1 /* 2 * Copyright 2012 The LibYuv Project Authors. All rights reserved. 3 * 4 * Use of this source code is governed by a BSD-style license 5 * that can be found in the LICENSE file in the root of the source 6 * tree. An additional intellectual property rights grant can be found 7 * in the file PATENTS. All contributing project authors may 8 * be found in the AUTHORS file in the root of the source tree. 9 */ 10 11 #include "libyuv/row.h" 12 13 #include <string.h> // For memset. 14 15 #include "libyuv/basic_types.h" 16 17 #ifdef __cplusplus 18 namespace libyuv { 19 extern "C" { 20 #endif 21 22 // memset for temp is meant to clear the source buffer (not dest) so that 23 // SIMD that reads full multiple of 16 bytes will not trigger msan errors. 24 // memset is not needed for production, as the garbage values are processed but 25 // not used, although there may be edge cases for subsampling. 26 // The size of the buffer is based on the largest read, which can be inferred 27 // by the source type (e.g. ARGB) and the mask (last parameter), or by examining 28 // the source code for how much the source pointers are advanced. 29 30 // Subsampled source needs to be increase by 1 of not even. 31 #define SS(width, shift) (((width) + (1 << (shift)) - 1) >> (shift)) 32 33 // Any 4 planes to 1 with yuvconstants 34 #define ANY41C(NAMEANY, ANY_SIMD, UVSHIFT, DUVSHIFT, BPP, MASK) \ 35 void NAMEANY(const uint8_t* y_buf, const uint8_t* u_buf, \ 36 const uint8_t* v_buf, const uint8_t* a_buf, uint8_t* dst_ptr, \ 37 const struct YuvConstants* yuvconstants, int width) { \ 38 SIMD_ALIGNED(uint8_t temp[64 * 5]); \ 39 memset(temp, 0, 64 * 4); /* for msan */ \ 40 int r = width & MASK; \ 41 int n = width & ~MASK; \ 42 if (n > 0) { \ 43 ANY_SIMD(y_buf, u_buf, v_buf, a_buf, dst_ptr, yuvconstants, n); \ 44 } \ 45 memcpy(temp, y_buf + n, r); \ 46 memcpy(temp + 64, u_buf + (n >> UVSHIFT), SS(r, UVSHIFT)); \ 47 memcpy(temp + 128, v_buf + (n >> UVSHIFT), SS(r, UVSHIFT)); \ 48 memcpy(temp + 192, a_buf + n, r); \ 49 ANY_SIMD(temp, temp + 64, temp + 128, temp + 192, temp + 256, \ 50 yuvconstants, MASK + 1); \ 51 memcpy(dst_ptr + (n >> DUVSHIFT) * BPP, temp + 256, \ 52 SS(r, DUVSHIFT) * BPP); \ 53 } 54 55 #ifdef HAS_I422ALPHATOARGBROW_SSSE3 56 ANY41C(I422AlphaToARGBRow_Any_SSSE3, I422AlphaToARGBRow_SSSE3, 1, 0, 4, 7) 57 #endif 58 #ifdef HAS_I422ALPHATOARGBROW_AVX2 59 ANY41C(I422AlphaToARGBRow_Any_AVX2, I422AlphaToARGBRow_AVX2, 1, 0, 4, 15) 60 #endif 61 #ifdef HAS_I422ALPHATOARGBROW_NEON 62 ANY41C(I422AlphaToARGBRow_Any_NEON, I422AlphaToARGBRow_NEON, 1, 0, 4, 7) 63 #endif 64 #ifdef HAS_I422ALPHATOARGBROW_MSA 65 ANY41C(I422AlphaToARGBRow_Any_MSA, I422AlphaToARGBRow_MSA, 1, 0, 4, 7) 66 #endif 67 #undef ANY41C 68 69 // Any 3 planes to 1. 70 #define ANY31(NAMEANY, ANY_SIMD, UVSHIFT, DUVSHIFT, BPP, MASK) \ 71 void NAMEANY(const uint8_t* y_buf, const uint8_t* u_buf, \ 72 const uint8_t* v_buf, uint8_t* dst_ptr, int width) { \ 73 SIMD_ALIGNED(uint8_t temp[64 * 4]); \ 74 memset(temp, 0, 64 * 3); /* for YUY2 and msan */ \ 75 int r = width & MASK; \ 76 int n = width & ~MASK; \ 77 if (n > 0) { \ 78 ANY_SIMD(y_buf, u_buf, v_buf, dst_ptr, n); \ 79 } \ 80 memcpy(temp, y_buf + n, r); \ 81 memcpy(temp + 64, u_buf + (n >> UVSHIFT), SS(r, UVSHIFT)); \ 82 memcpy(temp + 128, v_buf + (n >> UVSHIFT), SS(r, UVSHIFT)); \ 83 ANY_SIMD(temp, temp + 64, temp + 128, temp + 192, MASK + 1); \ 84 memcpy(dst_ptr + (n >> DUVSHIFT) * BPP, temp + 192, \ 85 SS(r, DUVSHIFT) * BPP); \ 86 } 87 88 // Merge functions. 89 #ifdef HAS_MERGERGBROW_SSSE3 90 ANY31(MergeRGBRow_Any_SSSE3, MergeRGBRow_SSSE3, 0, 0, 3, 15) 91 #endif 92 #ifdef HAS_MERGERGBROW_NEON 93 ANY31(MergeRGBRow_Any_NEON, MergeRGBRow_NEON, 0, 0, 3, 15) 94 #endif 95 #ifdef HAS_I422TOYUY2ROW_SSE2 96 ANY31(I422ToYUY2Row_Any_SSE2, I422ToYUY2Row_SSE2, 1, 1, 4, 15) 97 ANY31(I422ToUYVYRow_Any_SSE2, I422ToUYVYRow_SSE2, 1, 1, 4, 15) 98 #endif 99 #ifdef HAS_I422TOYUY2ROW_AVX2 100 ANY31(I422ToYUY2Row_Any_AVX2, I422ToYUY2Row_AVX2, 1, 1, 4, 31) 101 ANY31(I422ToUYVYRow_Any_AVX2, I422ToUYVYRow_AVX2, 1, 1, 4, 31) 102 #endif 103 #ifdef HAS_I422TOYUY2ROW_NEON 104 ANY31(I422ToYUY2Row_Any_NEON, I422ToYUY2Row_NEON, 1, 1, 4, 15) 105 #endif 106 #ifdef HAS_I422TOYUY2ROW_MSA 107 ANY31(I422ToYUY2Row_Any_MSA, I422ToYUY2Row_MSA, 1, 1, 4, 31) 108 #endif 109 #ifdef HAS_I422TOUYVYROW_NEON 110 ANY31(I422ToUYVYRow_Any_NEON, I422ToUYVYRow_NEON, 1, 1, 4, 15) 111 #endif 112 #ifdef HAS_I422TOUYVYROW_MSA 113 ANY31(I422ToUYVYRow_Any_MSA, I422ToUYVYRow_MSA, 1, 1, 4, 31) 114 #endif 115 #ifdef HAS_BLENDPLANEROW_AVX2 116 ANY31(BlendPlaneRow_Any_AVX2, BlendPlaneRow_AVX2, 0, 0, 1, 31) 117 #endif 118 #ifdef HAS_BLENDPLANEROW_SSSE3 119 ANY31(BlendPlaneRow_Any_SSSE3, BlendPlaneRow_SSSE3, 0, 0, 1, 7) 120 #endif 121 #undef ANY31 122 123 // Note that odd width replication includes 444 due to implementation 124 // on arm that subsamples 444 to 422 internally. 125 // Any 3 planes to 1 with yuvconstants 126 #define ANY31C(NAMEANY, ANY_SIMD, UVSHIFT, DUVSHIFT, BPP, MASK) \ 127 void NAMEANY(const uint8_t* y_buf, const uint8_t* u_buf, \ 128 const uint8_t* v_buf, uint8_t* dst_ptr, \ 129 const struct YuvConstants* yuvconstants, int width) { \ 130 SIMD_ALIGNED(uint8_t temp[128 * 4]); \ 131 memset(temp, 0, 128 * 3); /* for YUY2 and msan */ \ 132 int r = width & MASK; \ 133 int n = width & ~MASK; \ 134 if (n > 0) { \ 135 ANY_SIMD(y_buf, u_buf, v_buf, dst_ptr, yuvconstants, n); \ 136 } \ 137 memcpy(temp, y_buf + n, r); \ 138 memcpy(temp + 128, u_buf + (n >> UVSHIFT), SS(r, UVSHIFT)); \ 139 memcpy(temp + 256, v_buf + (n >> UVSHIFT), SS(r, UVSHIFT)); \ 140 if (width & 1) { \ 141 temp[128 + SS(r, UVSHIFT)] = temp[128 + SS(r, UVSHIFT) - 1]; \ 142 temp[256 + SS(r, UVSHIFT)] = temp[256 + SS(r, UVSHIFT) - 1]; \ 143 } \ 144 ANY_SIMD(temp, temp + 128, temp + 256, temp + 384, yuvconstants, \ 145 MASK + 1); \ 146 memcpy(dst_ptr + (n >> DUVSHIFT) * BPP, temp + 384, \ 147 SS(r, DUVSHIFT) * BPP); \ 148 } 149 150 #ifdef HAS_I422TOARGBROW_SSSE3 151 ANY31C(I422ToARGBRow_Any_SSSE3, I422ToARGBRow_SSSE3, 1, 0, 4, 7) 152 #endif 153 #ifdef HAS_I422TOAR30ROW_SSSE3 154 ANY31C(I422ToAR30Row_Any_SSSE3, I422ToAR30Row_SSSE3, 1, 0, 4, 7) 155 #endif 156 #ifdef HAS_I422TOAR30ROW_AVX2 157 ANY31C(I422ToAR30Row_Any_AVX2, I422ToAR30Row_AVX2, 1, 0, 4, 15) 158 #endif 159 #ifdef HAS_I444TOARGBROW_SSSE3 160 ANY31C(I444ToARGBRow_Any_SSSE3, I444ToARGBRow_SSSE3, 0, 0, 4, 7) 161 ANY31C(I422ToRGBARow_Any_SSSE3, I422ToRGBARow_SSSE3, 1, 0, 4, 7) 162 ANY31C(I422ToARGB4444Row_Any_SSSE3, I422ToARGB4444Row_SSSE3, 1, 0, 2, 7) 163 ANY31C(I422ToARGB1555Row_Any_SSSE3, I422ToARGB1555Row_SSSE3, 1, 0, 2, 7) 164 ANY31C(I422ToRGB565Row_Any_SSSE3, I422ToRGB565Row_SSSE3, 1, 0, 2, 7) 165 ANY31C(I422ToRGB24Row_Any_SSSE3, I422ToRGB24Row_SSSE3, 1, 0, 3, 15) 166 #endif // HAS_I444TOARGBROW_SSSE3 167 #ifdef HAS_I422TORGB24ROW_AVX2 168 ANY31C(I422ToRGB24Row_Any_AVX2, I422ToRGB24Row_AVX2, 1, 0, 3, 31) 169 #endif 170 #ifdef HAS_I422TOARGBROW_AVX2 171 ANY31C(I422ToARGBRow_Any_AVX2, I422ToARGBRow_AVX2, 1, 0, 4, 15) 172 #endif 173 #ifdef HAS_I422TORGBAROW_AVX2 174 ANY31C(I422ToRGBARow_Any_AVX2, I422ToRGBARow_AVX2, 1, 0, 4, 15) 175 #endif 176 #ifdef HAS_I444TOARGBROW_AVX2 177 ANY31C(I444ToARGBRow_Any_AVX2, I444ToARGBRow_AVX2, 0, 0, 4, 15) 178 #endif 179 #ifdef HAS_I422TOARGB4444ROW_AVX2 180 ANY31C(I422ToARGB4444Row_Any_AVX2, I422ToARGB4444Row_AVX2, 1, 0, 2, 15) 181 #endif 182 #ifdef HAS_I422TOARGB1555ROW_AVX2 183 ANY31C(I422ToARGB1555Row_Any_AVX2, I422ToARGB1555Row_AVX2, 1, 0, 2, 15) 184 #endif 185 #ifdef HAS_I422TORGB565ROW_AVX2 186 ANY31C(I422ToRGB565Row_Any_AVX2, I422ToRGB565Row_AVX2, 1, 0, 2, 15) 187 #endif 188 #ifdef HAS_I422TOARGBROW_NEON 189 ANY31C(I444ToARGBRow_Any_NEON, I444ToARGBRow_NEON, 0, 0, 4, 7) 190 ANY31C(I422ToARGBRow_Any_NEON, I422ToARGBRow_NEON, 1, 0, 4, 7) 191 ANY31C(I422ToRGBARow_Any_NEON, I422ToRGBARow_NEON, 1, 0, 4, 7) 192 ANY31C(I422ToRGB24Row_Any_NEON, I422ToRGB24Row_NEON, 1, 0, 3, 7) 193 ANY31C(I422ToARGB4444Row_Any_NEON, I422ToARGB4444Row_NEON, 1, 0, 2, 7) 194 ANY31C(I422ToARGB1555Row_Any_NEON, I422ToARGB1555Row_NEON, 1, 0, 2, 7) 195 ANY31C(I422ToRGB565Row_Any_NEON, I422ToRGB565Row_NEON, 1, 0, 2, 7) 196 #endif 197 #ifdef HAS_I422TOARGBROW_MSA 198 ANY31C(I444ToARGBRow_Any_MSA, I444ToARGBRow_MSA, 0, 0, 4, 7) 199 ANY31C(I422ToARGBRow_Any_MSA, I422ToARGBRow_MSA, 1, 0, 4, 7) 200 ANY31C(I422ToRGBARow_Any_MSA, I422ToRGBARow_MSA, 1, 0, 4, 7) 201 ANY31C(I422ToRGB24Row_Any_MSA, I422ToRGB24Row_MSA, 1, 0, 3, 15) 202 ANY31C(I422ToARGB4444Row_Any_MSA, I422ToARGB4444Row_MSA, 1, 0, 2, 7) 203 ANY31C(I422ToARGB1555Row_Any_MSA, I422ToARGB1555Row_MSA, 1, 0, 2, 7) 204 ANY31C(I422ToRGB565Row_Any_MSA, I422ToRGB565Row_MSA, 1, 0, 2, 7) 205 #endif 206 #undef ANY31C 207 208 // Any 3 planes of 16 bit to 1 with yuvconstants 209 // TODO(fbarchard): consider sharing this code with ANY31C 210 #define ANY31CT(NAMEANY, ANY_SIMD, UVSHIFT, DUVSHIFT, T, SBPP, BPP, MASK) \ 211 void NAMEANY(const T* y_buf, const T* u_buf, const T* v_buf, \ 212 uint8_t* dst_ptr, const struct YuvConstants* yuvconstants, \ 213 int width) { \ 214 SIMD_ALIGNED(T temp[16 * 3]); \ 215 SIMD_ALIGNED(uint8_t out[64]); \ 216 memset(temp, 0, 16 * 3 * SBPP); /* for YUY2 and msan */ \ 217 int r = width & MASK; \ 218 int n = width & ~MASK; \ 219 if (n > 0) { \ 220 ANY_SIMD(y_buf, u_buf, v_buf, dst_ptr, yuvconstants, n); \ 221 } \ 222 memcpy(temp, y_buf + n, r * SBPP); \ 223 memcpy(temp + 16, u_buf + (n >> UVSHIFT), SS(r, UVSHIFT) * SBPP); \ 224 memcpy(temp + 32, v_buf + (n >> UVSHIFT), SS(r, UVSHIFT) * SBPP); \ 225 ANY_SIMD(temp, temp + 16, temp + 32, out, yuvconstants, MASK + 1); \ 226 memcpy(dst_ptr + (n >> DUVSHIFT) * BPP, out, SS(r, DUVSHIFT) * BPP); \ 227 } 228 229 #ifdef HAS_I210TOAR30ROW_SSSE3 230 ANY31CT(I210ToAR30Row_Any_SSSE3, I210ToAR30Row_SSSE3, 1, 0, uint16_t, 2, 4, 7) 231 #endif 232 #ifdef HAS_I210TOARGBROW_SSSE3 233 ANY31CT(I210ToARGBRow_Any_SSSE3, I210ToARGBRow_SSSE3, 1, 0, uint16_t, 2, 4, 7) 234 #endif 235 #ifdef HAS_I210TOARGBROW_AVX2 236 ANY31CT(I210ToARGBRow_Any_AVX2, I210ToARGBRow_AVX2, 1, 0, uint16_t, 2, 4, 15) 237 #endif 238 #ifdef HAS_I210TOAR30ROW_AVX2 239 ANY31CT(I210ToAR30Row_Any_AVX2, I210ToAR30Row_AVX2, 1, 0, uint16_t, 2, 4, 15) 240 #endif 241 #undef ANY31CT 242 243 // Any 2 planes to 1. 244 #define ANY21(NAMEANY, ANY_SIMD, UVSHIFT, SBPP, SBPP2, BPP, MASK) \ 245 void NAMEANY(const uint8_t* y_buf, const uint8_t* uv_buf, uint8_t* dst_ptr, \ 246 int width) { \ 247 SIMD_ALIGNED(uint8_t temp[64 * 3]); \ 248 memset(temp, 0, 64 * 2); /* for msan */ \ 249 int r = width & MASK; \ 250 int n = width & ~MASK; \ 251 if (n > 0) { \ 252 ANY_SIMD(y_buf, uv_buf, dst_ptr, n); \ 253 } \ 254 memcpy(temp, y_buf + n * SBPP, r * SBPP); \ 255 memcpy(temp + 64, uv_buf + (n >> UVSHIFT) * SBPP2, \ 256 SS(r, UVSHIFT) * SBPP2); \ 257 ANY_SIMD(temp, temp + 64, temp + 128, MASK + 1); \ 258 memcpy(dst_ptr + n * BPP, temp + 128, r * BPP); \ 259 } 260 261 // Merge functions. 262 #ifdef HAS_MERGEUVROW_SSE2 263 ANY21(MergeUVRow_Any_SSE2, MergeUVRow_SSE2, 0, 1, 1, 2, 15) 264 #endif 265 #ifdef HAS_MERGEUVROW_AVX2 266 ANY21(MergeUVRow_Any_AVX2, MergeUVRow_AVX2, 0, 1, 1, 2, 31) 267 #endif 268 #ifdef HAS_MERGEUVROW_NEON 269 ANY21(MergeUVRow_Any_NEON, MergeUVRow_NEON, 0, 1, 1, 2, 15) 270 #endif 271 #ifdef HAS_MERGEUVROW_MSA 272 ANY21(MergeUVRow_Any_MSA, MergeUVRow_MSA, 0, 1, 1, 2, 15) 273 #endif 274 275 // Math functions. 276 #ifdef HAS_ARGBMULTIPLYROW_SSE2 277 ANY21(ARGBMultiplyRow_Any_SSE2, ARGBMultiplyRow_SSE2, 0, 4, 4, 4, 3) 278 #endif 279 #ifdef HAS_ARGBADDROW_SSE2 280 ANY21(ARGBAddRow_Any_SSE2, ARGBAddRow_SSE2, 0, 4, 4, 4, 3) 281 #endif 282 #ifdef HAS_ARGBSUBTRACTROW_SSE2 283 ANY21(ARGBSubtractRow_Any_SSE2, ARGBSubtractRow_SSE2, 0, 4, 4, 4, 3) 284 #endif 285 #ifdef HAS_ARGBMULTIPLYROW_AVX2 286 ANY21(ARGBMultiplyRow_Any_AVX2, ARGBMultiplyRow_AVX2, 0, 4, 4, 4, 7) 287 #endif 288 #ifdef HAS_ARGBADDROW_AVX2 289 ANY21(ARGBAddRow_Any_AVX2, ARGBAddRow_AVX2, 0, 4, 4, 4, 7) 290 #endif 291 #ifdef HAS_ARGBSUBTRACTROW_AVX2 292 ANY21(ARGBSubtractRow_Any_AVX2, ARGBSubtractRow_AVX2, 0, 4, 4, 4, 7) 293 #endif 294 #ifdef HAS_ARGBMULTIPLYROW_NEON 295 ANY21(ARGBMultiplyRow_Any_NEON, ARGBMultiplyRow_NEON, 0, 4, 4, 4, 7) 296 #endif 297 #ifdef HAS_ARGBADDROW_NEON 298 ANY21(ARGBAddRow_Any_NEON, ARGBAddRow_NEON, 0, 4, 4, 4, 7) 299 #endif 300 #ifdef HAS_ARGBSUBTRACTROW_NEON 301 ANY21(ARGBSubtractRow_Any_NEON, ARGBSubtractRow_NEON, 0, 4, 4, 4, 7) 302 #endif 303 #ifdef HAS_ARGBMULTIPLYROW_MSA 304 ANY21(ARGBMultiplyRow_Any_MSA, ARGBMultiplyRow_MSA, 0, 4, 4, 4, 3) 305 #endif 306 #ifdef HAS_ARGBADDROW_MSA 307 ANY21(ARGBAddRow_Any_MSA, ARGBAddRow_MSA, 0, 4, 4, 4, 7) 308 #endif 309 #ifdef HAS_ARGBSUBTRACTROW_MSA 310 ANY21(ARGBSubtractRow_Any_MSA, ARGBSubtractRow_MSA, 0, 4, 4, 4, 7) 311 #endif 312 #ifdef HAS_SOBELROW_SSE2 313 ANY21(SobelRow_Any_SSE2, SobelRow_SSE2, 0, 1, 1, 4, 15) 314 #endif 315 #ifdef HAS_SOBELROW_NEON 316 ANY21(SobelRow_Any_NEON, SobelRow_NEON, 0, 1, 1, 4, 7) 317 #endif 318 #ifdef HAS_SOBELROW_MSA 319 ANY21(SobelRow_Any_MSA, SobelRow_MSA, 0, 1, 1, 4, 15) 320 #endif 321 #ifdef HAS_SOBELTOPLANEROW_SSE2 322 ANY21(SobelToPlaneRow_Any_SSE2, SobelToPlaneRow_SSE2, 0, 1, 1, 1, 15) 323 #endif 324 #ifdef HAS_SOBELTOPLANEROW_NEON 325 ANY21(SobelToPlaneRow_Any_NEON, SobelToPlaneRow_NEON, 0, 1, 1, 1, 15) 326 #endif 327 #ifdef HAS_SOBELTOPLANEROW_MSA 328 ANY21(SobelToPlaneRow_Any_MSA, SobelToPlaneRow_MSA, 0, 1, 1, 1, 31) 329 #endif 330 #ifdef HAS_SOBELXYROW_SSE2 331 ANY21(SobelXYRow_Any_SSE2, SobelXYRow_SSE2, 0, 1, 1, 4, 15) 332 #endif 333 #ifdef HAS_SOBELXYROW_NEON 334 ANY21(SobelXYRow_Any_NEON, SobelXYRow_NEON, 0, 1, 1, 4, 7) 335 #endif 336 #ifdef HAS_SOBELXYROW_MSA 337 ANY21(SobelXYRow_Any_MSA, SobelXYRow_MSA, 0, 1, 1, 4, 15) 338 #endif 339 #undef ANY21 340 341 // Any 2 planes to 1 with yuvconstants 342 #define ANY21C(NAMEANY, ANY_SIMD, UVSHIFT, SBPP, SBPP2, BPP, MASK) \ 343 void NAMEANY(const uint8_t* y_buf, const uint8_t* uv_buf, uint8_t* dst_ptr, \ 344 const struct YuvConstants* yuvconstants, int width) { \ 345 SIMD_ALIGNED(uint8_t temp[128 * 3]); \ 346 memset(temp, 0, 128 * 2); /* for msan */ \ 347 int r = width & MASK; \ 348 int n = width & ~MASK; \ 349 if (n > 0) { \ 350 ANY_SIMD(y_buf, uv_buf, dst_ptr, yuvconstants, n); \ 351 } \ 352 memcpy(temp, y_buf + n * SBPP, r * SBPP); \ 353 memcpy(temp + 128, uv_buf + (n >> UVSHIFT) * SBPP2, \ 354 SS(r, UVSHIFT) * SBPP2); \ 355 ANY_SIMD(temp, temp + 128, temp + 256, yuvconstants, MASK + 1); \ 356 memcpy(dst_ptr + n * BPP, temp + 256, r * BPP); \ 357 } 358 359 // Biplanar to RGB. 360 #ifdef HAS_NV12TOARGBROW_SSSE3 361 ANY21C(NV12ToARGBRow_Any_SSSE3, NV12ToARGBRow_SSSE3, 1, 1, 2, 4, 7) 362 #endif 363 #ifdef HAS_NV12TOARGBROW_AVX2 364 ANY21C(NV12ToARGBRow_Any_AVX2, NV12ToARGBRow_AVX2, 1, 1, 2, 4, 15) 365 #endif 366 #ifdef HAS_NV12TOARGBROW_NEON 367 ANY21C(NV12ToARGBRow_Any_NEON, NV12ToARGBRow_NEON, 1, 1, 2, 4, 7) 368 #endif 369 #ifdef HAS_NV12TOARGBROW_MSA 370 ANY21C(NV12ToARGBRow_Any_MSA, NV12ToARGBRow_MSA, 1, 1, 2, 4, 7) 371 #endif 372 #ifdef HAS_NV21TOARGBROW_SSSE3 373 ANY21C(NV21ToARGBRow_Any_SSSE3, NV21ToARGBRow_SSSE3, 1, 1, 2, 4, 7) 374 #endif 375 #ifdef HAS_NV21TOARGBROW_AVX2 376 ANY21C(NV21ToARGBRow_Any_AVX2, NV21ToARGBRow_AVX2, 1, 1, 2, 4, 15) 377 #endif 378 #ifdef HAS_NV21TOARGBROW_NEON 379 ANY21C(NV21ToARGBRow_Any_NEON, NV21ToARGBRow_NEON, 1, 1, 2, 4, 7) 380 #endif 381 #ifdef HAS_NV21TOARGBROW_MSA 382 ANY21C(NV21ToARGBRow_Any_MSA, NV21ToARGBRow_MSA, 1, 1, 2, 4, 7) 383 #endif 384 #ifdef HAS_NV12TORGB24ROW_NEON 385 ANY21C(NV12ToRGB24Row_Any_NEON, NV12ToRGB24Row_NEON, 1, 1, 2, 3, 7) 386 #endif 387 #ifdef HAS_NV21TORGB24ROW_NEON 388 ANY21C(NV21ToRGB24Row_Any_NEON, NV21ToRGB24Row_NEON, 1, 1, 2, 3, 7) 389 #endif 390 #ifdef HAS_NV12TORGB24ROW_SSSE3 391 ANY21C(NV12ToRGB24Row_Any_SSSE3, NV12ToRGB24Row_SSSE3, 1, 1, 2, 3, 15) 392 #endif 393 #ifdef HAS_NV21TORGB24ROW_SSSE3 394 ANY21C(NV21ToRGB24Row_Any_SSSE3, NV21ToRGB24Row_SSSE3, 1, 1, 2, 3, 15) 395 #endif 396 #ifdef HAS_NV12TORGB24ROW_AVX2 397 ANY21C(NV12ToRGB24Row_Any_AVX2, NV12ToRGB24Row_AVX2, 1, 1, 2, 3, 31) 398 #endif 399 #ifdef HAS_NV21TORGB24ROW_AVX2 400 ANY21C(NV21ToRGB24Row_Any_AVX2, NV21ToRGB24Row_AVX2, 1, 1, 2, 3, 31) 401 #endif 402 #ifdef HAS_NV12TORGB565ROW_SSSE3 403 ANY21C(NV12ToRGB565Row_Any_SSSE3, NV12ToRGB565Row_SSSE3, 1, 1, 2, 2, 7) 404 #endif 405 #ifdef HAS_NV12TORGB565ROW_AVX2 406 ANY21C(NV12ToRGB565Row_Any_AVX2, NV12ToRGB565Row_AVX2, 1, 1, 2, 2, 15) 407 #endif 408 #ifdef HAS_NV12TORGB565ROW_NEON 409 ANY21C(NV12ToRGB565Row_Any_NEON, NV12ToRGB565Row_NEON, 1, 1, 2, 2, 7) 410 #endif 411 #ifdef HAS_NV12TORGB565ROW_MSA 412 ANY21C(NV12ToRGB565Row_Any_MSA, NV12ToRGB565Row_MSA, 1, 1, 2, 2, 7) 413 #endif 414 #undef ANY21C 415 416 // Any 1 to 1. 417 #define ANY11(NAMEANY, ANY_SIMD, UVSHIFT, SBPP, BPP, MASK) \ 418 void NAMEANY(const uint8_t* src_ptr, uint8_t* dst_ptr, int width) { \ 419 SIMD_ALIGNED(uint8_t temp[128 * 2]); \ 420 memset(temp, 0, 128); /* for YUY2 and msan */ \ 421 int r = width & MASK; \ 422 int n = width & ~MASK; \ 423 if (n > 0) { \ 424 ANY_SIMD(src_ptr, dst_ptr, n); \ 425 } \ 426 memcpy(temp, src_ptr + (n >> UVSHIFT) * SBPP, SS(r, UVSHIFT) * SBPP); \ 427 ANY_SIMD(temp, temp + 128, MASK + 1); \ 428 memcpy(dst_ptr + n * BPP, temp + 128, r * BPP); \ 429 } 430 431 #ifdef HAS_COPYROW_AVX 432 ANY11(CopyRow_Any_AVX, CopyRow_AVX, 0, 1, 1, 63) 433 #endif 434 #ifdef HAS_COPYROW_SSE2 435 ANY11(CopyRow_Any_SSE2, CopyRow_SSE2, 0, 1, 1, 31) 436 #endif 437 #ifdef HAS_COPYROW_NEON 438 ANY11(CopyRow_Any_NEON, CopyRow_NEON, 0, 1, 1, 31) 439 #endif 440 #if defined(HAS_ARGBTORGB24ROW_SSSE3) 441 ANY11(ARGBToRGB24Row_Any_SSSE3, ARGBToRGB24Row_SSSE3, 0, 4, 3, 15) 442 ANY11(ARGBToRAWRow_Any_SSSE3, ARGBToRAWRow_SSSE3, 0, 4, 3, 15) 443 ANY11(ARGBToRGB565Row_Any_SSE2, ARGBToRGB565Row_SSE2, 0, 4, 2, 3) 444 ANY11(ARGBToARGB1555Row_Any_SSE2, ARGBToARGB1555Row_SSE2, 0, 4, 2, 3) 445 ANY11(ARGBToARGB4444Row_Any_SSE2, ARGBToARGB4444Row_SSE2, 0, 4, 2, 3) 446 #endif 447 #if defined(HAS_ARGBTORGB24ROW_AVX2) 448 ANY11(ARGBToRGB24Row_Any_AVX2, ARGBToRGB24Row_AVX2, 0, 4, 3, 31) 449 #endif 450 #if defined(HAS_ARGBTORGB24ROW_AVX512VBMI) 451 ANY11(ARGBToRGB24Row_Any_AVX512VBMI, ARGBToRGB24Row_AVX512VBMI, 0, 4, 3, 31) 452 #endif 453 #if defined(HAS_ARGBTORAWROW_AVX2) 454 ANY11(ARGBToRAWRow_Any_AVX2, ARGBToRAWRow_AVX2, 0, 4, 3, 31) 455 #endif 456 #if defined(HAS_ARGBTORGB565ROW_AVX2) 457 ANY11(ARGBToRGB565Row_Any_AVX2, ARGBToRGB565Row_AVX2, 0, 4, 2, 7) 458 #endif 459 #if defined(HAS_ARGBTOARGB4444ROW_AVX2) 460 ANY11(ARGBToARGB1555Row_Any_AVX2, ARGBToARGB1555Row_AVX2, 0, 4, 2, 7) 461 ANY11(ARGBToARGB4444Row_Any_AVX2, ARGBToARGB4444Row_AVX2, 0, 4, 2, 7) 462 #endif 463 #if defined(HAS_ABGRTOAR30ROW_SSSE3) 464 ANY11(ABGRToAR30Row_Any_SSSE3, ABGRToAR30Row_SSSE3, 0, 4, 4, 3) 465 #endif 466 #if defined(HAS_ARGBTOAR30ROW_SSSE3) 467 ANY11(ARGBToAR30Row_Any_SSSE3, ARGBToAR30Row_SSSE3, 0, 4, 4, 3) 468 #endif 469 #if defined(HAS_ABGRTOAR30ROW_AVX2) 470 ANY11(ABGRToAR30Row_Any_AVX2, ABGRToAR30Row_AVX2, 0, 4, 4, 7) 471 #endif 472 #if defined(HAS_ARGBTOAR30ROW_AVX2) 473 ANY11(ARGBToAR30Row_Any_AVX2, ARGBToAR30Row_AVX2, 0, 4, 4, 7) 474 #endif 475 #if defined(HAS_J400TOARGBROW_SSE2) 476 ANY11(J400ToARGBRow_Any_SSE2, J400ToARGBRow_SSE2, 0, 1, 4, 7) 477 #endif 478 #if defined(HAS_J400TOARGBROW_AVX2) 479 ANY11(J400ToARGBRow_Any_AVX2, J400ToARGBRow_AVX2, 0, 1, 4, 15) 480 #endif 481 #if defined(HAS_I400TOARGBROW_SSE2) 482 ANY11(I400ToARGBRow_Any_SSE2, I400ToARGBRow_SSE2, 0, 1, 4, 7) 483 #endif 484 #if defined(HAS_I400TOARGBROW_AVX2) 485 ANY11(I400ToARGBRow_Any_AVX2, I400ToARGBRow_AVX2, 0, 1, 4, 15) 486 #endif 487 #if defined(HAS_RGB24TOARGBROW_SSSE3) 488 ANY11(RGB24ToARGBRow_Any_SSSE3, RGB24ToARGBRow_SSSE3, 0, 3, 4, 15) 489 ANY11(RAWToARGBRow_Any_SSSE3, RAWToARGBRow_SSSE3, 0, 3, 4, 15) 490 ANY11(RGB565ToARGBRow_Any_SSE2, RGB565ToARGBRow_SSE2, 0, 2, 4, 7) 491 ANY11(ARGB1555ToARGBRow_Any_SSE2, ARGB1555ToARGBRow_SSE2, 0, 2, 4, 7) 492 ANY11(ARGB4444ToARGBRow_Any_SSE2, ARGB4444ToARGBRow_SSE2, 0, 2, 4, 7) 493 #endif 494 #if defined(HAS_RAWTORGB24ROW_SSSE3) 495 ANY11(RAWToRGB24Row_Any_SSSE3, RAWToRGB24Row_SSSE3, 0, 3, 3, 7) 496 #endif 497 #if defined(HAS_RGB565TOARGBROW_AVX2) 498 ANY11(RGB565ToARGBRow_Any_AVX2, RGB565ToARGBRow_AVX2, 0, 2, 4, 15) 499 #endif 500 #if defined(HAS_ARGB1555TOARGBROW_AVX2) 501 ANY11(ARGB1555ToARGBRow_Any_AVX2, ARGB1555ToARGBRow_AVX2, 0, 2, 4, 15) 502 #endif 503 #if defined(HAS_ARGB4444TOARGBROW_AVX2) 504 ANY11(ARGB4444ToARGBRow_Any_AVX2, ARGB4444ToARGBRow_AVX2, 0, 2, 4, 15) 505 #endif 506 #if defined(HAS_ARGBTORGB24ROW_NEON) 507 ANY11(ARGBToRGB24Row_Any_NEON, ARGBToRGB24Row_NEON, 0, 4, 3, 7) 508 ANY11(ARGBToRAWRow_Any_NEON, ARGBToRAWRow_NEON, 0, 4, 3, 7) 509 ANY11(ARGBToRGB565Row_Any_NEON, ARGBToRGB565Row_NEON, 0, 4, 2, 7) 510 ANY11(ARGBToARGB1555Row_Any_NEON, ARGBToARGB1555Row_NEON, 0, 4, 2, 7) 511 ANY11(ARGBToARGB4444Row_Any_NEON, ARGBToARGB4444Row_NEON, 0, 4, 2, 7) 512 ANY11(J400ToARGBRow_Any_NEON, J400ToARGBRow_NEON, 0, 1, 4, 7) 513 ANY11(I400ToARGBRow_Any_NEON, I400ToARGBRow_NEON, 0, 1, 4, 7) 514 #endif 515 #if defined(HAS_ARGBTORGB24ROW_MSA) 516 ANY11(ARGBToRGB24Row_Any_MSA, ARGBToRGB24Row_MSA, 0, 4, 3, 15) 517 ANY11(ARGBToRAWRow_Any_MSA, ARGBToRAWRow_MSA, 0, 4, 3, 15) 518 ANY11(ARGBToRGB565Row_Any_MSA, ARGBToRGB565Row_MSA, 0, 4, 2, 7) 519 ANY11(ARGBToARGB1555Row_Any_MSA, ARGBToARGB1555Row_MSA, 0, 4, 2, 7) 520 ANY11(ARGBToARGB4444Row_Any_MSA, ARGBToARGB4444Row_MSA, 0, 4, 2, 7) 521 ANY11(J400ToARGBRow_Any_MSA, J400ToARGBRow_MSA, 0, 1, 4, 15) 522 ANY11(I400ToARGBRow_Any_MSA, I400ToARGBRow_MSA, 0, 1, 4, 15) 523 #endif 524 #if defined(HAS_RAWTORGB24ROW_NEON) 525 ANY11(RAWToRGB24Row_Any_NEON, RAWToRGB24Row_NEON, 0, 3, 3, 7) 526 #endif 527 #if defined(HAS_RAWTORGB24ROW_MSA) 528 ANY11(RAWToRGB24Row_Any_MSA, RAWToRGB24Row_MSA, 0, 3, 3, 15) 529 #endif 530 #ifdef HAS_ARGBTOYROW_AVX2 531 ANY11(ARGBToYRow_Any_AVX2, ARGBToYRow_AVX2, 0, 4, 1, 31) 532 #endif 533 #ifdef HAS_ARGBTOYJROW_AVX2 534 ANY11(ARGBToYJRow_Any_AVX2, ARGBToYJRow_AVX2, 0, 4, 1, 31) 535 #endif 536 #ifdef HAS_UYVYTOYROW_AVX2 537 ANY11(UYVYToYRow_Any_AVX2, UYVYToYRow_AVX2, 0, 2, 1, 31) 538 #endif 539 #ifdef HAS_YUY2TOYROW_AVX2 540 ANY11(YUY2ToYRow_Any_AVX2, YUY2ToYRow_AVX2, 1, 4, 1, 31) 541 #endif 542 #ifdef HAS_ARGBTOYROW_SSSE3 543 ANY11(ARGBToYRow_Any_SSSE3, ARGBToYRow_SSSE3, 0, 4, 1, 15) 544 #endif 545 #ifdef HAS_BGRATOYROW_SSSE3 546 ANY11(BGRAToYRow_Any_SSSE3, BGRAToYRow_SSSE3, 0, 4, 1, 15) 547 ANY11(ABGRToYRow_Any_SSSE3, ABGRToYRow_SSSE3, 0, 4, 1, 15) 548 ANY11(RGBAToYRow_Any_SSSE3, RGBAToYRow_SSSE3, 0, 4, 1, 15) 549 ANY11(YUY2ToYRow_Any_SSE2, YUY2ToYRow_SSE2, 1, 4, 1, 15) 550 ANY11(UYVYToYRow_Any_SSE2, UYVYToYRow_SSE2, 1, 4, 1, 15) 551 #endif 552 #ifdef HAS_ARGBTOYJROW_SSSE3 553 ANY11(ARGBToYJRow_Any_SSSE3, ARGBToYJRow_SSSE3, 0, 4, 1, 15) 554 #endif 555 #ifdef HAS_ARGBTOYROW_NEON 556 ANY11(ARGBToYRow_Any_NEON, ARGBToYRow_NEON, 0, 4, 1, 7) 557 #endif 558 #ifdef HAS_ARGBTOYROW_MSA 559 ANY11(ARGBToYRow_Any_MSA, ARGBToYRow_MSA, 0, 4, 1, 15) 560 #endif 561 #ifdef HAS_ARGBTOYJROW_NEON 562 ANY11(ARGBToYJRow_Any_NEON, ARGBToYJRow_NEON, 0, 4, 1, 7) 563 #endif 564 #ifdef HAS_ARGBTOYJROW_MSA 565 ANY11(ARGBToYJRow_Any_MSA, ARGBToYJRow_MSA, 0, 4, 1, 15) 566 #endif 567 #ifdef HAS_BGRATOYROW_NEON 568 ANY11(BGRAToYRow_Any_NEON, BGRAToYRow_NEON, 0, 4, 1, 7) 569 #endif 570 #ifdef HAS_BGRATOYROW_MSA 571 ANY11(BGRAToYRow_Any_MSA, BGRAToYRow_MSA, 0, 4, 1, 15) 572 #endif 573 #ifdef HAS_ABGRTOYROW_NEON 574 ANY11(ABGRToYRow_Any_NEON, ABGRToYRow_NEON, 0, 4, 1, 7) 575 #endif 576 #ifdef HAS_ABGRTOYROW_MSA 577 ANY11(ABGRToYRow_Any_MSA, ABGRToYRow_MSA, 0, 4, 1, 7) 578 #endif 579 #ifdef HAS_RGBATOYROW_NEON 580 ANY11(RGBAToYRow_Any_NEON, RGBAToYRow_NEON, 0, 4, 1, 7) 581 #endif 582 #ifdef HAS_RGBATOYROW_MSA 583 ANY11(RGBAToYRow_Any_MSA, RGBAToYRow_MSA, 0, 4, 1, 15) 584 #endif 585 #ifdef HAS_RGB24TOYROW_NEON 586 ANY11(RGB24ToYRow_Any_NEON, RGB24ToYRow_NEON, 0, 3, 1, 7) 587 #endif 588 #ifdef HAS_RGB24TOYROW_MSA 589 ANY11(RGB24ToYRow_Any_MSA, RGB24ToYRow_MSA, 0, 3, 1, 15) 590 #endif 591 #ifdef HAS_RAWTOYROW_NEON 592 ANY11(RAWToYRow_Any_NEON, RAWToYRow_NEON, 0, 3, 1, 7) 593 #endif 594 #ifdef HAS_RAWTOYROW_MSA 595 ANY11(RAWToYRow_Any_MSA, RAWToYRow_MSA, 0, 3, 1, 15) 596 #endif 597 #ifdef HAS_RGB565TOYROW_NEON 598 ANY11(RGB565ToYRow_Any_NEON, RGB565ToYRow_NEON, 0, 2, 1, 7) 599 #endif 600 #ifdef HAS_RGB565TOYROW_MSA 601 ANY11(RGB565ToYRow_Any_MSA, RGB565ToYRow_MSA, 0, 2, 1, 15) 602 #endif 603 #ifdef HAS_ARGB1555TOYROW_NEON 604 ANY11(ARGB1555ToYRow_Any_NEON, ARGB1555ToYRow_NEON, 0, 2, 1, 7) 605 #endif 606 #ifdef HAS_ARGB1555TOYROW_MSA 607 ANY11(ARGB1555ToYRow_Any_MSA, ARGB1555ToYRow_MSA, 0, 2, 1, 15) 608 #endif 609 #ifdef HAS_ARGB4444TOYROW_NEON 610 ANY11(ARGB4444ToYRow_Any_NEON, ARGB4444ToYRow_NEON, 0, 2, 1, 7) 611 #endif 612 #ifdef HAS_YUY2TOYROW_NEON 613 ANY11(YUY2ToYRow_Any_NEON, YUY2ToYRow_NEON, 1, 4, 1, 15) 614 #endif 615 #ifdef HAS_UYVYTOYROW_NEON 616 ANY11(UYVYToYRow_Any_NEON, UYVYToYRow_NEON, 1, 4, 1, 15) 617 #endif 618 #ifdef HAS_YUY2TOYROW_MSA 619 ANY11(YUY2ToYRow_Any_MSA, YUY2ToYRow_MSA, 1, 4, 1, 31) 620 #endif 621 #ifdef HAS_UYVYTOYROW_MSA 622 ANY11(UYVYToYRow_Any_MSA, UYVYToYRow_MSA, 1, 4, 1, 31) 623 #endif 624 #ifdef HAS_RGB24TOARGBROW_NEON 625 ANY11(RGB24ToARGBRow_Any_NEON, RGB24ToARGBRow_NEON, 0, 3, 4, 7) 626 #endif 627 #ifdef HAS_RGB24TOARGBROW_MSA 628 ANY11(RGB24ToARGBRow_Any_MSA, RGB24ToARGBRow_MSA, 0, 3, 4, 15) 629 #endif 630 #ifdef HAS_RAWTOARGBROW_NEON 631 ANY11(RAWToARGBRow_Any_NEON, RAWToARGBRow_NEON, 0, 3, 4, 7) 632 #endif 633 #ifdef HAS_RAWTOARGBROW_MSA 634 ANY11(RAWToARGBRow_Any_MSA, RAWToARGBRow_MSA, 0, 3, 4, 15) 635 #endif 636 #ifdef HAS_RGB565TOARGBROW_NEON 637 ANY11(RGB565ToARGBRow_Any_NEON, RGB565ToARGBRow_NEON, 0, 2, 4, 7) 638 #endif 639 #ifdef HAS_RGB565TOARGBROW_MSA 640 ANY11(RGB565ToARGBRow_Any_MSA, RGB565ToARGBRow_MSA, 0, 2, 4, 15) 641 #endif 642 #ifdef HAS_ARGB1555TOARGBROW_NEON 643 ANY11(ARGB1555ToARGBRow_Any_NEON, ARGB1555ToARGBRow_NEON, 0, 2, 4, 7) 644 #endif 645 #ifdef HAS_ARGB1555TOARGBROW_MSA 646 ANY11(ARGB1555ToARGBRow_Any_MSA, ARGB1555ToARGBRow_MSA, 0, 2, 4, 15) 647 #endif 648 #ifdef HAS_ARGB4444TOARGBROW_NEON 649 ANY11(ARGB4444ToARGBRow_Any_NEON, ARGB4444ToARGBRow_NEON, 0, 2, 4, 7) 650 #endif 651 #ifdef HAS_ARGB4444TOARGBROW_MSA 652 ANY11(ARGB4444ToARGBRow_Any_MSA, ARGB4444ToARGBRow_MSA, 0, 2, 4, 15) 653 #endif 654 #ifdef HAS_ARGBATTENUATEROW_SSSE3 655 ANY11(ARGBAttenuateRow_Any_SSSE3, ARGBAttenuateRow_SSSE3, 0, 4, 4, 3) 656 #endif 657 #ifdef HAS_ARGBUNATTENUATEROW_SSE2 658 ANY11(ARGBUnattenuateRow_Any_SSE2, ARGBUnattenuateRow_SSE2, 0, 4, 4, 3) 659 #endif 660 #ifdef HAS_ARGBATTENUATEROW_AVX2 661 ANY11(ARGBAttenuateRow_Any_AVX2, ARGBAttenuateRow_AVX2, 0, 4, 4, 7) 662 #endif 663 #ifdef HAS_ARGBUNATTENUATEROW_AVX2 664 ANY11(ARGBUnattenuateRow_Any_AVX2, ARGBUnattenuateRow_AVX2, 0, 4, 4, 7) 665 #endif 666 #ifdef HAS_ARGBATTENUATEROW_NEON 667 ANY11(ARGBAttenuateRow_Any_NEON, ARGBAttenuateRow_NEON, 0, 4, 4, 7) 668 #endif 669 #ifdef HAS_ARGBATTENUATEROW_MSA 670 ANY11(ARGBAttenuateRow_Any_MSA, ARGBAttenuateRow_MSA, 0, 4, 4, 7) 671 #endif 672 #ifdef HAS_ARGBEXTRACTALPHAROW_SSE2 673 ANY11(ARGBExtractAlphaRow_Any_SSE2, ARGBExtractAlphaRow_SSE2, 0, 4, 1, 7) 674 #endif 675 #ifdef HAS_ARGBEXTRACTALPHAROW_AVX2 676 ANY11(ARGBExtractAlphaRow_Any_AVX2, ARGBExtractAlphaRow_AVX2, 0, 4, 1, 31) 677 #endif 678 #ifdef HAS_ARGBEXTRACTALPHAROW_NEON 679 ANY11(ARGBExtractAlphaRow_Any_NEON, ARGBExtractAlphaRow_NEON, 0, 4, 1, 15) 680 #endif 681 #ifdef HAS_ARGBEXTRACTALPHAROW_MSA 682 ANY11(ARGBExtractAlphaRow_Any_MSA, ARGBExtractAlphaRow_MSA, 0, 4, 1, 15) 683 #endif 684 #undef ANY11 685 686 // Any 1 to 1 blended. Destination is read, modify, write. 687 #define ANY11B(NAMEANY, ANY_SIMD, UVSHIFT, SBPP, BPP, MASK) \ 688 void NAMEANY(const uint8_t* src_ptr, uint8_t* dst_ptr, int width) { \ 689 SIMD_ALIGNED(uint8_t temp[64 * 2]); \ 690 memset(temp, 0, 64 * 2); /* for msan */ \ 691 int r = width & MASK; \ 692 int n = width & ~MASK; \ 693 if (n > 0) { \ 694 ANY_SIMD(src_ptr, dst_ptr, n); \ 695 } \ 696 memcpy(temp, src_ptr + (n >> UVSHIFT) * SBPP, SS(r, UVSHIFT) * SBPP); \ 697 memcpy(temp + 64, dst_ptr + n * BPP, r * BPP); \ 698 ANY_SIMD(temp, temp + 64, MASK + 1); \ 699 memcpy(dst_ptr + n * BPP, temp + 64, r * BPP); \ 700 } 701 702 #ifdef HAS_ARGBCOPYALPHAROW_AVX2 703 ANY11B(ARGBCopyAlphaRow_Any_AVX2, ARGBCopyAlphaRow_AVX2, 0, 4, 4, 15) 704 #endif 705 #ifdef HAS_ARGBCOPYALPHAROW_SSE2 706 ANY11B(ARGBCopyAlphaRow_Any_SSE2, ARGBCopyAlphaRow_SSE2, 0, 4, 4, 7) 707 #endif 708 #ifdef HAS_ARGBCOPYYTOALPHAROW_AVX2 709 ANY11B(ARGBCopyYToAlphaRow_Any_AVX2, ARGBCopyYToAlphaRow_AVX2, 0, 1, 4, 15) 710 #endif 711 #ifdef HAS_ARGBCOPYYTOALPHAROW_SSE2 712 ANY11B(ARGBCopyYToAlphaRow_Any_SSE2, ARGBCopyYToAlphaRow_SSE2, 0, 1, 4, 7) 713 #endif 714 #undef ANY11B 715 716 // Any 1 to 1 with parameter. 717 #define ANY11P(NAMEANY, ANY_SIMD, T, SBPP, BPP, MASK) \ 718 void NAMEANY(const uint8_t* src_ptr, uint8_t* dst_ptr, T param, int width) { \ 719 SIMD_ALIGNED(uint8_t temp[64 * 2]); \ 720 memset(temp, 0, 64); /* for msan */ \ 721 int r = width & MASK; \ 722 int n = width & ~MASK; \ 723 if (n > 0) { \ 724 ANY_SIMD(src_ptr, dst_ptr, param, n); \ 725 } \ 726 memcpy(temp, src_ptr + n * SBPP, r * SBPP); \ 727 ANY_SIMD(temp, temp + 64, param, MASK + 1); \ 728 memcpy(dst_ptr + n * BPP, temp + 64, r * BPP); \ 729 } 730 731 #if defined(HAS_ARGBTORGB565DITHERROW_SSE2) 732 ANY11P(ARGBToRGB565DitherRow_Any_SSE2, 733 ARGBToRGB565DitherRow_SSE2, 734 const uint32_t, 735 4, 736 2, 737 3) 738 #endif 739 #if defined(HAS_ARGBTORGB565DITHERROW_AVX2) 740 ANY11P(ARGBToRGB565DitherRow_Any_AVX2, 741 ARGBToRGB565DitherRow_AVX2, 742 const uint32_t, 743 4, 744 2, 745 7) 746 #endif 747 #if defined(HAS_ARGBTORGB565DITHERROW_NEON) 748 ANY11P(ARGBToRGB565DitherRow_Any_NEON, 749 ARGBToRGB565DitherRow_NEON, 750 const uint32_t, 751 4, 752 2, 753 7) 754 #endif 755 #if defined(HAS_ARGBTORGB565DITHERROW_MSA) 756 ANY11P(ARGBToRGB565DitherRow_Any_MSA, 757 ARGBToRGB565DitherRow_MSA, 758 const uint32_t, 759 4, 760 2, 761 7) 762 #endif 763 #ifdef HAS_ARGBSHUFFLEROW_SSSE3 764 ANY11P(ARGBShuffleRow_Any_SSSE3, ARGBShuffleRow_SSSE3, const uint8_t*, 4, 4, 7) 765 #endif 766 #ifdef HAS_ARGBSHUFFLEROW_AVX2 767 ANY11P(ARGBShuffleRow_Any_AVX2, ARGBShuffleRow_AVX2, const uint8_t*, 4, 4, 15) 768 #endif 769 #ifdef HAS_ARGBSHUFFLEROW_NEON 770 ANY11P(ARGBShuffleRow_Any_NEON, ARGBShuffleRow_NEON, const uint8_t*, 4, 4, 3) 771 #endif 772 #ifdef HAS_ARGBSHUFFLEROW_MSA 773 ANY11P(ARGBShuffleRow_Any_MSA, ARGBShuffleRow_MSA, const uint8_t*, 4, 4, 7) 774 #endif 775 #undef ANY11P 776 777 // Any 1 to 1 with parameter and shorts. BPP measures in shorts. 778 #define ANY11C(NAMEANY, ANY_SIMD, SBPP, BPP, STYPE, DTYPE, MASK) \ 779 void NAMEANY(const STYPE* src_ptr, DTYPE* dst_ptr, int scale, int width) { \ 780 SIMD_ALIGNED(STYPE temp[32]); \ 781 SIMD_ALIGNED(DTYPE out[32]); \ 782 memset(temp, 0, 32 * SBPP); /* for msan */ \ 783 int r = width & MASK; \ 784 int n = width & ~MASK; \ 785 if (n > 0) { \ 786 ANY_SIMD(src_ptr, dst_ptr, scale, n); \ 787 } \ 788 memcpy(temp, src_ptr + n, r * SBPP); \ 789 ANY_SIMD(temp, out, scale, MASK + 1); \ 790 memcpy(dst_ptr + n, out, r * BPP); \ 791 } 792 793 #ifdef HAS_CONVERT16TO8ROW_SSSE3 794 ANY11C(Convert16To8Row_Any_SSSE3, 795 Convert16To8Row_SSSE3, 796 2, 797 1, 798 uint16_t, 799 uint8_t, 800 15) 801 #endif 802 #ifdef HAS_CONVERT16TO8ROW_AVX2 803 ANY11C(Convert16To8Row_Any_AVX2, 804 Convert16To8Row_AVX2, 805 2, 806 1, 807 uint16_t, 808 uint8_t, 809 31) 810 #endif 811 #ifdef HAS_CONVERT8TO16ROW_SSE2 812 ANY11C(Convert8To16Row_Any_SSE2, 813 Convert8To16Row_SSE2, 814 1, 815 2, 816 uint8_t, 817 uint16_t, 818 15) 819 #endif 820 #ifdef HAS_CONVERT8TO16ROW_AVX2 821 ANY11C(Convert8To16Row_Any_AVX2, 822 Convert8To16Row_AVX2, 823 1, 824 2, 825 uint8_t, 826 uint16_t, 827 31) 828 #endif 829 #undef ANY11C 830 831 // Any 1 to 1 with parameter and shorts to byte. BPP measures in shorts. 832 #define ANY11P16(NAMEANY, ANY_SIMD, ST, T, SBPP, BPP, MASK) \ 833 void NAMEANY(const ST* src_ptr, T* dst_ptr, float param, int width) { \ 834 SIMD_ALIGNED(ST temp[32]); \ 835 SIMD_ALIGNED(T out[32]); \ 836 memset(temp, 0, SBPP * 32); /* for msan */ \ 837 int r = width & MASK; \ 838 int n = width & ~MASK; \ 839 if (n > 0) { \ 840 ANY_SIMD(src_ptr, dst_ptr, param, n); \ 841 } \ 842 memcpy(temp, src_ptr + n, r * SBPP); \ 843 ANY_SIMD(temp, out, param, MASK + 1); \ 844 memcpy(dst_ptr + n, out, r * BPP); \ 845 } 846 847 #ifdef HAS_HALFFLOATROW_SSE2 848 ANY11P16(HalfFloatRow_Any_SSE2, HalfFloatRow_SSE2, uint16_t, uint16_t, 2, 2, 7) 849 #endif 850 #ifdef HAS_HALFFLOATROW_AVX2 851 ANY11P16(HalfFloatRow_Any_AVX2, HalfFloatRow_AVX2, uint16_t, uint16_t, 2, 2, 15) 852 #endif 853 #ifdef HAS_HALFFLOATROW_F16C 854 ANY11P16(HalfFloatRow_Any_F16C, HalfFloatRow_F16C, uint16_t, uint16_t, 2, 2, 15) 855 ANY11P16(HalfFloat1Row_Any_F16C, 856 HalfFloat1Row_F16C, 857 uint16_t, 858 uint16_t, 859 2, 860 2, 861 15) 862 #endif 863 #ifdef HAS_HALFFLOATROW_NEON 864 ANY11P16(HalfFloatRow_Any_NEON, HalfFloatRow_NEON, uint16_t, uint16_t, 2, 2, 7) 865 ANY11P16(HalfFloat1Row_Any_NEON, 866 HalfFloat1Row_NEON, 867 uint16_t, 868 uint16_t, 869 2, 870 2, 871 7) 872 #endif 873 #ifdef HAS_HALFFLOATROW_MSA 874 ANY11P16(HalfFloatRow_Any_MSA, HalfFloatRow_MSA, uint16_t, uint16_t, 2, 2, 31) 875 #endif 876 #ifdef HAS_BYTETOFLOATROW_NEON 877 ANY11P16(ByteToFloatRow_Any_NEON, ByteToFloatRow_NEON, uint8_t, float, 1, 3, 7) 878 #endif 879 #undef ANY11P16 880 881 // Any 1 to 1 with yuvconstants 882 #define ANY11C(NAMEANY, ANY_SIMD, UVSHIFT, SBPP, BPP, MASK) \ 883 void NAMEANY(const uint8_t* src_ptr, uint8_t* dst_ptr, \ 884 const struct YuvConstants* yuvconstants, int width) { \ 885 SIMD_ALIGNED(uint8_t temp[128 * 2]); \ 886 memset(temp, 0, 128); /* for YUY2 and msan */ \ 887 int r = width & MASK; \ 888 int n = width & ~MASK; \ 889 if (n > 0) { \ 890 ANY_SIMD(src_ptr, dst_ptr, yuvconstants, n); \ 891 } \ 892 memcpy(temp, src_ptr + (n >> UVSHIFT) * SBPP, SS(r, UVSHIFT) * SBPP); \ 893 ANY_SIMD(temp, temp + 128, yuvconstants, MASK + 1); \ 894 memcpy(dst_ptr + n * BPP, temp + 128, r * BPP); \ 895 } 896 #if defined(HAS_YUY2TOARGBROW_SSSE3) 897 ANY11C(YUY2ToARGBRow_Any_SSSE3, YUY2ToARGBRow_SSSE3, 1, 4, 4, 15) 898 ANY11C(UYVYToARGBRow_Any_SSSE3, UYVYToARGBRow_SSSE3, 1, 4, 4, 15) 899 #endif 900 #if defined(HAS_YUY2TOARGBROW_AVX2) 901 ANY11C(YUY2ToARGBRow_Any_AVX2, YUY2ToARGBRow_AVX2, 1, 4, 4, 31) 902 ANY11C(UYVYToARGBRow_Any_AVX2, UYVYToARGBRow_AVX2, 1, 4, 4, 31) 903 #endif 904 #if defined(HAS_YUY2TOARGBROW_NEON) 905 ANY11C(YUY2ToARGBRow_Any_NEON, YUY2ToARGBRow_NEON, 1, 4, 4, 7) 906 ANY11C(UYVYToARGBRow_Any_NEON, UYVYToARGBRow_NEON, 1, 4, 4, 7) 907 #endif 908 #if defined(HAS_YUY2TOARGBROW_MSA) 909 ANY11C(YUY2ToARGBRow_Any_MSA, YUY2ToARGBRow_MSA, 1, 4, 4, 7) 910 ANY11C(UYVYToARGBRow_Any_MSA, UYVYToARGBRow_MSA, 1, 4, 4, 7) 911 #endif 912 #undef ANY11C 913 914 // Any 1 to 1 interpolate. Takes 2 rows of source via stride. 915 #define ANY11T(NAMEANY, ANY_SIMD, SBPP, BPP, MASK) \ 916 void NAMEANY(uint8_t* dst_ptr, const uint8_t* src_ptr, \ 917 ptrdiff_t src_stride_ptr, int width, int source_y_fraction) { \ 918 SIMD_ALIGNED(uint8_t temp[64 * 3]); \ 919 memset(temp, 0, 64 * 2); /* for msan */ \ 920 int r = width & MASK; \ 921 int n = width & ~MASK; \ 922 if (n > 0) { \ 923 ANY_SIMD(dst_ptr, src_ptr, src_stride_ptr, n, source_y_fraction); \ 924 } \ 925 memcpy(temp, src_ptr + n * SBPP, r * SBPP); \ 926 memcpy(temp + 64, src_ptr + src_stride_ptr + n * SBPP, r * SBPP); \ 927 ANY_SIMD(temp + 128, temp, 64, MASK + 1, source_y_fraction); \ 928 memcpy(dst_ptr + n * BPP, temp + 128, r * BPP); \ 929 } 930 931 #ifdef HAS_INTERPOLATEROW_AVX2 932 ANY11T(InterpolateRow_Any_AVX2, InterpolateRow_AVX2, 1, 1, 31) 933 #endif 934 #ifdef HAS_INTERPOLATEROW_SSSE3 935 ANY11T(InterpolateRow_Any_SSSE3, InterpolateRow_SSSE3, 1, 1, 15) 936 #endif 937 #ifdef HAS_INTERPOLATEROW_NEON 938 ANY11T(InterpolateRow_Any_NEON, InterpolateRow_NEON, 1, 1, 15) 939 #endif 940 #ifdef HAS_INTERPOLATEROW_MSA 941 ANY11T(InterpolateRow_Any_MSA, InterpolateRow_MSA, 1, 1, 31) 942 #endif 943 #undef ANY11T 944 945 // Any 1 to 1 mirror. 946 #define ANY11M(NAMEANY, ANY_SIMD, BPP, MASK) \ 947 void NAMEANY(const uint8_t* src_ptr, uint8_t* dst_ptr, int width) { \ 948 SIMD_ALIGNED(uint8_t temp[64 * 2]); \ 949 memset(temp, 0, 64); /* for msan */ \ 950 int r = width & MASK; \ 951 int n = width & ~MASK; \ 952 if (n > 0) { \ 953 ANY_SIMD(src_ptr + r * BPP, dst_ptr, n); \ 954 } \ 955 memcpy(temp, src_ptr, r* BPP); \ 956 ANY_SIMD(temp, temp + 64, MASK + 1); \ 957 memcpy(dst_ptr + n * BPP, temp + 64 + (MASK + 1 - r) * BPP, r * BPP); \ 958 } 959 960 #ifdef HAS_MIRRORROW_AVX2 961 ANY11M(MirrorRow_Any_AVX2, MirrorRow_AVX2, 1, 31) 962 #endif 963 #ifdef HAS_MIRRORROW_SSSE3 964 ANY11M(MirrorRow_Any_SSSE3, MirrorRow_SSSE3, 1, 15) 965 #endif 966 #ifdef HAS_MIRRORROW_NEON 967 ANY11M(MirrorRow_Any_NEON, MirrorRow_NEON, 1, 15) 968 #endif 969 #ifdef HAS_MIRRORROW_MSA 970 ANY11M(MirrorRow_Any_MSA, MirrorRow_MSA, 1, 63) 971 #endif 972 #ifdef HAS_ARGBMIRRORROW_AVX2 973 ANY11M(ARGBMirrorRow_Any_AVX2, ARGBMirrorRow_AVX2, 4, 7) 974 #endif 975 #ifdef HAS_ARGBMIRRORROW_SSE2 976 ANY11M(ARGBMirrorRow_Any_SSE2, ARGBMirrorRow_SSE2, 4, 3) 977 #endif 978 #ifdef HAS_ARGBMIRRORROW_NEON 979 ANY11M(ARGBMirrorRow_Any_NEON, ARGBMirrorRow_NEON, 4, 3) 980 #endif 981 #ifdef HAS_ARGBMIRRORROW_MSA 982 ANY11M(ARGBMirrorRow_Any_MSA, ARGBMirrorRow_MSA, 4, 15) 983 #endif 984 #undef ANY11M 985 986 // Any 1 plane. (memset) 987 #define ANY1(NAMEANY, ANY_SIMD, T, BPP, MASK) \ 988 void NAMEANY(uint8_t* dst_ptr, T v32, int width) { \ 989 SIMD_ALIGNED(uint8_t temp[64]); \ 990 int r = width & MASK; \ 991 int n = width & ~MASK; \ 992 if (n > 0) { \ 993 ANY_SIMD(dst_ptr, v32, n); \ 994 } \ 995 ANY_SIMD(temp, v32, MASK + 1); \ 996 memcpy(dst_ptr + n * BPP, temp, r * BPP); \ 997 } 998 999 #ifdef HAS_SETROW_X86 1000 ANY1(SetRow_Any_X86, SetRow_X86, uint8_t, 1, 3) 1001 #endif 1002 #ifdef HAS_SETROW_NEON 1003 ANY1(SetRow_Any_NEON, SetRow_NEON, uint8_t, 1, 15) 1004 #endif 1005 #ifdef HAS_ARGBSETROW_NEON 1006 ANY1(ARGBSetRow_Any_NEON, ARGBSetRow_NEON, uint32_t, 4, 3) 1007 #endif 1008 #ifdef HAS_ARGBSETROW_MSA 1009 ANY1(ARGBSetRow_Any_MSA, ARGBSetRow_MSA, uint32_t, 4, 3) 1010 #endif 1011 #undef ANY1 1012 1013 // Any 1 to 2. Outputs UV planes. 1014 #define ANY12(NAMEANY, ANY_SIMD, UVSHIFT, BPP, DUVSHIFT, MASK) \ 1015 void NAMEANY(const uint8_t* src_ptr, uint8_t* dst_u, uint8_t* dst_v, \ 1016 int width) { \ 1017 SIMD_ALIGNED(uint8_t temp[128 * 3]); \ 1018 memset(temp, 0, 128); /* for msan */ \ 1019 int r = width & MASK; \ 1020 int n = width & ~MASK; \ 1021 if (n > 0) { \ 1022 ANY_SIMD(src_ptr, dst_u, dst_v, n); \ 1023 } \ 1024 memcpy(temp, src_ptr + (n >> UVSHIFT) * BPP, SS(r, UVSHIFT) * BPP); \ 1025 ANY_SIMD(temp, temp + 128, temp + 256, MASK + 1); \ 1026 memcpy(dst_u + (n >> DUVSHIFT), temp + 128, SS(r, DUVSHIFT)); \ 1027 memcpy(dst_v + (n >> DUVSHIFT), temp + 256, SS(r, DUVSHIFT)); \ 1028 } 1029 1030 #ifdef HAS_SPLITUVROW_SSE2 1031 ANY12(SplitUVRow_Any_SSE2, SplitUVRow_SSE2, 0, 2, 0, 15) 1032 #endif 1033 #ifdef HAS_SPLITUVROW_AVX2 1034 ANY12(SplitUVRow_Any_AVX2, SplitUVRow_AVX2, 0, 2, 0, 31) 1035 #endif 1036 #ifdef HAS_SPLITUVROW_NEON 1037 ANY12(SplitUVRow_Any_NEON, SplitUVRow_NEON, 0, 2, 0, 15) 1038 #endif 1039 #ifdef HAS_SPLITUVROW_MSA 1040 ANY12(SplitUVRow_Any_MSA, SplitUVRow_MSA, 0, 2, 0, 31) 1041 #endif 1042 #ifdef HAS_ARGBTOUV444ROW_SSSE3 1043 ANY12(ARGBToUV444Row_Any_SSSE3, ARGBToUV444Row_SSSE3, 0, 4, 0, 15) 1044 #endif 1045 #ifdef HAS_YUY2TOUV422ROW_AVX2 1046 ANY12(YUY2ToUV422Row_Any_AVX2, YUY2ToUV422Row_AVX2, 1, 4, 1, 31) 1047 ANY12(UYVYToUV422Row_Any_AVX2, UYVYToUV422Row_AVX2, 1, 4, 1, 31) 1048 #endif 1049 #ifdef HAS_YUY2TOUV422ROW_SSE2 1050 ANY12(YUY2ToUV422Row_Any_SSE2, YUY2ToUV422Row_SSE2, 1, 4, 1, 15) 1051 ANY12(UYVYToUV422Row_Any_SSE2, UYVYToUV422Row_SSE2, 1, 4, 1, 15) 1052 #endif 1053 #ifdef HAS_YUY2TOUV422ROW_NEON 1054 ANY12(ARGBToUV444Row_Any_NEON, ARGBToUV444Row_NEON, 0, 4, 0, 7) 1055 ANY12(YUY2ToUV422Row_Any_NEON, YUY2ToUV422Row_NEON, 1, 4, 1, 15) 1056 ANY12(UYVYToUV422Row_Any_NEON, UYVYToUV422Row_NEON, 1, 4, 1, 15) 1057 #endif 1058 #ifdef HAS_YUY2TOUV422ROW_MSA 1059 ANY12(ARGBToUV444Row_Any_MSA, ARGBToUV444Row_MSA, 0, 4, 0, 15) 1060 ANY12(YUY2ToUV422Row_Any_MSA, YUY2ToUV422Row_MSA, 1, 4, 1, 31) 1061 ANY12(UYVYToUV422Row_Any_MSA, UYVYToUV422Row_MSA, 1, 4, 1, 31) 1062 #endif 1063 #undef ANY12 1064 1065 // Any 1 to 3. Outputs RGB planes. 1066 #define ANY13(NAMEANY, ANY_SIMD, BPP, MASK) \ 1067 void NAMEANY(const uint8_t* src_ptr, uint8_t* dst_r, uint8_t* dst_g, \ 1068 uint8_t* dst_b, int width) { \ 1069 SIMD_ALIGNED(uint8_t temp[16 * 6]); \ 1070 memset(temp, 0, 16 * 3); /* for msan */ \ 1071 int r = width & MASK; \ 1072 int n = width & ~MASK; \ 1073 if (n > 0) { \ 1074 ANY_SIMD(src_ptr, dst_r, dst_g, dst_b, n); \ 1075 } \ 1076 memcpy(temp, src_ptr + n * BPP, r * BPP); \ 1077 ANY_SIMD(temp, temp + 16 * 3, temp + 16 * 4, temp + 16 * 5, MASK + 1); \ 1078 memcpy(dst_r + n, temp + 16 * 3, r); \ 1079 memcpy(dst_g + n, temp + 16 * 4, r); \ 1080 memcpy(dst_b + n, temp + 16 * 5, r); \ 1081 } 1082 1083 #ifdef HAS_SPLITRGBROW_SSSE3 1084 ANY13(SplitRGBRow_Any_SSSE3, SplitRGBRow_SSSE3, 3, 15) 1085 #endif 1086 #ifdef HAS_SPLITRGBROW_NEON 1087 ANY13(SplitRGBRow_Any_NEON, SplitRGBRow_NEON, 3, 15) 1088 #endif 1089 1090 // Any 1 to 2 with source stride (2 rows of source). Outputs UV planes. 1091 // 128 byte row allows for 32 avx ARGB pixels. 1092 #define ANY12S(NAMEANY, ANY_SIMD, UVSHIFT, BPP, MASK) \ 1093 void NAMEANY(const uint8_t* src_ptr, int src_stride_ptr, uint8_t* dst_u, \ 1094 uint8_t* dst_v, int width) { \ 1095 SIMD_ALIGNED(uint8_t temp[128 * 4]); \ 1096 memset(temp, 0, 128 * 2); /* for msan */ \ 1097 int r = width & MASK; \ 1098 int n = width & ~MASK; \ 1099 if (n > 0) { \ 1100 ANY_SIMD(src_ptr, src_stride_ptr, dst_u, dst_v, n); \ 1101 } \ 1102 memcpy(temp, src_ptr + (n >> UVSHIFT) * BPP, SS(r, UVSHIFT) * BPP); \ 1103 memcpy(temp + 128, src_ptr + src_stride_ptr + (n >> UVSHIFT) * BPP, \ 1104 SS(r, UVSHIFT) * BPP); \ 1105 if ((width & 1) && UVSHIFT == 0) { /* repeat last pixel for subsample */ \ 1106 memcpy(temp + SS(r, UVSHIFT) * BPP, temp + SS(r, UVSHIFT) * BPP - BPP, \ 1107 BPP); \ 1108 memcpy(temp + 128 + SS(r, UVSHIFT) * BPP, \ 1109 temp + 128 + SS(r, UVSHIFT) * BPP - BPP, BPP); \ 1110 } \ 1111 ANY_SIMD(temp, 128, temp + 256, temp + 384, MASK + 1); \ 1112 memcpy(dst_u + (n >> 1), temp + 256, SS(r, 1)); \ 1113 memcpy(dst_v + (n >> 1), temp + 384, SS(r, 1)); \ 1114 } 1115 1116 #ifdef HAS_ARGBTOUVROW_AVX2 1117 ANY12S(ARGBToUVRow_Any_AVX2, ARGBToUVRow_AVX2, 0, 4, 31) 1118 #endif 1119 #ifdef HAS_ARGBTOUVJROW_AVX2 1120 ANY12S(ARGBToUVJRow_Any_AVX2, ARGBToUVJRow_AVX2, 0, 4, 31) 1121 #endif 1122 #ifdef HAS_ARGBTOUVROW_SSSE3 1123 ANY12S(ARGBToUVRow_Any_SSSE3, ARGBToUVRow_SSSE3, 0, 4, 15) 1124 ANY12S(ARGBToUVJRow_Any_SSSE3, ARGBToUVJRow_SSSE3, 0, 4, 15) 1125 ANY12S(BGRAToUVRow_Any_SSSE3, BGRAToUVRow_SSSE3, 0, 4, 15) 1126 ANY12S(ABGRToUVRow_Any_SSSE3, ABGRToUVRow_SSSE3, 0, 4, 15) 1127 ANY12S(RGBAToUVRow_Any_SSSE3, RGBAToUVRow_SSSE3, 0, 4, 15) 1128 #endif 1129 #ifdef HAS_YUY2TOUVROW_AVX2 1130 ANY12S(YUY2ToUVRow_Any_AVX2, YUY2ToUVRow_AVX2, 1, 4, 31) 1131 ANY12S(UYVYToUVRow_Any_AVX2, UYVYToUVRow_AVX2, 1, 4, 31) 1132 #endif 1133 #ifdef HAS_YUY2TOUVROW_SSE2 1134 ANY12S(YUY2ToUVRow_Any_SSE2, YUY2ToUVRow_SSE2, 1, 4, 15) 1135 ANY12S(UYVYToUVRow_Any_SSE2, UYVYToUVRow_SSE2, 1, 4, 15) 1136 #endif 1137 #ifdef HAS_ARGBTOUVROW_NEON 1138 ANY12S(ARGBToUVRow_Any_NEON, ARGBToUVRow_NEON, 0, 4, 15) 1139 #endif 1140 #ifdef HAS_ARGBTOUVROW_MSA 1141 ANY12S(ARGBToUVRow_Any_MSA, ARGBToUVRow_MSA, 0, 4, 31) 1142 #endif 1143 #ifdef HAS_ARGBTOUVJROW_NEON 1144 ANY12S(ARGBToUVJRow_Any_NEON, ARGBToUVJRow_NEON, 0, 4, 15) 1145 #endif 1146 #ifdef HAS_ARGBTOUVJROW_MSA 1147 ANY12S(ARGBToUVJRow_Any_MSA, ARGBToUVJRow_MSA, 0, 4, 31) 1148 #endif 1149 #ifdef HAS_BGRATOUVROW_NEON 1150 ANY12S(BGRAToUVRow_Any_NEON, BGRAToUVRow_NEON, 0, 4, 15) 1151 #endif 1152 #ifdef HAS_BGRATOUVROW_MSA 1153 ANY12S(BGRAToUVRow_Any_MSA, BGRAToUVRow_MSA, 0, 4, 31) 1154 #endif 1155 #ifdef HAS_ABGRTOUVROW_NEON 1156 ANY12S(ABGRToUVRow_Any_NEON, ABGRToUVRow_NEON, 0, 4, 15) 1157 #endif 1158 #ifdef HAS_ABGRTOUVROW_MSA 1159 ANY12S(ABGRToUVRow_Any_MSA, ABGRToUVRow_MSA, 0, 4, 31) 1160 #endif 1161 #ifdef HAS_RGBATOUVROW_NEON 1162 ANY12S(RGBAToUVRow_Any_NEON, RGBAToUVRow_NEON, 0, 4, 15) 1163 #endif 1164 #ifdef HAS_RGBATOUVROW_MSA 1165 ANY12S(RGBAToUVRow_Any_MSA, RGBAToUVRow_MSA, 0, 4, 31) 1166 #endif 1167 #ifdef HAS_RGB24TOUVROW_NEON 1168 ANY12S(RGB24ToUVRow_Any_NEON, RGB24ToUVRow_NEON, 0, 3, 15) 1169 #endif 1170 #ifdef HAS_RGB24TOUVROW_MSA 1171 ANY12S(RGB24ToUVRow_Any_MSA, RGB24ToUVRow_MSA, 0, 3, 15) 1172 #endif 1173 #ifdef HAS_RAWTOUVROW_NEON 1174 ANY12S(RAWToUVRow_Any_NEON, RAWToUVRow_NEON, 0, 3, 15) 1175 #endif 1176 #ifdef HAS_RAWTOUVROW_MSA 1177 ANY12S(RAWToUVRow_Any_MSA, RAWToUVRow_MSA, 0, 3, 15) 1178 #endif 1179 #ifdef HAS_RGB565TOUVROW_NEON 1180 ANY12S(RGB565ToUVRow_Any_NEON, RGB565ToUVRow_NEON, 0, 2, 15) 1181 #endif 1182 #ifdef HAS_RGB565TOUVROW_MSA 1183 ANY12S(RGB565ToUVRow_Any_MSA, RGB565ToUVRow_MSA, 0, 2, 15) 1184 #endif 1185 #ifdef HAS_ARGB1555TOUVROW_NEON 1186 ANY12S(ARGB1555ToUVRow_Any_NEON, ARGB1555ToUVRow_NEON, 0, 2, 15) 1187 #endif 1188 #ifdef HAS_ARGB1555TOUVROW_MSA 1189 ANY12S(ARGB1555ToUVRow_Any_MSA, ARGB1555ToUVRow_MSA, 0, 2, 15) 1190 #endif 1191 #ifdef HAS_ARGB4444TOUVROW_NEON 1192 ANY12S(ARGB4444ToUVRow_Any_NEON, ARGB4444ToUVRow_NEON, 0, 2, 15) 1193 #endif 1194 #ifdef HAS_YUY2TOUVROW_NEON 1195 ANY12S(YUY2ToUVRow_Any_NEON, YUY2ToUVRow_NEON, 1, 4, 15) 1196 #endif 1197 #ifdef HAS_UYVYTOUVROW_NEON 1198 ANY12S(UYVYToUVRow_Any_NEON, UYVYToUVRow_NEON, 1, 4, 15) 1199 #endif 1200 #ifdef HAS_YUY2TOUVROW_MSA 1201 ANY12S(YUY2ToUVRow_Any_MSA, YUY2ToUVRow_MSA, 1, 4, 31) 1202 #endif 1203 #ifdef HAS_UYVYTOUVROW_MSA 1204 ANY12S(UYVYToUVRow_Any_MSA, UYVYToUVRow_MSA, 1, 4, 31) 1205 #endif 1206 #undef ANY12S 1207 1208 #ifdef __cplusplus 1209 } // extern "C" 1210 } // namespace libyuv 1211 #endif 1212