From ab0aa0e9d923296cd0e373d98b8586fe314b87f6 Mon Sep 17 00:00:00 2001 From: Michael Zuckerman Date: Mon, 4 Jan 2016 12:50:36 +0000 Subject: [PATCH] [AVX512] add PSRAW Intrinsic Differential Revision: http://reviews.llvm.org/D15850 git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@256751 91177308-0d34-0410-b5e6-96231b3b80d8 --- include/llvm/IR/IntrinsicsX86.td | 19 ++++++ lib/Target/X86/X86IntrinsicsInfo.h | 6 ++ test/CodeGen/X86/avx512bw-intrinsics.ll | 40 +++++++++++ test/CodeGen/X86/avx512bwvl-intrinsics.ll | 82 +++++++++++++++++++++++ 4 files changed, 147 insertions(+) diff --git a/include/llvm/IR/IntrinsicsX86.td b/include/llvm/IR/IntrinsicsX86.td index a20762e1f9b..0e6877268ec 100644 --- a/include/llvm/IR/IntrinsicsX86.td +++ b/include/llvm/IR/IntrinsicsX86.td @@ -2224,6 +2224,25 @@ let TargetPrefix = "x86" in { // All intrinsics start with "llvm.x86.". Intrinsic<[llvm_v32i16_ty], [llvm_v32i16_ty, llvm_i8_ty, llvm_v32i16_ty, llvm_i32_ty], [IntrNoMem]>; + def int_x86_avx512_mask_psra_w_128 : GCCBuiltin<"__builtin_ia32_psraw128_mask">, + Intrinsic<[llvm_v8i16_ty], [llvm_v8i16_ty, + llvm_v8i16_ty, llvm_v8i16_ty, llvm_i8_ty], [IntrNoMem]>; + def int_x86_avx512_mask_psra_w_256 : GCCBuiltin<"__builtin_ia32_psraw256_mask">, + Intrinsic<[llvm_v16i16_ty], [llvm_v16i16_ty, + llvm_v8i16_ty, llvm_v16i16_ty, llvm_i16_ty], [IntrNoMem]>; + def int_x86_avx512_mask_psra_w_512 : GCCBuiltin<"__builtin_ia32_psraw512_mask">, + Intrinsic<[llvm_v32i16_ty], [llvm_v32i16_ty, + llvm_v8i16_ty, llvm_v32i16_ty, llvm_i32_ty], [IntrNoMem]>; + def int_x86_avx512_mask_psra_wi_128 : GCCBuiltin<"__builtin_ia32_psrawi128_mask">, + Intrinsic<[llvm_v8i16_ty], [llvm_v8i16_ty, + llvm_i8_ty, llvm_v8i16_ty, llvm_i8_ty], [IntrNoMem]>; + def int_x86_avx512_mask_psra_wi_256 : GCCBuiltin<"__builtin_ia32_psrawi256_mask">, + Intrinsic<[llvm_v16i16_ty], [llvm_v16i16_ty, + llvm_i8_ty, llvm_v16i16_ty, llvm_i16_ty], [IntrNoMem]>; + def int_x86_avx512_mask_psra_wi_512 : GCCBuiltin<"__builtin_ia32_psrawi512_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]>; diff --git a/lib/Target/X86/X86IntrinsicsInfo.h b/lib/Target/X86/X86IntrinsicsInfo.h index 5a55e157079..c67844a2efc 100644 --- a/lib/Target/X86/X86IntrinsicsInfo.h +++ b/lib/Target/X86/X86IntrinsicsInfo.h @@ -1215,6 +1215,12 @@ static const IntrinsicData IntrinsicsWithoutChain[] = { X86_INTRINSIC_DATA(avx512_mask_psllv_q, INTR_TYPE_2OP_MASK, ISD::SHL, 0), X86_INTRINSIC_DATA(avx512_mask_psra_d, INTR_TYPE_2OP_MASK, X86ISD::VSRA, 0), X86_INTRINSIC_DATA(avx512_mask_psra_q, INTR_TYPE_2OP_MASK, X86ISD::VSRA, 0), + X86_INTRINSIC_DATA(avx512_mask_psra_w_128, INTR_TYPE_2OP_MASK, X86ISD::VSRA, 0), + X86_INTRINSIC_DATA(avx512_mask_psra_w_256, INTR_TYPE_2OP_MASK, X86ISD::VSRA, 0), + X86_INTRINSIC_DATA(avx512_mask_psra_w_512, INTR_TYPE_2OP_MASK, X86ISD::VSRA, 0), + X86_INTRINSIC_DATA(avx512_mask_psra_wi_128, INTR_TYPE_2OP_MASK, X86ISD::VSRAI, 0), + X86_INTRINSIC_DATA(avx512_mask_psra_wi_256, INTR_TYPE_2OP_MASK, X86ISD::VSRAI, 0), + X86_INTRINSIC_DATA(avx512_mask_psra_wi_512, INTR_TYPE_2OP_MASK, X86ISD::VSRAI, 0), X86_INTRINSIC_DATA(avx512_mask_psrai_d, VSHIFT_MASK, X86ISD::VSRAI, 0), X86_INTRINSIC_DATA(avx512_mask_psrai_q, VSHIFT_MASK, X86ISD::VSRAI, 0), X86_INTRINSIC_DATA(avx512_mask_psrav_d, INTR_TYPE_2OP_MASK, ISD::SRA, 0), diff --git a/test/CodeGen/X86/avx512bw-intrinsics.ll b/test/CodeGen/X86/avx512bw-intrinsics.ll index 2ee84c651e4..064652aa470 100644 --- a/test/CodeGen/X86/avx512bw-intrinsics.ll +++ b/test/CodeGen/X86/avx512bw-intrinsics.ll @@ -2926,3 +2926,43 @@ define <32 x i16>@test_int_x86_avx512_mask_psrlv32hi(<32 x i16> %x0, <32 x i16> %res4 = add <32 x i16> %res3, %res2 ret <32 x i16> %res4 } + +declare <32 x i16> @llvm.x86.avx512.mask.psra.w.512(<32 x i16>, <8 x i16>, <32 x i16>, i32) + +define <32 x i16>@test_int_x86_avx512_mask_psra_w_512(<32 x i16> %x0, <8 x i16> %x1, <32 x i16> %x2, i32 %x3) { +; CHECK-LABEL: test_int_x86_avx512_mask_psra_w_512: +; CHECK: ## BB#0: +; CHECK-NEXT: kmovd %edi, %k1 +; CHECK-NEXT: vpsraw %xmm1, %zmm0, %zmm2 {%k1} +; CHECK-NEXT: vpsraw %xmm1, %zmm0, %zmm3 {%k1} {z} +; CHECK-NEXT: vpsraw %xmm1, %zmm0, %zmm0 +; CHECK-NEXT: vpaddw %zmm3, %zmm2, %zmm1 +; CHECK-NEXT: vpaddw %zmm0, %zmm1, %zmm0 +; CHECK-NEXT: retq + %res = call <32 x i16> @llvm.x86.avx512.mask.psra.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.psra.w.512(<32 x i16> %x0, <8 x i16> %x1, <32 x i16> zeroinitializer, i32 %x3) + %res2 = call <32 x i16> @llvm.x86.avx512.mask.psra.w.512(<32 x i16> %x0, <8 x i16> %x1, <32 x i16> %x2, i32 -1) + %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.psra.wi.512(<32 x i16>, i8, <32 x i16>, i32) + +define <32 x i16>@test_int_x86_avx512_mask_psra_wi_512(<32 x i16> %x0, i8 %x1, <32 x i16> %x2, i32 %x3) { +; CHECK-LABEL: test_int_x86_avx512_mask_psra_wi_512: +; CHECK: ## BB#0: +; CHECK-NEXT: kmovd %esi, %k1 +; CHECK-NEXT: vpsraw $3, %zmm0, %zmm1 {%k1} +; CHECK-NEXT: vpsraw $3, %zmm0, %zmm2 {%k1} {z} +; CHECK-NEXT: vpsraw $3, %zmm0, %zmm0 +; CHECK-NEXT: vpaddw %zmm2, %zmm1, %zmm1 +; CHECK-NEXT: vpaddw %zmm0, %zmm1, %zmm0 +; CHECK-NEXT: retq + %res = call <32 x i16> @llvm.x86.avx512.mask.psra.wi.512(<32 x i16> %x0, i8 3, <32 x i16> %x2, i32 %x3) + %res1 = call <32 x i16> @llvm.x86.avx512.mask.psra.wi.512(<32 x i16> %x0, i8 3, <32 x i16> zeroinitializer, i32 %x3) + %res2 = call <32 x i16> @llvm.x86.avx512.mask.psra.wi.512(<32 x i16> %x0, i8 3, <32 x i16> %x2, i32 -1) + %res3 = add <32 x i16> %res, %res1 + %res4 = add <32 x i16> %res3, %res2 + ret <32 x i16> %res4 +} diff --git a/test/CodeGen/X86/avx512bwvl-intrinsics.ll b/test/CodeGen/X86/avx512bwvl-intrinsics.ll index 09ed302a910..6b2cb432f1c 100644 --- a/test/CodeGen/X86/avx512bwvl-intrinsics.ll +++ b/test/CodeGen/X86/avx512bwvl-intrinsics.ll @@ -4632,3 +4632,85 @@ define <8 x i16>@test_int_x86_avx512_mask_psrlv8_hi(<8 x i16> %x0, <8 x i16> %x1 %res4 = add <8 x i16> %res3, %res2 ret <8 x i16> %res4 } + +declare <8 x i16> @llvm.x86.avx512.mask.psra.w.128(<8 x i16>, <8 x i16>, <8 x i16>, i8) + +define <8 x i16>@test_int_x86_avx512_mask_psra_w_128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3) { +; CHECK-LABEL: test_int_x86_avx512_mask_psra_w_128: +; CHECK: ## BB#0: +; CHECK-NEXT: movzbl %dil, %eax +; CHECK-NEXT: kmovw %eax, %k1 +; CHECK-NEXT: vpsraw %xmm1, %xmm0, %xmm2 {%k1} +; CHECK-NEXT: vpsraw %xmm1, %xmm0, %xmm3 {%k1} {z} +; CHECK-NEXT: vpsraw %xmm1, %xmm0, %xmm0 +; CHECK-NEXT: vpaddw %xmm3, %xmm2, %xmm1 +; CHECK-NEXT: vpaddw %xmm0, %xmm1, %xmm0 +; CHECK-NEXT: retq + %res = call <8 x i16> @llvm.x86.avx512.mask.psra.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.psra.w.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> zeroinitializer, i8 %x3) + %res2 = call <8 x i16> @llvm.x86.avx512.mask.psra.w.128(<8 x i16> %x0, <8 x i16> %x1, <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 <8 x i16> @llvm.x86.avx512.mask.psra.wi.128(<8 x i16>, i8, <8 x i16>, i8) + +define <8 x i16>@test_int_x86_avx512_mask_psra_wi_128(<8 x i16> %x0, i8 %x1, <8 x i16> %x2, i8 %x3) { +; CHECK-LABEL: test_int_x86_avx512_mask_psra_wi_128: +; CHECK: ## BB#0: +; CHECK-NEXT: movzbl %sil, %eax +; CHECK-NEXT: kmovw %eax, %k1 +; CHECK-NEXT: vpsraw $3, %xmm0, %xmm1 {%k1} +; CHECK-NEXT: vpsraw $3, %xmm0, %xmm2 {%k1} {z} +; CHECK-NEXT: vpsraw $3, %xmm0, %xmm0 +; CHECK-NEXT: vpaddw %xmm2, %xmm1, %xmm1 +; CHECK-NEXT: vpaddw %xmm0, %xmm1, %xmm0 +; CHECK-NEXT: retq + %res = call <8 x i16> @llvm.x86.avx512.mask.psra.wi.128(<8 x i16> %x0, i8 3, <8 x i16> %x2, i8 %x3) + %res1 = call <8 x i16> @llvm.x86.avx512.mask.psra.wi.128(<8 x i16> %x0, i8 3, <8 x i16> zeroinitializer, i8 %x3) + %res2 = call <8 x i16> @llvm.x86.avx512.mask.psra.wi.128(<8 x i16> %x0, i8 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.psra.w.256(<16 x i16>, <8 x i16>, <16 x i16>, i16) + +define <16 x i16>@test_int_x86_avx512_mask_psra_w_256(<16 x i16> %x0, <8 x i16> %x1, <16 x i16> %x2, i16 %x3) { +; CHECK-LABEL: test_int_x86_avx512_mask_psra_w_256: +; CHECK: ## BB#0: +; CHECK-NEXT: kmovw %edi, %k1 +; CHECK-NEXT: vpsraw %xmm1, %ymm0, %ymm2 {%k1} +; CHECK-NEXT: vpsraw %xmm1, %ymm0, %ymm3 {%k1} {z} +; CHECK-NEXT: vpsraw %xmm1, %ymm0, %ymm0 +; CHECK-NEXT: vpaddw %ymm3, %ymm2, %ymm1 +; CHECK-NEXT: vpaddw %ymm0, %ymm1, %ymm0 +; CHECK-NEXT: retq + %res = call <16 x i16> @llvm.x86.avx512.mask.psra.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.psra.w.256(<16 x i16> %x0, <8 x i16> %x1, <16 x i16> zeroinitializer, i16 %x3) + %res2 = call <16 x i16> @llvm.x86.avx512.mask.psra.w.256(<16 x i16> %x0, <8 x i16> %x1, <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 <16 x i16> @llvm.x86.avx512.mask.psra.wi.256(<16 x i16>, i8, <16 x i16>, i16) + +define <16 x i16>@test_int_x86_avx512_mask_psra_wi_256(<16 x i16> %x0, i8 %x1, <16 x i16> %x2, i16 %x3) { +; CHECK-LABEL: test_int_x86_avx512_mask_psra_wi_256: +; CHECK: ## BB#0: +; CHECK-NEXT: kmovw %esi, %k1 +; CHECK-NEXT: vpsraw $3, %ymm0, %ymm1 {%k1} +; CHECK-NEXT: vpsraw $3, %ymm0, %ymm2 {%k1} {z} +; CHECK-NEXT: vpsraw $3, %ymm0, %ymm0 +; CHECK-NEXT: vpaddw %ymm2, %ymm1, %ymm1 +; CHECK-NEXT: vpaddw %ymm0, %ymm1, %ymm0 +; CHECK-NEXT: retq + %res = call <16 x i16> @llvm.x86.avx512.mask.psra.wi.256(<16 x i16> %x0, i8 3, <16 x i16> %x2, i16 %x3) + %res1 = call <16 x i16> @llvm.x86.avx512.mask.psra.wi.256(<16 x i16> %x0, i8 3, <16 x i16> zeroinitializer, i16 %x3) + %res2 = call <16 x i16> @llvm.x86.avx512.mask.psra.wi.256(<16 x i16> %x0, i8 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 +} -- 2.34.1