| 1 | //! Bit-oriented Algorithms (BITALG) |
| 2 | //! |
| 3 | //! The intrinsics here correspond to those in the `immintrin.h` C header. |
| 4 | //! |
| 5 | //! The reference is [Intel 64 and IA-32 Architectures Software Developer's |
| 6 | //! Manual Volume 2: Instruction Set Reference, A-Z][intel64_ref]. |
| 7 | //! |
| 8 | //! [intel64_ref]: https://www.intel.com/content/dam/www/public/us/en/documents/manuals/64-ia-32-architectures-software-developer-instruction-set-reference-manual-325383.pdf |
| 9 | |
| 10 | use crate::core_arch::simd::i8x16; |
| 11 | use crate::core_arch::simd::i8x32; |
| 12 | use crate::core_arch::simd::i8x64; |
| 13 | use crate::core_arch::simd::i16x8; |
| 14 | use crate::core_arch::simd::i16x16; |
| 15 | use crate::core_arch::simd::i16x32; |
| 16 | use crate::core_arch::x86::__m128i; |
| 17 | use crate::core_arch::x86::__m256i; |
| 18 | use crate::core_arch::x86::__m512i; |
| 19 | use crate::core_arch::x86::__mmask8; |
| 20 | use crate::core_arch::x86::__mmask16; |
| 21 | use crate::core_arch::x86::__mmask32; |
| 22 | use crate::core_arch::x86::__mmask64; |
| 23 | use crate::intrinsics::simd::{simd_ctpop, simd_select_bitmask}; |
| 24 | use crate::mem::transmute; |
| 25 | |
| 26 | #[cfg (test)] |
| 27 | use stdarch_test::assert_instr; |
| 28 | |
| 29 | #[allow (improper_ctypes)] |
| 30 | unsafe extern "C" { |
| 31 | #[link_name = "llvm.x86.avx512.mask.vpshufbitqmb.512" ] |
| 32 | unsafefn bitshuffle_512(data: i8x64, indices: i8x64, mask: __mmask64) -> __mmask64; |
| 33 | #[link_name = "llvm.x86.avx512.mask.vpshufbitqmb.256" ] |
| 34 | unsafefn bitshuffle_256(data: i8x32, indices: i8x32, mask: __mmask32) -> __mmask32; |
| 35 | #[link_name = "llvm.x86.avx512.mask.vpshufbitqmb.128" ] |
| 36 | unsafefn bitshuffle_128(data: i8x16, indices: i8x16, mask: __mmask16) -> __mmask16; |
| 37 | } |
| 38 | |
| 39 | /// For each packed 16-bit integer maps the value to the number of logical 1 bits. |
| 40 | /// |
| 41 | /// [Intel's documentation](https://www.intel.com/content/www/us/en/docs/intrinsics-guide/index.html#text=_mm512_popcnt_epi16) |
| 42 | #[inline ] |
| 43 | #[target_feature (enable = "avx512bitalg" )] |
| 44 | #[stable (feature = "stdarch_x86_avx512" , since = "1.89" )] |
| 45 | #[cfg_attr (test, assert_instr(vpopcntw))] |
| 46 | #[rustc_const_unstable (feature = "stdarch_const_x86" , issue = "149298" )] |
| 47 | pub const fn _mm512_popcnt_epi16(a: __m512i) -> __m512i { |
| 48 | unsafe { transmute(src:simd_ctpop(a.as_i16x32())) } |
| 49 | } |
| 50 | |
| 51 | /// For each packed 16-bit integer maps the value to the number of logical 1 bits. |
| 52 | /// |
| 53 | /// Uses the writemask in k - elements are zeroed in the result if the corresponding mask bit is not set. |
| 54 | /// Otherwise the computation result is written into the result. |
| 55 | /// |
| 56 | /// [Intel's documentation](https://www.intel.com/content/www/us/en/docs/intrinsics-guide/index.html#text=_mm512_maskz_popcnt_epi16) |
| 57 | #[inline ] |
| 58 | #[target_feature (enable = "avx512bitalg" )] |
| 59 | #[stable (feature = "stdarch_x86_avx512" , since = "1.89" )] |
| 60 | #[cfg_attr (test, assert_instr(vpopcntw))] |
| 61 | #[rustc_const_unstable (feature = "stdarch_const_x86" , issue = "149298" )] |
| 62 | pub const fn _mm512_maskz_popcnt_epi16(k: __mmask32, a: __m512i) -> __m512i { |
| 63 | unsafe { |
| 64 | transmute(src:simd_select_bitmask( |
| 65 | m:k, |
| 66 | yes:simd_ctpop(a.as_i16x32()), |
| 67 | no:i16x32::ZERO, |
| 68 | )) |
| 69 | } |
| 70 | } |
| 71 | |
| 72 | /// For each packed 16-bit integer maps the value to the number of logical 1 bits. |
| 73 | /// |
| 74 | /// Uses the writemask in k - elements are copied from src if the corresponding mask bit is not set. |
| 75 | /// Otherwise the computation result is written into the result. |
| 76 | /// |
| 77 | /// [Intel's documentation](https://www.intel.com/content/www/us/en/docs/intrinsics-guide/index.html#text=_mm512_mask_popcnt_epi16) |
| 78 | #[inline ] |
| 79 | #[target_feature (enable = "avx512bitalg" )] |
| 80 | #[stable (feature = "stdarch_x86_avx512" , since = "1.89" )] |
| 81 | #[cfg_attr (test, assert_instr(vpopcntw))] |
| 82 | #[rustc_const_unstable (feature = "stdarch_const_x86" , issue = "149298" )] |
| 83 | pub const fn _mm512_mask_popcnt_epi16(src: __m512i, k: __mmask32, a: __m512i) -> __m512i { |
| 84 | unsafe { |
| 85 | transmute(src:simd_select_bitmask( |
| 86 | m:k, |
| 87 | yes:simd_ctpop(a.as_i16x32()), |
| 88 | no:src.as_i16x32(), |
| 89 | )) |
| 90 | } |
| 91 | } |
| 92 | |
| 93 | /// For each packed 16-bit integer maps the value to the number of logical 1 bits. |
| 94 | /// |
| 95 | /// [Intel's documentation](https://www.intel.com/content/www/us/en/docs/intrinsics-guide/index.html#text=_mm256_popcnt_epi16) |
| 96 | #[inline ] |
| 97 | #[target_feature (enable = "avx512bitalg,avx512vl" )] |
| 98 | #[stable (feature = "stdarch_x86_avx512" , since = "1.89" )] |
| 99 | #[cfg_attr (test, assert_instr(vpopcntw))] |
| 100 | #[rustc_const_unstable (feature = "stdarch_const_x86" , issue = "149298" )] |
| 101 | pub const fn _mm256_popcnt_epi16(a: __m256i) -> __m256i { |
| 102 | unsafe { transmute(src:simd_ctpop(a.as_i16x16())) } |
| 103 | } |
| 104 | |
| 105 | /// For each packed 16-bit integer maps the value to the number of logical 1 bits. |
| 106 | /// |
| 107 | /// Uses the writemask in k - elements are zeroed in the result if the corresponding mask bit is not set. |
| 108 | /// Otherwise the computation result is written into the result. |
| 109 | /// |
| 110 | /// [Intel's documentation](https://www.intel.com/content/www/us/en/docs/intrinsics-guide/index.html#text=_mm256_maskz_popcnt_epi16) |
| 111 | #[inline ] |
| 112 | #[target_feature (enable = "avx512bitalg,avx512vl" )] |
| 113 | #[stable (feature = "stdarch_x86_avx512" , since = "1.89" )] |
| 114 | #[cfg_attr (test, assert_instr(vpopcntw))] |
| 115 | #[rustc_const_unstable (feature = "stdarch_const_x86" , issue = "149298" )] |
| 116 | pub const fn _mm256_maskz_popcnt_epi16(k: __mmask16, a: __m256i) -> __m256i { |
| 117 | unsafe { |
| 118 | transmute(src:simd_select_bitmask( |
| 119 | m:k, |
| 120 | yes:simd_ctpop(a.as_i16x16()), |
| 121 | no:i16x16::ZERO, |
| 122 | )) |
| 123 | } |
| 124 | } |
| 125 | |
| 126 | /// For each packed 16-bit integer maps the value to the number of logical 1 bits. |
| 127 | /// |
| 128 | /// Uses the writemask in k - elements are copied from src if the corresponding mask bit is not set. |
| 129 | /// Otherwise the computation result is written into the result. |
| 130 | /// |
| 131 | /// [Intel's documentation](https://www.intel.com/content/www/us/en/docs/intrinsics-guide/index.html#text=_mm256_mask_popcnt_epi16) |
| 132 | #[inline ] |
| 133 | #[target_feature (enable = "avx512bitalg,avx512vl" )] |
| 134 | #[stable (feature = "stdarch_x86_avx512" , since = "1.89" )] |
| 135 | #[cfg_attr (test, assert_instr(vpopcntw))] |
| 136 | #[rustc_const_unstable (feature = "stdarch_const_x86" , issue = "149298" )] |
| 137 | pub const fn _mm256_mask_popcnt_epi16(src: __m256i, k: __mmask16, a: __m256i) -> __m256i { |
| 138 | unsafe { |
| 139 | transmute(src:simd_select_bitmask( |
| 140 | m:k, |
| 141 | yes:simd_ctpop(a.as_i16x16()), |
| 142 | no:src.as_i16x16(), |
| 143 | )) |
| 144 | } |
| 145 | } |
| 146 | |
| 147 | /// For each packed 16-bit integer maps the value to the number of logical 1 bits. |
| 148 | /// |
| 149 | /// [Intel's documentation](https://www.intel.com/content/www/us/en/docs/intrinsics-guide/index.html#text=_mm_popcnt_epi16) |
| 150 | #[inline ] |
| 151 | #[target_feature (enable = "avx512bitalg,avx512vl" )] |
| 152 | #[stable (feature = "stdarch_x86_avx512" , since = "1.89" )] |
| 153 | #[cfg_attr (test, assert_instr(vpopcntw))] |
| 154 | #[rustc_const_unstable (feature = "stdarch_const_x86" , issue = "149298" )] |
| 155 | pub const fn _mm_popcnt_epi16(a: __m128i) -> __m128i { |
| 156 | unsafe { transmute(src:simd_ctpop(a.as_i16x8())) } |
| 157 | } |
| 158 | |
| 159 | /// For each packed 16-bit integer maps the value to the number of logical 1 bits. |
| 160 | /// |
| 161 | /// Uses the writemask in k - elements are zeroed in the result if the corresponding mask bit is not set. |
| 162 | /// Otherwise the computation result is written into the result. |
| 163 | /// |
| 164 | /// [Intel's documentation](https://www.intel.com/content/www/us/en/docs/intrinsics-guide/index.html#text=_mm_maskz_popcnt_epi16) |
| 165 | #[inline ] |
| 166 | #[target_feature (enable = "avx512bitalg,avx512vl" )] |
| 167 | #[stable (feature = "stdarch_x86_avx512" , since = "1.89" )] |
| 168 | #[cfg_attr (test, assert_instr(vpopcntw))] |
| 169 | #[rustc_const_unstable (feature = "stdarch_const_x86" , issue = "149298" )] |
| 170 | pub const fn _mm_maskz_popcnt_epi16(k: __mmask8, a: __m128i) -> __m128i { |
| 171 | unsafe { |
| 172 | transmute(src:simd_select_bitmask( |
| 173 | m:k, |
| 174 | yes:simd_ctpop(a.as_i16x8()), |
| 175 | no:i16x8::ZERO, |
| 176 | )) |
| 177 | } |
| 178 | } |
| 179 | |
| 180 | /// For each packed 16-bit integer maps the value to the number of logical 1 bits. |
| 181 | /// |
| 182 | /// Uses the writemask in k - elements are copied from src if the corresponding mask bit is not set. |
| 183 | /// Otherwise the computation result is written into the result. |
| 184 | /// |
| 185 | /// [Intel's documentation](https://www.intel.com/content/www/us/en/docs/intrinsics-guide/index.html#text=_mm_mask_popcnt_epi16) |
| 186 | #[inline ] |
| 187 | #[target_feature (enable = "avx512bitalg,avx512vl" )] |
| 188 | #[stable (feature = "stdarch_x86_avx512" , since = "1.89" )] |
| 189 | #[cfg_attr (test, assert_instr(vpopcntw))] |
| 190 | #[rustc_const_unstable (feature = "stdarch_const_x86" , issue = "149298" )] |
| 191 | pub const fn _mm_mask_popcnt_epi16(src: __m128i, k: __mmask8, a: __m128i) -> __m128i { |
| 192 | unsafe { |
| 193 | transmute(src:simd_select_bitmask( |
| 194 | m:k, |
| 195 | yes:simd_ctpop(a.as_i16x8()), |
| 196 | no:src.as_i16x8(), |
| 197 | )) |
| 198 | } |
| 199 | } |
| 200 | |
| 201 | /// For each packed 8-bit integer maps the value to the number of logical 1 bits. |
| 202 | /// |
| 203 | /// [Intel's documentation](https://www.intel.com/content/www/us/en/docs/intrinsics-guide/index.html#text=_mm512_popcnt_epi8) |
| 204 | #[inline ] |
| 205 | #[target_feature (enable = "avx512bitalg" )] |
| 206 | #[stable (feature = "stdarch_x86_avx512" , since = "1.89" )] |
| 207 | #[cfg_attr (test, assert_instr(vpopcntb))] |
| 208 | #[rustc_const_unstable (feature = "stdarch_const_x86" , issue = "149298" )] |
| 209 | pub const fn _mm512_popcnt_epi8(a: __m512i) -> __m512i { |
| 210 | unsafe { transmute(src:simd_ctpop(a.as_i8x64())) } |
| 211 | } |
| 212 | |
| 213 | /// For each packed 8-bit integer maps the value to the number of logical 1 bits. |
| 214 | /// |
| 215 | /// Uses the writemask in k - elements are zeroed in the result if the corresponding mask bit is not set. |
| 216 | /// Otherwise the computation result is written into the result. |
| 217 | /// |
| 218 | /// [Intel's documentation](https://www.intel.com/content/www/us/en/docs/intrinsics-guide/index.html#text=_mm512_maskz_popcnt_epi8) |
| 219 | #[inline ] |
| 220 | #[target_feature (enable = "avx512bitalg" )] |
| 221 | #[stable (feature = "stdarch_x86_avx512" , since = "1.89" )] |
| 222 | #[cfg_attr (test, assert_instr(vpopcntb))] |
| 223 | #[rustc_const_unstable (feature = "stdarch_const_x86" , issue = "149298" )] |
| 224 | pub const fn _mm512_maskz_popcnt_epi8(k: __mmask64, a: __m512i) -> __m512i { |
| 225 | unsafe { |
| 226 | transmute(src:simd_select_bitmask( |
| 227 | m:k, |
| 228 | yes:simd_ctpop(a.as_i8x64()), |
| 229 | no:i8x64::ZERO, |
| 230 | )) |
| 231 | } |
| 232 | } |
| 233 | |
| 234 | /// For each packed 8-bit integer maps the value to the number of logical 1 bits. |
| 235 | /// |
| 236 | /// Uses the writemask in k - elements are copied from src if the corresponding mask bit is not set. |
| 237 | /// Otherwise the computation result is written into the result. |
| 238 | /// |
| 239 | /// [Intel's documentation](https://www.intel.com/content/www/us/en/docs/intrinsics-guide/index.html#text=_mm512_mask_popcnt_epi8) |
| 240 | #[inline ] |
| 241 | #[target_feature (enable = "avx512bitalg" )] |
| 242 | #[stable (feature = "stdarch_x86_avx512" , since = "1.89" )] |
| 243 | #[cfg_attr (test, assert_instr(vpopcntb))] |
| 244 | #[rustc_const_unstable (feature = "stdarch_const_x86" , issue = "149298" )] |
| 245 | pub const fn _mm512_mask_popcnt_epi8(src: __m512i, k: __mmask64, a: __m512i) -> __m512i { |
| 246 | unsafe { |
| 247 | transmute(src:simd_select_bitmask( |
| 248 | m:k, |
| 249 | yes:simd_ctpop(a.as_i8x64()), |
| 250 | no:src.as_i8x64(), |
| 251 | )) |
| 252 | } |
| 253 | } |
| 254 | |
| 255 | /// For each packed 8-bit integer maps the value to the number of logical 1 bits. |
| 256 | /// |
| 257 | /// [Intel's documentation](https://www.intel.com/content/www/us/en/docs/intrinsics-guide/index.html#text=_mm256_popcnt_epi8) |
| 258 | #[inline ] |
| 259 | #[target_feature (enable = "avx512bitalg,avx512vl" )] |
| 260 | #[stable (feature = "stdarch_x86_avx512" , since = "1.89" )] |
| 261 | #[cfg_attr (test, assert_instr(vpopcntb))] |
| 262 | #[rustc_const_unstable (feature = "stdarch_const_x86" , issue = "149298" )] |
| 263 | pub const fn _mm256_popcnt_epi8(a: __m256i) -> __m256i { |
| 264 | unsafe { transmute(src:simd_ctpop(a.as_i8x32())) } |
| 265 | } |
| 266 | |
| 267 | /// For each packed 8-bit integer maps the value to the number of logical 1 bits. |
| 268 | /// |
| 269 | /// Uses the writemask in k - elements are zeroed in the result if the corresponding mask bit is not set. |
| 270 | /// Otherwise the computation result is written into the result. |
| 271 | /// |
| 272 | /// [Intel's documentation](https://www.intel.com/content/www/us/en/docs/intrinsics-guide/index.html#text=_mm256_maskz_popcnt_epi8) |
| 273 | #[inline ] |
| 274 | #[target_feature (enable = "avx512bitalg,avx512vl" )] |
| 275 | #[stable (feature = "stdarch_x86_avx512" , since = "1.89" )] |
| 276 | #[cfg_attr (test, assert_instr(vpopcntb))] |
| 277 | #[rustc_const_unstable (feature = "stdarch_const_x86" , issue = "149298" )] |
| 278 | pub const fn _mm256_maskz_popcnt_epi8(k: __mmask32, a: __m256i) -> __m256i { |
| 279 | unsafe { |
| 280 | transmute(src:simd_select_bitmask( |
| 281 | m:k, |
| 282 | yes:simd_ctpop(a.as_i8x32()), |
| 283 | no:i8x32::ZERO, |
| 284 | )) |
| 285 | } |
| 286 | } |
| 287 | |
| 288 | /// For each packed 8-bit integer maps the value to the number of logical 1 bits. |
| 289 | /// |
| 290 | /// Uses the writemask in k - elements are copied from src if the corresponding mask bit is not set. |
| 291 | /// Otherwise the computation result is written into the result. |
| 292 | /// |
| 293 | /// [Intel's documentation](https://www.intel.com/content/www/us/en/docs/intrinsics-guide/index.html#text=_mm256_mask_popcnt_epi8) |
| 294 | #[inline ] |
| 295 | #[target_feature (enable = "avx512bitalg,avx512vl" )] |
| 296 | #[stable (feature = "stdarch_x86_avx512" , since = "1.89" )] |
| 297 | #[cfg_attr (test, assert_instr(vpopcntb))] |
| 298 | #[rustc_const_unstable (feature = "stdarch_const_x86" , issue = "149298" )] |
| 299 | pub const fn _mm256_mask_popcnt_epi8(src: __m256i, k: __mmask32, a: __m256i) -> __m256i { |
| 300 | unsafe { |
| 301 | transmute(src:simd_select_bitmask( |
| 302 | m:k, |
| 303 | yes:simd_ctpop(a.as_i8x32()), |
| 304 | no:src.as_i8x32(), |
| 305 | )) |
| 306 | } |
| 307 | } |
| 308 | |
| 309 | /// For each packed 8-bit integer maps the value to the number of logical 1 bits. |
| 310 | /// |
| 311 | /// [Intel's documentation](https://www.intel.com/content/www/us/en/docs/intrinsics-guide/index.html#text=_mm_popcnt_epi8) |
| 312 | #[inline ] |
| 313 | #[target_feature (enable = "avx512bitalg,avx512vl" )] |
| 314 | #[stable (feature = "stdarch_x86_avx512" , since = "1.89" )] |
| 315 | #[cfg_attr (test, assert_instr(vpopcntb))] |
| 316 | #[rustc_const_unstable (feature = "stdarch_const_x86" , issue = "149298" )] |
| 317 | pub const fn _mm_popcnt_epi8(a: __m128i) -> __m128i { |
| 318 | unsafe { transmute(src:simd_ctpop(a.as_i8x16())) } |
| 319 | } |
| 320 | |
| 321 | /// For each packed 8-bit integer maps the value to the number of logical 1 bits. |
| 322 | /// |
| 323 | /// Uses the writemask in k - elements are zeroed in the result if the corresponding mask bit is not set. |
| 324 | /// Otherwise the computation result is written into the result. |
| 325 | /// |
| 326 | /// [Intel's documentation](https://www.intel.com/content/www/us/en/docs/intrinsics-guide/index.html#text=_mm_maskz_popcnt_epi8) |
| 327 | #[inline ] |
| 328 | #[target_feature (enable = "avx512bitalg,avx512vl" )] |
| 329 | #[stable (feature = "stdarch_x86_avx512" , since = "1.89" )] |
| 330 | #[cfg_attr (test, assert_instr(vpopcntb))] |
| 331 | #[rustc_const_unstable (feature = "stdarch_const_x86" , issue = "149298" )] |
| 332 | pub const fn _mm_maskz_popcnt_epi8(k: __mmask16, a: __m128i) -> __m128i { |
| 333 | unsafe { |
| 334 | transmute(src:simd_select_bitmask( |
| 335 | m:k, |
| 336 | yes:simd_ctpop(a.as_i8x16()), |
| 337 | no:i8x16::ZERO, |
| 338 | )) |
| 339 | } |
| 340 | } |
| 341 | |
| 342 | /// For each packed 8-bit integer maps the value to the number of logical 1 bits. |
| 343 | /// |
| 344 | /// Uses the writemask in k - elements are copied from src if the corresponding mask bit is not set. |
| 345 | /// Otherwise the computation result is written into the result. |
| 346 | /// |
| 347 | /// [Intel's documentation](https://www.intel.com/content/www/us/en/docs/intrinsics-guide/index.html#text=_mm_mask_popcnt_epi8) |
| 348 | #[inline ] |
| 349 | #[target_feature (enable = "avx512bitalg,avx512vl" )] |
| 350 | #[stable (feature = "stdarch_x86_avx512" , since = "1.89" )] |
| 351 | #[cfg_attr (test, assert_instr(vpopcntb))] |
| 352 | #[rustc_const_unstable (feature = "stdarch_const_x86" , issue = "149298" )] |
| 353 | pub const fn _mm_mask_popcnt_epi8(src: __m128i, k: __mmask16, a: __m128i) -> __m128i { |
| 354 | unsafe { |
| 355 | transmute(src:simd_select_bitmask( |
| 356 | m:k, |
| 357 | yes:simd_ctpop(a.as_i8x16()), |
| 358 | no:src.as_i8x16(), |
| 359 | )) |
| 360 | } |
| 361 | } |
| 362 | |
| 363 | /// Considers the input `b` as packed 64-bit integers and `c` as packed 8-bit integers. |
| 364 | /// Then groups 8 8-bit values from `c`as indices into the bits of the corresponding 64-bit integer. |
| 365 | /// It then selects these bits and packs them into the output. |
| 366 | /// |
| 367 | /// [Intel's documentation](https://www.intel.com/content/www/us/en/docs/intrinsics-guide/index.html#text=_mm512_bitshuffle_epi64_mask) |
| 368 | #[inline ] |
| 369 | #[target_feature (enable = "avx512bitalg" )] |
| 370 | #[stable (feature = "stdarch_x86_avx512" , since = "1.89" )] |
| 371 | #[cfg_attr (test, assert_instr(vpshufbitqmb))] |
| 372 | pub fn _mm512_bitshuffle_epi64_mask(b: __m512i, c: __m512i) -> __mmask64 { |
| 373 | unsafe { bitshuffle_512(data:b.as_i8x64(), indices:c.as_i8x64(), !0) } |
| 374 | } |
| 375 | |
| 376 | /// Considers the input `b` as packed 64-bit integers and `c` as packed 8-bit integers. |
| 377 | /// Then groups 8 8-bit values from `c`as indices into the bits of the corresponding 64-bit integer. |
| 378 | /// It then selects these bits and packs them into the output. |
| 379 | /// |
| 380 | /// Uses the writemask in k - elements are zeroed in the result if the corresponding mask bit is not set. |
| 381 | /// Otherwise the computation result is written into the result. |
| 382 | /// |
| 383 | /// [Intel's documentation](https://www.intel.com/content/www/us/en/docs/intrinsics-guide/index.html#text=_mm512_mask_bitshuffle_epi64_mask) |
| 384 | #[inline ] |
| 385 | #[target_feature (enable = "avx512bitalg" )] |
| 386 | #[stable (feature = "stdarch_x86_avx512" , since = "1.89" )] |
| 387 | #[cfg_attr (test, assert_instr(vpshufbitqmb))] |
| 388 | pub fn _mm512_mask_bitshuffle_epi64_mask(k: __mmask64, b: __m512i, c: __m512i) -> __mmask64 { |
| 389 | unsafe { bitshuffle_512(data:b.as_i8x64(), indices:c.as_i8x64(), mask:k) } |
| 390 | } |
| 391 | |
| 392 | /// Considers the input `b` as packed 64-bit integers and `c` as packed 8-bit integers. |
| 393 | /// Then groups 8 8-bit values from `c`as indices into the bits of the corresponding 64-bit integer. |
| 394 | /// It then selects these bits and packs them into the output. |
| 395 | /// |
| 396 | /// [Intel's documentation](https://www.intel.com/content/www/us/en/docs/intrinsics-guide/index.html#text=_mm256_bitshuffle_epi64_mask) |
| 397 | #[inline ] |
| 398 | #[target_feature (enable = "avx512bitalg,avx512vl" )] |
| 399 | #[stable (feature = "stdarch_x86_avx512" , since = "1.89" )] |
| 400 | #[cfg_attr (test, assert_instr(vpshufbitqmb))] |
| 401 | pub fn _mm256_bitshuffle_epi64_mask(b: __m256i, c: __m256i) -> __mmask32 { |
| 402 | unsafe { bitshuffle_256(data:b.as_i8x32(), indices:c.as_i8x32(), !0) } |
| 403 | } |
| 404 | |
| 405 | /// Considers the input `b` as packed 64-bit integers and `c` as packed 8-bit integers. |
| 406 | /// Then groups 8 8-bit values from `c`as indices into the bits of the corresponding 64-bit integer. |
| 407 | /// It then selects these bits and packs them into the output. |
| 408 | /// |
| 409 | /// Uses the writemask in k - elements are zeroed in the result if the corresponding mask bit is not set. |
| 410 | /// Otherwise the computation result is written into the result. |
| 411 | /// |
| 412 | /// [Intel's documentation](https://www.intel.com/content/www/us/en/docs/intrinsics-guide/index.html#text=_mm256_mask_bitshuffle_epi64_mask) |
| 413 | #[inline ] |
| 414 | #[target_feature (enable = "avx512bitalg,avx512vl" )] |
| 415 | #[stable (feature = "stdarch_x86_avx512" , since = "1.89" )] |
| 416 | #[cfg_attr (test, assert_instr(vpshufbitqmb))] |
| 417 | pub fn _mm256_mask_bitshuffle_epi64_mask(k: __mmask32, b: __m256i, c: __m256i) -> __mmask32 { |
| 418 | unsafe { bitshuffle_256(data:b.as_i8x32(), indices:c.as_i8x32(), mask:k) } |
| 419 | } |
| 420 | |
| 421 | /// Considers the input `b` as packed 64-bit integers and `c` as packed 8-bit integers. |
| 422 | /// Then groups 8 8-bit values from `c`as indices into the bits of the corresponding 64-bit integer. |
| 423 | /// It then selects these bits and packs them into the output. |
| 424 | /// |
| 425 | /// [Intel's documentation](https://www.intel.com/content/www/us/en/docs/intrinsics-guide/index.html#text=_mm_bitshuffle_epi64_mask) |
| 426 | #[inline ] |
| 427 | #[target_feature (enable = "avx512bitalg,avx512vl" )] |
| 428 | #[stable (feature = "stdarch_x86_avx512" , since = "1.89" )] |
| 429 | #[cfg_attr (test, assert_instr(vpshufbitqmb))] |
| 430 | pub fn _mm_bitshuffle_epi64_mask(b: __m128i, c: __m128i) -> __mmask16 { |
| 431 | unsafe { bitshuffle_128(data:b.as_i8x16(), indices:c.as_i8x16(), !0) } |
| 432 | } |
| 433 | |
| 434 | /// Considers the input `b` as packed 64-bit integers and `c` as packed 8-bit integers. |
| 435 | /// Then groups 8 8-bit values from `c`as indices into the bits of the corresponding 64-bit integer. |
| 436 | /// It then selects these bits and packs them into the output. |
| 437 | /// |
| 438 | /// Uses the writemask in k - elements are zeroed in the result if the corresponding mask bit is not set. |
| 439 | /// Otherwise the computation result is written into the result. |
| 440 | /// |
| 441 | /// [Intel's documentation](https://www.intel.com/content/www/us/en/docs/intrinsics-guide/index.html#text=_mm_mask_bitshuffle_epi64_mask) |
| 442 | #[inline ] |
| 443 | #[target_feature (enable = "avx512bitalg,avx512vl" )] |
| 444 | #[stable (feature = "stdarch_x86_avx512" , since = "1.89" )] |
| 445 | #[cfg_attr (test, assert_instr(vpshufbitqmb))] |
| 446 | pub fn _mm_mask_bitshuffle_epi64_mask(k: __mmask16, b: __m128i, c: __m128i) -> __mmask16 { |
| 447 | unsafe { bitshuffle_128(data:b.as_i8x16(), indices:c.as_i8x16(), mask:k) } |
| 448 | } |
| 449 | |
| 450 | #[cfg (test)] |
| 451 | mod tests { |
| 452 | // Some of the constants in the tests below are just bit patterns. They should not |
| 453 | // be interpreted as integers; signedness does not make sense for them, but |
| 454 | // __mXXXi happens to be defined in terms of signed integers. |
| 455 | #![allow (overflowing_literals)] |
| 456 | |
| 457 | use crate::core_arch::assert_eq_const as assert_eq; |
| 458 | use stdarch_test::simd_test; |
| 459 | |
| 460 | use crate::core_arch::x86::*; |
| 461 | |
| 462 | #[simd_test(enable = "avx512bitalg,avx512f" )] |
| 463 | const fn test_mm512_popcnt_epi16() { |
| 464 | let test_data = _mm512_set_epi16( |
| 465 | 0, 1, 3, 7, 0xF, 0x1F, 0x3F, 0x7F, 0xFF, 0x1_FF, 0x3_FF, 0x7_FF, 0xF_FF, 0x1F_FF, |
| 466 | 0x3F_FF, 0x7F_FF, 0xFF_FF, -1, -100, 255, 256, 2, 4, 8, 16, 32, 64, 128, 256, 512, |
| 467 | 1024, 2048, |
| 468 | ); |
| 469 | let actual_result = _mm512_popcnt_epi16(test_data); |
| 470 | let reference_result = _mm512_set_epi16( |
| 471 | 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 16, 12, 8, 1, 1, 1, 1, 1, 1, |
| 472 | 1, 1, 1, 1, 1, 1, |
| 473 | ); |
| 474 | assert_eq_m512i(actual_result, reference_result); |
| 475 | } |
| 476 | |
| 477 | #[simd_test(enable = "avx512bitalg,avx512f" )] |
| 478 | const fn test_mm512_maskz_popcnt_epi16() { |
| 479 | let test_data = _mm512_set_epi16( |
| 480 | 0, 1, 3, 7, 0xF, 0x1F, 0x3F, 0x7F, 0xFF, 0x1_FF, 0x3_FF, 0x7_FF, 0xF_FF, 0x1F_FF, |
| 481 | 0x3F_FF, 0x7F_FF, 0xFF_FF, -1, -100, 255, 256, 2, 4, 8, 16, 32, 64, 128, 256, 512, |
| 482 | 1024, 2048, |
| 483 | ); |
| 484 | let mask = 0xFF_FF_00_00; |
| 485 | let actual_result = _mm512_maskz_popcnt_epi16(mask, test_data); |
| 486 | let reference_result = _mm512_set_epi16( |
| 487 | 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, |
| 488 | 0, 0, 0, 0, 0, |
| 489 | ); |
| 490 | assert_eq_m512i(actual_result, reference_result); |
| 491 | } |
| 492 | |
| 493 | #[simd_test(enable = "avx512bitalg,avx512f" )] |
| 494 | const fn test_mm512_mask_popcnt_epi16() { |
| 495 | let test_data = _mm512_set_epi16( |
| 496 | 0, 1, 3, 7, 0xF, 0x1F, 0x3F, 0x7F, 0xFF, 0x1_FF, 0x3_FF, 0x7_FF, 0xF_FF, 0x1F_FF, |
| 497 | 0x3F_FF, 0x7F_FF, 0xFF_FF, -1, -100, 255, 256, 2, 4, 8, 16, 32, 64, 128, 256, 512, |
| 498 | 1024, 2048, |
| 499 | ); |
| 500 | let mask = 0xFF_FF_00_00; |
| 501 | let actual_result = _mm512_mask_popcnt_epi16(test_data, mask, test_data); |
| 502 | let reference_result = _mm512_set_epi16( |
| 503 | 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 0xFF_FF, -1, -100, 255, 256, 2, |
| 504 | 4, 8, 16, 32, 64, 128, 256, 512, 1024, 2048, |
| 505 | ); |
| 506 | assert_eq_m512i(actual_result, reference_result); |
| 507 | } |
| 508 | |
| 509 | #[simd_test(enable = "avx512bitalg,avx512f,avx512vl" )] |
| 510 | const fn test_mm256_popcnt_epi16() { |
| 511 | let test_data = _mm256_set_epi16( |
| 512 | 0, 1, 3, 7, 0xF, 0x1F, 0x3F, 0x7F, 0xFF, 0x1_FF, 0x3_FF, 0x7_FF, 0xF_FF, 0x1F_FF, |
| 513 | 0x3F_FF, 0x7F_FF, |
| 514 | ); |
| 515 | let actual_result = _mm256_popcnt_epi16(test_data); |
| 516 | let reference_result = |
| 517 | _mm256_set_epi16(0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15); |
| 518 | assert_eq_m256i(actual_result, reference_result); |
| 519 | } |
| 520 | |
| 521 | #[simd_test(enable = "avx512bitalg,avx512f,avx512vl" )] |
| 522 | const fn test_mm256_maskz_popcnt_epi16() { |
| 523 | let test_data = _mm256_set_epi16( |
| 524 | 0, 1, 3, 7, 0xF, 0x1F, 0x3F, 0x7F, 0xFF, 0x1_FF, 0x3_FF, 0x7_FF, 0xF_FF, 0x1F_FF, |
| 525 | 0x3F_FF, 0x7F_FF, |
| 526 | ); |
| 527 | let mask = 0xFF_00; |
| 528 | let actual_result = _mm256_maskz_popcnt_epi16(mask, test_data); |
| 529 | let reference_result = _mm256_set_epi16(0, 1, 2, 3, 4, 5, 6, 7, 0, 0, 0, 0, 0, 0, 0, 0); |
| 530 | assert_eq_m256i(actual_result, reference_result); |
| 531 | } |
| 532 | |
| 533 | #[simd_test(enable = "avx512bitalg,avx512f,avx512vl" )] |
| 534 | const fn test_mm256_mask_popcnt_epi16() { |
| 535 | let test_data = _mm256_set_epi16( |
| 536 | 0, 1, 3, 7, 0xF, 0x1F, 0x3F, 0x7F, 0xFF, 0x1_FF, 0x3_FF, 0x7_FF, 0xF_FF, 0x1F_FF, |
| 537 | 0x3F_FF, 0x7F_FF, |
| 538 | ); |
| 539 | let mask = 0xFF_00; |
| 540 | let actual_result = _mm256_mask_popcnt_epi16(test_data, mask, test_data); |
| 541 | let reference_result = _mm256_set_epi16( |
| 542 | 0, 1, 2, 3, 4, 5, 6, 7, 0xFF, 0x1_FF, 0x3_FF, 0x7_FF, 0xF_FF, 0x1F_FF, 0x3F_FF, 0x7F_FF, |
| 543 | ); |
| 544 | assert_eq_m256i(actual_result, reference_result); |
| 545 | } |
| 546 | |
| 547 | #[simd_test(enable = "avx512bitalg,avx512f,avx512vl" )] |
| 548 | const fn test_mm_popcnt_epi16() { |
| 549 | let test_data = _mm_set_epi16(0, 1, 3, 7, 0xF, 0x1F, 0x3F, 0x7F); |
| 550 | let actual_result = _mm_popcnt_epi16(test_data); |
| 551 | let reference_result = _mm_set_epi16(0, 1, 2, 3, 4, 5, 6, 7); |
| 552 | assert_eq_m128i(actual_result, reference_result); |
| 553 | } |
| 554 | |
| 555 | #[simd_test(enable = "avx512bitalg,avx512f,avx512vl" )] |
| 556 | const fn test_mm_maskz_popcnt_epi16() { |
| 557 | let test_data = _mm_set_epi16(0, 1, 3, 7, 0xF, 0x1F, 0x3F, 0x7F); |
| 558 | let mask = 0xF0; |
| 559 | let actual_result = _mm_maskz_popcnt_epi16(mask, test_data); |
| 560 | let reference_result = _mm_set_epi16(0, 1, 2, 3, 0, 0, 0, 0); |
| 561 | assert_eq_m128i(actual_result, reference_result); |
| 562 | } |
| 563 | |
| 564 | #[simd_test(enable = "avx512bitalg,avx512f,avx512vl" )] |
| 565 | const fn test_mm_mask_popcnt_epi16() { |
| 566 | let test_data = _mm_set_epi16(0, 1, 3, 7, 0xF, 0x1F, 0x3F, 0x7F); |
| 567 | let mask = 0xF0; |
| 568 | let actual_result = _mm_mask_popcnt_epi16(test_data, mask, test_data); |
| 569 | let reference_result = _mm_set_epi16(0, 1, 2, 3, 0xF, 0x1F, 0x3F, 0x7F); |
| 570 | assert_eq_m128i(actual_result, reference_result); |
| 571 | } |
| 572 | |
| 573 | #[simd_test(enable = "avx512bitalg,avx512f" )] |
| 574 | const fn test_mm512_popcnt_epi8() { |
| 575 | let test_data = _mm512_set_epi8( |
| 576 | 0, 1, 3, 7, 0xF, 0x1F, 0x3F, 0x7F, 0xFF, -1, 2, 4, 8, 16, 32, 64, 128, 171, 206, 100, |
| 577 | 217, 109, 253, 190, 177, 254, 179, 215, 230, 68, 201, 172, 183, 154, 84, 56, 227, 189, |
| 578 | 140, 35, 117, 219, 169, 226, 170, 13, 22, 159, 251, 73, 121, 143, 145, 85, 91, 137, 90, |
| 579 | 225, 21, 249, 211, 155, 228, 70, |
| 580 | ); |
| 581 | let actual_result = _mm512_popcnt_epi8(test_data); |
| 582 | let reference_result = _mm512_set_epi8( |
| 583 | 0, 1, 2, 3, 4, 5, 6, 7, 8, 8, 1, 1, 1, 1, 1, 1, 1, 5, 5, 3, 5, 5, 7, 6, 4, 7, 5, 6, 5, |
| 584 | 2, 4, 4, 6, 4, 3, 3, 5, 6, 3, 3, 5, 6, 4, 4, 4, 3, 3, 6, 7, 3, 5, 5, 3, 4, 5, 3, 4, 4, |
| 585 | 3, 6, 5, 5, 4, 3, |
| 586 | ); |
| 587 | assert_eq_m512i(actual_result, reference_result); |
| 588 | } |
| 589 | |
| 590 | #[simd_test(enable = "avx512bitalg,avx512f" )] |
| 591 | const fn test_mm512_maskz_popcnt_epi8() { |
| 592 | let test_data = _mm512_set_epi8( |
| 593 | 0, 1, 3, 7, 0xF, 0x1F, 0x3F, 0x7F, 0xFF, -1, 2, 4, 8, 16, 32, 64, 128, 171, 206, 100, |
| 594 | 217, 109, 253, 190, 177, 254, 179, 215, 230, 68, 201, 172, 183, 154, 84, 56, 227, 189, |
| 595 | 140, 35, 117, 219, 169, 226, 170, 13, 22, 159, 251, 73, 121, 143, 145, 85, 91, 137, 90, |
| 596 | 225, 21, 249, 211, 155, 228, 70, |
| 597 | ); |
| 598 | let mask = 0xFF_FF_FF_FF_00_00_00_00; |
| 599 | let actual_result = _mm512_maskz_popcnt_epi8(mask, test_data); |
| 600 | let reference_result = _mm512_set_epi8( |
| 601 | 0, 1, 2, 3, 4, 5, 6, 7, 8, 8, 1, 1, 1, 1, 1, 1, 1, 5, 5, 3, 5, 5, 7, 6, 4, 7, 5, 6, 5, |
| 602 | 2, 4, 4, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, |
| 603 | 0, 0, 0, 0, 0, 0, |
| 604 | ); |
| 605 | assert_eq_m512i(actual_result, reference_result); |
| 606 | } |
| 607 | |
| 608 | #[simd_test(enable = "avx512bitalg,avx512f" )] |
| 609 | const fn test_mm512_mask_popcnt_epi8() { |
| 610 | let test_data = _mm512_set_epi8( |
| 611 | 0, 1, 3, 7, 0xF, 0x1F, 0x3F, 0x7F, 0xFF, -1, 2, 4, 8, 16, 32, 64, 128, 171, 206, 100, |
| 612 | 217, 109, 253, 190, 177, 254, 179, 215, 230, 68, 201, 172, 183, 154, 84, 56, 227, 189, |
| 613 | 140, 35, 117, 219, 169, 226, 170, 13, 22, 159, 251, 73, 121, 143, 145, 85, 91, 137, 90, |
| 614 | 225, 21, 249, 211, 155, 228, 70, |
| 615 | ); |
| 616 | let mask = 0xFF_FF_FF_FF_00_00_00_00; |
| 617 | let actual_result = _mm512_mask_popcnt_epi8(test_data, mask, test_data); |
| 618 | let reference_result = _mm512_set_epi8( |
| 619 | 0, 1, 2, 3, 4, 5, 6, 7, 8, 8, 1, 1, 1, 1, 1, 1, 1, 5, 5, 3, 5, 5, 7, 6, 4, 7, 5, 6, 5, |
| 620 | 2, 4, 4, 183, 154, 84, 56, 227, 189, 140, 35, 117, 219, 169, 226, 170, 13, 22, 159, |
| 621 | 251, 73, 121, 143, 145, 85, 91, 137, 90, 225, 21, 249, 211, 155, 228, 70, |
| 622 | ); |
| 623 | assert_eq_m512i(actual_result, reference_result); |
| 624 | } |
| 625 | |
| 626 | #[simd_test(enable = "avx512bitalg,avx512f,avx512vl" )] |
| 627 | const fn test_mm256_popcnt_epi8() { |
| 628 | let test_data = _mm256_set_epi8( |
| 629 | 0, 1, 3, 7, 0xF, 0x1F, 0x3F, 0x7F, 0xFF, -1, 2, 4, 8, 16, 32, 64, 128, 171, 206, 100, |
| 630 | 217, 109, 253, 190, 177, 254, 179, 215, 230, 68, 201, 172, |
| 631 | ); |
| 632 | let actual_result = _mm256_popcnt_epi8(test_data); |
| 633 | let reference_result = _mm256_set_epi8( |
| 634 | 0, 1, 2, 3, 4, 5, 6, 7, 8, 8, 1, 1, 1, 1, 1, 1, 1, 5, 5, 3, 5, 5, 7, 6, 4, 7, 5, 6, 5, |
| 635 | 2, 4, 4, |
| 636 | ); |
| 637 | assert_eq_m256i(actual_result, reference_result); |
| 638 | } |
| 639 | |
| 640 | #[simd_test(enable = "avx512bitalg,avx512f,avx512vl" )] |
| 641 | const fn test_mm256_maskz_popcnt_epi8() { |
| 642 | let test_data = _mm256_set_epi8( |
| 643 | 0, 1, 3, 7, 0xF, 0x1F, 0x3F, 0x7F, 0xFF, -1, 2, 4, 8, 16, 32, 64, 251, 73, 121, 143, |
| 644 | 145, 85, 91, 137, 90, 225, 21, 249, 211, 155, 228, 70, |
| 645 | ); |
| 646 | let mask = 0xFF_FF_00_00; |
| 647 | let actual_result = _mm256_maskz_popcnt_epi8(mask, test_data); |
| 648 | let reference_result = _mm256_set_epi8( |
| 649 | 0, 1, 2, 3, 4, 5, 6, 7, 8, 8, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, |
| 650 | 0, 0, 0, |
| 651 | ); |
| 652 | assert_eq_m256i(actual_result, reference_result); |
| 653 | } |
| 654 | |
| 655 | #[simd_test(enable = "avx512bitalg,avx512f,avx512vl" )] |
| 656 | const fn test_mm256_mask_popcnt_epi8() { |
| 657 | let test_data = _mm256_set_epi8( |
| 658 | 0, 1, 3, 7, 0xF, 0x1F, 0x3F, 0x7F, 0xFF, -1, 2, 4, 8, 16, 32, 64, 251, 73, 121, 143, |
| 659 | 145, 85, 91, 137, 90, 225, 21, 249, 211, 155, 228, 70, |
| 660 | ); |
| 661 | let mask = 0xFF_FF_00_00; |
| 662 | let actual_result = _mm256_mask_popcnt_epi8(test_data, mask, test_data); |
| 663 | let reference_result = _mm256_set_epi8( |
| 664 | 0, 1, 2, 3, 4, 5, 6, 7, 8, 8, 1, 1, 1, 1, 1, 1, 251, 73, 121, 143, 145, 85, 91, 137, |
| 665 | 90, 225, 21, 249, 211, 155, 228, 70, |
| 666 | ); |
| 667 | assert_eq_m256i(actual_result, reference_result); |
| 668 | } |
| 669 | |
| 670 | #[simd_test(enable = "avx512bitalg,avx512f,avx512vl" )] |
| 671 | const fn test_mm_popcnt_epi8() { |
| 672 | let test_data = _mm_set_epi8( |
| 673 | 0, 1, 3, 7, 0xF, 0x1F, 0x3F, 0x7F, 0xFF, -1, 2, 4, 8, 16, 32, 64, |
| 674 | ); |
| 675 | let actual_result = _mm_popcnt_epi8(test_data); |
| 676 | let reference_result = _mm_set_epi8(0, 1, 2, 3, 4, 5, 6, 7, 8, 8, 1, 1, 1, 1, 1, 1); |
| 677 | assert_eq_m128i(actual_result, reference_result); |
| 678 | } |
| 679 | |
| 680 | #[simd_test(enable = "avx512bitalg,avx512f,avx512vl" )] |
| 681 | const fn test_mm_maskz_popcnt_epi8() { |
| 682 | let test_data = _mm_set_epi8( |
| 683 | 0, 1, 3, 7, 0xF, 0x1F, 0x3F, 0x7F, 90, 225, 21, 249, 211, 155, 228, 70, |
| 684 | ); |
| 685 | let mask = 0xFF_00; |
| 686 | let actual_result = _mm_maskz_popcnt_epi8(mask, test_data); |
| 687 | let reference_result = _mm_set_epi8(0, 1, 2, 3, 4, 5, 6, 7, 0, 0, 0, 0, 0, 0, 0, 0); |
| 688 | assert_eq_m128i(actual_result, reference_result); |
| 689 | } |
| 690 | |
| 691 | #[simd_test(enable = "avx512bitalg,avx512f,avx512vl" )] |
| 692 | const fn test_mm_mask_popcnt_epi8() { |
| 693 | let test_data = _mm_set_epi8( |
| 694 | 0, 1, 3, 7, 0xF, 0x1F, 0x3F, 0x7F, 90, 225, 21, 249, 211, 155, 228, 70, |
| 695 | ); |
| 696 | let mask = 0xFF_00; |
| 697 | let actual_result = _mm_mask_popcnt_epi8(test_data, mask, test_data); |
| 698 | let reference_result = |
| 699 | _mm_set_epi8(0, 1, 2, 3, 4, 5, 6, 7, 90, 225, 21, 249, 211, 155, 228, 70); |
| 700 | assert_eq_m128i(actual_result, reference_result); |
| 701 | } |
| 702 | |
| 703 | #[simd_test(enable = "avx512bitalg,avx512f" )] |
| 704 | fn test_mm512_bitshuffle_epi64_mask() { |
| 705 | let test_indices = _mm512_set_epi8( |
| 706 | 63, 62, 61, 60, 59, 58, 57, 56, 63, 62, 61, 60, 59, 58, 57, 56, 32, 32, 16, 16, 0, 0, |
| 707 | 8, 8, 56, 48, 40, 32, 24, 16, 8, 0, 63, 62, 61, 60, 59, 58, 57, 56, 63, 62, 61, 60, 59, |
| 708 | 58, 57, 56, 32, 32, 16, 16, 0, 0, 8, 8, 56, 48, 40, 32, 24, 16, 8, 0, |
| 709 | ); |
| 710 | let test_data = _mm512_setr_epi64( |
| 711 | 0xFF_FF_FF_FF_00_00_00_00, |
| 712 | 0xFF_00_FF_00_FF_00_FF_00, |
| 713 | 0xFF_00_00_00_00_00_00_00, |
| 714 | 0xAC_00_00_00_00_00_00_00, |
| 715 | 0xFF_FF_FF_FF_00_00_00_00, |
| 716 | 0xFF_00_FF_00_FF_00_FF_00, |
| 717 | 0xFF_00_00_00_00_00_00_00, |
| 718 | 0xAC_00_00_00_00_00_00_00, |
| 719 | ); |
| 720 | let actual_result = _mm512_bitshuffle_epi64_mask(test_data, test_indices); |
| 721 | let reference_result = 0xF0 << 0 |
| 722 | | 0x03 << 8 |
| 723 | | 0xFF << 16 |
| 724 | | 0xAC << 24 |
| 725 | | 0xF0 << 32 |
| 726 | | 0x03 << 40 |
| 727 | | 0xFF << 48 |
| 728 | | 0xAC << 56; |
| 729 | |
| 730 | assert_eq!(actual_result, reference_result); |
| 731 | } |
| 732 | |
| 733 | #[simd_test(enable = "avx512bitalg,avx512f" )] |
| 734 | fn test_mm512_mask_bitshuffle_epi64_mask() { |
| 735 | let test_indices = _mm512_set_epi8( |
| 736 | 63, 62, 61, 60, 59, 58, 57, 56, 63, 62, 61, 60, 59, 58, 57, 56, 32, 32, 16, 16, 0, 0, |
| 737 | 8, 8, 56, 48, 40, 32, 24, 16, 8, 0, 63, 62, 61, 60, 59, 58, 57, 56, 63, 62, 61, 60, 59, |
| 738 | 58, 57, 56, 32, 32, 16, 16, 0, 0, 8, 8, 56, 48, 40, 32, 24, 16, 8, 0, |
| 739 | ); |
| 740 | let test_data = _mm512_setr_epi64( |
| 741 | 0xFF_FF_FF_FF_00_00_00_00, |
| 742 | 0xFF_00_FF_00_FF_00_FF_00, |
| 743 | 0xFF_00_00_00_00_00_00_00, |
| 744 | 0xAC_00_00_00_00_00_00_00, |
| 745 | 0xFF_FF_FF_FF_00_00_00_00, |
| 746 | 0xFF_00_FF_00_FF_00_FF_00, |
| 747 | 0xFF_00_00_00_00_00_00_00, |
| 748 | 0xAC_00_00_00_00_00_00_00, |
| 749 | ); |
| 750 | let mask = 0xFF_FF_FF_FF_00_00_00_00; |
| 751 | let actual_result = _mm512_mask_bitshuffle_epi64_mask(mask, test_data, test_indices); |
| 752 | let reference_result = 0x00 << 0 |
| 753 | | 0x00 << 8 |
| 754 | | 0x00 << 16 |
| 755 | | 0x00 << 24 |
| 756 | | 0xF0 << 32 |
| 757 | | 0x03 << 40 |
| 758 | | 0xFF << 48 |
| 759 | | 0xAC << 56; |
| 760 | |
| 761 | assert_eq!(actual_result, reference_result); |
| 762 | } |
| 763 | |
| 764 | #[simd_test(enable = "avx512bitalg,avx512f,avx512vl" )] |
| 765 | fn test_mm256_bitshuffle_epi64_mask() { |
| 766 | let test_indices = _mm256_set_epi8( |
| 767 | 63, 62, 61, 60, 59, 58, 57, 56, 63, 62, 61, 60, 59, 58, 57, 56, 32, 32, 16, 16, 0, 0, |
| 768 | 8, 8, 56, 48, 40, 32, 24, 16, 8, 0, |
| 769 | ); |
| 770 | let test_data = _mm256_setr_epi64x( |
| 771 | 0xFF_FF_FF_FF_00_00_00_00, |
| 772 | 0xFF_00_FF_00_FF_00_FF_00, |
| 773 | 0xFF_00_00_00_00_00_00_00, |
| 774 | 0xAC_00_00_00_00_00_00_00, |
| 775 | ); |
| 776 | let actual_result = _mm256_bitshuffle_epi64_mask(test_data, test_indices); |
| 777 | let reference_result = 0xF0 << 0 | 0x03 << 8 | 0xFF << 16 | 0xAC << 24; |
| 778 | |
| 779 | assert_eq!(actual_result, reference_result); |
| 780 | } |
| 781 | |
| 782 | #[simd_test(enable = "avx512bitalg,avx512f,avx512vl" )] |
| 783 | fn test_mm256_mask_bitshuffle_epi64_mask() { |
| 784 | let test_indices = _mm256_set_epi8( |
| 785 | 63, 62, 61, 60, 59, 58, 57, 56, 63, 62, 61, 60, 59, 58, 57, 56, 32, 32, 16, 16, 0, 0, |
| 786 | 8, 8, 56, 48, 40, 32, 24, 16, 8, 0, |
| 787 | ); |
| 788 | let test_data = _mm256_setr_epi64x( |
| 789 | 0xFF_FF_FF_FF_00_00_00_00, |
| 790 | 0xFF_00_FF_00_FF_00_FF_00, |
| 791 | 0xFF_00_00_00_00_00_00_00, |
| 792 | 0xAC_00_00_00_00_00_00_00, |
| 793 | ); |
| 794 | let mask = 0xFF_FF_00_00; |
| 795 | let actual_result = _mm256_mask_bitshuffle_epi64_mask(mask, test_data, test_indices); |
| 796 | let reference_result = 0x00 << 0 | 0x00 << 8 | 0xFF << 16 | 0xAC << 24; |
| 797 | |
| 798 | assert_eq!(actual_result, reference_result); |
| 799 | } |
| 800 | |
| 801 | #[simd_test(enable = "avx512bitalg,avx512f,avx512vl" )] |
| 802 | fn test_mm_bitshuffle_epi64_mask() { |
| 803 | let test_indices = _mm_set_epi8( |
| 804 | 63, 62, 61, 60, 59, 58, 57, 56, 63, 62, 61, 60, 59, 58, 57, 56, |
| 805 | ); |
| 806 | let test_data = _mm_setr_epi64x(0xFF_00_00_00_00_00_00_00, 0xAC_00_00_00_00_00_00_00); |
| 807 | let actual_result = _mm_bitshuffle_epi64_mask(test_data, test_indices); |
| 808 | let reference_result = 0xFF << 0 | 0xAC << 8; |
| 809 | |
| 810 | assert_eq!(actual_result, reference_result); |
| 811 | } |
| 812 | |
| 813 | #[simd_test(enable = "avx512bitalg,avx512f,avx512vl" )] |
| 814 | fn test_mm_mask_bitshuffle_epi64_mask() { |
| 815 | let test_indices = _mm_set_epi8( |
| 816 | 63, 62, 61, 60, 59, 58, 57, 56, 63, 62, 61, 60, 59, 58, 57, 56, |
| 817 | ); |
| 818 | let test_data = _mm_setr_epi64x(0xFF_00_00_00_00_00_00_00, 0xAC_00_00_00_00_00_00_00); |
| 819 | let mask = 0xFF_00; |
| 820 | let actual_result = _mm_mask_bitshuffle_epi64_mask(mask, test_data, test_indices); |
| 821 | let reference_result = 0x00 << 0 | 0xAC << 8; |
| 822 | |
| 823 | assert_eq!(actual_result, reference_result); |
| 824 | } |
| 825 | } |
| 826 | |