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
test_cvtsh_ss(unsigned short a)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> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
17 // CHECK: bitcast <4 x i16> %{{.*}} to <4 x half>
18 // CHECK: call <4 x float> @llvm.experimental.constrained.fpext.v4f32.v4f16(<4 x half> %{{.*}}, metadata !"fpexcept.strict")
19 // CHECK: extractelement <4 x float> %{{.*}}, i32 0
20 return _cvtsh_ss(a);
21 }
22
test_cvtss_sh(float a)23 unsigned short test_cvtss_sh(float a) {
24 // CHECK-LABEL: test_cvtss_sh
25 // CHECK: insertelement <4 x float> undef, float %{{.*}}, i32 0
26 // CHECK: call float @llvm.experimental.constrained.sitofp.f32.i32(i32 0, metadata !"round.tonearest", metadata !"fpexcept.strict")
27 // CHECK: insertelement <4 x float> %{{.*}}, float %{{.*}}, i32 1
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 2
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 3
32 // CHECK: call <8 x i16> @llvm.x86.vcvtps2ph.128(<4 x float> %{{.*}}, i32 0)
33 // CHECK: extractelement <8 x i16> %{{.*}}, i32 0
34 return _cvtss_sh(a, 0);
35 }
36
test_mm_cvtph_ps(__m128i a)37 __m128 test_mm_cvtph_ps(__m128i a) {
38 // CHECK-LABEL: test_mm_cvtph_ps
39 // CHECK: bitcast <2 x i64> %{{.*}} to <8 x i16>
40 // CHECK: shufflevector <8 x i16> %{{.*}}, <8 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
41 // CHECK: bitcast <4 x i16> %{{.*}} to <4 x half>
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
test_mm256_cvtph_ps(__m128i a)46 __m256 test_mm256_cvtph_ps(__m128i a) {
47 // CHECK-LABEL: test_mm256_cvtph_ps
48 // CHECK: bitcast <2 x i64> %{{.*}} to <8 x i16>
49 // CHECK: bitcast <8 x i16> %{{.*}} to <8 x half>
50 // CHECK: call <8 x float> @llvm.experimental.constrained.fpext.v8f32.v8f16(<8 x half> %{{.*}}, metadata !"fpexcept.strict")
51 return _mm256_cvtph_ps(a);
52 }
53
test_mm_cvtps_ph(__m128 a)54 __m128i test_mm_cvtps_ph(__m128 a) {
55 // CHECK-LABEL: test_mm_cvtps_ph
56 // CHECK: call <8 x i16> @llvm.x86.vcvtps2ph.128(<4 x float> %{{.*}}, i32 0)
57 return _mm_cvtps_ph(a, 0);
58 }
59
test_mm256_cvtps_ph(__m256 a)60 __m128i test_mm256_cvtps_ph(__m256 a) {
61 // CHECK-LABEL: test_mm256_cvtps_ph
62 // CHECK: call <8 x i16> @llvm.x86.vcvtps2ph.256(<8 x float> %{{.*}}, i32 0)
63 return _mm256_cvtps_ph(a, 0);
64 }
65