54 lines · c
1// RUN: %clang_cc1 -x c -ffreestanding %s -triple=x86_64-apple-darwin -target-feature +f16c -emit-llvm -ffp-exception-behavior=strict -o - -Wall -Werror | FileCheck %s2// RUN: %clang_cc1 -x c -ffreestanding %s -triple=i386-apple-darwin -target-feature +f16c -emit-llvm -ffp-exception-behavior=strict -o - -Wall -Werror | FileCheck %s3// RUN: %clang_cc1 -x c++ -ffreestanding %s -triple=x86_64-apple-darwin -target-feature +f16c -emit-llvm -ffp-exception-behavior=strict -o - -Wall -Werror | FileCheck %s4// RUN: %clang_cc1 -x c++ -ffreestanding %s -triple=i386-apple-darwin -target-feature +f16c -emit-llvm -ffp-exception-behavior=strict -o - -Wall -Werror | FileCheck %s5 6 7#include <immintrin.h>8 9float test_cvtsh_ss(unsigned short a) {10 // CHECK-LABEL: test_cvtsh_ss11 // CHECK: [[CONV:%.*]] = call {{.*}}float @llvm.experimental.constrained.fpext.f32.f16(half %{{.*}}, metadata !"fpexcept.strict")12 // CHECK: ret float [[CONV]]13 return _cvtsh_ss(a);14}15 16unsigned short test_cvtss_sh(float a) {17 // CHECK-LABEL: test_cvtss_sh18 // CHECK: insertelement <4 x float> poison, float %{{.*}}, i32 019 // CHECK: call float @llvm.experimental.constrained.sitofp.f32.i32(i32 0, metadata !"round.tonearest", metadata !"fpexcept.strict")20 // CHECK: insertelement <4 x float> %{{.*}}, float %{{.*}}, i32 121 // CHECK: call float @llvm.experimental.constrained.sitofp.f32.i32(i32 0, metadata !"round.tonearest", metadata !"fpexcept.strict")22 // CHECK: insertelement <4 x float> %{{.*}}, float %{{.*}}, i32 223 // CHECK: call float @llvm.experimental.constrained.sitofp.f32.i32(i32 0, metadata !"round.tonearest", metadata !"fpexcept.strict")24 // CHECK: insertelement <4 x float> %{{.*}}, float %{{.*}}, i32 325 // CHECK: call <8 x i16> @llvm.x86.vcvtps2ph.128(<4 x float> %{{.*}}, i32 0)26 // CHECK: extractelement <8 x i16> %{{.*}}, i32 027 return _cvtss_sh(a, 0);28}29 30__m128 test_mm_cvtph_ps(__m128i a) {31 // CHECK-LABEL: test_mm_cvtph_ps32 // CHECK: shufflevector <8 x i16> %{{.*}}, <8 x i16> %{{.*}}, <4 x i32> <i32 0, i32 1, i32 2, i32 3>33 // CHECK: call {{.*}}<4 x float> @llvm.experimental.constrained.fpext.v4f32.v4f16(<4 x half> %{{.*}}, metadata !"fpexcept.strict")34 return _mm_cvtph_ps(a);35}36 37__m256 test_mm256_cvtph_ps(__m128i a) {38 // CHECK-LABEL: test_mm256_cvtph_ps39 // CHECK: call {{.*}}<8 x float> @llvm.experimental.constrained.fpext.v8f32.v8f16(<8 x half> %{{.*}}, metadata !"fpexcept.strict")40 return _mm256_cvtph_ps(a);41}42 43__m128i test_mm_cvtps_ph(__m128 a) {44 // CHECK-LABEL: test_mm_cvtps_ph45 // CHECK: call <8 x i16> @llvm.x86.vcvtps2ph.128(<4 x float> %{{.*}}, i32 0)46 return _mm_cvtps_ph(a, 0);47}48 49__m128i test_mm256_cvtps_ph(__m256 a) {50 // CHECK-LABEL: test_mm256_cvtps_ph51 // CHECK: call <8 x i16> @llvm.x86.vcvtps2ph.256(<8 x float> %{{.*}}, i32 0)52 return _mm256_cvtps_ph(a, 0);53}54