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 Libav.
  11  *
  12  * Libav 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  * Libav 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 Libav; 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 /*
 244  * I use less accurate approximation here by simply left-shifting the input
 245  * value and filling the low order bits with zeroes. This method improves PNG
 246  * compression but this scheme cannot reproduce white exactly, since it does
 247  * not generate an all-ones maximum value; the net effect is to darken the
 248  * image slightly.
 249  *
 250  * The better method should be "left bit replication":
 251  *
 252  *  4 3 2 1 0
 253  *  ---------
 254  *  1 1 0 1 1
 255  *
 256  *  7 6 5 4 3  2 1 0
 257  *  ----------------
 258  *  1 1 0 1 1  1 1 0
 259  *  |=======|  |===|
 260  *      |      leftmost bits repeated to fill open bits
 261  *      |
 262  *  original bits
 263  */
 264 static inline void rgb15tobgr24_c(const uint8_t *src, uint8_t *dst,
 265                                   int src_size)
 266 {
 267     uint8_t *d          = dst;
 268     const uint16_t *s   = (const uint16_t *)src;
 269     const uint16_t *end = s + src_size / 2;
 270
 271     while (s < end) {
 272         register uint16_t bgr = *s++;
 273         *d++ = (bgr & 0x1F)   << 3;
 274         *d++ = (bgr & 0x3E0)  >> 2;
 275         *d++ = (bgr & 0x7C00) >> 7;
 276     }
 277 }
 278
 279 static inline void rgb16tobgr24_c(const uint8_t *src, uint8_t *dst,
 280                                   int src_size)
 281 {
 282     uint8_t *d          = (uint8_t *)dst;
 283     const uint16_t *s   = (const uint16_t *)src;
 284     const uint16_t *end = s + src_size / 2;
 285
 286     while (s < end) {
 287         register uint16_t bgr = *s++;
 288         *d++ = (bgr & 0x1F)   << 3;
 289         *d++ = (bgr & 0x7E0)  >> 3;
 290         *d++ = (bgr & 0xF800) >> 8;
 291     }
 292 }
 293
 294 static inline void rgb15to32_c(const uint8_t *src, uint8_t *dst, int src_size)
 295 {
 296     uint8_t *d          = dst;
 297     const uint16_t *s   = (const uint16_t *)src;
 298     const uint16_t *end = s + src_size / 2;
 299
 300     while (s < end) {
 301         register uint16_t bgr = *s++;
 302 #if HAVE_BIGENDIAN
 303         *d++ = 255;
 304         *d++ = (bgr & 0x7C00) >> 7;
 305         *d++ = (bgr & 0x3E0)  >> 2;
 306         *d++ = (bgr & 0x1F)   << 3;
 307 #else
 308         *d++ = (bgr & 0x1F)   << 3;
 309         *d++ = (bgr & 0x3E0)  >> 2;
 310         *d++ = (bgr & 0x7C00) >> 7;
 311         *d++ = 255;
 312 #endif
 313     }
 314 }
 315
 316 static inline void rgb16to32_c(const uint8_t *src, uint8_t *dst, int src_size)
 317 {
 318     uint8_t *d          = dst;
 319     const uint16_t *s   = (const uint16_t *)src;
 320     const uint16_t *end = s + src_size / 2;
 321
 322     while (s < end) {
 323         register uint16_t bgr = *s++;
 324 #if HAVE_BIGENDIAN
 325         *d++ = 255;
 326         *d++ = (bgr & 0xF800) >> 8;
 327         *d++ = (bgr & 0x7E0)  >> 3;
 328         *d++ = (bgr & 0x1F)   << 3;
 329 #else
 330         *d++ = (bgr & 0x1F)   << 3;
 331         *d++ = (bgr & 0x7E0)  >> 3;
 332         *d++ = (bgr & 0xF800) >> 8;
 333         *d++ = 255;
 334 #endif
 335     }
 336 }
 337
 338 static inline void shuffle_bytes_2103_c(const uint8_t *src, uint8_t *dst,
 339                                         int src_size)
 340 {
 341     int idx          = 15  - src_size;
 342     const uint8_t *s = src - idx;
 343     uint8_t *d       = dst - idx;
 344
 345     for (; idx < 15; idx += 4) {
 346         register int v        = *(const uint32_t *)&s[idx], g = v & 0xff00ff00;
 347         v                    &= 0xff00ff;
 348         *(uint32_t *)&d[idx]  = (v >> 16) + g + (v << 16);
 349     }
 350 }
 351
 352 static inline void rgb24tobgr24_c(const uint8_t *src, uint8_t *dst, int src_size)
 353 {
 354     unsigned i;
 355
 356     for (i = 0; i < src_size; i += 3) {
 357         register uint8_t x = src[i + 2];
 358         dst[i + 1]         = src[i + 1];
 359         dst[i + 2]         = src[i + 0];
 360         dst[i + 0]         = x;
 361     }
 362 }
 363
 364 static inline void yuvPlanartoyuy2_c(const uint8_t *ysrc, const uint8_t *usrc,
 365                                      const uint8_t *vsrc, uint8_t *dst,
 366                                      int width, int height,
 367                                      int lumStride, int chromStride,
 368                                      int dstStride, int vertLumPerChroma)
 369 {
 370     int y, i;
 371     const int chromWidth = width >> 1;
 372
 373     for (y = 0; y < height; y++) {
 374 #if HAVE_FAST_64BIT
 375         uint64_t *ldst = (uint64_t *)dst;
 376         const uint8_t *yc = ysrc, *uc = usrc, *vc = vsrc;
 377         for (i = 0; i < chromWidth; i += 2) {
 378             uint64_t k = yc[0] + (uc[0] << 8) +
 379                          (yc[1] << 16) + ((unsigned) vc[0] << 24);
 380             uint64_t l = yc[2] + (uc[1] << 8) +
 381                          (yc[3] << 16) + ((unsigned) vc[1] << 24);
 382             *ldst++ = k + (l << 32);
 383             yc     += 4;
 384             uc     += 2;
 385             vc     += 2;
 386         }
 387
 388 #else
 389         int *idst = (int32_t *)dst;
 390         const uint8_t *yc = ysrc, *uc = usrc, *vc = vsrc;
 391
 392         for (i = 0; i < chromWidth; i++) {
 393 #if HAVE_BIGENDIAN
 394             *idst++ = (yc[0] << 24) + (uc[0] << 16) +
 395                       (yc[1] <<  8) + (vc[0] <<  0);
 396 #else
 397             *idst++ = yc[0] + (uc[0] << 8) +
 398                       (yc[1] << 16) + (vc[0] << 24);
 399 #endif
 400             yc += 2;
 401             uc++;
 402             vc++;
 403         }
 404 #endif
 405         if ((y & (vertLumPerChroma - 1)) == vertLumPerChroma - 1) {
 406             usrc += chromStride;
 407             vsrc += chromStride;
 408         }
 409         ysrc += lumStride;
 410         dst  += dstStride;
 411     }
 412 }
 413
 414 /**
 415  * Height should be a multiple of 2 and width should be a multiple of 16.
 416  * (If this is a problem for anyone then tell me, and I will fix it.)
 417  */
 418 static inline void yv12toyuy2_c(const uint8_t *ysrc, const uint8_t *usrc,
 419                                 const uint8_t *vsrc, uint8_t *dst,
 420                                 int width, int height, int lumStride,
 421                                 int chromStride, int dstStride)
 422 {
 423     //FIXME interpolate chroma
 424     yuvPlanartoyuy2_c(ysrc, usrc, vsrc, dst, width, height, lumStride,
 425                       chromStride, dstStride, 2);
 426 }
 427
 428 static inline void yuvPlanartouyvy_c(const uint8_t *ysrc, const uint8_t *usrc,
 429                                      const uint8_t *vsrc, uint8_t *dst,
 430                                      int width, int height,
 431                                      int lumStride, int chromStride,
 432                                      int dstStride, int vertLumPerChroma)
 433 {
 434     int y, i;
 435     const int chromWidth = width >> 1;
 436
 437     for (y = 0; y < height; y++) {
 438 #if HAVE_FAST_64BIT
 439         uint64_t *ldst = (uint64_t *)dst;
 440         const uint8_t *yc = ysrc, *uc = usrc, *vc = vsrc;
 441         for (i = 0; i < chromWidth; i += 2) {
 442             uint64_t k = uc[0] + (yc[0] << 8) +
 443                          (vc[0] << 16) + ((unsigned) yc[1] << 24);
 444             uint64_t l = uc[1] + (yc[2] << 8) +
 445                          (vc[1] << 16) + ((unsigned) yc[3] << 24);
 446             *ldst++ = k + (l << 32);
 447             yc     += 4;
 448             uc     += 2;
 449             vc     += 2;
 450         }
 451
 452 #else
 453         int *idst = (int32_t *)dst;
 454         const uint8_t *yc = ysrc, *uc = usrc, *vc = vsrc;
 455
 456         for (i = 0; i < chromWidth; i++) {
 457 #if HAVE_BIGENDIAN
 458             *idst++ = (uc[0] << 24) + (yc[0] << 16) +
 459                       (vc[0] <<  8) + (yc[1] <<  0);
 460 #else
 461             *idst++ = uc[0] + (yc[0] << 8) +
 462                       (vc[0] << 16) + (yc[1] << 24);
 463 #endif
 464             yc += 2;
 465             uc++;
 466             vc++;
 467         }
 468 #endif
 469         if ((y & (vertLumPerChroma - 1)) == vertLumPerChroma - 1) {
 470             usrc += chromStride;
 471             vsrc += chromStride;
 472         }
 473         ysrc += lumStride;
 474         dst  += dstStride;
 475     }
 476 }
 477
 478 /**
 479  * Height should be a multiple of 2 and width should be a multiple of 16
 480  * (If this is a problem for anyone then tell me, and I will fix it.)
 481  */
 482 static inline void yv12touyvy_c(const uint8_t *ysrc, const uint8_t *usrc,
 483                                 const uint8_t *vsrc, uint8_t *dst,
 484                                 int width, int height, int lumStride,
 485                                 int chromStride, int dstStride)
 486 {
 487     //FIXME interpolate chroma
 488     yuvPlanartouyvy_c(ysrc, usrc, vsrc, dst, width, height, lumStride,
 489                       chromStride, dstStride, 2);
 490 }
 491
 492 /**
 493  * Width should be a multiple of 16.
 494  */
 495 static inline void yuv422ptouyvy_c(const uint8_t *ysrc, const uint8_t *usrc,
 496                                    const uint8_t *vsrc, uint8_t *dst,
 497                                    int width, int height, int lumStride,
 498                                    int chromStride, int dstStride)
 499 {
 500     yuvPlanartouyvy_c(ysrc, usrc, vsrc, dst, width, height, lumStride,
 501                       chromStride, dstStride, 1);
 502 }
 503
 504 /**
 505  * Width should be a multiple of 16.
 506  */
 507 static inline void yuv422ptoyuy2_c(const uint8_t *ysrc, const uint8_t *usrc,
 508                                    const uint8_t *vsrc, uint8_t *dst,
 509                                    int width, int height, int lumStride,
 510                                    int chromStride, int dstStride)
 511 {
 512     yuvPlanartoyuy2_c(ysrc, usrc, vsrc, dst, width, height, lumStride,
 513                       chromStride, dstStride, 1);
 514 }
 515
 516 /**
 517  * Height should be a multiple of 2 and width should be a multiple of 16.
 518  * (If this is a problem for anyone then tell me, and I will fix it.)
 519  */
 520 static inline void yuy2toyv12_c(const uint8_t *src, uint8_t *ydst,
 521                                 uint8_t *udst, uint8_t *vdst,
 522                                 int width, int height, int lumStride,
 523                                 int chromStride, int srcStride)
 524 {
 525     int y;
 526     const int chromWidth = width >> 1;
 527
 528     for (y = 0; y < height; y += 2) {
 529         int i;
 530         for (i = 0; i < chromWidth; i++) {
 531             ydst[2 * i + 0] = src[4 * i + 0];
 532             udst[i]         = src[4 * i + 1];
 533             ydst[2 * i + 1] = src[4 * i + 2];
 534             vdst[i]         = src[4 * i + 3];
 535         }
 536         ydst += lumStride;
 537         src  += srcStride;
 538
 539         for (i = 0; i < chromWidth; i++) {
 540             ydst[2 * i + 0] = src[4 * i + 0];
 541             ydst[2 * i + 1] = src[4 * i + 2];
 542         }
 543         udst += chromStride;
 544         vdst += chromStride;
 545         ydst += lumStride;
 546         src  += srcStride;
 547     }
 548 }
 549
 550 static inline void planar2x_c(const uint8_t *src, uint8_t *dst, int srcWidth,
 551                               int srcHeight, int srcStride, int dstStride)
 552 {
 553     int x, y;
 554
 555     dst[0] = src[0];
 556
 557     // first line
 558     for (x = 0; x < srcWidth - 1; x++) {
 559         dst[2 * x + 1] = (3 * src[x] + src[x + 1]) >> 2;
 560         dst[2 * x + 2] = (src[x] + 3 * src[x + 1]) >> 2;
 561     }
 562     dst[2 * srcWidth - 1] = src[srcWidth - 1];
 563
 564     dst += dstStride;
 565
 566     for (y = 1; y < srcHeight; y++) {
 567         const int mmxSize = 1;
 568
 569         dst[0]         = (src[0] * 3 + src[srcStride]) >> 2;
 570         dst[dstStride] = (src[0] + 3 * src[srcStride]) >> 2;
 571
 572         for (x = mmxSize - 1; x < srcWidth - 1; x++) {
 573             dst[2 * x + 1]             = (src[x + 0] * 3 + src[x + srcStride + 1]) >> 2;
 574             dst[2 * x + dstStride + 2] = (src[x + 0] + 3 * src[x + srcStride + 1]) >> 2;
 575             dst[2 * x + dstStride + 1] = (src[x + 1] + 3 * src[x + srcStride])     >> 2;
 576             dst[2 * x + 2]             = (src[x + 1] * 3 + src[x + srcStride])     >> 2;
 577         }
 578         dst[srcWidth * 2 - 1]             = (src[srcWidth - 1] * 3 + src[srcWidth - 1 + srcStride]) >> 2;
 579         dst[srcWidth * 2 - 1 + dstStride] = (src[srcWidth - 1] + 3 * src[srcWidth - 1 + srcStride]) >> 2;
 580
 581         dst += dstStride * 2;
 582         src += srcStride;
 583     }
 584
 585     // last line
 586     dst[0] = src[0];
 587
 588     for (x = 0; x < srcWidth - 1; x++) {
 589         dst[2 * x + 1] = (src[x] * 3 + src[x + 1]) >> 2;
 590         dst[2 * x + 2] = (src[x] + 3 * src[x + 1]) >> 2;
 591     }
 592     dst[2 * srcWidth - 1] = src[srcWidth - 1];
 593 }
 594
 595 /**
 596  * Height should be a multiple of 2 and width should be a multiple of 16.
 597  * (If this is a problem for anyone then tell me, and I will fix it.)
 598  * Chrominance data is only taken from every second line, others are ignored.
 599  * FIXME: Write HQ version.
 600  */
 601 static inline void uyvytoyv12_c(const uint8_t *src, uint8_t *ydst,
 602                                 uint8_t *udst, uint8_t *vdst,
 603                                 int width, int height, int lumStride,
 604                                 int chromStride, int srcStride)
 605 {
 606     int y;
 607     const int chromWidth = width >> 1;
 608
 609     for (y = 0; y < height; y += 2) {
 610         int i;
 611         for (i = 0; i < chromWidth; i++) {
 612             udst[i]         = src[4 * i + 0];
 613             ydst[2 * i + 0] = src[4 * i + 1];
 614             vdst[i]         = src[4 * i + 2];
 615             ydst[2 * i + 1] = src[4 * i + 3];
 616         }
 617         ydst += lumStride;
 618         src  += srcStride;
 619
 620         for (i = 0; i < chromWidth; i++) {
 621             ydst[2 * i + 0] = src[4 * i + 1];
 622             ydst[2 * i + 1] = src[4 * i + 3];
 623         }
 624         udst += chromStride;
 625         vdst += chromStride;
 626         ydst += lumStride;
 627         src  += srcStride;
 628     }
 629 }
 630
 631 /**
 632  * Height should be a multiple of 2 and width should be a multiple of 2.
 633  * (If this is a problem for anyone then tell me, and I will fix it.)
 634  * Chrominance data is only taken from every second line,
 635  * others are ignored in the C version.
 636  * FIXME: Write HQ version.
 637  */
 638 void rgb24toyv12_c(const uint8_t *src, uint8_t *ydst, uint8_t *udst,
 639                    uint8_t *vdst, int width, int height, int lumStride,
 640                    int chromStride, int srcStride)
 641 {
 642     int y;
 643     const int chromWidth = width >> 1;
 644
 645     for (y = 0; y < height; y += 2) {
 646         int i;
 647         for (i = 0; i < chromWidth; i++) {
 648             unsigned int b = src[6 * i + 0];
 649             unsigned int g = src[6 * i + 1];
 650             unsigned int r = src[6 * i + 2];
 651
 652             unsigned int Y = ((RY * r + GY * g + BY * b) >> RGB2YUV_SHIFT) +  16;
 653             unsigned int V = ((RV * r + GV * g + BV * b) >> RGB2YUV_SHIFT) + 128;
 654             unsigned int U = ((RU * r + GU * g + BU * b) >> RGB2YUV_SHIFT) + 128;
 655
 656             udst[i]     = U;
 657             vdst[i]     = V;
 658             ydst[2 * i] = Y;
 659
 660             b = src[6 * i + 3];
 661             g = src[6 * i + 4];
 662             r = src[6 * i + 5];
 663
 664             Y = ((RY * r + GY * g + BY * b) >> RGB2YUV_SHIFT) + 16;
 665             ydst[2 * i + 1] = Y;
 666         }
 667         ydst += lumStride;
 668         src  += srcStride;
 669
 670         for (i = 0; i < chromWidth; i++) {
 671             unsigned int b = src[6 * i + 0];
 672             unsigned int g = src[6 * i + 1];
 673             unsigned int r = src[6 * i + 2];
 674
 675             unsigned int Y = ((RY * r + GY * g + BY * b) >> RGB2YUV_SHIFT) + 16;
 676
 677             ydst[2 * i] = Y;
 678
 679             b = src[6 * i + 3];
 680             g = src[6 * i + 4];
 681             r = src[6 * i + 5];
 682
 683             Y = ((RY * r + GY * g + BY * b) >> RGB2YUV_SHIFT) + 16;
 684             ydst[2 * i + 1] = Y;
 685         }
 686         udst += chromStride;
 687         vdst += chromStride;
 688         ydst += lumStride;
 689         src  += srcStride;
 690     }
 691 }
 692
 693 static void interleaveBytes_c(const uint8_t *src1, const uint8_t *src2,
 694                               uint8_t *dest, int width, int height,
 695                               int src1Stride, int src2Stride, int dstStride)
 696 {
 697     int h;
 698
 699     for (h = 0; h < height; h++) {
 700         int w;
 701         for (w = 0; w < width; w++) {
 702             dest[2 * w + 0] = src1[w];
 703             dest[2 * w + 1] = src2[w];
 704         }
 705         dest += dstStride;
 706         src1 += src1Stride;
 707         src2 += src2Stride;
 708     }
 709 }
 710
 711 static void deinterleaveBytes_c(const uint8_t *src, uint8_t *dst1, uint8_t *dst2,
 712                                 int width, int height, int srcStride,
 713                                 int dst1Stride, int dst2Stride)
 714 {
 715     int h;
 716
 717     for (h = 0; h < height; h++) {
 718         int w;
 719         for (w = 0; w < width; w++) {
 720             dst1[w] = src[2 * w + 0];
 721             dst2[w] = src[2 * w + 1];
 722         }
 723         src  += srcStride;
 724         dst1 += dst1Stride;
 725         dst2 += dst2Stride;
 726     }
 727 }
 728
 729 static inline void vu9_to_vu12_c(const uint8_t *src1, const uint8_t *src2,
 730                                  uint8_t *dst1, uint8_t *dst2,
 731                                  int width, int height,
 732                                  int srcStride1, int srcStride2,
 733                                  int dstStride1, int dstStride2)
 734 {
 735     int x, y;
 736     int w = width  / 2;
 737     int h = height / 2;
 738
 739     for (y = 0; y < h; y++) {
 740         const uint8_t *s1 = src1 + srcStride1 * (y >> 1);
 741         uint8_t *d        = dst1 + dstStride1 *  y;
 742         for (x = 0; x < w; x++)
 743             d[2 * x] = d[2 * x + 1] = s1[x];
 744     }
 745     for (y = 0; y < h; y++) {
 746         const uint8_t *s2 = src2 + srcStride2 * (y >> 1);
 747         uint8_t *d        = dst2 + dstStride2 *  y;
 748         for (x = 0; x < w; x++)
 749             d[2 * x] = d[2 * x + 1] = s2[x];
 750     }
 751 }
 752
 753 static inline void yvu9_to_yuy2_c(const uint8_t *src1, const uint8_t *src2,
 754                                   const uint8_t *src3, uint8_t *dst,
 755                                   int width, int height,
 756                                   int srcStride1, int srcStride2,
 757                                   int srcStride3, int dstStride)
 758 {
 759     int x, y;
 760     int w = width / 2;
 761     int h = height;
 762
 763     for (y = 0; y < h; y++) {
 764         const uint8_t *yp = src1 + srcStride1 *  y;
 765         const uint8_t *up = src2 + srcStride2 * (y >> 2);
 766         const uint8_t *vp = src3 + srcStride3 * (y >> 2);
 767         uint8_t *d        = dst  + dstStride  *  y;
 768         for (x = 0; x < w; x++) {
 769             const int x2 = x << 2;
 770             d[8 * x + 0] = yp[x2];
 771             d[8 * x + 1] = up[x];
 772             d[8 * x + 2] = yp[x2 + 1];
 773             d[8 * x + 3] = vp[x];
 774             d[8 * x + 4] = yp[x2 + 2];
 775             d[8 * x + 5] = up[x];
 776             d[8 * x + 6] = yp[x2 + 3];
 777             d[8 * x + 7] = vp[x];
 778         }
 779     }
 780 }
 781
 782 static void extract_even_c(const uint8_t *src, uint8_t *dst, int count)
 783 {
 784     dst   +=  count;
 785     src   +=  count * 2;
 786     count  = -count;
 787     while (count < 0) {
 788         dst[count] = src[2 * count];
 789         count++;
 790     }
 791 }
 792
 793 static void extract_even2_c(const uint8_t *src, uint8_t *dst0, uint8_t *dst1,
 794                             int count)
 795 {
 796     dst0  +=  count;
 797     dst1  +=  count;
 798     src   +=  count * 4;
 799     count  = -count;
 800     while (count < 0) {
 801         dst0[count] = src[4 * count + 0];
 802         dst1[count] = src[4 * count + 2];
 803         count++;
 804     }
 805 }
 806
 807 static void extract_even2avg_c(const uint8_t *src0, const uint8_t *src1,
 808                                uint8_t *dst0, uint8_t *dst1, int count)
 809 {
 810     dst0  +=  count;
 811     dst1  +=  count;
 812     src0  +=  count * 4;
 813     src1  +=  count * 4;
 814     count  = -count;
 815     while (count < 0) {
 816         dst0[count] = (src0[4 * count + 0] + src1[4 * count + 0]) >> 1;
 817         dst1[count] = (src0[4 * count + 2] + src1[4 * count + 2]) >> 1;
 818         count++;
 819     }
 820 }
 821
 822 static void extract_odd2_c(const uint8_t *src, uint8_t *dst0, uint8_t *dst1,
 823                            int count)
 824 {
 825     dst0  +=  count;
 826     dst1  +=  count;
 827     src   +=  count * 4;
 828     count  = -count;
 829     src++;
 830     while (count < 0) {
 831         dst0[count] = src[4 * count + 0];
 832         dst1[count] = src[4 * count + 2];
 833         count++;
 834     }
 835 }
 836
 837 static void extract_odd2avg_c(const uint8_t *src0, const uint8_t *src1,
 838                               uint8_t *dst0, uint8_t *dst1, int count)
 839 {
 840     dst0  +=  count;
 841     dst1  +=  count;
 842     src0  +=  count * 4;
 843     src1  +=  count * 4;
 844     count  = -count;
 845     src0++;
 846     src1++;
 847     while (count < 0) {
 848         dst0[count] = (src0[4 * count + 0] + src1[4 * count + 0]) >> 1;
 849         dst1[count] = (src0[4 * count + 2] + src1[4 * count + 2]) >> 1;
 850         count++;
 851     }
 852 }
 853
 854 static void yuyvtoyuv420_c(uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
 855                            const uint8_t *src, int width, int height,
 856                            int lumStride, int chromStride, int srcStride)
 857 {
 858     int y;
 859     const int chromWidth = AV_CEIL_RSHIFT(width, 1);
 860
 861     for (y = 0; y < height; y++) {
 862         extract_even_c(src, ydst, width);
 863         if (y & 1) {
 864             extract_odd2avg_c(src - srcStride, src, udst, vdst, chromWidth);
 865             udst += chromStride;
 866             vdst += chromStride;
 867         }
 868
 869         src  += srcStride;
 870         ydst += lumStride;
 871     }
 872 }
 873
 874 static void yuyvtoyuv422_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         extract_odd2_c(src, udst, vdst, chromWidth);
 884
 885         src  += srcStride;
 886         ydst += lumStride;
 887         udst += chromStride;
 888         vdst += chromStride;
 889     }
 890 }
 891
 892 static void uyvytoyuv420_c(uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
 893                            const uint8_t *src, int width, int height,
 894                            int lumStride, int chromStride, int srcStride)
 895 {
 896     int y;
 897     const int chromWidth = AV_CEIL_RSHIFT(width, 1);
 898
 899     for (y = 0; y < height; y++) {
 900         extract_even_c(src + 1, ydst, width);
 901         if (y & 1) {
 902             extract_even2avg_c(src - srcStride, src, udst, vdst, chromWidth);
 903             udst += chromStride;
 904             vdst += chromStride;
 905         }
 906
 907         src  += srcStride;
 908         ydst += lumStride;
 909     }
 910 }
 911
 912 static void uyvytoyuv422_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         extract_even2_c(src, udst, vdst, chromWidth);
 922
 923         src  += srcStride;
 924         ydst += lumStride;
 925         udst += chromStride;
 926         vdst += chromStride;
 927     }
 928 }
 929
 930 static av_cold void rgb2rgb_init_c(void)
 931 {
 932     rgb15to16          = rgb15to16_c;
 933     rgb15tobgr24       = rgb15tobgr24_c;
 934     rgb15to32          = rgb15to32_c;
 935     rgb16tobgr24       = rgb16tobgr24_c;
 936     rgb16to32          = rgb16to32_c;
 937     rgb16to15          = rgb16to15_c;
 938     rgb24tobgr16       = rgb24tobgr16_c;
 939     rgb24tobgr15       = rgb24tobgr15_c;
 940     rgb24tobgr32       = rgb24tobgr32_c;
 941     rgb32to16          = rgb32to16_c;
 942     rgb32to15          = rgb32to15_c;
 943     rgb32tobgr24       = rgb32tobgr24_c;
 944     rgb24to15          = rgb24to15_c;
 945     rgb24to16          = rgb24to16_c;
 946     rgb24tobgr24       = rgb24tobgr24_c;
 947     shuffle_bytes_2103 = shuffle_bytes_2103_c;
 948     rgb32tobgr16       = rgb32tobgr16_c;
 949     rgb32tobgr15       = rgb32tobgr15_c;
 950     yv12toyuy2         = yv12toyuy2_c;
 951     yv12touyvy         = yv12touyvy_c;
 952     yuv422ptoyuy2      = yuv422ptoyuy2_c;
 953     yuv422ptouyvy      = yuv422ptouyvy_c;
 954     yuy2toyv12         = yuy2toyv12_c;
 955     planar2x           = planar2x_c;
 956     rgb24toyv12        = rgb24toyv12_c;
 957     interleaveBytes    = interleaveBytes_c;
 958     deinterleaveBytes  = deinterleaveBytes_c;
 959     vu9_to_vu12        = vu9_to_vu12_c;
 960     yvu9_to_yuy2       = yvu9_to_yuy2_c;
 961
 962     uyvytoyuv420       = uyvytoyuv420_c;
 963     uyvytoyuv422       = uyvytoyuv422_c;
 964     yuyvtoyuv420       = yuyvtoyuv420_c;
 965     yuyvtoyuv422       = yuyvtoyuv422_c;
 966 }