falkTX
/
FFmpeg
mirror of https://github.com/falkTX/FFmpeg.git
1
0
Fork 0
Code Issues 0 Releases 338 Wiki Activity
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
41000 Commits
32 Branches
325MB
Tree: 33c827f632
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from '33c827f632'
${ noResults }
FFmpeg/libswscale/swscale_unscaled.c

1343 lines
52KB
Raw Blame History 

  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, static 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, static 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, static 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. DECLARE_ALIGNED(8, static const uint8_t, dither_8x8_256)[8][8] = {

  68.     {  72, 136, 120, 184,  68, 132, 116, 180,},

  69.     { 200,   8, 248,  56, 196,   4, 244,  52,},

  70.     { 104, 168,  88, 152, 100, 164,  84, 148,},

  71.     { 232,  40, 216,  24, 228,  36, 212,  20,},

  72.     {  64, 128, 102, 176,  76, 140, 124, 188,},

  73.     { 192,   0, 240,  48, 204,  12, 252,  60,},

  74.     {  96, 160,  80, 144, 108, 172,  92, 156,},

  75.     { 224,  32, 208,  16, 236,  44, 220,  28,},

  76. };

  77. 

  78. #define RGB2YUV_SHIFT 15

  79. #define BY ( (int) (0.114 * 219 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))

  80. #define BV (-(int) (0.081 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))

  81. #define BU ( (int) (0.500 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))

  82. #define GY ( (int) (0.587 * 219 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))

  83. #define GV (-(int) (0.419 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))

  84. #define GU (-(int) (0.331 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))

  85. #define RY ( (int) (0.299 * 219 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))

  86. #define RV ( (int) (0.500 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))

  87. #define RU (-(int) (0.169 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5))

  88. 

  89. static void fillPlane(uint8_t *plane, int stride, int width, int height, int y,

  90.                       uint8_t val)

  91. {

  92.     int i;

  93.     uint8_t *ptr = plane + stride * y;

  94.     for (i = 0; i < height; i++) {

  95.         memset(ptr, val, width);

  96.         ptr += stride;

  97.     }

  98. }

  99. 

  100. static void fill_plane9or10(uint8_t *plane, int stride, int width,

  101.                             int height, int y, uint8_t val,

  102.                             const int dst_depth, const int big_endian)

  103. {

  104.     int i, j;

  105.     uint16_t *dst = (uint16_t *) (plane + stride * y);

  106. #define FILL8TO9_OR_10(wfunc) \

  107.     for (i = 0; i < height; i++) { \

  108.         for (j = 0; j < width; j++) { \

  109.             wfunc(&dst[j], (val << (dst_depth - 8)) |  \

  110.                                (val >> (16 - dst_depth))); \

  111.         } \

  112.         dst += stride / 2; \

  113.     }

  114.     if (big_endian) {

  115.         FILL8TO9_OR_10(AV_WB16);

  116.     } else {

  117.         FILL8TO9_OR_10(AV_WL16);

  118.     }

  119. }

  120. 

  121. static void copyPlane(const uint8_t *src, int srcStride,

  122.                       int srcSliceY, int srcSliceH, int width,

  123.                       uint8_t *dst, int dstStride)

  124. {

  125.     dst += dstStride * srcSliceY;

  126.     if (dstStride == srcStride && srcStride > 0) {

  127.         memcpy(dst, src, srcSliceH * dstStride);

  128.     } else {

  129.         int i;

  130.         for (i = 0; i < srcSliceH; i++) {

  131.             memcpy(dst, src, width);

  132.             src += srcStride;

  133.             dst += dstStride;

  134.         }

  135.     }

  136. }

  137. 

  138. static int planarToNv12Wrapper(SwsContext *c, const uint8_t *src[],

  139.                                int srcStride[], int srcSliceY,

  140.                                int srcSliceH, uint8_t *dstParam[],

  141.                                int dstStride[])

  142. {

  143.     uint8_t *dst = dstParam[1] + dstStride[1] * srcSliceY / 2;

  144. 

  145.     copyPlane(src[0], srcStride[0], srcSliceY, srcSliceH, c->srcW,

  146.               dstParam[0], dstStride[0]);

  147. 

  148.     if (c->dstFormat == AV_PIX_FMT_NV12)

  149.         interleaveBytes(src[1], src[2], dst, c->srcW / 2, srcSliceH / 2,

  150.                         srcStride[1], srcStride[2], dstStride[1]);

  151.     else

  152.         interleaveBytes(src[2], src[1], dst, c->srcW / 2, srcSliceH / 2,

  153.                         srcStride[2], srcStride[1], dstStride[1]);

  154. 

  155.     return srcSliceH;

  156. }

  157. 

  158. static int nv12ToPlanarWrapper(SwsContext *c, const uint8_t *src[],

  159.                                int srcStride[], int srcSliceY,

  160.                                int srcSliceH, uint8_t *dstParam[],

  161.                                int dstStride[])

  162. {

  163.     uint8_t *dst1 = dstParam[1] + dstStride[1] * srcSliceY / 2;

  164.     uint8_t *dst2 = dstParam[2] + dstStride[2] * srcSliceY / 2;

  165. 

  166.     copyPlane(src[0], srcStride[0], srcSliceY, srcSliceH, c->srcW,

  167.               dstParam[0], dstStride[0]);

  168. 

  169.     if (c->srcFormat == AV_PIX_FMT_NV12)

  170.         deinterleaveBytes(src[1], dst1, dst2,c->srcW / 2, srcSliceH / 2,

  171.                           srcStride[1], dstStride[1], dstStride[2]);

  172.     else

  173.         deinterleaveBytes(src[1], dst2, dst1, c->srcW / 2, srcSliceH / 2,

  174.                           srcStride[1], dstStride[2], dstStride[1]);

  175. 

  176.     return srcSliceH;

  177. }

  178. 

  179. static int planarToYuy2Wrapper(SwsContext *c, const uint8_t *src[],

  180.                                int srcStride[], int srcSliceY, int srcSliceH,

  181.                                uint8_t *dstParam[], int dstStride[])

  182. {

  183.     uint8_t *dst = dstParam[0] + dstStride[0] * srcSliceY;

  184. 

  185.     yv12toyuy2(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0],

  186.                srcStride[1], dstStride[0]);

  187. 

  188.     return srcSliceH;

  189. }

  190. 

  191. static int planarToUyvyWrapper(SwsContext *c, const uint8_t *src[],

  192.                                int srcStride[], int srcSliceY, int srcSliceH,

  193.                                uint8_t *dstParam[], int dstStride[])

  194. {

  195.     uint8_t *dst = dstParam[0] + dstStride[0] * srcSliceY;

  196. 

  197.     yv12touyvy(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0],

  198.                srcStride[1], dstStride[0]);

  199. 

  200.     return srcSliceH;

  201. }

  202. 

  203. static int yuv422pToYuy2Wrapper(SwsContext *c, const uint8_t *src[],

  204.                                 int srcStride[], int srcSliceY, int srcSliceH,

  205.                                 uint8_t *dstParam[], int dstStride[])

  206. {

  207.     uint8_t *dst = dstParam[0] + dstStride[0] * srcSliceY;

  208. 

  209.     yuv422ptoyuy2(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0],

  210.                   srcStride[1], dstStride[0]);

  211. 

  212.     return srcSliceH;

  213. }

  214. 

  215. static int yuv422pToUyvyWrapper(SwsContext *c, const uint8_t *src[],

  216.                                 int srcStride[], int srcSliceY, int srcSliceH,

  217.                                 uint8_t *dstParam[], int dstStride[])

  218. {

  219.     uint8_t *dst = dstParam[0] + dstStride[0] * srcSliceY;

  220. 

  221.     yuv422ptouyvy(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0],

  222.                   srcStride[1], dstStride[0]);

  223. 

  224.     return srcSliceH;

  225. }

  226. 

  227. static int yuyvToYuv420Wrapper(SwsContext *c, const uint8_t *src[],

  228.                                int srcStride[], int srcSliceY, int srcSliceH,

  229.                                uint8_t *dstParam[], int dstStride[])

  230. {

  231.     uint8_t *ydst = dstParam[0] + dstStride[0] * srcSliceY;

  232.     uint8_t *udst = dstParam[1] + dstStride[1] * srcSliceY / 2;

  233.     uint8_t *vdst = dstParam[2] + dstStride[2] * srcSliceY / 2;

  234. 

  235.     yuyvtoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0],

  236.                  dstStride[1], srcStride[0]);

  237. 

  238.     if (dstParam[3])

  239.         fillPlane(dstParam[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);

  240. 

  241.     return srcSliceH;

  242. }

  243. 

  244. static int yuyvToYuv422Wrapper(SwsContext *c, const uint8_t *src[],

  245.                                int srcStride[], int srcSliceY, int srcSliceH,

  246.                                uint8_t *dstParam[], int dstStride[])

  247. {

  248.     uint8_t *ydst = dstParam[0] + dstStride[0] * srcSliceY;

  249.     uint8_t *udst = dstParam[1] + dstStride[1] * srcSliceY;

  250.     uint8_t *vdst = dstParam[2] + dstStride[2] * srcSliceY;

  251. 

  252.     yuyvtoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0],

  253.                  dstStride[1], srcStride[0]);

  254. 

  255.     return srcSliceH;

  256. }

  257. 

  258. static int uyvyToYuv420Wrapper(SwsContext *c, const uint8_t *src[],

  259.                                int srcStride[], int srcSliceY, int srcSliceH,

  260.                                uint8_t *dstParam[], int dstStride[])

  261. {

  262.     uint8_t *ydst = dstParam[0] + dstStride[0] * srcSliceY;

  263.     uint8_t *udst = dstParam[1] + dstStride[1] * srcSliceY / 2;

  264.     uint8_t *vdst = dstParam[2] + dstStride[2] * srcSliceY / 2;

  265. 

  266.     uyvytoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0],

  267.                  dstStride[1], srcStride[0]);

  268. 

  269.     if (dstParam[3])

  270.         fillPlane(dstParam[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);

  271. 

  272.     return srcSliceH;

  273. }

  274. 

  275. static int uyvyToYuv422Wrapper(SwsContext *c, const uint8_t *src[],

  276.                                int srcStride[], int srcSliceY, int srcSliceH,

  277.                                uint8_t *dstParam[], int dstStride[])

  278. {

  279.     uint8_t *ydst = dstParam[0] + dstStride[0] * srcSliceY;

  280.     uint8_t *udst = dstParam[1] + dstStride[1] * srcSliceY;

  281.     uint8_t *vdst = dstParam[2] + dstStride[2] * srcSliceY;

  282. 

  283.     uyvytoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0],

  284.                  dstStride[1], srcStride[0]);

  285. 

  286.     return srcSliceH;

  287. }

  288. 

  289. static void gray8aToPacked32(const uint8_t *src, uint8_t *dst, int num_pixels,

  290.                              const uint8_t *palette)

  291. {

  292.     int i;

  293.     for (i = 0; i < num_pixels; i++)

  294.         ((uint32_t *) dst)[i] = ((const uint32_t *) palette)[src[i << 1]] | (src[(i << 1) + 1] << 24);

  295. }

  296. 

  297. static void gray8aToPacked32_1(const uint8_t *src, uint8_t *dst, int num_pixels,

  298.                                const uint8_t *palette)

  299. {

  300.     int i;

  301. 

  302.     for (i = 0; i < num_pixels; i++)

  303.         ((uint32_t *) dst)[i] = ((const uint32_t *) palette)[src[i << 1]] | src[(i << 1) + 1];

  304. }

  305. 

  306. static void gray8aToPacked24(const uint8_t *src, uint8_t *dst, int num_pixels,

  307.                              const uint8_t *palette)

  308. {

  309.     int i;

  310. 

  311.     for (i = 0; i < num_pixels; i++) {

  312.         //FIXME slow?

  313.         dst[0] = palette[src[i << 1] * 4 + 0];

  314.         dst[1] = palette[src[i << 1] * 4 + 1];

  315.         dst[2] = palette[src[i << 1] * 4 + 2];

  316.         dst += 3;

  317.     }

  318. }

  319. 

  320. static int packed_16bpc_bswap(SwsContext *c, const uint8_t *src[],

  321.                               int srcStride[], int srcSliceY, int srcSliceH,

  322.                               uint8_t *dst[], int dstStride[])

  323. {

  324.     int i, j;

  325.     int srcstr = srcStride[0] >> 1;

  326.     int dststr = dstStride[0] >> 1;

  327.     uint16_t       *dstPtr =       (uint16_t *) dst[0];

  328.     const uint16_t *srcPtr = (const uint16_t *) src[0];

  329.     int min_stride         = FFMIN(srcstr, dststr);

  330. 

  331.     for (i = 0; i < srcSliceH; i++) {

  332.         for (j = 0; j < min_stride; j++) {

  333.             dstPtr[j] = av_bswap16(srcPtr[j]);

  334.         }

  335.         srcPtr += srcstr;

  336.         dstPtr += dststr;

  337.     }

  338. 

  339.     return srcSliceH;

  340. }

  341. 

  342. static int palToRgbWrapper(SwsContext *c, const uint8_t *src[], int srcStride[],

  343.                            int srcSliceY, int srcSliceH, uint8_t *dst[],

  344.                            int dstStride[])

  345. {

  346.     const enum AVPixelFormat srcFormat = c->srcFormat;

  347.     const enum AVPixelFormat dstFormat = c->dstFormat;

  348.     void (*conv)(const uint8_t *src, uint8_t *dst, int num_pixels,

  349.                  const uint8_t *palette) = NULL;

  350.     int i;

  351.     uint8_t *dstPtr = dst[0] + dstStride[0] * srcSliceY;

  352.     const uint8_t *srcPtr = src[0];

  353. 

  354.     if (srcFormat == AV_PIX_FMT_YA8) {

  355.         switch (dstFormat) {

  356.         case AV_PIX_FMT_RGB32  : conv = gray8aToPacked32; break;

  357.         case AV_PIX_FMT_BGR32  : conv = gray8aToPacked32; break;

  358.         case AV_PIX_FMT_BGR32_1: conv = gray8aToPacked32_1; break;

  359.         case AV_PIX_FMT_RGB32_1: conv = gray8aToPacked32_1; break;

  360.         case AV_PIX_FMT_RGB24  : conv = gray8aToPacked24; break;

  361.         case AV_PIX_FMT_BGR24  : conv = gray8aToPacked24; break;

  362.         }

  363.     } else if (usePal(srcFormat)) {

  364.         switch (dstFormat) {

  365.         case AV_PIX_FMT_RGB32  : conv = sws_convertPalette8ToPacked32; break;

  366.         case AV_PIX_FMT_BGR32  : conv = sws_convertPalette8ToPacked32; break;

  367.         case AV_PIX_FMT_BGR32_1: conv = sws_convertPalette8ToPacked32; break;

  368.         case AV_PIX_FMT_RGB32_1: conv = sws_convertPalette8ToPacked32; break;

  369.         case AV_PIX_FMT_RGB24  : conv = sws_convertPalette8ToPacked24; break;

  370.         case AV_PIX_FMT_BGR24  : conv = sws_convertPalette8ToPacked24; break;

  371.         }

  372.     }

  373. 

  374.     if (!conv)

  375.         av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",

  376.                sws_format_name(srcFormat), sws_format_name(dstFormat));

  377.     else {

  378.         for (i = 0; i < srcSliceH; i++) {

  379.             conv(srcPtr, dstPtr, c->srcW, (uint8_t *) c->pal_rgb);

  380.             srcPtr += srcStride[0];

  381.             dstPtr += dstStride[0];

  382.         }

  383.     }

  384. 

  385.     return srcSliceH;

  386. }

  387. 

  388. static void gbr24ptopacked24(const uint8_t *src[], int srcStride[],

  389.                              uint8_t *dst, int dstStride, int srcSliceH,

  390.                              int width)

  391. {

  392.     int x, h, i;

  393.     for (h = 0; h < srcSliceH; h++) {

  394.         uint8_t *dest = dst + dstStride * h;

  395.         for (x = 0; x < width; x++) {

  396.             *dest++ = src[0][x];

  397.             *dest++ = src[1][x];

  398.             *dest++ = src[2][x];

  399.         }

  400. 

  401.         for (i = 0; i < 3; i++)

  402.             src[i] += srcStride[i];

  403.     }

  404. }

  405. 

  406. static void gbr24ptopacked32(const uint8_t *src[], int srcStride[],

  407.                              uint8_t *dst, int dstStride, int srcSliceH,

  408.                              int alpha_first, int width)

  409. {

  410.     int x, h, i;

  411.     for (h = 0; h < srcSliceH; h++) {

  412.         uint8_t *dest = dst + dstStride * h;

  413. 

  414.         if (alpha_first) {

  415.             for (x = 0; x < width; x++) {

  416.                 *dest++ = 0xff;

  417.                 *dest++ = src[0][x];

  418.                 *dest++ = src[1][x];

  419.                 *dest++ = src[2][x];

  420.             }

  421.         } else {

  422.             for (x = 0; x < width; x++) {

  423.                 *dest++ = src[0][x];

  424.                 *dest++ = src[1][x];

  425.                 *dest++ = src[2][x];

  426.                 *dest++ = 0xff;

  427.             }

  428.         }

  429. 

  430.         for (i = 0; i < 3; i++)

  431.             src[i] += srcStride[i];

  432.     }

  433. }

  434. 

  435. static int planarRgbToRgbWrapper(SwsContext *c, const uint8_t *src[],

  436.                                  int srcStride[], int srcSliceY, int srcSliceH,

  437.                                  uint8_t *dst[], int dstStride[])

  438. {

  439.     int alpha_first = 0;

  440.     const uint8_t *src102[] = { src[1], src[0], src[2] };

  441.     const uint8_t *src201[] = { src[2], src[0], src[1] };

  442.     int stride102[] = { srcStride[1], srcStride[0], srcStride[2] };

  443.     int stride201[] = { srcStride[2], srcStride[0], srcStride[1] };

  444. 

  445.     if (c->srcFormat != AV_PIX_FMT_GBRP) {

  446.         av_log(c, AV_LOG_ERROR, "unsupported planar RGB conversion %s -> %s\n",

  447.                av_get_pix_fmt_name(c->srcFormat),

  448.                av_get_pix_fmt_name(c->dstFormat));

  449.         return srcSliceH;

  450.     }

  451. 

  452.     switch (c->dstFormat) {

  453.     case AV_PIX_FMT_BGR24:

  454.         gbr24ptopacked24(src102, stride102,

  455.                          dst[0] + srcSliceY * dstStride[0], dstStride[0],

  456.                          srcSliceH, c->srcW);

  457.         break;

  458. 

  459.     case AV_PIX_FMT_RGB24:

  460.         gbr24ptopacked24(src201, stride201,

  461.                          dst[0] + srcSliceY * dstStride[0], dstStride[0],

  462.                          srcSliceH, c->srcW);

  463.         break;

  464. 

  465.     case AV_PIX_FMT_ARGB:

  466.         alpha_first = 1;

  467.     case AV_PIX_FMT_RGBA:

  468.         gbr24ptopacked32(src201, stride201,

  469.                          dst[0] + srcSliceY * dstStride[0], dstStride[0],

  470.                          srcSliceH, alpha_first, c->srcW);

  471.         break;

  472. 

  473.     case AV_PIX_FMT_ABGR:

  474.         alpha_first = 1;

  475.     case AV_PIX_FMT_BGRA:

  476.         gbr24ptopacked32(src102, stride102,

  477.                          dst[0] + srcSliceY * dstStride[0], dstStride[0],

  478.                          srcSliceH, alpha_first, c->srcW);

  479.         break;

  480. 

  481.     default:

  482.         av_log(c, AV_LOG_ERROR,

  483.                "unsupported planar RGB conversion %s -> %s\n",

  484.                av_get_pix_fmt_name(c->srcFormat),

  485.                av_get_pix_fmt_name(c->dstFormat));

  486.     }

  487. 

  488.     return srcSliceH;

  489. }

  490. 

  491. static void packedtogbr24p(const uint8_t *src, int srcStride,

  492.                            uint8_t *dst[], int dstStride[], int srcSliceH,

  493.                            int alpha_first, int inc_size, int width)

  494. {

  495.     uint8_t *dest[3];

  496.     int x, h;

  497. 

  498.     dest[0] = dst[0];

  499.     dest[1] = dst[1];

  500.     dest[2] = dst[2];

  501. 

  502.     if (alpha_first)

  503.         src++;

  504. 

  505.     for (h = 0; h < srcSliceH; h++) {

  506.         for (x = 0; x < width; x++) {

  507.             dest[0][x] = src[0];

  508.             dest[1][x] = src[1];

  509.             dest[2][x] = src[2];

  510. 

  511.             src += inc_size;

  512.         }

  513.         src     += srcStride - width * inc_size;

  514.         dest[0] += dstStride[0];

  515.         dest[1] += dstStride[1];

  516.         dest[2] += dstStride[2];

  517.     }

  518. }

  519. 

  520. static int rgbToPlanarRgbWrapper(SwsContext *c, const uint8_t *src[],

  521.                                  int srcStride[], int srcSliceY, int srcSliceH,

  522.                                  uint8_t *dst[], int dstStride[])

  523. {

  524.     int alpha_first = 0;

  525.     int stride102[] = { dstStride[1], dstStride[0], dstStride[2] };

  526.     int stride201[] = { dstStride[2], dstStride[0], dstStride[1] };

  527.     uint8_t *dst102[] = { dst[1] + srcSliceY * dstStride[1],

  528.                           dst[0] + srcSliceY * dstStride[0],

  529.                           dst[2] + srcSliceY * dstStride[2] };

  530.     uint8_t *dst201[] = { dst[2] + srcSliceY * dstStride[2],

  531.                           dst[0] + srcSliceY * dstStride[0],

  532.                           dst[1] + srcSliceY * dstStride[1] };

  533. 

  534.     switch (c->srcFormat) {

  535.     case AV_PIX_FMT_RGB24:

  536.         packedtogbr24p((const uint8_t *) src[0], srcStride[0], dst201,

  537.                        stride201, srcSliceH, alpha_first, 3, c->srcW);

  538.         break;

  539.     case AV_PIX_FMT_BGR24:

  540.         packedtogbr24p((const uint8_t *) src[0], srcStride[0], dst102,

  541.                        stride102, srcSliceH, alpha_first, 3, c->srcW);

  542.         break;

  543.     case AV_PIX_FMT_ARGB:

  544.         alpha_first = 1;

  545.     case AV_PIX_FMT_RGBA:

  546.         packedtogbr24p((const uint8_t *) src[0], srcStride[0], dst201,

  547.                        stride201, srcSliceH, alpha_first, 4, c->srcW);

  548.         break;

  549.     case AV_PIX_FMT_ABGR:

  550.         alpha_first = 1;

  551.     case AV_PIX_FMT_BGRA:

  552.         packedtogbr24p((const uint8_t *) src[0], srcStride[0], dst102,

  553.                        stride102, srcSliceH, alpha_first, 4, c->srcW);

  554.         break;

  555.     default:

  556.         av_log(c, AV_LOG_ERROR,

  557.                "unsupported planar RGB conversion %s -> %s\n",

  558.                av_get_pix_fmt_name(c->srcFormat),

  559.                av_get_pix_fmt_name(c->dstFormat));

  560.     }

  561. 

  562.     return srcSliceH;

  563. }

  564. 

  565. #define isRGBA32(x) (            \

  566.            (x) == AV_PIX_FMT_ARGB   \

  567.         || (x) == AV_PIX_FMT_RGBA   \

  568.         || (x) == AV_PIX_FMT_BGRA   \

  569.         || (x) == AV_PIX_FMT_ABGR   \

  570.         )

  571. 

  572. /* {RGB,BGR}{15,16,24,32,32_1} -> {RGB,BGR}{15,16,24,32} */

  573. typedef void (* rgbConvFn) (const uint8_t *, uint8_t *, int);

  574. static rgbConvFn findRgbConvFn(SwsContext *c)

  575. {

  576.     const enum AVPixelFormat srcFormat = c->srcFormat;

  577.     const enum AVPixelFormat dstFormat = c->dstFormat;

  578.     const int srcId = c->srcFormatBpp;

  579.     const int dstId = c->dstFormatBpp;

  580.     rgbConvFn conv = NULL;

  581.     const AVPixFmtDescriptor *desc_src = av_pix_fmt_desc_get(srcFormat);

  582.     const AVPixFmtDescriptor *desc_dst = av_pix_fmt_desc_get(dstFormat);

  583. 

  584. #define IS_NOT_NE(bpp, desc) \

  585.     (((bpp + 7) >> 3) == 2 && \

  586.      (!(desc->flags & AV_PIX_FMT_FLAG_BE) != !HAVE_BIGENDIAN))

  587. 

  588.     /* if this is non-native rgb444/555/565, don't handle it here. */

  589.     if (IS_NOT_NE(srcId, desc_src) || IS_NOT_NE(dstId, desc_dst))

  590.         return NULL;

  591. 

  592. #define CONV_IS(src, dst) (srcFormat == AV_PIX_FMT_##src && dstFormat == AV_PIX_FMT_##dst)

  593. 

  594.     if (isRGBA32(srcFormat) && isRGBA32(dstFormat)) {

  595.         if (     CONV_IS(ABGR, RGBA)

  596.               || CONV_IS(ARGB, BGRA)

  597.               || CONV_IS(BGRA, ARGB)

  598.               || CONV_IS(RGBA, ABGR)) conv = shuffle_bytes_3210;

  599.         else if (CONV_IS(ABGR, ARGB)

  600.               || CONV_IS(ARGB, ABGR)) conv = shuffle_bytes_0321;

  601.         else if (CONV_IS(ABGR, BGRA)

  602.               || CONV_IS(ARGB, RGBA)) conv = shuffle_bytes_1230;

  603.         else if (CONV_IS(BGRA, RGBA)

  604.               || CONV_IS(RGBA, BGRA)) conv = shuffle_bytes_2103;

  605.         else if (CONV_IS(BGRA, ABGR)

  606.               || CONV_IS(RGBA, ARGB)) conv = shuffle_bytes_3012;

  607.     } else

  608.     /* BGR -> BGR */

  609.     if ((isBGRinInt(srcFormat) && isBGRinInt(dstFormat)) ||

  610.         (isRGBinInt(srcFormat) && isRGBinInt(dstFormat))) {

  611.         switch (srcId | (dstId << 16)) {

  612.         case 0x000F000C: conv = rgb12to15; break;

  613.         case 0x000F0010: conv = rgb16to15; break;

  614.         case 0x000F0018: conv = rgb24to15; break;

  615.         case 0x000F0020: conv = rgb32to15; break;

  616.         case 0x0010000F: conv = rgb15to16; break;

  617.         case 0x00100018: conv = rgb24to16; break;

  618.         case 0x00100020: conv = rgb32to16; break;

  619.         case 0x0018000F: conv = rgb15to24; break;

  620.         case 0x00180010: conv = rgb16to24; break;

  621.         case 0x00180020: conv = rgb32to24; break;

  622.         case 0x0020000F: conv = rgb15to32; break;

  623.         case 0x00200010: conv = rgb16to32; break;

  624.         case 0x00200018: conv = rgb24to32; break;

  625.         }

  626.     } else if ((isBGRinInt(srcFormat) && isRGBinInt(dstFormat)) ||

  627.                (isRGBinInt(srcFormat) && isBGRinInt(dstFormat))) {

  628.         switch (srcId | (dstId << 16)) {

  629.         case 0x000C000C: conv = rgb12tobgr12; break;

  630.         case 0x000F000F: conv = rgb15tobgr15; break;

  631.         case 0x000F0010: conv = rgb16tobgr15; break;

  632.         case 0x000F0018: conv = rgb24tobgr15; break;

  633.         case 0x000F0020: conv = rgb32tobgr15; break;

  634.         case 0x0010000F: conv = rgb15tobgr16; break;

  635.         case 0x00100010: conv = rgb16tobgr16; break;

  636.         case 0x00100018: conv = rgb24tobgr16; break;

  637.         case 0x00100020: conv = rgb32tobgr16; break;

  638.         case 0x0018000F: conv = rgb15tobgr24; break;

  639.         case 0x00180010: conv = rgb16tobgr24; break;

  640.         case 0x00180018: conv = rgb24tobgr24; break;

  641.         case 0x00180020: conv = rgb32tobgr24; break;

  642.         case 0x0020000F: conv = rgb15tobgr32; break;

  643.         case 0x00200010: conv = rgb16tobgr32; break;

  644.         case 0x00200018: conv = rgb24tobgr32; break;

  645.         }

  646.     }

  647. 

  648.     return conv;

  649. }

  650. 

  651. /* {RGB,BGR}{15,16,24,32,32_1} -> {RGB,BGR}{15,16,24,32} */

  652. static int rgbToRgbWrapper(SwsContext *c, const uint8_t *src[], int srcStride[],

  653.                            int srcSliceY, int srcSliceH, uint8_t *dst[],

  654.                            int dstStride[])

  655. 

  656. {

  657.     const enum AVPixelFormat srcFormat = c->srcFormat;

  658.     const enum AVPixelFormat dstFormat = c->dstFormat;

  659.     const int srcBpp = (c->srcFormatBpp + 7) >> 3;

  660.     const int dstBpp = (c->dstFormatBpp + 7) >> 3;

  661.     rgbConvFn conv = findRgbConvFn(c);

  662. 

  663.     if (!conv) {

  664.         av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",

  665.                sws_format_name(srcFormat), sws_format_name(dstFormat));

  666.     } else {

  667.         const uint8_t *srcPtr = src[0];

  668.               uint8_t *dstPtr = dst[0];

  669.         if ((srcFormat == AV_PIX_FMT_RGB32_1 || srcFormat == AV_PIX_FMT_BGR32_1) &&

  670.             !isRGBA32(dstFormat))

  671.             srcPtr += ALT32_CORR;

  672. 

  673.         if ((dstFormat == AV_PIX_FMT_RGB32_1 || dstFormat == AV_PIX_FMT_BGR32_1) &&

  674.             !isRGBA32(srcFormat))

  675.             dstPtr += ALT32_CORR;

  676. 

  677.         if (dstStride[0] * srcBpp == srcStride[0] * dstBpp && srcStride[0] > 0 &&

  678.             !(srcStride[0] % srcBpp))

  679.             conv(srcPtr, dstPtr + dstStride[0] * srcSliceY,

  680.                  (srcSliceH - 1) * srcStride[0] + c->srcW * srcBpp);

  681.         else {

  682.             int i;

  683.             dstPtr += dstStride[0] * srcSliceY;

  684. 

  685.             for (i = 0; i < srcSliceH; i++) {

  686.                 conv(srcPtr, dstPtr, c->srcW * srcBpp);

  687.                 srcPtr += srcStride[0];

  688.                 dstPtr += dstStride[0];

  689.             }

  690.         }

  691.     }

  692.     return srcSliceH;

  693. }

  694. 

  695. static int bgr24ToYv12Wrapper(SwsContext *c, const uint8_t *src[],

  696.                               int srcStride[], int srcSliceY, int srcSliceH,

  697.                               uint8_t *dst[], int dstStride[])

  698. {

  699.     rgb24toyv12(

  700.         src[0],

  701.         dst[0] +  srcSliceY       * dstStride[0],

  702.         dst[1] + (srcSliceY >> 1) * dstStride[1],

  703.         dst[2] + (srcSliceY >> 1) * dstStride[2],

  704.         c->srcW, srcSliceH,

  705.         dstStride[0], dstStride[1], srcStride[0]);

  706.     if (dst[3])

  707.         fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);

  708.     return srcSliceH;

  709. }

  710. 

  711. static int yvu9ToYv12Wrapper(SwsContext *c, const uint8_t *src[],

  712.                              int srcStride[], int srcSliceY, int srcSliceH,

  713.                              uint8_t *dst[], int dstStride[])

  714. {

  715.     copyPlane(src[0], srcStride[0], srcSliceY, srcSliceH, c->srcW,

  716.               dst[0], dstStride[0]);

  717. 

  718.     planar2x(src[1], dst[1] + dstStride[1] * (srcSliceY >> 1), c->chrSrcW,

  719.              srcSliceH >> 2, srcStride[1], dstStride[1]);

  720.     planar2x(src[2], dst[2] + dstStride[2] * (srcSliceY >> 1), c->chrSrcW,

  721.              srcSliceH >> 2, srcStride[2], dstStride[2]);

  722.     if (dst[3])

  723.         fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);

  724.     return srcSliceH;

  725. }

  726. 

  727. /* unscaled copy like stuff (assumes nearly identical formats) */

  728. static int packedCopyWrapper(SwsContext *c, const uint8_t *src[],

  729.                              int srcStride[], int srcSliceY, int srcSliceH,

  730.                              uint8_t *dst[], int dstStride[])

  731. {

  732.     if (dstStride[0] == srcStride[0] && srcStride[0] > 0)

  733.         memcpy(dst[0] + dstStride[0] * srcSliceY, src[0], srcSliceH * dstStride[0]);

  734.     else {

  735.         int i;

  736.         const uint8_t *srcPtr = src[0];

  737.         uint8_t *dstPtr = dst[0] + dstStride[0] * srcSliceY;

  738.         int length = 0;

  739. 

  740.         /* universal length finder */

  741.         while (length + c->srcW <= FFABS(dstStride[0]) &&

  742.                length + c->srcW <= FFABS(srcStride[0]))

  743.             length += c->srcW;

  744.         assert(length != 0);

  745. 

  746.         for (i = 0; i < srcSliceH; i++) {

  747.             memcpy(dstPtr, srcPtr, length);

  748.             srcPtr += srcStride[0];

  749.             dstPtr += dstStride[0];

  750.         }

  751.     }

  752.     return srcSliceH;

  753. }

  754. 

  755. #define clip9(x)  av_clip_uintp2(x,  9)

  756. #define clip10(x) av_clip_uintp2(x, 10)

  757. #define DITHER_COPY(dst, dstStride, wfunc, src, srcStride, rfunc, dithers, shift, clip) \

  758.     for (i = 0; i < height; i++) { \

  759.         const uint8_t *dither = dithers[i & 7]; \

  760.         for (j = 0; j < length - 7; j += 8) { \

  761.             wfunc(&dst[j + 0], clip((rfunc(&src[j + 0]) + dither[0]) >> shift)); \

  762.             wfunc(&dst[j + 1], clip((rfunc(&src[j + 1]) + dither[1]) >> shift)); \

  763.             wfunc(&dst[j + 2], clip((rfunc(&src[j + 2]) + dither[2]) >> shift)); \

  764.             wfunc(&dst[j + 3], clip((rfunc(&src[j + 3]) + dither[3]) >> shift)); \

  765.             wfunc(&dst[j + 4], clip((rfunc(&src[j + 4]) + dither[4]) >> shift)); \

  766.             wfunc(&dst[j + 5], clip((rfunc(&src[j + 5]) + dither[5]) >> shift)); \

  767.             wfunc(&dst[j + 6], clip((rfunc(&src[j + 6]) + dither[6]) >> shift)); \

  768.             wfunc(&dst[j + 7], clip((rfunc(&src[j + 7]) + dither[7]) >> shift)); \

  769.         } \

  770.         for (; j < length; j++) \

  771.             wfunc(&dst[j],     (rfunc(&src[j]) + dither[j & 7]) >> shift); \

  772.         dst += dstStride; \

  773.         src += srcStride; \

  774.     }

  775. 

  776. static int planarCopyWrapper(SwsContext *c, const uint8_t *src[],

  777.                              int srcStride[], int srcSliceY, int srcSliceH,

  778.                              uint8_t *dst[], int dstStride[])

  779. {

  780.     const AVPixFmtDescriptor *desc_src = av_pix_fmt_desc_get(c->srcFormat);

  781.     const AVPixFmtDescriptor *desc_dst = av_pix_fmt_desc_get(c->dstFormat);

  782.     int plane, i, j;

  783.     for (plane = 0; plane < 4; plane++) {

  784.         int length = (plane == 0 || plane == 3) ? c->srcW  : -((-c->srcW  ) >> c->chrDstHSubSample);

  785.         int y =      (plane == 0 || plane == 3) ? srcSliceY: -((-srcSliceY) >> c->chrDstVSubSample);

  786.         int height = (plane == 0 || plane == 3) ? srcSliceH: -((-srcSliceH) >> c->chrDstVSubSample);

  787.         const uint8_t *srcPtr = src[plane];

  788.         uint8_t *dstPtr = dst[plane] + dstStride[plane] * y;

  789.         int shiftonly = plane == 1 || plane == 2 || (!c->srcRange && plane == 0);

  790. 

  791.         if (!dst[plane])

  792.             continue;

  793.         // ignore palette for GRAY8

  794.         if (plane == 1 && !dst[2]) continue;

  795.         if (!src[plane] || (plane == 1 && !src[2])) {

  796.             int val = (plane == 3) ? 255 : 128;

  797.             if (is16BPS(c->dstFormat))

  798.                 length *= 2;

  799.             if (is9_OR_10BPS(c->dstFormat)) {

  800.                 fill_plane9or10(dst[plane], dstStride[plane],

  801.                                 length, height, y, val,

  802.                                 desc_dst->comp[plane].depth_minus1 + 1,

  803.                                 isBE(c->dstFormat));

  804.             } else

  805.                 fillPlane(dst[plane], dstStride[plane], length, height, y,

  806.                           val);

  807.         } else {

  808.             if (is9_OR_10BPS(c->srcFormat)) {

  809.                 const int src_depth = desc_src->comp[plane].depth_minus1 + 1;

  810.                 const int dst_depth = desc_dst->comp[plane].depth_minus1 + 1;

  811.                 const uint16_t *srcPtr2 = (const uint16_t *) srcPtr;

  812. 

  813.                 if (is16BPS(c->dstFormat)) {

  814.                     uint16_t *dstPtr2 = (uint16_t *) dstPtr;

  815. #define COPY9_OR_10TO16(rfunc, wfunc) \

  816.                     if (shiftonly) { \

  817.                         for (i = 0; i < height; i++) { \

  818.                             for (j = 0; j < length; j++) { \

  819.                                 int srcpx = rfunc(&srcPtr2[j]); \

  820.                                 wfunc(&dstPtr2[j], srcpx << (16 - src_depth)); \

  821.                             } \

  822.                             dstPtr2 += dstStride[plane] / 2; \

  823.                             srcPtr2 += srcStride[plane] / 2; \

  824.                         } \

  825.                     } else { \

  826.                         for (i = 0; i < height; i++) { \

  827.                             for (j = 0; j < length; j++) { \

  828.                                 int srcpx = rfunc(&srcPtr2[j]); \

  829.                                 wfunc(&dstPtr2[j], (srcpx << (16 - src_depth)) | (srcpx >> (2 * src_depth - 16))); \

  830.                             } \

  831.                             dstPtr2 += dstStride[plane] / 2; \

  832.                             srcPtr2 += srcStride[plane] / 2; \

  833.                         } \

  834.                     }

  835.                     if (isBE(c->dstFormat)) {

  836.                         if (isBE(c->srcFormat)) {

  837.                             COPY9_OR_10TO16(AV_RB16, AV_WB16);

  838.                         } else {

  839.                             COPY9_OR_10TO16(AV_RL16, AV_WB16);

  840.                         }

  841.                     } else {

  842.                         if (isBE(c->srcFormat)) {

  843.                             COPY9_OR_10TO16(AV_RB16, AV_WL16);

  844.                         } else {

  845.                             COPY9_OR_10TO16(AV_RL16, AV_WL16);

  846.                         }

  847.                     }

  848.                 } else if (is9_OR_10BPS(c->dstFormat)) {

  849.                     uint16_t *dstPtr2 = (uint16_t *) dstPtr;

  850. #define COPY9_OR_10TO9_OR_10(loop) \

  851.                     for (i = 0; i < height; i++) { \

  852.                         for (j = 0; j < length; j++) { \

  853.                             loop; \

  854.                         } \

  855.                         dstPtr2 += dstStride[plane] / 2; \

  856.                         srcPtr2 += srcStride[plane] / 2; \

  857.                     }

  858. #define COPY9_OR_10TO9_OR_10_2(rfunc, wfunc) \

  859.                     if (dst_depth > src_depth) { \

  860.                         COPY9_OR_10TO9_OR_10(int srcpx = rfunc(&srcPtr2[j]); \

  861.                             wfunc(&dstPtr2[j], (srcpx << 1) | (srcpx >> 9))); \

  862.                     } else if (dst_depth < src_depth) { \

  863.                         DITHER_COPY(dstPtr2, dstStride[plane] / 2, wfunc, \

  864.                                     srcPtr2, srcStride[plane] / 2, rfunc, \

  865.                                     dither_8x8_1, 1, clip9); \

  866.                     } else { \

  867.                         COPY9_OR_10TO9_OR_10(wfunc(&dstPtr2[j], rfunc(&srcPtr2[j]))); \

  868.                     }

  869.                     if (isBE(c->dstFormat)) {

  870.                         if (isBE(c->srcFormat)) {

  871.                             COPY9_OR_10TO9_OR_10_2(AV_RB16, AV_WB16);

  872.                         } else {

  873.                             COPY9_OR_10TO9_OR_10_2(AV_RL16, AV_WB16);

  874.                         }

  875.                     } else {

  876.                         if (isBE(c->srcFormat)) {

  877.                             COPY9_OR_10TO9_OR_10_2(AV_RB16, AV_WL16);

  878.                         } else {

  879.                             COPY9_OR_10TO9_OR_10_2(AV_RL16, AV_WL16);

  880.                         }

  881.                     }

  882.                 } else {

  883. #define W8(a, b) { *(a) = (b); }

  884. #define COPY9_OR_10TO8(rfunc) \

  885.                     if (src_depth == 9) { \

  886.                         DITHER_COPY(dstPtr,  dstStride[plane],   W8, \

  887.                                     srcPtr2, srcStride[plane] / 2, rfunc, \

  888.                                     dither_8x8_1, 1, av_clip_uint8); \

  889.                     } else { \

  890.                         DITHER_COPY(dstPtr,  dstStride[plane],   W8, \

  891.                                     srcPtr2, srcStride[plane] / 2, rfunc, \

  892.                                     dither_8x8_3, 2, av_clip_uint8); \

  893.                     }

  894.                     if (isBE(c->srcFormat)) {

  895.                         COPY9_OR_10TO8(AV_RB16);

  896.                     } else {

  897.                         COPY9_OR_10TO8(AV_RL16);

  898.                     }

  899.                 }

  900.             } else if (is9_OR_10BPS(c->dstFormat)) {

  901.                 const int dst_depth = desc_dst->comp[plane].depth_minus1 + 1;

  902.                 uint16_t *dstPtr2 = (uint16_t *) dstPtr;

  903. 

  904.                 if (is16BPS(c->srcFormat)) {

  905.                     const uint16_t *srcPtr2 = (const uint16_t *) srcPtr;

  906. #define COPY16TO9_OR_10(rfunc, wfunc) \

  907.                     if (dst_depth == 9) { \

  908.                         DITHER_COPY(dstPtr2, dstStride[plane] / 2, wfunc, \

  909.                                     srcPtr2, srcStride[plane] / 2, rfunc, \

  910.                                     ff_dither_8x8_128, 7, clip9); \

  911.                     } else { \

  912.                         DITHER_COPY(dstPtr2, dstStride[plane] / 2, wfunc, \

  913.                                     srcPtr2, srcStride[plane] / 2, rfunc, \

  914.                                     dither_8x8_64, 6, clip10); \

  915.                     }

  916.                     if (isBE(c->dstFormat)) {

  917.                         if (isBE(c->srcFormat)) {

  918.                             COPY16TO9_OR_10(AV_RB16, AV_WB16);

  919.                         } else {

  920.                             COPY16TO9_OR_10(AV_RL16, AV_WB16);

  921.                         }

  922.                     } else {

  923.                         if (isBE(c->srcFormat)) {

  924.                             COPY16TO9_OR_10(AV_RB16, AV_WL16);

  925.                         } else {

  926.                             COPY16TO9_OR_10(AV_RL16, AV_WL16);

  927.                         }

  928.                     }

  929.                 } else /* 8bit */ {

  930. #define COPY8TO9_OR_10(wfunc) \

  931.                     if (shiftonly) { \

  932.                         for (i = 0; i < height; i++) { \

  933.                             for (j = 0; j < length; j++) { \

  934.                                 const int srcpx = srcPtr[j]; \

  935.                                 wfunc(&dstPtr2[j], srcpx << (dst_depth - 8)); \

  936.                             } \

  937.                             dstPtr2 += dstStride[plane] / 2; \

  938.                             srcPtr  += srcStride[plane]; \

  939.                         } \

  940.                     } else { \

  941.                         for (i = 0; i < height; i++) { \

  942.                             for (j = 0; j < length; j++) { \

  943.                                 const int srcpx = srcPtr[j]; \

  944.                                 wfunc(&dstPtr2[j], (srcpx << (dst_depth - 8)) | (srcpx >> (16 - dst_depth))); \

  945.                             } \

  946.                             dstPtr2 += dstStride[plane] / 2; \

  947.                             srcPtr  += srcStride[plane]; \

  948.                         } \

  949.                     }

  950.                     if (isBE(c->dstFormat)) {

  951.                         COPY8TO9_OR_10(AV_WB16);

  952.                     } else {

  953.                         COPY8TO9_OR_10(AV_WL16);

  954.                     }

  955.                 }

  956.             } else if (is16BPS(c->srcFormat) && !is16BPS(c->dstFormat)) {

  957.                 const uint16_t *srcPtr2 = (const uint16_t *) srcPtr;

  958. #define COPY16TO8(rfunc) \

  959.                     DITHER_COPY(dstPtr,  dstStride[plane],   W8, \

  960.                                 srcPtr2, srcStride[plane] / 2, rfunc, \

  961.                                 dither_8x8_256, 8, av_clip_uint8);

  962.                 if (isBE(c->srcFormat)) {

  963.                     COPY16TO8(AV_RB16);

  964.                 } else {

  965.                     COPY16TO8(AV_RL16);

  966.                 }

  967.             } else if (!is16BPS(c->srcFormat) && is16BPS(c->dstFormat)) {

  968.                 for (i = 0; i < height; i++) {

  969.                     for (j = 0; j < length; j++) {

  970.                         dstPtr[ j << 1     ] = srcPtr[j];

  971.                         dstPtr[(j << 1) + 1] = srcPtr[j];

  972.                     }

  973.                     srcPtr += srcStride[plane];

  974.                     dstPtr += dstStride[plane];

  975.                 }

  976.             } else if (is16BPS(c->srcFormat) && is16BPS(c->dstFormat) &&

  977.                       isBE(c->srcFormat) != isBE(c->dstFormat)) {

  978. 

  979.                 for (i = 0; i < height; i++) {

  980.                     for (j = 0; j < length; j++)

  981.                         ((uint16_t *) dstPtr)[j] = av_bswap16(((const uint16_t *) srcPtr)[j]);

  982.                     srcPtr += srcStride[plane];

  983.                     dstPtr += dstStride[plane];

  984.                 }

  985.             } else if (dstStride[plane] == srcStride[plane] &&

  986.                        srcStride[plane] > 0 && srcStride[plane] == length) {

  987.                 memcpy(dst[plane] + dstStride[plane] * y, src[plane],

  988.                        height * dstStride[plane]);

  989.             } else {

  990.                 if (is16BPS(c->srcFormat) && is16BPS(c->dstFormat))

  991.                     length *= 2;

  992.                 else if (!desc_src->comp[0].depth_minus1)

  993.                     length >>= 3; // monowhite/black

  994.                 for (i = 0; i < height; i++) {

  995.                     memcpy(dstPtr, srcPtr, length);

  996.                     srcPtr += srcStride[plane];

  997.                     dstPtr += dstStride[plane];

  998.                 }

  999.             }

  1000.         }

  1001.     }

  1002.     return srcSliceH;

  1003. }

  1004. 

  1005. 

  1006. #define IS_DIFFERENT_ENDIANESS(src_fmt, dst_fmt, pix_fmt)          \

  1007.     ((src_fmt == pix_fmt ## BE && dst_fmt == pix_fmt ## LE) ||     \

  1008.      (src_fmt == pix_fmt ## LE && dst_fmt == pix_fmt ## BE))

  1009. 

  1010. 

  1011. void ff_get_unscaled_swscale(SwsContext *c)

  1012. {

  1013.     const enum AVPixelFormat srcFormat = c->srcFormat;

  1014.     const enum AVPixelFormat dstFormat = c->dstFormat;

  1015.     const int flags = c->flags;

  1016.     const int dstH = c->dstH;

  1017.     int needsDither;

  1018. 

  1019.     needsDither = isAnyRGB(dstFormat) &&

  1020.             c->dstFormatBpp < 24 &&

  1021.            (c->dstFormatBpp < c->srcFormatBpp || (!isAnyRGB(srcFormat)));

  1022. 

  1023.     /* yv12_to_nv12 */

  1024.     if ((srcFormat == AV_PIX_FMT_YUV420P || srcFormat == AV_PIX_FMT_YUVA420P) &&

  1025.         (dstFormat == AV_PIX_FMT_NV12 || dstFormat == AV_PIX_FMT_NV21)) {

  1026.         c->swscale = planarToNv12Wrapper;

  1027.     }

  1028.     /* nv12_to_yv12 */

  1029.     if (dstFormat == AV_PIX_FMT_YUV420P &&

  1030.         (srcFormat == AV_PIX_FMT_NV12 || srcFormat == AV_PIX_FMT_NV21)) {

  1031.         c->swscale = nv12ToPlanarWrapper;

  1032.     }

  1033.     /* yuv2bgr */

  1034.     if ((srcFormat == AV_PIX_FMT_YUV420P || srcFormat == AV_PIX_FMT_YUV422P ||

  1035.          srcFormat == AV_PIX_FMT_YUVA420P) && isAnyRGB(dstFormat) &&

  1036.         !(flags & SWS_ACCURATE_RND) && !(dstH & 1)) {

  1037.         c->swscale = ff_yuv2rgb_get_func_ptr(c);

  1038.     }

  1039. 

  1040.     if (srcFormat == AV_PIX_FMT_YUV410P &&

  1041.         (dstFormat == AV_PIX_FMT_YUV420P || dstFormat == AV_PIX_FMT_YUVA420P) &&

  1042.         !(flags & SWS_BITEXACT)) {

  1043.         c->swscale = yvu9ToYv12Wrapper;

  1044.     }

  1045. 

  1046.     /* bgr24toYV12 */

  1047.     if (srcFormat == AV_PIX_FMT_BGR24 &&

  1048.         (dstFormat == AV_PIX_FMT_YUV420P || dstFormat == AV_PIX_FMT_YUVA420P) &&

  1049.         !(flags & SWS_ACCURATE_RND))

  1050.         c->swscale = bgr24ToYv12Wrapper;

  1051. 

  1052.     /* RGB/BGR -> RGB/BGR (no dither needed forms) */

  1053.     if (isAnyRGB(srcFormat) && isAnyRGB(dstFormat) && findRgbConvFn(c)

  1054.         && (!needsDither || (c->flags&(SWS_FAST_BILINEAR|SWS_POINT))))

  1055.         c->swscale = rgbToRgbWrapper;

  1056. 

  1057. #define isByteRGB(f) (             \

  1058.         f == AV_PIX_FMT_RGB32   || \

  1059.         f == AV_PIX_FMT_RGB32_1 || \

  1060.         f == AV_PIX_FMT_RGB24   || \

  1061.         f == AV_PIX_FMT_BGR32   || \

  1062.         f == AV_PIX_FMT_BGR32_1 || \

  1063.         f == AV_PIX_FMT_BGR24)

  1064. 

  1065.     if (srcFormat == AV_PIX_FMT_GBRP && isPlanar(srcFormat) && isByteRGB(dstFormat))

  1066.         c->swscale = planarRgbToRgbWrapper;

  1067. 

  1068.     if (av_pix_fmt_desc_get(srcFormat)->comp[0].depth_minus1 == 7 &&

  1069.         isPackedRGB(srcFormat) && dstFormat == AV_PIX_FMT_GBRP)

  1070.         c->swscale = rgbToPlanarRgbWrapper;

  1071. 

  1072.     /* bswap 16 bits per pixel/component packed formats */

  1073.     if (IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_BGR444) ||

  1074.         IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_BGR48)  ||

  1075.         IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_BGR555) ||

  1076.         IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_BGR565) ||

  1077.         IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_BGRA64) ||

  1078.         IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_GRAY16) ||

  1079.         IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_YA16)   ||

  1080.         IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_RGB444) ||

  1081.         IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_RGB48)  ||

  1082.         IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_RGB555) ||

  1083.         IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_RGB565) ||

  1084.         IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_RGBA64) ||

  1085.         IS_DIFFERENT_ENDIANESS(srcFormat, dstFormat, AV_PIX_FMT_XYZ12))

  1086.         c->swscale = packed_16bpc_bswap;

  1087. 

  1088.     if ((usePal(srcFormat) && (

  1089.         dstFormat == AV_PIX_FMT_RGB32   ||

  1090.         dstFormat == AV_PIX_FMT_RGB32_1 ||

  1091.         dstFormat == AV_PIX_FMT_RGB24   ||

  1092.         dstFormat == AV_PIX_FMT_BGR32   ||

  1093.         dstFormat == AV_PIX_FMT_BGR32_1 ||

  1094.         dstFormat == AV_PIX_FMT_BGR24)))

  1095.         c->swscale = palToRgbWrapper;

  1096. 

  1097.     if (srcFormat == AV_PIX_FMT_YUV422P) {

  1098.         if (dstFormat == AV_PIX_FMT_YUYV422)

  1099.             c->swscale = yuv422pToYuy2Wrapper;

  1100.         else if (dstFormat == AV_PIX_FMT_UYVY422)

  1101.             c->swscale = yuv422pToUyvyWrapper;

  1102.     }

  1103. 

  1104.     /* LQ converters if -sws 0 or -sws 4*/

  1105.     if (c->flags&(SWS_FAST_BILINEAR|SWS_POINT)) {

  1106.         /* yv12_to_yuy2 */

  1107.         if (srcFormat == AV_PIX_FMT_YUV420P || srcFormat == AV_PIX_FMT_YUVA420P) {

  1108.             if (dstFormat == AV_PIX_FMT_YUYV422)

  1109.                 c->swscale = planarToYuy2Wrapper;

  1110.             else if (dstFormat == AV_PIX_FMT_UYVY422)

  1111.                 c->swscale = planarToUyvyWrapper;

  1112.         }

  1113.     }

  1114.     if (srcFormat == AV_PIX_FMT_YUYV422 &&

  1115.        (dstFormat == AV_PIX_FMT_YUV420P || dstFormat == AV_PIX_FMT_YUVA420P))

  1116.         c->swscale = yuyvToYuv420Wrapper;

  1117.     if (srcFormat == AV_PIX_FMT_UYVY422 &&

  1118.        (dstFormat == AV_PIX_FMT_YUV420P || dstFormat == AV_PIX_FMT_YUVA420P))

  1119.         c->swscale = uyvyToYuv420Wrapper;

  1120.     if (srcFormat == AV_PIX_FMT_YUYV422 && dstFormat == AV_PIX_FMT_YUV422P)

  1121.         c->swscale = yuyvToYuv422Wrapper;

  1122.     if (srcFormat == AV_PIX_FMT_UYVY422 && dstFormat == AV_PIX_FMT_YUV422P)

  1123.         c->swscale = uyvyToYuv422Wrapper;

  1124. 

  1125.     /* simple copy */

  1126.     if ( srcFormat == dstFormat ||

  1127.         (srcFormat == AV_PIX_FMT_YUVA420P && dstFormat == AV_PIX_FMT_YUV420P) ||

  1128.         (srcFormat == AV_PIX_FMT_YUV420P && dstFormat == AV_PIX_FMT_YUVA420P) ||

  1129.         (isPlanarYUV(srcFormat) && isGray(dstFormat)) ||

  1130.         (isPlanarYUV(dstFormat) && isGray(srcFormat)) ||

  1131.         (isGray(dstFormat) && isGray(srcFormat)) ||

  1132.         (isPlanarYUV(srcFormat) && isPlanarYUV(dstFormat) &&

  1133.          c->chrDstHSubSample == c->chrSrcHSubSample &&

  1134.          c->chrDstVSubSample == c->chrSrcVSubSample &&

  1135.          dstFormat != AV_PIX_FMT_NV12 && dstFormat != AV_PIX_FMT_NV21 &&

  1136.          srcFormat != AV_PIX_FMT_NV12 && srcFormat != AV_PIX_FMT_NV21))

  1137.     {

  1138.         if (isPacked(c->srcFormat))

  1139.             c->swscale = packedCopyWrapper;

  1140.         else /* Planar YUV or gray */

  1141.             c->swscale = planarCopyWrapper;

  1142.     }

  1143. 

  1144.     if (ARCH_PPC)

  1145.         ff_get_unscaled_swscale_ppc(c);

  1146. }

  1147. 

  1148. static void reset_ptr(const uint8_t *src[], int format)

  1149. {

  1150.     if (!isALPHA(format))

  1151.         src[3] = NULL;

  1152.     if (!isPlanar(format)) {

  1153.         src[3] = src[2] = NULL;

  1154. 

  1155.         if (!usePal(format))

  1156.             src[1] = NULL;

  1157.     }

  1158. }

  1159. 

  1160. static int check_image_pointers(uint8_t *data[4], enum AVPixelFormat pix_fmt,

  1161.                                 const int linesizes[4])

  1162. {

  1163.     const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(pix_fmt);

  1164.     int i;

  1165. 

  1166.     for (i = 0; i < 4; i++) {

  1167.         int plane = desc->comp[i].plane;

  1168.         if (!data[plane] || !linesizes[plane])

  1169.             return 0;

  1170.     }

  1171. 

  1172.     return 1;

  1173. }

  1174. 

  1175. /**

  1176.  * swscale wrapper, so we don't need to export the SwsContext.

  1177.  * Assumes planar YUV to be in YUV order instead of YVU.

  1178.  */

  1179. int attribute_align_arg sws_scale(struct SwsContext *c,

  1180.                                   const uint8_t * const srcSlice[],

  1181.                                   const int srcStride[], int srcSliceY,

  1182.                                   int srcSliceH, uint8_t *const dst[],

  1183.                                   const int dstStride[])

  1184. {

  1185.     int i;

  1186.     const uint8_t *src2[4] = { srcSlice[0], srcSlice[1], srcSlice[2], srcSlice[3] };

  1187.     uint8_t *dst2[4] = { dst[0], dst[1], dst[2], dst[3] };

  1188. 

  1189.     // do not mess up sliceDir if we have a "trailing" 0-size slice

  1190.     if (srcSliceH == 0)

  1191.         return 0;

  1192. 

  1193.     if (!check_image_pointers(srcSlice, c->srcFormat, srcStride)) {

  1194.         av_log(c, AV_LOG_ERROR, "bad src image pointers\n");

  1195.         return 0;

  1196.     }

  1197.     if (!check_image_pointers(dst, c->dstFormat, dstStride)) {

  1198.         av_log(c, AV_LOG_ERROR, "bad dst image pointers\n");

  1199.         return 0;

  1200.     }

  1201. 

  1202.     if (c->sliceDir == 0 && srcSliceY != 0 && srcSliceY + srcSliceH != c->srcH) {

  1203.         av_log(c, AV_LOG_ERROR, "Slices start in the middle!\n");

  1204.         return 0;

  1205.     }

  1206.     if (c->sliceDir == 0) {

  1207.         if (srcSliceY == 0) c->sliceDir = 1; else c->sliceDir = -1;

  1208.     }

  1209. 

  1210.     if (usePal(c->srcFormat)) {

  1211.         for (i = 0; i < 256; i++) {

  1212.             int r, g, b, y, u, v;

  1213.             if (c->srcFormat == AV_PIX_FMT_PAL8) {

  1214.                 uint32_t p = ((const uint32_t *)(srcSlice[1]))[i];

  1215.                 r = (p >> 16) & 0xFF;

  1216.                 g = (p >>  8) & 0xFF;

  1217.                 b =  p        & 0xFF;

  1218.             } else if (c->srcFormat == AV_PIX_FMT_RGB8) {

  1219.                 r = ( i >> 5     ) * 36;

  1220.                 g = ((i >> 2) & 7) * 36;

  1221.                 b = ( i       & 3) * 85;

  1222.             } else if (c->srcFormat == AV_PIX_FMT_BGR8) {

  1223.                 b = ( i >> 6     ) * 85;

  1224.                 g = ((i >> 3) & 7) * 36;

  1225.                 r = ( i       & 7) * 36;

  1226.             } else if (c->srcFormat == AV_PIX_FMT_RGB4_BYTE) {

  1227.                 r = ( i >> 3     ) * 255;

  1228.                 g = ((i >> 1) & 3) * 85;

  1229.                 b = ( i       & 1) * 255;

  1230.             } else if (c->srcFormat == AV_PIX_FMT_GRAY8 ||

  1231.                        c->srcFormat == AV_PIX_FMT_YA8) {

  1232.                 r = g = b = i;

  1233.             } else {

  1234.                 assert(c->srcFormat == AV_PIX_FMT_BGR4_BYTE);

  1235.                 b = ( i >> 3     ) * 255;

  1236.                 g = ((i >> 1) & 3) * 85;

  1237.                 r = ( i       & 1) * 255;

  1238.             }

  1239.             y = av_clip_uint8((RY * r + GY * g + BY * b + ( 33 << (RGB2YUV_SHIFT - 1))) >> RGB2YUV_SHIFT);

  1240.             u = av_clip_uint8((RU * r + GU * g + BU * b + (257 << (RGB2YUV_SHIFT - 1))) >> RGB2YUV_SHIFT);

  1241.             v = av_clip_uint8((RV * r + GV * g + BV * b + (257 << (RGB2YUV_SHIFT - 1))) >> RGB2YUV_SHIFT);

  1242.             c->pal_yuv[i] = y + (u << 8) + (v << 16) + (0xFFU << 24);

  1243. 

  1244.             switch (c->dstFormat) {

  1245.             case AV_PIX_FMT_BGR32:

  1246. #if !HAVE_BIGENDIAN

  1247.             case AV_PIX_FMT_RGB24:

  1248. #endif

  1249.                 c->pal_rgb[i] =  r + (g << 8) + (b << 16) + (0xFFU << 24);

  1250.                 break;

  1251.             case AV_PIX_FMT_BGR32_1:

  1252. #if HAVE_BIGENDIAN

  1253.             case AV_PIX_FMT_BGR24:

  1254. #endif

  1255.                 c->pal_rgb[i] = 0xFF + (r << 8) + (g << 16) + ((unsigned)b << 24);

  1256.                 break;

  1257.             case AV_PIX_FMT_RGB32_1:

  1258. #if HAVE_BIGENDIAN

  1259.             case AV_PIX_FMT_RGB24:

  1260. #endif

  1261.                 c->pal_rgb[i] = 0xFF + (b << 8) + (g << 16) + ((unsigned)r << 24);

  1262.                 break;

  1263.             case AV_PIX_FMT_RGB32:

  1264. #if !HAVE_BIGENDIAN

  1265.             case AV_PIX_FMT_BGR24:

  1266. #endif

  1267.             default:

  1268.                 c->pal_rgb[i] =  b + (g << 8) + (r << 16) + (0xFFU << 24);

  1269.             }

  1270.         }

  1271.     }

  1272. 

  1273.     // copy strides, so they can safely be modified

  1274.     if (c->sliceDir == 1) {

  1275.         // slices go from top to bottom

  1276.         int srcStride2[4] = { srcStride[0], srcStride[1], srcStride[2],

  1277.                               srcStride[3] };

  1278.         int dstStride2[4] = { dstStride[0], dstStride[1], dstStride[2],

  1279.                               dstStride[3] };

  1280. 

  1281.         reset_ptr(src2, c->srcFormat);

  1282.         reset_ptr((const uint8_t **) dst2, c->dstFormat);

  1283. 

  1284.         /* reset slice direction at end of frame */

  1285.         if (srcSliceY + srcSliceH == c->srcH)

  1286.             c->sliceDir = 0;

  1287. 

  1288.         return c->swscale(c, src2, srcStride2, srcSliceY, srcSliceH, dst2,

  1289.                           dstStride2);

  1290.     } else {

  1291.         // slices go from bottom to top => we flip the image internally

  1292.         int srcStride2[4] = { -srcStride[0], -srcStride[1], -srcStride[2],

  1293.                               -srcStride[3] };

  1294.         int dstStride2[4] = { -dstStride[0], -dstStride[1], -dstStride[2],

  1295.                               -dstStride[3] };

  1296. 

  1297.         src2[0] += (srcSliceH - 1) * srcStride[0];

  1298.         if (!usePal(c->srcFormat))

  1299.             src2[1] += ((srcSliceH >> c->chrSrcVSubSample) - 1) * srcStride[1];

  1300.         src2[2] += ((srcSliceH >> c->chrSrcVSubSample) - 1) * srcStride[2];

  1301.         src2[3] += (srcSliceH - 1) * srcStride[3];

  1302.         dst2[0] += ( c->dstH                         - 1) * dstStride[0];

  1303.         dst2[1] += ((c->dstH >> c->chrDstVSubSample) - 1) * dstStride[1];

  1304.         dst2[2] += ((c->dstH >> c->chrDstVSubSample) - 1) * dstStride[2];

  1305.         dst2[3] += ( c->dstH                         - 1) * dstStride[3];

  1306. 

  1307.         reset_ptr(src2, c->srcFormat);

  1308.         reset_ptr((const uint8_t **) dst2, c->dstFormat);

  1309. 

  1310.         /* reset slice direction at end of frame */

  1311.         if (!srcSliceY)

  1312.             c->sliceDir = 0;

  1313. 

  1314.         return c->swscale(c, src2, srcStride2, c->srcH-srcSliceY-srcSliceH,

  1315.                           srcSliceH, dst2, dstStride2);

  1316.     }

  1317. }

  1318. 

  1319. /* Convert the palette to the same packed 32-bit format as the palette */

  1320. void sws_convertPalette8ToPacked32(const uint8_t *src, uint8_t *dst,

  1321.                                    int num_pixels, const uint8_t *palette)

  1322. {

  1323.     int i;

  1324. 

  1325.     for (i = 0; i < num_pixels; i++)

  1326.         ((uint32_t *) dst)[i] = ((const uint32_t *) palette)[src[i]];

  1327. }

  1328. 

  1329. /* Palette format: ABCD -> dst format: ABC */

  1330. void sws_convertPalette8ToPacked24(const uint8_t *src, uint8_t *dst,

  1331.                                    int num_pixels, const uint8_t *palette)

  1332. {

  1333.     int i;

  1334. 

  1335.     for (i = 0; i < num_pixels; i++) {

  1336.         //FIXME slow?

  1337.         dst[0] = palette[src[i] * 4 + 0];

  1338.         dst[1] = palette[src[i] * 4 + 1];

  1339.         dst[2] = palette[src[i] * 4 + 2];

  1340.         dst += 3;

  1341.     }

  1342. }