clang/test/CodeGen/X86/f16c-builtins-constrained.c

   1 // RUN: %clang_cc1 -ffreestanding %s -triple=x86_64-apple-darwin -target-feature +f16c -emit-llvm -ffp-exception-behavior=strict -o - -Wall -Werror | FileCheck %s
   2
   3
   4 #include <immintrin.h>
   5
   6 float test_cvtsh_ss(unsigned short a) {
   7   // CHECK-LABEL: test_cvtsh_ss
   8   // CHECK: insertelement <8 x i16> undef, i16 %{{.*}}, i32 0
   9   // CHECK: insertelement <8 x i16> %{{.*}}, i16 0, i32 1
  10   // CHECK: insertelement <8 x i16> %{{.*}}, i16 0, i32 2
  11   // CHECK: insertelement <8 x i16> %{{.*}}, i16 0, i32 3
  12   // CHECK: insertelement <8 x i16> %{{.*}}, i16 0, i32 4
  13   // CHECK: insertelement <8 x i16> %{{.*}}, i16 0, i32 5
  14   // CHECK: insertelement <8 x i16> %{{.*}}, i16 0, i32 6
  15   // CHECK: insertelement <8 x i16> %{{.*}}, i16 0, i32 7
  16   // CHECK: shufflevector <8 x i16> %{{.*}}, <8 x i16> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  17   // CHECK: call <4 x float> @llvm.experimental.constrained.fpext.v4f32.v4f16(<4 x half> %{{.*}}, metadata !"fpexcept.strict")
  18   // CHECK: extractelement <4 x float> %{{.*}}, i32 0
  19   return _cvtsh_ss(a);
  20 }
  21
  22 unsigned short test_cvtss_sh(float a) {
  23   // CHECK-LABEL: test_cvtss_sh
  24   // CHECK: insertelement <4 x float> undef, float %{{.*}}, i32 0
  25   // CHECK: call float @llvm.experimental.constrained.sitofp.f32.i32(i32 0, metadata !"round.tonearest", metadata !"fpexcept.strict")
  26   // CHECK: insertelement <4 x float> %{{.*}}, float %{{.*}}, i32 1
  27   // CHECK: call float @llvm.experimental.constrained.sitofp.f32.i32(i32 0, metadata !"round.tonearest", metadata !"fpexcept.strict")
  28   // CHECK: insertelement <4 x float> %{{.*}}, float %{{.*}}, i32 2
  29   // CHECK: call float @llvm.experimental.constrained.sitofp.f32.i32(i32 0, metadata !"round.tonearest", metadata !"fpexcept.strict")
  30   // CHECK: insertelement <4 x float> %{{.*}}, float %{{.*}}, i32 3
  31   // CHECK: call <8 x i16> @llvm.x86.vcvtps2ph.128(<4 x float> %{{.*}}, i32 0)
  32   // CHECK: extractelement <8 x i16> %{{.*}}, i32 0
  33   return _cvtss_sh(a, 0);
  34 }
  35
  36 __m128 test_mm_cvtph_ps(__m128i a) {
  37   // CHECK-LABEL: test_mm_cvtph_ps
  38   // CHECK: shufflevector <8 x i16> %{{.*}}, <8 x i16> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
  39   // CHECK: call <4 x float> @llvm.experimental.constrained.fpext.v4f32.v4f16(<4 x half> %{{.*}}, metadata !"fpexcept.strict")
  40   return _mm_cvtph_ps(a);
  41 }
  42
  43 __m256 test_mm256_cvtph_ps(__m128i a) {
  44   // CHECK-LABEL: test_mm256_cvtph_ps
  45   // CHECK: call <8 x float> @llvm.experimental.constrained.fpext.v8f32.v8f16(<8 x half> %{{.*}}, metadata !"fpexcept.strict")
  46   return _mm256_cvtph_ps(a);
  47 }
  48
  49 __m128i test_mm_cvtps_ph(__m128 a) {
  50   // CHECK-LABEL: test_mm_cvtps_ph
  51   // CHECK: call <8 x i16> @llvm.x86.vcvtps2ph.128(<4 x float> %{{.*}}, i32 0)
  52   return _mm_cvtps_ph(a, 0);
  53 }
  54
  55 __m128i test_mm256_cvtps_ph(__m256 a) {
  56   // CHECK-LABEL: test_mm256_cvtps_ph
  57   // CHECK: call <8 x i16> @llvm.x86.vcvtps2ph.256(<8 x float> %{{.*}}, i32 0)
  58   return _mm256_cvtps_ph(a, 0);
  59 }