RPCS3/llvm
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm.git synced 2025-02-12 13:48:45 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

[AVX512] add PSRLW Intrinsic

Browse Source 
Differential Revision: http://reviews.llvm.org/D15751


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256558 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
Michael Zuckerman 2015-12-29 13:04:35 +00:00
parent df1cf99600
commit 48d1be172b
4 changed files with 148 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

21
include/llvm/IR/IntrinsicsX86.td
View File

@ -2191,7 +2191,26 @@ let TargetPrefix = "x86" in { // All intrinsics start with "llvm.x86.".
def int_x86_avx512_mask_psrai_q : GCCBuiltin<"__builtin_ia32_psraqi512">,
Intrinsic<[llvm_v8i64_ty], [llvm_v8i64_ty,
llvm_i32_ty, llvm_v8i64_ty, llvm_i8_ty], [IntrNoMem]>;
def int_x86_avx512_mask_psrl_w_128 : GCCBuiltin<"__builtin_ia32_psrlw128_mask">,
Intrinsic<[llvm_v8i16_ty], [llvm_v8i16_ty,
llvm_v8i16_ty, llvm_v8i16_ty, llvm_i8_ty], [IntrNoMem]>;
def int_x86_avx512_mask_psrl_w_256 : GCCBuiltin<"__builtin_ia32_psrlw256_mask">,
Intrinsic<[llvm_v16i16_ty], [llvm_v16i16_ty,
llvm_v8i16_ty, llvm_v16i16_ty, llvm_i16_ty], [IntrNoMem]>;
def int_x86_avx512_mask_psrl_w_512 : GCCBuiltin<"__builtin_ia32_psrlw512_mask">,
Intrinsic<[llvm_v32i16_ty], [llvm_v32i16_ty,
llvm_v8i16_ty, llvm_v32i16_ty, llvm_i32_ty], [IntrNoMem]>;
def int_x86_avx512_mask_psrl_wi_128 : GCCBuiltin<"__builtin_ia32_psrlwi128_mask">,
Intrinsic<[llvm_v8i16_ty], [llvm_v8i16_ty,
llvm_i8_ty, llvm_v8i16_ty, llvm_i8_ty], [IntrNoMem]>;
def int_x86_avx512_mask_psrl_wi_256 : GCCBuiltin<"__builtin_ia32_psrlwi256_mask">,
Intrinsic<[llvm_v16i16_ty], [llvm_v16i16_ty,
llvm_i8_ty, llvm_v16i16_ty, llvm_i16_ty], [IntrNoMem]>;
def int_x86_avx512_mask_psrl_wi_512 : GCCBuiltin<"__builtin_ia32_psrlwi512_mask">,
Intrinsic<[llvm_v32i16_ty], [llvm_v32i16_ty,
llvm_i8_ty, llvm_v32i16_ty, llvm_i32_ty], [IntrNoMem]>;
def int_x86_avx512_mask_psll_d : GCCBuiltin<"__builtin_ia32_pslld512_mask">,
Intrinsic<[llvm_v16i32_ty], [llvm_v16i32_ty,
llvm_v4i32_ty, llvm_v16i32_ty, llvm_i16_ty], [IntrNoMem]>;

6
lib/Target/X86/X86IntrinsicsInfo.h
View File

@ -1221,6 +1221,12 @@ static const IntrinsicData IntrinsicsWithoutChain[] = {
X86_INTRINSIC_DATA(avx512_mask_psrav_q, INTR_TYPE_2OP_MASK, ISD::SRA, 0),
X86_INTRINSIC_DATA(avx512_mask_psrl_d, INTR_TYPE_2OP_MASK, X86ISD::VSRL, 0),
X86_INTRINSIC_DATA(avx512_mask_psrl_q, INTR_TYPE_2OP_MASK, X86ISD::VSRL, 0),
X86_INTRINSIC_DATA(avx512_mask_psrl_w_128, INTR_TYPE_2OP_MASK, X86ISD::VSRL, 0),
X86_INTRINSIC_DATA(avx512_mask_psrl_w_256, INTR_TYPE_2OP_MASK, X86ISD::VSRL, 0),
X86_INTRINSIC_DATA(avx512_mask_psrl_w_512, INTR_TYPE_2OP_MASK, X86ISD::VSRL, 0),
X86_INTRINSIC_DATA(avx512_mask_psrl_wi_128, INTR_TYPE_2OP_MASK, X86ISD::VSRLI, 0),
X86_INTRINSIC_DATA(avx512_mask_psrl_wi_256, INTR_TYPE_2OP_MASK, X86ISD::VSRLI, 0),
X86_INTRINSIC_DATA(avx512_mask_psrl_wi_512, INTR_TYPE_2OP_MASK, X86ISD::VSRLI, 0),
X86_INTRINSIC_DATA(avx512_mask_psrli_d, VSHIFT_MASK, X86ISD::VSRLI, 0),
X86_INTRINSIC_DATA(avx512_mask_psrli_q, VSHIFT_MASK, X86ISD::VSRLI, 0),
X86_INTRINSIC_DATA(avx512_mask_psrlv_d, INTR_TYPE_2OP_MASK, ISD::SRL, 0),

40
test/CodeGen/X86/avx512bw-intrinsics.ll
View File

@ -2866,3 +2866,43 @@ define <32 x i16>@test_int_x86_avx512_cvtmask2w_512(i32 %x0) {
%res = call <32 x i16> @llvm.x86.avx512.cvtmask2w.512(i32 %x0)
ret <32 x i16> %res
}
declare <32 x i16> @llvm.x86.avx512.mask.psrl.w.512(<32 x i16>, <8 x i16>, <32 x i16>, i32)
define <32 x i16>@test_int_x86_avx512_mask_psrl_w_512(<32 x i16> %x0, <8 x i16> %x1, <32 x i16> %x2, i32 %x3) {
; AVX512BW-LABEL: test_int_x86_avx512_mask_psrl_w_512:
; AVX512BW: ## BB#0:
; AVX512BW-NEXT: kmovd %edi, %k1
; AVX512BW-NEXT: vpsrlw %xmm1, %zmm0, %zmm2 {%k1}
; AVX512BW-NEXT: vpsrlw %xmm1, %zmm0, %zmm3 {%k1} {z}
; AVX512BW-NEXT: vpsrlw %xmm1, %zmm0, %zmm0
; AVX512BW-NEXT: vpaddw %zmm0, %zmm2, %zmm0
; AVX512BW-NEXT: vpaddw %zmm3, %zmm0, %zmm0
; AVX512BW-NEXT: retq
%res = call <32 x i16> @llvm.x86.avx512.mask.psrl.w.512(<32 x i16> %x0, <8 x i16> %x1, <32 x i16> %x2, i32 %x3)
%res1 = call <32 x i16> @llvm.x86.avx512.mask.psrl.w.512(<32 x i16> %x0, <8 x i16> %x1, <32 x i16> %x2, i32 -1)
%res2 = call <32 x i16> @llvm.x86.avx512.mask.psrl.w.512(<32 x i16> %x0, <8 x i16> %x1, <32 x i16> zeroinitializer, i32 %x3)
%res3 = add <32 x i16> %res, %res1
%res4 = add <32 x i16> %res3, %res2
ret <32 x i16> %res4
}
declare <32 x i16> @llvm.x86.avx512.mask.psrl.wi.512(<32 x i16>, i8, <32 x i16>, i32)
define <32 x i16>@test_int_x86_avx512_mask_psrl_wi_512(<32 x i16> %x0, i8 %x1, <32 x i16> %x2, i32 %x3) {
; AVX512BW-LABEL: test_int_x86_avx512_mask_psrl_wi_512:
; AVX512BW: ## BB#0:
; AVX512BW-NEXT: kmovd %esi, %k1
; AVX512BW-NEXT: vpsrlw $3, %zmm0, %zmm1 {%k1}
; AVX512BW-NEXT: vpsrlw $3, %zmm0, %zmm2 {%k1} {z}
; AVX512BW-NEXT: vpsrlw $3, %zmm0, %zmm0
; AVX512BW-NEXT: vpaddw %zmm0, %zmm1, %zmm0
; AVX512BW-NEXT: vpaddw %zmm2, %zmm0, %zmm0
; AVX512BW-NEXT: retq
%res = call <32 x i16> @llvm.x86.avx512.mask.psrl.wi.512(<32 x i16> %x0, i8 3, <32 x i16> %x2, i32 %x3)
%res1 = call <32 x i16> @llvm.x86.avx512.mask.psrl.wi.512(<32 x i16> %x0, i8 3, <32 x i16> %x2, i32 -1)
%res2 = call <32 x i16> @llvm.x86.avx512.mask.psrl.wi.512(<32 x i16> %x0, i8 3, <32 x i16> zeroinitializer, i32 %x3)
%res3 = add <32 x i16> %res, %res1
%res4 = add <32 x i16> %res3, %res2
ret <32 x i16> %res4
}

82
test/CodeGen/X86/avx512bwvl-intrinsics.ll
View File

@ -4509,3 +4509,85 @@ define <16 x i16>@test_int_x86_avx512_cvtmask2w_256(i16 %x0) {
%res = call <16 x i16> @llvm.x86.avx512.cvtmask2w.256(i16 %x0)
ret <16 x i16> %res
}
declare <8 x i16> @llvm.x86.avx512.mask.psrl.w.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)
define <8 x i16>@test_int_x86_avx512_mask_psrl_w_128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_psrl_w_128:
; CHECK: ## BB#0:
; CHECK-NEXT: movzbl %dil, %eax
; CHECK-NEXT: kmovw %eax, %k1
; CHECK-NEXT: vpsrlw %xmm1, %xmm0, %xmm2 {%k1}
; CHECK-NEXT: vpsrlw %xmm1, %xmm0, %xmm3 {%k1} {z}
; CHECK-NEXT: vpsrlw %xmm1, %xmm0, %xmm0
; CHECK-NEXT: vpaddw %xmm0, %xmm2, %xmm0
; CHECK-NEXT: vpaddw %xmm0, %xmm3, %xmm0
; CHECK-NEXT: retq
%res = call <8 x i16> @llvm.x86.avx512.mask.psrl.w.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3)
%res1 = call <8 x i16> @llvm.x86.avx512.mask.psrl.w.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 -1)
%res2 = call <8 x i16> @llvm.x86.avx512.mask.psrl.w.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> zeroinitializer, i8 %x3)
%res3 = add <8 x i16> %res, %res1
%res4 = add <8 x i16> %res2, %res3
ret <8 x i16> %res4
}
declare <16 x i16> @llvm.x86.avx512.mask.psrl.w.256(<16 x i16>, <8 x i16>, <16 x i16>, i16)
define <16 x i16>@test_int_x86_avx512_mask_psrl_w_256(<16 x i16> %x0, <8 x i16> %x1, <16 x i16> %x2, i16 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_psrl_w_256:
; CHECK: ## BB#0:
; CHECK-NEXT: kmovw %edi, %k1
; CHECK-NEXT: vpsrlw %xmm1, %ymm0, %ymm2 {%k1}
; CHECK-NEXT: vpsrlw %xmm1, %ymm0, %ymm3 {%k1} {z}
; CHECK-NEXT: vpsrlw %xmm1, %ymm0, %ymm0
; CHECK-NEXT: vpaddw %ymm0, %ymm2, %ymm0
; CHECK-NEXT: vpaddw %ymm3, %ymm0, %ymm0
; CHECK-NEXT: retq
%res = call <16 x i16> @llvm.x86.avx512.mask.psrl.w.256(<16 x i16> %x0, <8 x i16> %x1, <16 x i16> %x2, i16 %x3)
%res1 = call <16 x i16> @llvm.x86.avx512.mask.psrl.w.256(<16 x i16> %x0, <8 x i16> %x1, <16 x i16> %x2, i16 -1)
%res2 = call <16 x i16> @llvm.x86.avx512.mask.psrl.w.256(<16 x i16> %x0, <8 x i16> %x1, <16 x i16> zeroinitializer, i16 %x3)
%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.psrl.wi.128(<8 x i16>, i8, <8 x i16>, i8)
define <8 x i16>@test_int_x86_avx512_mask_psrl_wi_128(<8 x i16> %x0, i8 %x1, <8 x i16> %x2, i8 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_psrl_wi_128:
; CHECK: ## BB#0:
; CHECK-NEXT: movzbl %sil, %eax
; CHECK-NEXT: kmovw %eax, %k1
; CHECK-NEXT: vpsrlw $3, %xmm0, %xmm1 {%k1}
; CHECK-NEXT: vpsrlw $3, %xmm0, %xmm2 {%k1} {z}
; CHECK-NEXT: vpsrlw $3, %xmm0, %xmm0
; CHECK-NEXT: vpaddw %xmm0, %xmm1, %xmm0
; CHECK-NEXT: vpaddw %xmm0, %xmm2, %xmm0
; CHECK-NEXT: retq
%res = call <8 x i16> @llvm.x86.avx512.mask.psrl.wi.128(<8 x i16> %x0, i8 3, <8 x i16> %x2, i8 %x3)
%res1 = call <8 x i16> @llvm.x86.avx512.mask.psrl.wi.128(<8 x i16> %x0, i8 3, <8 x i16> %x2, i8 -1)
%res2 = call <8 x i16> @llvm.x86.avx512.mask.psrl.wi.128(<8 x i16> %x0, i8 3, <8 x i16> zeroinitializer, i8 %x3)
%res3 = add <8 x i16> %res, %res1
%res4 = add <8 x i16> %res2, %res3
ret <8 x i16> %res4
}
declare <16 x i16> @llvm.x86.avx512.mask.psrl.wi.256(<16 x i16>, i8, <16 x i16>, i16)
define <16 x i16>@test_int_x86_avx512_mask_psrl_wi_256(<16 x i16> %x0, i8 %x1, <16 x i16> %x2, i16 %x3) {
; CHECK-LABEL: test_int_x86_avx512_mask_psrl_wi_256:
; CHECK: ## BB#0:
; CHECK-NEXT: kmovw %esi, %k1
; CHECK-NEXT: vpsrlw $3, %ymm0, %ymm1 {%k1}
; CHECK-NEXT: vpsrlw $3, %ymm0, %ymm2 {%k1} {z}
; CHECK-NEXT: vpsrlw $3, %ymm0, %ymm0
; CHECK-NEXT: vpaddw %ymm0, %ymm1, %ymm0
; CHECK-NEXT: vpaddw %ymm2, %ymm0, %ymm0
; CHECK-NEXT: retq
%res = call <16 x i16> @llvm.x86.avx512.mask.psrl.wi.256(<16 x i16> %x0, i8 3, <16 x i16> %x2, i16 %x3)
%res1 = call <16 x i16> @llvm.x86.avx512.mask.psrl.wi.256(<16 x i16> %x0, i8 3, <16 x i16> %x2, i16 -1)
%res2 = call <16 x i16> @llvm.x86.avx512.mask.psrl.wi.256(<16 x i16> %x0, i8 3, <16 x i16> zeroinitializer, i16 %x3)
%res3 = add <16 x i16> %res, %res1
%res4 = add <16 x i16> %res3, %res2
ret <16 x i16> %res4
}