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_MERGERGBROW_MMI 96 ANY31(MergeRGBRow_Any_MMI, MergeRGBRow_MMI, 0, 0, 3, 7) 97 #endif 98 #ifdef HAS_I422TOYUY2ROW_SSE2 99 ANY31(I422ToYUY2Row_Any_SSE2, I422ToYUY2Row_SSE2, 1, 1, 4, 15) 100 ANY31(I422ToUYVYRow_Any_SSE2, I422ToUYVYRow_SSE2, 1, 1, 4, 15) 101 #endif 102 #ifdef HAS_I422TOYUY2ROW_AVX2 103 ANY31(I422ToYUY2Row_Any_AVX2, I422ToYUY2Row_AVX2, 1, 1, 4, 31) 104 ANY31(I422ToUYVYRow_Any_AVX2, I422ToUYVYRow_AVX2, 1, 1, 4, 31) 105 #endif 106 #ifdef HAS_I422TOYUY2ROW_NEON 107 ANY31(I422ToYUY2Row_Any_NEON, I422ToYUY2Row_NEON, 1, 1, 4, 15) 108 #endif 109 #ifdef HAS_I422TOYUY2ROW_MSA 110 ANY31(I422ToYUY2Row_Any_MSA, I422ToYUY2Row_MSA, 1, 1, 4, 31) 111 #endif 112 #ifdef HAS_I422TOYUY2ROW_MMI 113 ANY31(I422ToYUY2Row_Any_MMI, I422ToYUY2Row_MMI, 1, 1, 4, 7) 114 #endif 115 #ifdef HAS_I422TOUYVYROW_NEON 116 ANY31(I422ToUYVYRow_Any_NEON, I422ToUYVYRow_NEON, 1, 1, 4, 15) 117 #endif 118 #ifdef HAS_I422TOUYVYROW_MSA 119 ANY31(I422ToUYVYRow_Any_MSA, I422ToUYVYRow_MSA, 1, 1, 4, 31) 120 #endif 121 #ifdef HAS_I422TOUYVYROW_MMI 122 ANY31(I422ToUYVYRow_Any_MMI, I422ToUYVYRow_MMI, 1, 1, 4, 7) 123 #endif 124 #ifdef HAS_BLENDPLANEROW_AVX2 125 ANY31(BlendPlaneRow_Any_AVX2, BlendPlaneRow_AVX2, 0, 0, 1, 31) 126 #endif 127 #ifdef HAS_BLENDPLANEROW_SSSE3 128 ANY31(BlendPlaneRow_Any_SSSE3, BlendPlaneRow_SSSE3, 0, 0, 1, 7) 129 #endif 130 #ifdef HAS_BLENDPLANEROW_MMI 131 ANY31(BlendPlaneRow_Any_MMI, BlendPlaneRow_MMI, 0, 0, 1, 7) 132 #endif 133 #undef ANY31 134 135 // Note that odd width replication includes 444 due to implementation 136 // on arm that subsamples 444 to 422 internally. 137 // Any 3 planes to 1 with yuvconstants 138 #define ANY31C(NAMEANY, ANY_SIMD, UVSHIFT, DUVSHIFT, BPP, MASK) \ 139 void NAMEANY(const uint8_t* y_buf, const uint8_t* u_buf, \ 140 const uint8_t* v_buf, uint8_t* dst_ptr, \ 141 const struct YuvConstants* yuvconstants, int width) { \ 142 SIMD_ALIGNED(uint8_t temp[128 * 4]); \ 143 memset(temp, 0, 128 * 3); /* for YUY2 and msan */ \ 144 int r = width & MASK; \ 145 int n = width & ~MASK; \ 146 if (n > 0) { \ 147 ANY_SIMD(y_buf, u_buf, v_buf, dst_ptr, yuvconstants, n); \ 148 } \ 149 memcpy(temp, y_buf + n, r); \ 150 memcpy(temp + 128, u_buf + (n >> UVSHIFT), SS(r, UVSHIFT)); \ 151 memcpy(temp + 256, v_buf + (n >> UVSHIFT), SS(r, UVSHIFT)); \ 152 if (width & 1) { \ 153 temp[128 + SS(r, UVSHIFT)] = temp[128 + SS(r, UVSHIFT) - 1]; \ 154 temp[256 + SS(r, UVSHIFT)] = temp[256 + SS(r, UVSHIFT) - 1]; \ 155 } \ 156 ANY_SIMD(temp, temp + 128, temp + 256, temp + 384, yuvconstants, \ 157 MASK + 1); \ 158 memcpy(dst_ptr + (n >> DUVSHIFT) * BPP, temp + 384, \ 159 SS(r, DUVSHIFT) * BPP); \ 160 } 161 162 #ifdef HAS_I422TOARGBROW_SSSE3 163 ANY31C(I422ToARGBRow_Any_SSSE3, I422ToARGBRow_SSSE3, 1, 0, 4, 7) 164 #endif 165 #ifdef HAS_I422TOAR30ROW_SSSE3 166 ANY31C(I422ToAR30Row_Any_SSSE3, I422ToAR30Row_SSSE3, 1, 0, 4, 7) 167 #endif 168 #ifdef HAS_I422TOAR30ROW_AVX2 169 ANY31C(I422ToAR30Row_Any_AVX2, I422ToAR30Row_AVX2, 1, 0, 4, 15) 170 #endif 171 #ifdef HAS_I444TOARGBROW_SSSE3 172 ANY31C(I444ToARGBRow_Any_SSSE3, I444ToARGBRow_SSSE3, 0, 0, 4, 7) 173 ANY31C(I422ToRGBARow_Any_SSSE3, I422ToRGBARow_SSSE3, 1, 0, 4, 7) 174 ANY31C(I422ToARGB4444Row_Any_SSSE3, I422ToARGB4444Row_SSSE3, 1, 0, 2, 7) 175 ANY31C(I422ToARGB1555Row_Any_SSSE3, I422ToARGB1555Row_SSSE3, 1, 0, 2, 7) 176 ANY31C(I422ToRGB565Row_Any_SSSE3, I422ToRGB565Row_SSSE3, 1, 0, 2, 7) 177 ANY31C(I422ToRGB24Row_Any_SSSE3, I422ToRGB24Row_SSSE3, 1, 0, 3, 15) 178 #endif // HAS_I444TOARGBROW_SSSE3 179 #ifdef HAS_I422TORGB24ROW_AVX2 180 ANY31C(I422ToRGB24Row_Any_AVX2, I422ToRGB24Row_AVX2, 1, 0, 3, 31) 181 #endif 182 #ifdef HAS_I422TOARGBROW_AVX2 183 ANY31C(I422ToARGBRow_Any_AVX2, I422ToARGBRow_AVX2, 1, 0, 4, 15) 184 #endif 185 #ifdef HAS_I422TORGBAROW_AVX2 186 ANY31C(I422ToRGBARow_Any_AVX2, I422ToRGBARow_AVX2, 1, 0, 4, 15) 187 #endif 188 #ifdef HAS_I444TOARGBROW_AVX2 189 ANY31C(I444ToARGBRow_Any_AVX2, I444ToARGBRow_AVX2, 0, 0, 4, 15) 190 #endif 191 #ifdef HAS_I422TOARGB4444ROW_AVX2 192 ANY31C(I422ToARGB4444Row_Any_AVX2, I422ToARGB4444Row_AVX2, 1, 0, 2, 15) 193 #endif 194 #ifdef HAS_I422TOARGB1555ROW_AVX2 195 ANY31C(I422ToARGB1555Row_Any_AVX2, I422ToARGB1555Row_AVX2, 1, 0, 2, 15) 196 #endif 197 #ifdef HAS_I422TORGB565ROW_AVX2 198 ANY31C(I422ToRGB565Row_Any_AVX2, I422ToRGB565Row_AVX2, 1, 0, 2, 15) 199 #endif 200 #ifdef HAS_I422TOARGBROW_NEON 201 ANY31C(I444ToARGBRow_Any_NEON, I444ToARGBRow_NEON, 0, 0, 4, 7) 202 ANY31C(I422ToARGBRow_Any_NEON, I422ToARGBRow_NEON, 1, 0, 4, 7) 203 ANY31C(I422ToRGBARow_Any_NEON, I422ToRGBARow_NEON, 1, 0, 4, 7) 204 ANY31C(I422ToRGB24Row_Any_NEON, I422ToRGB24Row_NEON, 1, 0, 3, 7) 205 ANY31C(I422ToARGB4444Row_Any_NEON, I422ToARGB4444Row_NEON, 1, 0, 2, 7) 206 ANY31C(I422ToARGB1555Row_Any_NEON, I422ToARGB1555Row_NEON, 1, 0, 2, 7) 207 ANY31C(I422ToRGB565Row_Any_NEON, I422ToRGB565Row_NEON, 1, 0, 2, 7) 208 #endif 209 #ifdef HAS_I422TOARGBROW_MSA 210 ANY31C(I444ToARGBRow_Any_MSA, I444ToARGBRow_MSA, 0, 0, 4, 7) 211 ANY31C(I422ToARGBRow_Any_MSA, I422ToARGBRow_MSA, 1, 0, 4, 7) 212 ANY31C(I422ToRGBARow_Any_MSA, I422ToRGBARow_MSA, 1, 0, 4, 7) 213 ANY31C(I422ToRGB24Row_Any_MSA, I422ToRGB24Row_MSA, 1, 0, 3, 15) 214 ANY31C(I422ToARGB4444Row_Any_MSA, I422ToARGB4444Row_MSA, 1, 0, 2, 7) 215 ANY31C(I422ToARGB1555Row_Any_MSA, I422ToARGB1555Row_MSA, 1, 0, 2, 7) 216 ANY31C(I422ToRGB565Row_Any_MSA, I422ToRGB565Row_MSA, 1, 0, 2, 7) 217 #endif 218 #undef ANY31C 219 220 // Any 3 planes of 16 bit to 1 with yuvconstants 221 // TODO(fbarchard): consider sharing this code with ANY31C 222 #define ANY31CT(NAMEANY, ANY_SIMD, UVSHIFT, DUVSHIFT, T, SBPP, BPP, MASK) \ 223 void NAMEANY(const T* y_buf, const T* u_buf, const T* v_buf, \ 224 uint8_t* dst_ptr, const struct YuvConstants* yuvconstants, \ 225 int width) { \ 226 SIMD_ALIGNED(T temp[16 * 3]); \ 227 SIMD_ALIGNED(uint8_t out[64]); \ 228 memset(temp, 0, 16 * 3 * SBPP); /* for YUY2 and msan */ \ 229 int r = width & MASK; \ 230 int n = width & ~MASK; \ 231 if (n > 0) { \ 232 ANY_SIMD(y_buf, u_buf, v_buf, dst_ptr, yuvconstants, n); \ 233 } \ 234 memcpy(temp, y_buf + n, r * SBPP); \ 235 memcpy(temp + 16, u_buf + (n >> UVSHIFT), SS(r, UVSHIFT) * SBPP); \ 236 memcpy(temp + 32, v_buf + (n >> UVSHIFT), SS(r, UVSHIFT) * SBPP); \ 237 ANY_SIMD(temp, temp + 16, temp + 32, out, yuvconstants, MASK + 1); \ 238 memcpy(dst_ptr + (n >> DUVSHIFT) * BPP, out, SS(r, DUVSHIFT) * BPP); \ 239 } 240 241 #ifdef HAS_I210TOAR30ROW_SSSE3 242 ANY31CT(I210ToAR30Row_Any_SSSE3, I210ToAR30Row_SSSE3, 1, 0, uint16_t, 2, 4, 7) 243 #endif 244 #ifdef HAS_I210TOARGBROW_SSSE3 245 ANY31CT(I210ToARGBRow_Any_SSSE3, I210ToARGBRow_SSSE3, 1, 0, uint16_t, 2, 4, 7) 246 #endif 247 #ifdef HAS_I210TOARGBROW_AVX2 248 ANY31CT(I210ToARGBRow_Any_AVX2, I210ToARGBRow_AVX2, 1, 0, uint16_t, 2, 4, 15) 249 #endif 250 #ifdef HAS_I210TOAR30ROW_AVX2 251 ANY31CT(I210ToAR30Row_Any_AVX2, I210ToAR30Row_AVX2, 1, 0, uint16_t, 2, 4, 15) 252 #endif 253 #undef ANY31CT 254 255 // Any 2 planes to 1. 256 #define ANY21(NAMEANY, ANY_SIMD, UVSHIFT, SBPP, SBPP2, BPP, MASK) \ 257 void NAMEANY(const uint8_t* y_buf, const uint8_t* uv_buf, uint8_t* dst_ptr, \ 258 int width) { \ 259 SIMD_ALIGNED(uint8_t temp[64 * 3]); \ 260 memset(temp, 0, 64 * 2); /* for msan */ \ 261 int r = width & MASK; \ 262 int n = width & ~MASK; \ 263 if (n > 0) { \ 264 ANY_SIMD(y_buf, uv_buf, dst_ptr, n); \ 265 } \ 266 memcpy(temp, y_buf + n * SBPP, r * SBPP); \ 267 memcpy(temp + 64, uv_buf + (n >> UVSHIFT) * SBPP2, \ 268 SS(r, UVSHIFT) * SBPP2); \ 269 ANY_SIMD(temp, temp + 64, temp + 128, MASK + 1); \ 270 memcpy(dst_ptr + n * BPP, temp + 128, r * BPP); \ 271 } 272 273 // Merge functions. 274 #ifdef HAS_MERGEUVROW_SSE2 275 ANY21(MergeUVRow_Any_SSE2, MergeUVRow_SSE2, 0, 1, 1, 2, 15) 276 #endif 277 #ifdef HAS_MERGEUVROW_AVX2 278 ANY21(MergeUVRow_Any_AVX2, MergeUVRow_AVX2, 0, 1, 1, 2, 31) 279 #endif 280 #ifdef HAS_MERGEUVROW_NEON 281 ANY21(MergeUVRow_Any_NEON, MergeUVRow_NEON, 0, 1, 1, 2, 15) 282 #endif 283 #ifdef HAS_MERGEUVROW_MSA 284 ANY21(MergeUVRow_Any_MSA, MergeUVRow_MSA, 0, 1, 1, 2, 15) 285 #endif 286 #ifdef HAS_MERGEUVROW_MMI 287 ANY21(MergeUVRow_Any_MMI, MergeUVRow_MMI, 0, 1, 1, 2, 7) 288 #endif 289 #ifdef HAS_NV21TOYUV24ROW_NEON 290 ANY21(NV21ToYUV24Row_Any_NEON, NV21ToYUV24Row_NEON, 1, 1, 2, 3, 15) 291 #endif 292 #ifdef HAS_NV21TOYUV24ROW_AVX2 293 ANY21(NV21ToYUV24Row_Any_AVX2, NV21ToYUV24Row_AVX2, 1, 1, 2, 3, 31) 294 #endif 295 // Math functions. 296 #ifdef HAS_ARGBMULTIPLYROW_SSE2 297 ANY21(ARGBMultiplyRow_Any_SSE2, ARGBMultiplyRow_SSE2, 0, 4, 4, 4, 3) 298 #endif 299 #ifdef HAS_ARGBADDROW_SSE2 300 ANY21(ARGBAddRow_Any_SSE2, ARGBAddRow_SSE2, 0, 4, 4, 4, 3) 301 #endif 302 #ifdef HAS_ARGBSUBTRACTROW_SSE2 303 ANY21(ARGBSubtractRow_Any_SSE2, ARGBSubtractRow_SSE2, 0, 4, 4, 4, 3) 304 #endif 305 #ifdef HAS_ARGBMULTIPLYROW_AVX2 306 ANY21(ARGBMultiplyRow_Any_AVX2, ARGBMultiplyRow_AVX2, 0, 4, 4, 4, 7) 307 #endif 308 #ifdef HAS_ARGBADDROW_AVX2 309 ANY21(ARGBAddRow_Any_AVX2, ARGBAddRow_AVX2, 0, 4, 4, 4, 7) 310 #endif 311 #ifdef HAS_ARGBSUBTRACTROW_AVX2 312 ANY21(ARGBSubtractRow_Any_AVX2, ARGBSubtractRow_AVX2, 0, 4, 4, 4, 7) 313 #endif 314 #ifdef HAS_ARGBMULTIPLYROW_NEON 315 ANY21(ARGBMultiplyRow_Any_NEON, ARGBMultiplyRow_NEON, 0, 4, 4, 4, 7) 316 #endif 317 #ifdef HAS_ARGBADDROW_NEON 318 ANY21(ARGBAddRow_Any_NEON, ARGBAddRow_NEON, 0, 4, 4, 4, 7) 319 #endif 320 #ifdef HAS_ARGBSUBTRACTROW_NEON 321 ANY21(ARGBSubtractRow_Any_NEON, ARGBSubtractRow_NEON, 0, 4, 4, 4, 7) 322 #endif 323 #ifdef HAS_ARGBMULTIPLYROW_MSA 324 ANY21(ARGBMultiplyRow_Any_MSA, ARGBMultiplyRow_MSA, 0, 4, 4, 4, 3) 325 #endif 326 #ifdef HAS_ARGBMULTIPLYROW_MMI 327 ANY21(ARGBMultiplyRow_Any_MMI, ARGBMultiplyRow_MMI, 0, 4, 4, 4, 1) 328 #endif 329 #ifdef HAS_ARGBADDROW_MSA 330 ANY21(ARGBAddRow_Any_MSA, ARGBAddRow_MSA, 0, 4, 4, 4, 7) 331 #endif 332 #ifdef HAS_ARGBADDROW_MMI 333 ANY21(ARGBAddRow_Any_MMI, ARGBAddRow_MMI, 0, 4, 4, 4, 1) 334 #endif 335 #ifdef HAS_ARGBSUBTRACTROW_MSA 336 ANY21(ARGBSubtractRow_Any_MSA, ARGBSubtractRow_MSA, 0, 4, 4, 4, 7) 337 #endif 338 #ifdef HAS_ARGBSUBTRACTROW_MMI 339 ANY21(ARGBSubtractRow_Any_MMI, ARGBSubtractRow_MMI, 0, 4, 4, 4, 1) 340 #endif 341 #ifdef HAS_SOBELROW_SSE2 342 ANY21(SobelRow_Any_SSE2, SobelRow_SSE2, 0, 1, 1, 4, 15) 343 #endif 344 #ifdef HAS_SOBELROW_NEON 345 ANY21(SobelRow_Any_NEON, SobelRow_NEON, 0, 1, 1, 4, 7) 346 #endif 347 #ifdef HAS_SOBELROW_MSA 348 ANY21(SobelRow_Any_MSA, SobelRow_MSA, 0, 1, 1, 4, 15) 349 #endif 350 #ifdef HAS_SOBELROW_MMI 351 ANY21(SobelRow_Any_MMI, SobelRow_MMI, 0, 1, 1, 4, 7) 352 #endif 353 #ifdef HAS_SOBELTOPLANEROW_SSE2 354 ANY21(SobelToPlaneRow_Any_SSE2, SobelToPlaneRow_SSE2, 0, 1, 1, 1, 15) 355 #endif 356 #ifdef HAS_SOBELTOPLANEROW_NEON 357 ANY21(SobelToPlaneRow_Any_NEON, SobelToPlaneRow_NEON, 0, 1, 1, 1, 15) 358 #endif 359 #ifdef HAS_SOBELTOPLANEROW_MSA 360 ANY21(SobelToPlaneRow_Any_MSA, SobelToPlaneRow_MSA, 0, 1, 1, 1, 31) 361 #endif 362 #ifdef HAS_SOBELTOPLANEROW_MMI 363 ANY21(SobelToPlaneRow_Any_MMI, SobelToPlaneRow_MMI, 0, 1, 1, 1, 7) 364 #endif 365 #ifdef HAS_SOBELXYROW_SSE2 366 ANY21(SobelXYRow_Any_SSE2, SobelXYRow_SSE2, 0, 1, 1, 4, 15) 367 #endif 368 #ifdef HAS_SOBELXYROW_NEON 369 ANY21(SobelXYRow_Any_NEON, SobelXYRow_NEON, 0, 1, 1, 4, 7) 370 #endif 371 #ifdef HAS_SOBELXYROW_MSA 372 ANY21(SobelXYRow_Any_MSA, SobelXYRow_MSA, 0, 1, 1, 4, 15) 373 #endif 374 #ifdef HAS_SOBELXYROW_MMI 375 ANY21(SobelXYRow_Any_MMI, SobelXYRow_MMI, 0, 1, 1, 4, 7) 376 #endif 377 #undef ANY21 378 379 // Any 2 planes to 1 with yuvconstants 380 #define ANY21C(NAMEANY, ANY_SIMD, UVSHIFT, SBPP, SBPP2, BPP, MASK) \ 381 void NAMEANY(const uint8_t* y_buf, const uint8_t* uv_buf, uint8_t* dst_ptr, \ 382 const struct YuvConstants* yuvconstants, int width) { \ 383 SIMD_ALIGNED(uint8_t temp[128 * 3]); \ 384 memset(temp, 0, 128 * 2); /* for msan */ \ 385 int r = width & MASK; \ 386 int n = width & ~MASK; \ 387 if (n > 0) { \ 388 ANY_SIMD(y_buf, uv_buf, dst_ptr, yuvconstants, n); \ 389 } \ 390 memcpy(temp, y_buf + n * SBPP, r * SBPP); \ 391 memcpy(temp + 128, uv_buf + (n >> UVSHIFT) * SBPP2, \ 392 SS(r, UVSHIFT) * SBPP2); \ 393 ANY_SIMD(temp, temp + 128, temp + 256, yuvconstants, MASK + 1); \ 394 memcpy(dst_ptr + n * BPP, temp + 256, r * BPP); \ 395 } 396 397 // Biplanar to RGB. 398 #ifdef HAS_NV12TOARGBROW_SSSE3 399 ANY21C(NV12ToARGBRow_Any_SSSE3, NV12ToARGBRow_SSSE3, 1, 1, 2, 4, 7) 400 #endif 401 #ifdef HAS_NV12TOARGBROW_AVX2 402 ANY21C(NV12ToARGBRow_Any_AVX2, NV12ToARGBRow_AVX2, 1, 1, 2, 4, 15) 403 #endif 404 #ifdef HAS_NV12TOARGBROW_NEON 405 ANY21C(NV12ToARGBRow_Any_NEON, NV12ToARGBRow_NEON, 1, 1, 2, 4, 7) 406 #endif 407 #ifdef HAS_NV12TOARGBROW_MSA 408 ANY21C(NV12ToARGBRow_Any_MSA, NV12ToARGBRow_MSA, 1, 1, 2, 4, 7) 409 #endif 410 #ifdef HAS_NV21TOARGBROW_SSSE3 411 ANY21C(NV21ToARGBRow_Any_SSSE3, NV21ToARGBRow_SSSE3, 1, 1, 2, 4, 7) 412 #endif 413 #ifdef HAS_NV21TOARGBROW_AVX2 414 ANY21C(NV21ToARGBRow_Any_AVX2, NV21ToARGBRow_AVX2, 1, 1, 2, 4, 15) 415 #endif 416 #ifdef HAS_NV21TOARGBROW_NEON 417 ANY21C(NV21ToARGBRow_Any_NEON, NV21ToARGBRow_NEON, 1, 1, 2, 4, 7) 418 #endif 419 #ifdef HAS_NV21TOARGBROW_MSA 420 ANY21C(NV21ToARGBRow_Any_MSA, NV21ToARGBRow_MSA, 1, 1, 2, 4, 7) 421 #endif 422 #ifdef HAS_NV12TORGB24ROW_NEON 423 ANY21C(NV12ToRGB24Row_Any_NEON, NV12ToRGB24Row_NEON, 1, 1, 2, 3, 7) 424 #endif 425 #ifdef HAS_NV21TORGB24ROW_NEON 426 ANY21C(NV21ToRGB24Row_Any_NEON, NV21ToRGB24Row_NEON, 1, 1, 2, 3, 7) 427 #endif 428 #ifdef HAS_NV12TORGB24ROW_SSSE3 429 ANY21C(NV12ToRGB24Row_Any_SSSE3, NV12ToRGB24Row_SSSE3, 1, 1, 2, 3, 15) 430 #endif 431 #ifdef HAS_NV21TORGB24ROW_SSSE3 432 ANY21C(NV21ToRGB24Row_Any_SSSE3, NV21ToRGB24Row_SSSE3, 1, 1, 2, 3, 15) 433 #endif 434 #ifdef HAS_NV12TORGB24ROW_AVX2 435 ANY21C(NV12ToRGB24Row_Any_AVX2, NV12ToRGB24Row_AVX2, 1, 1, 2, 3, 31) 436 #endif 437 #ifdef HAS_NV21TORGB24ROW_AVX2 438 ANY21C(NV21ToRGB24Row_Any_AVX2, NV21ToRGB24Row_AVX2, 1, 1, 2, 3, 31) 439 #endif 440 #ifdef HAS_NV12TORGB565ROW_SSSE3 441 ANY21C(NV12ToRGB565Row_Any_SSSE3, NV12ToRGB565Row_SSSE3, 1, 1, 2, 2, 7) 442 #endif 443 #ifdef HAS_NV12TORGB565ROW_AVX2 444 ANY21C(NV12ToRGB565Row_Any_AVX2, NV12ToRGB565Row_AVX2, 1, 1, 2, 2, 15) 445 #endif 446 #ifdef HAS_NV12TORGB565ROW_NEON 447 ANY21C(NV12ToRGB565Row_Any_NEON, NV12ToRGB565Row_NEON, 1, 1, 2, 2, 7) 448 #endif 449 #ifdef HAS_NV12TORGB565ROW_MSA 450 ANY21C(NV12ToRGB565Row_Any_MSA, NV12ToRGB565Row_MSA, 1, 1, 2, 2, 7) 451 #endif 452 #undef ANY21C 453 454 // Any 1 to 1. 455 #define ANY11(NAMEANY, ANY_SIMD, UVSHIFT, SBPP, BPP, MASK) \ 456 void NAMEANY(const uint8_t* src_ptr, uint8_t* dst_ptr, int width) { \ 457 SIMD_ALIGNED(uint8_t temp[128 * 2]); \ 458 memset(temp, 0, 128); /* for YUY2 and msan */ \ 459 int r = width & MASK; \ 460 int n = width & ~MASK; \ 461 if (n > 0) { \ 462 ANY_SIMD(src_ptr, dst_ptr, n); \ 463 } \ 464 memcpy(temp, src_ptr + (n >> UVSHIFT) * SBPP, SS(r, UVSHIFT) * SBPP); \ 465 ANY_SIMD(temp, temp + 128, MASK + 1); \ 466 memcpy(dst_ptr + n * BPP, temp + 128, r * BPP); \ 467 } 468 469 #ifdef HAS_COPYROW_AVX 470 ANY11(CopyRow_Any_AVX, CopyRow_AVX, 0, 1, 1, 63) 471 #endif 472 #ifdef HAS_COPYROW_SSE2 473 ANY11(CopyRow_Any_SSE2, CopyRow_SSE2, 0, 1, 1, 31) 474 #endif 475 #ifdef HAS_COPYROW_NEON 476 ANY11(CopyRow_Any_NEON, CopyRow_NEON, 0, 1, 1, 31) 477 #endif 478 #if defined(HAS_ARGBTORGB24ROW_SSSE3) 479 ANY11(ARGBToRGB24Row_Any_SSSE3, ARGBToRGB24Row_SSSE3, 0, 4, 3, 15) 480 ANY11(ARGBToRAWRow_Any_SSSE3, ARGBToRAWRow_SSSE3, 0, 4, 3, 15) 481 ANY11(ARGBToRGB565Row_Any_SSE2, ARGBToRGB565Row_SSE2, 0, 4, 2, 3) 482 ANY11(ARGBToARGB1555Row_Any_SSE2, ARGBToARGB1555Row_SSE2, 0, 4, 2, 3) 483 ANY11(ARGBToARGB4444Row_Any_SSE2, ARGBToARGB4444Row_SSE2, 0, 4, 2, 3) 484 #endif 485 #if defined(HAS_ARGBTORGB24ROW_AVX2) 486 ANY11(ARGBToRGB24Row_Any_AVX2, ARGBToRGB24Row_AVX2, 0, 4, 3, 31) 487 #endif 488 #if defined(HAS_ARGBTORGB24ROW_AVX512VBMI) 489 ANY11(ARGBToRGB24Row_Any_AVX512VBMI, ARGBToRGB24Row_AVX512VBMI, 0, 4, 3, 31) 490 #endif 491 #if defined(HAS_ARGBTORAWROW_AVX2) 492 ANY11(ARGBToRAWRow_Any_AVX2, ARGBToRAWRow_AVX2, 0, 4, 3, 31) 493 #endif 494 #if defined(HAS_ARGBTORGB565ROW_AVX2) 495 ANY11(ARGBToRGB565Row_Any_AVX2, ARGBToRGB565Row_AVX2, 0, 4, 2, 7) 496 #endif 497 #if defined(HAS_ARGBTOARGB4444ROW_AVX2) 498 ANY11(ARGBToARGB1555Row_Any_AVX2, ARGBToARGB1555Row_AVX2, 0, 4, 2, 7) 499 ANY11(ARGBToARGB4444Row_Any_AVX2, ARGBToARGB4444Row_AVX2, 0, 4, 2, 7) 500 #endif 501 #if defined(HAS_ABGRTOAR30ROW_SSSE3) 502 ANY11(ABGRToAR30Row_Any_SSSE3, ABGRToAR30Row_SSSE3, 0, 4, 4, 3) 503 #endif 504 #if defined(HAS_ARGBTOAR30ROW_SSSE3) 505 ANY11(ARGBToAR30Row_Any_SSSE3, ARGBToAR30Row_SSSE3, 0, 4, 4, 3) 506 #endif 507 #if defined(HAS_ABGRTOAR30ROW_AVX2) 508 ANY11(ABGRToAR30Row_Any_AVX2, ABGRToAR30Row_AVX2, 0, 4, 4, 7) 509 #endif 510 #if defined(HAS_ARGBTOAR30ROW_AVX2) 511 ANY11(ARGBToAR30Row_Any_AVX2, ARGBToAR30Row_AVX2, 0, 4, 4, 7) 512 #endif 513 #if defined(HAS_J400TOARGBROW_SSE2) 514 ANY11(J400ToARGBRow_Any_SSE2, J400ToARGBRow_SSE2, 0, 1, 4, 7) 515 #endif 516 #if defined(HAS_J400TOARGBROW_AVX2) 517 ANY11(J400ToARGBRow_Any_AVX2, J400ToARGBRow_AVX2, 0, 1, 4, 15) 518 #endif 519 #if defined(HAS_I400TOARGBROW_SSE2) 520 ANY11(I400ToARGBRow_Any_SSE2, I400ToARGBRow_SSE2, 0, 1, 4, 7) 521 #endif 522 #if defined(HAS_I400TOARGBROW_AVX2) 523 ANY11(I400ToARGBRow_Any_AVX2, I400ToARGBRow_AVX2, 0, 1, 4, 15) 524 #endif 525 #if defined(HAS_RGB24TOARGBROW_SSSE3) 526 ANY11(RGB24ToARGBRow_Any_SSSE3, RGB24ToARGBRow_SSSE3, 0, 3, 4, 15) 527 ANY11(RAWToARGBRow_Any_SSSE3, RAWToARGBRow_SSSE3, 0, 3, 4, 15) 528 ANY11(RGB565ToARGBRow_Any_SSE2, RGB565ToARGBRow_SSE2, 0, 2, 4, 7) 529 ANY11(ARGB1555ToARGBRow_Any_SSE2, ARGB1555ToARGBRow_SSE2, 0, 2, 4, 7) 530 ANY11(ARGB4444ToARGBRow_Any_SSE2, ARGB4444ToARGBRow_SSE2, 0, 2, 4, 7) 531 #endif 532 #if defined(HAS_RAWTORGB24ROW_SSSE3) 533 ANY11(RAWToRGB24Row_Any_SSSE3, RAWToRGB24Row_SSSE3, 0, 3, 3, 7) 534 #endif 535 #if defined(HAS_RGB565TOARGBROW_AVX2) 536 ANY11(RGB565ToARGBRow_Any_AVX2, RGB565ToARGBRow_AVX2, 0, 2, 4, 15) 537 #endif 538 #if defined(HAS_ARGB1555TOARGBROW_AVX2) 539 ANY11(ARGB1555ToARGBRow_Any_AVX2, ARGB1555ToARGBRow_AVX2, 0, 2, 4, 15) 540 #endif 541 #if defined(HAS_ARGB4444TOARGBROW_AVX2) 542 ANY11(ARGB4444ToARGBRow_Any_AVX2, ARGB4444ToARGBRow_AVX2, 0, 2, 4, 15) 543 #endif 544 #if defined(HAS_ARGBTORGB24ROW_NEON) 545 ANY11(ARGBToRGB24Row_Any_NEON, ARGBToRGB24Row_NEON, 0, 4, 3, 7) 546 ANY11(ARGBToRAWRow_Any_NEON, ARGBToRAWRow_NEON, 0, 4, 3, 7) 547 ANY11(ARGBToRGB565Row_Any_NEON, ARGBToRGB565Row_NEON, 0, 4, 2, 7) 548 ANY11(ARGBToARGB1555Row_Any_NEON, ARGBToARGB1555Row_NEON, 0, 4, 2, 7) 549 ANY11(ARGBToARGB4444Row_Any_NEON, ARGBToARGB4444Row_NEON, 0, 4, 2, 7) 550 ANY11(J400ToARGBRow_Any_NEON, J400ToARGBRow_NEON, 0, 1, 4, 7) 551 ANY11(I400ToARGBRow_Any_NEON, I400ToARGBRow_NEON, 0, 1, 4, 7) 552 #endif 553 #if defined(HAS_ARGBTORGB24ROW_MSA) 554 ANY11(ARGBToRGB24Row_Any_MSA, ARGBToRGB24Row_MSA, 0, 4, 3, 15) 555 ANY11(ARGBToRAWRow_Any_MSA, ARGBToRAWRow_MSA, 0, 4, 3, 15) 556 ANY11(ARGBToRGB565Row_Any_MSA, ARGBToRGB565Row_MSA, 0, 4, 2, 7) 557 ANY11(ARGBToARGB1555Row_Any_MSA, ARGBToARGB1555Row_MSA, 0, 4, 2, 7) 558 ANY11(ARGBToARGB4444Row_Any_MSA, ARGBToARGB4444Row_MSA, 0, 4, 2, 7) 559 ANY11(J400ToARGBRow_Any_MSA, J400ToARGBRow_MSA, 0, 1, 4, 15) 560 ANY11(I400ToARGBRow_Any_MSA, I400ToARGBRow_MSA, 0, 1, 4, 15) 561 #endif 562 #if defined(HAS_ARGBTORGB24ROW_MMI) 563 ANY11(ARGBToRGB24Row_Any_MMI, ARGBToRGB24Row_MMI, 0, 4, 3, 3) 564 ANY11(ARGBToRAWRow_Any_MMI, ARGBToRAWRow_MMI, 0, 4, 3, 3) 565 ANY11(ARGBToRGB565Row_Any_MMI, ARGBToRGB565Row_MMI, 0, 4, 2, 3) 566 ANY11(ARGBToARGB1555Row_Any_MMI, ARGBToARGB1555Row_MMI, 0, 4, 2, 3) 567 ANY11(ARGBToARGB4444Row_Any_MMI, ARGBToARGB4444Row_MMI, 0, 4, 2, 3) 568 ANY11(J400ToARGBRow_Any_MMI, J400ToARGBRow_MMI, 0, 1, 4, 3) 569 ANY11(I400ToARGBRow_Any_MMI, I400ToARGBRow_MMI, 0, 1, 4, 7) 570 #endif 571 #if defined(HAS_RAWTORGB24ROW_NEON) 572 ANY11(RAWToRGB24Row_Any_NEON, RAWToRGB24Row_NEON, 0, 3, 3, 7) 573 #endif 574 #if defined(HAS_RAWTORGB24ROW_MSA) 575 ANY11(RAWToRGB24Row_Any_MSA, RAWToRGB24Row_MSA, 0, 3, 3, 15) 576 #endif 577 #if defined(HAS_RAWTORGB24ROW_MMI) 578 ANY11(RAWToRGB24Row_Any_MMI, RAWToRGB24Row_MMI, 0, 3, 3, 3) 579 #endif 580 #ifdef HAS_ARGBTOYROW_AVX2 581 ANY11(ARGBToYRow_Any_AVX2, ARGBToYRow_AVX2, 0, 4, 1, 31) 582 #endif 583 #ifdef HAS_ARGBTOYJROW_AVX2 584 ANY11(ARGBToYJRow_Any_AVX2, ARGBToYJRow_AVX2, 0, 4, 1, 31) 585 #endif 586 #ifdef HAS_UYVYTOYROW_AVX2 587 ANY11(UYVYToYRow_Any_AVX2, UYVYToYRow_AVX2, 0, 2, 1, 31) 588 #endif 589 #ifdef HAS_YUY2TOYROW_AVX2 590 ANY11(YUY2ToYRow_Any_AVX2, YUY2ToYRow_AVX2, 1, 4, 1, 31) 591 #endif 592 #ifdef HAS_ARGBTOYROW_SSSE3 593 ANY11(ARGBToYRow_Any_SSSE3, ARGBToYRow_SSSE3, 0, 4, 1, 15) 594 #endif 595 #ifdef HAS_BGRATOYROW_SSSE3 596 ANY11(BGRAToYRow_Any_SSSE3, BGRAToYRow_SSSE3, 0, 4, 1, 15) 597 ANY11(ABGRToYRow_Any_SSSE3, ABGRToYRow_SSSE3, 0, 4, 1, 15) 598 ANY11(RGBAToYRow_Any_SSSE3, RGBAToYRow_SSSE3, 0, 4, 1, 15) 599 ANY11(YUY2ToYRow_Any_SSE2, YUY2ToYRow_SSE2, 1, 4, 1, 15) 600 ANY11(UYVYToYRow_Any_SSE2, UYVYToYRow_SSE2, 1, 4, 1, 15) 601 #endif 602 #ifdef HAS_ARGBTOYJROW_SSSE3 603 ANY11(ARGBToYJRow_Any_SSSE3, ARGBToYJRow_SSSE3, 0, 4, 1, 15) 604 #endif 605 #ifdef HAS_ARGBTOYROW_NEON 606 ANY11(ARGBToYRow_Any_NEON, ARGBToYRow_NEON, 0, 4, 1, 7) 607 #endif 608 #ifdef HAS_ARGBTOYROW_MSA 609 ANY11(ARGBToYRow_Any_MSA, ARGBToYRow_MSA, 0, 4, 1, 15) 610 #endif 611 #ifdef HAS_ARGBTOYROW_MMI 612 ANY11(ARGBToYRow_Any_MMI, ARGBToYRow_MMI, 0, 4, 1, 7) 613 #endif 614 #ifdef HAS_ARGBTOYJROW_NEON 615 ANY11(ARGBToYJRow_Any_NEON, ARGBToYJRow_NEON, 0, 4, 1, 7) 616 #endif 617 #ifdef HAS_ARGBTOYJROW_MSA 618 ANY11(ARGBToYJRow_Any_MSA, ARGBToYJRow_MSA, 0, 4, 1, 15) 619 #endif 620 #ifdef HAS_ARGBTOYJROW_MMI 621 ANY11(ARGBToYJRow_Any_MMI, ARGBToYJRow_MMI, 0, 4, 1, 7) 622 #endif 623 #ifdef HAS_BGRATOYROW_NEON 624 ANY11(BGRAToYRow_Any_NEON, BGRAToYRow_NEON, 0, 4, 1, 7) 625 #endif 626 #ifdef HAS_BGRATOYROW_MSA 627 ANY11(BGRAToYRow_Any_MSA, BGRAToYRow_MSA, 0, 4, 1, 15) 628 #endif 629 #ifdef HAS_BGRATOYROW_MMI 630 ANY11(BGRAToYRow_Any_MMI, BGRAToYRow_MMI, 0, 4, 1, 7) 631 #endif 632 #ifdef HAS_ABGRTOYROW_NEON 633 ANY11(ABGRToYRow_Any_NEON, ABGRToYRow_NEON, 0, 4, 1, 7) 634 #endif 635 #ifdef HAS_ABGRTOYROW_MSA 636 ANY11(ABGRToYRow_Any_MSA, ABGRToYRow_MSA, 0, 4, 1, 7) 637 #endif 638 #ifdef HAS_ABGRTOYROW_MMI 639 ANY11(ABGRToYRow_Any_MMI, ABGRToYRow_MMI, 0, 4, 1, 7) 640 #endif 641 #ifdef HAS_RGBATOYROW_NEON 642 ANY11(RGBAToYRow_Any_NEON, RGBAToYRow_NEON, 0, 4, 1, 7) 643 #endif 644 #ifdef HAS_RGBATOYROW_MSA 645 ANY11(RGBAToYRow_Any_MSA, RGBAToYRow_MSA, 0, 4, 1, 15) 646 #endif 647 #ifdef HAS_RGBATOYROW_MMI 648 ANY11(RGBAToYRow_Any_MMI, RGBAToYRow_MMI, 0, 4, 1, 7) 649 #endif 650 #ifdef HAS_RGB24TOYROW_NEON 651 ANY11(RGB24ToYRow_Any_NEON, RGB24ToYRow_NEON, 0, 3, 1, 7) 652 #endif 653 #ifdef HAS_RGB24TOYROW_MSA 654 ANY11(RGB24ToYRow_Any_MSA, RGB24ToYRow_MSA, 0, 3, 1, 15) 655 #endif 656 #ifdef HAS_RGB24TOYROW_MMI 657 ANY11(RGB24ToYRow_Any_MMI, RGB24ToYRow_MMI, 0, 3, 1, 7) 658 #endif 659 #ifdef HAS_RAWTOYROW_NEON 660 ANY11(RAWToYRow_Any_NEON, RAWToYRow_NEON, 0, 3, 1, 7) 661 #endif 662 #ifdef HAS_RAWTOYROW_MSA 663 ANY11(RAWToYRow_Any_MSA, RAWToYRow_MSA, 0, 3, 1, 15) 664 #endif 665 #ifdef HAS_RAWTOYROW_MMI 666 ANY11(RAWToYRow_Any_MMI, RAWToYRow_MMI, 0, 3, 1, 7) 667 #endif 668 #ifdef HAS_RGB565TOYROW_NEON 669 ANY11(RGB565ToYRow_Any_NEON, RGB565ToYRow_NEON, 0, 2, 1, 7) 670 #endif 671 #ifdef HAS_RGB565TOYROW_MSA 672 ANY11(RGB565ToYRow_Any_MSA, RGB565ToYRow_MSA, 0, 2, 1, 15) 673 #endif 674 #ifdef HAS_RGB565TOYROW_MMI 675 ANY11(RGB565ToYRow_Any_MMI, RGB565ToYRow_MMI, 0, 2, 1, 7) 676 #endif 677 #ifdef HAS_ARGB1555TOYROW_NEON 678 ANY11(ARGB1555ToYRow_Any_NEON, ARGB1555ToYRow_NEON, 0, 2, 1, 7) 679 #endif 680 #ifdef HAS_ARGB1555TOYROW_MSA 681 ANY11(ARGB1555ToYRow_Any_MSA, ARGB1555ToYRow_MSA, 0, 2, 1, 15) 682 #endif 683 #ifdef HAS_ARGB1555TOYROW_MMI 684 ANY11(ARGB1555ToYRow_Any_MMI, ARGB1555ToYRow_MMI, 0, 2, 1, 7) 685 #endif 686 #ifdef HAS_ARGB4444TOYROW_NEON 687 ANY11(ARGB4444ToYRow_Any_NEON, ARGB4444ToYRow_NEON, 0, 2, 1, 7) 688 #endif 689 #ifdef HAS_ARGB4444TOYROW_MMI 690 ANY11(ARGB4444ToYRow_Any_MMI, ARGB4444ToYRow_MMI, 0, 2, 1, 7) 691 #endif 692 #ifdef HAS_YUY2TOYROW_NEON 693 ANY11(YUY2ToYRow_Any_NEON, YUY2ToYRow_NEON, 1, 4, 1, 15) 694 #endif 695 #ifdef HAS_UYVYTOYROW_NEON 696 ANY11(UYVYToYRow_Any_NEON, UYVYToYRow_NEON, 1, 4, 1, 15) 697 #endif 698 #ifdef HAS_YUY2TOYROW_MSA 699 ANY11(YUY2ToYRow_Any_MSA, YUY2ToYRow_MSA, 1, 4, 1, 31) 700 #endif 701 #ifdef HAS_YUY2TOYROW_MMI 702 ANY11(YUY2ToYRow_Any_MMI, YUY2ToYRow_MMI, 1, 4, 1, 7) 703 #endif 704 #ifdef HAS_UYVYTOYROW_MSA 705 ANY11(UYVYToYRow_Any_MSA, UYVYToYRow_MSA, 1, 4, 1, 31) 706 #endif 707 #ifdef HAS_UYVYTOYROW_MMI 708 ANY11(UYVYToYRow_Any_MMI, UYVYToYRow_MMI, 1, 4, 1, 15) 709 #endif 710 #ifdef HAS_AYUVTOYROW_NEON 711 ANY11(AYUVToYRow_Any_NEON, AYUVToYRow_NEON, 0, 4, 1, 15) 712 #endif 713 #ifdef HAS_AYUVTOYROW_NEON 714 ANY11(UVToVURow_Any_NEON, UVToVURow_NEON, 0, 2, 2, 15) 715 #endif 716 #ifdef HAS_RGB24TOARGBROW_NEON 717 ANY11(RGB24ToARGBRow_Any_NEON, RGB24ToARGBRow_NEON, 0, 3, 4, 7) 718 #endif 719 #ifdef HAS_RGB24TOARGBROW_MSA 720 ANY11(RGB24ToARGBRow_Any_MSA, RGB24ToARGBRow_MSA, 0, 3, 4, 15) 721 #endif 722 #ifdef HAS_RGB24TOARGBROW_MMI 723 ANY11(RGB24ToARGBRow_Any_MMI, RGB24ToARGBRow_MMI, 0, 3, 4, 3) 724 #endif 725 #ifdef HAS_RAWTOARGBROW_NEON 726 ANY11(RAWToARGBRow_Any_NEON, RAWToARGBRow_NEON, 0, 3, 4, 7) 727 #endif 728 #ifdef HAS_RAWTOARGBROW_MSA 729 ANY11(RAWToARGBRow_Any_MSA, RAWToARGBRow_MSA, 0, 3, 4, 15) 730 #endif 731 #ifdef HAS_RAWTOARGBROW_MMI 732 ANY11(RAWToARGBRow_Any_MMI, RAWToARGBRow_MMI, 0, 3, 4, 3) 733 #endif 734 #ifdef HAS_RGB565TOARGBROW_NEON 735 ANY11(RGB565ToARGBRow_Any_NEON, RGB565ToARGBRow_NEON, 0, 2, 4, 7) 736 #endif 737 #ifdef HAS_RGB565TOARGBROW_MSA 738 ANY11(RGB565ToARGBRow_Any_MSA, RGB565ToARGBRow_MSA, 0, 2, 4, 15) 739 #endif 740 #ifdef HAS_RGB565TOARGBROW_MMI 741 ANY11(RGB565ToARGBRow_Any_MMI, RGB565ToARGBRow_MMI, 0, 2, 4, 3) 742 #endif 743 #ifdef HAS_ARGB1555TOARGBROW_NEON 744 ANY11(ARGB1555ToARGBRow_Any_NEON, ARGB1555ToARGBRow_NEON, 0, 2, 4, 7) 745 #endif 746 #ifdef HAS_ARGB1555TOARGBROW_MSA 747 ANY11(ARGB1555ToARGBRow_Any_MSA, ARGB1555ToARGBRow_MSA, 0, 2, 4, 15) 748 #endif 749 #ifdef HAS_ARGB1555TOARGBROW_MMI 750 ANY11(ARGB1555ToARGBRow_Any_MMI, ARGB1555ToARGBRow_MMI, 0, 2, 4, 3) 751 #endif 752 #ifdef HAS_ARGB4444TOARGBROW_NEON 753 ANY11(ARGB4444ToARGBRow_Any_NEON, ARGB4444ToARGBRow_NEON, 0, 2, 4, 7) 754 #endif 755 #ifdef HAS_ARGB4444TOARGBROW_MSA 756 ANY11(ARGB4444ToARGBRow_Any_MSA, ARGB4444ToARGBRow_MSA, 0, 2, 4, 15) 757 #endif 758 #ifdef HAS_ARGB4444TOARGBROW_MMI 759 ANY11(ARGB4444ToARGBRow_Any_MMI, ARGB4444ToARGBRow_MMI, 0, 2, 4, 3) 760 #endif 761 #ifdef HAS_ARGBATTENUATEROW_SSSE3 762 ANY11(ARGBAttenuateRow_Any_SSSE3, ARGBAttenuateRow_SSSE3, 0, 4, 4, 3) 763 #endif 764 #ifdef HAS_ARGBUNATTENUATEROW_SSE2 765 ANY11(ARGBUnattenuateRow_Any_SSE2, ARGBUnattenuateRow_SSE2, 0, 4, 4, 3) 766 #endif 767 #ifdef HAS_ARGBATTENUATEROW_AVX2 768 ANY11(ARGBAttenuateRow_Any_AVX2, ARGBAttenuateRow_AVX2, 0, 4, 4, 7) 769 #endif 770 #ifdef HAS_ARGBUNATTENUATEROW_AVX2 771 ANY11(ARGBUnattenuateRow_Any_AVX2, ARGBUnattenuateRow_AVX2, 0, 4, 4, 7) 772 #endif 773 #ifdef HAS_ARGBATTENUATEROW_NEON 774 ANY11(ARGBAttenuateRow_Any_NEON, ARGBAttenuateRow_NEON, 0, 4, 4, 7) 775 #endif 776 #ifdef HAS_ARGBATTENUATEROW_MSA 777 ANY11(ARGBAttenuateRow_Any_MSA, ARGBAttenuateRow_MSA, 0, 4, 4, 7) 778 #endif 779 #ifdef HAS_ARGBATTENUATEROW_MMI 780 ANY11(ARGBAttenuateRow_Any_MMI, ARGBAttenuateRow_MMI, 0, 4, 4, 1) 781 #endif 782 #ifdef HAS_ARGBEXTRACTALPHAROW_SSE2 783 ANY11(ARGBExtractAlphaRow_Any_SSE2, ARGBExtractAlphaRow_SSE2, 0, 4, 1, 7) 784 #endif 785 #ifdef HAS_ARGBEXTRACTALPHAROW_AVX2 786 ANY11(ARGBExtractAlphaRow_Any_AVX2, ARGBExtractAlphaRow_AVX2, 0, 4, 1, 31) 787 #endif 788 #ifdef HAS_ARGBEXTRACTALPHAROW_NEON 789 ANY11(ARGBExtractAlphaRow_Any_NEON, ARGBExtractAlphaRow_NEON, 0, 4, 1, 15) 790 #endif 791 #ifdef HAS_ARGBEXTRACTALPHAROW_MSA 792 ANY11(ARGBExtractAlphaRow_Any_MSA, ARGBExtractAlphaRow_MSA, 0, 4, 1, 15) 793 #endif 794 #ifdef HAS_ARGBEXTRACTALPHAROW_MMI 795 ANY11(ARGBExtractAlphaRow_Any_MMI, ARGBExtractAlphaRow_MMI, 0, 4, 1, 7) 796 #endif 797 #undef ANY11 798 799 // Any 1 to 1 blended. Destination is read, modify, write. 800 #define ANY11B(NAMEANY, ANY_SIMD, UVSHIFT, SBPP, BPP, MASK) \ 801 void NAMEANY(const uint8_t* src_ptr, uint8_t* dst_ptr, int width) { \ 802 SIMD_ALIGNED(uint8_t temp[64 * 2]); \ 803 memset(temp, 0, 64 * 2); /* for msan */ \ 804 int r = width & MASK; \ 805 int n = width & ~MASK; \ 806 if (n > 0) { \ 807 ANY_SIMD(src_ptr, dst_ptr, n); \ 808 } \ 809 memcpy(temp, src_ptr + (n >> UVSHIFT) * SBPP, SS(r, UVSHIFT) * SBPP); \ 810 memcpy(temp + 64, dst_ptr + n * BPP, r * BPP); \ 811 ANY_SIMD(temp, temp + 64, MASK + 1); \ 812 memcpy(dst_ptr + n * BPP, temp + 64, r * BPP); \ 813 } 814 815 #ifdef HAS_ARGBCOPYALPHAROW_AVX2 816 ANY11B(ARGBCopyAlphaRow_Any_AVX2, ARGBCopyAlphaRow_AVX2, 0, 4, 4, 15) 817 #endif 818 #ifdef HAS_ARGBCOPYALPHAROW_SSE2 819 ANY11B(ARGBCopyAlphaRow_Any_SSE2, ARGBCopyAlphaRow_SSE2, 0, 4, 4, 7) 820 #endif 821 #ifdef HAS_ARGBCOPYALPHAROW_MMI 822 ANY11B(ARGBCopyAlphaRow_Any_MMI, ARGBCopyAlphaRow_MMI, 0, 4, 4, 1) 823 #endif 824 #ifdef HAS_ARGBCOPYYTOALPHAROW_AVX2 825 ANY11B(ARGBCopyYToAlphaRow_Any_AVX2, ARGBCopyYToAlphaRow_AVX2, 0, 1, 4, 15) 826 #endif 827 #ifdef HAS_ARGBCOPYYTOALPHAROW_SSE2 828 ANY11B(ARGBCopyYToAlphaRow_Any_SSE2, ARGBCopyYToAlphaRow_SSE2, 0, 1, 4, 7) 829 #endif 830 #ifdef HAS_ARGBCOPYYTOALPHAROW_MMI 831 ANY11B(ARGBCopyYToAlphaRow_Any_MMI, ARGBCopyYToAlphaRow_MMI, 0, 1, 4, 7) 832 #endif 833 #undef ANY11B 834 835 // Any 1 to 1 with parameter. 836 #define ANY11P(NAMEANY, ANY_SIMD, T, SBPP, BPP, MASK) \ 837 void NAMEANY(const uint8_t* src_ptr, uint8_t* dst_ptr, T param, int width) { \ 838 SIMD_ALIGNED(uint8_t temp[64 * 2]); \ 839 memset(temp, 0, 64); /* for msan */ \ 840 int r = width & MASK; \ 841 int n = width & ~MASK; \ 842 if (n > 0) { \ 843 ANY_SIMD(src_ptr, dst_ptr, param, n); \ 844 } \ 845 memcpy(temp, src_ptr + n * SBPP, r * SBPP); \ 846 ANY_SIMD(temp, temp + 64, param, MASK + 1); \ 847 memcpy(dst_ptr + n * BPP, temp + 64, r * BPP); \ 848 } 849 850 #if defined(HAS_ARGBTORGB565DITHERROW_SSE2) 851 ANY11P(ARGBToRGB565DitherRow_Any_SSE2, 852 ARGBToRGB565DitherRow_SSE2, 853 const uint32_t, 854 4, 855 2, 856 3) 857 #endif 858 #if defined(HAS_ARGBTORGB565DITHERROW_AVX2) 859 ANY11P(ARGBToRGB565DitherRow_Any_AVX2, 860 ARGBToRGB565DitherRow_AVX2, 861 const uint32_t, 862 4, 863 2, 864 7) 865 #endif 866 #if defined(HAS_ARGBTORGB565DITHERROW_NEON) 867 ANY11P(ARGBToRGB565DitherRow_Any_NEON, 868 ARGBToRGB565DitherRow_NEON, 869 const uint32_t, 870 4, 871 2, 872 7) 873 #endif 874 #if defined(HAS_ARGBTORGB565DITHERROW_MSA) 875 ANY11P(ARGBToRGB565DitherRow_Any_MSA, 876 ARGBToRGB565DitherRow_MSA, 877 const uint32_t, 878 4, 879 2, 880 7) 881 #endif 882 #if defined(HAS_ARGBTORGB565DITHERROW_MMI) 883 ANY11P(ARGBToRGB565DitherRow_Any_MMI, 884 ARGBToRGB565DitherRow_MMI, 885 const uint32_t, 886 4, 887 2, 888 3) 889 #endif 890 #ifdef HAS_ARGBSHUFFLEROW_SSSE3 891 ANY11P(ARGBShuffleRow_Any_SSSE3, ARGBShuffleRow_SSSE3, const uint8_t*, 4, 4, 7) 892 #endif 893 #ifdef HAS_ARGBSHUFFLEROW_AVX2 894 ANY11P(ARGBShuffleRow_Any_AVX2, ARGBShuffleRow_AVX2, const uint8_t*, 4, 4, 15) 895 #endif 896 #ifdef HAS_ARGBSHUFFLEROW_NEON 897 ANY11P(ARGBShuffleRow_Any_NEON, ARGBShuffleRow_NEON, const uint8_t*, 4, 4, 3) 898 #endif 899 #ifdef HAS_ARGBSHUFFLEROW_MSA 900 ANY11P(ARGBShuffleRow_Any_MSA, ARGBShuffleRow_MSA, const uint8_t*, 4, 4, 7) 901 #endif 902 #ifdef HAS_ARGBSHUFFLEROW_MMI 903 ANY11P(ARGBShuffleRow_Any_MMI, ARGBShuffleRow_MMI, const uint8_t*, 4, 4, 1) 904 #endif 905 #undef ANY11P 906 #undef ANY11P 907 908 // Any 1 to 1 with parameter and shorts. BPP measures in shorts. 909 #define ANY11C(NAMEANY, ANY_SIMD, SBPP, BPP, STYPE, DTYPE, MASK) \ 910 void NAMEANY(const STYPE* src_ptr, DTYPE* dst_ptr, int scale, int width) { \ 911 SIMD_ALIGNED(STYPE temp[32]); \ 912 SIMD_ALIGNED(DTYPE out[32]); \ 913 memset(temp, 0, 32 * SBPP); /* for msan */ \ 914 int r = width & MASK; \ 915 int n = width & ~MASK; \ 916 if (n > 0) { \ 917 ANY_SIMD(src_ptr, dst_ptr, scale, n); \ 918 } \ 919 memcpy(temp, src_ptr + n, r * SBPP); \ 920 ANY_SIMD(temp, out, scale, MASK + 1); \ 921 memcpy(dst_ptr + n, out, r * BPP); \ 922 } 923 924 #ifdef HAS_CONVERT16TO8ROW_SSSE3 925 ANY11C(Convert16To8Row_Any_SSSE3, 926 Convert16To8Row_SSSE3, 927 2, 928 1, 929 uint16_t, 930 uint8_t, 931 15) 932 #endif 933 #ifdef HAS_CONVERT16TO8ROW_AVX2 934 ANY11C(Convert16To8Row_Any_AVX2, 935 Convert16To8Row_AVX2, 936 2, 937 1, 938 uint16_t, 939 uint8_t, 940 31) 941 #endif 942 #ifdef HAS_CONVERT8TO16ROW_SSE2 943 ANY11C(Convert8To16Row_Any_SSE2, 944 Convert8To16Row_SSE2, 945 1, 946 2, 947 uint8_t, 948 uint16_t, 949 15) 950 #endif 951 #ifdef HAS_CONVERT8TO16ROW_AVX2 952 ANY11C(Convert8To16Row_Any_AVX2, 953 Convert8To16Row_AVX2, 954 1, 955 2, 956 uint8_t, 957 uint16_t, 958 31) 959 #endif 960 #undef ANY11C 961 962 // Any 1 to 1 with parameter and shorts to byte. BPP measures in shorts. 963 #define ANY11P16(NAMEANY, ANY_SIMD, ST, T, SBPP, BPP, MASK) \ 964 void NAMEANY(const ST* src_ptr, T* dst_ptr, float param, int width) { \ 965 SIMD_ALIGNED(ST temp[32]); \ 966 SIMD_ALIGNED(T out[32]); \ 967 memset(temp, 0, SBPP * 32); /* for msan */ \ 968 int r = width & MASK; \ 969 int n = width & ~MASK; \ 970 if (n > 0) { \ 971 ANY_SIMD(src_ptr, dst_ptr, param, n); \ 972 } \ 973 memcpy(temp, src_ptr + n, r * SBPP); \ 974 ANY_SIMD(temp, out, param, MASK + 1); \ 975 memcpy(dst_ptr + n, out, r * BPP); \ 976 } 977 978 #ifdef HAS_HALFFLOATROW_SSE2 979 ANY11P16(HalfFloatRow_Any_SSE2, HalfFloatRow_SSE2, uint16_t, uint16_t, 2, 2, 7) 980 #endif 981 #ifdef HAS_HALFFLOATROW_AVX2 982 ANY11P16(HalfFloatRow_Any_AVX2, HalfFloatRow_AVX2, uint16_t, uint16_t, 2, 2, 15) 983 #endif 984 #ifdef HAS_HALFFLOATROW_F16C 985 ANY11P16(HalfFloatRow_Any_F16C, HalfFloatRow_F16C, uint16_t, uint16_t, 2, 2, 15) 986 ANY11P16(HalfFloat1Row_Any_F16C, 987 HalfFloat1Row_F16C, 988 uint16_t, 989 uint16_t, 990 2, 991 2, 992 15) 993 #endif 994 #ifdef HAS_HALFFLOATROW_NEON 995 ANY11P16(HalfFloatRow_Any_NEON, HalfFloatRow_NEON, uint16_t, uint16_t, 2, 2, 7) 996 ANY11P16(HalfFloat1Row_Any_NEON, 997 HalfFloat1Row_NEON, 998 uint16_t, 999 uint16_t, 1000 2, 1001 2, 1002 7) 1003 #endif 1004 #ifdef HAS_HALFFLOATROW_MSA 1005 ANY11P16(HalfFloatRow_Any_MSA, HalfFloatRow_MSA, uint16_t, uint16_t, 2, 2, 31) 1006 #endif 1007 #ifdef HAS_BYTETOFLOATROW_NEON 1008 ANY11P16(ByteToFloatRow_Any_NEON, ByteToFloatRow_NEON, uint8_t, float, 1, 3, 7) 1009 #endif 1010 #undef ANY11P16 1011 1012 // Any 1 to 1 with yuvconstants 1013 #define ANY11C(NAMEANY, ANY_SIMD, UVSHIFT, SBPP, BPP, MASK) \ 1014 void NAMEANY(const uint8_t* src_ptr, uint8_t* dst_ptr, \ 1015 const struct YuvConstants* yuvconstants, int width) { \ 1016 SIMD_ALIGNED(uint8_t temp[128 * 2]); \ 1017 memset(temp, 0, 128); /* for YUY2 and msan */ \ 1018 int r = width & MASK; \ 1019 int n = width & ~MASK; \ 1020 if (n > 0) { \ 1021 ANY_SIMD(src_ptr, dst_ptr, yuvconstants, n); \ 1022 } \ 1023 memcpy(temp, src_ptr + (n >> UVSHIFT) * SBPP, SS(r, UVSHIFT) * SBPP); \ 1024 ANY_SIMD(temp, temp + 128, yuvconstants, MASK + 1); \ 1025 memcpy(dst_ptr + n * BPP, temp + 128, r * BPP); \ 1026 } 1027 #if defined(HAS_YUY2TOARGBROW_SSSE3) 1028 ANY11C(YUY2ToARGBRow_Any_SSSE3, YUY2ToARGBRow_SSSE3, 1, 4, 4, 15) 1029 ANY11C(UYVYToARGBRow_Any_SSSE3, UYVYToARGBRow_SSSE3, 1, 4, 4, 15) 1030 #endif 1031 #if defined(HAS_YUY2TOARGBROW_AVX2) 1032 ANY11C(YUY2ToARGBRow_Any_AVX2, YUY2ToARGBRow_AVX2, 1, 4, 4, 31) 1033 ANY11C(UYVYToARGBRow_Any_AVX2, UYVYToARGBRow_AVX2, 1, 4, 4, 31) 1034 #endif 1035 #if defined(HAS_YUY2TOARGBROW_NEON) 1036 ANY11C(YUY2ToARGBRow_Any_NEON, YUY2ToARGBRow_NEON, 1, 4, 4, 7) 1037 ANY11C(UYVYToARGBRow_Any_NEON, UYVYToARGBRow_NEON, 1, 4, 4, 7) 1038 #endif 1039 #if defined(HAS_YUY2TOARGBROW_MSA) 1040 ANY11C(YUY2ToARGBRow_Any_MSA, YUY2ToARGBRow_MSA, 1, 4, 4, 7) 1041 ANY11C(UYVYToARGBRow_Any_MSA, UYVYToARGBRow_MSA, 1, 4, 4, 7) 1042 #endif 1043 #undef ANY11C 1044 1045 // Any 1 to 1 interpolate. Takes 2 rows of source via stride. 1046 #define ANY11T(NAMEANY, ANY_SIMD, SBPP, BPP, MASK) \ 1047 void NAMEANY(uint8_t* dst_ptr, const uint8_t* src_ptr, \ 1048 ptrdiff_t src_stride_ptr, int width, int source_y_fraction) { \ 1049 SIMD_ALIGNED(uint8_t temp[64 * 3]); \ 1050 memset(temp, 0, 64 * 2); /* for msan */ \ 1051 int r = width & MASK; \ 1052 int n = width & ~MASK; \ 1053 if (n > 0) { \ 1054 ANY_SIMD(dst_ptr, src_ptr, src_stride_ptr, n, source_y_fraction); \ 1055 } \ 1056 memcpy(temp, src_ptr + n * SBPP, r * SBPP); \ 1057 memcpy(temp + 64, src_ptr + src_stride_ptr + n * SBPP, r * SBPP); \ 1058 ANY_SIMD(temp + 128, temp, 64, MASK + 1, source_y_fraction); \ 1059 memcpy(dst_ptr + n * BPP, temp + 128, r * BPP); \ 1060 } 1061 1062 #ifdef HAS_INTERPOLATEROW_AVX2 1063 ANY11T(InterpolateRow_Any_AVX2, InterpolateRow_AVX2, 1, 1, 31) 1064 #endif 1065 #ifdef HAS_INTERPOLATEROW_SSSE3 1066 ANY11T(InterpolateRow_Any_SSSE3, InterpolateRow_SSSE3, 1, 1, 15) 1067 #endif 1068 #ifdef HAS_INTERPOLATEROW_NEON 1069 ANY11T(InterpolateRow_Any_NEON, InterpolateRow_NEON, 1, 1, 15) 1070 #endif 1071 #ifdef HAS_INTERPOLATEROW_MSA 1072 ANY11T(InterpolateRow_Any_MSA, InterpolateRow_MSA, 1, 1, 31) 1073 #endif 1074 #ifdef HAS_INTERPOLATEROW_MMI 1075 ANY11T(InterpolateRow_Any_MMI, InterpolateRow_MMI, 1, 1, 7) 1076 #endif 1077 #undef ANY11T 1078 1079 // Any 1 to 1 mirror. 1080 #define ANY11M(NAMEANY, ANY_SIMD, BPP, MASK) \ 1081 void NAMEANY(const uint8_t* src_ptr, uint8_t* dst_ptr, int width) { \ 1082 SIMD_ALIGNED(uint8_t temp[64 * 2]); \ 1083 memset(temp, 0, 64); /* for msan */ \ 1084 int r = width & MASK; \ 1085 int n = width & ~MASK; \ 1086 if (n > 0) { \ 1087 ANY_SIMD(src_ptr + r * BPP, dst_ptr, n); \ 1088 } \ 1089 memcpy(temp, src_ptr, r* BPP); \ 1090 ANY_SIMD(temp, temp + 64, MASK + 1); \ 1091 memcpy(dst_ptr + n * BPP, temp + 64 + (MASK + 1 - r) * BPP, r * BPP); \ 1092 } 1093 1094 #ifdef HAS_MIRRORROW_AVX2 1095 ANY11M(MirrorRow_Any_AVX2, MirrorRow_AVX2, 1, 31) 1096 #endif 1097 #ifdef HAS_MIRRORROW_SSSE3 1098 ANY11M(MirrorRow_Any_SSSE3, MirrorRow_SSSE3, 1, 15) 1099 #endif 1100 #ifdef HAS_MIRRORROW_NEON 1101 ANY11M(MirrorRow_Any_NEON, MirrorRow_NEON, 1, 15) 1102 #endif 1103 #ifdef HAS_MIRRORROW_MSA 1104 ANY11M(MirrorRow_Any_MSA, MirrorRow_MSA, 1, 63) 1105 #endif 1106 #ifdef HAS_MIRRORROW_MMI 1107 ANY11M(MirrorRow_Any_MMI, MirrorRow_MMI, 1, 7) 1108 #endif 1109 #ifdef HAS_ARGBMIRRORROW_AVX2 1110 ANY11M(ARGBMirrorRow_Any_AVX2, ARGBMirrorRow_AVX2, 4, 7) 1111 #endif 1112 #ifdef HAS_ARGBMIRRORROW_SSE2 1113 ANY11M(ARGBMirrorRow_Any_SSE2, ARGBMirrorRow_SSE2, 4, 3) 1114 #endif 1115 #ifdef HAS_ARGBMIRRORROW_NEON 1116 ANY11M(ARGBMirrorRow_Any_NEON, ARGBMirrorRow_NEON, 4, 3) 1117 #endif 1118 #ifdef HAS_ARGBMIRRORROW_MSA 1119 ANY11M(ARGBMirrorRow_Any_MSA, ARGBMirrorRow_MSA, 4, 15) 1120 #endif 1121 #ifdef HAS_ARGBMIRRORROW_MMI 1122 ANY11M(ARGBMirrorRow_Any_MMI, ARGBMirrorRow_MMI, 4, 1) 1123 #endif 1124 #undef ANY11M 1125 1126 // Any 1 plane. (memset) 1127 #define ANY1(NAMEANY, ANY_SIMD, T, BPP, MASK) \ 1128 void NAMEANY(uint8_t* dst_ptr, T v32, int width) { \ 1129 SIMD_ALIGNED(uint8_t temp[64]); \ 1130 int r = width & MASK; \ 1131 int n = width & ~MASK; \ 1132 if (n > 0) { \ 1133 ANY_SIMD(dst_ptr, v32, n); \ 1134 } \ 1135 ANY_SIMD(temp, v32, MASK + 1); \ 1136 memcpy(dst_ptr + n * BPP, temp, r * BPP); \ 1137 } 1138 1139 #ifdef HAS_SETROW_X86 1140 ANY1(SetRow_Any_X86, SetRow_X86, uint8_t, 1, 3) 1141 #endif 1142 #ifdef HAS_SETROW_NEON 1143 ANY1(SetRow_Any_NEON, SetRow_NEON, uint8_t, 1, 15) 1144 #endif 1145 #ifdef HAS_ARGBSETROW_NEON 1146 ANY1(ARGBSetRow_Any_NEON, ARGBSetRow_NEON, uint32_t, 4, 3) 1147 #endif 1148 #ifdef HAS_ARGBSETROW_MSA 1149 ANY1(ARGBSetRow_Any_MSA, ARGBSetRow_MSA, uint32_t, 4, 3) 1150 #endif 1151 #undef ANY1 1152 1153 // Any 1 to 2. Outputs UV planes. 1154 #define ANY12(NAMEANY, ANY_SIMD, UVSHIFT, BPP, DUVSHIFT, MASK) \ 1155 void NAMEANY(const uint8_t* src_ptr, uint8_t* dst_u, uint8_t* dst_v, \ 1156 int width) { \ 1157 SIMD_ALIGNED(uint8_t temp[128 * 3]); \ 1158 memset(temp, 0, 128); /* for msan */ \ 1159 int r = width & MASK; \ 1160 int n = width & ~MASK; \ 1161 if (n > 0) { \ 1162 ANY_SIMD(src_ptr, dst_u, dst_v, n); \ 1163 } \ 1164 memcpy(temp, src_ptr + (n >> UVSHIFT) * BPP, SS(r, UVSHIFT) * BPP); \ 1165 ANY_SIMD(temp, temp + 128, temp + 256, MASK + 1); \ 1166 memcpy(dst_u + (n >> DUVSHIFT), temp + 128, SS(r, DUVSHIFT)); \ 1167 memcpy(dst_v + (n >> DUVSHIFT), temp + 256, SS(r, DUVSHIFT)); \ 1168 } 1169 1170 #ifdef HAS_SPLITUVROW_SSE2 1171 ANY12(SplitUVRow_Any_SSE2, SplitUVRow_SSE2, 0, 2, 0, 15) 1172 #endif 1173 #ifdef HAS_SPLITUVROW_AVX2 1174 ANY12(SplitUVRow_Any_AVX2, SplitUVRow_AVX2, 0, 2, 0, 31) 1175 #endif 1176 #ifdef HAS_SPLITUVROW_NEON 1177 ANY12(SplitUVRow_Any_NEON, SplitUVRow_NEON, 0, 2, 0, 15) 1178 #endif 1179 #ifdef HAS_SPLITUVROW_MSA 1180 ANY12(SplitUVRow_Any_MSA, SplitUVRow_MSA, 0, 2, 0, 31) 1181 #endif 1182 #ifdef HAS_SPLITUVROW_MMI 1183 ANY12(SplitUVRow_Any_MMI, SplitUVRow_MMI, 0, 2, 0, 7) 1184 #endif 1185 #ifdef HAS_ARGBTOUV444ROW_SSSE3 1186 ANY12(ARGBToUV444Row_Any_SSSE3, ARGBToUV444Row_SSSE3, 0, 4, 0, 15) 1187 #endif 1188 #ifdef HAS_YUY2TOUV422ROW_AVX2 1189 ANY12(YUY2ToUV422Row_Any_AVX2, YUY2ToUV422Row_AVX2, 1, 4, 1, 31) 1190 ANY12(UYVYToUV422Row_Any_AVX2, UYVYToUV422Row_AVX2, 1, 4, 1, 31) 1191 #endif 1192 #ifdef HAS_YUY2TOUV422ROW_SSE2 1193 ANY12(YUY2ToUV422Row_Any_SSE2, YUY2ToUV422Row_SSE2, 1, 4, 1, 15) 1194 ANY12(UYVYToUV422Row_Any_SSE2, UYVYToUV422Row_SSE2, 1, 4, 1, 15) 1195 #endif 1196 #ifdef HAS_YUY2TOUV422ROW_NEON 1197 ANY12(ARGBToUV444Row_Any_NEON, ARGBToUV444Row_NEON, 0, 4, 0, 7) 1198 ANY12(YUY2ToUV422Row_Any_NEON, YUY2ToUV422Row_NEON, 1, 4, 1, 15) 1199 ANY12(UYVYToUV422Row_Any_NEON, UYVYToUV422Row_NEON, 1, 4, 1, 15) 1200 #endif 1201 #ifdef HAS_YUY2TOUV422ROW_MSA 1202 ANY12(ARGBToUV444Row_Any_MSA, ARGBToUV444Row_MSA, 0, 4, 0, 15) 1203 ANY12(YUY2ToUV422Row_Any_MSA, YUY2ToUV422Row_MSA, 1, 4, 1, 31) 1204 ANY12(UYVYToUV422Row_Any_MSA, UYVYToUV422Row_MSA, 1, 4, 1, 31) 1205 #endif 1206 #ifdef HAS_YUY2TOUV422ROW_MMI 1207 ANY12(ARGBToUV444Row_Any_MMI, ARGBToUV444Row_MMI, 0, 4, 0, 7) 1208 ANY12(UYVYToUV422Row_Any_MMI, UYVYToUV422Row_MMI, 1, 4, 1, 15) 1209 ANY12(YUY2ToUV422Row_Any_MMI, YUY2ToUV422Row_MMI, 1, 4, 1, 15) 1210 #endif 1211 #undef ANY12 1212 1213 // Any 1 to 3. Outputs RGB planes. 1214 #define ANY13(NAMEANY, ANY_SIMD, BPP, MASK) \ 1215 void NAMEANY(const uint8_t* src_ptr, uint8_t* dst_r, uint8_t* dst_g, \ 1216 uint8_t* dst_b, int width) { \ 1217 SIMD_ALIGNED(uint8_t temp[16 * 6]); \ 1218 memset(temp, 0, 16 * 3); /* for msan */ \ 1219 int r = width & MASK; \ 1220 int n = width & ~MASK; \ 1221 if (n > 0) { \ 1222 ANY_SIMD(src_ptr, dst_r, dst_g, dst_b, n); \ 1223 } \ 1224 memcpy(temp, src_ptr + n * BPP, r * BPP); \ 1225 ANY_SIMD(temp, temp + 16 * 3, temp + 16 * 4, temp + 16 * 5, MASK + 1); \ 1226 memcpy(dst_r + n, temp + 16 * 3, r); \ 1227 memcpy(dst_g + n, temp + 16 * 4, r); \ 1228 memcpy(dst_b + n, temp + 16 * 5, r); \ 1229 } 1230 1231 #ifdef HAS_SPLITRGBROW_SSSE3 1232 ANY13(SplitRGBRow_Any_SSSE3, SplitRGBRow_SSSE3, 3, 15) 1233 #endif 1234 #ifdef HAS_SPLITRGBROW_NEON 1235 ANY13(SplitRGBRow_Any_NEON, SplitRGBRow_NEON, 3, 15) 1236 #endif 1237 #ifdef HAS_SPLITRGBROW_MMI 1238 ANY13(SplitRGBRow_Any_MMI, SplitRGBRow_MMI, 3, 3) 1239 #endif 1240 1241 // Any 1 to 2 with source stride (2 rows of source). Outputs UV planes. 1242 // 128 byte row allows for 32 avx ARGB pixels. 1243 #define ANY12S(NAMEANY, ANY_SIMD, UVSHIFT, BPP, MASK) \ 1244 void NAMEANY(const uint8_t* src_ptr, int src_stride_ptr, uint8_t* dst_u, \ 1245 uint8_t* dst_v, int width) { \ 1246 SIMD_ALIGNED(uint8_t temp[128 * 4]); \ 1247 memset(temp, 0, 128 * 2); /* for msan */ \ 1248 int r = width & MASK; \ 1249 int n = width & ~MASK; \ 1250 if (n > 0) { \ 1251 ANY_SIMD(src_ptr, src_stride_ptr, dst_u, dst_v, n); \ 1252 } \ 1253 memcpy(temp, src_ptr + (n >> UVSHIFT) * BPP, SS(r, UVSHIFT) * BPP); \ 1254 memcpy(temp + 128, src_ptr + src_stride_ptr + (n >> UVSHIFT) * BPP, \ 1255 SS(r, UVSHIFT) * BPP); \ 1256 if ((width & 1) && UVSHIFT == 0) { /* repeat last pixel for subsample */ \ 1257 memcpy(temp + SS(r, UVSHIFT) * BPP, temp + SS(r, UVSHIFT) * BPP - BPP, \ 1258 BPP); \ 1259 memcpy(temp + 128 + SS(r, UVSHIFT) * BPP, \ 1260 temp + 128 + SS(r, UVSHIFT) * BPP - BPP, BPP); \ 1261 } \ 1262 ANY_SIMD(temp, 128, temp + 256, temp + 384, MASK + 1); \ 1263 memcpy(dst_u + (n >> 1), temp + 256, SS(r, 1)); \ 1264 memcpy(dst_v + (n >> 1), temp + 384, SS(r, 1)); \ 1265 } 1266 1267 #ifdef HAS_ARGBTOUVROW_AVX2 1268 ANY12S(ARGBToUVRow_Any_AVX2, ARGBToUVRow_AVX2, 0, 4, 31) 1269 #endif 1270 #ifdef HAS_ARGBTOUVJROW_AVX2 1271 ANY12S(ARGBToUVJRow_Any_AVX2, ARGBToUVJRow_AVX2, 0, 4, 31) 1272 #endif 1273 #ifdef HAS_ARGBTOUVROW_SSSE3 1274 ANY12S(ARGBToUVRow_Any_SSSE3, ARGBToUVRow_SSSE3, 0, 4, 15) 1275 ANY12S(ARGBToUVJRow_Any_SSSE3, ARGBToUVJRow_SSSE3, 0, 4, 15) 1276 ANY12S(BGRAToUVRow_Any_SSSE3, BGRAToUVRow_SSSE3, 0, 4, 15) 1277 ANY12S(ABGRToUVRow_Any_SSSE3, ABGRToUVRow_SSSE3, 0, 4, 15) 1278 ANY12S(RGBAToUVRow_Any_SSSE3, RGBAToUVRow_SSSE3, 0, 4, 15) 1279 #endif 1280 #ifdef HAS_YUY2TOUVROW_AVX2 1281 ANY12S(YUY2ToUVRow_Any_AVX2, YUY2ToUVRow_AVX2, 1, 4, 31) 1282 ANY12S(UYVYToUVRow_Any_AVX2, UYVYToUVRow_AVX2, 1, 4, 31) 1283 #endif 1284 #ifdef HAS_YUY2TOUVROW_SSE2 1285 ANY12S(YUY2ToUVRow_Any_SSE2, YUY2ToUVRow_SSE2, 1, 4, 15) 1286 ANY12S(UYVYToUVRow_Any_SSE2, UYVYToUVRow_SSE2, 1, 4, 15) 1287 #endif 1288 #ifdef HAS_ARGBTOUVROW_NEON 1289 ANY12S(ARGBToUVRow_Any_NEON, ARGBToUVRow_NEON, 0, 4, 15) 1290 #endif 1291 #ifdef HAS_ARGBTOUVROW_MSA 1292 ANY12S(ARGBToUVRow_Any_MSA, ARGBToUVRow_MSA, 0, 4, 31) 1293 #endif 1294 #ifdef HAS_ARGBTOUVROW_MMI 1295 ANY12S(ARGBToUVRow_Any_MMI, ARGBToUVRow_MMI, 0, 4, 15) 1296 #endif 1297 #ifdef HAS_ARGBTOUVJROW_NEON 1298 ANY12S(ARGBToUVJRow_Any_NEON, ARGBToUVJRow_NEON, 0, 4, 15) 1299 #endif 1300 #ifdef HAS_ARGBTOUVJROW_MSA 1301 ANY12S(ARGBToUVJRow_Any_MSA, ARGBToUVJRow_MSA, 0, 4, 31) 1302 #endif 1303 #ifdef HAS_ARGBTOUVJROW_MMI 1304 ANY12S(ARGBToUVJRow_Any_MMI, ARGBToUVJRow_MMI, 0, 4, 15) 1305 #endif 1306 #ifdef HAS_BGRATOUVROW_NEON 1307 ANY12S(BGRAToUVRow_Any_NEON, BGRAToUVRow_NEON, 0, 4, 15) 1308 #endif 1309 #ifdef HAS_BGRATOUVROW_MSA 1310 ANY12S(BGRAToUVRow_Any_MSA, BGRAToUVRow_MSA, 0, 4, 31) 1311 #endif 1312 #ifdef HAS_BGRATOUVROW_MMI 1313 ANY12S(BGRAToUVRow_Any_MMI, BGRAToUVRow_MMI, 0, 4, 15) 1314 #endif 1315 #ifdef HAS_ABGRTOUVROW_NEON 1316 ANY12S(ABGRToUVRow_Any_NEON, ABGRToUVRow_NEON, 0, 4, 15) 1317 #endif 1318 #ifdef HAS_ABGRTOUVROW_MSA 1319 ANY12S(ABGRToUVRow_Any_MSA, ABGRToUVRow_MSA, 0, 4, 31) 1320 #endif 1321 #ifdef HAS_ABGRTOUVROW_MMI 1322 ANY12S(ABGRToUVRow_Any_MMI, ABGRToUVRow_MMI, 0, 4, 15) 1323 #endif 1324 #ifdef HAS_RGBATOUVROW_NEON 1325 ANY12S(RGBAToUVRow_Any_NEON, RGBAToUVRow_NEON, 0, 4, 15) 1326 #endif 1327 #ifdef HAS_RGBATOUVROW_MSA 1328 ANY12S(RGBAToUVRow_Any_MSA, RGBAToUVRow_MSA, 0, 4, 31) 1329 #endif 1330 #ifdef HAS_RGBATOUVROW_MMI 1331 ANY12S(RGBAToUVRow_Any_MMI, RGBAToUVRow_MMI, 0, 4, 15) 1332 #endif 1333 #ifdef HAS_RGB24TOUVROW_NEON 1334 ANY12S(RGB24ToUVRow_Any_NEON, RGB24ToUVRow_NEON, 0, 3, 15) 1335 #endif 1336 #ifdef HAS_RGB24TOUVROW_MSA 1337 ANY12S(RGB24ToUVRow_Any_MSA, RGB24ToUVRow_MSA, 0, 3, 15) 1338 #endif 1339 #ifdef HAS_RGB24TOUVROW_MMI 1340 ANY12S(RGB24ToUVRow_Any_MMI, RGB24ToUVRow_MMI, 0, 3, 15) 1341 #endif 1342 #ifdef HAS_RAWTOUVROW_NEON 1343 ANY12S(RAWToUVRow_Any_NEON, RAWToUVRow_NEON, 0, 3, 15) 1344 #endif 1345 #ifdef HAS_RAWTOUVROW_MSA 1346 ANY12S(RAWToUVRow_Any_MSA, RAWToUVRow_MSA, 0, 3, 15) 1347 #endif 1348 #ifdef HAS_RAWTOUVROW_MMI 1349 ANY12S(RAWToUVRow_Any_MMI, RAWToUVRow_MMI, 0, 3, 15) 1350 #endif 1351 #ifdef HAS_RGB565TOUVROW_NEON 1352 ANY12S(RGB565ToUVRow_Any_NEON, RGB565ToUVRow_NEON, 0, 2, 15) 1353 #endif 1354 #ifdef HAS_RGB565TOUVROW_MSA 1355 ANY12S(RGB565ToUVRow_Any_MSA, RGB565ToUVRow_MSA, 0, 2, 15) 1356 #endif 1357 #ifdef HAS_RGB565TOUVROW_MMI 1358 ANY12S(RGB565ToUVRow_Any_MMI, RGB565ToUVRow_MMI, 0, 2, 15) 1359 #endif 1360 #ifdef HAS_ARGB1555TOUVROW_NEON 1361 ANY12S(ARGB1555ToUVRow_Any_NEON, ARGB1555ToUVRow_NEON, 0, 2, 15) 1362 #endif 1363 #ifdef HAS_ARGB1555TOUVROW_MSA 1364 ANY12S(ARGB1555ToUVRow_Any_MSA, ARGB1555ToUVRow_MSA, 0, 2, 15) 1365 #endif 1366 #ifdef HAS_ARGB1555TOUVROW_MMI 1367 ANY12S(ARGB1555ToUVRow_Any_MMI, ARGB1555ToUVRow_MMI, 0, 2, 15) 1368 #endif 1369 #ifdef HAS_ARGB4444TOUVROW_NEON 1370 ANY12S(ARGB4444ToUVRow_Any_NEON, ARGB4444ToUVRow_NEON, 0, 2, 15) 1371 #endif 1372 #ifdef HAS_ARGB4444TOUVROW_MMI 1373 ANY12S(ARGB4444ToUVRow_Any_MMI, ARGB4444ToUVRow_MMI, 0, 2, 15) 1374 #endif 1375 #ifdef HAS_YUY2TOUVROW_NEON 1376 ANY12S(YUY2ToUVRow_Any_NEON, YUY2ToUVRow_NEON, 1, 4, 15) 1377 #endif 1378 #ifdef HAS_UYVYTOUVROW_NEON 1379 ANY12S(UYVYToUVRow_Any_NEON, UYVYToUVRow_NEON, 1, 4, 15) 1380 #endif 1381 #ifdef HAS_YUY2TOUVROW_MSA 1382 ANY12S(YUY2ToUVRow_Any_MSA, YUY2ToUVRow_MSA, 1, 4, 31) 1383 #endif 1384 #ifdef HAS_YUY2TOUVROW_MMI 1385 ANY12S(YUY2ToUVRow_Any_MMI, YUY2ToUVRow_MMI, 1, 4, 15) 1386 #endif 1387 #ifdef HAS_UYVYTOUVROW_MSA 1388 ANY12S(UYVYToUVRow_Any_MSA, UYVYToUVRow_MSA, 1, 4, 31) 1389 #endif 1390 #ifdef HAS_UYVYTOUVROW_MMI 1391 ANY12S(UYVYToUVRow_Any_MMI, UYVYToUVRow_MMI, 1, 4, 15) 1392 #endif 1393 #undef ANY12S 1394 1395 // Any 1 to 1 with source stride (2 rows of source). Outputs UV plane. 1396 // 128 byte row allows for 32 avx ARGB pixels. 1397 #define ANY11S(NAMEANY, ANY_SIMD, UVSHIFT, BPP, MASK) \ 1398 void NAMEANY(const uint8_t* src_ptr, int src_stride_ptr, uint8_t* dst_vu, \ 1399 int width) { \ 1400 SIMD_ALIGNED(uint8_t temp[128 * 3]); \ 1401 memset(temp, 0, 128 * 2); /* for msan */ \ 1402 int r = width & MASK; \ 1403 int n = width & ~MASK; \ 1404 if (n > 0) { \ 1405 ANY_SIMD(src_ptr, src_stride_ptr, dst_vu, n); \ 1406 } \ 1407 memcpy(temp, src_ptr + (n >> UVSHIFT) * BPP, SS(r, UVSHIFT) * BPP); \ 1408 memcpy(temp + 128, src_ptr + src_stride_ptr + (n >> UVSHIFT) * BPP, \ 1409 SS(r, UVSHIFT) * BPP); \ 1410 if ((width & 1) && UVSHIFT == 0) { /* repeat last pixel for subsample */ \ 1411 memcpy(temp + SS(r, UVSHIFT) * BPP, temp + SS(r, UVSHIFT) * BPP - BPP, \ 1412 BPP); \ 1413 memcpy(temp + 128 + SS(r, UVSHIFT) * BPP, \ 1414 temp + 128 + SS(r, UVSHIFT) * BPP - BPP, BPP); \ 1415 } \ 1416 ANY_SIMD(temp, 128, temp + 256, MASK + 1); \ 1417 memcpy(dst_vu + (n >> 1) * 2, temp + 256, SS(r, 1) * 2); \ 1418 } 1419 1420 #ifdef HAS_AYUVTOVUROW_NEON 1421 ANY11S(AYUVToUVRow_Any_NEON, AYUVToUVRow_NEON, 0, 4, 15) 1422 ANY11S(AYUVToVURow_Any_NEON, AYUVToVURow_NEON, 0, 4, 15) 1423 #endif 1424 #undef ANY11S 1425 1426 #ifdef __cplusplus 1427 } // extern "C" 1428 } // namespace libyuv 1429 #endif 1430