1 // RUN: %clang_cc1 -ffreestanding %s -O3 -triple=x86_64-apple-darwin -target-feature +avx -emit-llvm -o - | FileCheck %s
2 // FIXME: This is testing optimized generation of shuffle instructions and should be fixed.
3
4
5 #include <immintrin.h>
6
7 //
8 // Test LLVM IR codegen of shuffle instructions
test_mm256_add_pd(__m256d A,__m256d B)9 //
10
11 __m256 x(__m256 a, __m256 b) {
12 // Check if the mask is correct
13 // CHECK: shufflevector{{.*}}<i32 3, i32 2, i32 8, i32 11, i32 7, i32 6, i32 12, i32 15>
14 return _mm256_shuffle_ps(a, b, 203);
15 }
16
17 __m128d test_mm_permute_pd(__m128d a) {
18 // Check if the mask is correct
19 // CHECK: shufflevector{{.*}}<i32 1, i32 0>
20 return _mm_permute_pd(a, 1);
test_mm256_addsub_pd(__m256d A,__m256d B)21 }
22
23 __m256d test_mm256_permute_pd(__m256d a) {
24 // Check if the mask is correct
25 // CHECK: shufflevector{{.*}}<i32 1, i32 0, i32 3, i32 2>
26 return _mm256_permute_pd(a, 5);
27 }
28
29 __m128 test_mm_permute_ps(__m128 a) {
30 // Check if the mask is correct
31 // CHECK: shufflevector{{.*}}<i32 3, i32 2, i32 1, i32 0>
32 return _mm_permute_ps(a, 0x1b);
test_mm256_and_pd(__m256d A,__m256d B)33 }
34
35 // Test case for PR12401
36 __m128 test_mm_permute_ps2(__m128 a) {
37 // Check if the mask is correct
38 // CHECK: shufflevector{{.*}}<i32 2, i32 1, i32 2, i32 3>
39 return _mm_permute_ps(a, 0xe6);
40 }
41
42 __m256 test_mm256_permute_ps(__m256 a) {
43 // Check if the mask is correct
44 // CHECK: shufflevector{{.*}}<i32 3, i32 2, i32 1, i32 0, i32 7, i32 6, i32 5, i32 4>
test_mm256_andnot_pd(__m256d A,__m256d B)45 return _mm256_permute_ps(a, 0x1b);
46 }
47
48 __m256d test_mm256_permute2f128_pd(__m256d a, __m256d b) {
49 // Check if the mask is correct
50 // CHECK: shufflevector{{.*}}<i32 2, i32 3, i32 6, i32 7>
51 return _mm256_permute2f128_pd(a, b, 0x31);
test_mm256_andnot_ps(__m256 A,__m256 B)52 }
53
54 __m256 test_mm256_permute2f128_ps(__m256 a, __m256 b) {
55 // Check if the mask is correct
56 // CHECK: shufflevector{{.*}}<i32 4, i32 5, i32 6, i32 7, i32 12, i32 13, i32 14, i32 15>
57 return _mm256_permute2f128_ps(a, b, 0x13);
58 }
test_mm256_blend_pd(__m256d A,__m256d B)59
60 __m256i test_mm256_permute2f128_si256(__m256i a, __m256i b) {
61 // Check if the mask is correct
62 // CHECK: shufflevector{{.*}} <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 8, i32 9, i32 10, i32 11>
63 return _mm256_permute2f128_si256(a, b, 0x20);
64 }
test_mm256_blend_ps(__m256 A,__m256 B)65
66 __m128
67 test_mm_broadcast_ss(float const *__a) {
68 // CHECK-LABEL: @test_mm_broadcast_ss
69 // CHECK: insertelement <4 x float> {{.*}}, i32 0
70 // CHECK: shufflevector <4 x float> {{.*}}, <4 x float> undef, <4 x i32> zeroinitializer
71 return _mm_broadcast_ss(__a);
72 }
73
74 __m256d
75 test_mm256_broadcast_sd(double const *__a) {
76 // CHECK-LABEL: @test_mm256_broadcast_sd
test_mm256_blendv_ps(__m256 V1,__m256 V2,__m256 V3)77 // CHECK: insertelement <4 x double> {{.*}}, i32 0
78 // CHECK: shufflevector <4 x double> {{.*}}, <4 x double> undef, <4 x i32> zeroinitializer
79 return _mm256_broadcast_sd(__a);
80 }
81
82 __m256
test_mm256_broadcast_pd(__m128d * A)83 test_mm256_broadcast_ss(float const *__a) {
84 // CHECK-LABEL: @test_mm256_broadcast_ss
85 // CHECK: insertelement <8 x float> {{.*}}, i32 0
86 // CHECK: shufflevector <8 x float> {{.*}}, <8 x float> undef, <8 x i32> zeroinitializer
87 return _mm256_broadcast_ss(__a);
88 }
89
test_mm256_broadcast_ps(__m128 * A)90 // Make sure we have the correct mask for each insertf128 case.
91
92 __m256 test_mm256_insertf128_ps_0(__m256 a, __m128 b) {
93 // CHECK-LABEL: @test_mm256_insertf128_ps_0
94 // CHECK: shufflevector{{.*}}<i32 8, i32 9, i32 10, i32 11, i32 4, i32 5, i32 6, i32 7>
95 return _mm256_insertf128_ps(a, b, 0);
96 }
test_mm256_broadcast_sd(double * A)97
98 __m256d test_mm256_insertf128_pd_0(__m256d a, __m128d b) {
99 // CHECK-LABEL: @test_mm256_insertf128_pd_0
100 // CHECK: shufflevector{{.*}}<i32 4, i32 5, i32 2, i32 3>
101 return _mm256_insertf128_pd(a, b, 0);
102 }
103
104 __m256i test_mm256_insertf128_si256_0(__m256i a, __m128i b) {
105 // CHECK-LABEL: @test_mm256_insertf128_si256_0
106 // CHECK: shufflevector{{.*}}<i32 8, i32 9, i32 10, i32 11, i32 4, i32 5, i32 6, i32 7>
test_mm_broadcast_ss(float * A)107 return _mm256_insertf128_si256(a, b, 0);
108 }
109
110 __m256 test_mm256_insertf128_ps_1(__m256 a, __m128 b) {
111 // CHECK-LABEL: @test_mm256_insertf128_ps_1
112 // CHECK: shufflevector{{.*}}<i32 0, i32 1, i32 2, i32 3, i32 8, i32 9, i32 10, i32 11>
113 return _mm256_insertf128_ps(a, b, 1);
114 }
115
116 __m256d test_mm256_insertf128_pd_1(__m256d a, __m128d b) {
test_mm256_broadcast_ss(float * A)117 // CHECK-LABEL: @test_mm256_insertf128_pd_1
118 // CHECK: shufflevector{{.*}}<i32 0, i32 1, i32 4, i32 5>
119 return _mm256_insertf128_pd(a, b, 1);
120 }
121
122 __m256i test_mm256_insertf128_si256_1(__m256i a, __m128i b) {
123 // CHECK-LABEL: @test_mm256_insertf128_si256_1
124 // CHECK: shufflevector{{.*}}<i32 0, i32 1, i32 2, i32 3, i32 8, i32 9, i32 10, i32 11>
125 return _mm256_insertf128_si256(a, b, 1);
126 }
127
128 // Make sure we have the correct mask for each extractf128 case.
129
130 __m128 test_mm256_extractf128_ps_0(__m256 a) {
test_mm256_castpd_ps(__m256d A)131 // CHECK-LABEL: @test_mm256_extractf128_ps_0
132 // CHECK: shufflevector{{.*}}<i32 0, i32 1, i32 2, i32 3>
133 return _mm256_extractf128_ps(a, 0);
134 }
135
136 __m128d test_mm256_extractf128_pd_0(__m256d a) {
test_mm256_castpd_si256(__m256d A)137 // CHECK-LABEL: @test_mm256_extractf128_pd_0
138 // CHECK: shufflevector{{.*}}<i32 0, i32 1>
139 return _mm256_extractf128_pd(a, 0);
140 }
141
142 __m128i test_mm256_extractf128_si256_0(__m256i a) {
test_mm256_castpd128_pd256(__m128d A)143 // CHECK-LABEL: @test_mm256_extractf128_si256_0
144 // CHECK: shufflevector{{.*}}<i32 0, i32 1, i32 2, i32 3>
145 return _mm256_extractf128_si256(a, 0);
146 }
147
148 __m128 test_mm256_extractf128_ps_1(__m256 a) {
test_mm256_castpd256_pd128(__m256d A)149 // CHECK-LABEL: @test_mm256_extractf128_ps_1
150 // CHECK: shufflevector{{.*}}<i32 4, i32 5, i32 6, i32 7>
151 return _mm256_extractf128_ps(a, 1);
152 }
153
154 __m128d test_mm256_extractf128_pd_1(__m256d a) {
test_mm256_castps_pd(__m256 A)155 // CHECK-LABEL: @test_mm256_extractf128_pd_1
156 // CHECK: shufflevector{{.*}}<i32 2, i32 3>
157 return _mm256_extractf128_pd(a, 1);
158 }
159
160 __m128i test_mm256_extractf128_si256_1(__m256i a) {
test_mm256_castps_si256(__m256 A)161 // CHECK-LABEL: @test_mm256_extractf128_si256_1
162 // CHECK: shufflevector{{.*}}<i32 4, i32 5, i32 6, i32 7>
163 return _mm256_extractf128_si256(a, 1);
164 }
165
166 __m256 test_mm256_set_m128(__m128 hi, __m128 lo) {
test_mm256_castps128_ps256(__m128 A)167 // CHECK-LABEL: @test_mm256_set_m128
168 // CHECK: shufflevector{{.*}}<i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
169 return _mm256_set_m128(hi, lo);
170 }
171
172 __m256d test_mm256_set_m128d(__m128d hi, __m128d lo) {
test_mm256_castps256_ps128(__m256 A)173 // CHECK-LABEL: @test_mm256_set_m128d
174 // CHECK: shufflevector{{.*}}<i32 0, i32 1, i32 2, i32 3>
175 return _mm256_set_m128d(hi, lo);
176 }
177
178 __m256i test_mm256_set_m128i(__m128i hi, __m128i lo) {
test_mm256_castsi128_si256(__m128i A)179 // CHECK-LABEL: @test_mm256_set_m128i
180 // CHECK: shufflevector{{.*}}<i32 0, i32 1, i32 2, i32 3>
181 return _mm256_set_m128i(hi, lo);
182 }
183
184 __m256 test_mm256_setr_m128(__m128 hi, __m128 lo) {
test_mm256_castsi256_pd(__m256i A)185 // CHECK-LABEL: @test_mm256_setr_m128
186 // CHECK: shufflevector{{.*}}<i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
187 return _mm256_setr_m128(lo, hi);
188 }
189
190 __m256d test_mm256_setr_m128d(__m128d hi, __m128d lo) {
test_mm256_castsi256_ps(__m256i A)191 // CHECK-LABEL: @test_mm256_setr_m128d
192 // CHECK: shufflevector{{.*}}<i32 0, i32 1, i32 2, i32 3>
193 return _mm256_setr_m128d(lo, hi);
194 }
195
196 __m256i test_mm256_setr_m128i(__m128i hi, __m128i lo) {
test_mm256_castsi256_si128(__m256i A)197 // CHECK-LABEL: @test_mm256_setr_m128i
198 // CHECK: shufflevector{{.*}}<i32 0, i32 1, i32 2, i32 3>
199 return _mm256_setr_m128i(lo, hi);
200 }
201