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 static inline void rgb24tobgr32_c(const uint8_t *src, uint8_t *dst,
  30                                   int src_size)
  31 {
  32     uint8_t *dest      = dst;
  33     const uint8_t *s   = src;
  34     const uint8_t *end = s + src_size;
  35
  36     while (s < end) {
  37 #if HAVE_BIGENDIAN
  38         /* RGB24 (= R, G, B) -> RGB32 (= A, B, G, R) */
  39         *dest++  = 255;
  40         *dest++  = s[2];
  41         *dest++  = s[1];
  42         *dest++  = s[0];
  43         s       += 3;
  44 #else
  45         *dest++  = *s++;
  46         *dest++  = *s++;
  47         *dest++  = *s++;
  48         *dest++  = 255;
  49 #endif
  50     }
  51 }
  52
  53 static inline void rgb32tobgr24_c(const uint8_t *src, uint8_t *dst,
  54                                   int src_size)
  55 {
  56     uint8_t *dest      = dst;
  57     const uint8_t *s   = src;
  58     const uint8_t *end = s + src_size;
  59
  60     while (s < end) {
  61 #if HAVE_BIGENDIAN
  62         /* RGB32 (= A, B, G, R) -> RGB24 (= R, G, B) */
  63         s++;
  64         dest[2]  = *s++;
  65         dest[1]  = *s++;
  66         dest[0]  = *s++;
  67         dest    += 3;
  68 #else
  69         *dest++  = *s++;
  70         *dest++  = *s++;
  71         *dest++  = *s++;
  72         s++;
  73 #endif
  74     }
  75 }
  76
  77 /*
  78  * original by Strepto/Astral
  79  * ported to gcc & bugfixed: A'rpi
  80  * MMXEXT, 3DNOW optimization by Nick Kurshev
  81  * 32-bit C version, and and&add trick by Michael Niedermayer
  82  */
  83 static inline void rgb15to16_c(const uint8_t *src, uint8_t *dst, int src_size)
  84 {
  85     register uint8_t *d         = dst;
  86     register const uint8_t *s   = src;
  87     register const uint8_t *end = s + src_size;
  88     const uint8_t *mm_end       = end - 3;
  89
  90     while (s < mm_end) {
  91         register unsigned x = *((const uint32_t *)s);
  92         *((uint32_t *)d)    = (x & 0x7FFF7FFF) + (x & 0x7FE07FE0);
  93         d += 4;
  94         s += 4;
  95     }
  96     if (s < end) {
  97         register unsigned short x = *((const uint16_t *)s);
  98         *((uint16_t *)d)          = (x & 0x7FFF) + (x & 0x7FE0);
  99     }
 100 }
 101
 102 static inline void rgb16to15_c(const uint8_t *src, uint8_t *dst, int src_size)
 103 {
 104     register uint8_t *d         = dst;
 105     register const uint8_t *s   = src;
 106     register const uint8_t *end = s + src_size;
 107     const uint8_t *mm_end       = end - 3;
 108
 109     while (s < mm_end) {
 110         register uint32_t x  = *((const uint32_t *)s);
 111         *((uint32_t *)d)     = ((x >> 1) & 0x7FE07FE0) | (x & 0x001F001F);
 112         s                   += 4;
 113         d                   += 4;
 114     }
 115     if (s < end) {
 116         register uint16_t x = *((const uint16_t *)s);
 117         *((uint16_t *)d)    = ((x >> 1) & 0x7FE0) | (x & 0x001F);
 118     }
 119 }
 120
 121 static inline void rgb32to16_c(const uint8_t *src, uint8_t *dst, int src_size)
 122 {
 123     uint16_t *d        = (uint16_t *)dst;
 124     const uint8_t *s   = src;
 125     const uint8_t *end = s + src_size;
 126
 127     while (s < end) {
 128         register int rgb  = *(const uint32_t *)s;
 129         s                += 4;
 130         *d++              = ((rgb & 0xFF)     >> 3) +
 131                             ((rgb & 0xFC00)   >> 5) +
 132                             ((rgb & 0xF80000) >> 8);
 133     }
 134 }
 135
 136 static inline void rgb32tobgr16_c(const uint8_t *src, uint8_t *dst,
 137                                   int src_size)
 138 {
 139     uint16_t *d        = (uint16_t *)dst;
 140     const uint8_t *s   = src;
 141     const uint8_t *end = s + src_size;
 142
 143     while (s < end) {
 144         register int rgb  = *(const uint32_t *)s;
 145         s                += 4;
 146         *d++              = ((rgb & 0xF8)     << 8) +
 147                             ((rgb & 0xFC00)   >> 5) +
 148                             ((rgb & 0xF80000) >> 19);
 149     }
 150 }
 151
 152 static inline void rgb32to15_c(const uint8_t *src, uint8_t *dst, int src_size)
 153 {
 154     uint16_t *d        = (uint16_t *)dst;
 155     const uint8_t *s   = src;
 156     const uint8_t *end = s + src_size;
 157
 158     while (s < end) {
 159         register int rgb  = *(const uint32_t *)s;
 160         s                += 4;
 161         *d++              = ((rgb & 0xFF)     >> 3) +
 162                             ((rgb & 0xF800)   >> 6) +
 163                             ((rgb & 0xF80000) >> 9);
 164     }
 165 }
 166
 167 static inline void rgb32tobgr15_c(const uint8_t *src, uint8_t *dst,
 168                                   int src_size)
 169 {
 170     uint16_t *d        = (uint16_t *)dst;
 171     const uint8_t *s   = src;
 172     const uint8_t *end = s + src_size;
 173
 174     while (s < end) {
 175         register int rgb  = *(const uint32_t *)s;
 176         s                += 4;
 177         *d++              = ((rgb & 0xF8)     <<  7) +
 178                             ((rgb & 0xF800)   >>  6) +
 179                             ((rgb & 0xF80000) >> 19);
 180     }
 181 }
 182
 183 static inline void rgb24tobgr16_c(const uint8_t *src, uint8_t *dst,
 184                                   int src_size)
 185 {
 186     uint16_t *d        = (uint16_t *)dst;
 187     const uint8_t *s   = src;
 188     const uint8_t *end = s + src_size;
 189
 190     while (s < end) {
 191         const int b = *s++;
 192         const int g = *s++;
 193         const int r = *s++;
 194         *d++        = (b >> 3) | ((g & 0xFC) << 3) | ((r & 0xF8) << 8);
 195     }
 196 }
 197
 198 static inline void rgb24to16_c(const uint8_t *src, uint8_t *dst, int src_size)
 199 {
 200     uint16_t *d        = (uint16_t *)dst;
 201     const uint8_t *s   = src;
 202     const uint8_t *end = s + src_size;
 203
 204     while (s < end) {
 205         const int r = *s++;
 206         const int g = *s++;
 207         const int b = *s++;
 208         *d++        = (b >> 3) | ((g & 0xFC) << 3) | ((r & 0xF8) << 8);
 209     }
 210 }
 211
 212 static inline void rgb24tobgr15_c(const uint8_t *src, uint8_t *dst,
 213                                   int src_size)
 214 {
 215     uint16_t *d        = (uint16_t *)dst;
 216     const uint8_t *s   = src;
 217     const uint8_t *end = s + src_size;
 218
 219     while (s < end) {
 220         const int b = *s++;
 221         const int g = *s++;
 222         const int r = *s++;
 223         *d++        = (b >> 3) | ((g & 0xF8) << 2) | ((r & 0xF8) << 7);
 224     }
 225 }
 226
 227 static inline void rgb24to15_c(const uint8_t *src, uint8_t *dst, int src_size)
 228 {
 229     uint16_t *d        = (uint16_t *)dst;
 230     const uint8_t *s   = src;
 231     const uint8_t *end = s + src_size;
 232
 233     while (s < end) {
 234         const int r = *s++;
 235         const int g = *s++;
 236         const int b = *s++;
 237         *d++        = (b >> 3) | ((g & 0xF8) << 2) | ((r & 0xF8) << 7);
 238     }
 239 }
 240
 241 /*
 242  * I use less accurate approximation here by simply left-shifting the input
 243  * value and filling the low order bits with zeroes. This method improves PNG
 244  * compression but this scheme cannot reproduce white exactly, since it does
 245  * not generate an all-ones maximum value; the net effect is to darken the
 246  * image slightly.
 247  *
 248  * The better method should be "left bit replication":
 249  *
 250  *  4 3 2 1 0
 251  *  ---------
 252  *  1 1 0 1 1
 253  *
 254  *  7 6 5 4 3  2 1 0
 255  *  ----------------
 256  *  1 1 0 1 1  1 1 0
 257  *  |=======|  |===|
 258  *      |      leftmost bits repeated to fill open bits
 259  *      |
 260  *  original bits
 261  */
 262 static inline void rgb15tobgr24_c(const uint8_t *src, uint8_t *dst,
 263                                   int src_size)
 264 {
 265     uint8_t *d          = dst;
 266     const uint16_t *s   = (const uint16_t *)src;
 267     const uint16_t *end = s + src_size / 2;
 268
 269     while (s < end) {
 270         register uint16_t bgr = *s++;
 271         *d++ = (bgr & 0x1F)   << 3;
 272         *d++ = (bgr & 0x3E0)  >> 2;
 273         *d++ = (bgr & 0x7C00) >> 7;
 274     }
 275 }
 276
 277 static inline void rgb16tobgr24_c(const uint8_t *src, uint8_t *dst,
 278                                   int src_size)
 279 {
 280     uint8_t *d          = (uint8_t *)dst;
 281     const uint16_t *s   = (const uint16_t *)src;
 282     const uint16_t *end = s + src_size / 2;
 283
 284     while (s < end) {
 285         register uint16_t bgr = *s++;
 286         *d++ = (bgr & 0x1F)   << 3;
 287         *d++ = (bgr & 0x7E0)  >> 3;
 288         *d++ = (bgr & 0xF800) >> 8;
 289     }
 290 }
 291
 292 static inline void rgb15to32_c(const uint8_t *src, uint8_t *dst, int src_size)
 293 {
 294     uint8_t *d          = dst;
 295     const uint16_t *s   = (const uint16_t *)src;
 296     const uint16_t *end = s + src_size / 2;
 297
 298     while (s < end) {
 299         register uint16_t bgr = *s++;
 300 #if HAVE_BIGENDIAN
 301         *d++ = 255;
 302         *d++ = (bgr & 0x7C00) >> 7;
 303         *d++ = (bgr & 0x3E0)  >> 2;
 304         *d++ = (bgr & 0x1F)   << 3;
 305 #else
 306         *d++ = (bgr & 0x1F)   << 3;
 307         *d++ = (bgr & 0x3E0)  >> 2;
 308         *d++ = (bgr & 0x7C00) >> 7;
 309         *d++ = 255;
 310 #endif
 311     }
 312 }
 313
 314 static inline void rgb16to32_c(const uint8_t *src, uint8_t *dst, int src_size)
 315 {
 316     uint8_t *d          = dst;
 317     const uint16_t *s   = (const uint16_t *)src;
 318     const uint16_t *end = s + src_size / 2;
 319
 320     while (s < end) {
 321         register uint16_t bgr = *s++;
 322 #if HAVE_BIGENDIAN
 323         *d++ = 255;
 324         *d++ = (bgr & 0xF800) >> 8;
 325         *d++ = (bgr & 0x7E0)  >> 3;
 326         *d++ = (bgr & 0x1F)   << 3;
 327 #else
 328         *d++ = (bgr & 0x1F)   << 3;
 329         *d++ = (bgr & 0x7E0)  >> 3;
 330         *d++ = (bgr & 0xF800) >> 8;
 331         *d++ = 255;
 332 #endif
 333     }
 334 }
 335
 336 static inline void shuffle_bytes_2103_c(const uint8_t *src, uint8_t *dst,
 337                                         int src_size)
 338 {
 339     int idx          = 15  - src_size;
 340     const uint8_t *s = src - idx;
 341     uint8_t *d       = dst - idx;
 342
 343     for (; idx < 15; idx += 4) {
 344         register int v        = *(const uint32_t *)&s[idx], g = v & 0xff00ff00;
 345         v                    &= 0xff00ff;
 346         *(uint32_t *)&d[idx]  = (v >> 16) + g + (v << 16);
 347     }
 348 }
 349
 350 static inline void rgb24tobgr24_c(const uint8_t *src, uint8_t *dst, int src_size)
 351 {
 352     unsigned i;
 353
 354     for (i = 0; i < src_size; i += 3) {
 355         register uint8_t x = src[i + 2];
 356         dst[i + 1]         = src[i + 1];
 357         dst[i + 2]         = src[i + 0];
 358         dst[i + 0]         = x;
 359     }
 360 }
 361
 362 static inline void yuvPlanartoyuy2_c(const uint8_t *ysrc, const uint8_t *usrc,
 363                                      const uint8_t *vsrc, uint8_t *dst,
 364                                      int width, int height,
 365                                      int lumStride, int chromStride,
 366                                      int dstStride, int vertLumPerChroma)
 367 {
 368     int y, i;
 369     const int chromWidth = width >> 1;
 370
 371     for (y = 0; y < height; y++) {
 372 #if HAVE_FAST_64BIT
 373         uint64_t *ldst = (uint64_t *)dst;
 374         const uint8_t *yc = ysrc, *uc = usrc, *vc = vsrc;
 375         for (i = 0; i < chromWidth; i += 2) {
 376             uint64_t k = yc[0] + (uc[0] << 8) +
 377                          (yc[1] << 16) + (vc[0] << 24);
 378             uint64_t l = yc[2] + (uc[1] << 8) +
 379                          (yc[3] << 16) + (vc[1] << 24);
 380             *ldst++ = k + (l << 32);
 381             yc     += 4;
 382             uc     += 2;
 383             vc     += 2;
 384         }
 385
 386 #else
 387         int *idst = (int32_t *)dst;
 388         const uint8_t *yc = ysrc, *uc = usrc, *vc = vsrc;
 389
 390         for (i = 0; i < chromWidth; i++) {
 391 #if HAVE_BIGENDIAN
 392             *idst++ = (yc[0] << 24) + (uc[0] << 16) +
 393                       (yc[1] <<  8) + (vc[0] <<  0);
 394 #else
 395             *idst++ = yc[0] + (uc[0] << 8) +
 396                       (yc[1] << 16) + (vc[0] << 24);
 397 #endif
 398             yc += 2;
 399             uc++;
 400             vc++;
 401         }
 402 #endif
 403         if ((y & (vertLumPerChroma - 1)) == vertLumPerChroma - 1) {
 404             usrc += chromStride;
 405             vsrc += chromStride;
 406         }
 407         ysrc += lumStride;
 408         dst  += dstStride;
 409     }
 410 }
 411
 412 /**
 413  * Height should be a multiple of 2 and width should be a multiple of 16.
 414  * (If this is a problem for anyone then tell me, and I will fix it.)
 415  */
 416 static inline void yv12toyuy2_c(const uint8_t *ysrc, const uint8_t *usrc,
 417                                 const uint8_t *vsrc, uint8_t *dst,
 418                                 int width, int height, int lumStride,
 419                                 int chromStride, int dstStride)
 420 {
 421     //FIXME interpolate chroma
 422     yuvPlanartoyuy2_c(ysrc, usrc, vsrc, dst, width, height, lumStride,
 423                       chromStride, dstStride, 2);
 424 }
 425
 426 static inline void yuvPlanartouyvy_c(const uint8_t *ysrc, const uint8_t *usrc,
 427                                      const uint8_t *vsrc, uint8_t *dst,
 428                                      int width, int height,
 429                                      int lumStride, int chromStride,
 430                                      int dstStride, int vertLumPerChroma)
 431 {
 432     int y, i;
 433     const int chromWidth = width >> 1;
 434
 435     for (y = 0; y < height; y++) {
 436 #if HAVE_FAST_64BIT
 437         uint64_t *ldst = (uint64_t *)dst;
 438         const uint8_t *yc = ysrc, *uc = usrc, *vc = vsrc;
 439         for (i = 0; i < chromWidth; i += 2) {
 440             uint64_t k = uc[0] + (yc[0] << 8) +
 441                          (vc[0] << 16) + (yc[1] << 24);
 442             uint64_t l = uc[1] + (yc[2] << 8) +
 443                          (vc[1] << 16) + (yc[3] << 24);
 444             *ldst++ = k + (l << 32);
 445             yc     += 4;
 446             uc     += 2;
 447             vc     += 2;
 448         }
 449
 450 #else
 451         int *idst = (int32_t *)dst;
 452         const uint8_t *yc = ysrc, *uc = usrc, *vc = vsrc;
 453
 454         for (i = 0; i < chromWidth; i++) {
 455 #if HAVE_BIGENDIAN
 456             *idst++ = (uc[0] << 24) + (yc[0] << 16) +
 457                       (vc[0] <<  8) + (yc[1] <<  0);
 458 #else
 459             *idst++ = uc[0] + (yc[0] << 8) +
 460                       (vc[0] << 16) + (yc[1] << 24);
 461 #endif
 462             yc += 2;
 463             uc++;
 464             vc++;
 465         }
 466 #endif
 467         if ((y & (vertLumPerChroma - 1)) == vertLumPerChroma - 1) {
 468             usrc += chromStride;
 469             vsrc += chromStride;
 470         }
 471         ysrc += lumStride;
 472         dst  += dstStride;
 473     }
 474 }
 475
 476 /**
 477  * Height should be a multiple of 2 and width should be a multiple of 16
 478  * (If this is a problem for anyone then tell me, and I will fix it.)
 479  */
 480 static inline void yv12touyvy_c(const uint8_t *ysrc, const uint8_t *usrc,
 481                                 const uint8_t *vsrc, uint8_t *dst,
 482                                 int width, int height, int lumStride,
 483                                 int chromStride, int dstStride)
 484 {
 485     //FIXME interpolate chroma
 486     yuvPlanartouyvy_c(ysrc, usrc, vsrc, dst, width, height, lumStride,
 487                       chromStride, dstStride, 2);
 488 }
 489
 490 /**
 491  * Width should be a multiple of 16.
 492  */
 493 static inline void yuv422ptouyvy_c(const uint8_t *ysrc, const uint8_t *usrc,
 494                                    const uint8_t *vsrc, uint8_t *dst,
 495                                    int width, int height, int lumStride,
 496                                    int chromStride, int dstStride)
 497 {
 498     yuvPlanartouyvy_c(ysrc, usrc, vsrc, dst, width, height, lumStride,
 499                       chromStride, dstStride, 1);
 500 }
 501
 502 /**
 503  * Width should be a multiple of 16.
 504  */
 505 static inline void yuv422ptoyuy2_c(const uint8_t *ysrc, const uint8_t *usrc,
 506                                    const uint8_t *vsrc, uint8_t *dst,
 507                                    int width, int height, int lumStride,
 508                                    int chromStride, int dstStride)
 509 {
 510     yuvPlanartoyuy2_c(ysrc, usrc, vsrc, dst, width, height, lumStride,
 511                       chromStride, dstStride, 1);
 512 }
 513
 514 /**
 515  * Height should be a multiple of 2 and width should be a multiple of 16.
 516  * (If this is a problem for anyone then tell me, and I will fix it.)
 517  */
 518 static inline void yuy2toyv12_c(const uint8_t *src, uint8_t *ydst,
 519                                 uint8_t *udst, uint8_t *vdst,
 520                                 int width, int height, int lumStride,
 521                                 int chromStride, int srcStride)
 522 {
 523     int y;
 524     const int chromWidth = width >> 1;
 525
 526     for (y = 0; y < height; y += 2) {
 527         int i;
 528         for (i = 0; i < chromWidth; i++) {
 529             ydst[2 * i + 0] = src[4 * i + 0];
 530             udst[i]         = src[4 * i + 1];
 531             ydst[2 * i + 1] = src[4 * i + 2];
 532             vdst[i]         = src[4 * i + 3];
 533         }
 534         ydst += lumStride;
 535         src  += srcStride;
 536
 537         for (i = 0; i < chromWidth; i++) {
 538             ydst[2 * i + 0] = src[4 * i + 0];
 539             ydst[2 * i + 1] = src[4 * i + 2];
 540         }
 541         udst += chromStride;
 542         vdst += chromStride;
 543         ydst += lumStride;
 544         src  += srcStride;
 545     }
 546 }
 547
 548 static inline void planar2x_c(const uint8_t *src, uint8_t *dst, int srcWidth,
 549                               int srcHeight, int srcStride, int dstStride)
 550 {
 551     int x, y;
 552
 553     dst[0] = src[0];
 554
 555     // first line
 556     for (x = 0; x < srcWidth - 1; x++) {
 557         dst[2 * x + 1] = (3 * src[x] + src[x + 1]) >> 2;
 558         dst[2 * x + 2] = (src[x] + 3 * src[x + 1]) >> 2;
 559     }
 560     dst[2 * srcWidth - 1] = src[srcWidth - 1];
 561
 562     dst += dstStride;
 563
 564     for (y = 1; y < srcHeight; y++) {
 565         const int mmxSize = 1;
 566
 567         dst[0]         = (src[0] * 3 + src[srcStride]) >> 2;
 568         dst[dstStride] = (src[0] + 3 * src[srcStride]) >> 2;
 569
 570         for (x = mmxSize - 1; x < srcWidth - 1; x++) {
 571             dst[2 * x + 1]             = (src[x + 0] * 3 + src[x + srcStride + 1]) >> 2;
 572             dst[2 * x + dstStride + 2] = (src[x + 0] + 3 * src[x + srcStride + 1]) >> 2;
 573             dst[2 * x + dstStride + 1] = (src[x + 1] + 3 * src[x + srcStride])     >> 2;
 574             dst[2 * x + 2]             = (src[x + 1] * 3 + src[x + srcStride])     >> 2;
 575         }
 576         dst[srcWidth * 2 - 1]             = (src[srcWidth - 1] * 3 + src[srcWidth - 1 + srcStride]) >> 2;
 577         dst[srcWidth * 2 - 1 + dstStride] = (src[srcWidth - 1] + 3 * src[srcWidth - 1 + srcStride]) >> 2;
 578
 579         dst += dstStride * 2;
 580         src += srcStride;
 581     }
 582
 583     // last line
 584     dst[0] = src[0];
 585
 586     for (x = 0; x < srcWidth - 1; x++) {
 587         dst[2 * x + 1] = (src[x] * 3 + src[x + 1]) >> 2;
 588         dst[2 * x + 2] = (src[x] + 3 * src[x + 1]) >> 2;
 589     }
 590     dst[2 * srcWidth - 1] = src[srcWidth - 1];
 591 }
 592
 593 /**
 594  * Height should be a multiple of 2 and width should be a multiple of 16.
 595  * (If this is a problem for anyone then tell me, and I will fix it.)
 596  * Chrominance data is only taken from every second line, others are ignored.
 597  * FIXME: Write HQ version.
 598  */
 599 static inline void uyvytoyv12_c(const uint8_t *src, uint8_t *ydst,
 600                                 uint8_t *udst, uint8_t *vdst,
 601                                 int width, int height, int lumStride,
 602                                 int chromStride, int srcStride)
 603 {
 604     int y;
 605     const int chromWidth = width >> 1;
 606
 607     for (y = 0; y < height; y += 2) {
 608         int i;
 609         for (i = 0; i < chromWidth; i++) {
 610             udst[i]         = src[4 * i + 0];
 611             ydst[2 * i + 0] = src[4 * i + 1];
 612             vdst[i]         = src[4 * i + 2];
 613             ydst[2 * i + 1] = src[4 * i + 3];
 614         }
 615         ydst += lumStride;
 616         src  += srcStride;
 617
 618         for (i = 0; i < chromWidth; i++) {
 619             ydst[2 * i + 0] = src[4 * i + 1];
 620             ydst[2 * i + 1] = src[4 * i + 3];
 621         }
 622         udst += chromStride;
 623         vdst += chromStride;
 624         ydst += lumStride;
 625         src  += srcStride;
 626     }
 627 }
 628
 629 /**
 630  * Height should be a multiple of 2 and width should be a multiple of 2.
 631  * (If this is a problem for anyone then tell me, and I will fix it.)
 632  * Chrominance data is only taken from every second line,
 633  * others are ignored in the C version.
 634  * FIXME: Write HQ version.
 635  */
 636 void rgb24toyv12_c(const uint8_t *src, uint8_t *ydst, uint8_t *udst,
 637                    uint8_t *vdst, int width, int height, int lumStride,
 638                    int chromStride, int srcStride)
 639 {
 640     int y;
 641     const int chromWidth = width >> 1;
 642
 643     for (y = 0; y < height; y += 2) {
 644         int i;
 645         for (i = 0; i < chromWidth; i++) {
 646             unsigned int b = src[6 * i + 0];
 647             unsigned int g = src[6 * i + 1];
 648             unsigned int r = src[6 * i + 2];
 649
 650             unsigned int Y = ((RY * r + GY * g + BY * b) >> RGB2YUV_SHIFT) +  16;
 651             unsigned int V = ((RV * r + GV * g + BV * b) >> RGB2YUV_SHIFT) + 128;
 652             unsigned int U = ((RU * r + GU * g + BU * b) >> RGB2YUV_SHIFT) + 128;
 653
 654             udst[i]     = U;
 655             vdst[i]     = V;
 656             ydst[2 * i] = Y;
 657
 658             b = src[6 * i + 3];
 659             g = src[6 * i + 4];
 660             r = src[6 * i + 5];
 661
 662             Y = ((RY * r + GY * g + BY * b) >> RGB2YUV_SHIFT) + 16;
 663             ydst[2 * i + 1] = Y;
 664         }
 665         ydst += lumStride;
 666         src  += srcStride;
 667
 668         for (i = 0; i < chromWidth; i++) {
 669             unsigned int b = src[6 * i + 0];
 670             unsigned int g = src[6 * i + 1];
 671             unsigned int r = src[6 * i + 2];
 672
 673             unsigned int Y = ((RY * r + GY * g + BY * b) >> RGB2YUV_SHIFT) + 16;
 674
 675             ydst[2 * i] = Y;
 676
 677             b = src[6 * i + 3];
 678             g = src[6 * i + 4];
 679             r = src[6 * i + 5];
 680
 681             Y = ((RY * r + GY * g + BY * b) >> RGB2YUV_SHIFT) + 16;
 682             ydst[2 * i + 1] = Y;
 683         }
 684         udst += chromStride;
 685         vdst += chromStride;
 686         ydst += lumStride;
 687         src  += srcStride;
 688     }
 689 }
 690
 691 static void interleaveBytes_c(const uint8_t *src1, const uint8_t *src2,
 692                               uint8_t *dest, int width, int height,
 693                               int src1Stride, int src2Stride, int dstStride)
 694 {
 695     int h;
 696
 697     for (h = 0; h < height; h++) {
 698         int w;
 699         for (w = 0; w < width; w++) {
 700             dest[2 * w + 0] = src1[w];
 701             dest[2 * w + 1] = src2[w];
 702         }
 703         dest += dstStride;
 704         src1 += src1Stride;
 705         src2 += src2Stride;
 706     }
 707 }
 708
 709 static inline void vu9_to_vu12_c(const uint8_t *src1, const uint8_t *src2,
 710                                  uint8_t *dst1, uint8_t *dst2,
 711                                  int width, int height,
 712                                  int srcStride1, int srcStride2,
 713                                  int dstStride1, int dstStride2)
 714 {
 715     int x, y;
 716     int w = width  / 2;
 717     int h = height / 2;
 718
 719     for (y = 0; y < h; y++) {
 720         const uint8_t *s1 = src1 + srcStride1 * (y >> 1);
 721         uint8_t *d        = dst1 + dstStride1 *  y;
 722         for (x = 0; x < w; x++)
 723             d[2 * x] = d[2 * x + 1] = s1[x];
 724     }
 725     for (y = 0; y < h; y++) {
 726         const uint8_t *s2 = src2 + srcStride2 * (y >> 1);
 727         uint8_t *d        = dst2 + dstStride2 *  y;
 728         for (x = 0; x < w; x++)
 729             d[2 * x] = d[2 * x + 1] = s2[x];
 730     }
 731 }
 732
 733 static inline void yvu9_to_yuy2_c(const uint8_t *src1, const uint8_t *src2,
 734                                   const uint8_t *src3, uint8_t *dst,
 735                                   int width, int height,
 736                                   int srcStride1, int srcStride2,
 737                                   int srcStride3, int dstStride)
 738 {
 739     int x, y;
 740     int w = width / 2;
 741     int h = height;
 742
 743     for (y = 0; y < h; y++) {
 744         const uint8_t *yp = src1 + srcStride1 *  y;
 745         const uint8_t *up = src2 + srcStride2 * (y >> 2);
 746         const uint8_t *vp = src3 + srcStride3 * (y >> 2);
 747         uint8_t *d        = dst  + dstStride  *  y;
 748         for (x = 0; x < w; x++) {
 749             const int x2 = x << 2;
 750             d[8 * x + 0] = yp[x2];
 751             d[8 * x + 1] = up[x];
 752             d[8 * x + 2] = yp[x2 + 1];
 753             d[8 * x + 3] = vp[x];
 754             d[8 * x + 4] = yp[x2 + 2];
 755             d[8 * x + 5] = up[x];
 756             d[8 * x + 6] = yp[x2 + 3];
 757             d[8 * x + 7] = vp[x];
 758         }
 759     }
 760 }
 761
 762 static void extract_even_c(const uint8_t *src, uint8_t *dst, int count)
 763 {
 764     dst   +=  count;
 765     src   +=  count * 2;
 766     count  = -count;
 767     while (count < 0) {
 768         dst[count] = src[2 * count];
 769         count++;
 770     }
 771 }
 772
 773 static void extract_even2_c(const uint8_t *src, uint8_t *dst0, uint8_t *dst1,
 774                             int count)
 775 {
 776     dst0  +=  count;
 777     dst1  +=  count;
 778     src   +=  count * 4;
 779     count  = -count;
 780     while (count < 0) {
 781         dst0[count] = src[4 * count + 0];
 782         dst1[count] = src[4 * count + 2];
 783         count++;
 784     }
 785 }
 786
 787 static void extract_even2avg_c(const uint8_t *src0, const uint8_t *src1,
 788                                uint8_t *dst0, uint8_t *dst1, int count)
 789 {
 790     dst0  +=  count;
 791     dst1  +=  count;
 792     src0  +=  count * 4;
 793     src1  +=  count * 4;
 794     count  = -count;
 795     while (count < 0) {
 796         dst0[count] = (src0[4 * count + 0] + src1[4 * count + 0]) >> 1;
 797         dst1[count] = (src0[4 * count + 2] + src1[4 * count + 2]) >> 1;
 798         count++;
 799     }
 800 }
 801
 802 static void extract_odd2_c(const uint8_t *src, uint8_t *dst0, uint8_t *dst1,
 803                            int count)
 804 {
 805     dst0  +=  count;
 806     dst1  +=  count;
 807     src   +=  count * 4;
 808     count  = -count;
 809     src++;
 810     while (count < 0) {
 811         dst0[count] = src[4 * count + 0];
 812         dst1[count] = src[4 * count + 2];
 813         count++;
 814     }
 815 }
 816
 817 static void extract_odd2avg_c(const uint8_t *src0, const uint8_t *src1,
 818                               uint8_t *dst0, uint8_t *dst1, int count)
 819 {
 820     dst0  +=  count;
 821     dst1  +=  count;
 822     src0  +=  count * 4;
 823     src1  +=  count * 4;
 824     count  = -count;
 825     src0++;
 826     src1++;
 827     while (count < 0) {
 828         dst0[count] = (src0[4 * count + 0] + src1[4 * count + 0]) >> 1;
 829         dst1[count] = (src0[4 * count + 2] + src1[4 * count + 2]) >> 1;
 830         count++;
 831     }
 832 }
 833
 834 static void yuyvtoyuv420_c(uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
 835                            const uint8_t *src, int width, int height,
 836                            int lumStride, int chromStride, int srcStride)
 837 {
 838     int y;
 839     const int chromWidth = -((-width) >> 1);
 840
 841     for (y = 0; y < height; y++) {
 842         extract_even_c(src, ydst, width);
 843         if (y & 1) {
 844             extract_odd2avg_c(src - srcStride, src, udst, vdst, chromWidth);
 845             udst += chromStride;
 846             vdst += chromStride;
 847         }
 848
 849         src  += srcStride;
 850         ydst += lumStride;
 851     }
 852 }
 853
 854 static void yuyvtoyuv422_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 = -((-width) >> 1);
 860
 861     for (y = 0; y < height; y++) {
 862         extract_even_c(src, ydst, width);
 863         extract_odd2_c(src, udst, vdst, chromWidth);
 864
 865         src  += srcStride;
 866         ydst += lumStride;
 867         udst += chromStride;
 868         vdst += chromStride;
 869     }
 870 }
 871
 872 static void uyvytoyuv420_c(uint8_t *ydst, uint8_t *udst, uint8_t *vdst,
 873                            const uint8_t *src, int width, int height,
 874                            int lumStride, int chromStride, int srcStride)
 875 {
 876     int y;
 877     const int chromWidth = -((-width) >> 1);
 878
 879     for (y = 0; y < height; y++) {
 880         extract_even_c(src + 1, ydst, width);
 881         if (y & 1) {
 882             extract_even2avg_c(src - srcStride, src, udst, vdst, chromWidth);
 883             udst += chromStride;
 884             vdst += chromStride;
 885         }
 886
 887         src  += srcStride;
 888         ydst += lumStride;
 889     }
 890 }
 891
 892 static void uyvytoyuv422_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 = -((-width) >> 1);
 898
 899     for (y = 0; y < height; y++) {
 900         extract_even_c(src + 1, ydst, width);
 901         extract_even2_c(src, udst, vdst, chromWidth);
 902
 903         src  += srcStride;
 904         ydst += lumStride;
 905         udst += chromStride;
 906         vdst += chromStride;
 907     }
 908 }
 909
 910 static inline void rgb2rgb_init_c(void)
 911 {
 912     rgb15to16          = rgb15to16_c;
 913     rgb15tobgr24       = rgb15tobgr24_c;
 914     rgb15to32          = rgb15to32_c;
 915     rgb16tobgr24       = rgb16tobgr24_c;
 916     rgb16to32          = rgb16to32_c;
 917     rgb16to15          = rgb16to15_c;
 918     rgb24tobgr16       = rgb24tobgr16_c;
 919     rgb24tobgr15       = rgb24tobgr15_c;
 920     rgb24tobgr32       = rgb24tobgr32_c;
 921     rgb32to16          = rgb32to16_c;
 922     rgb32to15          = rgb32to15_c;
 923     rgb32tobgr24       = rgb32tobgr24_c;
 924     rgb24to15          = rgb24to15_c;
 925     rgb24to16          = rgb24to16_c;
 926     rgb24tobgr24       = rgb24tobgr24_c;
 927     shuffle_bytes_2103 = shuffle_bytes_2103_c;
 928     rgb32tobgr16       = rgb32tobgr16_c;
 929     rgb32tobgr15       = rgb32tobgr15_c;
 930     yv12toyuy2         = yv12toyuy2_c;
 931     yv12touyvy         = yv12touyvy_c;
 932     yuv422ptoyuy2      = yuv422ptoyuy2_c;
 933     yuv422ptouyvy      = yuv422ptouyvy_c;
 934     yuy2toyv12         = yuy2toyv12_c;
 935     planar2x           = planar2x_c;
 936     rgb24toyv12        = rgb24toyv12_c;
 937     interleaveBytes    = interleaveBytes_c;
 938     vu9_to_vu12        = vu9_to_vu12_c;
 939     yvu9_to_yuy2       = yvu9_to_yuy2_c;
 940
 941     uyvytoyuv420       = uyvytoyuv420_c;
 942     uyvytoyuv422       = uyvytoyuv422_c;
 943     yuyvtoyuv420       = yuyvtoyuv420_c;
 944     yuyvtoyuv422       = yuyvtoyuv422_c;
 945 }