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 // Subsampled source needs to be increase by 1 of not even. 23 #define SS(width, shift) (((width) + (1 << (shift)) - 1) >> (shift)) 24 25 // Any 4 planes to 1 with yuvconstants 26 #define ANY41C(NAMEANY, ANY_SIMD, UVSHIFT, DUVSHIFT, BPP, MASK) \ 27 void NAMEANY(const uint8* y_buf, const uint8* u_buf, const uint8* v_buf, \ 28 const uint8* a_buf, uint8* dst_ptr, \ 29 const struct YuvConstants* yuvconstants, int width) { \ 30 SIMD_ALIGNED(uint8 temp[64 * 5]); \ 31 memset(temp, 0, 64 * 4); /* for msan */ \ 32 int r = width & MASK; \ 33 int n = width & ~MASK; \ 34 if (n > 0) { \ 35 ANY_SIMD(y_buf, u_buf, v_buf, a_buf, dst_ptr, yuvconstants, n); \ 36 } \ 37 memcpy(temp, y_buf + n, r); \ 38 memcpy(temp + 64, u_buf + (n >> UVSHIFT), SS(r, UVSHIFT)); \ 39 memcpy(temp + 128, v_buf + (n >> UVSHIFT), SS(r, UVSHIFT)); \ 40 memcpy(temp + 192, a_buf + n, r); \ 41 ANY_SIMD(temp, temp + 64, temp + 128, temp + 192, temp + 256, \ 42 yuvconstants, MASK + 1); \ 43 memcpy(dst_ptr + (n >> DUVSHIFT) * BPP, temp + 256, \ 44 SS(r, DUVSHIFT) * BPP); \ 45 } 46 47 #ifdef HAS_I422ALPHATOARGBROW_SSSE3 48 ANY41C(I422AlphaToARGBRow_Any_SSSE3, I422AlphaToARGBRow_SSSE3, 1, 0, 4, 7) 49 #endif 50 #ifdef HAS_I422ALPHATOARGBROW_AVX2 51 ANY41C(I422AlphaToARGBRow_Any_AVX2, I422AlphaToARGBRow_AVX2, 1, 0, 4, 15) 52 #endif 53 #ifdef HAS_I422ALPHATOARGBROW_NEON 54 ANY41C(I422AlphaToARGBRow_Any_NEON, I422AlphaToARGBRow_NEON, 1, 0, 4, 7) 55 #endif 56 #undef ANY41C 57 58 // Any 3 planes to 1. 59 #define ANY31(NAMEANY, ANY_SIMD, UVSHIFT, DUVSHIFT, BPP, MASK) \ 60 void NAMEANY(const uint8* y_buf, const uint8* u_buf, const uint8* v_buf, \ 61 uint8* dst_ptr, int width) { \ 62 SIMD_ALIGNED(uint8 temp[64 * 4]); \ 63 memset(temp, 0, 64 * 3); /* for YUY2 and msan */ \ 64 int r = width & MASK; \ 65 int n = width & ~MASK; \ 66 if (n > 0) { \ 67 ANY_SIMD(y_buf, u_buf, v_buf, dst_ptr, n); \ 68 } \ 69 memcpy(temp, y_buf + n, r); \ 70 memcpy(temp + 64, u_buf + (n >> UVSHIFT), SS(r, UVSHIFT)); \ 71 memcpy(temp + 128, v_buf + (n >> UVSHIFT), SS(r, UVSHIFT)); \ 72 ANY_SIMD(temp, temp + 64, temp + 128, temp + 192, MASK + 1); \ 73 memcpy(dst_ptr + (n >> DUVSHIFT) * BPP, temp + 192, \ 74 SS(r, DUVSHIFT) * BPP); \ 75 } 76 #ifdef HAS_I422TOYUY2ROW_SSE2 77 ANY31(I422ToYUY2Row_Any_SSE2, I422ToYUY2Row_SSE2, 1, 1, 4, 15) 78 ANY31(I422ToUYVYRow_Any_SSE2, I422ToUYVYRow_SSE2, 1, 1, 4, 15) 79 #endif 80 #ifdef HAS_I422TOYUY2ROW_NEON 81 ANY31(I422ToYUY2Row_Any_NEON, I422ToYUY2Row_NEON, 1, 1, 4, 15) 82 #endif 83 #ifdef HAS_I422TOUYVYROW_NEON 84 ANY31(I422ToUYVYRow_Any_NEON, I422ToUYVYRow_NEON, 1, 1, 4, 15) 85 #endif 86 #ifdef HAS_BLENDPLANEROW_AVX2 87 ANY31(BlendPlaneRow_Any_AVX2, BlendPlaneRow_AVX2, 0, 0, 1, 31) 88 #endif 89 #ifdef HAS_BLENDPLANEROW_SSSE3 90 ANY31(BlendPlaneRow_Any_SSSE3, BlendPlaneRow_SSSE3, 0, 0, 1, 7) 91 #endif 92 #undef ANY31 93 94 // Note that odd width replication includes 444 due to implementation 95 // on arm that subsamples 444 to 422 internally. 96 // Any 3 planes to 1 with yuvconstants 97 #define ANY31C(NAMEANY, ANY_SIMD, UVSHIFT, DUVSHIFT, BPP, MASK) \ 98 void NAMEANY(const uint8* y_buf, const uint8* u_buf, const uint8* v_buf, \ 99 uint8* dst_ptr, const struct YuvConstants* yuvconstants, \ 100 int width) { \ 101 SIMD_ALIGNED(uint8 temp[64 * 4]); \ 102 memset(temp, 0, 64 * 3); /* for YUY2 and msan */ \ 103 int r = width & MASK; \ 104 int n = width & ~MASK; \ 105 if (n > 0) { \ 106 ANY_SIMD(y_buf, u_buf, v_buf, dst_ptr, yuvconstants, n); \ 107 } \ 108 memcpy(temp, y_buf + n, r); \ 109 memcpy(temp + 64, u_buf + (n >> UVSHIFT), SS(r, UVSHIFT)); \ 110 memcpy(temp + 128, v_buf + (n >> UVSHIFT), SS(r, UVSHIFT)); \ 111 if (width & 1) { \ 112 temp[64 + SS(r, UVSHIFT)] = temp[64 + SS(r, UVSHIFT) - 1]; \ 113 temp[128 + SS(r, UVSHIFT)] = temp[128 + SS(r, UVSHIFT) - 1]; \ 114 } \ 115 ANY_SIMD(temp, temp + 64, temp + 128, temp + 192, \ 116 yuvconstants, MASK + 1); \ 117 memcpy(dst_ptr + (n >> DUVSHIFT) * BPP, temp + 192, \ 118 SS(r, DUVSHIFT) * BPP); \ 119 } 120 121 #ifdef HAS_I422TOARGBROW_SSSE3 122 ANY31C(I422ToARGBRow_Any_SSSE3, I422ToARGBRow_SSSE3, 1, 0, 4, 7) 123 #endif 124 #ifdef HAS_I411TOARGBROW_SSSE3 125 ANY31C(I411ToARGBRow_Any_SSSE3, I411ToARGBRow_SSSE3, 2, 0, 4, 7) 126 #endif 127 #ifdef HAS_I444TOARGBROW_SSSE3 128 ANY31C(I444ToARGBRow_Any_SSSE3, I444ToARGBRow_SSSE3, 0, 0, 4, 7) 129 ANY31C(I422ToRGBARow_Any_SSSE3, I422ToRGBARow_SSSE3, 1, 0, 4, 7) 130 ANY31C(I422ToARGB4444Row_Any_SSSE3, I422ToARGB4444Row_SSSE3, 1, 0, 2, 7) 131 ANY31C(I422ToARGB1555Row_Any_SSSE3, I422ToARGB1555Row_SSSE3, 1, 0, 2, 7) 132 ANY31C(I422ToRGB565Row_Any_SSSE3, I422ToRGB565Row_SSSE3, 1, 0, 2, 7) 133 ANY31C(I422ToRGB24Row_Any_SSSE3, I422ToRGB24Row_SSSE3, 1, 0, 3, 7) 134 #endif // HAS_I444TOARGBROW_SSSE3 135 #ifdef HAS_I422TORGB24ROW_AVX2 136 ANY31C(I422ToRGB24Row_Any_AVX2, I422ToRGB24Row_AVX2, 1, 0, 3, 15) 137 #endif 138 #ifdef HAS_I422TOARGBROW_AVX2 139 ANY31C(I422ToARGBRow_Any_AVX2, I422ToARGBRow_AVX2, 1, 0, 4, 15) 140 #endif 141 #ifdef HAS_I422TORGBAROW_AVX2 142 ANY31C(I422ToRGBARow_Any_AVX2, I422ToRGBARow_AVX2, 1, 0, 4, 15) 143 #endif 144 #ifdef HAS_I444TOARGBROW_AVX2 145 ANY31C(I444ToARGBRow_Any_AVX2, I444ToARGBRow_AVX2, 0, 0, 4, 15) 146 #endif 147 #ifdef HAS_I411TOARGBROW_AVX2 148 ANY31C(I411ToARGBRow_Any_AVX2, I411ToARGBRow_AVX2, 2, 0, 4, 15) 149 #endif 150 #ifdef HAS_I422TOARGB4444ROW_AVX2 151 ANY31C(I422ToARGB4444Row_Any_AVX2, I422ToARGB4444Row_AVX2, 1, 0, 2, 7) 152 #endif 153 #ifdef HAS_I422TOARGB1555ROW_AVX2 154 ANY31C(I422ToARGB1555Row_Any_AVX2, I422ToARGB1555Row_AVX2, 1, 0, 2, 7) 155 #endif 156 #ifdef HAS_I422TORGB565ROW_AVX2 157 ANY31C(I422ToRGB565Row_Any_AVX2, I422ToRGB565Row_AVX2, 1, 0, 2, 7) 158 #endif 159 #ifdef HAS_I422TOARGBROW_NEON 160 ANY31C(I444ToARGBRow_Any_NEON, I444ToARGBRow_NEON, 0, 0, 4, 7) 161 ANY31C(I422ToARGBRow_Any_NEON, I422ToARGBRow_NEON, 1, 0, 4, 7) 162 ANY31C(I411ToARGBRow_Any_NEON, I411ToARGBRow_NEON, 2, 0, 4, 7) 163 ANY31C(I422ToRGBARow_Any_NEON, I422ToRGBARow_NEON, 1, 0, 4, 7) 164 ANY31C(I422ToRGB24Row_Any_NEON, I422ToRGB24Row_NEON, 1, 0, 3, 7) 165 ANY31C(I422ToARGB4444Row_Any_NEON, I422ToARGB4444Row_NEON, 1, 0, 2, 7) 166 ANY31C(I422ToARGB1555Row_Any_NEON, I422ToARGB1555Row_NEON, 1, 0, 2, 7) 167 ANY31C(I422ToRGB565Row_Any_NEON, I422ToRGB565Row_NEON, 1, 0, 2, 7) 168 #endif 169 #undef ANY31C 170 171 // Any 2 planes to 1. 172 #define ANY21(NAMEANY, ANY_SIMD, UVSHIFT, SBPP, SBPP2, BPP, MASK) \ 173 void NAMEANY(const uint8* y_buf, const uint8* uv_buf, \ 174 uint8* dst_ptr, int width) { \ 175 SIMD_ALIGNED(uint8 temp[64 * 3]); \ 176 memset(temp, 0, 64 * 2); /* for msan */ \ 177 int r = width & MASK; \ 178 int n = width & ~MASK; \ 179 if (n > 0) { \ 180 ANY_SIMD(y_buf, uv_buf, dst_ptr, n); \ 181 } \ 182 memcpy(temp, y_buf + n * SBPP, r * SBPP); \ 183 memcpy(temp + 64, uv_buf + (n >> UVSHIFT) * SBPP2, \ 184 SS(r, UVSHIFT) * SBPP2); \ 185 ANY_SIMD(temp, temp + 64, temp + 128, MASK + 1); \ 186 memcpy(dst_ptr + n * BPP, temp + 128, r * BPP); \ 187 } 188 189 // Merge functions. 190 #ifdef HAS_MERGEUVROW_SSE2 191 ANY21(MergeUVRow_Any_SSE2, MergeUVRow_SSE2, 0, 1, 1, 2, 15) 192 #endif 193 #ifdef HAS_MERGEUVROW_AVX2 194 ANY21(MergeUVRow_Any_AVX2, MergeUVRow_AVX2, 0, 1, 1, 2, 31) 195 #endif 196 #ifdef HAS_MERGEUVROW_NEON 197 ANY21(MergeUVRow_Any_NEON, MergeUVRow_NEON, 0, 1, 1, 2, 15) 198 #endif 199 200 // Math functions. 201 #ifdef HAS_ARGBMULTIPLYROW_SSE2 202 ANY21(ARGBMultiplyRow_Any_SSE2, ARGBMultiplyRow_SSE2, 0, 4, 4, 4, 3) 203 #endif 204 #ifdef HAS_ARGBADDROW_SSE2 205 ANY21(ARGBAddRow_Any_SSE2, ARGBAddRow_SSE2, 0, 4, 4, 4, 3) 206 #endif 207 #ifdef HAS_ARGBSUBTRACTROW_SSE2 208 ANY21(ARGBSubtractRow_Any_SSE2, ARGBSubtractRow_SSE2, 0, 4, 4, 4, 3) 209 #endif 210 #ifdef HAS_ARGBMULTIPLYROW_AVX2 211 ANY21(ARGBMultiplyRow_Any_AVX2, ARGBMultiplyRow_AVX2, 0, 4, 4, 4, 7) 212 #endif 213 #ifdef HAS_ARGBADDROW_AVX2 214 ANY21(ARGBAddRow_Any_AVX2, ARGBAddRow_AVX2, 0, 4, 4, 4, 7) 215 #endif 216 #ifdef HAS_ARGBSUBTRACTROW_AVX2 217 ANY21(ARGBSubtractRow_Any_AVX2, ARGBSubtractRow_AVX2, 0, 4, 4, 4, 7) 218 #endif 219 #ifdef HAS_ARGBMULTIPLYROW_NEON 220 ANY21(ARGBMultiplyRow_Any_NEON, ARGBMultiplyRow_NEON, 0, 4, 4, 4, 7) 221 #endif 222 #ifdef HAS_ARGBADDROW_NEON 223 ANY21(ARGBAddRow_Any_NEON, ARGBAddRow_NEON, 0, 4, 4, 4, 7) 224 #endif 225 #ifdef HAS_ARGBSUBTRACTROW_NEON 226 ANY21(ARGBSubtractRow_Any_NEON, ARGBSubtractRow_NEON, 0, 4, 4, 4, 7) 227 #endif 228 #ifdef HAS_SOBELROW_SSE2 229 ANY21(SobelRow_Any_SSE2, SobelRow_SSE2, 0, 1, 1, 4, 15) 230 #endif 231 #ifdef HAS_SOBELROW_NEON 232 ANY21(SobelRow_Any_NEON, SobelRow_NEON, 0, 1, 1, 4, 7) 233 #endif 234 #ifdef HAS_SOBELTOPLANEROW_SSE2 235 ANY21(SobelToPlaneRow_Any_SSE2, SobelToPlaneRow_SSE2, 0, 1, 1, 1, 15) 236 #endif 237 #ifdef HAS_SOBELTOPLANEROW_NEON 238 ANY21(SobelToPlaneRow_Any_NEON, SobelToPlaneRow_NEON, 0, 1, 1, 1, 15) 239 #endif 240 #ifdef HAS_SOBELXYROW_SSE2 241 ANY21(SobelXYRow_Any_SSE2, SobelXYRow_SSE2, 0, 1, 1, 4, 15) 242 #endif 243 #ifdef HAS_SOBELXYROW_NEON 244 ANY21(SobelXYRow_Any_NEON, SobelXYRow_NEON, 0, 1, 1, 4, 7) 245 #endif 246 #undef ANY21 247 248 // Any 2 planes to 1 with yuvconstants 249 #define ANY21C(NAMEANY, ANY_SIMD, UVSHIFT, SBPP, SBPP2, BPP, MASK) \ 250 void NAMEANY(const uint8* y_buf, const uint8* uv_buf, \ 251 uint8* dst_ptr, const struct YuvConstants* yuvconstants, \ 252 int width) { \ 253 SIMD_ALIGNED(uint8 temp[64 * 3]); \ 254 memset(temp, 0, 64 * 2); /* for msan */ \ 255 int r = width & MASK; \ 256 int n = width & ~MASK; \ 257 if (n > 0) { \ 258 ANY_SIMD(y_buf, uv_buf, dst_ptr, yuvconstants, n); \ 259 } \ 260 memcpy(temp, y_buf + n * SBPP, r * SBPP); \ 261 memcpy(temp + 64, uv_buf + (n >> UVSHIFT) * SBPP2, \ 262 SS(r, UVSHIFT) * SBPP2); \ 263 ANY_SIMD(temp, temp + 64, temp + 128, yuvconstants, MASK + 1); \ 264 memcpy(dst_ptr + n * BPP, temp + 128, r * BPP); \ 265 } 266 267 // Biplanar to RGB. 268 #ifdef HAS_NV12TOARGBROW_SSSE3 269 ANY21C(NV12ToARGBRow_Any_SSSE3, NV12ToARGBRow_SSSE3, 1, 1, 2, 4, 7) 270 #endif 271 #ifdef HAS_NV12TOARGBROW_AVX2 272 ANY21C(NV12ToARGBRow_Any_AVX2, NV12ToARGBRow_AVX2, 1, 1, 2, 4, 15) 273 #endif 274 #ifdef HAS_NV12TOARGBROW_NEON 275 ANY21C(NV12ToARGBRow_Any_NEON, NV12ToARGBRow_NEON, 1, 1, 2, 4, 7) 276 #endif 277 #ifdef HAS_NV21TOARGBROW_SSSE3 278 ANY21C(NV21ToARGBRow_Any_SSSE3, NV21ToARGBRow_SSSE3, 1, 1, 2, 4, 7) 279 #endif 280 #ifdef HAS_NV21TOARGBROW_AVX2 281 ANY21C(NV21ToARGBRow_Any_AVX2, NV21ToARGBRow_AVX2, 1, 1, 2, 4, 15) 282 #endif 283 #ifdef HAS_NV21TOARGBROW_NEON 284 ANY21C(NV21ToARGBRow_Any_NEON, NV21ToARGBRow_NEON, 1, 1, 2, 4, 7) 285 #endif 286 #ifdef HAS_NV12TORGB565ROW_SSSE3 287 ANY21C(NV12ToRGB565Row_Any_SSSE3, NV12ToRGB565Row_SSSE3, 1, 1, 2, 2, 7) 288 #endif 289 #ifdef HAS_NV12TORGB565ROW_AVX2 290 ANY21C(NV12ToRGB565Row_Any_AVX2, NV12ToRGB565Row_AVX2, 1, 1, 2, 2, 15) 291 #endif 292 #ifdef HAS_NV12TORGB565ROW_NEON 293 ANY21C(NV12ToRGB565Row_Any_NEON, NV12ToRGB565Row_NEON, 1, 1, 2, 2, 7) 294 #endif 295 #undef ANY21C 296 297 // Any 1 to 1. 298 #define ANY11(NAMEANY, ANY_SIMD, UVSHIFT, SBPP, BPP, MASK) \ 299 void NAMEANY(const uint8* src_ptr, uint8* dst_ptr, int width) { \ 300 SIMD_ALIGNED(uint8 temp[128 * 2]); \ 301 memset(temp, 0, 128); /* for YUY2 and msan */ \ 302 int r = width & MASK; \ 303 int n = width & ~MASK; \ 304 if (n > 0) { \ 305 ANY_SIMD(src_ptr, dst_ptr, n); \ 306 } \ 307 memcpy(temp, src_ptr + (n >> UVSHIFT) * SBPP, SS(r, UVSHIFT) * SBPP); \ 308 ANY_SIMD(temp, temp + 128, MASK + 1); \ 309 memcpy(dst_ptr + n * BPP, temp + 128, r * BPP); \ 310 } 311 312 #ifdef HAS_COPYROW_AVX 313 ANY11(CopyRow_Any_AVX, CopyRow_AVX, 0, 1, 1, 63) 314 #endif 315 #ifdef HAS_COPYROW_SSE2 316 ANY11(CopyRow_Any_SSE2, CopyRow_SSE2, 0, 1, 1, 31) 317 #endif 318 #ifdef HAS_COPYROW_NEON 319 ANY11(CopyRow_Any_NEON, CopyRow_NEON, 0, 1, 1, 31) 320 #endif 321 #if defined(HAS_ARGBTORGB24ROW_SSSE3) 322 ANY11(ARGBToRGB24Row_Any_SSSE3, ARGBToRGB24Row_SSSE3, 0, 4, 3, 15) 323 ANY11(ARGBToRAWRow_Any_SSSE3, ARGBToRAWRow_SSSE3, 0, 4, 3, 15) 324 ANY11(ARGBToRGB565Row_Any_SSE2, ARGBToRGB565Row_SSE2, 0, 4, 2, 3) 325 ANY11(ARGBToARGB1555Row_Any_SSE2, ARGBToARGB1555Row_SSE2, 0, 4, 2, 3) 326 ANY11(ARGBToARGB4444Row_Any_SSE2, ARGBToARGB4444Row_SSE2, 0, 4, 2, 3) 327 #endif 328 #if defined(HAS_ARGBTORGB565ROW_AVX2) 329 ANY11(ARGBToRGB565Row_Any_AVX2, ARGBToRGB565Row_AVX2, 0, 4, 2, 7) 330 #endif 331 #if defined(HAS_ARGBTOARGB4444ROW_AVX2) 332 ANY11(ARGBToARGB1555Row_Any_AVX2, ARGBToARGB1555Row_AVX2, 0, 4, 2, 7) 333 ANY11(ARGBToARGB4444Row_Any_AVX2, ARGBToARGB4444Row_AVX2, 0, 4, 2, 7) 334 #endif 335 #if defined(HAS_J400TOARGBROW_SSE2) 336 ANY11(J400ToARGBRow_Any_SSE2, J400ToARGBRow_SSE2, 0, 1, 4, 7) 337 #endif 338 #if defined(HAS_J400TOARGBROW_AVX2) 339 ANY11(J400ToARGBRow_Any_AVX2, J400ToARGBRow_AVX2, 0, 1, 4, 15) 340 #endif 341 #if defined(HAS_I400TOARGBROW_SSE2) 342 ANY11(I400ToARGBRow_Any_SSE2, I400ToARGBRow_SSE2, 0, 1, 4, 7) 343 #endif 344 #if defined(HAS_I400TOARGBROW_AVX2) 345 ANY11(I400ToARGBRow_Any_AVX2, I400ToARGBRow_AVX2, 0, 1, 4, 15) 346 #endif 347 #if defined(HAS_RGB24TOARGBROW_SSSE3) 348 ANY11(RGB24ToARGBRow_Any_SSSE3, RGB24ToARGBRow_SSSE3, 0, 3, 4, 15) 349 ANY11(RAWToARGBRow_Any_SSSE3, RAWToARGBRow_SSSE3, 0, 3, 4, 15) 350 ANY11(RGB565ToARGBRow_Any_SSE2, RGB565ToARGBRow_SSE2, 0, 2, 4, 7) 351 ANY11(ARGB1555ToARGBRow_Any_SSE2, ARGB1555ToARGBRow_SSE2, 0, 2, 4, 7) 352 ANY11(ARGB4444ToARGBRow_Any_SSE2, ARGB4444ToARGBRow_SSE2, 0, 2, 4, 7) 353 #endif 354 #if defined(HAS_RAWTORGB24ROW_SSSE3) 355 ANY11(RAWToRGB24Row_Any_SSSE3, RAWToRGB24Row_SSSE3, 0, 3, 3, 7) 356 #endif 357 #if defined(HAS_RGB565TOARGBROW_AVX2) 358 ANY11(RGB565ToARGBRow_Any_AVX2, RGB565ToARGBRow_AVX2, 0, 2, 4, 15) 359 #endif 360 #if defined(HAS_ARGB1555TOARGBROW_AVX2) 361 ANY11(ARGB1555ToARGBRow_Any_AVX2, ARGB1555ToARGBRow_AVX2, 0, 2, 4, 15) 362 #endif 363 #if defined(HAS_ARGB4444TOARGBROW_AVX2) 364 ANY11(ARGB4444ToARGBRow_Any_AVX2, ARGB4444ToARGBRow_AVX2, 0, 2, 4, 15) 365 #endif 366 #if defined(HAS_ARGBTORGB24ROW_NEON) 367 ANY11(ARGBToRGB24Row_Any_NEON, ARGBToRGB24Row_NEON, 0, 4, 3, 7) 368 ANY11(ARGBToRAWRow_Any_NEON, ARGBToRAWRow_NEON, 0, 4, 3, 7) 369 ANY11(ARGBToRGB565Row_Any_NEON, ARGBToRGB565Row_NEON, 0, 4, 2, 7) 370 ANY11(ARGBToARGB1555Row_Any_NEON, ARGBToARGB1555Row_NEON, 0, 4, 2, 7) 371 ANY11(ARGBToARGB4444Row_Any_NEON, ARGBToARGB4444Row_NEON, 0, 4, 2, 7) 372 ANY11(J400ToARGBRow_Any_NEON, J400ToARGBRow_NEON, 0, 1, 4, 7) 373 ANY11(I400ToARGBRow_Any_NEON, I400ToARGBRow_NEON, 0, 1, 4, 7) 374 #endif 375 #if defined(HAS_RAWTORGB24ROW_NEON) 376 ANY11(RAWToRGB24Row_Any_NEON, RAWToRGB24Row_NEON, 0, 3, 3, 7) 377 #endif 378 #ifdef HAS_ARGBTOYROW_AVX2 379 ANY11(ARGBToYRow_Any_AVX2, ARGBToYRow_AVX2, 0, 4, 1, 31) 380 #endif 381 #ifdef HAS_ARGBTOYJROW_AVX2 382 ANY11(ARGBToYJRow_Any_AVX2, ARGBToYJRow_AVX2, 0, 4, 1, 31) 383 #endif 384 #ifdef HAS_UYVYTOYROW_AVX2 385 ANY11(UYVYToYRow_Any_AVX2, UYVYToYRow_AVX2, 0, 2, 1, 31) 386 #endif 387 #ifdef HAS_YUY2TOYROW_AVX2 388 ANY11(YUY2ToYRow_Any_AVX2, YUY2ToYRow_AVX2, 1, 4, 1, 31) 389 #endif 390 #ifdef HAS_ARGBTOYROW_SSSE3 391 ANY11(ARGBToYRow_Any_SSSE3, ARGBToYRow_SSSE3, 0, 4, 1, 15) 392 #endif 393 #ifdef HAS_BGRATOYROW_SSSE3 394 ANY11(BGRAToYRow_Any_SSSE3, BGRAToYRow_SSSE3, 0, 4, 1, 15) 395 ANY11(ABGRToYRow_Any_SSSE3, ABGRToYRow_SSSE3, 0, 4, 1, 15) 396 ANY11(RGBAToYRow_Any_SSSE3, RGBAToYRow_SSSE3, 0, 4, 1, 15) 397 ANY11(YUY2ToYRow_Any_SSE2, YUY2ToYRow_SSE2, 1, 4, 1, 15) 398 ANY11(UYVYToYRow_Any_SSE2, UYVYToYRow_SSE2, 1, 4, 1, 15) 399 #endif 400 #ifdef HAS_ARGBTOYJROW_SSSE3 401 ANY11(ARGBToYJRow_Any_SSSE3, ARGBToYJRow_SSSE3, 0, 4, 1, 15) 402 #endif 403 #ifdef HAS_ARGBTOYROW_NEON 404 ANY11(ARGBToYRow_Any_NEON, ARGBToYRow_NEON, 0, 4, 1, 7) 405 #endif 406 #ifdef HAS_ARGBTOYJROW_NEON 407 ANY11(ARGBToYJRow_Any_NEON, ARGBToYJRow_NEON, 0, 4, 1, 7) 408 #endif 409 #ifdef HAS_BGRATOYROW_NEON 410 ANY11(BGRAToYRow_Any_NEON, BGRAToYRow_NEON, 0, 4, 1, 7) 411 #endif 412 #ifdef HAS_ABGRTOYROW_NEON 413 ANY11(ABGRToYRow_Any_NEON, ABGRToYRow_NEON, 0, 4, 1, 7) 414 #endif 415 #ifdef HAS_RGBATOYROW_NEON 416 ANY11(RGBAToYRow_Any_NEON, RGBAToYRow_NEON, 0, 4, 1, 7) 417 #endif 418 #ifdef HAS_RGB24TOYROW_NEON 419 ANY11(RGB24ToYRow_Any_NEON, RGB24ToYRow_NEON, 0, 3, 1, 7) 420 #endif 421 #ifdef HAS_RAWTOYROW_NEON 422 ANY11(RAWToYRow_Any_NEON, RAWToYRow_NEON, 0, 3, 1, 7) 423 #endif 424 #ifdef HAS_RGB565TOYROW_NEON 425 ANY11(RGB565ToYRow_Any_NEON, RGB565ToYRow_NEON, 0, 2, 1, 7) 426 #endif 427 #ifdef HAS_ARGB1555TOYROW_NEON 428 ANY11(ARGB1555ToYRow_Any_NEON, ARGB1555ToYRow_NEON, 0, 2, 1, 7) 429 #endif 430 #ifdef HAS_ARGB4444TOYROW_NEON 431 ANY11(ARGB4444ToYRow_Any_NEON, ARGB4444ToYRow_NEON, 0, 2, 1, 7) 432 #endif 433 #ifdef HAS_YUY2TOYROW_NEON 434 ANY11(YUY2ToYRow_Any_NEON, YUY2ToYRow_NEON, 1, 4, 1, 15) 435 #endif 436 #ifdef HAS_UYVYTOYROW_NEON 437 ANY11(UYVYToYRow_Any_NEON, UYVYToYRow_NEON, 0, 2, 1, 15) 438 #endif 439 #ifdef HAS_RGB24TOARGBROW_NEON 440 ANY11(RGB24ToARGBRow_Any_NEON, RGB24ToARGBRow_NEON, 0, 3, 4, 7) 441 #endif 442 #ifdef HAS_RAWTOARGBROW_NEON 443 ANY11(RAWToARGBRow_Any_NEON, RAWToARGBRow_NEON, 0, 3, 4, 7) 444 #endif 445 #ifdef HAS_RGB565TOARGBROW_NEON 446 ANY11(RGB565ToARGBRow_Any_NEON, RGB565ToARGBRow_NEON, 0, 2, 4, 7) 447 #endif 448 #ifdef HAS_ARGB1555TOARGBROW_NEON 449 ANY11(ARGB1555ToARGBRow_Any_NEON, ARGB1555ToARGBRow_NEON, 0, 2, 4, 7) 450 #endif 451 #ifdef HAS_ARGB4444TOARGBROW_NEON 452 ANY11(ARGB4444ToARGBRow_Any_NEON, ARGB4444ToARGBRow_NEON, 0, 2, 4, 7) 453 #endif 454 #ifdef HAS_ARGBATTENUATEROW_SSSE3 455 ANY11(ARGBAttenuateRow_Any_SSSE3, ARGBAttenuateRow_SSSE3, 0, 4, 4, 3) 456 #endif 457 #ifdef HAS_ARGBUNATTENUATEROW_SSE2 458 ANY11(ARGBUnattenuateRow_Any_SSE2, ARGBUnattenuateRow_SSE2, 0, 4, 4, 3) 459 #endif 460 #ifdef HAS_ARGBATTENUATEROW_AVX2 461 ANY11(ARGBAttenuateRow_Any_AVX2, ARGBAttenuateRow_AVX2, 0, 4, 4, 7) 462 #endif 463 #ifdef HAS_ARGBUNATTENUATEROW_AVX2 464 ANY11(ARGBUnattenuateRow_Any_AVX2, ARGBUnattenuateRow_AVX2, 0, 4, 4, 7) 465 #endif 466 #ifdef HAS_ARGBATTENUATEROW_NEON 467 ANY11(ARGBAttenuateRow_Any_NEON, ARGBAttenuateRow_NEON, 0, 4, 4, 7) 468 #endif 469 #ifdef HAS_ARGBEXTRACTALPHAROW_SSE2 470 ANY11(ARGBExtractAlphaRow_Any_SSE2, ARGBExtractAlphaRow_SSE2, 0, 4, 1, 7) 471 #endif 472 #ifdef HAS_ARGBEXTRACTALPHAROW_NEON 473 ANY11(ARGBExtractAlphaRow_Any_NEON, ARGBExtractAlphaRow_NEON, 0, 4, 1, 15) 474 #endif 475 #undef ANY11 476 477 // Any 1 to 1 blended. Destination is read, modify, write. 478 #define ANY11B(NAMEANY, ANY_SIMD, UVSHIFT, SBPP, BPP, MASK) \ 479 void NAMEANY(const uint8* src_ptr, uint8* dst_ptr, int width) { \ 480 SIMD_ALIGNED(uint8 temp[128 * 2]); \ 481 memset(temp, 0, 128 * 2); /* for YUY2 and msan */ \ 482 int r = width & MASK; \ 483 int n = width & ~MASK; \ 484 if (n > 0) { \ 485 ANY_SIMD(src_ptr, dst_ptr, n); \ 486 } \ 487 memcpy(temp, src_ptr + (n >> UVSHIFT) * SBPP, SS(r, UVSHIFT) * SBPP); \ 488 memcpy(temp + 128, dst_ptr + n * BPP, r * BPP); \ 489 ANY_SIMD(temp, temp + 128, MASK + 1); \ 490 memcpy(dst_ptr + n * BPP, temp + 128, r * BPP); \ 491 } 492 493 #ifdef HAS_ARGBCOPYALPHAROW_AVX2 494 ANY11B(ARGBCopyAlphaRow_Any_AVX2, ARGBCopyAlphaRow_AVX2, 0, 4, 4, 15) 495 #endif 496 #ifdef HAS_ARGBCOPYALPHAROW_SSE2 497 ANY11B(ARGBCopyAlphaRow_Any_SSE2, ARGBCopyAlphaRow_SSE2, 0, 4, 4, 7) 498 #endif 499 #ifdef HAS_ARGBCOPYYTOALPHAROW_AVX2 500 ANY11B(ARGBCopyYToAlphaRow_Any_AVX2, ARGBCopyYToAlphaRow_AVX2, 0, 1, 4, 15) 501 #endif 502 #ifdef HAS_ARGBCOPYYTOALPHAROW_SSE2 503 ANY11B(ARGBCopyYToAlphaRow_Any_SSE2, ARGBCopyYToAlphaRow_SSE2, 0, 1, 4, 7) 504 #endif 505 #undef ANY11B 506 507 // Any 1 to 1 with parameter. 508 #define ANY11P(NAMEANY, ANY_SIMD, T, SBPP, BPP, MASK) \ 509 void NAMEANY(const uint8* src_ptr, uint8* dst_ptr, \ 510 T shuffler, int width) { \ 511 SIMD_ALIGNED(uint8 temp[64 * 2]); \ 512 memset(temp, 0, 64); /* for msan */ \ 513 int r = width & MASK; \ 514 int n = width & ~MASK; \ 515 if (n > 0) { \ 516 ANY_SIMD(src_ptr, dst_ptr, shuffler, n); \ 517 } \ 518 memcpy(temp, src_ptr + n * SBPP, r * SBPP); \ 519 ANY_SIMD(temp, temp + 64, shuffler, MASK + 1); \ 520 memcpy(dst_ptr + n * BPP, temp + 64, r * BPP); \ 521 } 522 523 #if defined(HAS_ARGBTORGB565DITHERROW_SSE2) 524 ANY11P(ARGBToRGB565DitherRow_Any_SSE2, ARGBToRGB565DitherRow_SSE2, 525 const uint32, 4, 2, 3) 526 #endif 527 #if defined(HAS_ARGBTORGB565DITHERROW_AVX2) 528 ANY11P(ARGBToRGB565DitherRow_Any_AVX2, ARGBToRGB565DitherRow_AVX2, 529 const uint32, 4, 2, 7) 530 #endif 531 #if defined(HAS_ARGBTORGB565DITHERROW_NEON) 532 ANY11P(ARGBToRGB565DitherRow_Any_NEON, ARGBToRGB565DitherRow_NEON, 533 const uint32, 4, 2, 7) 534 #endif 535 #ifdef HAS_ARGBSHUFFLEROW_SSE2 536 ANY11P(ARGBShuffleRow_Any_SSE2, ARGBShuffleRow_SSE2, const uint8*, 4, 4, 3) 537 #endif 538 #ifdef HAS_ARGBSHUFFLEROW_SSSE3 539 ANY11P(ARGBShuffleRow_Any_SSSE3, ARGBShuffleRow_SSSE3, const uint8*, 4, 4, 7) 540 #endif 541 #ifdef HAS_ARGBSHUFFLEROW_AVX2 542 ANY11P(ARGBShuffleRow_Any_AVX2, ARGBShuffleRow_AVX2, const uint8*, 4, 4, 15) 543 #endif 544 #ifdef HAS_ARGBSHUFFLEROW_NEON 545 ANY11P(ARGBShuffleRow_Any_NEON, ARGBShuffleRow_NEON, const uint8*, 4, 4, 3) 546 #endif 547 #undef ANY11P 548 549 // Any 1 to 1 with yuvconstants 550 #define ANY11C(NAMEANY, ANY_SIMD, UVSHIFT, SBPP, BPP, MASK) \ 551 void NAMEANY(const uint8* src_ptr, uint8* dst_ptr, \ 552 const struct YuvConstants* yuvconstants, int width) { \ 553 SIMD_ALIGNED(uint8 temp[128 * 2]); \ 554 memset(temp, 0, 128); /* for YUY2 and msan */ \ 555 int r = width & MASK; \ 556 int n = width & ~MASK; \ 557 if (n > 0) { \ 558 ANY_SIMD(src_ptr, dst_ptr, yuvconstants, n); \ 559 } \ 560 memcpy(temp, src_ptr + (n >> UVSHIFT) * SBPP, SS(r, UVSHIFT) * SBPP); \ 561 ANY_SIMD(temp, temp + 128, yuvconstants, MASK + 1); \ 562 memcpy(dst_ptr + n * BPP, temp + 128, r * BPP); \ 563 } 564 #if defined(HAS_YUY2TOARGBROW_SSSE3) 565 ANY11C(YUY2ToARGBRow_Any_SSSE3, YUY2ToARGBRow_SSSE3, 1, 4, 4, 15) 566 ANY11C(UYVYToARGBRow_Any_SSSE3, UYVYToARGBRow_SSSE3, 1, 4, 4, 15) 567 #endif 568 #if defined(HAS_YUY2TOARGBROW_AVX2) 569 ANY11C(YUY2ToARGBRow_Any_AVX2, YUY2ToARGBRow_AVX2, 1, 4, 4, 31) 570 ANY11C(UYVYToARGBRow_Any_AVX2, UYVYToARGBRow_AVX2, 1, 4, 4, 31) 571 #endif 572 #if defined(HAS_YUY2TOARGBROW_NEON) 573 ANY11C(YUY2ToARGBRow_Any_NEON, YUY2ToARGBRow_NEON, 1, 4, 4, 7) 574 ANY11C(UYVYToARGBRow_Any_NEON, UYVYToARGBRow_NEON, 1, 4, 4, 7) 575 #endif 576 #undef ANY11C 577 578 // Any 1 to 1 interpolate. Takes 2 rows of source via stride. 579 #define ANY11T(NAMEANY, ANY_SIMD, SBPP, BPP, MASK) \ 580 void NAMEANY(uint8* dst_ptr, const uint8* src_ptr, \ 581 ptrdiff_t src_stride_ptr, int width, \ 582 int source_y_fraction) { \ 583 SIMD_ALIGNED(uint8 temp[64 * 3]); \ 584 memset(temp, 0, 64 * 2); /* for msan */ \ 585 int r = width & MASK; \ 586 int n = width & ~MASK; \ 587 if (n > 0) { \ 588 ANY_SIMD(dst_ptr, src_ptr, src_stride_ptr, n, source_y_fraction); \ 589 } \ 590 memcpy(temp, src_ptr + n * SBPP, r * SBPP); \ 591 memcpy(temp + 64, src_ptr + src_stride_ptr + n * SBPP, r * SBPP); \ 592 ANY_SIMD(temp + 128, temp, 64, MASK + 1, source_y_fraction); \ 593 memcpy(dst_ptr + n * BPP, temp + 128, r * BPP); \ 594 } 595 596 #ifdef HAS_INTERPOLATEROW_AVX2 597 ANY11T(InterpolateRow_Any_AVX2, InterpolateRow_AVX2, 1, 1, 31) 598 #endif 599 #ifdef HAS_INTERPOLATEROW_SSSE3 600 ANY11T(InterpolateRow_Any_SSSE3, InterpolateRow_SSSE3, 1, 1, 15) 601 #endif 602 #ifdef HAS_INTERPOLATEROW_NEON 603 ANY11T(InterpolateRow_Any_NEON, InterpolateRow_NEON, 1, 1, 15) 604 #endif 605 #ifdef HAS_INTERPOLATEROW_DSPR2 606 ANY11T(InterpolateRow_Any_DSPR2, InterpolateRow_DSPR2, 1, 1, 3) 607 #endif 608 #undef ANY11T 609 610 // Any 1 to 1 mirror. 611 #define ANY11M(NAMEANY, ANY_SIMD, BPP, MASK) \ 612 void NAMEANY(const uint8* src_ptr, uint8* dst_ptr, int width) { \ 613 SIMD_ALIGNED(uint8 temp[64 * 2]); \ 614 memset(temp, 0, 64); /* for msan */ \ 615 int r = width & MASK; \ 616 int n = width & ~MASK; \ 617 if (n > 0) { \ 618 ANY_SIMD(src_ptr + r * BPP, dst_ptr, n); \ 619 } \ 620 memcpy(temp, src_ptr, r * BPP); \ 621 ANY_SIMD(temp, temp + 64, MASK + 1); \ 622 memcpy(dst_ptr + n * BPP, temp + 64 + (MASK + 1 - r) * BPP, r * BPP); \ 623 } 624 625 #ifdef HAS_MIRRORROW_AVX2 626 ANY11M(MirrorRow_Any_AVX2, MirrorRow_AVX2, 1, 31) 627 #endif 628 #ifdef HAS_MIRRORROW_SSSE3 629 ANY11M(MirrorRow_Any_SSSE3, MirrorRow_SSSE3, 1, 15) 630 #endif 631 #ifdef HAS_MIRRORROW_NEON 632 ANY11M(MirrorRow_Any_NEON, MirrorRow_NEON, 1, 15) 633 #endif 634 #ifdef HAS_ARGBMIRRORROW_AVX2 635 ANY11M(ARGBMirrorRow_Any_AVX2, ARGBMirrorRow_AVX2, 4, 7) 636 #endif 637 #ifdef HAS_ARGBMIRRORROW_SSE2 638 ANY11M(ARGBMirrorRow_Any_SSE2, ARGBMirrorRow_SSE2, 4, 3) 639 #endif 640 #ifdef HAS_ARGBMIRRORROW_NEON 641 ANY11M(ARGBMirrorRow_Any_NEON, ARGBMirrorRow_NEON, 4, 3) 642 #endif 643 #undef ANY11M 644 645 // Any 1 plane. (memset) 646 #define ANY1(NAMEANY, ANY_SIMD, T, BPP, MASK) \ 647 void NAMEANY(uint8* dst_ptr, T v32, int width) { \ 648 SIMD_ALIGNED(uint8 temp[64]); \ 649 int r = width & MASK; \ 650 int n = width & ~MASK; \ 651 if (n > 0) { \ 652 ANY_SIMD(dst_ptr, v32, n); \ 653 } \ 654 ANY_SIMD(temp, v32, MASK + 1); \ 655 memcpy(dst_ptr + n * BPP, temp, r * BPP); \ 656 } 657 658 #ifdef HAS_SETROW_X86 659 ANY1(SetRow_Any_X86, SetRow_X86, uint8, 1, 3) 660 #endif 661 #ifdef HAS_SETROW_NEON 662 ANY1(SetRow_Any_NEON, SetRow_NEON, uint8, 1, 15) 663 #endif 664 #ifdef HAS_ARGBSETROW_NEON 665 ANY1(ARGBSetRow_Any_NEON, ARGBSetRow_NEON, uint32, 4, 3) 666 #endif 667 #undef ANY1 668 669 // Any 1 to 2. Outputs UV planes. 670 #define ANY12(NAMEANY, ANY_SIMD, UVSHIFT, BPP, DUVSHIFT, MASK) \ 671 void NAMEANY(const uint8* src_ptr, uint8* dst_u, uint8* dst_v, int width) {\ 672 SIMD_ALIGNED(uint8 temp[128 * 3]); \ 673 memset(temp, 0, 128); /* for msan */ \ 674 int r = width & MASK; \ 675 int n = width & ~MASK; \ 676 if (n > 0) { \ 677 ANY_SIMD(src_ptr, dst_u, dst_v, n); \ 678 } \ 679 memcpy(temp, src_ptr + (n >> UVSHIFT) * BPP, SS(r, UVSHIFT) * BPP); \ 680 /* repeat last 4 bytes for 422 subsampler */ \ 681 if ((width & 1) && BPP == 4 && DUVSHIFT == 1) { \ 682 memcpy(temp + SS(r, UVSHIFT) * BPP, \ 683 temp + SS(r, UVSHIFT) * BPP - BPP, BPP); \ 684 } \ 685 /* repeat last 4 - 12 bytes for 411 subsampler */ \ 686 if (((width & 3) == 1) && BPP == 4 && DUVSHIFT == 2) { \ 687 memcpy(temp + SS(r, UVSHIFT) * BPP, \ 688 temp + SS(r, UVSHIFT) * BPP - BPP, BPP); \ 689 memcpy(temp + SS(r, UVSHIFT) * BPP + BPP, \ 690 temp + SS(r, UVSHIFT) * BPP - BPP, BPP * 2); \ 691 } \ 692 if (((width & 3) == 2) && BPP == 4 && DUVSHIFT == 2) { \ 693 memcpy(temp + SS(r, UVSHIFT) * BPP, \ 694 temp + SS(r, UVSHIFT) * BPP - BPP * 2, BPP * 2); \ 695 } \ 696 if (((width & 3) == 3) && BPP == 4 && DUVSHIFT == 2) { \ 697 memcpy(temp + SS(r, UVSHIFT) * BPP, \ 698 temp + SS(r, UVSHIFT) * BPP - BPP, BPP); \ 699 } \ 700 ANY_SIMD(temp, temp + 128, temp + 256, MASK + 1); \ 701 memcpy(dst_u + (n >> DUVSHIFT), temp + 128, SS(r, DUVSHIFT)); \ 702 memcpy(dst_v + (n >> DUVSHIFT), temp + 256, SS(r, DUVSHIFT)); \ 703 } 704 705 #ifdef HAS_SPLITUVROW_SSE2 706 ANY12(SplitUVRow_Any_SSE2, SplitUVRow_SSE2, 0, 2, 0, 15) 707 #endif 708 #ifdef HAS_SPLITUVROW_AVX2 709 ANY12(SplitUVRow_Any_AVX2, SplitUVRow_AVX2, 0, 2, 0, 31) 710 #endif 711 #ifdef HAS_SPLITUVROW_NEON 712 ANY12(SplitUVRow_Any_NEON, SplitUVRow_NEON, 0, 2, 0, 15) 713 #endif 714 #ifdef HAS_SPLITUVROW_DSPR2 715 ANY12(SplitUVRow_Any_DSPR2, SplitUVRow_DSPR2, 0, 2, 0, 15) 716 #endif 717 #ifdef HAS_ARGBTOUV444ROW_SSSE3 718 ANY12(ARGBToUV444Row_Any_SSSE3, ARGBToUV444Row_SSSE3, 0, 4, 0, 15) 719 #endif 720 #ifdef HAS_YUY2TOUV422ROW_AVX2 721 ANY12(YUY2ToUV422Row_Any_AVX2, YUY2ToUV422Row_AVX2, 1, 4, 1, 31) 722 ANY12(UYVYToUV422Row_Any_AVX2, UYVYToUV422Row_AVX2, 1, 4, 1, 31) 723 #endif 724 #ifdef HAS_YUY2TOUV422ROW_SSE2 725 ANY12(YUY2ToUV422Row_Any_SSE2, YUY2ToUV422Row_SSE2, 1, 4, 1, 15) 726 ANY12(UYVYToUV422Row_Any_SSE2, UYVYToUV422Row_SSE2, 1, 4, 1, 15) 727 #endif 728 #ifdef HAS_YUY2TOUV422ROW_NEON 729 ANY12(ARGBToUV444Row_Any_NEON, ARGBToUV444Row_NEON, 0, 4, 0, 7) 730 ANY12(ARGBToUV411Row_Any_NEON, ARGBToUV411Row_NEON, 0, 4, 2, 31) 731 ANY12(YUY2ToUV422Row_Any_NEON, YUY2ToUV422Row_NEON, 1, 4, 1, 15) 732 ANY12(UYVYToUV422Row_Any_NEON, UYVYToUV422Row_NEON, 1, 4, 1, 15) 733 #endif 734 #undef ANY12 735 736 // Any 1 to 2 with source stride (2 rows of source). Outputs UV planes. 737 // 128 byte row allows for 32 avx ARGB pixels. 738 #define ANY12S(NAMEANY, ANY_SIMD, UVSHIFT, BPP, MASK) \ 739 void NAMEANY(const uint8* src_ptr, int src_stride_ptr, \ 740 uint8* dst_u, uint8* dst_v, int width) { \ 741 SIMD_ALIGNED(uint8 temp[128 * 4]); \ 742 memset(temp, 0, 128 * 2); /* for msan */ \ 743 int r = width & MASK; \ 744 int n = width & ~MASK; \ 745 if (n > 0) { \ 746 ANY_SIMD(src_ptr, src_stride_ptr, dst_u, dst_v, n); \ 747 } \ 748 memcpy(temp, src_ptr + (n >> UVSHIFT) * BPP, SS(r, UVSHIFT) * BPP); \ 749 memcpy(temp + 128, src_ptr + src_stride_ptr + (n >> UVSHIFT) * BPP, \ 750 SS(r, UVSHIFT) * BPP); \ 751 if ((width & 1) && UVSHIFT == 0) { /* repeat last pixel for subsample */\ 752 memcpy(temp + SS(r, UVSHIFT) * BPP, \ 753 temp + SS(r, UVSHIFT) * BPP - BPP, BPP); \ 754 memcpy(temp + 128 + SS(r, UVSHIFT) * BPP, \ 755 temp + 128 + SS(r, UVSHIFT) * BPP - BPP, BPP); \ 756 } \ 757 ANY_SIMD(temp, 128, temp + 256, temp + 384, MASK + 1); \ 758 memcpy(dst_u + (n >> 1), temp + 256, SS(r, 1)); \ 759 memcpy(dst_v + (n >> 1), temp + 384, SS(r, 1)); \ 760 } 761 762 #ifdef HAS_ARGBTOUVROW_AVX2 763 ANY12S(ARGBToUVRow_Any_AVX2, ARGBToUVRow_AVX2, 0, 4, 31) 764 #endif 765 #ifdef HAS_ARGBTOUVJROW_AVX2 766 ANY12S(ARGBToUVJRow_Any_AVX2, ARGBToUVJRow_AVX2, 0, 4, 31) 767 #endif 768 #ifdef HAS_ARGBTOUVROW_SSSE3 769 ANY12S(ARGBToUVRow_Any_SSSE3, ARGBToUVRow_SSSE3, 0, 4, 15) 770 ANY12S(ARGBToUVJRow_Any_SSSE3, ARGBToUVJRow_SSSE3, 0, 4, 15) 771 ANY12S(BGRAToUVRow_Any_SSSE3, BGRAToUVRow_SSSE3, 0, 4, 15) 772 ANY12S(ABGRToUVRow_Any_SSSE3, ABGRToUVRow_SSSE3, 0, 4, 15) 773 ANY12S(RGBAToUVRow_Any_SSSE3, RGBAToUVRow_SSSE3, 0, 4, 15) 774 #endif 775 #ifdef HAS_YUY2TOUVROW_AVX2 776 ANY12S(YUY2ToUVRow_Any_AVX2, YUY2ToUVRow_AVX2, 1, 4, 31) 777 ANY12S(UYVYToUVRow_Any_AVX2, UYVYToUVRow_AVX2, 1, 4, 31) 778 #endif 779 #ifdef HAS_YUY2TOUVROW_SSE2 780 ANY12S(YUY2ToUVRow_Any_SSE2, YUY2ToUVRow_SSE2, 1, 4, 15) 781 ANY12S(UYVYToUVRow_Any_SSE2, UYVYToUVRow_SSE2, 1, 4, 15) 782 #endif 783 #ifdef HAS_ARGBTOUVROW_NEON 784 ANY12S(ARGBToUVRow_Any_NEON, ARGBToUVRow_NEON, 0, 4, 15) 785 #endif 786 #ifdef HAS_ARGBTOUVJROW_NEON 787 ANY12S(ARGBToUVJRow_Any_NEON, ARGBToUVJRow_NEON, 0, 4, 15) 788 #endif 789 #ifdef HAS_BGRATOUVROW_NEON 790 ANY12S(BGRAToUVRow_Any_NEON, BGRAToUVRow_NEON, 0, 4, 15) 791 #endif 792 #ifdef HAS_ABGRTOUVROW_NEON 793 ANY12S(ABGRToUVRow_Any_NEON, ABGRToUVRow_NEON, 0, 4, 15) 794 #endif 795 #ifdef HAS_RGBATOUVROW_NEON 796 ANY12S(RGBAToUVRow_Any_NEON, RGBAToUVRow_NEON, 0, 4, 15) 797 #endif 798 #ifdef HAS_RGB24TOUVROW_NEON 799 ANY12S(RGB24ToUVRow_Any_NEON, RGB24ToUVRow_NEON, 0, 3, 15) 800 #endif 801 #ifdef HAS_RAWTOUVROW_NEON 802 ANY12S(RAWToUVRow_Any_NEON, RAWToUVRow_NEON, 0, 3, 15) 803 #endif 804 #ifdef HAS_RGB565TOUVROW_NEON 805 ANY12S(RGB565ToUVRow_Any_NEON, RGB565ToUVRow_NEON, 0, 2, 15) 806 #endif 807 #ifdef HAS_ARGB1555TOUVROW_NEON 808 ANY12S(ARGB1555ToUVRow_Any_NEON, ARGB1555ToUVRow_NEON, 0, 2, 15) 809 #endif 810 #ifdef HAS_ARGB4444TOUVROW_NEON 811 ANY12S(ARGB4444ToUVRow_Any_NEON, ARGB4444ToUVRow_NEON, 0, 2, 15) 812 #endif 813 #ifdef HAS_YUY2TOUVROW_NEON 814 ANY12S(YUY2ToUVRow_Any_NEON, YUY2ToUVRow_NEON, 1, 4, 15) 815 #endif 816 #ifdef HAS_UYVYTOUVROW_NEON 817 ANY12S(UYVYToUVRow_Any_NEON, UYVYToUVRow_NEON, 1, 4, 15) 818 #endif 819 #undef ANY12S 820 821 #ifdef __cplusplus 822 } // extern "C" 823 } // namespace libyuv 824 #endif 825