1 // RUN: %clang_cc1 %s -triple=x86_64-apple-darwin -target-feature +sse4.1 -emit-llvm -o - -Werror | FileCheck %s 2 // RUN: %clang_cc1 %s -triple=x86_64-apple-darwin -target-feature +sse4.1 -fno-signed-char -emit-llvm -o - -Werror | FileCheck %s 3 4 // Don't include mm_malloc.h, it's system specific. 5 #define __MM_MALLOC_H 6 7 #include <x86intrin.h> 8 9 // NOTE: This should match the tests in llvm/test/CodeGen/X86/sse41-intrinsics-fast-isel.ll 10 11 __m128i test_mm_blend_epi16(__m128i V1, __m128i V2) { 12 // CHECK-LABEL: test_mm_blend_epi16 13 // CHECK: shufflevector <8 x i16> %{{.*}}, <8 x i16> %{{.*}}, <8 x i32> <i32 0, i32 9, i32 2, i32 11, i32 4, i32 13, i32 6, i32 7> 14 return _mm_blend_epi16(V1, V2, 42); 15 } 16 17 __m128d test_mm_blend_pd(__m128d V1, __m128d V2) { 18 // CHECK-LABEL: test_mm_blend_pd 19 // CHECK: shufflevector <2 x double> %{{.*}}, <2 x double> %{{.*}}, <2 x i32> <i32 0, i32 3> 20 return _mm_blend_pd(V1, V2, 2); 21 } 22 23 __m128 test_mm_blend_ps(__m128 V1, __m128 V2) { 24 // CHECK-LABEL: test_mm_blend_ps 25 // CHECK: shufflevector <4 x float> %{{.*}}, <4 x float> %{{.*}}, <4 x i32> <i32 0, i32 5, i32 6, i32 3> 26 return _mm_blend_ps(V1, V2, 6); 27 } 28 29 __m128i test_mm_blendv_epi8(__m128i V1, __m128i V2, __m128i V3) { 30 // CHECK-LABEL: test_mm_blendv_epi8 31 // CHECK: call <16 x i8> @llvm.x86.sse41.pblendvb(<16 x i8> %{{.*}}, <16 x i8> %{{.*}}, <16 x i8> %{{.*}}) 32 return _mm_blendv_epi8(V1, V2, V3); 33 } 34 35 __m128d test_mm_blendv_pd(__m128d V1, __m128d V2, __m128d V3) { 36 // CHECK-LABEL: test_mm_blendv_pd 37 // CHECK: call <2 x double> @llvm.x86.sse41.blendvpd(<2 x double> %{{.*}}, <2 x double> %{{.*}}, <2 x double> %{{.*}}) 38 return _mm_blendv_pd(V1, V2, V3); 39 } 40 41 __m128 test_mm_blendv_ps(__m128 V1, __m128 V2, __m128 V3) { 42 // CHECK-LABEL: test_mm_blendv_ps 43 // CHECK: call <4 x float> @llvm.x86.sse41.blendvps(<4 x float> %{{.*}}, <4 x float> %{{.*}}, <4 x float> %{{.*}}) 44 return _mm_blendv_ps(V1, V2, V3); 45 } 46 47 __m128d test_mm_ceil_pd(__m128d x) { 48 // CHECK-LABEL: test_mm_ceil_pd 49 // CHECK: call <2 x double> @llvm.x86.sse41.round.pd(<2 x double> %{{.*}}, i32 2) 50 return _mm_ceil_pd(x); 51 } 52 53 __m128 test_mm_ceil_ps(__m128 x) { 54 // CHECK-LABEL: test_mm_ceil_ps 55 // CHECK: call <4 x float> @llvm.x86.sse41.round.ps(<4 x float> %{{.*}}, i32 2) 56 return _mm_ceil_ps(x); 57 } 58 59 __m128d test_mm_ceil_sd(__m128d x, __m128d y) { 60 // CHECK-LABEL: test_mm_ceil_sd 61 // CHECK: call <2 x double> @llvm.x86.sse41.round.sd(<2 x double> %{{.*}}, <2 x double> %{{.*}}, i32 2) 62 return _mm_ceil_sd(x, y); 63 } 64 65 __m128 test_mm_ceil_ss(__m128 x, __m128 y) { 66 // CHECK-LABEL: test_mm_ceil_ss 67 // CHECK: call <4 x float> @llvm.x86.sse41.round.ss(<4 x float> %{{.*}}, <4 x float> %{{.*}}, i32 2) 68 return _mm_ceil_ss(x, y); 69 } 70 71 __m128i test_mm_cmpeq_epi64(__m128i A, __m128i B) { 72 // CHECK-LABEL: test_mm_cmpeq_epi64 73 // CHECK: icmp eq <2 x i64> 74 // CHECK: sext <2 x i1> %{{.*}} to <2 x i64> 75 return _mm_cmpeq_epi64(A, B); 76 } 77 78 __m128i test_mm_cvtepi8_epi16(__m128i a) { 79 // CHECK-LABEL: test_mm_cvtepi8_epi16 80 // CHECK: shufflevector <16 x i8> {{.*}}, <16 x i8> {{.*}}, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7> 81 // CHECK: sext <8 x i8> {{.*}} to <8 x i16> 82 return _mm_cvtepi8_epi16(a); 83 } 84 85 __m128i test_mm_cvtepi8_epi32(__m128i a) { 86 // CHECK-LABEL: test_mm_cvtepi8_epi32 87 // CHECK: shufflevector <16 x i8> {{.*}}, <16 x i8> {{.*}}, <4 x i32> <i32 0, i32 1, i32 2, i32 3> 88 // CHECK: sext <4 x i8> {{.*}} to <4 x i32> 89 return _mm_cvtepi8_epi32(a); 90 } 91 92 __m128i test_mm_cvtepi8_epi64(__m128i a) { 93 // CHECK-LABEL: test_mm_cvtepi8_epi64 94 // CHECK: shufflevector <16 x i8> {{.*}}, <16 x i8> {{.*}}, <2 x i32> <i32 0, i32 1> 95 // CHECK: sext <2 x i8> {{.*}} to <2 x i64> 96 return _mm_cvtepi8_epi64(a); 97 } 98 99 __m128i test_mm_cvtepi16_epi32(__m128i a) { 100 // CHECK-LABEL: test_mm_cvtepi16_epi32 101 // CHECK: shufflevector <8 x i16> {{.*}}, <8 x i16> {{.*}}, <4 x i32> <i32 0, i32 1, i32 2, i32 3> 102 // CHECK: sext <4 x i16> {{.*}} to <4 x i32> 103 return _mm_cvtepi16_epi32(a); 104 } 105 106 __m128i test_mm_cvtepi16_epi64(__m128i a) { 107 // CHECK-LABEL: test_mm_cvtepi16_epi64 108 // CHECK: shufflevector <8 x i16> {{.*}}, <8 x i16> {{.*}}, <2 x i32> <i32 0, i32 1> 109 // CHECK: sext <2 x i16> {{.*}} to <2 x i64> 110 return _mm_cvtepi16_epi64(a); 111 } 112 113 __m128i test_mm_cvtepi32_epi64(__m128i a) { 114 // CHECK-LABEL: test_mm_cvtepi32_epi64 115 // CHECK: shufflevector <4 x i32> {{.*}}, <4 x i32> {{.*}}, <2 x i32> <i32 0, i32 1> 116 // CHECK: sext <2 x i32> {{.*}} to <2 x i64> 117 return _mm_cvtepi32_epi64(a); 118 } 119 120 __m128i test_mm_cvtepu8_epi16(__m128i a) { 121 // CHECK-LABEL: test_mm_cvtepu8_epi16 122 // CHECK: shufflevector <16 x i8> {{.*}}, <16 x i8> {{.*}}, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7> 123 // CHECK: zext <8 x i8> {{.*}} to <8 x i16> 124 return _mm_cvtepu8_epi16(a); 125 } 126 127 __m128i test_mm_cvtepu8_epi32(__m128i a) { 128 // CHECK-LABEL: test_mm_cvtepu8_epi32 129 // CHECK: shufflevector <16 x i8> {{.*}}, <16 x i8> {{.*}}, <4 x i32> <i32 0, i32 1, i32 2, i32 3> 130 // CHECK: zext <4 x i8> {{.*}} to <4 x i32> 131 return _mm_cvtepu8_epi32(a); 132 } 133 134 __m128i test_mm_cvtepu8_epi64(__m128i a) { 135 // CHECK-LABEL: test_mm_cvtepu8_epi64 136 // CHECK: shufflevector <16 x i8> {{.*}}, <16 x i8> {{.*}}, <2 x i32> <i32 0, i32 1> 137 // CHECK: zext <2 x i8> {{.*}} to <2 x i64> 138 return _mm_cvtepu8_epi64(a); 139 } 140 141 __m128i test_mm_cvtepu16_epi32(__m128i a) { 142 // CHECK-LABEL: test_mm_cvtepu16_epi32 143 // CHECK: shufflevector <8 x i16> {{.*}}, <8 x i16> {{.*}}, <4 x i32> <i32 0, i32 1, i32 2, i32 3> 144 // CHECK: zext <4 x i16> {{.*}} to <4 x i32> 145 return _mm_cvtepu16_epi32(a); 146 } 147 148 __m128i test_mm_cvtepu16_epi64(__m128i a) { 149 // CHECK-LABEL: test_mm_cvtepu16_epi64 150 // CHECK: shufflevector <8 x i16> {{.*}}, <8 x i16> {{.*}}, <2 x i32> <i32 0, i32 1> 151 // CHECK: zext <2 x i16> {{.*}} to <2 x i64> 152 return _mm_cvtepu16_epi64(a); 153 } 154 155 __m128i test_mm_cvtepu32_epi64(__m128i a) { 156 // CHECK-LABEL: test_mm_cvtepu32_epi64 157 // CHECK: shufflevector <4 x i32> {{.*}}, <4 x i32> {{.*}}, <2 x i32> <i32 0, i32 1> 158 // CHECK: zext <2 x i32> {{.*}} to <2 x i64> 159 return _mm_cvtepu32_epi64(a); 160 } 161 162 __m128d test_mm_dp_pd(__m128d x, __m128d y) { 163 // CHECK-LABEL: test_mm_dp_pd 164 // CHECK: call <2 x double> @llvm.x86.sse41.dppd(<2 x double> {{.*}}, <2 x double> {{.*}}, i8 7) 165 return _mm_dp_pd(x, y, 7); 166 } 167 168 __m128 test_mm_dp_ps(__m128 x, __m128 y) { 169 // CHECK-LABEL: test_mm_dp_ps 170 // CHECK: call <4 x float> @llvm.x86.sse41.dpps(<4 x float> {{.*}}, <4 x float> {{.*}}, i8 7) 171 return _mm_dp_ps(x, y, 7); 172 } 173 174 int test_mm_extract_epi8(__m128i x) { 175 // CHECK-LABEL: test_mm_extract_epi8 176 // CHECK: extractelement <16 x i8> %{{.*}}, i32 1 177 // CHECK: zext i8 %{{.*}} to i32 178 return _mm_extract_epi8(x, 1); 179 } 180 181 int test_mm_extract_epi32(__m128i x) { 182 // CHECK-LABEL: test_mm_extract_epi32 183 // CHECK: extractelement <4 x i32> %{{.*}}, i32 1 184 return _mm_extract_epi32(x, 1); 185 } 186 187 long long test_mm_extract_epi64(__m128i x) { 188 // CHECK-LABEL: test_mm_extract_epi64 189 // CHECK: extractelement <2 x i64> %{{.*}}, i32 1 190 return _mm_extract_epi64(x, 1); 191 } 192 193 int test_mm_extract_ps(__m128 x) { 194 // CHECK-LABEL: test_mm_extract_ps 195 // CHECK: extractelement <4 x float> %{{.*}}, i32 1 196 return _mm_extract_ps(x, 1); 197 } 198 199 __m128d test_mm_floor_pd(__m128d x) { 200 // CHECK-LABEL: test_mm_floor_pd 201 // CHECK: call <2 x double> @llvm.x86.sse41.round.pd(<2 x double> %{{.*}}, i32 1) 202 return _mm_floor_pd(x); 203 } 204 205 __m128 test_mm_floor_ps(__m128 x) { 206 // CHECK-LABEL: test_mm_floor_ps 207 // CHECK: call <4 x float> @llvm.x86.sse41.round.ps(<4 x float> %{{.*}}, i32 1) 208 return _mm_floor_ps(x); 209 } 210 211 __m128d test_mm_floor_sd(__m128d x, __m128d y) { 212 // CHECK-LABEL: test_mm_floor_sd 213 // CHECK: call <2 x double> @llvm.x86.sse41.round.sd(<2 x double> %{{.*}}, <2 x double> %{{.*}}, i32 1) 214 return _mm_floor_sd(x, y); 215 } 216 217 __m128 test_mm_floor_ss(__m128 x, __m128 y) { 218 // CHECK-LABEL: test_mm_floor_ss 219 // CHECK: call <4 x float> @llvm.x86.sse41.round.ss(<4 x float> %{{.*}}, <4 x float> %{{.*}}, i32 1) 220 return _mm_floor_ss(x, y); 221 } 222 223 __m128i test_mm_insert_epi8(__m128i x, char b) { 224 // CHECK-LABEL: test_mm_insert_epi8 225 // CHECK: insertelement <16 x i8> %{{.*}}, i8 %{{.*}}, i32 0 226 return _mm_insert_epi8(x, b, 16); 227 } 228 229 __m128i test_mm_insert_epi32(__m128i x, int b) { 230 // CHECK-LABEL: test_mm_insert_epi32 231 // CHECK: insertelement <4 x i32> %{{.*}}, i32 %{{.*}}, i32 0 232 return _mm_insert_epi32(x, b, 4); 233 } 234 235 __m128i test_mm_insert_epi64(__m128i x, long long b) { 236 // CHECK-LABEL: test_mm_insert_epi64 237 // CHECK: insertelement <2 x i64> %{{.*}}, i64 %{{.*}}, i32 0 238 return _mm_insert_epi64(x, b, 2); 239 } 240 241 __m128 test_mm_insert_ps(__m128 x, __m128 y) { 242 // CHECK-LABEL: test_mm_insert_ps 243 // CHECK: call <4 x float> @llvm.x86.sse41.insertps(<4 x float> %{{.*}}, <4 x float> %{{.*}}, i8 4) 244 return _mm_insert_ps(x, y, 4); 245 } 246 247 __m128i test_mm_max_epi8(__m128i x, __m128i y) { 248 // CHECK-LABEL: test_mm_max_epi8 249 // CHECK: [[CMP:%.*]] = icmp sgt <16 x i8> [[X:%.*]], [[Y:%.*]] 250 // CHECK-NEXT: select <16 x i1> [[CMP]], <16 x i8> [[X]], <16 x i8> [[Y]] 251 return _mm_max_epi8(x, y); 252 } 253 254 __m128i test_mm_max_epi32(__m128i x, __m128i y) { 255 // CHECK-LABEL: test_mm_max_epi32 256 // CHECK: [[CMP:%.*]] = icmp sgt <4 x i32> [[X:%.*]], [[Y:%.*]] 257 // CHECK-NEXT: select <4 x i1> [[CMP]], <4 x i32> [[X]], <4 x i32> [[Y]] 258 return _mm_max_epi32(x, y); 259 } 260 261 __m128i test_mm_max_epu16(__m128i x, __m128i y) { 262 // CHECK-LABEL: test_mm_max_epu16 263 // CHECK: [[CMP:%.*]] = icmp ugt <8 x i16> [[X:%.*]], [[Y:%.*]] 264 // CHECK-NEXT: select <8 x i1> [[CMP]], <8 x i16> [[X]], <8 x i16> [[Y]] 265 return _mm_max_epu16(x, y); 266 } 267 268 __m128i test_mm_max_epu32(__m128i x, __m128i y) { 269 // CHECK-LABEL: test_mm_max_epu32 270 // CHECK: [[CMP:%.*]] = icmp ugt <4 x i32> [[X:%.*]], [[Y:%.*]] 271 // CHECK-NEXT: select <4 x i1> [[CMP]], <4 x i32> [[X]], <4 x i32> [[Y]] 272 return _mm_max_epu32(x, y); 273 } 274 275 __m128i test_mm_min_epi8(__m128i x, __m128i y) { 276 // CHECK-LABEL: test_mm_min_epi8 277 // CHECK: [[CMP:%.*]] = icmp slt <16 x i8> [[X:%.*]], [[Y:%.*]] 278 // CHECK-NEXT: select <16 x i1> [[CMP]], <16 x i8> [[X]], <16 x i8> [[Y]] 279 return _mm_min_epi8(x, y); 280 } 281 282 __m128i test_mm_min_epi32(__m128i x, __m128i y) { 283 // CHECK-LABEL: test_mm_min_epi32 284 // CHECK: [[CMP:%.*]] = icmp slt <4 x i32> [[X:%.*]], [[Y:%.*]] 285 // CHECK-NEXT: select <4 x i1> [[CMP]], <4 x i32> [[X]], <4 x i32> [[Y]] 286 return _mm_min_epi32(x, y); 287 } 288 289 __m128i test_mm_min_epu16(__m128i x, __m128i y) { 290 // CHECK-LABEL: test_mm_min_epu16 291 // CHECK: [[CMP:%.*]] = icmp ult <8 x i16> [[X:%.*]], [[Y:%.*]] 292 // CHECK-NEXT: select <8 x i1> [[CMP]], <8 x i16> [[X]], <8 x i16> [[Y]] 293 return _mm_min_epu16(x, y); 294 } 295 296 __m128i test_mm_min_epu32(__m128i x, __m128i y) { 297 // CHECK-LABEL: test_mm_min_epu32 298 // CHECK: [[CMP:%.*]] = icmp ult <4 x i32> [[X:%.*]], [[Y:%.*]] 299 // CHECK-NEXT: select <4 x i1> [[CMP]], <4 x i32> [[X]], <4 x i32> [[Y]] 300 return _mm_min_epu32(x, y); 301 } 302 303 __m128i test_mm_minpos_epu16(__m128i x) { 304 // CHECK-LABEL: test_mm_minpos_epu16 305 // CHECK: call <8 x i16> @llvm.x86.sse41.phminposuw(<8 x i16> %{{.*}}) 306 return _mm_minpos_epu16(x); 307 } 308 309 __m128i test_mm_mpsadbw_epu8(__m128i x, __m128i y) { 310 // CHECK-LABEL: test_mm_mpsadbw_epu8 311 // CHECK: call <8 x i16> @llvm.x86.sse41.mpsadbw(<16 x i8> %{{.*}}, <16 x i8> %{{.*}}, i8 1) 312 return _mm_mpsadbw_epu8(x, y, 1); 313 } 314 315 __m128i test_mm_mul_epi32(__m128i x, __m128i y) { 316 // CHECK-LABEL: test_mm_mul_epi32 317 // CHECK: call <2 x i64> @llvm.x86.sse41.pmuldq(<4 x i32> %{{.*}}, <4 x i32> %{{.*}}) 318 return _mm_mul_epi32(x, y); 319 } 320 321 __m128i test_mm_mullo_epi32(__m128i x, __m128i y) { 322 // CHECK-LABEL: test_mm_mullo_epi32 323 // CHECK: mul <4 x i32> 324 return _mm_mullo_epi32(x, y); 325 } 326 327 __m128i test_mm_packus_epi32(__m128i x, __m128i y) { 328 // CHECK-LABEL: test_mm_packus_epi32 329 // CHECK: call <8 x i16> @llvm.x86.sse41.packusdw(<4 x i32> %{{.*}}, <4 x i32> %{{.*}}) 330 return _mm_packus_epi32(x, y); 331 } 332 333 __m128d test_mm_round_pd(__m128d x) { 334 // CHECK-LABEL: test_mm_round_pd 335 // CHECK: call <2 x double> @llvm.x86.sse41.round.pd(<2 x double> %{{.*}}, i32 4) 336 return _mm_round_pd(x, 4); 337 } 338 339 __m128 test_mm_round_ps(__m128 x) { 340 // CHECK-LABEL: test_mm_round_ps 341 // CHECK: call <4 x float> @llvm.x86.sse41.round.ps(<4 x float> %{{.*}}, i32 4) 342 return _mm_round_ps(x, 4); 343 } 344 345 __m128d test_mm_round_sd(__m128d x, __m128d y) { 346 // CHECK-LABEL: test_mm_round_sd 347 // CHECK: call <2 x double> @llvm.x86.sse41.round.sd(<2 x double> %{{.*}}, <2 x double> %{{.*}}, i32 4) 348 return _mm_round_sd(x, y, 4); 349 } 350 351 __m128 test_mm_round_ss(__m128 x, __m128 y) { 352 // CHECK-LABEL: test_mm_round_ss 353 // CHECK: call <4 x float> @llvm.x86.sse41.round.ss(<4 x float> %{{.*}}, <4 x float> %{{.*}}, i32 4) 354 return _mm_round_ss(x, y, 4); 355 } 356 357 __m128i test_mm_stream_load_si128(__m128i const *a) { 358 // CHECK-LABEL: test_mm_stream_load_si128 359 // CHECK: call <2 x i64> @llvm.x86.sse41.movntdqa(i8* %{{.*}}) 360 return _mm_stream_load_si128(a); 361 } 362 363 int test_mm_test_all_ones(__m128i x) { 364 // CHECK-LABEL: test_mm_test_all_ones 365 // CHECK: call i32 @llvm.x86.sse41.ptestc(<2 x i64> %{{.*}}, <2 x i64> %{{.*}}) 366 return _mm_test_all_ones(x); 367 } 368 369 int test_mm_test_all_zeros(__m128i x, __m128i y) { 370 // CHECK-LABEL: test_mm_test_all_zeros 371 // CHECK: call i32 @llvm.x86.sse41.ptestz(<2 x i64> %{{.*}}, <2 x i64> %{{.*}}) 372 return _mm_test_all_zeros(x, y); 373 } 374 375 int test_mm_test_mix_ones_zeros(__m128i x, __m128i y) { 376 // CHECK-LABEL: test_mm_test_mix_ones_zeros 377 // CHECK: call i32 @llvm.x86.sse41.ptestnzc(<2 x i64> %{{.*}}, <2 x i64> %{{.*}}) 378 return _mm_test_mix_ones_zeros(x, y); 379 } 380 381 int test_mm_testc_si128(__m128i x, __m128i y) { 382 // CHECK-LABEL: test_mm_testc_si128 383 // CHECK: call i32 @llvm.x86.sse41.ptestc(<2 x i64> %{{.*}}, <2 x i64> %{{.*}}) 384 return _mm_testc_si128(x, y); 385 } 386 387 int test_mm_testnzc_si128(__m128i x, __m128i y) { 388 // CHECK-LABEL: test_mm_testnzc_si128 389 // CHECK: call i32 @llvm.x86.sse41.ptestnzc(<2 x i64> %{{.*}}, <2 x i64> %{{.*}}) 390 return _mm_testnzc_si128(x, y); 391 } 392 393 int test_mm_testz_si128(__m128i x, __m128i y) { 394 // CHECK-LABEL: test_mm_testz_si128 395 // CHECK: call i32 @llvm.x86.sse41.ptestz(<2 x i64> %{{.*}}, <2 x i64> %{{.*}}) 396 return _mm_testz_si128(x, y); 397 } 398