libswscale/rgb2rgb_template.c

   1 /*
   2  * software RGB to RGB converter
   3  * pluralize by software PAL8 to RGB converter
   4  *              software YUV to YUV converter
   5  *              software YUV to RGB converter
   6  * Written by Nick Kurshev.
   7  * palette & YUV & runtime CPU stuff by Michael (michaelni@gmx.at)
   8  * lot of big-endian byte order fixes by Alex Beregszaszi
   9  *
  10  * This file is part of FFmpeg.
  11  *
  12  * FFmpeg is free software; you can redistribute it and/or
  13  * modify it under the terms of the GNU Lesser General Public
  14  * License as published by the Free Software Foundation; either
  15  * version 2.1 of the License, or (at your option) any later version.
  16  *
  17  * FFmpeg is distributed in the hope that it will be useful,
  18  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  19  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  20  * Lesser General Public License for more details.
  21  *
  22  * You should have received a copy of the GNU Lesser General Public
  23  * License along with FFmpeg; if not, write to the Free Software
  24  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  25  */
  26
  27 #include <stddef.h>
  28
  29 #include "libavutil/attributes.h"
  30
  31 static inline void rgb24tobgr32_c(const uint8_t *src, uint8_t *dst,
  32                                   int src_size)
  33 {
  34     uint8_t *dest      = dst;
  35     const uint8_t *s   = src;
  36     const uint8_t *end = s + src_size;
  37
  38     while (s < end) {
  39 #if HAVE_BIGENDIAN
  40         /* RGB24 (= R, G, B) -> RGB32 (= A, B, G, R) */
  41         *dest++  = 255;
  42         *dest++  = s[2];
  43         *dest++  = s[1];
  44         *dest++  = s[0];
  45         s       += 3;
  46 #else
  47         *dest++  = *s++;
  48         *dest++  = *s++;
  49         *dest++  = *s++;
  50         *dest++  = 255;
  51 #endif
  52     }
  53 }
  54
  55 static inline void rgb32tobgr24_c(const uint8_t *src, uint8_t *dst,
  56                                   int src_size)
  57 {
  58     uint8_t *dest      = dst;
  59     const uint8_t *s   = src;
  60     const uint8_t *end = s + src_size;
  61
  62     while (s < end) {
  63 #if HAVE_BIGENDIAN
  64         /* RGB32 (= A, B, G, R) -> RGB24 (= R, G, B) */
  65         s++;
  66         dest[2]  = *s++;
  67         dest[1]  = *s++;
  68         dest[0]  = *s++;
  69         dest    += 3;
  70 #else
  71         *dest++  = *s++;
  72         *dest++  = *s++;
  73         *dest++  = *s++;
  74         s++;
  75 #endif
  76     }
  77 }
  78
  79 /*
  80  * original by Strepto/Astral
  81  * ported to gcc & bugfixed: A'rpi
  82  * MMXEXT, 3DNOW optimization by Nick Kurshev
  83  * 32-bit C version, and and&add trick by Michael Niedermayer
  84  */
  85 static inline void rgb15to16_c(const uint8_t *src, uint8_t *dst, int src_size)
  86 {
  87     register uint8_t *d         = dst;
  88     register const uint8_t *s   = src;
  89     register const uint8_t *end = s + src_size;
  90     const uint8_t *mm_end       = end - 3;
  91
  92     while (s < mm_end) {
  93         register unsigned x = *((const uint32_t *)s);
  94         *((uint32_t *)d)    = (x & 0x7FFF7FFF) + (x & 0x7FE07FE0);
  95         d += 4;
  96         s += 4;
  97     }
  98     if (s < end) {
  99         register unsigned short x = *((const uint16_t *)s);
 100         *((uint16_t *)d)          = (x & 0x7FFF) + (x & 0x7FE0);
 101     }
 102 }
 103
 104 static inline void rgb16to15_c(const uint8_t *src, uint8_t *dst, int src_size)
 105 {
 106     register uint8_t *d         = dst;
 107     register const uint8_t *s   = src;
 108     register const uint8_t *end = s + src_size;
 109     const uint8_t *mm_end       = end - 3;
 110
 111     while (s < mm_end) {
 112         register uint32_t x  = *((const uint32_t *)s);
 113         *((uint32_t *)d)     = ((x >> 1) & 0x7FE07FE0) | (x & 0x001F001F);
 114         s                   += 4;
 115         d                   += 4;
 116     }
 117     if (s < end) {
 118         register uint16_t x = *((const uint16_t *)s);
 119         *((uint16_t *)d)    = ((x >> 1) & 0x7FE0) | (x & 0x001F);
 120     }
 121 }
 122
 123 static inline void rgb32to16_c(const uint8_t *src, uint8_t *dst, int src_size)
 124 {
 125     uint16_t *d        = (uint16_t *)dst;
 126     const uint8_t *s   = src;
 127     const uint8_t *end = s + src_size;
 128
 129     while (s < end) {
 130         register int rgb  = *(const uint32_t *)s;
 131         s                += 4;
 132         *d++              = ((rgb & 0xFF)     >> 3) +
 133                             ((rgb & 0xFC00)   >> 5) +
 134                             ((rgb & 0xF80000) >> 8);
 135     }
 136 }
 137
 138 static inline void rgb32tobgr16_c(const uint8_t *src, uint8_t *dst,
 139                                   int src_size)
 140 {
 141     uint16_t *d        = (uint16_t *)dst;
 142     const uint8_t *s   = src;
 143     const uint8_t *end = s + src_size;
 144
 145     while (s < end) {
 146         register int rgb  = *(const uint32_t *)s;
 147         s                += 4;
 148         *d++              = ((rgb & 0xF8)     << 8) +
 149                             ((rgb & 0xFC00)   >> 5) +
 150                             ((rgb & 0xF80000) >> 19);
 151     }
 152 }
 153
 154 static inline void rgb32to15_c(const uint8_t *src, uint8_t *dst, int src_size)
 155 {
 156     uint16_t *d        = (uint16_t *)dst;
 157     const uint8_t *s   = src;
 158     const uint8_t *end = s + src_size;
 159
 160     while (s < end) {
 161         register int rgb  = *(const uint32_t *)s;
 162         s                += 4;
 163         *d++              = ((rgb & 0xFF)     >> 3) +
 164                             ((rgb & 0xF800)   >> 6) +
 165                             ((rgb & 0xF80000) >> 9);
 166     }
 167 }
 168
 169 static inline void rgb32tobgr15_c(const uint8_t *src, uint8_t *dst,
 170                                   int src_size)
 171 {
 172     uint16_t *d        = (uint16_t *)dst;
 173     const uint8_t *s   = src;
 174     const uint8_t *end = s + src_size;
 175
 176     while (s < end) {
 177         register int rgb  = *(const uint32_t *)s;
 178         s                += 4;
 179         *d++              = ((rgb & 0xF8)     <<  7) +
 180                             ((rgb & 0xF800)   >>  6) +
 181                             ((rgb & 0xF80000) >> 19);
 182     }
 183 }
 184
 185 static inline void rgb24tobgr16_c(const uint8_t *src, uint8_t *dst,
 186                                   int src_size)
 187 {
 188     uint16_t *d        = (uint16_t *)dst;
 189     const uint8_t *s   = src;
 190     const uint8_t *end = s + src_size;
 191
 192     while (s < end) {
 193         const int b = *s++;
 194         const int g = *s++;
 195         const int r = *s++;
 196         *d++        = (b >> 3) | ((g & 0xFC) << 3) | ((r & 0xF8) << 8);
 197     }
 198 }
 199
 200 static inline void rgb24to16_c(const uint8_t *src, uint8_t *dst, int src_size)
 201 {
 202     uint16_t *d        = (uint16_t *)dst;
 203     const uint8_t *s   = src;
 204     const uint8_t *end = s + src_size;
 205
 206     while (s < end) {
 207         const int r = *s++;
 208         const int g = *s++;
 209         const int b = *s++;
 210         *d++        = (b >> 3) | ((g & 0xFC) << 3) | ((r & 0xF8) << 8);
 211     }
 212 }
 213
 214 static inline void rgb24tobgr15_c(const uint8_t *src, uint8_t *dst,
 215                                   int src_size)
 216 {
 217     uint16_t *d        = (uint16_t *)dst;
 218     const uint8_t *s   = src;
 219     const uint8_t *end = s + src_size;
 220
 221     while (s < end) {
 222         const int b = *s++;
 223         const int g = *s++;
 224         const int r = *s++;
 225         *d++        = (b >> 3) | ((g & 0xF8) << 2) | ((r & 0xF8) << 7);
 226     }
 227 }
 228
 229 static inline void rgb24to15_c(const uint8_t *src, uint8_t *dst, int src_size)
 230 {
 231     uint16_t *d        = (uint16_t *)dst;
 232     const uint8_t *s   = src;
 233     const uint8_t *end = s + src_size;
 234
 235     while (s < end) {
 236         const int r = *s++;
 237         const int g = *s++;
 238         const int b = *s++;
 239         *d++        = (b >> 3) | ((g & 0xF8) << 2) | ((r & 0xF8) << 7);
 240     }
 241 }
 242
 243 static inline void rgb15tobgr24_c(const uint8_t *src, uint8_t *dst,
 244                                   int src_size)
 245 {
 246     uint8_t *d          = dst;
 247     const uint16_t *s   = (const uint16_t *)src;
 248     const uint16_t *end = s + src_size / 2;
 249
 250     while (s < end) {
 251         register uint16_t bgr = *s++;
 252         *d++ = ((bgr&0x001F)<<3) | ((bgr&0x001F)>> 2);
 253         *d++ = ((bgr&0x03E0)>>2) | ((bgr&0x03E0)>> 7);
 254         *d++ = ((bgr&0x7C00)>>7) | ((bgr&0x7C00)>>12);
 255     }
 256 }
 257
 258 static inline void rgb16tobgr24_c(const uint8_t *src, uint8_t *dst,
 259                                   int src_size)
 260 {
 261     uint8_t *d          = (uint8_t *)dst;
 262     const uint16_t *s   = (const uint16_t *)src;
 263     const uint16_t *end = s + src_size / 2;
 264
 265     while (s < end) {
 266         register uint16_t bgr = *s++;
 267         *d++ = ((bgr&0x001F)<<3) | ((bgr&0x001F)>> 2);
 268         *d++ = ((bgr&0x07E0)>>3) | ((bgr&0x07E0)>> 9);
 269         *d++ = ((bgr&0xF800)>>8) | ((bgr&0xF800)>>13);
 270     }
 271 }
 272
 273 static inline void rgb15to32_c(const uint8_t *src, uint8_t *dst, int src_size)
 274 {
 275     uint8_t *d          = dst;
 276     const uint16_t *s   = (const uint16_t *)src;
 277     const uint16_t *end = s + src_size / 2;
 278
 279     while (s < end) {
 280         register uint16_t bgr = *s++;
 281 #if HAVE_BIGENDIAN
 282         *d++ = 255;
 283         *d++ = ((bgr&0x7C00)>>7) | ((bgr&0x7C00)>>12);
 284         *d++ = ((bgr&0x03E0)>>2) | ((bgr&0x03E0)>> 7);
 285         *d++ = ((bgr&0x001F)<<3) | ((bgr&0x001F)>> 2);
 286 #else
 287         *d++ = ((bgr&0x001F)<<3) | ((bgr&0x001F)>> 2);
 288         *d++ = ((bgr&0x03E0)>>2) | ((bgr&0x03E0)>> 7);
 289         *d++ = ((bgr&0x7C00)>>7) | ((bgr&0x7C00)>>12);
 290         *d++ = 255;
 291 #endif
 292     }
 293 }
 294
 295 static inline void rgb16to32_c(const uint8_t *src, uint8_t *dst, int src_size)
 296 {
 297     uint8_t *d          = dst;
 298     const uint16_t *s   = (const uint16_t *)src;
 299     const uint16_t *end = s + src_size / 2;
 300
 301     while (s < end) {
 302         register uint16_t bgr = *s++;
 303 #if HAVE_BIGENDIAN
 304         *d++ = 255;
 305         *d++ = ((bgr&0xF800)>>8) | ((bgr&0xF800)>>13);
 306         *d++ = ((bgr&0x07E0)>>3) | ((bgr&0x07E0)>> 9);
 307         *d++ = ((bgr&0x001F)<<3) | ((bgr&0x001F)>> 2);
 308 #else
 309         *d++ = ((bgr&0x001F)<<3) | ((bgr&0x001F)>> 2);
 310         *d++ = ((bgr&0x07E0)>>3) | ((bgr&0x07E0)>> 9);
 311         *d++ = ((bgr&0xF800)>>8) | ((bgr&0xF800)>>13);
 312         *d++ = 255;
 313 #endif
 314     }
 315 }
 316
 317 static inline void shuffle_bytes_2103_c(const uint8_t *src, uint8_t *dst,
 318                                         int src_size)
 319 {
 320     int idx          = 15  - src_size;
 321     const uint8_t *s = src - idx;
 322     uint8_t *d       = dst - idx;
 323
 324     for (; idx < 15; idx += 4) {
 325         register unsigned v   = *(const uint32_t *)&s[idx], g = v & 0xff00ff00;
 326         v                    &= 0xff00ff;
 327         *(uint32_t *)&d[idx]  = (v >> 16) + g + (v << 16);
 328     }
 329 }
 330
 331 static inline void shuffle_bytes_0321_c(const uint8_t *src, uint8_t *dst,
 332                                         int src_size)
 333 {
 334     int idx          = 15  - src_size;
 335     const uint8_t *s = src - idx;
 336     uint8_t *d       = dst - idx;
 337
 338     for (; idx < 15; idx += 4) {
 339         register unsigned v   = *(const uint32_t *)&s[idx], g = v & 0x00ff00ff;
 340         v                    &= 0xff00ff00;
 341         *(uint32_t *)&d[idx]  = (v >> 16) + g + (v << 16);
 342     }
 343 }
 344
 345 #define DEFINE_SHUFFLE_BYTES(name, a, b, c, d)                          \
 346 static void shuffle_bytes_##name (const uint8_t *src,                   \
 347                                         uint8_t *dst, int src_size)     \
 348 {                                                                       \
 349     int i;                                                              \
 350                                                                         \
 351     for (i = 0; i < src_size; i += 4) {                                 \
 352         dst[i + 0] = src[i + a];                                        \
 353         dst[i + 1] = src[i + b];                                        \
 354         dst[i + 2] = src[i + c];                                        \
 355         dst[i + 3] = src[i + d];                                        \
 356     }                                                                   \
 357 }
 358
 359 DEFINE_SHUFFLE_BYTES(1230_c, 1, 2, 3, 0)
 360 DEFINE_SHUFFLE_BYTES(3012_c, 3, 0, 1, 2)
 361 DEFINE_SHUFFLE_BYTES(3210_c, 3, 2, 1, 0)
 362 DEFINE_SHUFFLE_BYTES(3102_c, 3, 1, 0, 2)
 363 DEFINE_SHUFFLE_BYTES(2013_c, 2, 0, 1, 3)
 364 DEFINE_SHUFFLE_BYTES(2130_c, 2, 1, 3, 0)
 365 DEFINE_SHUFFLE_BYTES(1203_c, 1, 2, 0, 3)
 366
 367 static inline void rgb24tobgr24_c(const uint8_t *src, uint8_t *dst, int src_size)
 368 {
 369     unsigned i;
 370
 371     for (i = 0; i < src_size; i += 3) {
 372         register uint8_t x = src[i + 2];
 373         dst[i + 1]         = src[i + 1];
 374         dst[i + 2]         = src[i + 0];
 375         dst[i + 0]         = x;
 376     }
 377 }
 378
 379 static inline void yuvPlanartoyuy2_c(const uint8_t *ysrc, const uint8_t *usrc,
 380                                      const uint8_t *vsrc, uint8_t *dst,
 381                                      int width, int height,
 382                                      int lumStride, int chromStride,
 383                                      int dstStride, int vertLumPerChroma)
 384 {
 385     int y, i;
 386     const int chromWidth = width >> 1;
 387
 388     for (y = 0; y < height; y++) {
 389 #if HAVE_FAST_64BIT
 390         uint64_t *ldst = (uint64_t *)dst;
 391         const uint8_t *yc = ysrc, *uc = usrc, *vc = vsrc;
 392         for (i = 0; i < chromWidth; i += 2) {
 393             uint64_t k = yc[0] + (uc[0] << 8) +
 394                          (yc[1] << 16) + ((unsigned) vc[0] << 24);
 395             uint64_t l = yc[2] + (uc[1] << 8) +
 396                          (yc[3] << 16) + ((unsigned) vc[1] << 24);
 397             *ldst++ = k + (l << 32);
 398             yc     += 4;
 399             uc     += 2;
 400             vc     += 2;
 401         }
 402
 403 #else
 404         int *idst = (int32_t *)dst;
 405         const uint8_t *yc = ysrc, *uc = usrc, *vc = vsrc;
 406
 407         for (i = 0; i < chromWidth; i++) {
 408 #if HAVE_BIGENDIAN
 409             *idst++ = (yc[0] << 24) + (uc[0] << 16) +
 410                       (yc[1] <<  8) + (vc[0] <<  0);
 411 #else
 412             *idst++ = yc[0] + (uc[0] << 8) +
 413                       (yc[1] << 16) + (vc[0] << 24);
 414 #endif
 415             yc += 2;
 416             uc++;
 417             vc++;
 418         }
 419 #endif
 420         if ((y & (vertLumPerChroma - 1)) == vertLumPerChroma - 1) {
 421             usrc += chromStride;
 422             vsrc += chromStride;
 423         }
 424         ysrc += lumStride;
 425         dst  += dstStride;
 426     }
 427 }
 428
 429 /**
 430  * Height should be a multiple of 2 and width should be a multiple of 16.
 431  * (If this is a problem for anyone then tell me, and I will fix it.)
 432  */
 433 static inline void yv12toyuy2_c(const uint8_t *ysrc, const uint8_t *usrc,
 434                                 const uint8_t *vsrc, uint8_t *dst,
 435                                 int width, int height, int lumStride,
 436                                 int chromStride, int dstStride)
 437 {
 438     //FIXME interpolate chroma
 439     yuvPlanartoyuy2_c(ysrc, usrc, vsrc, dst, width, height, lumStride,
 440                       chromStride, dstStride, 2);
 441 }
 442
 443 static inline void yuvPlanartouyvy_c(const uint8_t *ysrc, const uint8_t *usrc,
 444                                      const uint8_t *vsrc, uint8_t *dst,
 445                                      int width, int height,
 446                                      int lumStride, int chromStride,
 447                                      int dstStride, int vertLumPerChroma)
 448 {
 449     int y, i;
 450     const int chromWidth = width >> 1;
 451
 452     for (y = 0; y < height; y++) {
 453 #if HAVE_FAST_64BIT
 454         uint64_t *ldst = (uint64_t *)dst;
 455         const uint8_t *yc = ysrc, *uc = usrc, *vc = vsrc;
 456         for (i = 0; i < chromWidth; i += 2) {
 457             uint64_t k = uc[0] + (yc[0] << 8) +
 458                          (vc[0] << 16) + ((unsigned) yc[1] << 24);
 459             uint64_t l = uc[1] + (yc[2] << 8) +
 460                          (vc[1] << 16) + ((unsigned) yc[3] << 24);
 461             *ldst++ = k + (l << 32);
 462             yc     += 4;
 463             uc     += 2;
 464             vc     += 2;
 465         }
 466
 467 #else
 468         int *idst = (int32_t *)dst;
 469         const uint8_t *yc = ysrc, *uc = usrc, *vc = vsrc;
 470
 471         for (i = 0; i < chromWidth; i++) {
 472 #if HAVE_BIGENDIAN
 473             *idst++ = (uc[0] << 24) + (yc[0] << 16) +
 474                       (vc[0] <<  8) + (yc[1] <<  0);
 475 #else
 476             *idst++ = uc[0] + (yc[0] << 8) +
 477                       (vc[0] << 16) + (yc[1] << 24);
 478 #endif
 479             yc += 2;
 480             uc++;
 481             vc++;
 482         }
 483 #endif
 484         if ((y & (vertLumPerChroma - 1)) == vertLumPerChroma - 1) {
 485             usrc += chromStride;
 486             vsrc += chromStride;
 487         }
 488         ysrc += lumStride;
 489         dst  += dstStride;
 490     }
 491 }
 492
 493 /**
 494  * Height should be a multiple of 2 and width should be a multiple of 16
 495  * (If this is a problem for anyone then tell me, and I will fix it.)
 496  */
 497 static inline void yv12touyvy_c(const uint8_t *ysrc, const uint8_t *usrc,
 498                                 const uint8_t *vsrc, uint8_t *dst,
 499                                 int width, int height, int lumStride,
 500                                 int chromStride, int dstStride)
 501 {
 502     //FIXME interpolate chroma
 503     yuvPlanartouyvy_c(ysrc, usrc, vsrc, dst, width, height, lumStride,
 504                       chromStride, dstStride, 2);
 505 }
 506
 507 /**
 508  * Width should be a multiple of 16.
 509  */
 510 static inline void yuv422ptouyvy_c(const uint8_t *ysrc, const uint8_t *usrc,
 511                                    const uint8_t *vsrc, uint8_t *dst,
 512                                    int width, int height, int lumStride,
 513                                    int chromStride, int dstStride)
 514 {
 515     yuvPlanartouyvy_c(ysrc, usrc, vsrc, dst, width, height, lumStride,
 516                       chromStride, dstStride, 1);
 517 }
 518
 519 /**
 520  * Width should be a multiple of 16.
 521  */
 522 static inline void yuv422ptoyuy2_c(const uint8_t *ysrc, const uint8_t *usrc,
 523                                    const uint8_t *vsrc, uint8_t *dst,
 524                                    int width, int height, int lumStride,
 525                                    int chromStride, int dstStride)
 526 {
 527     yuvPlanartoyuy2_c(ysrc, usrc, vsrc, dst, width, height, lumStride,
 528                       chromStride, dstStride, 1);
 529 }
 530
 531 /**
 532  * Height should be a multiple of 2 and width should be a multiple of 16.
 533  * (If this is a problem for anyone then tell me, and I will fix it.)
 534  */
 535 static inline void yuy2toyv12_c(const uint8_t *src, uint8_t *ydst,
 536                                 uint8_t *udst, uint8_t *vdst,
 537                                 int width, int height, int lumStride,
 538                                 int chromStride, int srcStride)
 539 {
 540     int y;
 541     const int chromWidth = width >> 1;
 542
 543     for (y = 0; y < height; y += 2) {
 544         int i;
 545         for (i = 0; i < chromWidth; i++) {
 546             ydst[2 * i + 0] = src[4 * i + 0];
 547             udst[i]         = src[4 * i + 1];
 548             ydst[2 * i + 1] = src[4 * i + 2];
 549             vdst[i]         = src[4 * i + 3];
 550         }
 551         ydst += lumStride;
 552         src  += srcStride;
 553
 554         for (i = 0; i < chromWidth; i++) {
 555             ydst[2 * i + 0] = src[4 * i + 0];
 556             ydst[2 * i + 1] = src[4 * i + 2];
 557         }
 558         udst += chromStride;
 559         vdst += chromStride;
 560         ydst += lumStride;
 561         src  += srcStride;
 562     }
 563 }
 564
 565 static inline void planar2x_c(const uint8_t *src, uint8_t *dst, int srcWidth,
 566                               int srcHeight, int srcStride, int dstStride)
 567 {
 568     int x, y;
 569
 570     dst[0] = src[0];
 571
 572     // first line
 573     for (x = 0; x < srcWidth - 1; x++) {
 574         dst[2 * x + 1] = (3 * src[x] + src[x + 1]) >> 2;
 575         dst[2 * x + 2] = (src[x] + 3 * src[x + 1]) >> 2;
 576     }
 577     dst[2 * srcWidth - 1] = src[srcWidth - 1];
 578
 579     dst += dstStride;
 580
 581     for (y = 1; y < srcHeight; y++) {
 582         const int mmxSize = 1;
 583
 584         dst[0]         = (src[0] * 3 + src[srcStride]) >> 2;
 585         dst[dstStride] = (src[0] + 3 * src[srcStride]) >> 2;
 586
 587         for (x = mmxSize - 1; x < srcWidth - 1; x++) {
 588             dst[2 * x + 1]             = (src[x + 0] * 3 + src[x + srcStride + 1]) >> 2;
 589             dst[2 * x + dstStride + 2] = (src[x + 0] + 3 * src[x + srcStride + 1]) >> 2;
 590             dst[2 * x + dstStride + 1] = (src[x + 1] + 3 * src[x + srcStride])     >> 2;
 591             dst[2 * x + 2]             = (src[x + 1] * 3 + src[x + srcStride])     >> 2;
 592         }
 593         dst[srcWidth * 2 - 1]             = (src[srcWidth - 1] * 3 + src[srcWidth - 1 + srcStride]) >> 2;
 594         dst[srcWidth * 2 - 1 + dstStride] = (src[srcWidth - 1] + 3 * src[srcWidth - 1 + srcStride]) >> 2;
 595
 596         dst += dstStride * 2;
 597         src += srcStride;
 598     }
 599
 600     // last line
 601     dst[0] = src[0];
 602
 603     for (x = 0; x < srcWidth - 1; x++) {
 604         dst[2 * x + 1] = (src[x] * 3 + src[x + 1]) >> 2;
 605         dst[2 * x + 2] = (src[x] + 3 * src[x + 1]) >> 2;
 606     }
 607     dst[2 * srcWidth - 1] = src[srcWidth - 1];
 608 }
 609
 610 /**
 611  * Height should be a multiple of 2 and width should be a multiple of 16.
 612  * (If this is a problem for anyone then tell me, and I will fix it.)
 613  * Chrominance data is only taken from every second line, others are ignored.
 614  * FIXME: Write HQ version.
 615  */
 616 static inline void uyvytoyv12_c(const uint8_t *src, uint8_t *ydst,
 617                                 uint8_t *udst, uint8_t *vdst,
 618                                 int width, int height, int lumStride,
 619                                 int chromStride, int srcStride)
 620 {
 621     int y;
 622     const int chromWidth = width >> 1;
 623
 624     for (y = 0; y < height; y += 2) {
 625         int i;
 626         for (i = 0; i < chromWidth; i++) {
 627             udst[i]         = src[4 * i + 0];
 628             ydst[2 * i + 0] = src[4 * i + 1];
 629             vdst[i]         = src[4 * i + 2];
 630             ydst[2 * i + 1] = src[4 * i + 3];
 631         }
 632         ydst += lumStride;
 633         src  += srcStride;
 634
 635         for (i = 0; i < chromWidth; i++) {
 636             ydst[2 * i + 0] = src[4 * i + 1];
 637             ydst[2 * i + 1] = src[4 * i + 3];
 638         }
 639         udst += chromStride;
 640         vdst += chromStride;
 641         ydst += lumStride;
 642         src  += srcStride;
 643     }
 644 }
 645
 646 /**
 647  * width should be a multiple of 2.
 648  * (If this is a problem for anyone then tell me, and I will fix it.)
 649  */
 650 void ff_rgb24toyv12_c(const uint8_t *src, uint8_t *ydst, uint8_t *udst,
 651                    uint8_t *vdst, int width, int height, int lumStride,
 652                    int chromStride, int srcStride, const int32_t *rgb2yuv)
 653 {
 654     int32_t ry = rgb2yuv[RY_IDX], gy = rgb2yuv[GY_IDX], by = rgb2yuv[BY_IDX];
 655     int32_t ru = rgb2yuv[RU_IDX], gu = rgb2yuv[GU_IDX], bu = rgb2yuv[BU_IDX];
 656     int32_t rv = rgb2yuv[RV_IDX], gv = rgb2yuv[GV_IDX], bv = rgb2yuv[BV_IDX];
 657     int y;
 658     const int chromWidth = width >> 1;
 659     const uint8_t *src1 = src;
 660     const uint8_t *src2 = src1 + srcStride;
 661     uint8_t *ydst1 = ydst;
 662     uint8_t *ydst2 = ydst + lumStride;
 663
 664     for (y = 0; y < height; y += 2) {
 665         int i;
 666         if (y + 1 == height) {
 667             ydst2 = ydst1;
 668             src2  = src1;
 669         }
 670
 671         for (i = 0; i < chromWidth; i++) {
 672             unsigned int b11 = src1[6 * i + 0];
 673             unsigned int g11 = src1[6 * i + 1];
 674             unsigned int r11 = src1[6 * i + 2];
 675             unsigned int b12 = src1[6 * i + 3];
 676             unsigned int g12 = src1[6 * i + 4];
 677             unsigned int r12 = src1[6 * i + 5];
 678             unsigned int b21 = src2[6 * i + 0];
 679             unsigned int g21 = src2[6 * i + 1];
 680             unsigned int r21 = src2[6 * i + 2];
 681             unsigned int b22 = src2[6 * i + 3];
 682             unsigned int g22 = src2[6 * i + 4];
 683             unsigned int r22 = src2[6 * i + 5];
 684
 685             unsigned int Y11 = ((ry * r11 + gy * g11 + by * b11) >> RGB2YUV_SHIFT) + 16;
 686             unsigned int Y12 = ((ry * r12 + gy * g12 + by * b12) >> RGB2YUV_SHIFT) + 16;
 687             unsigned int Y21 = ((ry * r21 + gy * g21 + by * b21) >> RGB2YUV_SHIFT) + 16;
 688             unsigned int Y22 = ((ry * r22 + gy * g22 + by * b22) >> RGB2YUV_SHIFT) + 16;
 689
 690             unsigned int bx = (b11 + b12 + b21 + b22) >> 2;
 691             unsigned int gx = (g11 + g12 + g21 + g22) >> 2;
 692             unsigned int rx = (r11 + r12 + r21 + r22) >> 2;
 693
 694             unsigned int U  = ((ru * rx + gu * gx + bu * bx) >> RGB2YUV_SHIFT) + 128;
 695             unsigned int V  = ((rv * rx + gv * gx + bv * bx) >> RGB2YUV_SHIFT) + 128;
 696
 697             ydst1[2 * i + 0] = Y11;
 698             ydst1[2 * i + 1] = Y12;
 699             ydst2[2 * i + 0] = Y21;
 700             ydst2[2 * i + 1] = Y22;
 701             udst[i]          = U;
 702             vdst[i]          = V;
 703         }
 704         src1  += srcStride * 2;
 705         src2  += srcStride * 2;
 706         ydst1 += lumStride * 2;
 707         ydst2 += lumStride * 2;
 708         udst  += chromStride;
 709         vdst  += chromStride;
 710     }
 711 }
 712
 713 static void interleaveBytes_c(const uint8_t *src1, const uint8_t *src2,
 714                               uint8_t *dest, int width, int height,
 715                               int src1Stride, int src2Stride, int dstStride)
 716 {
 717     int h;
 718
 719     for (h = 0; h < height; h++) {
 720         int w;
 721         for (w = 0; w < width; w++) {
 722             dest[2 * w + 0] = src1[w];
 723             dest[2 * w + 1] = src2[w];
 724         }
 725         dest += dstStride;
 726         src1 += src1Stride;
 727         src2 += src2Stride;
 728     }
 729 }
 730
 731 static void deinterleaveBytes_c(const uint8_t *src, uint8_t *dst1, uint8_t *dst2,
 732                                 int width, int height, int srcStride,
 733                                 int dst1Stride, int dst2Stride)
 734 {
 735     int h;
 736
 737     for (h = 0; h < height; h++) {
 738         int w;
 739         for (w = 0; w < width; w++) {
 740             dst1[w] = src[2 * w + 0];
 741             dst2[w] = src[2 * w + 1];
 742         }
 743         src  += srcStride;
 744         dst1 += dst1Stride;
 745         dst2 += dst2Stride;
 746     }
 747 }
 748
 749 static inline void vu9_to_vu12_c(const uint8_t *src1, const uint8_t *src2,
 750                                  uint8_t *dst1, uint8_t *dst2,
 751                                  int width, int height,
 752                                  int srcStride1, int srcStride2,
 753                                  int dstStride1, int dstStride2)
 754 {
 755     int x, y;
 756     int w = width  / 2;
 757     int h = height / 2;
 758
 759     for (y = 0; y < h; y++) {
 760         const uint8_t *s1 = src1 + srcStride1 * (y >> 1);
 761         uint8_t *d        = dst1 + dstStride1 *  y;
 762         for (x = 0; x < w; x++)
 763             d[2 * x] = d[2 * x + 1] = s1[x];
 764     }
 765     for (y = 0; y < h; y++) {
 766         const uint8_t *s2 = src2 + srcStride2 * (y >> 1);
 767         uint8_t *d        = dst2 + dstStride2 *  y;
 768         for (x = 0; x < w; x++)
 769             d[2 * x] = d[2 * x + 1] = s2[x];
 770     }
 771 }
 772
 773 static inline void yvu9_to_yuy2_c(const uint8_t *src1, const uint8_t *src2,
 774                                   const uint8_t *src3, uint8_t *dst,
 775                                   int width, int height,
 776                                   int srcStride1, int srcStride2,
 777                                   int srcStride3, int dstStride)
 778 {
 779     int x, y;
 780     int w = width / 2;
 781     int h = height;
 782
 783     for (y = 0; y < h; y++) {
 784         const uint8_t *yp = src1 + srcStride1 *  y;
 785         const uint8_t *up = src2 + srcStride2 * (y >> 2);
 786         const uint8_t *vp = src3 + srcStride3 * (y >> 2);
 787         uint8_t *d        = dst  + dstStride  *  y;
 788         for (x = 0; x < w; x++) {
 789             const int x2 = x << 2;
 790             d[8 * x + 0] = yp[x2];
 791             d[8 * x + 1] = up[x];
 792             d[8 * x + 2] = yp[x2 + 1];
 793             d[8 * x + 3] = vp[x];
 794             d[8 * x + 4] = yp[x2 + 2];
 795             d[8 * x + 5] = up[x];
 796             d[8 * x + 6] = yp[x2 + 3];
 797             d[8 * x + 7] = vp[x];
 798         }
 799     }
 800 }
 801
 802 static void extract_even_c(const uint8_t *src, uint8_t *dst, int count)
 803 {
 804     dst   +=  count;
 805     src   +=  count * 2;
 806     count  = -count;
 807     while (count < 0) {
 808         dst[count] = src[2 * count];
 809         count++;
 810     }
 811 }
 812
 813 static void extract_even2_c(const uint8_t *src, uint8_t *dst0, uint8_t *dst1,
 814                             int count)
 815 {
 816     dst0  +=  count;
 817     dst1  +=  count;
 818     src   +=  count * 4;
 819     count  = -count;
 820     while (count < 0) {
 821         dst0[count] = src[4 * count + 0];
 822         dst1[count] = src[4 * count + 2];
 823         count++;
 824     }
 825 }
 826
 827 static void extract_even2avg_c(const uint8_t *src0, const uint8_t *src1,
 828                                uint8_t *dst0, uint8_t *dst1, int count)
 829 {
 830     dst0  +=  count;
 831     dst1  +=  count;
 832     src0  +=  count * 4;
 833     src1  +=  count * 4;
 834     count  = -count;
 835     while (count < 0) {
 836         dst0[count] = (src0[4 * count + 0] + src1[4 * count + 0]) >> 1;
 837         dst1[count] = (src0[4 * count + 2] + src1[4 * count + 2]) >> 1;
 838         count++;
 839     }
 840 }
 841
 842 static void extract_odd2_c(const uint8_t *src, uint8_t *dst0, uint8_t *dst1,
 843                            int count)
 844 {
 845     dst0  +=  count;
 846     dst1  +=  count;
 847     src   +=  count * 4;
 848     count  = -count;
 849     src++;
 850     while (count < 0) {
 851         dst0[count] = src[4 * count + 0];
 852         dst1[count] = src[4 * count + 2];
 853         count++;
 854     }
 855 }
 856
 857 static void extract_odd2avg_c(const uint8_t *src0, const uint8_t *src1,
 858                               uint8_t *dst0, uint8_t *dst1, int count)
 859 {
 860     dst0  +=  count;
 861     dst1  +=  count;
 862     src0  +=  count * 4;
 863     src1  +=  count * 4;
 864     count  = -count;
 865     src0++;
 866     src1++;
 867     while (count < 0) {
 868         dst0[count] = (src0[4 * count + 0] + src1[4 * count + 0]) >> 1;
 869         dst1[count] = (src0[4 * count + 2] + src1[4 * count + 2]) >> 1;
 870         count++;
 871     }
 872 }
 873
 874 static void yuyvtoyuv420_c(uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
 875                            const uint8_t *src, int width, int height,
 876                            int lumStride, int chromStride, int srcStride)
 877 {
 878     int y;
 879     const int chromWidth = AV_CEIL_RSHIFT(width, 1);
 880
 881     for (y = 0; y < height; y++) {
 882         extract_even_c(src, ydst, width);
 883         if (y & 1) {
 884             extract_odd2avg_c(src - srcStride, src, udst, vdst, chromWidth);
 885             udst += chromStride;
 886             vdst += chromStride;
 887         }
 888
 889         src  += srcStride;
 890         ydst += lumStride;
 891     }
 892 }
 893
 894 static void yuyvtoyuv422_c(uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
 895                            const uint8_t *src, int width, int height,
 896                            int lumStride, int chromStride, int srcStride)
 897 {
 898     int y;
 899     const int chromWidth = AV_CEIL_RSHIFT(width, 1);
 900
 901     for (y = 0; y < height; y++) {
 902         extract_even_c(src, ydst, width);
 903         extract_odd2_c(src, udst, vdst, chromWidth);
 904
 905         src  += srcStride;
 906         ydst += lumStride;
 907         udst += chromStride;
 908         vdst += chromStride;
 909     }
 910 }
 911
 912 static void uyvytoyuv420_c(uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
 913                            const uint8_t *src, int width, int height,
 914                            int lumStride, int chromStride, int srcStride)
 915 {
 916     int y;
 917     const int chromWidth = AV_CEIL_RSHIFT(width, 1);
 918
 919     for (y = 0; y < height; y++) {
 920         extract_even_c(src + 1, ydst, width);
 921         if (y & 1) {
 922             extract_even2avg_c(src - srcStride, src, udst, vdst, chromWidth);
 923             udst += chromStride;
 924             vdst += chromStride;
 925         }
 926
 927         src  += srcStride;
 928         ydst += lumStride;
 929     }
 930 }
 931
 932 static void uyvytoyuv422_c(uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
 933                            const uint8_t *src, int width, int height,
 934                            int lumStride, int chromStride, int srcStride)
 935 {
 936     int y;
 937     const int chromWidth = AV_CEIL_RSHIFT(width, 1);
 938
 939     for (y = 0; y < height; y++) {
 940         extract_even_c(src + 1, ydst, width);
 941         extract_even2_c(src, udst, vdst, chromWidth);
 942
 943         src  += srcStride;
 944         ydst += lumStride;
 945         udst += chromStride;
 946         vdst += chromStride;
 947     }
 948 }
 949
 950 static av_cold void rgb2rgb_init_c(void)
 951 {
 952     rgb15to16          = rgb15to16_c;
 953     rgb15tobgr24       = rgb15tobgr24_c;
 954     rgb15to32          = rgb15to32_c;
 955     rgb16tobgr24       = rgb16tobgr24_c;
 956     rgb16to32          = rgb16to32_c;
 957     rgb16to15          = rgb16to15_c;
 958     rgb24tobgr16       = rgb24tobgr16_c;
 959     rgb24tobgr15       = rgb24tobgr15_c;
 960     rgb24tobgr32       = rgb24tobgr32_c;
 961     rgb32to16          = rgb32to16_c;
 962     rgb32to15          = rgb32to15_c;
 963     rgb32tobgr24       = rgb32tobgr24_c;
 964     rgb24to15          = rgb24to15_c;
 965     rgb24to16          = rgb24to16_c;
 966     rgb24tobgr24       = rgb24tobgr24_c;
 967 #if HAVE_BIGENDIAN
 968     shuffle_bytes_0321 = shuffle_bytes_2103_c;
 969     shuffle_bytes_2103 = shuffle_bytes_0321_c;
 970 #else
 971     shuffle_bytes_0321 = shuffle_bytes_0321_c;
 972     shuffle_bytes_2103 = shuffle_bytes_2103_c;
 973 #endif
 974     shuffle_bytes_1230 = shuffle_bytes_1230_c;
 975     shuffle_bytes_3012 = shuffle_bytes_3012_c;
 976     shuffle_bytes_3210 = shuffle_bytes_3210_c;
 977     shuffle_bytes_3102 = shuffle_bytes_3102_c;
 978     shuffle_bytes_2013 = shuffle_bytes_2013_c;
 979     shuffle_bytes_2130 = shuffle_bytes_2130_c;
 980     shuffle_bytes_1203 = shuffle_bytes_1203_c;
 981     rgb32tobgr16       = rgb32tobgr16_c;
 982     rgb32tobgr15       = rgb32tobgr15_c;
 983     yv12toyuy2         = yv12toyuy2_c;
 984     yv12touyvy         = yv12touyvy_c;
 985     yuv422ptoyuy2      = yuv422ptoyuy2_c;
 986     yuv422ptouyvy      = yuv422ptouyvy_c;
 987     yuy2toyv12         = yuy2toyv12_c;
 988     planar2x           = planar2x_c;
 989     ff_rgb24toyv12     = ff_rgb24toyv12_c;
 990     interleaveBytes    = interleaveBytes_c;
 991     deinterleaveBytes  = deinterleaveBytes_c;
 992     vu9_to_vu12        = vu9_to_vu12_c;
 993     yvu9_to_yuy2       = yvu9_to_yuy2_c;
 994
 995     uyvytoyuv420       = uyvytoyuv420_c;
 996     uyvytoyuv422       = uyvytoyuv422_c;
 997     yuyvtoyuv420       = yuyvtoyuv420_c;
 998     yuyvtoyuv422       = yuyvtoyuv422_c;
 999 }