]
; CHECK-NEXT: vpaddb %xmm3, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xfc,0xc3]
; CHECK-NEXT: retq ## encoding: [0xc3]
%res = call <16 x i8> @llvm.x86.avx512.mask.palignr.128(<16 x i8> %x0, <16 x i8> %x1, i32 2, <16 x i8> %x3, i16 %x4)
%res1 = call <16 x i8> @llvm.x86.avx512.mask.palignr.128(<16 x i8> %x0, <16 x i8> %x1, i32 2, <16 x i8> zeroinitializer, i16 %x4)
%res2 = call <16 x i8> @llvm.x86.avx512.mask.palignr.128(<16 x i8> %x0, <16 x i8> %x1, i32 2, <16 x i8> %x3, i16 -1)
%res3 = add <16 x i8> %res, %res1
%res4 = add <16 x i8> %res3, %res2
ret <16 x i8> %res4
}
declare <32 x i8> @llvm.x86.avx512.mask.palignr.256(<32 x i8>, <32 x i8>, i32, <32 x i8>, i32)
define <32 x i8>@test_int_x86_avx512_mask_palignr_256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x3, i32 %x4) {
; CHECK-LABEL: test_int_x86_avx512_mask_palignr_256:
; CHECK: ## BB#0:
; CHECK-NEXT: vpalignr $2, %ymm1, %ymm0, %ymm3 ## encoding: [0x62,0xf3,0x7d,0x28,0x0f,0xd9,0x02]
[X86] Fix a test I failed to re-generate in r272249. llvm-svn: 272250
2016-06-09 15:10:34 +08:00
; CHECK-NEXT: ## ymm3 = ymm1[2,3,4,5,6,7,8,9,10,11,12,13,14,15],ymm0[0,1],ymm1[18,19,20,21,22,23,24,25,26,27,28,29,30,31],ymm0[16,17]
[AVX512] Remove masked palignr intrinsics and auto-upgrade them to native IR of vector shuffle and select. llvm-svn: 271872
2016-06-06 14:12:54 +08:00
; CHECK-NEXT: kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
; CHECK-NEXT: vpalignr $2, %ymm1, %ymm0, %ymm2 {%k1} ## encoding: [0x62,0xf3,0x7d,0x29,0x0f,0xd1,0x02]
[AVX512] Fix shuffle comment printing to handle the masked versions of some shuffles. Previously we were printing the mask operands as the register names. llvm-svn: 272367
2016-06-10 12:48:05 +08:00
; CHECK-NEXT: ## ymm2 = ymm1[2,3,4,5,6,7,8,9,10,11,12,13,14,15],ymm0[0,1],ymm1[18,19,20,21,22,23,24,25,26,27,28,29,30,31],ymm0[16,17]
[AVX512] Remove masked palignr intrinsics and auto-upgrade them to native IR of vector shuffle and select. llvm-svn: 271872
2016-06-06 14:12:54 +08:00
; CHECK-NEXT: vpalignr $2, %ymm1, %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf3,0x7d,0xa9,0x0f,0xc1,0x02]
[AVX512] Fix shuffle comment printing to handle the masked versions of some shuffles. Previously we were printing the mask operands as the register names. llvm-svn: 272367
2016-06-10 12:48:05 +08:00
; CHECK-NEXT: ## ymm0 = ymm1[2,3,4,5,6,7,8,9,10,11,12,13,14,15],ymm0[0,1],ymm1[18,19,20,21,22,23,24,25,26,27,28,29,30,31],ymm0[16,17]
[AVX512] Remove masked palignr intrinsics and auto-upgrade them to native IR of vector shuffle and select. llvm-svn: 271872
2016-06-06 14:12:54 +08:00
; CHECK-NEXT: vpaddb %ymm0, %ymm2, %ymm0 ## encoding: [0x62,0xf1,0x6d,0x28,0xfc,0xc0]
; CHECK-NEXT: vpaddb %ymm3, %ymm0, %ymm0 ## encoding: [0x62,0xf1,0x7d,0x28,0xfc,0xc3]
; CHECK-NEXT: retq ## encoding: [0xc3]
%res = call <32 x i8> @llvm.x86.avx512.mask.palignr.256(<32 x i8> %x0, <32 x i8> %x1, i32 2, <32 x i8> %x3, i32 %x4)
%res1 = call <32 x i8> @llvm.x86.avx512.mask.palignr.256(<32 x i8> %x0, <32 x i8> %x1, i32 2, <32 x i8> zeroinitializer, i32 %x4)
%res2 = call <32 x i8> @llvm.x86.avx512.mask.palignr.256(<32 x i8> %x0, <32 x i8> %x1, i32 2, <32 x i8> %x3, i32 -1)
%res3 = add <32 x i8> %res, %res1
%res4 = add <32 x i8> %res3, %res2
ret <32 x i8> %res4
}
[AVX512] Remove maksed pshufd, pshuflw, and phufhw intrinsics and autoupgrade them to selects and shufflevector. llvm-svn: 272527
2016-06-13 10:36:48 +08:00
declare <8 x i16> @llvm.x86.avx512.mask.pshufh.w.128(<8 x i16>, i32, <8 x i16>, i8)
define <8 x i16>@test_int_x86_avx512_mask_pshufh_w_128(<8 x i16> %x0, i32 %x1, <8 x i16> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_pshufh_w_128:
; CHECK: ## BB#0:
; CHECK-NEXT: vpshufhw $3, %xmm0, %xmm2 ## encoding: [0x62,0xf1,0x7e,0x08,0x70,0xd0,0x03]
; CHECK-NEXT: ## xmm2 = xmm0[0,1,2,3,7,4,4,4]
; CHECK-NEXT: kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT: vpshufhw $3, %xmm0, %xmm1 {%k1} ## encoding: [0x62,0xf1,0x7e,0x09,0x70,0xc8,0x03]
; CHECK-NEXT: ## xmm1 = xmm0[0,1,2,3,7,4,4,4]
; CHECK-NEXT: vpshufhw $3, %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7e,0x89,0x70,0xc0,0x03]
; CHECK-NEXT: ## xmm0 = xmm0[0,1,2,3,7,4,4,4]
; CHECK-NEXT: vpaddw %xmm0, %xmm1, %xmm0 ## encoding: [0x62,0xf1,0x75,0x08,0xfd,0xc0]
; CHECK-NEXT: vpaddw %xmm2, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xfd,0xc2]
; CHECK-NEXT: retq ## encoding: [0xc3]
%res = call <8 x i16> @llvm.x86.avx512.mask.pshufh.w.128(<8 x i16> %x0, i32 3, <8 x i16> %x2, i8 %x3)
%res1 = call <8 x i16> @llvm.x86.avx512.mask.pshufh.w.128(<8 x i16> %x0, i32 3, <8 x i16> zeroinitializer, i8 %x3)
%res2 = call <8 x i16> @llvm.x86.avx512.mask.pshufh.w.128(<8 x i16> %x0, i32 3, <8 x i16> %x2, i8 -1)
%res3 = add <8 x i16> %res, %res1
%res4 = add <8 x i16> %res3, %res2
ret <8 x i16> %res4
}
declare <16 x i16> @llvm.x86.avx512.mask.pshufh.w.256(<16 x i16>, i32, <16 x i16>, i16)
define <16 x i16>@test_int_x86_avx512_mask_pshufh_w_256(<16 x i16> %x0, i32 %x1, <16 x i16> %x2, i16 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_pshufh_w_256:
; CHECK: ## BB#0:
; CHECK-NEXT: vpshufhw $3, %ymm0, %ymm2 ## encoding: [0x62,0xf1,0x7e,0x28,0x70,0xd0,0x03]
; CHECK-NEXT: ## ymm2 = ymm0[0,1,2,3,7,4,4,4,8,9,10,11,15,12,12,12]
; CHECK-NEXT: kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT: vpshufhw $3, %ymm0, %ymm1 {%k1} ## encoding: [0x62,0xf1,0x7e,0x29,0x70,0xc8,0x03]
; CHECK-NEXT: ## ymm1 = ymm0[0,1,2,3,7,4,4,4,8,9,10,11,15,12,12,12]
; CHECK-NEXT: vpshufhw $3, %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7e,0xa9,0x70,0xc0,0x03]
; CHECK-NEXT: ## ymm0 = ymm0[0,1,2,3,7,4,4,4,8,9,10,11,15,12,12,12]
; CHECK-NEXT: vpaddw %ymm0, %ymm1, %ymm0 ## encoding: [0x62,0xf1,0x75,0x28,0xfd,0xc0]
; CHECK-NEXT: vpaddw %ymm2, %ymm0, %ymm0 ## encoding: [0x62,0xf1,0x7d,0x28,0xfd,0xc2]
; CHECK-NEXT: retq ## encoding: [0xc3]
%res = call <16 x i16> @llvm.x86.avx512.mask.pshufh.w.256(<16 x i16> %x0, i32 3, <16 x i16> %x2, i16 %x3)
%res1 = call <16 x i16> @llvm.x86.avx512.mask.pshufh.w.256(<16 x i16> %x0, i32 3, <16 x i16> zeroinitializer, i16 %x3)
%res2 = call <16 x i16> @llvm.x86.avx512.mask.pshufh.w.256(<16 x i16> %x0, i32 3, <16 x i16> %x2, i16 -1)
%res3 = add <16 x i16> %res, %res1
%res4 = add <16 x i16> %res3, %res2
ret <16 x i16> %res4
}
declare <8 x i16> @llvm.x86.avx512.mask.pshufl.w.128(<8 x i16>, i32, <8 x i16>, i8)
define <8 x i16>@test_int_x86_avx512_mask_pshufl_w_128(<8 x i16> %x0, i32 %x1, <8 x i16> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_pshufl_w_128:
; CHECK: ## BB#0:
; CHECK-NEXT: vpshuflw $3, %xmm0, %xmm2 ## encoding: [0x62,0xf1,0x7f,0x08,0x70,0xd0,0x03]
; CHECK-NEXT: ## xmm2 = xmm0[3,0,0,0,4,5,6,7]
; CHECK-NEXT: kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT: vpshuflw $3, %xmm0, %xmm1 {%k1} ## encoding: [0x62,0xf1,0x7f,0x09,0x70,0xc8,0x03]
; CHECK-NEXT: ## xmm1 = xmm0[3,0,0,0,4,5,6,7]
; CHECK-NEXT: vpshuflw $3, %xmm0, %xmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7f,0x89,0x70,0xc0,0x03]
; CHECK-NEXT: ## xmm0 = xmm0[3,0,0,0,4,5,6,7]
; CHECK-NEXT: vpaddw %xmm0, %xmm1, %xmm0 ## encoding: [0x62,0xf1,0x75,0x08,0xfd,0xc0]
; CHECK-NEXT: vpaddw %xmm2, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xfd,0xc2]
; CHECK-NEXT: retq ## encoding: [0xc3]
%res = call <8 x i16> @llvm.x86.avx512.mask.pshufl.w.128(<8 x i16> %x0, i32 3, <8 x i16> %x2, i8 %x3)
%res1 = call <8 x i16> @llvm.x86.avx512.mask.pshufl.w.128(<8 x i16> %x0, i32 3, <8 x i16> zeroinitializer, i8 %x3)
%res2 = call <8 x i16> @llvm.x86.avx512.mask.pshufl.w.128(<8 x i16> %x0, i32 3, <8 x i16> %x2, i8 -1)
%res3 = add <8 x i16> %res, %res1
%res4 = add <8 x i16> %res3, %res2
ret <8 x i16> %res4
}
declare <16 x i16> @llvm.x86.avx512.mask.pshufl.w.256(<16 x i16>, i32, <16 x i16>, i16)
define <16 x i16>@test_int_x86_avx512_mask_pshufl_w_256(<16 x i16> %x0, i32 %x1, <16 x i16> %x2, i16 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_pshufl_w_256:
; CHECK: ## BB#0:
; CHECK-NEXT: vpshuflw $3, %ymm0, %ymm2 ## encoding: [0x62,0xf1,0x7f,0x28,0x70,0xd0,0x03]
; CHECK-NEXT: ## ymm2 = ymm0[3,0,0,0,4,5,6,7,11,8,8,8,12,13,14,15]
; CHECK-NEXT: kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
; CHECK-NEXT: vpshuflw $3, %ymm0, %ymm1 {%k1} ## encoding: [0x62,0xf1,0x7f,0x29,0x70,0xc8,0x03]
; CHECK-NEXT: ## ymm1 = ymm0[3,0,0,0,4,5,6,7,11,8,8,8,12,13,14,15]
; CHECK-NEXT: vpshuflw $3, %ymm0, %ymm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7f,0xa9,0x70,0xc0,0x03]
; CHECK-NEXT: ## ymm0 = ymm0[3,0,0,0,4,5,6,7,11,8,8,8,12,13,14,15]
; CHECK-NEXT: vpaddw %ymm0, %ymm1, %ymm0 ## encoding: [0x62,0xf1,0x75,0x28,0xfd,0xc0]
; CHECK-NEXT: vpaddw %ymm2, %ymm0, %ymm0 ## encoding: [0x62,0xf1,0x7d,0x28,0xfd,0xc2]
; CHECK-NEXT: retq ## encoding: [0xc3]
%res = call <16 x i16> @llvm.x86.avx512.mask.pshufl.w.256(<16 x i16> %x0, i32 3, <16 x i16> %x2, i16 %x3)
%res1 = call <16 x i16> @llvm.x86.avx512.mask.pshufl.w.256(<16 x i16> %x0, i32 3, <16 x i16> zeroinitializer, i16 %x3)
%res2 = call <16 x i16> @llvm.x86.avx512.mask.pshufl.w.256(<16 x i16> %x0, i32 3, <16 x i16> %x2, i16 -1)
%res3 = add <16 x i16> %res, %res1
%res4 = add <16 x i16> %res3, %res2
ret <16 x i16> %res4
}