clang/test/CodeGen/X86/f16c-builtins-constrained.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 %s
   2 // RUN: %clang_cc1 -x c -ffreestanding %s -triple=i386-apple-darwin -target-feature +f16c -emit-llvm -ffp-exception-behavior=strict -o - -Wall -Werror | FileCheck %s
   3 // 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 %s
   4 // RUN: %clang_cc1 -x c++ -ffreestanding %s -triple=i386-apple-darwin -target-feature +f16c -emit-llvm -ffp-exception-behavior=strict -o - -Wall -Werror | FileCheck %s
   5
   6
   7 #include <immintrin.h>
   8
   9 float test_cvtsh_ss(unsigned short a) {
  10   // CHECK-LABEL: test_cvtsh_ss
  11   // CHECK: insertelement <8 x i16> poison, i16 %{{.*}}, i32 0
  12   // CHECK: insertelement <8 x i16> %{{.*}}, i16 0, i32 1
  13   // CHECK: insertelement <8 x i16> %{{.*}}, i16 0, i32 2
  14   // CHECK: insertelement <8 x i16> %{{.*}}, i16 0, i32 3
  15   // CHECK: insertelement <8 x i16> %{{.*}}, i16 0, i32 4
  16   // CHECK: insertelement <8 x i16> %{{.*}}, i16 0, i32 5
  17   // CHECK: insertelement <8 x i16> %{{.*}}, i16 0, i32 6
  18   // CHECK: insertelement <8 x i16> %{{.*}}, i16 0, i32 7
  19   // CHECK: shufflevector <8 x i16> %{{.*}}, <8 x i16> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  20   // CHECK: call <4 x float> @llvm.experimental.constrained.fpext.v4f32.v4f16(<4 x half> %{{.*}}, metadata !"fpexcept.strict")
  21   // CHECK: extractelement <4 x float> %{{.*}}, i32 0
  22   return _cvtsh_ss(a);
  23 }
  24
  25 unsigned short test_cvtss_sh(float a) {
  26   // CHECK-LABEL: test_cvtss_sh
  27   // CHECK: insertelement <4 x float> poison, float %{{.*}}, i32 0
  28   // CHECK: call float @llvm.experimental.constrained.sitofp.f32.i32(i32 0, metadata !"round.tonearest", metadata !"fpexcept.strict")
  29   // CHECK: insertelement <4 x float> %{{.*}}, float %{{.*}}, i32 1
  30   // CHECK: call float @llvm.experimental.constrained.sitofp.f32.i32(i32 0, metadata !"round.tonearest", metadata !"fpexcept.strict")
  31   // CHECK: insertelement <4 x float> %{{.*}}, float %{{.*}}, i32 2
  32   // CHECK: call float @llvm.experimental.constrained.sitofp.f32.i32(i32 0, metadata !"round.tonearest", metadata !"fpexcept.strict")
  33   // CHECK: insertelement <4 x float> %{{.*}}, float %{{.*}}, i32 3
  34   // CHECK: call <8 x i16> @llvm.x86.vcvtps2ph.128(<4 x float> %{{.*}}, i32 0)
  35   // CHECK: extractelement <8 x i16> %{{.*}}, i32 0
  36   return _cvtss_sh(a, 0);
  37 }
  38
  39 __m128 test_mm_cvtph_ps(__m128i a) {
  40   // CHECK-LABEL: test_mm_cvtph_ps
  41   // CHECK: shufflevector <8 x i16> %{{.*}}, <8 x i16> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  42   // CHECK: call {{.*}}<4 x float> @llvm.experimental.constrained.fpext.v4f32.v4f16(<4 x half> %{{.*}}, metadata !"fpexcept.strict")
  43   return _mm_cvtph_ps(a);
  44 }
  45
  46 __m256 test_mm256_cvtph_ps(__m128i a) {
  47   // CHECK-LABEL: test_mm256_cvtph_ps
  48   // CHECK: call {{.*}}<8 x float> @llvm.experimental.constrained.fpext.v8f32.v8f16(<8 x half> %{{.*}}, metadata !"fpexcept.strict")
  49   return _mm256_cvtph_ps(a);
  50 }
  51
  52 __m128i test_mm_cvtps_ph(__m128 a) {
  53   // CHECK-LABEL: test_mm_cvtps_ph
  54   // CHECK: call <8 x i16> @llvm.x86.vcvtps2ph.128(<4 x float> %{{.*}}, i32 0)
  55   return _mm_cvtps_ph(a, 0);
  56 }
  57
  58 __m128i test_mm256_cvtps_ph(__m256 a) {
  59   // CHECK-LABEL: test_mm256_cvtps_ph
  60   // CHECK: call <8 x i16> @llvm.x86.vcvtps2ph.256(<8 x float> %{{.*}}, i32 0)
  61   return _mm256_cvtps_ph(a, 0);
  62 }