1 // RUN: %clang_cc1 %s -O3 -triple=x86_64-apple-darwin -target-feature +avx -emit-llvm -o - | FileCheck %s
2
3 // Don't include mm_malloc.h, it's system specific.
4 #define __MM_MALLOC_H
5
6 #include <immintrin.h>
7
8 //
9 // Test LLVM IR codegen of shuffle instructions
10 //
11
x(__m256 a,__m256 b)12 __m256 x(__m256 a, __m256 b) {
13 // Check if the mask is correct
14 // CHECK: shufflevector{{.*}}<i32 3, i32 2, i32 8, i32 11, i32 7, i32 6, i32 12, i32 15>
15 return _mm256_shuffle_ps(a, b, 203);
16 }
17
test_mm_permute_pd(__m128d a)18 __m128d test_mm_permute_pd(__m128d a) {
19 // Check if the mask is correct
20 // CHECK: shufflevector{{.*}}<i32 1, i32 0>
21 return _mm_permute_pd(a, 1);
22 }
23
test_mm256_permute_pd(__m256d a)24 __m256d test_mm256_permute_pd(__m256d a) {
25 // Check if the mask is correct
26 // CHECK: shufflevector{{.*}}<i32 1, i32 0, i32 3, i32 2>
27 return _mm256_permute_pd(a, 5);
28 }
29
test_mm_permute_ps(__m128 a)30 __m128 test_mm_permute_ps(__m128 a) {
31 // Check if the mask is correct
32 // CHECK: shufflevector{{.*}}<i32 3, i32 2, i32 1, i32 0>
33 return _mm_permute_ps(a, 0x1b);
34 }
35
36 // Test case for PR12401
test_mm_permute_ps2(__m128 a)37 __m128 test_mm_permute_ps2(__m128 a) {
38 // Check if the mask is correct
39 // CHECK: shufflevector{{.*}}<i32 2, i32 1, i32 2, i32 3>
40 return _mm_permute_ps(a, 0xe6);
41 }
42
test_mm256_permute_ps(__m256 a)43 __m256 test_mm256_permute_ps(__m256 a) {
44 // Check if the mask is correct
45 // CHECK: shufflevector{{.*}}<i32 3, i32 2, i32 1, i32 0, i32 7, i32 6, i32 5, i32 4>
46 return _mm256_permute_ps(a, 0x1b);
47 }
48
test_mm256_permute2f128_pd(__m256d a,__m256d b)49 __m256d test_mm256_permute2f128_pd(__m256d a, __m256d b) {
50 // Check if the mask is correct
51 // CHECK: @llvm.x86.avx.vperm2f128.pd.256
52 return _mm256_permute2f128_pd(a, b, 0x31);
53 }
54
test_mm256_permute2f128_ps(__m256 a,__m256 b)55 __m256 test_mm256_permute2f128_ps(__m256 a, __m256 b) {
56 // Check if the mask is correct
57 // CHECK: @llvm.x86.avx.vperm2f128.ps.256
58 return _mm256_permute2f128_ps(a, b, 0x13);
59 }
60
test_mm256_permute2f128_si256(__m256i a,__m256i b)61 __m256i test_mm256_permute2f128_si256(__m256i a, __m256i b) {
62 // Check if the mask is correct
63 // CHECK: @llvm.x86.avx.vperm2f128.si.256
64 return _mm256_permute2f128_si256(a, b, 0x20);
65 }
66
67 __m128
test_mm_broadcast_ss(float const * __a)68 test_mm_broadcast_ss(float const *__a) {
69 // CHECK-LABEL: @test_mm_broadcast_ss
70 // CHECK: insertelement <4 x float> {{.*}}, i32 0
71 // CHECK: insertelement <4 x float> {{.*}}, i32 1
72 // CHECK: insertelement <4 x float> {{.*}}, i32 2
73 // CHECK: insertelement <4 x float> {{.*}}, i32 3
74 return _mm_broadcast_ss(__a);
75 }
76
77 __m256d
test_mm256_broadcast_sd(double const * __a)78 test_mm256_broadcast_sd(double const *__a) {
79 // CHECK-LABEL: @test_mm256_broadcast_sd
80 // CHECK: insertelement <4 x double> {{.*}}, i32 0
81 // CHECK: insertelement <4 x double> {{.*}}, i32 1
82 // CHECK: insertelement <4 x double> {{.*}}, i32 2
83 // CHECK: insertelement <4 x double> {{.*}}, i32 3
84 return _mm256_broadcast_sd(__a);
85 }
86
87 __m256
test_mm256_broadcast_ss(float const * __a)88 test_mm256_broadcast_ss(float const *__a) {
89 // CHECK-LABEL: @test_mm256_broadcast_ss
90 // CHECK: insertelement <8 x float> {{.*}}, i32 0
91 // CHECK: insertelement <8 x float> {{.*}}, i32 1
92 // CHECK: insertelement <8 x float> {{.*}}, i32 2
93 // CHECK: insertelement <8 x float> {{.*}}, i32 3
94 // CHECK: insertelement <8 x float> {{.*}}, i32 4
95 // CHECK: insertelement <8 x float> {{.*}}, i32 5
96 // CHECK: insertelement <8 x float> {{.*}}, i32 6
97 // CHECK: insertelement <8 x float> {{.*}}, i32 7
98 return _mm256_broadcast_ss(__a);
99 }
100