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.
28654 Commits
32 Branches
380MB
Tree: 11ffefefdb
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from '11ffefefdb'
${ noResults }
FFmpeg/libswscale/swscale.c

2102 lines
76KB
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. /*

  22.   supported Input formats: YV12, I420/IYUV, YUY2, UYVY, BGR32, BGR32_1, BGR24, BGR16, BGR15, RGB32, RGB32_1, RGB24, Y8/Y800, YVU9/IF09, PAL8

  23.   supported output formats: YV12, I420/IYUV, YUY2, UYVY, {BGR,RGB}{1,4,8,15,16,24,32}, Y8/Y800, YVU9/IF09

  24.   {BGR,RGB}{1,4,8,15,16} support dithering

  25. 

  26.   unscaled special converters (YV12=I420=IYUV, Y800=Y8)

  27.   YV12 -> {BGR,RGB}{1,4,8,12,15,16,24,32}

  28.   x -> x

  29.   YUV9 -> YV12

  30.   YUV9/YV12 -> Y800

  31.   Y800 -> YUV9/YV12

  32.   BGR24 -> BGR32 & RGB24 -> RGB32

  33.   BGR32 -> BGR24 & RGB32 -> RGB24

  34.   BGR15 -> BGR16

  35. */

  36. 

  37. /*

  38. tested special converters (most are tested actually, but I did not write it down ...)

  39.  YV12 -> BGR12/BGR16

  40.  YV12 -> YV12

  41.  BGR15 -> BGR16

  42.  BGR16 -> BGR16

  43.  YVU9 -> YV12

  44. 

  45. untested special converters

  46.   YV12/I420 -> BGR15/BGR24/BGR32 (it is the yuv2rgb stuff, so it should be OK)

  47.   YV12/I420 -> YV12/I420

  48.   YUY2/BGR15/BGR24/BGR32/RGB24/RGB32 -> same format

  49.   BGR24 -> BGR32 & RGB24 -> RGB32

  50.   BGR32 -> BGR24 & RGB32 -> RGB24

  51.   BGR24 -> YV12

  52. */

  53. 

  54. #include <inttypes.h>

  55. #include <string.h>

  56. #include <math.h>

  57. #include <stdio.h>

  58. #include "config.h"

  59. #include <assert.h>

  60. #include "swscale.h"

  61. #include "swscale_internal.h"

  62. #include "rgb2rgb.h"

  63. #include "libavutil/intreadwrite.h"

  64. #include "libavutil/x86_cpu.h"

  65. #include "libavutil/cpu.h"

  66. #include "libavutil/avutil.h"

  67. #include "libavutil/mathematics.h"

  68. #include "libavutil/bswap.h"

  69. #include "libavutil/pixdesc.h"

  70. 

  71. #undef MOVNTQ

  72. #undef PAVGB

  73. 

  74. #define DITHER1XBPP

  75. 

  76. #define isPacked(x)         (       \

  77.            (x)==PIX_FMT_PAL8        \

  78.         || (x)==PIX_FMT_YUYV422     \

  79.         || (x)==PIX_FMT_UYVY422     \

  80.         || (x)==PIX_FMT_Y400A       \

  81.         || isAnyRGB(x)              \

  82.     )

  83. 

  84. #define RGB2YUV_SHIFT 15

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

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

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

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

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

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

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

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

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

  94. 

  95. static const double rgb2yuv_table[8][9]={

  96.     {0.7152, 0.0722, 0.2126, -0.386, 0.5, -0.115, -0.454, -0.046, 0.5}, //ITU709

  97.     {0.7152, 0.0722, 0.2126, -0.386, 0.5, -0.115, -0.454, -0.046, 0.5}, //ITU709

  98.     {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5}, //DEFAULT / ITU601 / ITU624 / SMPTE 170M

  99.     {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5}, //DEFAULT / ITU601 / ITU624 / SMPTE 170M

  100.     {0.59  , 0.11  , 0.30  , -0.331, 0.5, -0.169, -0.421, -0.079, 0.5}, //FCC

  101.     {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5}, //DEFAULT / ITU601 / ITU624 / SMPTE 170M

  102.     {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5}, //DEFAULT / ITU601 / ITU624 / SMPTE 170M

  103.     {0.701 , 0.087 , 0.212 , -0.384, 0.5, -0.116, -0.445, -0.055, 0.5}, //SMPTE 240M

  104. };

  105. 

  106. /*

  107. NOTES

  108. Special versions: fast Y 1:1 scaling (no interpolation in y direction)

  109. 

  110. TODO

  111. more intelligent misalignment avoidance for the horizontal scaler

  112. write special vertical cubic upscale version

  113. optimize C code (YV12 / minmax)

  114. add support for packed pixel YUV input & output

  115. add support for Y8 output

  116. optimize BGR24 & BGR32

  117. add BGR4 output support

  118. write special BGR->BGR scaler

  119. */

  120. 

  121. DECLARE_ALIGNED(8, static const uint8_t, dither_2x2_4)[2][8]={

  122. {  1,   3,   1,   3,   1,   3,   1,   3, },

  123. {  2,   0,   2,   0,   2,   0,   2,   0, },

  124. };

  125. 

  126. DECLARE_ALIGNED(8, static const uint8_t, dither_2x2_8)[2][8]={

  127. {  6,   2,   6,   2,   6,   2,   6,   2, },

  128. {  0,   4,   0,   4,   0,   4,   0,   4, },

  129. };

  130. 

  131. DECLARE_ALIGNED(8, const uint8_t, dither_4x4_16)[4][8]={

  132. {  8,   4,  11,   7,   8,   4,  11,   7, },

  133. {  2,  14,   1,  13,   2,  14,   1,  13, },

  134. { 10,   6,   9,   5,  10,   6,   9,   5, },

  135. {  0,  12,   3,  15,   0,  12,   3,  15, },

  136. };

  137. 

  138. DECLARE_ALIGNED(8, const uint8_t, dither_8x8_32)[8][8]={

  139. { 17,   9,  23,  15,  16,   8,  22,  14, },

  140. {  5,  29,   3,  27,   4,  28,   2,  26, },

  141. { 21,  13,  19,  11,  20,  12,  18,  10, },

  142. {  0,  24,   6,  30,   1,  25,   7,  31, },

  143. { 16,   8,  22,  14,  17,   9,  23,  15, },

  144. {  4,  28,   2,  26,   5,  29,   3,  27, },

  145. { 20,  12,  18,  10,  21,  13,  19,  11, },

  146. {  1,  25,   7,  31,   0,  24,   6,  30, },

  147. };

  148. 

  149. DECLARE_ALIGNED(8, const uint8_t, dither_8x8_73)[8][8]={

  150. {  0,  55,  14,  68,   3,  58,  17,  72, },

  151. { 37,  18,  50,  32,  40,  22,  54,  35, },

  152. {  9,  64,   5,  59,  13,  67,   8,  63, },

  153. { 46,  27,  41,  23,  49,  31,  44,  26, },

  154. {  2,  57,  16,  71,   1,  56,  15,  70, },

  155. { 39,  21,  52,  34,  38,  19,  51,  33, },

  156. { 11,  66,   7,  62,  10,  65,   6,  60, },

  157. { 48,  30,  43,  25,  47,  29,  42,  24, },

  158. };

  159. 

  160. #if 1

  161. DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220)[8][8]={

  162. {117,  62, 158, 103, 113,  58, 155, 100, },

  163. { 34, 199,  21, 186,  31, 196,  17, 182, },

  164. {144,  89, 131,  76, 141,  86, 127,  72, },

  165. {  0, 165,  41, 206,  10, 175,  52, 217, },

  166. {110,  55, 151,  96, 120,  65, 162, 107, },

  167. { 28, 193,  14, 179,  38, 203,  24, 189, },

  168. {138,  83, 124,  69, 148,  93, 134,  79, },

  169. {  7, 172,  48, 213,   3, 168,  45, 210, },

  170. };

  171. #elif 1

  172. // tries to correct a gamma of 1.5

  173. DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220)[8][8]={

  174. {  0, 143,  18, 200,   2, 156,  25, 215, },

  175. { 78,  28, 125,  64,  89,  36, 138,  74, },

  176. { 10, 180,   3, 161,  16, 195,   8, 175, },

  177. {109,  51,  93,  38, 121,  60, 105,  47, },

  178. {  1, 152,  23, 210,   0, 147,  20, 205, },

  179. { 85,  33, 134,  71,  81,  30, 130,  67, },

  180. { 14, 190,   6, 171,  12, 185,   5, 166, },

  181. {117,  57, 101,  44, 113,  54,  97,  41, },

  182. };

  183. #elif 1

  184. // tries to correct a gamma of 2.0

  185. DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220)[8][8]={

  186. {  0, 124,   8, 193,   0, 140,  12, 213, },

  187. { 55,  14, 104,  42,  66,  19, 119,  52, },

  188. {  3, 168,   1, 145,   6, 187,   3, 162, },

  189. { 86,  31,  70,  21,  99,  39,  82,  28, },

  190. {  0, 134,  11, 206,   0, 129,   9, 200, },

  191. { 62,  17, 114,  48,  58,  16, 109,  45, },

  192. {  5, 181,   2, 157,   4, 175,   1, 151, },

  193. { 95,  36,  78,  26,  90,  34,  74,  24, },

  194. };

  195. #else

  196. // tries to correct a gamma of 2.5

  197. DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220)[8][8]={

  198. {  0, 107,   3, 187,   0, 125,   6, 212, },

  199. { 39,   7,  86,  28,  49,  11, 102,  36, },

  200. {  1, 158,   0, 131,   3, 180,   1, 151, },

  201. { 68,  19,  52,  12,  81,  25,  64,  17, },

  202. {  0, 119,   5, 203,   0, 113,   4, 195, },

  203. { 45,   9,  96,  33,  42,   8,  91,  30, },

  204. {  2, 172,   1, 144,   2, 165,   0, 137, },

  205. { 77,  23,  60,  15,  72,  21,  56,  14, },

  206. };

  207. #endif

  208. 

  209. static av_always_inline void yuv2yuvX16inC_template(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,

  210.                                                     const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,

  211.                                                     const int16_t **alpSrc, uint16_t *dest, uint16_t *uDest, uint16_t *vDest, uint16_t *aDest,

  212.                                                     int dstW, int chrDstW, int big_endian, int output_bits)

  213. {

  214.     //FIXME Optimize (just quickly written not optimized..)

  215.     int i;

  216.     int shift = 11 + 16 - output_bits;

  217. 

  218. #define output_pixel(pos, val) \

  219.     if (big_endian) { \

  220.         if (output_bits == 16) { \

  221.             AV_WB16(pos, av_clip_uint16(val >> shift)); \

  222.         } else { \

  223.             AV_WB16(pos, av_clip_uintp2(val >> shift, output_bits)); \

  224.         } \

  225.     } else { \

  226.         if (output_bits == 16) { \

  227.             AV_WL16(pos, av_clip_uint16(val >> shift)); \

  228.         } else { \

  229.             AV_WL16(pos, av_clip_uintp2(val >> shift, output_bits)); \

  230.         } \

  231.     }

  232.     for (i = 0; i < dstW; i++) {

  233.         int val = 1 << (26-output_bits);

  234.         int j;

  235. 

  236.         for (j = 0; j < lumFilterSize; j++)

  237.             val += lumSrc[j][i] * lumFilter[j];

  238. 

  239.         output_pixel(&dest[i], val);

  240.     }

  241. 

  242.     if (uDest) {

  243.         for (i = 0; i < chrDstW; i++) {

  244.             int u = 1 << (26-output_bits);

  245.             int v = 1 << (26-output_bits);

  246.             int j;

  247. 

  248.             for (j = 0; j < chrFilterSize; j++) {

  249.                 u += chrSrc[j][i       ] * chrFilter[j];

  250.                 v += chrSrc[j][i + VOFW] * chrFilter[j];

  251.             }

  252. 

  253.             output_pixel(&uDest[i], u);

  254.             output_pixel(&vDest[i], v);

  255.         }

  256.     }

  257. 

  258.     if (CONFIG_SWSCALE_ALPHA && aDest) {

  259.         for (i = 0; i < dstW; i++) {

  260.             int val = 1 << (26-output_bits);

  261.             int j;

  262. 

  263.             for (j = 0; j < lumFilterSize; j++)

  264.                 val += alpSrc[j][i] * lumFilter[j];

  265. 

  266.             output_pixel(&aDest[i], val);

  267.         }

  268.     }

  269. }

  270. 

  271. #define yuv2NBPS(bits, BE_LE, is_be) \

  272. static void yuv2yuvX ## bits ## BE_LE ## _c(const int16_t *lumFilter, \

  273.                               const int16_t **lumSrc, int lumFilterSize, \

  274.                               const int16_t *chrFilter, const int16_t **chrSrc, \

  275.                               int chrFilterSize, const int16_t **alpSrc, \

  276.                               uint16_t *dest, uint16_t *uDest, uint16_t *vDest, \

  277.                               uint16_t *aDest, int dstW, int chrDstW) \

  278. { \

  279.     yuv2yuvX16inC_template(lumFilter, lumSrc, lumFilterSize, \

  280.                            chrFilter, chrSrc, chrFilterSize, \

  281.                            alpSrc, \

  282.                            dest, uDest, vDest, aDest, \

  283.                            dstW, chrDstW, is_be, bits); \

  284. }

  285. yuv2NBPS( 9, BE, 1);

  286. yuv2NBPS( 9, LE, 0);

  287. yuv2NBPS(10, BE, 1);

  288. yuv2NBPS(10, LE, 0);

  289. yuv2NBPS(16, BE, 1);

  290. yuv2NBPS(16, LE, 0);

  291. 

  292. static inline void yuv2yuvX16inC(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,

  293.                                  const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,

  294.                                  const int16_t **alpSrc, uint16_t *dest, uint16_t *uDest, uint16_t *vDest, uint16_t *aDest, int dstW, int chrDstW,

  295.                                  enum PixelFormat dstFormat)

  296. {

  297. #define conv16(bits) \

  298.     if (isBE(dstFormat)) { \

  299.         yuv2yuvX ## bits ## BE_c(lumFilter, lumSrc, lumFilterSize, \

  300.                                chrFilter, chrSrc, chrFilterSize, \

  301.                                alpSrc, \

  302.                                dest, uDest, vDest, aDest, \

  303.                                dstW, chrDstW); \

  304.     } else { \

  305.         yuv2yuvX ## bits ## LE_c(lumFilter, lumSrc, lumFilterSize, \

  306.                                chrFilter, chrSrc, chrFilterSize, \

  307.                                alpSrc, \

  308.                                dest, uDest, vDest, aDest, \

  309.                                dstW, chrDstW); \

  310.     }

  311.     if (is16BPS(dstFormat)) {

  312.         conv16(16);

  313.     } else if (av_pix_fmt_descriptors[dstFormat].comp[0].depth_minus1 == 8) {

  314.         conv16(9);

  315.     } else {

  316.         conv16(10);

  317.     }

  318. #undef conv16

  319. }

  320. 

  321. static inline void yuv2yuvXinC(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,

  322.                                const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,

  323.                                const int16_t **alpSrc, uint8_t *dest, uint8_t *uDest, uint8_t *vDest, uint8_t *aDest, int dstW, int chrDstW)

  324. {

  325.     //FIXME Optimize (just quickly written not optimized..)

  326.     int i;

  327.     for (i=0; i<dstW; i++) {

  328.         int val=1<<18;

  329.         int j;

  330.         for (j=0; j<lumFilterSize; j++)

  331.             val += lumSrc[j][i] * lumFilter[j];

  332. 

  333.         dest[i]= av_clip_uint8(val>>19);

  334.     }

  335. 

  336.     if (uDest)

  337.         for (i=0; i<chrDstW; i++) {

  338.             int u=1<<18;

  339.             int v=1<<18;

  340.             int j;

  341.             for (j=0; j<chrFilterSize; j++) {

  342.                 u += chrSrc[j][i] * chrFilter[j];

  343.                 v += chrSrc[j][i + VOFW] * chrFilter[j];

  344.             }

  345. 

  346.             uDest[i]= av_clip_uint8(u>>19);

  347.             vDest[i]= av_clip_uint8(v>>19);

  348.         }

  349. 

  350.     if (CONFIG_SWSCALE_ALPHA && aDest)

  351.         for (i=0; i<dstW; i++) {

  352.             int val=1<<18;

  353.             int j;

  354.             for (j=0; j<lumFilterSize; j++)

  355.                 val += alpSrc[j][i] * lumFilter[j];

  356. 

  357.             aDest[i]= av_clip_uint8(val>>19);

  358.         }

  359. 

  360. }

  361. 

  362. static inline void yuv2nv12XinC(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,

  363.                                 const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,

  364.                                 uint8_t *dest, uint8_t *uDest, int dstW, int chrDstW, int dstFormat)

  365. {

  366.     //FIXME Optimize (just quickly written not optimized..)

  367.     int i;

  368.     for (i=0; i<dstW; i++) {

  369.         int val=1<<18;

  370.         int j;

  371.         for (j=0; j<lumFilterSize; j++)

  372.             val += lumSrc[j][i] * lumFilter[j];

  373. 

  374.         dest[i]= av_clip_uint8(val>>19);

  375.     }

  376. 

  377.     if (!uDest)

  378.         return;

  379. 

  380.     if (dstFormat == PIX_FMT_NV12)

  381.         for (i=0; i<chrDstW; i++) {

  382.             int u=1<<18;

  383.             int v=1<<18;

  384.             int j;

  385.             for (j=0; j<chrFilterSize; j++) {

  386.                 u += chrSrc[j][i] * chrFilter[j];

  387.                 v += chrSrc[j][i + VOFW] * chrFilter[j];

  388.             }

  389. 

  390.             uDest[2*i]= av_clip_uint8(u>>19);

  391.             uDest[2*i+1]= av_clip_uint8(v>>19);

  392.         }

  393.     else

  394.         for (i=0; i<chrDstW; i++) {

  395.             int u=1<<18;

  396.             int v=1<<18;

  397.             int j;

  398.             for (j=0; j<chrFilterSize; j++) {

  399.                 u += chrSrc[j][i] * chrFilter[j];

  400.                 v += chrSrc[j][i + VOFW] * chrFilter[j];

  401.             }

  402. 

  403.             uDest[2*i]= av_clip_uint8(v>>19);

  404.             uDest[2*i+1]= av_clip_uint8(u>>19);

  405.         }

  406. }

  407. 

  408. #define YSCALE_YUV_2_PACKEDX_NOCLIP_C(type,alpha) \

  409.     for (i=0; i<(dstW>>1); i++) {\

  410.         int j;\

  411.         int Y1 = 1<<18;\

  412.         int Y2 = 1<<18;\

  413.         int U  = 1<<18;\

  414.         int V  = 1<<18;\

  415.         int av_unused A1, A2;\

  416.         type av_unused *r, *b, *g;\

  417.         const int i2= 2*i;\

  418.         \

  419.         for (j=0; j<lumFilterSize; j++) {\

  420.             Y1 += lumSrc[j][i2] * lumFilter[j];\

  421.             Y2 += lumSrc[j][i2+1] * lumFilter[j];\

  422.         }\

  423.         for (j=0; j<chrFilterSize; j++) {\

  424.             U += chrSrc[j][i] * chrFilter[j];\

  425.             V += chrSrc[j][i+VOFW] * chrFilter[j];\

  426.         }\

  427.         Y1>>=19;\

  428.         Y2>>=19;\

  429.         U >>=19;\

  430.         V >>=19;\

  431.         if (alpha) {\

  432.             A1 = 1<<18;\

  433.             A2 = 1<<18;\

  434.             for (j=0; j<lumFilterSize; j++) {\

  435.                 A1 += alpSrc[j][i2  ] * lumFilter[j];\

  436.                 A2 += alpSrc[j][i2+1] * lumFilter[j];\

  437.             }\

  438.             A1>>=19;\

  439.             A2>>=19;\

  440.         }

  441. 

  442. #define YSCALE_YUV_2_PACKEDX_C(type,alpha) \

  443.         YSCALE_YUV_2_PACKEDX_NOCLIP_C(type,alpha)\

  444.         if ((Y1|Y2|U|V)&256) {\

  445.             if (Y1>255)   Y1=255; \

  446.             else if (Y1<0)Y1=0;   \

  447.             if (Y2>255)   Y2=255; \

  448.             else if (Y2<0)Y2=0;   \

  449.             if (U>255)    U=255;  \

  450.             else if (U<0) U=0;    \

  451.             if (V>255)    V=255;  \

  452.             else if (V<0) V=0;    \

  453.         }\

  454.         if (alpha && ((A1|A2)&256)) {\

  455.             A1=av_clip_uint8(A1);\

  456.             A2=av_clip_uint8(A2);\

  457.         }

  458. 

  459. #define YSCALE_YUV_2_PACKEDX_FULL_C(rnd,alpha) \

  460.     for (i=0; i<dstW; i++) {\

  461.         int j;\

  462.         int Y = 0;\

  463.         int U = -128<<19;\

  464.         int V = -128<<19;\

  465.         int av_unused A;\

  466.         int R,G,B;\

  467.         \

  468.         for (j=0; j<lumFilterSize; j++) {\

  469.             Y += lumSrc[j][i     ] * lumFilter[j];\

  470.         }\

  471.         for (j=0; j<chrFilterSize; j++) {\

  472.             U += chrSrc[j][i     ] * chrFilter[j];\

  473.             V += chrSrc[j][i+VOFW] * chrFilter[j];\

  474.         }\

  475.         Y >>=10;\

  476.         U >>=10;\

  477.         V >>=10;\

  478.         if (alpha) {\

  479.             A = rnd;\

  480.             for (j=0; j<lumFilterSize; j++)\

  481.                 A += alpSrc[j][i     ] * lumFilter[j];\

  482.             A >>=19;\

  483.             if (A&256)\

  484.                 A = av_clip_uint8(A);\

  485.         }

  486. 

  487. #define YSCALE_YUV_2_RGBX_FULL_C(rnd,alpha) \

  488.     YSCALE_YUV_2_PACKEDX_FULL_C(rnd>>3,alpha)\

  489.         Y-= c->yuv2rgb_y_offset;\

  490.         Y*= c->yuv2rgb_y_coeff;\

  491.         Y+= rnd;\

  492.         R= Y + V*c->yuv2rgb_v2r_coeff;\

  493.         G= Y + V*c->yuv2rgb_v2g_coeff + U*c->yuv2rgb_u2g_coeff;\

  494.         B= Y +                          U*c->yuv2rgb_u2b_coeff;\

  495.         if ((R|G|B)&(0xC0000000)) {\

  496.             if (R>=(256<<22))   R=(256<<22)-1; \

  497.             else if (R<0)R=0;   \

  498.             if (G>=(256<<22))   G=(256<<22)-1; \

  499.             else if (G<0)G=0;   \

  500.             if (B>=(256<<22))   B=(256<<22)-1; \

  501.             else if (B<0)B=0;   \

  502.         }

  503. 

  504. #define YSCALE_YUV_2_GRAY16_C \

  505.     for (i=0; i<(dstW>>1); i++) {\

  506.         int j;\

  507.         int Y1 = 1<<18;\

  508.         int Y2 = 1<<18;\

  509.         int U  = 1<<18;\

  510.         int V  = 1<<18;\

  511.         \

  512.         const int i2= 2*i;\

  513.         \

  514.         for (j=0; j<lumFilterSize; j++) {\

  515.             Y1 += lumSrc[j][i2] * lumFilter[j];\

  516.             Y2 += lumSrc[j][i2+1] * lumFilter[j];\

  517.         }\

  518.         Y1>>=11;\

  519.         Y2>>=11;\

  520.         if ((Y1|Y2|U|V)&65536) {\

  521.             if (Y1>65535)   Y1=65535; \

  522.             else if (Y1<0)Y1=0;   \

  523.             if (Y2>65535)   Y2=65535; \

  524.             else if (Y2<0)Y2=0;   \

  525.         }

  526. 

  527. #define YSCALE_YUV_2_RGBX_C(type,alpha) \

  528.     YSCALE_YUV_2_PACKEDX_C(type,alpha)  /* FIXME fix tables so that clipping is not needed and then use _NOCLIP*/\

  529.     r = (type *)c->table_rV[V];   \

  530.     g = (type *)(c->table_gU[U] + c->table_gV[V]); \

  531.     b = (type *)c->table_bU[U];

  532. 

  533. #define YSCALE_YUV_2_PACKED2_C(type,alpha)   \

  534.     for (i=0; i<(dstW>>1); i++) { \

  535.         const int i2= 2*i;       \

  536.         int Y1= (buf0[i2  ]*yalpha1+buf1[i2  ]*yalpha)>>19;           \

  537.         int Y2= (buf0[i2+1]*yalpha1+buf1[i2+1]*yalpha)>>19;           \

  538.         int U= (uvbuf0[i     ]*uvalpha1+uvbuf1[i     ]*uvalpha)>>19;  \

  539.         int V= (uvbuf0[i+VOFW]*uvalpha1+uvbuf1[i+VOFW]*uvalpha)>>19;  \

  540.         type av_unused *r, *b, *g;                                    \

  541.         int av_unused A1, A2;                                         \

  542.         if (alpha) {\

  543.             A1= (abuf0[i2  ]*yalpha1+abuf1[i2  ]*yalpha)>>19;         \

  544.             A2= (abuf0[i2+1]*yalpha1+abuf1[i2+1]*yalpha)>>19;         \

  545.         }

  546. 

  547. #define YSCALE_YUV_2_GRAY16_2_C   \

  548.     for (i=0; i<(dstW>>1); i++) { \

  549.         const int i2= 2*i;       \

  550.         int Y1= (buf0[i2  ]*yalpha1+buf1[i2  ]*yalpha)>>11;           \

  551.         int Y2= (buf0[i2+1]*yalpha1+buf1[i2+1]*yalpha)>>11;

  552. 

  553. #define YSCALE_YUV_2_RGB2_C(type,alpha) \

  554.     YSCALE_YUV_2_PACKED2_C(type,alpha)\

  555.     r = (type *)c->table_rV[V];\

  556.     g = (type *)(c->table_gU[U] + c->table_gV[V]);\

  557.     b = (type *)c->table_bU[U];

  558. 

  559. #define YSCALE_YUV_2_PACKED1_C(type,alpha) \

  560.     for (i=0; i<(dstW>>1); i++) {\

  561.         const int i2= 2*i;\

  562.         int Y1= buf0[i2  ]>>7;\

  563.         int Y2= buf0[i2+1]>>7;\

  564.         int U= (uvbuf1[i     ])>>7;\

  565.         int V= (uvbuf1[i+VOFW])>>7;\

  566.         type av_unused *r, *b, *g;\

  567.         int av_unused A1, A2;\

  568.         if (alpha) {\

  569.             A1= abuf0[i2  ]>>7;\

  570.             A2= abuf0[i2+1]>>7;\

  571.         }

  572. 

  573. #define YSCALE_YUV_2_GRAY16_1_C \

  574.     for (i=0; i<(dstW>>1); i++) {\

  575.         const int i2= 2*i;\

  576.         int Y1= buf0[i2  ]<<1;\

  577.         int Y2= buf0[i2+1]<<1;

  578. 

  579. #define YSCALE_YUV_2_RGB1_C(type,alpha) \

  580.     YSCALE_YUV_2_PACKED1_C(type,alpha)\

  581.     r = (type *)c->table_rV[V];\

  582.     g = (type *)(c->table_gU[U] + c->table_gV[V]);\

  583.     b = (type *)c->table_bU[U];

  584. 

  585. #define YSCALE_YUV_2_PACKED1B_C(type,alpha) \

  586.     for (i=0; i<(dstW>>1); i++) {\

  587.         const int i2= 2*i;\

  588.         int Y1= buf0[i2  ]>>7;\

  589.         int Y2= buf0[i2+1]>>7;\

  590.         int U= (uvbuf0[i     ] + uvbuf1[i     ])>>8;\

  591.         int V= (uvbuf0[i+VOFW] + uvbuf1[i+VOFW])>>8;\

  592.         type av_unused *r, *b, *g;\

  593.         int av_unused A1, A2;\

  594.         if (alpha) {\

  595.             A1= abuf0[i2  ]>>7;\

  596.             A2= abuf0[i2+1]>>7;\

  597.         }

  598. 

  599. #define YSCALE_YUV_2_RGB1B_C(type,alpha) \

  600.     YSCALE_YUV_2_PACKED1B_C(type,alpha)\

  601.     r = (type *)c->table_rV[V];\

  602.     g = (type *)(c->table_gU[U] + c->table_gV[V]);\

  603.     b = (type *)c->table_bU[U];

  604. 

  605. #define YSCALE_YUV_2_MONO2_C \

  606.     const uint8_t * const d128=dither_8x8_220[y&7];\

  607.     uint8_t *g= c->table_gU[128] + c->table_gV[128];\

  608.     for (i=0; i<dstW-7; i+=8) {\

  609.         int acc;\

  610.         acc =       g[((buf0[i  ]*yalpha1+buf1[i  ]*yalpha)>>19) + d128[0]];\

  611.         acc+= acc + g[((buf0[i+1]*yalpha1+buf1[i+1]*yalpha)>>19) + d128[1]];\

  612.         acc+= acc + g[((buf0[i+2]*yalpha1+buf1[i+2]*yalpha)>>19) + d128[2]];\

  613.         acc+= acc + g[((buf0[i+3]*yalpha1+buf1[i+3]*yalpha)>>19) + d128[3]];\

  614.         acc+= acc + g[((buf0[i+4]*yalpha1+buf1[i+4]*yalpha)>>19) + d128[4]];\

  615.         acc+= acc + g[((buf0[i+5]*yalpha1+buf1[i+5]*yalpha)>>19) + d128[5]];\

  616.         acc+= acc + g[((buf0[i+6]*yalpha1+buf1[i+6]*yalpha)>>19) + d128[6]];\

  617.         acc+= acc + g[((buf0[i+7]*yalpha1+buf1[i+7]*yalpha)>>19) + d128[7]];\

  618.         ((uint8_t*)dest)[0]= c->dstFormat == PIX_FMT_MONOBLACK ? acc : ~acc;\

  619.         dest++;\

  620.     }

  621. 

  622. #define YSCALE_YUV_2_MONOX_C \

  623.     const uint8_t * const d128=dither_8x8_220[y&7];\

  624.     uint8_t *g= c->table_gU[128] + c->table_gV[128];\

  625.     int acc=0;\

  626.     for (i=0; i<dstW-1; i+=2) {\

  627.         int j;\

  628.         int Y1=1<<18;\

  629.         int Y2=1<<18;\

  630. \

  631.         for (j=0; j<lumFilterSize; j++) {\

  632.             Y1 += lumSrc[j][i] * lumFilter[j];\

  633.             Y2 += lumSrc[j][i+1] * lumFilter[j];\

  634.         }\

  635.         Y1>>=19;\

  636.         Y2>>=19;\

  637.         if ((Y1|Y2)&256) {\

  638.             if (Y1>255)   Y1=255;\

  639.             else if (Y1<0)Y1=0;\

  640.             if (Y2>255)   Y2=255;\

  641.             else if (Y2<0)Y2=0;\

  642.         }\

  643.         acc+= acc + g[Y1+d128[(i+0)&7]];\

  644.         acc+= acc + g[Y2+d128[(i+1)&7]];\

  645.         if ((i&7)==6) {\

  646.             ((uint8_t*)dest)[0]= c->dstFormat == PIX_FMT_MONOBLACK ? acc : ~acc;\

  647.             dest++;\

  648.         }\

  649.     }

  650. 

  651. #define YSCALE_YUV_2_ANYRGB_C(func, func2, func_g16, func_monoblack)\

  652.     switch(c->dstFormat) {\

  653.     case PIX_FMT_RGB48BE:\

  654.     case PIX_FMT_RGB48LE:\

  655.         func(uint8_t,0)\

  656.             ((uint8_t*)dest)[ 0]= r[Y1];\

  657.             ((uint8_t*)dest)[ 1]= r[Y1];\

  658.             ((uint8_t*)dest)[ 2]= g[Y1];\

  659.             ((uint8_t*)dest)[ 3]= g[Y1];\

  660.             ((uint8_t*)dest)[ 4]= b[Y1];\

  661.             ((uint8_t*)dest)[ 5]= b[Y1];\

  662.             ((uint8_t*)dest)[ 6]= r[Y2];\

  663.             ((uint8_t*)dest)[ 7]= r[Y2];\

  664.             ((uint8_t*)dest)[ 8]= g[Y2];\

  665.             ((uint8_t*)dest)[ 9]= g[Y2];\

  666.             ((uint8_t*)dest)[10]= b[Y2];\

  667.             ((uint8_t*)dest)[11]= b[Y2];\

  668.             dest+=12;\

  669.         }\

  670.         break;\

  671.     case PIX_FMT_BGR48BE:\

  672.     case PIX_FMT_BGR48LE:\

  673.         func(uint8_t,0)\

  674.             ((uint8_t*)dest)[ 0] = ((uint8_t*)dest)[ 1] = b[Y1];\

  675.             ((uint8_t*)dest)[ 2] = ((uint8_t*)dest)[ 3] = g[Y1];\

  676.             ((uint8_t*)dest)[ 4] = ((uint8_t*)dest)[ 5] = r[Y1];\

  677.             ((uint8_t*)dest)[ 6] = ((uint8_t*)dest)[ 7] = b[Y2];\

  678.             ((uint8_t*)dest)[ 8] = ((uint8_t*)dest)[ 9] = g[Y2];\

  679.             ((uint8_t*)dest)[10] = ((uint8_t*)dest)[11] = r[Y2];\

  680.             dest+=12;\

  681.         }\

  682.         break;\

  683.     case PIX_FMT_RGBA:\

  684.     case PIX_FMT_BGRA:\

  685.         if (CONFIG_SMALL) {\

  686.             int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;\

  687.             func(uint32_t,needAlpha)\

  688.                 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (needAlpha ? (A1<<24) : 0);\

  689.                 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (needAlpha ? (A2<<24) : 0);\

  690.             }\

  691.         } else {\

  692.             if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {\

  693.                 func(uint32_t,1)\

  694.                     ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (A1<<24);\

  695.                     ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (A2<<24);\

  696.                 }\

  697.             } else {\

  698.                 func(uint32_t,0)\

  699.                     ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1];\

  700.                     ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2];\

  701.                 }\

  702.             }\

  703.         }\

  704.         break;\

  705.     case PIX_FMT_ARGB:\

  706.     case PIX_FMT_ABGR:\

  707.         if (CONFIG_SMALL) {\

  708.             int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;\

  709.             func(uint32_t,needAlpha)\

  710.                 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (needAlpha ? A1 : 0);\

  711.                 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (needAlpha ? A2 : 0);\

  712.             }\

  713.         } else {\

  714.             if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {\

  715.                 func(uint32_t,1)\

  716.                     ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + A1;\

  717.                     ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + A2;\

  718.                 }\

  719.             } else {\

  720.                 func(uint32_t,0)\

  721.                     ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1];\

  722.                     ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2];\

  723.                 }\

  724.             }\

  725.         }                \

  726.         break;\

  727.     case PIX_FMT_RGB24:\

  728.         func(uint8_t,0)\

  729.             ((uint8_t*)dest)[0]= r[Y1];\

  730.             ((uint8_t*)dest)[1]= g[Y1];\

  731.             ((uint8_t*)dest)[2]= b[Y1];\

  732.             ((uint8_t*)dest)[3]= r[Y2];\

  733.             ((uint8_t*)dest)[4]= g[Y2];\

  734.             ((uint8_t*)dest)[5]= b[Y2];\

  735.             dest+=6;\

  736.         }\

  737.         break;\

  738.     case PIX_FMT_BGR24:\

  739.         func(uint8_t,0)\

  740.             ((uint8_t*)dest)[0]= b[Y1];\

  741.             ((uint8_t*)dest)[1]= g[Y1];\

  742.             ((uint8_t*)dest)[2]= r[Y1];\

  743.             ((uint8_t*)dest)[3]= b[Y2];\

  744.             ((uint8_t*)dest)[4]= g[Y2];\

  745.             ((uint8_t*)dest)[5]= r[Y2];\

  746.             dest+=6;\

  747.         }\

  748.         break;\

  749.     case PIX_FMT_RGB565BE:\

  750.     case PIX_FMT_RGB565LE:\

  751.     case PIX_FMT_BGR565BE:\

  752.     case PIX_FMT_BGR565LE:\

  753.         {\

  754.             const int dr1= dither_2x2_8[y&1    ][0];\

  755.             const int dg1= dither_2x2_4[y&1    ][0];\

  756.             const int db1= dither_2x2_8[(y&1)^1][0];\

  757.             const int dr2= dither_2x2_8[y&1    ][1];\

  758.             const int dg2= dither_2x2_4[y&1    ][1];\

  759.             const int db2= dither_2x2_8[(y&1)^1][1];\

  760.             func(uint16_t,0)\

  761.                 ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\

  762.                 ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\

  763.             }\

  764.         }\

  765.         break;\

  766.     case PIX_FMT_RGB555BE:\

  767.     case PIX_FMT_RGB555LE:\

  768.     case PIX_FMT_BGR555BE:\

  769.     case PIX_FMT_BGR555LE:\

  770.         {\

  771.             const int dr1= dither_2x2_8[y&1    ][0];\

  772.             const int dg1= dither_2x2_8[y&1    ][1];\

  773.             const int db1= dither_2x2_8[(y&1)^1][0];\

  774.             const int dr2= dither_2x2_8[y&1    ][1];\

  775.             const int dg2= dither_2x2_8[y&1    ][0];\

  776.             const int db2= dither_2x2_8[(y&1)^1][1];\

  777.             func(uint16_t,0)\

  778.                 ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\

  779.                 ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\

  780.             }\

  781.         }\

  782.         break;\

  783.     case PIX_FMT_RGB444BE:\

  784.     case PIX_FMT_RGB444LE:\

  785.     case PIX_FMT_BGR444BE:\

  786.     case PIX_FMT_BGR444LE:\

  787.         {\

  788.             const int dr1= dither_4x4_16[y&3    ][0];\

  789.             const int dg1= dither_4x4_16[y&3    ][1];\

  790.             const int db1= dither_4x4_16[(y&3)^3][0];\

  791.             const int dr2= dither_4x4_16[y&3    ][1];\

  792.             const int dg2= dither_4x4_16[y&3    ][0];\

  793.             const int db2= dither_4x4_16[(y&3)^3][1];\

  794.             func(uint16_t,0)\

  795.                 ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\

  796.                 ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\

  797.             }\

  798.         }\

  799.         break;\

  800.     case PIX_FMT_RGB8:\

  801.     case PIX_FMT_BGR8:\

  802.         {\

  803.             const uint8_t * const d64= dither_8x8_73[y&7];\

  804.             const uint8_t * const d32= dither_8x8_32[y&7];\

  805.             func(uint8_t,0)\

  806.                 ((uint8_t*)dest)[i2+0]= r[Y1+d32[(i2+0)&7]] + g[Y1+d32[(i2+0)&7]] + b[Y1+d64[(i2+0)&7]];\

  807.                 ((uint8_t*)dest)[i2+1]= r[Y2+d32[(i2+1)&7]] + g[Y2+d32[(i2+1)&7]] + b[Y2+d64[(i2+1)&7]];\

  808.             }\

  809.         }\

  810.         break;\

  811.     case PIX_FMT_RGB4:\

  812.     case PIX_FMT_BGR4:\

  813.         {\

  814.             const uint8_t * const d64= dither_8x8_73 [y&7];\

  815.             const uint8_t * const d128=dither_8x8_220[y&7];\

  816.             func(uint8_t,0)\

  817.                 ((uint8_t*)dest)[i]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]]\

  818.                                  + ((r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]])<<4);\

  819.             }\

  820.         }\

  821.         break;\

  822.     case PIX_FMT_RGB4_BYTE:\

  823.     case PIX_FMT_BGR4_BYTE:\

  824.         {\

  825.             const uint8_t * const d64= dither_8x8_73 [y&7];\

  826.             const uint8_t * const d128=dither_8x8_220[y&7];\

  827.             func(uint8_t,0)\

  828.                 ((uint8_t*)dest)[i2+0]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]];\

  829.                 ((uint8_t*)dest)[i2+1]= r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]];\

  830.             }\

  831.         }\

  832.         break;\

  833.     case PIX_FMT_MONOBLACK:\

  834.     case PIX_FMT_MONOWHITE:\

  835.         {\

  836.             func_monoblack\

  837.         }\

  838.         break;\

  839.     case PIX_FMT_YUYV422:\

  840.         func2\

  841.             ((uint8_t*)dest)[2*i2+0]= Y1;\

  842.             ((uint8_t*)dest)[2*i2+1]= U;\

  843.             ((uint8_t*)dest)[2*i2+2]= Y2;\

  844.             ((uint8_t*)dest)[2*i2+3]= V;\

  845.         }                \

  846.         break;\

  847.     case PIX_FMT_UYVY422:\

  848.         func2\

  849.             ((uint8_t*)dest)[2*i2+0]= U;\

  850.             ((uint8_t*)dest)[2*i2+1]= Y1;\

  851.             ((uint8_t*)dest)[2*i2+2]= V;\

  852.             ((uint8_t*)dest)[2*i2+3]= Y2;\

  853.         }                \

  854.         break;\

  855.     case PIX_FMT_GRAY16BE:\

  856.         func_g16\

  857.             ((uint8_t*)dest)[2*i2+0]= Y1>>8;\

  858.             ((uint8_t*)dest)[2*i2+1]= Y1;\

  859.             ((uint8_t*)dest)[2*i2+2]= Y2>>8;\

  860.             ((uint8_t*)dest)[2*i2+3]= Y2;\

  861.         }                \

  862.         break;\

  863.     case PIX_FMT_GRAY16LE:\

  864.         func_g16\

  865.             ((uint8_t*)dest)[2*i2+0]= Y1;\

  866.             ((uint8_t*)dest)[2*i2+1]= Y1>>8;\

  867.             ((uint8_t*)dest)[2*i2+2]= Y2;\

  868.             ((uint8_t*)dest)[2*i2+3]= Y2>>8;\

  869.         }                \

  870.         break;\

  871.     }

  872. 

  873. static inline void yuv2packedXinC(SwsContext *c, const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,

  874.                                   const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,

  875.                                   const int16_t **alpSrc, uint8_t *dest, int dstW, int y)

  876. {

  877.     int i;

  878.     YSCALE_YUV_2_ANYRGB_C(YSCALE_YUV_2_RGBX_C, YSCALE_YUV_2_PACKEDX_C(void,0), YSCALE_YUV_2_GRAY16_C, YSCALE_YUV_2_MONOX_C)

  879. }

  880. 

  881. static inline void yuv2rgbXinC_full(SwsContext *c, const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,

  882.                                     const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,

  883.                                     const int16_t **alpSrc, uint8_t *dest, int dstW, int y)

  884. {

  885.     int i;

  886.     int step= c->dstFormatBpp/8;

  887.     int aidx= 3;

  888. 

  889.     switch(c->dstFormat) {

  890.     case PIX_FMT_ARGB:

  891.         dest++;

  892.         aidx= 0;

  893.     case PIX_FMT_RGB24:

  894.         aidx--;

  895.     case PIX_FMT_RGBA:

  896.         if (CONFIG_SMALL) {

  897.             int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;

  898.             YSCALE_YUV_2_RGBX_FULL_C(1<<21, needAlpha)

  899.                 dest[aidx]= needAlpha ? A : 255;

  900.                 dest[0]= R>>22;

  901.                 dest[1]= G>>22;

  902.                 dest[2]= B>>22;

  903.                 dest+= step;

  904.             }

  905.         } else {

  906.             if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {

  907.                 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 1)

  908.                     dest[aidx]= A;

  909.                     dest[0]= R>>22;

  910.                     dest[1]= G>>22;

  911.                     dest[2]= B>>22;

  912.                     dest+= step;

  913.                 }

  914.             } else {

  915.                 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 0)

  916.                     dest[aidx]= 255;

  917.                     dest[0]= R>>22;

  918.                     dest[1]= G>>22;

  919.                     dest[2]= B>>22;

  920.                     dest+= step;

  921.                 }

  922.             }

  923.         }

  924.         break;

  925.     case PIX_FMT_ABGR:

  926.         dest++;

  927.         aidx= 0;

  928.     case PIX_FMT_BGR24:

  929.         aidx--;

  930.     case PIX_FMT_BGRA:

  931.         if (CONFIG_SMALL) {

  932.             int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;

  933.             YSCALE_YUV_2_RGBX_FULL_C(1<<21, needAlpha)

  934.                 dest[aidx]= needAlpha ? A : 255;

  935.                 dest[0]= B>>22;

  936.                 dest[1]= G>>22;

  937.                 dest[2]= R>>22;

  938.                 dest+= step;

  939.             }

  940.         } else {

  941.             if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {

  942.                 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 1)

  943.                     dest[aidx]= A;

  944.                     dest[0]= B>>22;

  945.                     dest[1]= G>>22;

  946.                     dest[2]= R>>22;

  947.                     dest+= step;

  948.                 }

  949.             } else {

  950.                 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 0)

  951.                     dest[aidx]= 255;

  952.                     dest[0]= B>>22;

  953.                     dest[1]= G>>22;

  954.                     dest[2]= R>>22;

  955.                     dest+= step;

  956.                 }

  957.             }

  958.         }

  959.         break;

  960.     default:

  961.         assert(0);

  962.     }

  963. }

  964. 

  965. static void fillPlane(uint8_t* plane, int stride, int width, int height, int y, uint8_t val)

  966. {

  967.     int i;

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

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

  970.         memset(ptr, val, width);

  971.         ptr += stride;

  972.     }

  973. }

  974. 

  975. static inline void rgb48ToY(uint8_t *dst, const uint8_t *src, long width,

  976.                             uint32_t *unused)

  977. {

  978.     int i;

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

  980.         int r = src[i*6+0];

  981.         int g = src[i*6+2];

  982.         int b = src[i*6+4];

  983. 

  984.         dst[i] = (RY*r + GY*g + BY*b + (33<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;

  985.     }

  986. }

  987. 

  988. static inline void rgb48ToUV(uint8_t *dstU, uint8_t *dstV,

  989.                              const uint8_t *src1, const uint8_t *src2,

  990.                              long width, uint32_t *unused)

  991. {

  992.     int i;

  993.     assert(src1==src2);

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

  995.         int r = src1[6*i + 0];

  996.         int g = src1[6*i + 2];

  997.         int b = src1[6*i + 4];

  998. 

  999.         dstU[i] = (RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;

  1000.         dstV[i] = (RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;

  1001.     }

  1002. }

  1003. 

  1004. static inline void rgb48ToUV_half(uint8_t *dstU, uint8_t *dstV,

  1005.                                   const uint8_t *src1, const uint8_t *src2,

  1006.                                   long width, uint32_t *unused)

  1007. {

  1008.     int i;

  1009.     assert(src1==src2);

  1010.     for (i = 0; i < width; i++) {

  1011.         int r= src1[12*i + 0] + src1[12*i + 6];

  1012.         int g= src1[12*i + 2] + src1[12*i + 8];

  1013.         int b= src1[12*i + 4] + src1[12*i + 10];

  1014. 

  1015.         dstU[i]= (RU*r + GU*g + BU*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);

  1016.         dstV[i]= (RV*r + GV*g + BV*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);

  1017.     }

  1018. }

  1019. 

  1020. static inline void bgr48ToY(uint8_t *dst, const uint8_t *src, long width,

  1021.                             uint32_t *unused)

  1022. {

  1023.     int i;

  1024.     for (i = 0; i < width; i++) {

  1025.         int b = src[i*6+0];

  1026.         int g = src[i*6+2];

  1027.         int r = src[i*6+4];

  1028. 

  1029.         dst[i] = (RY*r + GY*g + BY*b + (33<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;

  1030.     }

  1031. }

  1032. 

  1033. static inline void bgr48ToUV(uint8_t *dstU, uint8_t *dstV,

  1034.                              const uint8_t *src1, const uint8_t *src2,

  1035.                              long width, uint32_t *unused)

  1036. {

  1037.     int i;

  1038.     for (i = 0; i < width; i++) {

  1039.         int b = src1[6*i + 0];

  1040.         int g = src1[6*i + 2];

  1041.         int r = src1[6*i + 4];

  1042. 

  1043.         dstU[i] = (RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;

  1044.         dstV[i] = (RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;

  1045.     }

  1046. }

  1047. 

  1048. static inline void bgr48ToUV_half(uint8_t *dstU, uint8_t *dstV,

  1049.                                   const uint8_t *src1, const uint8_t *src2,

  1050.                                   long width, uint32_t *unused)

  1051. {

  1052.     int i;

  1053.     for (i = 0; i < width; i++) {

  1054.         int b= src1[12*i + 0] + src1[12*i + 6];

  1055.         int g= src1[12*i + 2] + src1[12*i + 8];

  1056.         int r= src1[12*i + 4] + src1[12*i + 10];

  1057. 

  1058.         dstU[i]= (RU*r + GU*g + BU*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);

  1059.         dstV[i]= (RV*r + GV*g + BV*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);

  1060.     }

  1061. }

  1062. 

  1063. #define BGR2Y(type, name, shr, shg, shb, maskr, maskg, maskb, RY, GY, BY, S)\

  1064. static inline void name(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)\

  1065. {\

  1066.     int i;\

  1067.     for (i=0; i<width; i++) {\

  1068.         int b= (((const type*)src)[i]>>shb)&maskb;\

  1069.         int g= (((const type*)src)[i]>>shg)&maskg;\

  1070.         int r= (((const type*)src)[i]>>shr)&maskr;\

  1071. \

  1072.         dst[i]= (((RY)*r + (GY)*g + (BY)*b + (33<<((S)-1)))>>(S));\

  1073.     }\

  1074. }

  1075. 

  1076. BGR2Y(uint32_t, bgr32ToY,16, 0, 0, 0x00FF, 0xFF00, 0x00FF, RY<< 8, GY   , BY<< 8, RGB2YUV_SHIFT+8)

  1077. BGR2Y(uint32_t,bgr321ToY,16,16, 0, 0xFF00, 0x00FF, 0xFF00, RY    , GY<<8, BY    , RGB2YUV_SHIFT+8)

  1078. BGR2Y(uint32_t, rgb32ToY, 0, 0,16, 0x00FF, 0xFF00, 0x00FF, RY<< 8, GY   , BY<< 8, RGB2YUV_SHIFT+8)

  1079. BGR2Y(uint32_t,rgb321ToY, 0,16,16, 0xFF00, 0x00FF, 0xFF00, RY    , GY<<8, BY    , RGB2YUV_SHIFT+8)

  1080. BGR2Y(uint16_t, bgr16ToY, 0, 0, 0, 0x001F, 0x07E0, 0xF800, RY<<11, GY<<5, BY    , RGB2YUV_SHIFT+8)

  1081. BGR2Y(uint16_t, bgr15ToY, 0, 0, 0, 0x001F, 0x03E0, 0x7C00, RY<<10, GY<<5, BY    , RGB2YUV_SHIFT+7)

  1082. BGR2Y(uint16_t, rgb16ToY, 0, 0, 0, 0xF800, 0x07E0, 0x001F, RY    , GY<<5, BY<<11, RGB2YUV_SHIFT+8)

  1083. BGR2Y(uint16_t, rgb15ToY, 0, 0, 0, 0x7C00, 0x03E0, 0x001F, RY    , GY<<5, BY<<10, RGB2YUV_SHIFT+7)

  1084. 

  1085. static inline void abgrToA(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)

  1086. {

  1087.     int i;

  1088.     for (i=0; i<width; i++) {

  1089.         dst[i]= src[4*i];

  1090.     }

  1091. }

  1092. 

  1093. #define BGR2UV(type, name, shr, shg, shb, shp, maskr, maskg, maskb, RU, GU, BU, RV, GV, BV, S) \

  1094. static inline void name(uint8_t *dstU, uint8_t *dstV, const uint8_t *src, const uint8_t *dummy, long width, uint32_t *unused)\

  1095. {\

  1096.     int i;\

  1097.     for (i=0; i<width; i++) {\

  1098.         int b= ((((const type*)src)[i]>>shp)&maskb)>>shb;\

  1099.         int g= ((((const type*)src)[i]>>shp)&maskg)>>shg;\

  1100.         int r= ((((const type*)src)[i]>>shp)&maskr)>>shr;\

  1101. \

  1102.         dstU[i]= ((RU)*r + (GU)*g + (BU)*b + (257<<((S)-1)))>>(S);\

  1103.         dstV[i]= ((RV)*r + (GV)*g + (BV)*b + (257<<((S)-1)))>>(S);\

  1104.     }\

  1105. }\

  1106. static inline void name ## _half(uint8_t *dstU, uint8_t *dstV, const uint8_t *src, const uint8_t *dummy, long width, uint32_t *unused)\

  1107. {\

  1108.     int i;\

  1109.     for (i=0; i<width; i++) {\

  1110.         int pix0= ((const type*)src)[2*i+0]>>shp;\

  1111.         int pix1= ((const type*)src)[2*i+1]>>shp;\

  1112.         int g= (pix0&~(maskr|maskb))+(pix1&~(maskr|maskb));\

  1113.         int b= ((pix0+pix1-g)&(maskb|(2*maskb)))>>shb;\

  1114.         int r= ((pix0+pix1-g)&(maskr|(2*maskr)))>>shr;\

  1115.         g&= maskg|(2*maskg);\

  1116. \

  1117.         g>>=shg;\

  1118. \

  1119.         dstU[i]= ((RU)*r + (GU)*g + (BU)*b + (257<<(S)))>>((S)+1);\

  1120.         dstV[i]= ((RV)*r + (GV)*g + (BV)*b + (257<<(S)))>>((S)+1);\

  1121.     }\

  1122. }

  1123. 

  1124. BGR2UV(uint32_t, bgr32ToUV,16, 0, 0, 0, 0xFF0000, 0xFF00,   0x00FF, RU<< 8, GU   , BU<< 8, RV<< 8, GV   , BV<< 8, RGB2YUV_SHIFT+8)

  1125. BGR2UV(uint32_t,bgr321ToUV,16, 0, 0, 8, 0xFF0000, 0xFF00,   0x00FF, RU<< 8, GU   , BU<< 8, RV<< 8, GV   , BV<< 8, RGB2YUV_SHIFT+8)

  1126. BGR2UV(uint32_t, rgb32ToUV, 0, 0,16, 0,   0x00FF, 0xFF00, 0xFF0000, RU<< 8, GU   , BU<< 8, RV<< 8, GV   , BV<< 8, RGB2YUV_SHIFT+8)

  1127. BGR2UV(uint32_t,rgb321ToUV, 0, 0,16, 8,   0x00FF, 0xFF00, 0xFF0000, RU<< 8, GU   , BU<< 8, RV<< 8, GV   , BV<< 8, RGB2YUV_SHIFT+8)

  1128. BGR2UV(uint16_t, bgr16ToUV, 0, 0, 0, 0,   0x001F, 0x07E0,   0xF800, RU<<11, GU<<5, BU    , RV<<11, GV<<5, BV    , RGB2YUV_SHIFT+8)

  1129. BGR2UV(uint16_t, bgr15ToUV, 0, 0, 0, 0,   0x001F, 0x03E0,   0x7C00, RU<<10, GU<<5, BU    , RV<<10, GV<<5, BV    , RGB2YUV_SHIFT+7)

  1130. BGR2UV(uint16_t, rgb16ToUV, 0, 0, 0, 0,   0xF800, 0x07E0,   0x001F, RU    , GU<<5, BU<<11, RV    , GV<<5, BV<<11, RGB2YUV_SHIFT+8)

  1131. BGR2UV(uint16_t, rgb15ToUV, 0, 0, 0, 0,   0x7C00, 0x03E0,   0x001F, RU    , GU<<5, BU<<10, RV    , GV<<5, BV<<10, RGB2YUV_SHIFT+7)

  1132. 

  1133. static inline void palToY(uint8_t *dst, const uint8_t *src, long width, uint32_t *pal)

  1134. {

  1135.     int i;

  1136.     for (i=0; i<width; i++) {

  1137.         int d= src[i];

  1138. 

  1139.         dst[i]= pal[d] & 0xFF;

  1140.     }

  1141. }

  1142. 

  1143. static inline void palToUV(uint8_t *dstU, uint8_t *dstV,

  1144.                            const uint8_t *src1, const uint8_t *src2,

  1145.                            long width, uint32_t *pal)

  1146. {

  1147.     int i;

  1148.     assert(src1 == src2);

  1149.     for (i=0; i<width; i++) {

  1150.         int p= pal[src1[i]];

  1151. 

  1152.         dstU[i]= p>>8;

  1153.         dstV[i]= p>>16;

  1154.     }

  1155. }

  1156. 

  1157. static inline void monowhite2Y(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)

  1158. {

  1159.     int i, j;

  1160.     for (i=0; i<width/8; i++) {

  1161.         int d= ~src[i];

  1162.         for(j=0; j<8; j++)

  1163.             dst[8*i+j]= ((d>>(7-j))&1)*255;

  1164.     }

  1165. }

  1166. 

  1167. static inline void monoblack2Y(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)

  1168. {

  1169.     int i, j;

  1170.     for (i=0; i<width/8; i++) {

  1171.         int d= src[i];

  1172.         for(j=0; j<8; j++)

  1173.             dst[8*i+j]= ((d>>(7-j))&1)*255;

  1174.     }

  1175. }

  1176. 

  1177. //Note: we have C, MMX, MMX2, 3DNOW versions, there is no 3DNOW+MMX2 one

  1178. //Plain C versions

  1179. 

  1180. #define COMPILE_TEMPLATE_MMX2 0

  1181. #define COMPILE_TEMPLATE_AMD3DNOW 0

  1182. #define COMPILE_TEMPLATE_ALTIVEC 0

  1183. 

  1184. #include "swscale_template.c"

  1185. 

  1186. #if HAVE_ALTIVEC

  1187. #undef RENAME

  1188. #undef COMPILE_TEMPLATE_ALTIVEC

  1189. #define COMPILE_TEMPLATE_ALTIVEC 1

  1190. #define RENAME(a) a ## _altivec

  1191. #include "ppc/swscale_template.c"

  1192. #endif

  1193. 

  1194. //MMX versions

  1195. #if HAVE_MMX

  1196. #undef RENAME

  1197. #undef COMPILE_TEMPLATE_MMX2

  1198. #undef COMPILE_TEMPLATE_AMD3DNOW

  1199. #define COMPILE_TEMPLATE_MMX2 0

  1200. #define COMPILE_TEMPLATE_AMD3DNOW 0

  1201. #define RENAME(a) a ## _MMX

  1202. #include "x86/swscale_template.c"

  1203. #endif

  1204. 

  1205. //MMX2 versions

  1206. #if HAVE_MMX2

  1207. #undef RENAME

  1208. #undef COMPILE_TEMPLATE_MMX2

  1209. #undef COMPILE_TEMPLATE_AMD3DNOW

  1210. #define COMPILE_TEMPLATE_MMX2 1

  1211. #define COMPILE_TEMPLATE_AMD3DNOW 0

  1212. #define RENAME(a) a ## _MMX2

  1213. #include "x86/swscale_template.c"

  1214. #endif

  1215. 

  1216. //3DNOW versions

  1217. #if HAVE_AMD3DNOW

  1218. #undef RENAME

  1219. #undef COMPILE_TEMPLATE_MMX2

  1220. #undef COMPILE_TEMPLATE_AMD3DNOW

  1221. #define COMPILE_TEMPLATE_MMX2 0

  1222. #define COMPILE_TEMPLATE_AMD3DNOW 1

  1223. #define RENAME(a) a ## _3DNow

  1224. #include "x86/swscale_template.c"

  1225. #endif

  1226. 

  1227. SwsFunc ff_getSwsFunc(SwsContext *c)

  1228. {

  1229.     int cpu_flags = av_get_cpu_flags();

  1230. 

  1231.     sws_init_swScale_c(c);

  1232. 

  1233. #if HAVE_MMX2

  1234.     // ordered per speed fastest first

  1235.     if (cpu_flags & AV_CPU_FLAG_MMX2) {

  1236.         sws_init_swScale_MMX2(c);

  1237.         return swScale_MMX2;

  1238.     } else

  1239. #endif

  1240. #if HAVE_AMD3DNOW

  1241.     if (cpu_flags & AV_CPU_FLAG_3DNOW) {

  1242.         sws_init_swScale_3DNow(c);

  1243.         return swScale_3DNow;

  1244.     } else

  1245. #endif

  1246. #if HAVE_MMX

  1247.     if (cpu_flags & AV_CPU_FLAG_MMX) {

  1248.         sws_init_swScale_MMX(c);

  1249.         return swScale_MMX;

  1250.     } else

  1251. #endif

  1252. #if HAVE_ALTIVEC

  1253.     if (cpu_flags & AV_CPU_FLAG_ALTIVEC) {

  1254.         sws_init_swScale_altivec(c);

  1255.         return swScale_altivec;

  1256.     } else

  1257. #endif

  1258. 

  1259.     return swScale_c;

  1260. }

  1261. 

  1262. static void copyPlane(const uint8_t *src, int srcStride,

  1263.                       int srcSliceY, int srcSliceH, int width,

  1264.                       uint8_t *dst, int dstStride)

  1265. {

  1266.     dst += dstStride * srcSliceY;

  1267.     if (dstStride == srcStride && srcStride > 0) {

  1268.         memcpy(dst, src, srcSliceH * dstStride);

  1269.     } else {

  1270.         int i;

  1271.         for (i=0; i<srcSliceH; i++) {

  1272.             memcpy(dst, src, width);

  1273.             src += srcStride;

  1274.             dst += dstStride;

  1275.         }

  1276.     }

  1277. }

  1278. 

  1279. static int planarToNv12Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,

  1280.                                int srcSliceH, uint8_t* dstParam[], int dstStride[])

  1281. {

  1282.     uint8_t *dst = dstParam[1] + dstStride[1]*srcSliceY/2;

  1283. 

  1284.     copyPlane(src[0], srcStride[0], srcSliceY, srcSliceH, c->srcW,

  1285.               dstParam[0], dstStride[0]);

  1286. 

  1287.     if (c->dstFormat == PIX_FMT_NV12)

  1288.         interleaveBytes(src[1], src[2], dst, c->srcW/2, srcSliceH/2, srcStride[1], srcStride[2], dstStride[0]);

  1289.     else

  1290.         interleaveBytes(src[2], src[1], dst, c->srcW/2, srcSliceH/2, srcStride[2], srcStride[1], dstStride[0]);

  1291. 

  1292.     return srcSliceH;

  1293. }

  1294. 

  1295. static int planarToYuy2Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,

  1296.                                int srcSliceH, uint8_t* dstParam[], int dstStride[])

  1297. {

  1298.     uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;

  1299. 

  1300.     yv12toyuy2(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);

  1301. 

  1302.     return srcSliceH;

  1303. }

  1304. 

  1305. static int planarToUyvyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,

  1306.                                int srcSliceH, uint8_t* dstParam[], int dstStride[])

  1307. {

  1308.     uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;

  1309. 

  1310.     yv12touyvy(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);

  1311. 

  1312.     return srcSliceH;

  1313. }

  1314. 

  1315. static int yuv422pToYuy2Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,

  1316.                                 int srcSliceH, uint8_t* dstParam[], int dstStride[])

  1317. {

  1318.     uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;

  1319. 

  1320.     yuv422ptoyuy2(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);

  1321. 

  1322.     return srcSliceH;

  1323. }

  1324. 

  1325. static int yuv422pToUyvyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,

  1326.                                 int srcSliceH, uint8_t* dstParam[], int dstStride[])

  1327. {

  1328.     uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;

  1329. 

  1330.     yuv422ptouyvy(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);

  1331. 

  1332.     return srcSliceH;

  1333. }

  1334. 

  1335. static int yuyvToYuv420Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,

  1336.                                int srcSliceH, uint8_t* dstParam[], int dstStride[])

  1337. {

  1338.     uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;

  1339.     uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY/2;

  1340.     uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY/2;

  1341. 

  1342.     yuyvtoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);

  1343. 

  1344.     if (dstParam[3])

  1345.         fillPlane(dstParam[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);

  1346. 

  1347.     return srcSliceH;

  1348. }

  1349. 

  1350. static int yuyvToYuv422Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,

  1351.                                int srcSliceH, uint8_t* dstParam[], int dstStride[])

  1352. {

  1353.     uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;

  1354.     uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY;

  1355.     uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY;

  1356. 

  1357.     yuyvtoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);

  1358. 

  1359.     return srcSliceH;

  1360. }

  1361. 

  1362. static int uyvyToYuv420Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,

  1363.                                int srcSliceH, uint8_t* dstParam[], int dstStride[])

  1364. {

  1365.     uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;

  1366.     uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY/2;

  1367.     uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY/2;

  1368. 

  1369.     uyvytoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);

  1370. 

  1371.     if (dstParam[3])

  1372.         fillPlane(dstParam[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);

  1373. 

  1374.     return srcSliceH;

  1375. }

  1376. 

  1377. static int uyvyToYuv422Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,

  1378.                                int srcSliceH, uint8_t* dstParam[], int dstStride[])

  1379. {

  1380.     uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;

  1381.     uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY;

  1382.     uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY;

  1383. 

  1384.     uyvytoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);

  1385. 

  1386.     return srcSliceH;

  1387. }

  1388. 

  1389. static void gray8aToPacked32(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)

  1390. {

  1391.     long i;

  1392.     for (i=0; i<num_pixels; i++)

  1393.         ((uint32_t *) dst)[i] = ((const uint32_t *)palette)[src[i<<1]] | (src[(i<<1)+1] << 24);

  1394. }

  1395. 

  1396. static void gray8aToPacked32_1(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)

  1397. {

  1398.     long i;

  1399. 

  1400.     for (i=0; i<num_pixels; i++)

  1401.         ((uint32_t *) dst)[i] = ((const uint32_t *)palette)[src[i<<1]] | src[(i<<1)+1];

  1402. }

  1403. 

  1404. static void gray8aToPacked24(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)

  1405. {

  1406.     long i;

  1407. 

  1408.     for (i=0; i<num_pixels; i++) {

  1409.         //FIXME slow?

  1410.         dst[0]= palette[src[i<<1]*4+0];

  1411.         dst[1]= palette[src[i<<1]*4+1];

  1412.         dst[2]= palette[src[i<<1]*4+2];

  1413.         dst+= 3;

  1414.     }

  1415. }

  1416. 

  1417. static int palToRgbWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,

  1418.                            int srcSliceH, uint8_t* dst[], int dstStride[])

  1419. {

  1420.     const enum PixelFormat srcFormat= c->srcFormat;

  1421.     const enum PixelFormat dstFormat= c->dstFormat;

  1422.     void (*conv)(const uint8_t *src, uint8_t *dst, long num_pixels,

  1423.                  const uint8_t *palette)=NULL;

  1424.     int i;

  1425.     uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;

  1426.     const uint8_t *srcPtr= src[0];

  1427. 

  1428.     if (srcFormat == PIX_FMT_Y400A) {

  1429.         switch (dstFormat) {

  1430.         case PIX_FMT_RGB32  : conv = gray8aToPacked32; break;

  1431.         case PIX_FMT_BGR32  : conv = gray8aToPacked32; break;

  1432.         case PIX_FMT_BGR32_1: conv = gray8aToPacked32_1; break;

  1433.         case PIX_FMT_RGB32_1: conv = gray8aToPacked32_1; break;

  1434.         case PIX_FMT_RGB24  : conv = gray8aToPacked24; break;

  1435.         case PIX_FMT_BGR24  : conv = gray8aToPacked24; break;

  1436.         }

  1437.     } else if (usePal(srcFormat)) {

  1438.         switch (dstFormat) {

  1439.         case PIX_FMT_RGB32  : conv = sws_convertPalette8ToPacked32; break;

  1440.         case PIX_FMT_BGR32  : conv = sws_convertPalette8ToPacked32; break;

  1441.         case PIX_FMT_BGR32_1: conv = sws_convertPalette8ToPacked32; break;

  1442.         case PIX_FMT_RGB32_1: conv = sws_convertPalette8ToPacked32; break;

  1443.         case PIX_FMT_RGB24  : conv = sws_convertPalette8ToPacked24; break;

  1444.         case PIX_FMT_BGR24  : conv = sws_convertPalette8ToPacked24; break;

  1445.         }

  1446.     }

  1447. 

  1448.     if (!conv)

  1449.         av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",

  1450.                sws_format_name(srcFormat), sws_format_name(dstFormat));

  1451.     else {

  1452.         for (i=0; i<srcSliceH; i++) {

  1453.             conv(srcPtr, dstPtr, c->srcW, (uint8_t *) c->pal_rgb);

  1454.             srcPtr+= srcStride[0];

  1455.             dstPtr+= dstStride[0];

  1456.         }

  1457.     }

  1458. 

  1459.     return srcSliceH;

  1460. }

  1461. 

  1462. #define isRGBA32(x) (            \

  1463.            (x) == PIX_FMT_ARGB   \

  1464.         || (x) == PIX_FMT_RGBA   \

  1465.         || (x) == PIX_FMT_BGRA   \

  1466.         || (x) == PIX_FMT_ABGR   \

  1467.         )

  1468. 

  1469. /* {RGB,BGR}{15,16,24,32,32_1} -> {RGB,BGR}{15,16,24,32} */

  1470. static int rgbToRgbWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,

  1471.                            int srcSliceH, uint8_t* dst[], int dstStride[])

  1472. {

  1473.     const enum PixelFormat srcFormat= c->srcFormat;

  1474.     const enum PixelFormat dstFormat= c->dstFormat;

  1475.     const int srcBpp= (c->srcFormatBpp + 7) >> 3;

  1476.     const int dstBpp= (c->dstFormatBpp + 7) >> 3;

  1477.     const int srcId= c->srcFormatBpp >> 2; /* 1:0, 4:1, 8:2, 15:3, 16:4, 24:6, 32:8 */

  1478.     const int dstId= c->dstFormatBpp >> 2;

  1479.     void (*conv)(const uint8_t *src, uint8_t *dst, long src_size)=NULL;

  1480. 

  1481. #define CONV_IS(src, dst) (srcFormat == PIX_FMT_##src && dstFormat == PIX_FMT_##dst)

  1482. 

  1483.     if (isRGBA32(srcFormat) && isRGBA32(dstFormat)) {

  1484.         if (     CONV_IS(ABGR, RGBA)

  1485.               || CONV_IS(ARGB, BGRA)

  1486.               || CONV_IS(BGRA, ARGB)

  1487.               || CONV_IS(RGBA, ABGR)) conv = shuffle_bytes_3210;

  1488.         else if (CONV_IS(ABGR, ARGB)

  1489.               || CONV_IS(ARGB, ABGR)) conv = shuffle_bytes_0321;

  1490.         else if (CONV_IS(ABGR, BGRA)

  1491.               || CONV_IS(ARGB, RGBA)) conv = shuffle_bytes_1230;

  1492.         else if (CONV_IS(BGRA, RGBA)

  1493.               || CONV_IS(RGBA, BGRA)) conv = shuffle_bytes_2103;

  1494.         else if (CONV_IS(BGRA, ABGR)

  1495.               || CONV_IS(RGBA, ARGB)) conv = shuffle_bytes_3012;

  1496.     } else

  1497.     /* BGR -> BGR */

  1498.     if (  (isBGRinInt(srcFormat) && isBGRinInt(dstFormat))

  1499.        || (isRGBinInt(srcFormat) && isRGBinInt(dstFormat))) {

  1500.         switch(srcId | (dstId<<4)) {

  1501.         case 0x34: conv= rgb16to15; break;

  1502.         case 0x36: conv= rgb24to15; break;

  1503.         case 0x38: conv= rgb32to15; break;

  1504.         case 0x43: conv= rgb15to16; break;

  1505.         case 0x46: conv= rgb24to16; break;

  1506.         case 0x48: conv= rgb32to16; break;

  1507.         case 0x63: conv= rgb15to24; break;

  1508.         case 0x64: conv= rgb16to24; break;

  1509.         case 0x68: conv= rgb32to24; break;

  1510.         case 0x83: conv= rgb15to32; break;

  1511.         case 0x84: conv= rgb16to32; break;

  1512.         case 0x86: conv= rgb24to32; break;

  1513.         }

  1514.     } else if (  (isBGRinInt(srcFormat) && isRGBinInt(dstFormat))

  1515.              || (isRGBinInt(srcFormat) && isBGRinInt(dstFormat))) {

  1516.         switch(srcId | (dstId<<4)) {

  1517.         case 0x33: conv= rgb15tobgr15; break;

  1518.         case 0x34: conv= rgb16tobgr15; break;

  1519.         case 0x36: conv= rgb24tobgr15; break;

  1520.         case 0x38: conv= rgb32tobgr15; break;

  1521.         case 0x43: conv= rgb15tobgr16; break;

  1522.         case 0x44: conv= rgb16tobgr16; break;

  1523.         case 0x46: conv= rgb24tobgr16; break;

  1524.         case 0x48: conv= rgb32tobgr16; break;

  1525.         case 0x63: conv= rgb15tobgr24; break;

  1526.         case 0x64: conv= rgb16tobgr24; break;

  1527.         case 0x66: conv= rgb24tobgr24; break;

  1528.         case 0x68: conv= rgb32tobgr24; break;

  1529.         case 0x83: conv= rgb15tobgr32; break;

  1530.         case 0x84: conv= rgb16tobgr32; break;

  1531.         case 0x86: conv= rgb24tobgr32; break;

  1532.         }

  1533.     }

  1534. 

  1535.     if (!conv) {

  1536.         av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",

  1537.                sws_format_name(srcFormat), sws_format_name(dstFormat));

  1538.     } else {

  1539.         const uint8_t *srcPtr= src[0];

  1540.               uint8_t *dstPtr= dst[0];

  1541.         if ((srcFormat == PIX_FMT_RGB32_1 || srcFormat == PIX_FMT_BGR32_1) && !isRGBA32(dstFormat))

  1542.             srcPtr += ALT32_CORR;

  1543. 

  1544.         if ((dstFormat == PIX_FMT_RGB32_1 || dstFormat == PIX_FMT_BGR32_1) && !isRGBA32(srcFormat))

  1545.             dstPtr += ALT32_CORR;

  1546. 

  1547.         if (dstStride[0]*srcBpp == srcStride[0]*dstBpp && srcStride[0] > 0)

  1548.             conv(srcPtr, dstPtr + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]);

  1549.         else {

  1550.             int i;

  1551.             dstPtr += dstStride[0]*srcSliceY;

  1552. 

  1553.             for (i=0; i<srcSliceH; i++) {

  1554.                 conv(srcPtr, dstPtr, c->srcW*srcBpp);

  1555.                 srcPtr+= srcStride[0];

  1556.                 dstPtr+= dstStride[0];

  1557.             }

  1558.         }

  1559.     }

  1560.     return srcSliceH;

  1561. }

  1562. 

  1563. static int bgr24ToYv12Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,

  1564.                               int srcSliceH, uint8_t* dst[], int dstStride[])

  1565. {

  1566.     rgb24toyv12(

  1567.         src[0],

  1568.         dst[0]+ srcSliceY    *dstStride[0],

  1569.         dst[1]+(srcSliceY>>1)*dstStride[1],

  1570.         dst[2]+(srcSliceY>>1)*dstStride[2],

  1571.         c->srcW, srcSliceH,

  1572.         dstStride[0], dstStride[1], srcStride[0]);

  1573.     if (dst[3])

  1574.         fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);

  1575.     return srcSliceH;

  1576. }

  1577. 

  1578. static int yvu9ToYv12Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,

  1579.                              int srcSliceH, uint8_t* dst[], int dstStride[])

  1580. {

  1581.     copyPlane(src[0], srcStride[0], srcSliceY, srcSliceH, c->srcW,

  1582.               dst[0], dstStride[0]);

  1583. 

  1584.     planar2x(src[1], dst[1] + dstStride[1]*(srcSliceY >> 1), c->chrSrcW,

  1585.              srcSliceH >> 2, srcStride[1], dstStride[1]);

  1586.     planar2x(src[2], dst[2] + dstStride[2]*(srcSliceY >> 1), c->chrSrcW,

  1587.              srcSliceH >> 2, srcStride[2], dstStride[2]);

  1588.     if (dst[3])

  1589.         fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);

  1590.     return srcSliceH;

  1591. }

  1592. 

  1593. /* unscaled copy like stuff (assumes nearly identical formats) */

  1594. static int packedCopyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,

  1595.                              int srcSliceH, uint8_t* dst[], int dstStride[])

  1596. {

  1597.     if (dstStride[0]==srcStride[0] && srcStride[0] > 0)

  1598.         memcpy(dst[0] + dstStride[0]*srcSliceY, src[0], srcSliceH*dstStride[0]);

  1599.     else {

  1600.         int i;

  1601.         const uint8_t *srcPtr= src[0];

  1602.         uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;

  1603.         int length=0;

  1604. 

  1605.         /* universal length finder */

  1606.         while(length+c->srcW <= FFABS(dstStride[0])

  1607.            && length+c->srcW <= FFABS(srcStride[0])) length+= c->srcW;

  1608.         assert(length!=0);

  1609. 

  1610.         for (i=0; i<srcSliceH; i++) {

  1611.             memcpy(dstPtr, srcPtr, length);

  1612.             srcPtr+= srcStride[0];

  1613.             dstPtr+= dstStride[0];

  1614.         }

  1615.     }

  1616.     return srcSliceH;

  1617. }

  1618. 

  1619. static int planarCopyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,

  1620.                              int srcSliceH, uint8_t* dst[], int dstStride[])

  1621. {

  1622.     int plane, i, j;

  1623.     for (plane=0; plane<4; plane++) {

  1624.         int length= (plane==0 || plane==3) ? c->srcW  : -((-c->srcW  )>>c->chrDstHSubSample);

  1625.         int y=      (plane==0 || plane==3) ? srcSliceY: -((-srcSliceY)>>c->chrDstVSubSample);

  1626.         int height= (plane==0 || plane==3) ? srcSliceH: -((-srcSliceH)>>c->chrDstVSubSample);

  1627.         const uint8_t *srcPtr= src[plane];

  1628.         uint8_t *dstPtr= dst[plane] + dstStride[plane]*y;

  1629. 

  1630.         if (!dst[plane]) continue;

  1631.         // ignore palette for GRAY8

  1632.         if (plane == 1 && !dst[2]) continue;

  1633.         if (!src[plane] || (plane == 1 && !src[2])) {

  1634.             if(is16BPS(c->dstFormat))

  1635.                 length*=2;

  1636.             fillPlane(dst[plane], dstStride[plane], length, height, y, (plane==3) ? 255 : 128);

  1637.         } else {

  1638.             if(is9_OR_10BPS(c->srcFormat)) {

  1639.                 const int src_depth = av_pix_fmt_descriptors[c->srcFormat].comp[plane].depth_minus1+1;

  1640.                 const int dst_depth = av_pix_fmt_descriptors[c->dstFormat].comp[plane].depth_minus1+1;

  1641.                 const uint16_t *srcPtr2 = (const uint16_t*)srcPtr;

  1642. 

  1643.                 if (is16BPS(c->dstFormat)) {

  1644.                     uint16_t *dstPtr2 = (uint16_t*)dstPtr;

  1645. #define COPY9_OR_10TO16(rfunc, wfunc) \

  1646.                     for (i = 0; i < height; i++) { \

  1647.                         for (j = 0; j < length; j++) { \

  1648.                             int srcpx = rfunc(&srcPtr2[j]); \

  1649.                             wfunc(&dstPtr2[j], (srcpx<<(16-src_depth)) | (srcpx>>(2*src_depth-16))); \

  1650.                         } \

  1651.                         dstPtr2 += dstStride[plane]/2; \

  1652.                         srcPtr2 += srcStride[plane]/2; \

  1653.                     }

  1654.                     if (isBE(c->dstFormat)) {

  1655.                         if (isBE(c->srcFormat)) {

  1656.                             COPY9_OR_10TO16(AV_RB16, AV_WB16);

  1657.                         } else {

  1658.                             COPY9_OR_10TO16(AV_RL16, AV_WB16);

  1659.                         }

  1660.                     } else {

  1661.                         if (isBE(c->srcFormat)) {

  1662.                             COPY9_OR_10TO16(AV_RB16, AV_WL16);

  1663.                         } else {

  1664.                             COPY9_OR_10TO16(AV_RL16, AV_WL16);

  1665.                         }

  1666.                     }

  1667.                 } else if (is9_OR_10BPS(c->dstFormat)) {

  1668.                     uint16_t *dstPtr2 = (uint16_t*)dstPtr;

  1669. #define COPY9_OR_10TO9_OR_10(loop) \

  1670.                     for (i = 0; i < height; i++) { \

  1671.                         for (j = 0; j < length; j++) { \

  1672.                             loop; \

  1673.                         } \

  1674.                         dstPtr2 += dstStride[plane]/2; \

  1675.                         srcPtr2 += srcStride[plane]/2; \

  1676.                     }

  1677. #define COPY9_OR_10TO9_OR_10_2(rfunc, wfunc) \

  1678.                     if (dst_depth > src_depth) { \

  1679.                         COPY9_OR_10TO9_OR_10(int srcpx = rfunc(&srcPtr2[j]); \

  1680.                             wfunc(&dstPtr2[j], (srcpx << 1) | (srcpx >> 9))); \

  1681.                     } else if (dst_depth < src_depth) { \

  1682.                         COPY9_OR_10TO9_OR_10(wfunc(&dstPtr2[j], rfunc(&srcPtr2[j]) >> 1)); \

  1683.                     } else { \

  1684.                         COPY9_OR_10TO9_OR_10(wfunc(&dstPtr2[j], rfunc(&srcPtr2[j]))); \

  1685.                     }

  1686.                     if (isBE(c->dstFormat)) {

  1687.                         if (isBE(c->srcFormat)) {

  1688.                             COPY9_OR_10TO9_OR_10_2(AV_RB16, AV_WB16);

  1689.                         } else {

  1690.                             COPY9_OR_10TO9_OR_10_2(AV_RL16, AV_WB16);

  1691.                         }

  1692.                     } else {

  1693.                         if (isBE(c->srcFormat)) {

  1694.                             COPY9_OR_10TO9_OR_10_2(AV_RB16, AV_WL16);

  1695.                         } else {

  1696.                             COPY9_OR_10TO9_OR_10_2(AV_RL16, AV_WL16);

  1697.                         }

  1698.                     }

  1699.                 } else {

  1700.                     // FIXME Maybe dither instead.

  1701. #define COPY9_OR_10TO8(rfunc) \

  1702.                     for (i = 0; i < height; i++) { \

  1703.                         for (j = 0; j < length; j++) { \

  1704.                             dstPtr[j] = rfunc(&srcPtr2[j])>>(src_depth-8); \

  1705.                         } \

  1706.                         dstPtr  += dstStride[plane]; \

  1707.                         srcPtr2 += srcStride[plane]/2; \

  1708.                     }

  1709.                     if (isBE(c->srcFormat)) {

  1710.                         COPY9_OR_10TO8(AV_RB16);

  1711.                     } else {

  1712.                         COPY9_OR_10TO8(AV_RL16);

  1713.                     }

  1714.                 }

  1715.             } else if(is9_OR_10BPS(c->dstFormat)) {

  1716.                 const int dst_depth = av_pix_fmt_descriptors[c->dstFormat].comp[plane].depth_minus1+1;

  1717.                 uint16_t *dstPtr2 = (uint16_t*)dstPtr;

  1718. 

  1719.                 if (is16BPS(c->srcFormat)) {

  1720.                     const uint16_t *srcPtr2 = (const uint16_t*)srcPtr;

  1721. #define COPY16TO9_OR_10(rfunc, wfunc) \

  1722.                     for (i = 0; i < height; i++) { \

  1723.                         for (j = 0; j < length; j++) { \

  1724.                             wfunc(&dstPtr2[j], rfunc(&srcPtr2[j])>>(16-dst_depth)); \

  1725.                         } \

  1726.                         dstPtr2 += dstStride[plane]/2; \

  1727.                         srcPtr2 += srcStride[plane]/2; \

  1728.                     }

  1729.                     if (isBE(c->dstFormat)) {

  1730.                         if (isBE(c->srcFormat)) {

  1731.                             COPY16TO9_OR_10(AV_RB16, AV_WB16);

  1732.                         } else {

  1733.                             COPY16TO9_OR_10(AV_RL16, AV_WB16);

  1734.                         }

  1735.                     } else {

  1736.                         if (isBE(c->srcFormat)) {

  1737.                             COPY16TO9_OR_10(AV_RB16, AV_WL16);

  1738.                         } else {

  1739.                             COPY16TO9_OR_10(AV_RL16, AV_WL16);

  1740.                         }

  1741.                     }

  1742.                 } else /* 8bit */ {

  1743. #define COPY8TO9_OR_10(wfunc) \

  1744.                     for (i = 0; i < height; i++) { \

  1745.                         for (j = 0; j < length; j++) { \

  1746.                             const int srcpx = srcPtr[j]; \

  1747.                             wfunc(&dstPtr2[j], (srcpx<<(dst_depth-8)) | (srcpx >> (16-dst_depth))); \

  1748.                         } \

  1749.                         dstPtr2 += dstStride[plane]/2; \

  1750.                         srcPtr  += srcStride[plane]; \

  1751.                     }

  1752.                     if (isBE(c->dstFormat)) {

  1753.                         COPY8TO9_OR_10(AV_WB16);

  1754.                     } else {

  1755.                         COPY8TO9_OR_10(AV_WL16);

  1756.                     }

  1757.                 }

  1758.             } else if(is16BPS(c->srcFormat) && !is16BPS(c->dstFormat)) {

  1759.                 if (!isBE(c->srcFormat)) srcPtr++;

  1760.                 for (i=0; i<height; i++) {

  1761.                     for (j=0; j<length; j++) dstPtr[j] = srcPtr[j<<1];

  1762.                     srcPtr+= srcStride[plane];

  1763.                     dstPtr+= dstStride[plane];

  1764.                 }

  1765.             } else if(!is16BPS(c->srcFormat) && is16BPS(c->dstFormat)) {

  1766.                 for (i=0; i<height; i++) {

  1767.                     for (j=0; j<length; j++) {

  1768.                         dstPtr[ j<<1   ] = srcPtr[j];

  1769.                         dstPtr[(j<<1)+1] = srcPtr[j];

  1770.                     }

  1771.                     srcPtr+= srcStride[plane];

  1772.                     dstPtr+= dstStride[plane];

  1773.                 }

  1774.             } else if(is16BPS(c->srcFormat) && is16BPS(c->dstFormat)

  1775.                   && isBE(c->srcFormat) != isBE(c->dstFormat)) {

  1776. 

  1777.                 for (i=0; i<height; i++) {

  1778.                     for (j=0; j<length; j++)

  1779.                         ((uint16_t*)dstPtr)[j] = av_bswap16(((const uint16_t*)srcPtr)[j]);

  1780.                     srcPtr+= srcStride[plane];

  1781.                     dstPtr+= dstStride[plane];

  1782.                 }

  1783.             } else if (dstStride[plane] == srcStride[plane] &&

  1784.                        srcStride[plane] > 0 && srcStride[plane] == length) {

  1785.                 memcpy(dst[plane] + dstStride[plane]*y, src[plane],

  1786.                        height*dstStride[plane]);

  1787.             } else {

  1788.                 if(is16BPS(c->srcFormat) && is16BPS(c->dstFormat))

  1789.                     length*=2;

  1790.                 for (i=0; i<height; i++) {

  1791.                     memcpy(dstPtr, srcPtr, length);

  1792.                     srcPtr+= srcStride[plane];

  1793.                     dstPtr+= dstStride[plane];

  1794.                 }

  1795.             }

  1796.         }

  1797.     }

  1798.     return srcSliceH;

  1799. }

  1800. 

  1801. void ff_get_unscaled_swscale(SwsContext *c)

  1802. {

  1803.     const enum PixelFormat srcFormat = c->srcFormat;

  1804.     const enum PixelFormat dstFormat = c->dstFormat;

  1805.     const int flags = c->flags;

  1806.     const int dstH = c->dstH;

  1807.     int needsDither;

  1808. 

  1809.     needsDither= isAnyRGB(dstFormat)

  1810.         &&  c->dstFormatBpp < 24

  1811.         && (c->dstFormatBpp < c->srcFormatBpp || (!isAnyRGB(srcFormat)));

  1812. 

  1813.     /* yv12_to_nv12 */

  1814.     if ((srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUVA420P) && (dstFormat == PIX_FMT_NV12 || dstFormat == PIX_FMT_NV21)) {

  1815.         c->swScale= planarToNv12Wrapper;

  1816.     }

  1817.     /* yuv2bgr */

  1818.     if ((srcFormat==PIX_FMT_YUV420P || srcFormat==PIX_FMT_YUV422P || srcFormat==PIX_FMT_YUVA420P) && isAnyRGB(dstFormat)

  1819.         && !(flags & SWS_ACCURATE_RND) && !(dstH&1)) {

  1820.         c->swScale= ff_yuv2rgb_get_func_ptr(c);

  1821.     }

  1822. 

  1823.     if (srcFormat==PIX_FMT_YUV410P && (dstFormat==PIX_FMT_YUV420P || dstFormat==PIX_FMT_YUVA420P) && !(flags & SWS_BITEXACT)) {

  1824.         c->swScale= yvu9ToYv12Wrapper;

  1825.     }

  1826. 

  1827.     /* bgr24toYV12 */

  1828.     if (srcFormat==PIX_FMT_BGR24 && (dstFormat==PIX_FMT_YUV420P || dstFormat==PIX_FMT_YUVA420P) && !(flags & SWS_ACCURATE_RND))

  1829.         c->swScale= bgr24ToYv12Wrapper;

  1830. 

  1831.     /* RGB/BGR -> RGB/BGR (no dither needed forms) */

  1832.     if (   isAnyRGB(srcFormat)

  1833.         && isAnyRGB(dstFormat)

  1834.         && srcFormat != PIX_FMT_BGR8      && dstFormat != PIX_FMT_BGR8

  1835.         && srcFormat != PIX_FMT_RGB8      && dstFormat != PIX_FMT_RGB8

  1836.         && srcFormat != PIX_FMT_BGR4      && dstFormat != PIX_FMT_BGR4

  1837.         && srcFormat != PIX_FMT_RGB4      && dstFormat != PIX_FMT_RGB4

  1838.         && srcFormat != PIX_FMT_BGR4_BYTE && dstFormat != PIX_FMT_BGR4_BYTE

  1839.         && srcFormat != PIX_FMT_RGB4_BYTE && dstFormat != PIX_FMT_RGB4_BYTE

  1840.         && srcFormat != PIX_FMT_MONOBLACK && dstFormat != PIX_FMT_MONOBLACK

  1841.         && srcFormat != PIX_FMT_MONOWHITE && dstFormat != PIX_FMT_MONOWHITE

  1842.         && srcFormat != PIX_FMT_RGB48LE   && dstFormat != PIX_FMT_RGB48LE

  1843.         && srcFormat != PIX_FMT_RGB48BE   && dstFormat != PIX_FMT_RGB48BE

  1844.         && srcFormat != PIX_FMT_BGR48LE   && dstFormat != PIX_FMT_BGR48LE

  1845.         && srcFormat != PIX_FMT_BGR48BE   && dstFormat != PIX_FMT_BGR48BE

  1846.         && (!needsDither || (c->flags&(SWS_FAST_BILINEAR|SWS_POINT))))

  1847.         c->swScale= rgbToRgbWrapper;

  1848. 

  1849.     if ((usePal(srcFormat) && (

  1850.         dstFormat == PIX_FMT_RGB32   ||

  1851.         dstFormat == PIX_FMT_RGB32_1 ||

  1852.         dstFormat == PIX_FMT_RGB24   ||

  1853.         dstFormat == PIX_FMT_BGR32   ||

  1854.         dstFormat == PIX_FMT_BGR32_1 ||

  1855.         dstFormat == PIX_FMT_BGR24)))

  1856.         c->swScale= palToRgbWrapper;

  1857. 

  1858.     if (srcFormat == PIX_FMT_YUV422P) {

  1859.         if (dstFormat == PIX_FMT_YUYV422)

  1860.             c->swScale= yuv422pToYuy2Wrapper;

  1861.         else if (dstFormat == PIX_FMT_UYVY422)

  1862.             c->swScale= yuv422pToUyvyWrapper;

  1863.     }

  1864. 

  1865.     /* LQ converters if -sws 0 or -sws 4*/

  1866.     if (c->flags&(SWS_FAST_BILINEAR|SWS_POINT)) {

  1867.         /* yv12_to_yuy2 */

  1868.         if (srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUVA420P) {

  1869.             if (dstFormat == PIX_FMT_YUYV422)

  1870.                 c->swScale= planarToYuy2Wrapper;

  1871.             else if (dstFormat == PIX_FMT_UYVY422)

  1872.                 c->swScale= planarToUyvyWrapper;

  1873.         }

  1874.     }

  1875.     if(srcFormat == PIX_FMT_YUYV422 && (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P))

  1876.         c->swScale= yuyvToYuv420Wrapper;

  1877.     if(srcFormat == PIX_FMT_UYVY422 && (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P))

  1878.         c->swScale= uyvyToYuv420Wrapper;

  1879.     if(srcFormat == PIX_FMT_YUYV422 && dstFormat == PIX_FMT_YUV422P)

  1880.         c->swScale= yuyvToYuv422Wrapper;

  1881.     if(srcFormat == PIX_FMT_UYVY422 && dstFormat == PIX_FMT_YUV422P)

  1882.         c->swScale= uyvyToYuv422Wrapper;

  1883. 

  1884. #if HAVE_ALTIVEC

  1885.     if ((av_get_cpu_flags() & AV_CPU_FLAG_ALTIVEC) &&

  1886.         !(c->flags & SWS_BITEXACT) &&

  1887.         srcFormat == PIX_FMT_YUV420P) {

  1888.         // unscaled YV12 -> packed YUV, we want speed

  1889.         if (dstFormat == PIX_FMT_YUYV422)

  1890.             c->swScale= yv12toyuy2_unscaled_altivec;

  1891.         else if (dstFormat == PIX_FMT_UYVY422)

  1892.             c->swScale= yv12touyvy_unscaled_altivec;

  1893.     }

  1894. #endif

  1895. 

  1896.     /* simple copy */

  1897.     if (  srcFormat == dstFormat

  1898.         || (srcFormat == PIX_FMT_YUVA420P && dstFormat == PIX_FMT_YUV420P)

  1899.         || (srcFormat == PIX_FMT_YUV420P && dstFormat == PIX_FMT_YUVA420P)

  1900.         || (isPlanarYUV(srcFormat) && isGray(dstFormat))

  1901.         || (isPlanarYUV(dstFormat) && isGray(srcFormat))

  1902.         || (isGray(dstFormat) && isGray(srcFormat))

  1903.         || (isPlanarYUV(srcFormat) && isPlanarYUV(dstFormat)

  1904.             && c->chrDstHSubSample == c->chrSrcHSubSample

  1905.             && c->chrDstVSubSample == c->chrSrcVSubSample

  1906.             && dstFormat != PIX_FMT_NV12 && dstFormat != PIX_FMT_NV21

  1907.             && srcFormat != PIX_FMT_NV12 && srcFormat != PIX_FMT_NV21))

  1908.     {

  1909.         if (isPacked(c->srcFormat))

  1910.             c->swScale= packedCopyWrapper;

  1911.         else /* Planar YUV or gray */

  1912.             c->swScale= planarCopyWrapper;

  1913.     }

  1914. #if ARCH_BFIN

  1915.     ff_bfin_get_unscaled_swscale (c);

  1916. #endif

  1917. }

  1918. 

  1919. static void reset_ptr(const uint8_t* src[], int format)

  1920. {

  1921.     if(!isALPHA(format))

  1922.         src[3]=NULL;

  1923.     if(!isPlanarYUV(format)) {

  1924.         src[3]=src[2]=NULL;

  1925. 

  1926.         if (!usePal(format))

  1927.             src[1]= NULL;

  1928.     }

  1929. }

  1930. 

  1931. static int check_image_pointers(uint8_t *data[4], enum PixelFormat pix_fmt,

  1932.                                 const int linesizes[4])

  1933. {

  1934.     const AVPixFmtDescriptor *desc = &av_pix_fmt_descriptors[pix_fmt];

  1935.     int i;

  1936. 

  1937.     for (i = 0; i < 4; i++) {

  1938.         int plane = desc->comp[i].plane;

  1939.         if (!data[plane] || !linesizes[plane])

  1940.             return 0;

  1941.     }

  1942. 

  1943.     return 1;

  1944. }

  1945. 

  1946. /**

  1947.  * swscale wrapper, so we don't need to export the SwsContext.

  1948.  * Assumes planar YUV to be in YUV order instead of YVU.

  1949.  */

  1950. int sws_scale(SwsContext *c, const uint8_t* const src[], const int srcStride[], int srcSliceY,

  1951.               int srcSliceH, uint8_t* const dst[], const int dstStride[])

  1952. {

  1953.     int i;

  1954.     const uint8_t* src2[4]= {src[0], src[1], src[2], src[3]};

  1955.     uint8_t* dst2[4]= {dst[0], dst[1], dst[2], dst[3]};

  1956. 

  1957.     // do not mess up sliceDir if we have a "trailing" 0-size slice

  1958.     if (srcSliceH == 0)

  1959.         return 0;

  1960. 

  1961.     if (!check_image_pointers(src, c->srcFormat, srcStride)) {

  1962.         av_log(c, AV_LOG_ERROR, "bad src image pointers\n");

  1963.         return 0;

  1964.     }

  1965.     if (!check_image_pointers(dst, c->dstFormat, dstStride)) {

  1966.         av_log(c, AV_LOG_ERROR, "bad dst image pointers\n");

  1967.         return 0;

  1968.     }

  1969. 

  1970.     if (c->sliceDir == 0 && srcSliceY != 0 && srcSliceY + srcSliceH != c->srcH) {

  1971.         av_log(c, AV_LOG_ERROR, "Slices start in the middle!\n");

  1972.         return 0;

  1973.     }

  1974.     if (c->sliceDir == 0) {

  1975.         if (srcSliceY == 0) c->sliceDir = 1; else c->sliceDir = -1;

  1976.     }

  1977. 

  1978.     if (usePal(c->srcFormat)) {

  1979.         for (i=0; i<256; i++) {

  1980.             int p, r, g, b,y,u,v;

  1981.             if(c->srcFormat == PIX_FMT_PAL8) {

  1982.                 p=((const uint32_t*)(src[1]))[i];

  1983.                 r= (p>>16)&0xFF;

  1984.                 g= (p>> 8)&0xFF;

  1985.                 b=  p     &0xFF;

  1986.             } else if(c->srcFormat == PIX_FMT_RGB8) {

  1987.                 r= (i>>5    )*36;

  1988.                 g= ((i>>2)&7)*36;

  1989.                 b= (i&3     )*85;

  1990.             } else if(c->srcFormat == PIX_FMT_BGR8) {

  1991.                 b= (i>>6    )*85;

  1992.                 g= ((i>>3)&7)*36;

  1993.                 r= (i&7     )*36;

  1994.             } else if(c->srcFormat == PIX_FMT_RGB4_BYTE) {

  1995.                 r= (i>>3    )*255;

  1996.                 g= ((i>>1)&3)*85;

  1997.                 b= (i&1     )*255;

  1998.             } else if(c->srcFormat == PIX_FMT_GRAY8 || c->srcFormat == PIX_FMT_Y400A) {

  1999.                 r = g = b = i;

  2000.             } else {

  2001.                 assert(c->srcFormat == PIX_FMT_BGR4_BYTE);

  2002.                 b= (i>>3    )*255;

  2003.                 g= ((i>>1)&3)*85;

  2004.                 r= (i&1     )*255;

  2005.             }

  2006.             y= av_clip_uint8((RY*r + GY*g + BY*b + ( 33<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);

  2007.             u= av_clip_uint8((RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);

  2008.             v= av_clip_uint8((RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);

  2009.             c->pal_yuv[i]= y + (u<<8) + (v<<16);

  2010. 

  2011.             switch(c->dstFormat) {

  2012.             case PIX_FMT_BGR32:

  2013. #if !HAVE_BIGENDIAN

  2014.             case PIX_FMT_RGB24:

  2015. #endif

  2016.                 c->pal_rgb[i]=  r + (g<<8) + (b<<16);

  2017.                 break;

  2018.             case PIX_FMT_BGR32_1:

  2019. #if HAVE_BIGENDIAN

  2020.             case PIX_FMT_BGR24:

  2021. #endif

  2022.                 c->pal_rgb[i]= (r + (g<<8) + (b<<16)) << 8;

  2023.                 break;

  2024.             case PIX_FMT_RGB32_1:

  2025. #if HAVE_BIGENDIAN

  2026.             case PIX_FMT_RGB24:

  2027. #endif

  2028.                 c->pal_rgb[i]= (b + (g<<8) + (r<<16)) << 8;

  2029.                 break;

  2030.             case PIX_FMT_RGB32:

  2031. #if !HAVE_BIGENDIAN

  2032.             case PIX_FMT_BGR24:

  2033. #endif

  2034.             default:

  2035.                 c->pal_rgb[i]=  b + (g<<8) + (r<<16);

  2036.             }

  2037.         }

  2038.     }

  2039. 

  2040.     // copy strides, so they can safely be modified

  2041.     if (c->sliceDir == 1) {

  2042.         // slices go from top to bottom

  2043.         int srcStride2[4]= {srcStride[0], srcStride[1], srcStride[2], srcStride[3]};

  2044.         int dstStride2[4]= {dstStride[0], dstStride[1], dstStride[2], dstStride[3]};

  2045. 

  2046.         reset_ptr(src2, c->srcFormat);

  2047.         reset_ptr((const uint8_t**)dst2, c->dstFormat);

  2048. 

  2049.         /* reset slice direction at end of frame */

  2050.         if (srcSliceY + srcSliceH == c->srcH)

  2051.             c->sliceDir = 0;

  2052. 

  2053.         return c->swScale(c, src2, srcStride2, srcSliceY, srcSliceH, dst2, dstStride2);

  2054.     } else {

  2055.         // slices go from bottom to top => we flip the image internally

  2056.         int srcStride2[4]= {-srcStride[0], -srcStride[1], -srcStride[2], -srcStride[3]};

  2057.         int dstStride2[4]= {-dstStride[0], -dstStride[1], -dstStride[2], -dstStride[3]};

  2058. 

  2059.         src2[0] += (srcSliceH-1)*srcStride[0];

  2060.         if (!usePal(c->srcFormat))

  2061.             src2[1] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[1];

  2062.         src2[2] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[2];

  2063.         src2[3] += (srcSliceH-1)*srcStride[3];

  2064.         dst2[0] += ( c->dstH                      -1)*dstStride[0];

  2065.         dst2[1] += ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[1];

  2066.         dst2[2] += ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[2];

  2067.         dst2[3] += ( c->dstH                      -1)*dstStride[3];

  2068. 

  2069.         reset_ptr(src2, c->srcFormat);

  2070.         reset_ptr((const uint8_t**)dst2, c->dstFormat);

  2071. 

  2072.         /* reset slice direction at end of frame */

  2073.         if (!srcSliceY)

  2074.             c->sliceDir = 0;

  2075. 

  2076.         return c->swScale(c, src2, srcStride2, c->srcH-srcSliceY-srcSliceH, srcSliceH, dst2, dstStride2);

  2077.     }

  2078. }

  2079. 

  2080. /* Convert the palette to the same packed 32-bit format as the palette */

  2081. void sws_convertPalette8ToPacked32(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)

  2082. {

  2083.     long i;

  2084. 

  2085.     for (i=0; i<num_pixels; i++)

  2086.         ((uint32_t *) dst)[i] = ((const uint32_t *) palette)[src[i]];

  2087. }

  2088. 

  2089. /* Palette format: ABCD -> dst format: ABC */

  2090. void sws_convertPalette8ToPacked24(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)

  2091. {

  2092.     long i;

  2093. 

  2094.     for (i=0; i<num_pixels; i++) {

  2095.         //FIXME slow?

  2096.         dst[0]= palette[src[i]*4+0];

  2097.         dst[1]= palette[src[i]*4+1];

  2098.         dst[2]= palette[src[i]*4+2];

  2099.         dst+= 3;

  2100.     }

  2101. }