libswscale/swscale_unscaled.c

   1 /*
   2  * Copyright (C) 2001-2003 Michael Niedermayer <michaelni@gmx.at>
   3  *
   4  * This file is part of Libav.
   5  *
   6  * Libav is free software; you can redistribute it and/or
   7  * modify it under the terms of the GNU Lesser General Public
   8  * License as published by the Free Software Foundation; either
   9  * version 2.1 of the License, or (at your option) any later version.
  10  *
  11  * Libav is distributed in the hope that it will be useful,
  12  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  14  * Lesser General Public License for more details.
  15  *
  16  * You should have received a copy of the GNU Lesser General Public
  17  * License along with Libav; if not, write to the Free Software
  18  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  19  */
  20
  21 #include <inttypes.h>
  22 #include <string.h>
  23 #include <math.h>
  24 #include <stdio.h>
  25 #include "config.h"
  26 #include <assert.h>
  27 #include "swscale.h"
  28 #include "swscale_internal.h"
  29 #include "rgb2rgb.h"
  30 #include "libavutil/intreadwrite.h"
  31 #include "libavutil/cpu.h"
  32 #include "libavutil/avutil.h"
  33 #include "libavutil/mathematics.h"
  34 #include "libavutil/bswap.h"
  35 #include "libavutil/pixdesc.h"
  36
  37 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_1)[8][8] = {
  38     {   0,  1,  0,  1,  0,  1,  0,  1,},
  39     {   1,  0,  1,  0,  1,  0,  1,  0,},
  40     {   0,  1,  0,  1,  0,  1,  0,  1,},
  41     {   1,  0,  1,  0,  1,  0,  1,  0,},
  42     {   0,  1,  0,  1,  0,  1,  0,  1,},
  43     {   1,  0,  1,  0,  1,  0,  1,  0,},
  44     {   0,  1,  0,  1,  0,  1,  0,  1,},
  45     {   1,  0,  1,  0,  1,  0,  1,  0,},
  46 };
  47 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_3)[8][8] = {
  48     {   1,  2,  1,  2,  1,  2,  1,  2,},
  49     {   3,  0,  3,  0,  3,  0,  3,  0,},
  50     {   1,  2,  1,  2,  1,  2,  1,  2,},
  51     {   3,  0,  3,  0,  3,  0,  3,  0,},
  52     {   1,  2,  1,  2,  1,  2,  1,  2,},
  53     {   3,  0,  3,  0,  3,  0,  3,  0,},
  54     {   1,  2,  1,  2,  1,  2,  1,  2,},
  55     {   3,  0,  3,  0,  3,  0,  3,  0,},
  56 };
  57 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_64)[8][8] = {
  58     {  18, 34, 30, 46, 17, 33, 29, 45,},
  59     {  50,  2, 62, 14, 49,  1, 61, 13,},
  60     {  26, 42, 22, 38, 25, 41, 21, 37,},
  61     {  58, 10, 54,  6, 57,  9, 53,  5,},
  62     {  16, 32, 28, 44, 19, 35, 31, 47,},
  63     {  48,  0, 60, 12, 51,  3, 63, 15,},
  64     {  24, 40, 20, 36, 27, 43, 23, 39,},
  65     {  56,  8, 52,  4, 59, 11, 55,  7,},
  66 };
  67 extern const uint8_t dither_8x8_128[8][8];
  68 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_256)[8][8] = {
  69     {  72, 136, 120, 184,  68, 132, 116, 180,},
  70     { 200,   8, 248,  56, 196,   4, 244,  52,},
  71     { 104, 168,  88, 152, 100, 164,  84, 148,},
  72     { 232,  40, 216,  24, 228,  36, 212,  20,},
  73     {  64, 128, 102, 176,  76, 140, 124, 188,},
  74     { 192,   0, 240,  48, 204,  12, 252,  60,},
  75     {  96, 160,  80, 144, 108, 172,  92, 156,},
  76     { 224,  32, 208,  16, 236,  44, 220,  28,},
  77 };
  78
  79 #define RGB2YUV_SHIFT 15
  80 #define BY ( (int) (0.114 * 219 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))
  81 #define BV (-(int) (0.081 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))
  82 #define BU ( (int) (0.500 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))
  83 #define GY ( (int) (0.587 * 219 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))
  84 #define GV (-(int) (0.419 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))
  85 #define GU (-(int) (0.331 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))
  86 #define RY ( (int) (0.299 * 219 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))
  87 #define RV ( (int) (0.500 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))
  88 #define RU (-(int) (0.169 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))
  89
  90 static void fillPlane(uint8_t *plane, int stride, int width, int height, int y,
  91                       uint8_t val)
  92 {
  93     int i;
  94     uint8_t *ptr = plane + stride * y;
  95     for (i = 0; i < height; i++) {
  96         memset(ptr, val, width);
  97         ptr += stride;
  98     }
  99 }
 100
 101 static void fill_plane9or10(uint8_t *plane, int stride, int width,
 102                             int height, int y, uint8_t val,
 103                             const int dst_depth, const int big_endian)
 104 {
 105     int i, j;
 106     uint16_t *dst = (uint16_t *) (plane + stride * y);
 107 #define FILL8TO9_OR_10(wfunc) \
 108     for (i = 0; i < height; i++) { \
 109         for (j = 0; j < width; j++) { \
 110             wfunc(&dst[j], (val << (dst_depth - 8)) |  \
 111                                (val >> (16 - dst_depth))); \
 112         } \
 113         dst += stride / 2; \
 114     }
 115     if (big_endian) {
 116         FILL8TO9_OR_10(AV_WB16);
 117     } else {
 118         FILL8TO9_OR_10(AV_WL16);
 119     }
 120 }
 121
 122 static void copyPlane(const uint8_t *src, int srcStride,
 123                       int srcSliceY, int srcSliceH, int width,
 124                       uint8_t *dst, int dstStride)
 125 {
 126     dst += dstStride * srcSliceY;
 127     if (dstStride == srcStride && srcStride > 0) {
 128         memcpy(dst, src, srcSliceH * dstStride);
 129     } else {
 130         int i;
 131         for (i = 0; i < srcSliceH; i++) {
 132             memcpy(dst, src, width);
 133             src += srcStride;
 134             dst += dstStride;
 135         }
 136     }
 137 }
 138
 139 static int planarToNv12Wrapper(SwsContext *c, const uint8_t *src[],
 140                                int srcStride[], int srcSliceY,
 141                                int srcSliceH, uint8_t *dstParam[],
 142                                int dstStride[])
 143 {
 144     uint8_t *dst = dstParam[1] + dstStride[1] * srcSliceY / 2;
 145
 146     copyPlane(src[0], srcStride[0], srcSliceY, srcSliceH, c->srcW,
 147               dstParam[0], dstStride[0]);
 148
 149     if (c->dstFormat == AV_PIX_FMT_NV12)
 150         interleaveBytes(src[1], src[2], dst, c->srcW / 2, srcSliceH / 2,
 151                         srcStride[1], srcStride[2], dstStride[0]);
 152     else
 153         interleaveBytes(src[2], src[1], dst, c->srcW / 2, srcSliceH / 2,
 154                         srcStride[2], srcStride[1], dstStride[0]);
 155
 156     return srcSliceH;
 157 }
 158
 159 static int planarToYuy2Wrapper(SwsContext *c, const uint8_t *src[],
 160                                int srcStride[], int srcSliceY, int srcSliceH,
 161                                uint8_t *dstParam[], int dstStride[])
 162 {
 163     uint8_t *dst = dstParam[0] + dstStride[0] * srcSliceY;
 164
 165     yv12toyuy2(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0],
 166                srcStride[1], dstStride[0]);
 167
 168     return srcSliceH;
 169 }
 170
 171 static int planarToUyvyWrapper(SwsContext *c, const uint8_t *src[],
 172                                int srcStride[], int srcSliceY, int srcSliceH,
 173                                uint8_t *dstParam[], int dstStride[])
 174 {
 175     uint8_t *dst = dstParam[0] + dstStride[0] * srcSliceY;
 176
 177     yv12touyvy(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0],
 178                srcStride[1], dstStride[0]);
 179
 180     return srcSliceH;
 181 }
 182
 183 static int yuv422pToYuy2Wrapper(SwsContext *c, const uint8_t *src[],
 184                                 int srcStride[], int srcSliceY, int srcSliceH,
 185                                 uint8_t *dstParam[], int dstStride[])
 186 {
 187     uint8_t *dst = dstParam[0] + dstStride[0] * srcSliceY;
 188
 189     yuv422ptoyuy2(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0],
 190                   srcStride[1], dstStride[0]);
 191
 192     return srcSliceH;
 193 }
 194
 195 static int yuv422pToUyvyWrapper(SwsContext *c, const uint8_t *src[],
 196                                 int srcStride[], int srcSliceY, int srcSliceH,
 197                                 uint8_t *dstParam[], int dstStride[])
 198 {
 199     uint8_t *dst = dstParam[0] + dstStride[0] * srcSliceY;
 200
 201     yuv422ptouyvy(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0],
 202                   srcStride[1], dstStride[0]);
 203
 204     return srcSliceH;
 205 }
 206
 207 static int yuyvToYuv420Wrapper(SwsContext *c, const uint8_t *src[],
 208                                int srcStride[], int srcSliceY, int srcSliceH,
 209                                uint8_t *dstParam[], int dstStride[])
 210 {
 211     uint8_t *ydst = dstParam[0] + dstStride[0] * srcSliceY;
 212     uint8_t *udst = dstParam[1] + dstStride[1] * srcSliceY / 2;
 213     uint8_t *vdst = dstParam[2] + dstStride[2] * srcSliceY / 2;
 214
 215     yuyvtoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0],
 216                  dstStride[1], srcStride[0]);
 217
 218     if (dstParam[3])
 219         fillPlane(dstParam[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
 220
 221     return srcSliceH;
 222 }
 223
 224 static int yuyvToYuv422Wrapper(SwsContext *c, const uint8_t *src[],
 225                                int srcStride[], int srcSliceY, int srcSliceH,
 226                                uint8_t *dstParam[], int dstStride[])
 227 {
 228     uint8_t *ydst = dstParam[0] + dstStride[0] * srcSliceY;
 229     uint8_t *udst = dstParam[1] + dstStride[1] * srcSliceY;
 230     uint8_t *vdst = dstParam[2] + dstStride[2] * srcSliceY;
 231
 232     yuyvtoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0],
 233                  dstStride[1], srcStride[0]);
 234
 235     return srcSliceH;
 236 }
 237
 238 static int uyvyToYuv420Wrapper(SwsContext *c, const uint8_t *src[],
 239                                int srcStride[], int srcSliceY, int srcSliceH,
 240                                uint8_t *dstParam[], int dstStride[])
 241 {
 242     uint8_t *ydst = dstParam[0] + dstStride[0] * srcSliceY;
 243     uint8_t *udst = dstParam[1] + dstStride[1] * srcSliceY / 2;
 244     uint8_t *vdst = dstParam[2] + dstStride[2] * srcSliceY / 2;
 245
 246     uyvytoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0],
 247                  dstStride[1], srcStride[0]);
 248
 249     if (dstParam[3])
 250         fillPlane(dstParam[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
 251
 252     return srcSliceH;
 253 }
 254
 255 static int uyvyToYuv422Wrapper(SwsContext *c, const uint8_t *src[],
 256                                int srcStride[], int srcSliceY, int srcSliceH,
 257                                uint8_t *dstParam[], int dstStride[])
 258 {
 259     uint8_t *ydst = dstParam[0] + dstStride[0] * srcSliceY;
 260     uint8_t *udst = dstParam[1] + dstStride[1] * srcSliceY;
 261     uint8_t *vdst = dstParam[2] + dstStride[2] * srcSliceY;
 262
 263     uyvytoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0],
 264                  dstStride[1], srcStride[0]);
 265
 266     return srcSliceH;
 267 }
 268
 269 static void gray8aToPacked32(const uint8_t *src, uint8_t *dst, int num_pixels,
 270                              const uint8_t *palette)
 271 {
 272     int i;
 273     for (i = 0; i < num_pixels; i++)
 274         ((uint32_t *) dst)[i] = ((const uint32_t *) palette)[src[i << 1]] | (src[(i << 1) + 1] << 24);
 275 }
 276
 277 static void gray8aToPacked32_1(const uint8_t *src, uint8_t *dst, int num_pixels,
 278                                const uint8_t *palette)
 279 {
 280     int i;
 281
 282     for (i = 0; i < num_pixels; i++)
 283         ((uint32_t *) dst)[i] = ((const uint32_t *) palette)[src[i << 1]] | src[(i << 1) + 1];
 284 }
 285
 286 static void gray8aToPacked24(const uint8_t *src, uint8_t *dst, int num_pixels,
 287                              const uint8_t *palette)
 288 {
 289     int i;
 290
 291     for (i = 0; i < num_pixels; i++) {
 292         //FIXME slow?
 293         dst[0] = palette[src[i << 1] * 4 + 0];
 294         dst[1] = palette[src[i << 1] * 4 + 1];
 295         dst[2] = palette[src[i << 1] * 4 + 2];
 296         dst += 3;
 297     }
 298 }
 299
 300 static int packed_16bpc_bswap(SwsContext *c, const uint8_t *src[],
 301                               int srcStride[], int srcSliceY, int srcSliceH,
 302                               uint8_t *dst[], int dstStride[])
 303 {
 304     int i, j;
 305     int srcstr = srcStride[0] >> 1;
 306     int dststr = dstStride[0] >> 1;
 307     uint16_t       *dstPtr =       (uint16_t *) dst[0];
 308     const uint16_t *srcPtr = (const uint16_t *) src[0];
 309     int min_stride         = FFMIN(srcstr, dststr);
 310
 311     for (i = 0; i < srcSliceH; i++) {
 312         for (j = 0; j < min_stride; j++) {
 313             dstPtr[j] = av_bswap16(srcPtr[j]);
 314         }
 315         srcPtr += srcstr;
 316         dstPtr += dststr;
 317     }
 318
 319     return srcSliceH;
 320 }
 321
 322 static int palToRgbWrapper(SwsContext *c, const uint8_t *src[], int srcStride[],
 323                            int srcSliceY, int srcSliceH, uint8_t *dst[],
 324                            int dstStride[])
 325 {
 326     const enum AVPixelFormat srcFormat = c->srcFormat;
 327     const enum AVPixelFormat dstFormat = c->dstFormat;
 328     void (*conv)(const uint8_t *src, uint8_t *dst, int num_pixels,
 329                  const uint8_t *palette) = NULL;
 330     int i;
 331     uint8_t *dstPtr = dst[0] + dstStride[0] * srcSliceY;
 332     const uint8_t *srcPtr = src[0];
 333
 334     if (srcFormat == AV_PIX_FMT_Y400A) {
 335         switch (dstFormat) {
 336         case AV_PIX_FMT_RGB32  : conv = gray8aToPacked32; break;
 337         case AV_PIX_FMT_BGR32  : conv = gray8aToPacked32; break;
 338         case AV_PIX_FMT_BGR32_1: conv = gray8aToPacked32_1; break;
 339         case AV_PIX_FMT_RGB32_1: conv = gray8aToPacked32_1; break;
 340         case AV_PIX_FMT_RGB24  : conv = gray8aToPacked24; break;
 341         case AV_PIX_FMT_BGR24  : conv = gray8aToPacked24; break;
 342         }
 343     } else if (usePal(srcFormat)) {
 344         switch (dstFormat) {
 345         case AV_PIX_FMT_RGB32  : conv = sws_convertPalette8ToPacked32; break;
 346         case AV_PIX_FMT_BGR32  : conv = sws_convertPalette8ToPacked32; break;
 347         case AV_PIX_FMT_BGR32_1: conv = sws_convertPalette8ToPacked32; break;
 348         case AV_PIX_FMT_RGB32_1: conv = sws_convertPalette8ToPacked32; break;
 349         case AV_PIX_FMT_RGB24  : conv = sws_convertPalette8ToPacked24; break;
 350         case AV_PIX_FMT_BGR24  : conv = sws_convertPalette8ToPacked24; break;
 351         }
 352     }
 353
 354     if (!conv)
 355         av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
 356                sws_format_name(srcFormat), sws_format_name(dstFormat));
 357     else {
 358         for (i = 0; i < srcSliceH; i++) {
 359             conv(srcPtr, dstPtr, c->srcW, (uint8_t *) c->pal_rgb);
 360             srcPtr += srcStride[0];
 361             dstPtr += dstStride[0];
 362         }
 363     }
 364
 365     return srcSliceH;
 366 }
 367
 368 static void gbr24ptopacked24(const uint8_t *src[], int srcStride[],
 369                              uint8_t *dst, int dstStride, int srcSliceH,
 370                              int width)
 371 {
 372     int x, h, i;
 373     for (h = 0; h < srcSliceH; h++) {
 374         uint8_t *dest = dst + dstStride * h;
 375         for (x = 0; x < width; x++) {
 376             *dest++ = src[0][x];
 377             *dest++ = src[1][x];
 378             *dest++ = src[2][x];
 379         }
 380
 381         for (i = 0; i < 3; i++)
 382             src[i] += srcStride[i];
 383     }
 384 }
 385
 386 static void gbr24ptopacked32(const uint8_t *src[], int srcStride[],
 387                              uint8_t *dst, int dstStride, int srcSliceH,
 388                              int alpha_first, int width)
 389 {
 390     int x, h, i;
 391     for (h = 0; h < srcSliceH; h++) {
 392         uint8_t *dest = dst + dstStride * h;
 393
 394         if (alpha_first) {
 395             for (x = 0; x < width; x++) {
 396                 *dest++ = 0xff;
 397                 *dest++ = src[0][x];
 398                 *dest++ = src[1][x];
 399                 *dest++ = src[2][x];
 400             }
 401         } else {
 402             for (x = 0; x < width; x++) {
 403                 *dest++ = src[0][x];
 404                 *dest++ = src[1][x];
 405                 *dest++ = src[2][x];
 406                 *dest++ = 0xff;
 407             }
 408         }
 409
 410         for (i = 0; i < 3; i++)
 411             src[i] += srcStride[i];
 412     }
 413 }
 414
 415 static int planarRgbToRgbWrapper(SwsContext *c, const uint8_t *src[],
 416                                  int srcStride[], int srcSliceY, int srcSliceH,
 417                                  uint8_t *dst[], int dstStride[])
 418 {
 419     int alpha_first = 0;
 420     const uint8_t *src102[] = { src[1], src[0], src[2] };
 421     const uint8_t *src201[] = { src[2], src[0], src[1] };
 422     int stride102[] = { srcStride[1], srcStride[0], srcStride[2] };
 423     int stride201[] = { srcStride[2], srcStride[0], srcStride[1] };
 424
 425     if (c->srcFormat != AV_PIX_FMT_GBRP) {
 426         av_log(c, AV_LOG_ERROR, "unsupported planar RGB conversion %s -> %s\n",
 427                av_get_pix_fmt_name(c->srcFormat),
 428                av_get_pix_fmt_name(c->dstFormat));
 429         return srcSliceH;
 430     }
 431
 432     switch (c->dstFormat) {
 433     case AV_PIX_FMT_BGR24:
 434         gbr24ptopacked24(src102, stride102,
 435                          dst[0] + srcSliceY * dstStride[0], dstStride[0],
 436                          srcSliceH, c->srcW);
 437         break;
 438
 439     case AV_PIX_FMT_RGB24:
 440         gbr24ptopacked24(src201, stride201,
 441                          dst[0] + srcSliceY * dstStride[0], dstStride[0],
 442                          srcSliceH, c->srcW);
 443         break;
 444
 445     case AV_PIX_FMT_ARGB:
 446         alpha_first = 1;
 447     case AV_PIX_FMT_RGBA:
 448         gbr24ptopacked32(src201, stride201,
 449                          dst[0] + srcSliceY * dstStride[0], dstStride[0],
 450                          srcSliceH, alpha_first, c->srcW);
 451         break;
 452
 453     case AV_PIX_FMT_ABGR:
 454         alpha_first = 1;
 455     case AV_PIX_FMT_BGRA:
 456         gbr24ptopacked32(src102, stride102,
 457                          dst[0] + srcSliceY * dstStride[0], dstStride[0],
 458                          srcSliceH, alpha_first, c->srcW);
 459         break;
 460
 461     default:
 462         av_log(c, AV_LOG_ERROR,
 463                "unsupported planar RGB conversion %s -> %s\n",
 464                av_get_pix_fmt_name(c->srcFormat),
 465                av_get_pix_fmt_name(c->dstFormat));
 466     }
 467
 468     return srcSliceH;
 469 }
 470
 471 static void packedtogbr24p(const uint8_t *src, int srcStride,
 472                            uint8_t *dst[], int dstStride[], int srcSliceH,
 473                            int alpha_first, int inc_size, int width)
 474 {
 475     uint8_t *dest[3];
 476     int x, h;
 477
 478     dest[0] = dst[0];
 479     dest[1] = dst[1];
 480     dest[2] = dst[2];
 481
 482     if (alpha_first)
 483         src++;
 484
 485     for (h = 0; h < srcSliceH; h++) {
 486         for (x = 0; x < width; x++) {
 487             dest[0][x] = src[0];
 488             dest[1][x] = src[1];
 489             dest[2][x] = src[2];
 490
 491             src += inc_size;
 492         }
 493         src     += srcStride - width * inc_size;
 494         dest[0] += dstStride[0];
 495         dest[1] += dstStride[1];
 496         dest[2] += dstStride[2];
 497     }
 498 }
 499
 500 static int rgbToPlanarRgbWrapper(SwsContext *c, const uint8_t *src[],
 501                                  int srcStride[], int srcSliceY, int srcSliceH,
 502                                  uint8_t *dst[], int dstStride[])
 503 {
 504     int alpha_first = 0;
 505     int stride102[] = { dstStride[1], dstStride[0], dstStride[2] };
 506     int stride201[] = { dstStride[2], dstStride[0], dstStride[1] };
 507     uint8_t *dst102[] = { dst[1] + srcSliceY * dstStride[1],
 508                           dst[0] + srcSliceY * dstStride[0],
 509                           dst[2] + srcSliceY * dstStride[2] };
 510     uint8_t *dst201[] = { dst[2] + srcSliceY * dstStride[2],
 511                           dst[0] + srcSliceY * dstStride[0],
 512                           dst[1] + srcSliceY * dstStride[1] };
 513
 514     switch (c->srcFormat) {
 515     case PIX_FMT_RGB24:
 516         packedtogbr24p((const uint8_t *) src[0], srcStride[0], dst201,
 517                        stride201, srcSliceH, alpha_first, 3, c->srcW);
 518         break;
 519     case PIX_FMT_BGR24:
 520         packedtogbr24p((const uint8_t *) src[0], srcStride[0], dst102,
 521                        stride102, srcSliceH, alpha_first, 3, c->srcW);
 522         break;
 523     case PIX_FMT_ARGB:
 524         alpha_first = 1;
 525     case PIX_FMT_RGBA:
 526         packedtogbr24p((const uint8_t *) src[0], srcStride[0], dst201,
 527                        stride201, srcSliceH, alpha_first, 4, c->srcW);
 528         break;
 529     case PIX_FMT_ABGR:
 530         alpha_first = 1;
 531     case PIX_FMT_BGRA:
 532         packedtogbr24p((const uint8_t *) src[0], srcStride[0], dst102,
 533                        stride102, srcSliceH, alpha_first, 4, c->srcW);
 534         break;
 535     default:
 536         av_log(c, AV_LOG_ERROR,
 537                "unsupported planar RGB conversion %s -> %s\n",
 538                av_get_pix_fmt_name(c->srcFormat),
 539                av_get_pix_fmt_name(c->dstFormat));
 540     }
 541
 542     return srcSliceH;
 543 }
 544
 545 #define isRGBA32(x) (            \
 546            (x) == AV_PIX_FMT_ARGB   \
 547         || (x) == AV_PIX_FMT_RGBA   \
 548         || (x) == AV_PIX_FMT_BGRA   \
 549         || (x) == AV_PIX_FMT_ABGR   \
 550         )
 551
 552 /* {RGB,BGR}{15,16,24,32,32_1} -> {RGB,BGR}{15,16,24,32} */
 553 typedef void (* rgbConvFn) (const uint8_t *, uint8_t *, int);
 554 static rgbConvFn findRgbConvFn(SwsContext *c)
 555 {
 556     const enum AVPixelFormat srcFormat = c->srcFormat;
 557     const enum AVPixelFormat dstFormat = c->dstFormat;
 558     const int srcId = c->srcFormatBpp;
 559     const int dstId = c->dstFormatBpp;
 560     rgbConvFn conv = NULL;
 561     const AVPixFmtDescriptor *desc_src = av_pix_fmt_desc_get(srcFormat);
 562     const AVPixFmtDescriptor *desc_dst = av_pix_fmt_desc_get(dstFormat);
 563
 564 #define IS_NOT_NE(bpp, desc) \
 565     (((bpp + 7) >> 3) == 2 && \
 566      (!(desc->flags & PIX_FMT_BE) != !HAVE_BIGENDIAN))
 567
 568     /* if this is non-native rgb444/555/565, don't handle it here. */
 569     if (IS_NOT_NE(srcId, desc_src) || IS_NOT_NE(dstId, desc_dst))
 570         return NULL;
 571
 572 #define CONV_IS(src, dst) (srcFormat == AV_PIX_FMT_##src && dstFormat == AV_PIX_FMT_##dst)
 573
 574     if (isRGBA32(srcFormat) && isRGBA32(dstFormat)) {
 575         if (     CONV_IS(ABGR, RGBA)
 576               || CONV_IS(ARGB, BGRA)
 577               || CONV_IS(BGRA, ARGB)
 578               || CONV_IS(RGBA, ABGR)) conv = shuffle_bytes_3210;
 579         else if (CONV_IS(ABGR, ARGB)
 580               || CONV_IS(ARGB, ABGR)) conv = shuffle_bytes_0321;
 581         else if (CONV_IS(ABGR, BGRA)
 582               || CONV_IS(ARGB, RGBA)) conv = shuffle_bytes_1230;
 583         else if (CONV_IS(BGRA, RGBA)
 584               || CONV_IS(RGBA, BGRA)) conv = shuffle_bytes_2103;
 585         else if (CONV_IS(BGRA, ABGR)
 586               || CONV_IS(RGBA, ARGB)) conv = shuffle_bytes_3012;
 587     } else
 588     /* BGR -> BGR */
 589     if ((isBGRinInt(srcFormat) && isBGRinInt(dstFormat)) ||
 590         (isRGBinInt(srcFormat) && isRGBinInt(dstFormat))) {
 591         switch (srcId | (dstId << 16)) {
 592         case 0x000F000C: conv = rgb12to15; break;
 593         case 0x000F0010: conv = rgb16to15; break;
 594         case 0x000F0018: conv = rgb24to15; break;
 595         case 0x000F0020: conv = rgb32to15; break;
 596         case 0x0010000F: conv = rgb15to16; break;
 597         case 0x00100018: conv = rgb24to16; break;
 598         case 0x00100020: conv = rgb32to16; break;
 599         case 0x0018000F: conv = rgb15to24; break;
 600         case 0x00180010: conv = rgb16to24; break;
 601         case 0x00180020: conv = rgb32to24; break;
 602         case 0x0020000F: conv = rgb15to32; break;
 603         case 0x00200010: conv = rgb16to32; break;
 604         case 0x00200018: conv = rgb24to32; break;
 605         }
 606     } else if ((isBGRinInt(srcFormat) && isRGBinInt(dstFormat)) ||
 607                (isRGBinInt(srcFormat) && isBGRinInt(dstFormat))) {
 608         switch (srcId | (dstId << 16)) {
 609         case 0x000C000C: conv = rgb12tobgr12; break;
 610         case 0x000F000F: conv = rgb15tobgr15; break;
 611         case 0x000F0010: conv = rgb16tobgr15; break;
 612         case 0x000F0018: conv = rgb24tobgr15; break;
 613         case 0x000F0020: conv = rgb32tobgr15; break;
 614         case 0x0010000F: conv = rgb15tobgr16; break;
 615         case 0x00100010: conv = rgb16tobgr16; break;
 616         case 0x00100018: conv = rgb24tobgr16; break;
 617         case 0x00100020: conv = rgb32tobgr16; break;
 618         case 0x0018000F: conv = rgb15tobgr24; break;
 619         case 0x00180010: conv = rgb16tobgr24; break;
 620         case 0x00180018: conv = rgb24tobgr24; break;
 621         case 0x00180020: conv = rgb32tobgr24; break;
 622         case 0x0020000F: conv = rgb15tobgr32; break;
 623         case 0x00200010: conv = rgb16tobgr32; break;
 624         case 0x00200018: conv = rgb24tobgr32; break;
 625         }
 626     }
 627
 628     return conv;
 629 }
 630
 631 /* {RGB,BGR}{15,16,24,32,32_1} -> {RGB,BGR}{15,16,24,32} */
 632 static int rgbToRgbWrapper(SwsContext *c, const uint8_t *src[], int srcStride[],
 633                            int srcSliceY, int srcSliceH, uint8_t *dst[],
 634                            int dstStride[])
 635
 636 {
 637     const enum AVPixelFormat srcFormat = c->srcFormat;
 638     const enum AVPixelFormat dstFormat = c->dstFormat;
 639     const int srcBpp = (c->srcFormatBpp + 7) >> 3;
 640     const int dstBpp = (c->dstFormatBpp + 7) >> 3;
 641     rgbConvFn conv = findRgbConvFn(c);
 642
 643     if (!conv) {
 644         av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
 645                sws_format_name(srcFormat), sws_format_name(dstFormat));
 646     } else {
 647         const uint8_t *srcPtr = src[0];
 648               uint8_t *dstPtr = dst[0];
 649         if ((srcFormat == AV_PIX_FMT_RGB32_1 || srcFormat == AV_PIX_FMT_BGR32_1) &&
 650             !isRGBA32(dstFormat))
 651             srcPtr += ALT32_CORR;
 652
 653         if ((dstFormat == AV_PIX_FMT_RGB32_1 || dstFormat == AV_PIX_FMT_BGR32_1) &&
 654             !isRGBA32(srcFormat))
 655             dstPtr += ALT32_CORR;
 656
 657         if (dstStride[0] * srcBpp == srcStride[0] * dstBpp && srcStride[0] > 0 &&
 658             !(srcStride[0] % srcBpp))
 659             conv(srcPtr, dstPtr + dstStride[0] * srcSliceY,
 660                  srcSliceH * srcStride[0]);
 661         else {
 662             int i;
 663             dstPtr += dstStride[0] * srcSliceY;
 664
 665             for (i = 0; i < srcSliceH; i++) {
 666                 conv(srcPtr, dstPtr, c->srcW * srcBpp);
 667                 srcPtr += srcStride[0];
 668                 dstPtr += dstStride[0];
 669             }
 670         }
 671     }
 672     return srcSliceH;
 673 }
 674
 675 static int bgr24ToYv12Wrapper(SwsContext *c, const uint8_t *src[],
 676                               int srcStride[], int srcSliceY, int srcSliceH,
 677                               uint8_t *dst[], int dstStride[])
 678 {
 679     rgb24toyv12(
 680         src[0],
 681         dst[0] +  srcSliceY       * dstStride[0],
 682         dst[1] + (srcSliceY >> 1) * dstStride[1],
 683         dst[2] + (srcSliceY >> 1) * dstStride[2],
 684         c->srcW, srcSliceH,
 685         dstStride[0], dstStride[1], srcStride[0]);
 686     if (dst[3])
 687         fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
 688     return srcSliceH;
 689 }
 690
 691 static int yvu9ToYv12Wrapper(SwsContext *c, const uint8_t *src[],
 692                              int srcStride[], int srcSliceY, int srcSliceH,
 693                              uint8_t *dst[], int dstStride[])
 694 {
 695     copyPlane(src[0], srcStride[0], srcSliceY, srcSliceH, c->srcW,
 696               dst[0], dstStride[0]);
 697
 698     planar2x(src[1], dst[1] + dstStride[1] * (srcSliceY >> 1), c->chrSrcW,
 699              srcSliceH >> 2, srcStride[1], dstStride[1]);
 700     planar2x(src[2], dst[2] + dstStride[2] * (srcSliceY >> 1), c->chrSrcW,
 701              srcSliceH >> 2, srcStride[2], dstStride[2]);
 702     if (dst[3])
 703         fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
 704     return srcSliceH;
 705 }
 706
 707 /* unscaled copy like stuff (assumes nearly identical formats) */
 708 static int packedCopyWrapper(SwsContext *c, const uint8_t *src[],
 709                              int srcStride[], int srcSliceY, int srcSliceH,
 710                              uint8_t *dst[], int dstStride[])
 711 {
 712     if (dstStride[0] == srcStride[0] && srcStride[0] > 0)
 713         memcpy(dst[0] + dstStride[0] * srcSliceY, src[0], srcSliceH * dstStride[0]);
 714     else {
 715         int i;
 716         const uint8_t *srcPtr = src[0];
 717         uint8_t *dstPtr = dst[0] + dstStride[0] * srcSliceY;
 718         int length = 0;
 719
 720         /* universal length finder */
 721         while (length + c->srcW <= FFABS(dstStride[0]) &&
 722                length + c->srcW <= FFABS(srcStride[0]))
 723             length += c->srcW;
 724         assert(length != 0);
 725
 726         for (i = 0; i < srcSliceH; i++) {
 727             memcpy(dstPtr, srcPtr, length);
 728             srcPtr += srcStride[0];
 729             dstPtr += dstStride[0];
 730         }
 731     }
 732     return srcSliceH;
 733 }
 734
 735 #define clip9(x)  av_clip_uintp2(x,  9)
 736 #define clip10(x) av_clip_uintp2(x, 10)
 737 #define DITHER_COPY(dst, dstStride, wfunc, src, srcStride, rfunc, dithers, shift, clip) \
 738     for (i = 0; i < height; i++) { \
 739         const uint8_t *dither = dithers[i & 7]; \
 740         for (j = 0; j < length - 7; j += 8) { \
 741             wfunc(&dst[j + 0], clip((rfunc(&src[j + 0]) + dither[0]) >> shift)); \
 742             wfunc(&dst[j + 1], clip((rfunc(&src[j + 1]) + dither[1]) >> shift)); \
 743             wfunc(&dst[j + 2], clip((rfunc(&src[j + 2]) + dither[2]) >> shift)); \
 744             wfunc(&dst[j + 3], clip((rfunc(&src[j + 3]) + dither[3]) >> shift)); \
 745             wfunc(&dst[j + 4], clip((rfunc(&src[j + 4]) + dither[4]) >> shift)); \
 746             wfunc(&dst[j + 5], clip((rfunc(&src[j + 5]) + dither[5]) >> shift)); \
 747             wfunc(&dst[j + 6], clip((rfunc(&src[j + 6]) + dither[6]) >> shift)); \
 748             wfunc(&dst[j + 7], clip((rfunc(&src[j + 7]) + dither[7]) >> shift)); \
 749         } \
 750         for (; j < length; j++) \
 751             wfunc(&dst[j],     (rfunc(&src[j]) + dither[j & 7]) >> shift); \
 752         dst += dstStride; \
 753         src += srcStride; \
 754     }
 755
 756 static int planarCopyWrapper(SwsContext *c, const uint8_t *src[],
 757                              int srcStride[], int srcSliceY, int srcSliceH,
 758                              uint8_t *dst[], int dstStride[])
 759 {
 760     const AVPixFmtDescriptor *desc_src = av_pix_fmt_desc_get(c->srcFormat);
 761     const AVPixFmtDescriptor *desc_dst = av_pix_fmt_desc_get(c->dstFormat);
 762     int plane, i, j;
 763     for (plane = 0; plane < 4; plane++) {
 764         int length = (plane == 0 || plane == 3) ? c->srcW  : -((-c->srcW  ) >> c->chrDstHSubSample);
 765         int y =      (plane == 0 || plane == 3) ? srcSliceY: -((-srcSliceY) >> c->chrDstVSubSample);
 766         int height = (plane == 0 || plane == 3) ? srcSliceH: -((-srcSliceH) >> c->chrDstVSubSample);
 767         const uint8_t *srcPtr = src[plane];
 768         uint8_t *dstPtr = dst[plane] + dstStride[plane] * y;
 769
 770         if (!dst[plane])
 771             continue;
 772         // ignore palette for GRAY8
 773         if (plane == 1 && !dst[2]) continue;
 774         if (!src[plane] || (plane == 1 && !src[2])) {
 775             int val = (plane == 3) ? 255 : 128;
 776             if (is16BPS(c->dstFormat))
 777                 length *= 2;
 778             if (is9_OR_10BPS(c->dstFormat)) {
 779                 fill_plane9or10(dst[plane], dstStride[plane],
 780                                 length, height, y, val,
 781                                 desc_dst->comp[plane].depth_minus1 + 1,
 782                                 isBE(c->dstFormat));
 783             } else
 784                 fillPlane(dst[plane], dstStride[plane], length, height, y,
 785                           val);
 786         } else {
 787             if (is9_OR_10BPS(c->srcFormat)) {
 788                 const int src_depth = desc_src->comp[plane].depth_minus1 + 1;
 789                 const int dst_depth = desc_dst->comp[plane].depth_minus1 + 1;
 790                 const uint16_t *srcPtr2 = (const uint16_t *) srcPtr;
 791
 792                 if (is16BPS(c->dstFormat)) {
 793                     uint16_t *dstPtr2 = (uint16_t *) dstPtr;
 794 #define COPY9_OR_10TO16(rfunc, wfunc) \
 795                     for (i = 0; i < height; i++) { \
 796                         for (j = 0; j < length; j++) { \
 797                             int srcpx = rfunc(&srcPtr2[j]); \
 798                             wfunc(&dstPtr2[j], (srcpx << (16 - src_depth)) | (srcpx >> (2 * src_depth - 16))); \
 799                         } \
 800                         dstPtr2 += dstStride[plane] / 2; \
 801                         srcPtr2 += srcStride[plane] / 2; \
 802                     }
 803                     if (isBE(c->dstFormat)) {
 804                         if (isBE(c->srcFormat)) {
 805                             COPY9_OR_10TO16(AV_RB16, AV_WB16);
 806                         } else {
 807                             COPY9_OR_10TO16(AV_RL16, AV_WB16);
 808                         }
 809                     } else {
 810                         if (isBE(c->srcFormat)) {
 811                             COPY9_OR_10TO16(AV_RB16, AV_WL16);
 812                         } else {
 813                             COPY9_OR_10TO16(AV_RL16, AV_WL16);
 814                         }
 815                     }
 816                 } else if (is9_OR_10BPS(c->dstFormat)) {
 817                     uint16_t *dstPtr2 = (uint16_t *) dstPtr;
 818 #define COPY9_OR_10TO9_OR_10(loop) \
 819                     for (i = 0; i < height; i++) { \
 820                         for (j = 0; j < length; j++) { \
 821                             loop; \
 822                         } \
 823                         dstPtr2 += dstStride[plane] / 2; \
 824                         srcPtr2 += srcStride[plane] / 2; \
 825                     }
 826 #define COPY9_OR_10TO9_OR_10_2(rfunc, wfunc) \
 827                     if (dst_depth > src_depth) { \
 828                         COPY9_OR_10TO9_OR_10(int srcpx = rfunc(&srcPtr2[j]); \
 829                             wfunc(&dstPtr2[j], (srcpx << 1) | (srcpx >> 9))); \
 830                     } else if (dst_depth < src_depth) { \
 831                         DITHER_COPY(dstPtr2, dstStride[plane] / 2, wfunc, \
 832                                     srcPtr2, srcStride[plane] / 2, rfunc, \
 833                                     dither_8x8_1, 1, clip9); \
 834                     } else { \
 835                         COPY9_OR_10TO9_OR_10(wfunc(&dstPtr2[j], rfunc(&srcPtr2[j]))); \
 836                     }
 837                     if (isBE(c->dstFormat)) {
 838                         if (isBE(c->srcFormat)) {
 839                             COPY9_OR_10TO9_OR_10_2(AV_RB16, AV_WB16);
 840                         } else {
 841                             COPY9_OR_10TO9_OR_10_2(AV_RL16, AV_WB16);
 842                         }
 843                     } else {
 844                         if (isBE(c->srcFormat)) {
 845                             COPY9_OR_10TO9_OR_10_2(AV_RB16, AV_WL16);
 846                         } else {
 847                             COPY9_OR_10TO9_OR_10_2(AV_RL16, AV_WL16);
 848                         }
 849                     }
 850                 } else {
 851 #define W8(a, b) { *(a) = (b); }
 852 #define COPY9_OR_10TO8(rfunc) \
 853                     if (src_depth == 9) { \
 854                         DITHER_COPY(dstPtr,  dstStride[plane],   W8, \
 855                                     srcPtr2, srcStride[plane] / 2, rfunc, \
 856                                     dither_8x8_1, 1, av_clip_uint8); \
 857                     } else { \
 858                         DITHER_COPY(dstPtr,  dstStride[plane],   W8, \
 859                                     srcPtr2, srcStride[plane] / 2, rfunc, \
 860                                     dither_8x8_3, 2, av_clip_uint8); \
 861                     }
 862                     if (isBE(c->srcFormat)) {
 863                         COPY9_OR_10TO8(AV_RB16);
 864                     } else {
 865                         COPY9_OR_10TO8(AV_RL16);
 866                     }
 867                 }
 868             } else if (is9_OR_10BPS(c->dstFormat)) {
 869                 const int dst_depth = desc_dst->comp[plane].depth_minus1 + 1;
 870                 uint16_t *dstPtr2 = (uint16_t *) dstPtr;
 871
 872                 if (is16BPS(c->srcFormat)) {
 873                     const uint16_t *srcPtr2 = (const uint16_t *) srcPtr;
 874 #define COPY16TO9_OR_10(rfunc, wfunc) \
 875                     if (dst_depth == 9) { \
 876                         DITHER_COPY(dstPtr2, dstStride[plane] / 2, wfunc, \
 877                                     srcPtr2, srcStride[plane] / 2, rfunc, \
 878                                     dither_8x8_128, 7, clip9); \
 879                     } else { \
 880                         DITHER_COPY(dstPtr2, dstStride[plane] / 2, wfunc, \
 881                                     srcPtr2, srcStride[plane] / 2, rfunc, \
 882                                     dither_8x8_64, 6, clip10); \
 883                     }
 884                     if (isBE(c->dstFormat)) {
 885                         if (isBE(c->srcFormat)) {
 886                             COPY16TO9_OR_10(AV_RB16, AV_WB16);
 887                         } else {
 888                             COPY16TO9_OR_10(AV_RL16, AV_WB16);
 889                         }
 890                     } else {
 891                         if (isBE(c->srcFormat)) {
 892                             COPY16TO9_OR_10(AV_RB16, AV_WL16);
 893                         } else {
 894                             COPY16TO9_OR_10(AV_RL16, AV_WL16);
 895                         }
 896                     }
 897                 } else /* 8bit */ {
 898 #define COPY8TO9_OR_10(wfunc) \
 899                     for (i = 0; i < height; i++) { \
 900                         for (j = 0; j < length; j++) { \
 901                             const int srcpx = srcPtr[j]; \
 902                             wfunc(&dstPtr2[j], (srcpx << (dst_depth - 8)) | (srcpx >> (16 - dst_depth))); \
 903                         } \
 904                         dstPtr2 += dstStride[plane] / 2; \
 905                         srcPtr  += srcStride[plane]; \
 906                     }
 907                     if (isBE(c->dstFormat)) {
 908                         COPY8TO9_OR_10(AV_WB16);
 909                     } else {
 910                         COPY8TO9_OR_10(AV_WL16);
 911                     }
 912                 }
 913             } else if (is16BPS(c->srcFormat) && !is16BPS(c->dstFormat)) {
 914                 const uint16_t *srcPtr2 = (const uint16_t *) srcPtr;
 915 #define COPY16TO8(rfunc) \
 916                     DITHER_COPY(dstPtr,  dstStride[plane],   W8, \
 917                                 srcPtr2, srcStride[plane] / 2, rfunc, \
 918                                 dither_8x8_256, 8, av_clip_uint8);
 919                 if (isBE(c->srcFormat)) {
 920                     COPY16TO8(AV_RB16);
 921                 } else {
 922                     COPY16TO8(AV_RL16);
 923                 }
 924             } else if (!is16BPS(c->srcFormat) && is16BPS(c->dstFormat)) {
 925                 for (i = 0; i < height; i++) {
 926                     for (j = 0; j < length; j++) {
 927                         dstPtr[ j << 1     ] = srcPtr[j];
 928                         dstPtr[(j << 1) + 1] = srcPtr[j];
 929                     }
 930                     srcPtr += srcStride[plane];
 931                     dstPtr += dstStride[plane];
 932                 }
 933             } else if (is16BPS(c->srcFormat) && is16BPS(c->dstFormat) &&
 934                       isBE(c->srcFormat) != isBE(c->dstFormat)) {
 935
 936                 for (i = 0; i < height; i++) {
 937                     for (j = 0; j < length; j++)
 938                         ((uint16_t *) dstPtr)[j] = av_bswap16(((const uint16_t *) srcPtr)[j]);
 939                     srcPtr += srcStride[plane];
 940                     dstPtr += dstStride[plane];
 941                 }
 942             } else if (dstStride[plane] == srcStride[plane] &&
 943                        srcStride[plane] > 0 && srcStride[plane] == length) {
 944                 memcpy(dst[plane] + dstStride[plane] * y, src[plane],
 945                        height * dstStride[plane]);
 946             } else {
 947                 if (is16BPS(c->srcFormat) && is16BPS(c->dstFormat))
 948                     length *= 2;
 949                 else if (!desc_src->comp[0].depth_minus1)
 950                     length >>= 3; // monowhite/black
 951                 for (i = 0; i < height; i++) {
 952                     memcpy(dstPtr, srcPtr, length);
 953                     srcPtr += srcStride[plane];
 954                     dstPtr += dstStride[plane];
 955                 }
 956             }
 957         }
 958     }
 959     return srcSliceH;
 960 }
 961
 962
 963 #define IS_DIFFERENT_ENDIANESS(src_fmt, dst_fmt, pix_fmt)          \
 964     ((src_fmt == pix_fmt ## BE && dst_fmt == pix_fmt ## LE) ||     \
 965      (src_fmt == pix_fmt ## LE && dst_fmt == pix_fmt ## BE))
 966
 967
 968 void ff_get_unscaled_swscale(SwsContext *c)
 969 {
 970     const enum AVPixelFormat srcFormat = c->srcFormat;
 971     const enum AVPixelFormat dstFormat = c->dstFormat;
 972     const int flags = c->flags;
 973     const int dstH = c->dstH;
 974     int needsDither;
 975
 976     needsDither = isAnyRGB(dstFormat) &&
 977             c->dstFormatBpp < 24 &&
 978            (c->dstFormatBpp < c->srcFormatBpp || (!isAnyRGB(srcFormat)));
 979
 980     /* yv12_to_nv12 */
 981     if ((srcFormat == AV_PIX_FMT_YUV420P || srcFormat == AV_PIX_FMT_YUVA420P) &&
 982         (dstFormat == AV_PIX_FMT_NV12 || dstFormat == AV_PIX_FMT_NV21)) {
 983         c->swScale = planarToNv12Wrapper;
 984     }
 985     /* yuv2bgr */
 986     if ((srcFormat == AV_PIX_FMT_YUV420P || srcFormat == AV_PIX_FMT_YUV422P ||
 987          srcFormat == AV_PIX_FMT_YUVA420P) && isAnyRGB(dstFormat) &&
 988         !(flags & SWS_ACCURATE_RND) && !(dstH & 1)) {
 989         c->swScale = ff_yuv2rgb_get_func_ptr(c);
 990     }
 991
 992     if (srcFormat == AV_PIX_FMT_YUV410P &&
 993         (dstFormat == AV_PIX_FMT_YUV420P || dstFormat == AV_PIX_FMT_YUVA420P) &&
 994         !(flags & SWS_BITEXACT)) {
 995         c->swScale = yvu9ToYv12Wrapper;
 996     }
 997
 998     /* bgr24toYV12 */
 999     if (srcFormat == AV_PIX_FMT_BGR24 &&
1000         (dstFormat == AV_PIX_FMT_YUV420P || dstFormat == AV_PIX_FMT_YUVA420P) &&
1001         !(flags & SWS_ACCURATE_RND))
1002         c->swScale = bgr24ToYv12Wrapper;
1003
1004     /* RGB/BGR -> RGB/BGR (no dither needed forms) */
1005     if (isAnyRGB(srcFormat) && isAnyRGB(dstFormat) && findRgbConvFn(c)
1006         && (!needsDither || (c->flags&(SWS_FAST_BILINEAR|SWS_POINT))))
1007         c->swScale= rgbToRgbWrapper;
1008
1009 #define isByteRGB(f) (             \
1010         f == AV_PIX_FMT_RGB32   || \
1011         f == AV_PIX_FMT_RGB32_1 || \
1012         f == AV_PIX_FMT_RGB24   || \
1013         f == AV_PIX_FMT_BGR32   || \
1014         f == AV_PIX_FMT_BGR32_1 || \
1015         f == AV_PIX_FMT_BGR24)
1016
1017     if (srcFormat == AV_PIX_FMT_GBRP && isPlanar(srcFormat) && isByteRGB(dstFormat))
1018         c->swScale = planarRgbToRgbWrapper;
1019
1020     if (av_pix_fmt_desc_get(srcFormat)->comp[0].depth_minus1 == 7 &&
1021         isPackedRGB(srcFormat) && dstFormat == AV_PIX_FMT_GBRP)
1022         c->swScale = rgbToPlanarRgbWrapper;
1023
1024     /* bswap 16 bits per pixel/component packed formats */
1025     if (IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_BGR444) ||
1026         IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_BGR48)  ||
1027         IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_BGR555) ||
1028         IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_BGR565) ||
1029         IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_GRAY16) ||
1030         IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_RGB444) ||
1031         IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_RGB48)  ||
1032         IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_RGB555) ||
1033         IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_RGB565))
1034         c->swScale = packed_16bpc_bswap;
1035
1036     if ((usePal(srcFormat) && (
1037         dstFormat == AV_PIX_FMT_RGB32   ||
1038         dstFormat == AV_PIX_FMT_RGB32_1 ||
1039         dstFormat == AV_PIX_FMT_RGB24   ||
1040         dstFormat == AV_PIX_FMT_BGR32   ||
1041         dstFormat == AV_PIX_FMT_BGR32_1 ||
1042         dstFormat == AV_PIX_FMT_BGR24)))
1043         c->swScale = palToRgbWrapper;
1044
1045     if (srcFormat == AV_PIX_FMT_YUV422P) {
1046         if (dstFormat == AV_PIX_FMT_YUYV422)
1047             c->swScale = yuv422pToYuy2Wrapper;
1048         else if (dstFormat == AV_PIX_FMT_UYVY422)
1049             c->swScale = yuv422pToUyvyWrapper;
1050     }
1051
1052     /* LQ converters if -sws 0 or -sws 4*/
1053     if (c->flags&(SWS_FAST_BILINEAR|SWS_POINT)) {
1054         /* yv12_to_yuy2 */
1055         if (srcFormat == AV_PIX_FMT_YUV420P || srcFormat == AV_PIX_FMT_YUVA420P) {
1056             if (dstFormat == AV_PIX_FMT_YUYV422)
1057                 c->swScale = planarToYuy2Wrapper;
1058             else if (dstFormat == AV_PIX_FMT_UYVY422)
1059                 c->swScale = planarToUyvyWrapper;
1060         }
1061     }
1062     if (srcFormat == AV_PIX_FMT_YUYV422 &&
1063        (dstFormat == AV_PIX_FMT_YUV420P || dstFormat == AV_PIX_FMT_YUVA420P))
1064         c->swScale = yuyvToYuv420Wrapper;
1065     if (srcFormat == AV_PIX_FMT_UYVY422 &&
1066        (dstFormat == AV_PIX_FMT_YUV420P || dstFormat == AV_PIX_FMT_YUVA420P))
1067         c->swScale = uyvyToYuv420Wrapper;
1068     if (srcFormat == AV_PIX_FMT_YUYV422 && dstFormat == AV_PIX_FMT_YUV422P)
1069         c->swScale = yuyvToYuv422Wrapper;
1070     if (srcFormat == AV_PIX_FMT_UYVY422 && dstFormat == AV_PIX_FMT_YUV422P)
1071         c->swScale = uyvyToYuv422Wrapper;
1072
1073     /* simple copy */
1074     if ( srcFormat == dstFormat ||
1075         (srcFormat == AV_PIX_FMT_YUVA420P && dstFormat == AV_PIX_FMT_YUV420P) ||
1076         (srcFormat == AV_PIX_FMT_YUV420P && dstFormat == AV_PIX_FMT_YUVA420P) ||
1077         (isPlanarYUV(srcFormat) && isGray(dstFormat)) ||
1078         (isPlanarYUV(dstFormat) && isGray(srcFormat)) ||
1079         (isGray(dstFormat) && isGray(srcFormat)) ||
1080         (isPlanarYUV(srcFormat) && isPlanarYUV(dstFormat) &&
1081          c->chrDstHSubSample == c->chrSrcHSubSample &&
1082          c->chrDstVSubSample == c->chrSrcVSubSample &&
1083          dstFormat != AV_PIX_FMT_NV12 && dstFormat != AV_PIX_FMT_NV21 &&
1084          srcFormat != AV_PIX_FMT_NV12 && srcFormat != AV_PIX_FMT_NV21))
1085     {
1086         if (isPacked(c->srcFormat))
1087             c->swScale = packedCopyWrapper;
1088         else /* Planar YUV or gray */
1089             c->swScale = planarCopyWrapper;
1090     }
1091
1092     if (ARCH_BFIN)
1093         ff_bfin_get_unscaled_swscale(c);
1094     if (HAVE_ALTIVEC)
1095         ff_swscale_get_unscaled_altivec(c);
1096 }
1097
1098 static void reset_ptr(const uint8_t *src[], int format)
1099 {
1100     if (!isALPHA(format))
1101         src[3] = NULL;
1102     if (!isPlanar(format)) {
1103         src[3] = src[2] = NULL;
1104
1105         if (!usePal(format))
1106             src[1] = NULL;
1107     }
1108 }
1109
1110 static int check_image_pointers(uint8_t *data[4], enum AVPixelFormat pix_fmt,
1111                                 const int linesizes[4])
1112 {
1113     const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(pix_fmt);
1114     int i;
1115
1116     for (i = 0; i < 4; i++) {
1117         int plane = desc->comp[i].plane;
1118         if (!data[plane] || !linesizes[plane])
1119             return 0;
1120     }
1121
1122     return 1;
1123 }
1124
1125 /**
1126  * swscale wrapper, so we don't need to export the SwsContext.
1127  * Assumes planar YUV to be in YUV order instead of YVU.
1128  */
1129 int attribute_align_arg sws_scale(struct SwsContext *c,
1130                                   const uint8_t * const srcSlice[],
1131                                   const int srcStride[], int srcSliceY,
1132                                   int srcSliceH, uint8_t *const dst[],
1133                                   const int dstStride[])
1134 {
1135     int i;
1136     const uint8_t *src2[4] = { srcSlice[0], srcSlice[1], srcSlice[2], srcSlice[3] };
1137     uint8_t *dst2[4] = { dst[0], dst[1], dst[2], dst[3] };
1138
1139     // do not mess up sliceDir if we have a "trailing" 0-size slice
1140     if (srcSliceH == 0)
1141         return 0;
1142
1143     if (!check_image_pointers(srcSlice, c->srcFormat, srcStride)) {
1144         av_log(c, AV_LOG_ERROR, "bad src image pointers\n");
1145         return 0;
1146     }
1147     if (!check_image_pointers(dst, c->dstFormat, dstStride)) {
1148         av_log(c, AV_LOG_ERROR, "bad dst image pointers\n");
1149         return 0;
1150     }
1151
1152     if (c->sliceDir == 0 && srcSliceY != 0 && srcSliceY + srcSliceH != c->srcH) {
1153         av_log(c, AV_LOG_ERROR, "Slices start in the middle!\n");
1154         return 0;
1155     }
1156     if (c->sliceDir == 0) {
1157         if (srcSliceY == 0) c->sliceDir = 1; else c->sliceDir = -1;
1158     }
1159
1160     if (usePal(c->srcFormat)) {
1161         for (i = 0; i < 256; i++) {
1162             int p, r, g, b, y, u, v;
1163             if (c->srcFormat == AV_PIX_FMT_PAL8) {
1164                 p = ((const uint32_t *)(srcSlice[1]))[i];
1165                 r = (p >> 16) & 0xFF;
1166                 g = (p >>  8) & 0xFF;
1167                 b =  p        & 0xFF;
1168             } else if (c->srcFormat == AV_PIX_FMT_RGB8) {
1169                 r = ( i >> 5     ) * 36;
1170                 g = ((i >> 2) & 7) * 36;
1171                 b = ( i       & 3) * 85;
1172             } else if (c->srcFormat == AV_PIX_FMT_BGR8) {
1173                 b = ( i >> 6     ) * 85;
1174                 g = ((i >> 3) & 7) * 36;
1175                 r = ( i       & 7) * 36;
1176             } else if (c->srcFormat == AV_PIX_FMT_RGB4_BYTE) {
1177                 r = ( i >> 3     ) * 255;
1178                 g = ((i >> 1) & 3) * 85;
1179                 b = ( i       & 1) * 255;
1180             } else if (c->srcFormat == AV_PIX_FMT_GRAY8 ||
1181                       c->srcFormat == AV_PIX_FMT_Y400A) {
1182                 r = g = b = i;
1183             } else {
1184                 assert(c->srcFormat == AV_PIX_FMT_BGR4_BYTE);
1185                 b = ( i >> 3     ) * 255;
1186                 g = ((i >> 1) & 3) * 85;
1187                 r = ( i       & 1) * 255;
1188             }
1189             y = av_clip_uint8((RY * r + GY * g + BY * b + ( 33 << (RGB2YUV_SHIFT - 1))) >> RGB2YUV_SHIFT);
1190             u = av_clip_uint8((RU * r + GU * g + BU * b + (257 << (RGB2YUV_SHIFT - 1))) >> RGB2YUV_SHIFT);
1191             v = av_clip_uint8((RV * r + GV * g + BV * b + (257 << (RGB2YUV_SHIFT - 1))) >> RGB2YUV_SHIFT);
1192             c->pal_yuv[i] = y + (u << 8) + (v << 16);
1193
1194             switch (c->dstFormat) {
1195             case AV_PIX_FMT_BGR32:
1196 #if !HAVE_BIGENDIAN
1197             case AV_PIX_FMT_RGB24:
1198 #endif
1199                 c->pal_rgb[i] =  r + (g << 8) + (b << 16);
1200                 break;
1201             case AV_PIX_FMT_BGR32_1:
1202 #if HAVE_BIGENDIAN
1203             case AV_PIX_FMT_BGR24:
1204 #endif
1205                 c->pal_rgb[i] = (r + (g << 8) + (b << 16)) << 8;
1206                 break;
1207             case AV_PIX_FMT_RGB32_1:
1208 #if HAVE_BIGENDIAN
1209             case AV_PIX_FMT_RGB24:
1210 #endif
1211                 c->pal_rgb[i] = (b + (g << 8) + (r << 16)) << 8;
1212                 break;
1213             case AV_PIX_FMT_RGB32:
1214 #if !HAVE_BIGENDIAN
1215             case AV_PIX_FMT_BGR24:
1216 #endif
1217             default:
1218                 c->pal_rgb[i] =  b + (g << 8) + (r << 16);
1219             }
1220         }
1221     }
1222
1223     // copy strides, so they can safely be modified
1224     if (c->sliceDir == 1) {
1225         // slices go from top to bottom
1226         int srcStride2[4] = { srcStride[0], srcStride[1], srcStride[2],
1227                               srcStride[3] };
1228         int dstStride2[4] = { dstStride[0], dstStride[1], dstStride[2],
1229                               dstStride[3] };
1230
1231         reset_ptr(src2, c->srcFormat);
1232         reset_ptr((const uint8_t **) dst2, c->dstFormat);
1233
1234         /* reset slice direction at end of frame */
1235         if (srcSliceY + srcSliceH == c->srcH)
1236             c->sliceDir = 0;
1237
1238         return c->swScale(c, src2, srcStride2, srcSliceY, srcSliceH, dst2,
1239                           dstStride2);
1240     } else {
1241         // slices go from bottom to top => we flip the image internally
1242         int srcStride2[4] = { -srcStride[0], -srcStride[1], -srcStride[2],
1243                               -srcStride[3] };
1244         int dstStride2[4] = { -dstStride[0], -dstStride[1], -dstStride[2],
1245                               -dstStride[3] };
1246
1247         src2[0] += (srcSliceH - 1) * srcStride[0];
1248         if (!usePal(c->srcFormat))
1249             src2[1] += ((srcSliceH >> c->chrSrcVSubSample) - 1) * srcStride[1];
1250         src2[2] += ((srcSliceH >> c->chrSrcVSubSample) - 1) * srcStride[2];
1251         src2[3] += (srcSliceH - 1) * srcStride[3];
1252         dst2[0] += ( c->dstH                         - 1) * dstStride[0];
1253         dst2[1] += ((c->dstH >> c->chrDstVSubSample) - 1) * dstStride[1];
1254         dst2[2] += ((c->dstH >> c->chrDstVSubSample) - 1) * dstStride[2];
1255         dst2[3] += ( c->dstH                         - 1) * dstStride[3];
1256
1257         reset_ptr(src2, c->srcFormat);
1258         reset_ptr((const uint8_t **) dst2, c->dstFormat);
1259
1260         /* reset slice direction at end of frame */
1261         if (!srcSliceY)
1262             c->sliceDir = 0;
1263
1264         return c->swScale(c, src2, srcStride2, c->srcH-srcSliceY-srcSliceH,
1265                           srcSliceH, dst2, dstStride2);
1266     }
1267 }
1268
1269 /* Convert the palette to the same packed 32-bit format as the palette */
1270 void sws_convertPalette8ToPacked32(const uint8_t *src, uint8_t *dst,
1271                                    int num_pixels, const uint8_t *palette)
1272 {
1273     int i;
1274
1275     for (i = 0; i < num_pixels; i++)
1276         ((uint32_t *) dst)[i] = ((const uint32_t *) palette)[src[i]];
1277 }
1278
1279 /* Palette format: ABCD -> dst format: ABC */
1280 void sws_convertPalette8ToPacked24(const uint8_t *src, uint8_t *dst,
1281                                    int num_pixels, const uint8_t *palette)
1282 {
1283     int i;
1284
1285     for (i = 0; i < num_pixels; i++) {
1286         //FIXME slow?
1287         dst[0] = palette[src[i] * 4 + 0];
1288         dst[1] = palette[src[i] * 4 + 1];
1289         dst[2] = palette[src[i] * 4 + 2];
1290         dst += 3;
1291     }
1292 }