From 97cf2229cd96da1a3d2c0839ee59080b152b690e Mon Sep 17 00:00:00 2001 From: "Ivan A. Kosarev" Date: Fri, 13 Apr 2018 12:46:02 +0000 Subject: [PATCH] [NEON] Support vrndns_f32 intrinsic Differential Revision: https://reviews.llvm.org/D45515 git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@330012 91177308-0d34-0410-b5e6-96231b3b80d8 --- include/clang/Basic/arm_neon.td | 6 ++++++ lib/CodeGen/CGBuiltin.cpp | 6 ++++++ test/CodeGen/arm-neon-directed-rounding.c | 11 ++++++++++- 3 files changed, 22 insertions(+), 1 deletion(-) diff --git a/include/clang/Basic/arm_neon.td b/include/clang/Basic/arm_neon.td index 94d39e92f2..afa2d32ef0 100644 --- a/include/clang/Basic/arm_neon.td +++ b/include/clang/Basic/arm_neon.td @@ -1115,6 +1115,12 @@ def SCALAR_FCVTZS_N_S64 : SInst<"vcvt_n_s64", "$si", "Sd">; def SCALAR_FCVTZU_N_U64 : SInst<"vcvt_n_u64", "bsi", "Sd">; } +//////////////////////////////////////////////////////////////////////////////// +// Scalar Floating-point Round to Integral +let ArchGuard = "__ARM_ARCH >= 8 && defined(__ARM_FEATURE_DIRECTED_ROUNDING)" in { +def SCALAR_FRINTN_S32 : SInst<"vrndn", "ss", "Sf">; +} + //////////////////////////////////////////////////////////////////////////////// // Scalar Reduce Pairwise Addition (Scalar and Floating Point) def SCALAR_ADDP : SInst<"vpadd", "sd", "SfSHlSHdSHUl">; diff --git a/lib/CodeGen/CGBuiltin.cpp b/lib/CodeGen/CGBuiltin.cpp index 4baa5aff72..e194a5de5e 100644 --- a/lib/CodeGen/CGBuiltin.cpp +++ b/lib/CodeGen/CGBuiltin.cpp @@ -5590,6 +5590,12 @@ Value *CodeGenFunction::EmitARMBuiltinExpr(unsigned BuiltinID, case NEON::BI__builtin_neon_vgetq_lane_f32: return Builder.CreateExtractElement(Ops[0], Ops[1], "vget_lane"); + case NEON::BI__builtin_neon_vrndns_f32: { + Value *Arg = EmitScalarExpr(E->getArg(0)); + llvm::Type *Tys[] = {Arg->getType()}; + Function *F = CGM.getIntrinsic(Intrinsic::arm_neon_vrintn, Tys); + return Builder.CreateCall(F, {Arg}, "vrndn"); } + case NEON::BI__builtin_neon_vset_lane_i8: case NEON::BI__builtin_neon_vset_lane_i16: case NEON::BI__builtin_neon_vset_lane_i32: diff --git a/test/CodeGen/arm-neon-directed-rounding.c b/test/CodeGen/arm-neon-directed-rounding.c index b06808a77c..9f30ac7cb0 100644 --- a/test/CodeGen/arm-neon-directed-rounding.c +++ b/test/CodeGen/arm-neon-directed-rounding.c @@ -1,4 +1,6 @@ -// RUN: %clang_cc1 -triple thumbv8-linux-gnueabihf -target-cpu cortex-a57 -ffreestanding -disable-O0-optnone -emit-llvm %s -o - | opt -S -mem2reg | FileCheck %s +// RUN: %clang_cc1 -triple thumbv8-linux-gnueabihf -target-cpu cortex-a57 \ +// RUN: -ffreestanding -disable-O0-optnone -emit-llvm %s -o - | \ +// RUN: opt -S -mem2reg | FileCheck %s #include @@ -85,3 +87,10 @@ float32x2_t test_vrnd_f32(float32x2_t a) { float32x4_t test_vrndq_f32(float32x4_t a) { return vrndq_f32(a); } + +// CHECK-LABEL: define float @test_vrndns_f32(float %a) #0 { +// CHECK: [[VRNDN_I:%.*]] = call float @llvm.arm.neon.vrintn.f32(float %a) #2 +// CHECK: ret float [[VRNDN_I]] +float32_t test_vrndns_f32(float32_t a) { + return vrndns_f32(a); +} -- 2.40.0