clang/test/CodeGen/X86/avx-shuffle-builtins.c

   1 // RUN: %clang_cc1 -ffreestanding %s -O3 -triple=x86_64-apple-darwin -target-feature +avx -emit-llvm -o - | FileCheck %s --check-prefixes=CHECK,X64
   2 // RUN: %clang_cc1 -ffreestanding %s -O3 -triple=i386-apple-darwin -target-feature +avx -emit-llvm -o - | FileCheck %s --check-prefixes=CHECK,X86
   3 // FIXME: This is testing optimized generation of shuffle instructions and should be fixed.
   4
   5
   6 #include <immintrin.h>
   7
   8 //
   9 // Test LLVM IR codegen of shuffle instructions, checking if the masks are correct
  10 //
  11
  12 __m256 x(__m256 a, __m256 b) {
  13   // CHECK-LABEL: x
  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
  18 __m128d test_mm_permute_pd(__m128d a) {
  19   // CHECK-LABEL: test_mm_permute_pd
  20   // CHECK: shufflevector{{.*}}<i32 1, i32 0>
  21   return _mm_permute_pd(a, 1);
  22 }
  23
  24 __m256d test_mm256_permute_pd(__m256d a) {
  25   // CHECK-LABEL: test_mm256_permute_pd
  26   // CHECK: shufflevector{{.*}}<i32 1, i32 0, i32 3, i32 2>
  27   return _mm256_permute_pd(a, 5);
  28 }
  29
  30 __m128 test_mm_permute_ps(__m128 a) {
  31   // CHECK-LABEL: test_mm_permute_ps
  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
  37 __m128 test_mm_permute_ps2(__m128 a) {
  38   // CHECK-LABEL: test_mm_permute_ps2
  39   // CHECK: shufflevector{{.*}}<i32 2, i32 1, i32 2, i32 3>
  40   return _mm_permute_ps(a, 0xe6);
  41 }
  42
  43 __m256 test_mm256_permute_ps(__m256 a) {
  44   // CHECK-LABEL: test_mm256_permute_ps
  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
  49 __m256d test_mm256_permute2f128_pd(__m256d a, __m256d b) {
  50   // CHECK-LABEL: test_mm256_permute2f128_pd
  51   // CHECK: shufflevector{{.*}}<i32 2, i32 3, i32 6, i32 7>
  52   return _mm256_permute2f128_pd(a, b, 0x31);
  53 }
  54
  55 __m256 test_mm256_permute2f128_ps(__m256 a, __m256 b) {
  56   // CHECK-LABEL: test_mm256_permute2f128_ps
  57   // CHECK: shufflevector{{.*}}<i32 4, i32 5, i32 6, i32 7, i32 12, i32 13, i32 14, i32 15>
  58   return _mm256_permute2f128_ps(a, b, 0x13);
  59 }
  60
  61 __m256i test_mm256_permute2f128_si256(__m256i a, __m256i b) {
  62   // CHECK-LABEL: test_mm256_permute2f128_si256
  63   // CHECK: shufflevector{{.*}} <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 8, i32 9, i32 10, i32 11>
  64   return _mm256_permute2f128_si256(a, b, 0x20);
  65 }
  66
  67 __m128
  68 test_mm_broadcast_ss(float const *__a) {
  69   // CHECK-LABEL: test_mm_broadcast_ss
  70   // CHECK: insertelement <4 x float> {{.*}}, i64 0
  71   // CHECK: shufflevector <4 x float> {{.*}}, <4 x float> poison, <4 x i32> zeroinitializer
  72   return _mm_broadcast_ss(__a);
  73 }
  74
  75 __m256d
  76 test_mm256_broadcast_sd(double const *__a) {
  77   // CHECK-LABEL: test_mm256_broadcast_sd
  78   // CHECK: insertelement <4 x double> {{.*}}, i64 0
  79   // CHECK: shufflevector <4 x double> {{.*}}, <4 x double> poison, <4 x i32> zeroinitializer
  80   return _mm256_broadcast_sd(__a);
  81 }
  82
  83 __m256
  84 test_mm256_broadcast_ss(float const *__a) {
  85   // CHECK-LABEL: test_mm256_broadcast_ss
  86   // CHECK: insertelement <8 x float> {{.*}}, i64 0
  87   // CHECK: shufflevector <8 x float> {{.*}}, <8 x float> poison, <8 x i32> zeroinitializer
  88   return _mm256_broadcast_ss(__a);
  89 }
  90
  91 // Make sure we have the correct mask for each insertf128 case.
  92
  93 __m256 test_mm256_insertf128_ps_0(__m256 a, __m128 b) {
  94   // CHECK-LABEL: test_mm256_insertf128_ps_0
  95   // CHECK: shufflevector{{.*}}<i32 0, i32 1, i32 2, i32 3, i32 12, i32 13, i32 14, i32 15>
  96   return _mm256_insertf128_ps(a, b, 0);
  97 }
  98
  99 __m256d test_mm256_insertf128_pd_0(__m256d a, __m128d b) {
 100   // CHECK-LABEL: test_mm256_insertf128_pd_0
 101   // CHECK: shufflevector{{.*}}<i32 0, i32 1, i32 6, i32 7>
 102   return _mm256_insertf128_pd(a, b, 0);
 103 }
 104
 105 __m256i test_mm256_insertf128_si256_0(__m256i a, __m128i b) {
 106   // CHECK-LABEL: test_mm256_insertf128_si256_0
 107   // CHECK: shufflevector{{.*}}<i32 0, i32 1, i32 2, i32 3, i32 12, i32 13, i32 14, i32 15>
 108   return _mm256_insertf128_si256(a, b, 0);
 109 }
 110
 111 __m256 test_mm256_insertf128_ps_1(__m256 a, __m128 b) {
 112   // CHECK-LABEL: test_mm256_insertf128_ps_1
 113   // CHECK: shufflevector{{.*}}<i32 0, i32 1, i32 2, i32 3, i32 8, i32 9, i32 10, i32 11>
 114   return _mm256_insertf128_ps(a, b, 1);
 115 }
 116
 117 __m256d test_mm256_insertf128_pd_1(__m256d a, __m128d b) {
 118   // CHECK-LABEL: test_mm256_insertf128_pd_1
 119   // CHECK: shufflevector{{.*}}<i32 0, i32 1, i32 4, i32 5>
 120   return _mm256_insertf128_pd(a, b, 1);
 121 }
 122
 123 __m256i test_mm256_insertf128_si256_1(__m256i a, __m128i b) {
 124   // CHECK-LABEL: test_mm256_insertf128_si256_1
 125   // CHECK: shufflevector{{.*}}<i32 0, i32 1, i32 2, i32 3, i32 8, i32 9, i32 10, i32 11>
 126   return _mm256_insertf128_si256(a, b, 1);
 127 }
 128
 129 // Make sure we have the correct mask for each extractf128 case.
 130
 131 __m128 test_mm256_extractf128_ps_0(__m256 a) {
 132   // X64-LABEL: test_mm256_extractf128_ps_0
 133   // X64: shufflevector{{.*}}<i32 0, i32 1, i32 2, i32 3>
 134   //
 135   // X86-LABEL: test_mm256_extractf128_ps_0
 136   // X86: shufflevector{{.*}}<i32 0, i32 1>
 137   return _mm256_extractf128_ps(a, 0);
 138 }
 139
 140 __m128d test_mm256_extractf128_pd_0(__m256d a) {
 141   // CHECK-LABEL: test_mm256_extractf128_pd_0
 142   // CHECK: shufflevector{{.*}}<i32 0, i32 1>
 143   return _mm256_extractf128_pd(a, 0);
 144 }
 145
 146 __m128i test_mm256_extractf128_si256_0(__m256i a) {
 147   // CHECK-LABEL: test_mm256_extractf128_si256_0
 148   // CHECK: shufflevector{{.*}}<i32 0, i32 1>
 149   return _mm256_extractf128_si256(a, 0);
 150 }
 151
 152 __m128 test_mm256_extractf128_ps_1(__m256 a) {
 153   // X64-LABEL: test_mm256_extractf128_ps_1
 154   // X64: shufflevector{{.*}}<i32 4, i32 5, i32 6, i32 7>
 155   //
 156   // X86-LABEL: test_mm256_extractf128_ps_1
 157   // X86: shufflevector{{.*}}<i32 2, i32 3>
 158   return _mm256_extractf128_ps(a, 1);
 159 }
 160
 161 __m128d test_mm256_extractf128_pd_1(__m256d a) {
 162   // CHECK-LABEL: test_mm256_extractf128_pd_1
 163   // CHECK: shufflevector{{.*}}<i32 2, i32 3>
 164   return _mm256_extractf128_pd(a, 1);
 165 }
 166
 167 __m128i test_mm256_extractf128_si256_1(__m256i a) {
 168   // CHECK-LABEL: test_mm256_extractf128_si256_1
 169   // CHECK: shufflevector{{.*}}<i32 2, i32 3>
 170   return _mm256_extractf128_si256(a, 1);
 171 }
 172
 173 __m256 test_mm256_set_m128(__m128 hi, __m128 lo) {
 174   // CHECK-LABEL: test_mm256_set_m128
 175   // CHECK: shufflevector{{.*}}<i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
 176   return _mm256_set_m128(hi, lo);
 177 }
 178
 179 __m256d test_mm256_set_m128d(__m128d hi, __m128d lo) {
 180   // CHECK-LABEL: test_mm256_set_m128d
 181   // CHECK: shufflevector{{.*}}<i32 0, i32 1, i32 2, i32 3>
 182   return _mm256_set_m128d(hi, lo);
 183 }
 184
 185 __m256i test_mm256_set_m128i(__m128i hi, __m128i lo) {
 186   // CHECK-LABEL: test_mm256_set_m128i
 187   // CHECK: shufflevector{{.*}}<i32 0, i32 1, i32 2, i32 3>
 188   return _mm256_set_m128i(hi, lo);
 189 }
 190
 191 __m256 test_mm256_setr_m128(__m128 hi, __m128 lo) {
 192   // CHECK-LABEL: test_mm256_setr_m128
 193   // CHECK: shufflevector{{.*}}<i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
 194   return _mm256_setr_m128(lo, hi);
 195 }
 196
 197 __m256d test_mm256_setr_m128d(__m128d hi, __m128d lo) {
 198   // CHECK-LABEL: test_mm256_setr_m128d
 199   // CHECK: shufflevector{{.*}}<i32 0, i32 1, i32 2, i32 3>
 200   return _mm256_setr_m128d(lo, hi);
 201 }
 202
 203 __m256i test_mm256_setr_m128i(__m128i hi, __m128i lo) {
 204   // CHECK-LABEL: test_mm256_setr_m128i
 205   // CHECK: shufflevector{{.*}}<i32 0, i32 1, i32 2, i32 3>
 206   return _mm256_setr_m128i(lo, hi);
 207 }