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.
28806 Commits
32 Branches
367MB
Tree: 836f47d34b
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from '836f47d34b'
${ noResults }
FFmpeg/libswscale/swscale.c

2081 lines
75KB
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 **chrUSrc,

  211.                                                     const int16_t **chrVSrc, int chrFilterSize,

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

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

  214. {

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

  216.     int i;

  217.     int shift = 11 + 16 - output_bits;

  218. 

  219. #define output_pixel(pos, val) \

  220.     if (big_endian) { \

  221.         if (output_bits == 16) { \

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

  223.         } else { \

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

  225.         } \

  226.     } else { \

  227.         if (output_bits == 16) { \

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

  229.         } else { \

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

  231.         } \

  232.     }

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

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

  235.         int j;

  236. 

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

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

  239. 

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

  241.     }

  242. 

  243.     if (uDest) {

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

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

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

  247.             int j;

  248. 

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

  250.                 u += chrUSrc[j][i] * chrFilter[j];

  251.                 v += chrVSrc[j][i] * chrFilter[j];

  252.             }

  253. 

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

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

  256.         }

  257.     }

  258. 

  259.     if (CONFIG_SWSCALE_ALPHA && aDest) {

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

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

  262.             int j;

  263. 

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

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

  266. 

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

  268.         }

  269.     }

  270. }

  271. 

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

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

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

  275.                               const int16_t *chrFilter, const int16_t **chrUSrc, \

  276.                               const int16_t **chrVSrc, \

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

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

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

  280. { \

  281.     yuv2yuvX16inC_template(lumFilter, lumSrc, lumFilterSize, \

  282.                            chrFilter, chrUSrc, chrVSrc, chrFilterSize, \

  283.                            alpSrc, \

  284.                            dest, uDest, vDest, aDest, \

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

  286. }

  287. yuv2NBPS( 9, BE, 1);

  288. yuv2NBPS( 9, LE, 0);

  289. yuv2NBPS(10, BE, 1);

  290. yuv2NBPS(10, LE, 0);

  291. yuv2NBPS(16, BE, 1);

  292. yuv2NBPS(16, LE, 0);

  293. 

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

  295.                                  const int16_t *chrFilter, const int16_t **chrUSrc, const int16_t **chrVSrc, int chrFilterSize,

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

  297.                                  enum PixelFormat dstFormat)

  298. {

  299. #define conv16(bits) \

  300.     if (isBE(dstFormat)) { \

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

  302.                                chrFilter, chrUSrc, chrVSrc, chrFilterSize, \

  303.                                alpSrc, \

  304.                                dest, uDest, vDest, aDest, \

  305.                                dstW, chrDstW); \

  306.     } else { \

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

  308.                                chrFilter, chrUSrc, chrVSrc, chrFilterSize, \

  309.                                alpSrc, \

  310.                                dest, uDest, vDest, aDest, \

  311.                                dstW, chrDstW); \

  312.     }

  313.     if (is16BPS(dstFormat)) {

  314.         conv16(16);

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

  316.         conv16(9);

  317.     } else {

  318.         conv16(10);

  319.     }

  320. #undef conv16

  321. }

  322. 

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

  324.                                const int16_t *chrFilter, const int16_t **chrUSrc,

  325.                                const int16_t **chrVSrc, int chrFilterSize,

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

  327. {

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

  329.     int i;

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

  331.         int val=1<<18;

  332.         int j;

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

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

  335. 

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

  337.     }

  338. 

  339.     if (uDest)

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

  341.             int u=1<<18;

  342.             int v=1<<18;

  343.             int j;

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

  345.                 u += chrUSrc[j][i] * chrFilter[j];

  346.                 v += chrVSrc[j][i] * chrFilter[j];

  347.             }

  348. 

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

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

  351.         }

  352. 

  353.     if (CONFIG_SWSCALE_ALPHA && aDest)

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

  355.             int val=1<<18;

  356.             int j;

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

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

  359. 

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

  361.         }

  362. 

  363. }

  364. 

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

  366.                                 const int16_t *chrFilter, const int16_t **chrUSrc,

  367.                                 const int16_t **chrVSrc, int chrFilterSize,

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

  369. {

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

  371.     int i;

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

  373.         int val=1<<18;

  374.         int j;

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

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

  377. 

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

  379.     }

  380. 

  381.     if (!uDest)

  382.         return;

  383. 

  384.     if (dstFormat == PIX_FMT_NV12)

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

  386.             int u=1<<18;

  387.             int v=1<<18;

  388.             int j;

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

  390.                 u += chrUSrc[j][i] * chrFilter[j];

  391.                 v += chrVSrc[j][i] * chrFilter[j];

  392.             }

  393. 

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

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

  396.         }

  397.     else

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

  399.             int u=1<<18;

  400.             int v=1<<18;

  401.             int j;

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

  403.                 u += chrUSrc[j][i] * chrFilter[j];

  404.                 v += chrVSrc[j][i] * chrFilter[j];

  405.             }

  406. 

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

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

  409.         }

  410. }

  411. 

  412. #define YSCALE_YUV_2_PACKEDX_NOCLIP_C(type,alpha) \

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

  414.         int j;\

  415.         int Y1 = 1<<18;\

  416.         int Y2 = 1<<18;\

  417.         int U  = 1<<18;\

  418.         int V  = 1<<18;\

  419.         int av_unused A1, A2;\

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

  421.         const int i2= 2*i;\

  422.         \

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

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

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

  426.         }\

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

  428.             U += chrUSrc[j][i] * chrFilter[j];\

  429.             V += chrVSrc[j][i] * chrFilter[j];\

  430.         }\

  431.         Y1>>=19;\

  432.         Y2>>=19;\

  433.         U >>=19;\

  434.         V >>=19;\

  435.         if (alpha) {\

  436.             A1 = 1<<18;\

  437.             A2 = 1<<18;\

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

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

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

  441.             }\

  442.             A1>>=19;\

  443.             A2>>=19;\

  444.         }

  445. 

  446. #define YSCALE_YUV_2_PACKEDX_C(type,alpha) \

  447.         YSCALE_YUV_2_PACKEDX_NOCLIP_C(type,alpha)\

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

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

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

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

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

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

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

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

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

  457.         }\

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

  459.             A1=av_clip_uint8(A1);\

  460.             A2=av_clip_uint8(A2);\

  461.         }

  462. 

  463. #define YSCALE_YUV_2_PACKEDX_FULL_C(rnd,alpha) \

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

  465.         int j;\

  466.         int Y = 0;\

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

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

  469.         int av_unused A;\

  470.         int R,G,B;\

  471.         \

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

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

  474.         }\

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

  476.             U += chrUSrc[j][i] * chrFilter[j];\

  477.             V += chrVSrc[j][i] * chrFilter[j];\

  478.         }\

  479.         Y >>=10;\

  480.         U >>=10;\

  481.         V >>=10;\

  482.         if (alpha) {\

  483.             A = rnd;\

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

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

  486.             A >>=19;\

  487.             if (A&256)\

  488.                 A = av_clip_uint8(A);\

  489.         }

  490. 

  491. #define YSCALE_YUV_2_RGBX_FULL_C(rnd,alpha) \

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

  493.         Y-= c->yuv2rgb_y_offset;\

  494.         Y*= c->yuv2rgb_y_coeff;\

  495.         Y+= rnd;\

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

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

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

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

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

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

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

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

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

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

  506.         }

  507. 

  508. #define YSCALE_YUV_2_GRAY16_C \

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

  510.         int j;\

  511.         int Y1 = 1<<18;\

  512.         int Y2 = 1<<18;\

  513.         int U  = 1<<18;\

  514.         int V  = 1<<18;\

  515.         \

  516.         const int i2= 2*i;\

  517.         \

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

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

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

  521.         }\

  522.         Y1>>=11;\

  523.         Y2>>=11;\

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

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

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

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

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

  529.         }

  530. 

  531. #define YSCALE_YUV_2_RGBX_C(type,alpha) \

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

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

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

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

  536. 

  537. #define YSCALE_YUV_2_PACKED2_C(type,alpha)   \

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

  539.         const int i2= 2*i;       \

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

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

  542.         int U= (ubuf0[i]*uvalpha1+ubuf1[i]*uvalpha)>>19;              \

  543.         int V= (vbuf0[i]*uvalpha1+vbuf1[i]*uvalpha)>>19;              \

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

  545.         int av_unused A1, A2;                                         \

  546.         if (alpha) {\

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

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

  549.         }

  550. 

  551. #define YSCALE_YUV_2_GRAY16_2_C   \

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

  553.         const int i2= 2*i;       \

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

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

  556. 

  557. #define YSCALE_YUV_2_RGB2_C(type,alpha) \

  558.     YSCALE_YUV_2_PACKED2_C(type,alpha)\

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

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

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

  562. 

  563. #define YSCALE_YUV_2_PACKED1_C(type,alpha) \

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

  565.         const int i2= 2*i;\

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

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

  568.         int U= (ubuf1[i])>>7;\

  569.         int V= (vbuf1[i])>>7;\

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

  571.         int av_unused A1, A2;\

  572.         if (alpha) {\

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

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

  575.         }

  576. 

  577. #define YSCALE_YUV_2_GRAY16_1_C \

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

  579.         const int i2= 2*i;\

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

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

  582. 

  583. #define YSCALE_YUV_2_RGB1_C(type,alpha) \

  584.     YSCALE_YUV_2_PACKED1_C(type,alpha)\

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

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

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

  588. 

  589. #define YSCALE_YUV_2_PACKED1B_C(type,alpha) \

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

  591.         const int i2= 2*i;\

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

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

  594.         int U= (ubuf0[i] + ubuf1[i])>>8;\

  595.         int V= (vbuf0[i] + vbuf1[i])>>8;\

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

  597.         int av_unused A1, A2;\

  598.         if (alpha) {\

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

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

  601.         }

  602. 

  603. #define YSCALE_YUV_2_RGB1B_C(type,alpha) \

  604.     YSCALE_YUV_2_PACKED1B_C(type,alpha)\

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

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

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

  608. 

  609. #define YSCALE_YUV_2_MONO2_C \

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

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

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

  613.         int acc;\

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

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

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

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

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

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

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

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

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

  623.         dest++;\

  624.     }

  625. 

  626. #define YSCALE_YUV_2_MONOX_C \

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

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

  629.     int acc=0;\

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

  631.         int j;\

  632.         int Y1=1<<18;\

  633.         int Y2=1<<18;\

  634. \

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

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

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

  638.         }\

  639.         Y1>>=19;\

  640.         Y2>>=19;\

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

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

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

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

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

  646.         }\

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

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

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

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

  651.             dest++;\

  652.         }\

  653.     }

  654. 

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

  656.     switch(c->dstFormat) {\

  657.     case PIX_FMT_RGB48BE:\

  658.     case PIX_FMT_RGB48LE:\

  659.         func(uint8_t,0)\

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

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

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

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

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

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

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

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

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

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

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

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

  672.             dest+=12;\

  673.         }\

  674.         break;\

  675.     case PIX_FMT_BGR48BE:\

  676.     case PIX_FMT_BGR48LE:\

  677.         func(uint8_t,0)\

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

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

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

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

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

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

  684.             dest+=12;\

  685.         }\

  686.         break;\

  687.     case PIX_FMT_RGBA:\

  688.     case PIX_FMT_BGRA:\

  689.         if (CONFIG_SMALL) {\

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

  691.             func(uint32_t,needAlpha)\

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

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

  694.             }\

  695.         } else {\

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

  697.                 func(uint32_t,1)\

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

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

  700.                 }\

  701.             } else {\

  702.                 func(uint32_t,0)\

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

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

  705.                 }\

  706.             }\

  707.         }\

  708.         break;\

  709.     case PIX_FMT_ARGB:\

  710.     case PIX_FMT_ABGR:\

  711.         if (CONFIG_SMALL) {\

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

  713.             func(uint32_t,needAlpha)\

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

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

  716.             }\

  717.         } else {\

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

  719.                 func(uint32_t,1)\

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

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

  722.                 }\

  723.             } else {\

  724.                 func(uint32_t,0)\

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

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

  727.                 }\

  728.             }\

  729.         }                \

  730.         break;\

  731.     case PIX_FMT_RGB24:\

  732.         func(uint8_t,0)\

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

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

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

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

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

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

  739.             dest+=6;\

  740.         }\

  741.         break;\

  742.     case PIX_FMT_BGR24:\

  743.         func(uint8_t,0)\

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

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

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

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

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

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

  750.             dest+=6;\

  751.         }\

  752.         break;\

  753.     case PIX_FMT_RGB565BE:\

  754.     case PIX_FMT_RGB565LE:\

  755.     case PIX_FMT_BGR565BE:\

  756.     case PIX_FMT_BGR565LE:\

  757.         {\

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

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

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

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

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

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

  764.             func(uint16_t,0)\

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

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

  767.             }\

  768.         }\

  769.         break;\

  770.     case PIX_FMT_RGB555BE:\

  771.     case PIX_FMT_RGB555LE:\

  772.     case PIX_FMT_BGR555BE:\

  773.     case PIX_FMT_BGR555LE:\

  774.         {\

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

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

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

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

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

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

  781.             func(uint16_t,0)\

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

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

  784.             }\

  785.         }\

  786.         break;\

  787.     case PIX_FMT_RGB444BE:\

  788.     case PIX_FMT_RGB444LE:\

  789.     case PIX_FMT_BGR444BE:\

  790.     case PIX_FMT_BGR444LE:\

  791.         {\

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

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

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

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

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

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

  798.             func(uint16_t,0)\

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

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

  801.             }\

  802.         }\

  803.         break;\

  804.     case PIX_FMT_RGB8:\

  805.     case PIX_FMT_BGR8:\

  806.         {\

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

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

  809.             func(uint8_t,0)\

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

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

  812.             }\

  813.         }\

  814.         break;\

  815.     case PIX_FMT_RGB4:\

  816.     case PIX_FMT_BGR4:\

  817.         {\

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

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

  820.             func(uint8_t,0)\

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

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

  823.             }\

  824.         }\

  825.         break;\

  826.     case PIX_FMT_RGB4_BYTE:\

  827.     case PIX_FMT_BGR4_BYTE:\

  828.         {\

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

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

  831.             func(uint8_t,0)\

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

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

  834.             }\

  835.         }\

  836.         break;\

  837.     case PIX_FMT_MONOBLACK:\

  838.     case PIX_FMT_MONOWHITE:\

  839.         {\

  840.             func_monoblack\

  841.         }\

  842.         break;\

  843.     case PIX_FMT_YUYV422:\

  844.         func2\

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

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

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

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

  849.         }                \

  850.         break;\

  851.     case PIX_FMT_UYVY422:\

  852.         func2\

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

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

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

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

  857.         }                \

  858.         break;\

  859.     case PIX_FMT_GRAY16BE:\

  860.         func_g16\

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

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

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

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

  865.         }                \

  866.         break;\

  867.     case PIX_FMT_GRAY16LE:\

  868.         func_g16\

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

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

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

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

  873.         }                \

  874.         break;\

  875.     }

  876. 

  877. static inline void yuv2packedXinC(SwsContext *c, const int16_t *lumFilter,

  878.                                   const int16_t **lumSrc, int lumFilterSize,

  879.                                   const int16_t *chrFilter, const int16_t **chrUSrc,

  880.                                   const int16_t **chrVSrc, int chrFilterSize,

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

  882. {

  883.     int i;

  884.     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)

  885. }

  886. 

  887. static inline void yuv2rgbXinC_full(SwsContext *c, const int16_t *lumFilter,

  888.                                     const int16_t **lumSrc, int lumFilterSize,

  889.                                     const int16_t *chrFilter, const int16_t **chrUSrc,

  890.                                     const int16_t **chrVSrc, int chrFilterSize,

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

  892. {

  893.     int i;

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

  895.     int aidx= 3;

  896. 

  897.     switch(c->dstFormat) {

  898.     case PIX_FMT_ARGB:

  899.         dest++;

  900.         aidx= 0;

  901.     case PIX_FMT_RGB24:

  902.         aidx--;

  903.     case PIX_FMT_RGBA:

  904.         if (CONFIG_SMALL) {

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

  906.             YSCALE_YUV_2_RGBX_FULL_C(1<<21, needAlpha)

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

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

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

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

  911.                 dest+= step;

  912.             }

  913.         } else {

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

  915.                 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 1)

  916.                     dest[aidx]= A;

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

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

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

  920.                     dest+= step;

  921.                 }

  922.             } else {

  923.                 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 0)

  924.                     dest[aidx]= 255;

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

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

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

  928.                     dest+= step;

  929.                 }

  930.             }

  931.         }

  932.         break;

  933.     case PIX_FMT_ABGR:

  934.         dest++;

  935.         aidx= 0;

  936.     case PIX_FMT_BGR24:

  937.         aidx--;

  938.     case PIX_FMT_BGRA:

  939.         if (CONFIG_SMALL) {

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

  941.             YSCALE_YUV_2_RGBX_FULL_C(1<<21, needAlpha)

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

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

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

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

  946.                 dest+= step;

  947.             }

  948.         } else {

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

  950.                 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 1)

  951.                     dest[aidx]= A;

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

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

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

  955.                     dest+= step;

  956.                 }

  957.             } else {

  958.                 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 0)

  959.                     dest[aidx]= 255;

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

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

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

  963.                     dest+= step;

  964.                 }

  965.             }

  966.         }

  967.         break;

  968.     default:

  969.         assert(0);

  970.     }

  971. }

  972. 

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

  974. {

  975.     int i;

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

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

  978.         memset(ptr, val, width);

  979.         ptr += stride;

  980.     }

  981. }

  982. 

  983. static inline void rgb48ToY(uint8_t *dst, const uint8_t *src, int width,

  984.                             uint32_t *unused)

  985. {

  986.     int i;

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

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

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

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

  991. 

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

  993.     }

  994. }

  995. 

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

  997.                              const uint8_t *src1, const uint8_t *src2,

  998.                              int width, uint32_t *unused)

  999. {

  1000.     int i;

  1001.     assert(src1==src2);

  1002.     for (i = 0; i < width; i++) {

  1003.         int r = src1[6*i + 0];

  1004.         int g = src1[6*i + 2];

  1005.         int b = src1[6*i + 4];

  1006. 

  1007.         dstU[i] = (RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;

  1008.         dstV[i] = (RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;

  1009.     }

  1010. }

  1011. 

  1012. static inline void rgb48ToUV_half(uint8_t *dstU, uint8_t *dstV,

  1013.                                   const uint8_t *src1, const uint8_t *src2,

  1014.                                   int width, uint32_t *unused)

  1015. {

  1016.     int i;

  1017.     assert(src1==src2);

  1018.     for (i = 0; i < width; i++) {

  1019.         int r= src1[12*i + 0] + src1[12*i + 6];

  1020.         int g= src1[12*i + 2] + src1[12*i + 8];

  1021.         int b= src1[12*i + 4] + src1[12*i + 10];

  1022. 

  1023.         dstU[i]= (RU*r + GU*g + BU*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);

  1024.         dstV[i]= (RV*r + GV*g + BV*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);

  1025.     }

  1026. }

  1027. 

  1028. static inline void bgr48ToY(uint8_t *dst, const uint8_t *src, int width,

  1029.                             uint32_t *unused)

  1030. {

  1031.     int i;

  1032.     for (i = 0; i < width; i++) {

  1033.         int b = src[i*6+0];

  1034.         int g = src[i*6+2];

  1035.         int r = src[i*6+4];

  1036. 

  1037.         dst[i] = (RY*r + GY*g + BY*b + (33<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;

  1038.     }

  1039. }

  1040. 

  1041. static inline void bgr48ToUV(uint8_t *dstU, uint8_t *dstV,

  1042.                              const uint8_t *src1, const uint8_t *src2,

  1043.                              int width, uint32_t *unused)

  1044. {

  1045.     int i;

  1046.     for (i = 0; i < width; i++) {

  1047.         int b = src1[6*i + 0];

  1048.         int g = src1[6*i + 2];

  1049.         int r = src1[6*i + 4];

  1050. 

  1051.         dstU[i] = (RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;

  1052.         dstV[i] = (RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;

  1053.     }

  1054. }

  1055. 

  1056. static inline void bgr48ToUV_half(uint8_t *dstU, uint8_t *dstV,

  1057.                                   const uint8_t *src1, const uint8_t *src2,

  1058.                                   int width, uint32_t *unused)

  1059. {

  1060.     int i;

  1061.     for (i = 0; i < width; i++) {

  1062.         int b= src1[12*i + 0] + src1[12*i + 6];

  1063.         int g= src1[12*i + 2] + src1[12*i + 8];

  1064.         int r= src1[12*i + 4] + src1[12*i + 10];

  1065. 

  1066.         dstU[i]= (RU*r + GU*g + BU*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);

  1067.         dstV[i]= (RV*r + GV*g + BV*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);

  1068.     }

  1069. }

  1070. 

  1071. #define BGR2Y(type, name, shr, shg, shb, maskr, maskg, maskb, RY, GY, BY, S)\

  1072. static inline void name(uint8_t *dst, const uint8_t *src, int width, uint32_t *unused)\

  1073. {\

  1074.     int i;\

  1075.     for (i=0; i<width; i++) {\

  1076.         int b= (((const type*)src)[i]>>shb)&maskb;\

  1077.         int g= (((const type*)src)[i]>>shg)&maskg;\

  1078.         int r= (((const type*)src)[i]>>shr)&maskr;\

  1079. \

  1080.         dst[i]= (((RY)*r + (GY)*g + (BY)*b + (33<<((S)-1)))>>(S));\

  1081.     }\

  1082. }

  1083. 

  1084. BGR2Y(uint32_t, bgr32ToY,16, 0, 0, 0x00FF, 0xFF00, 0x00FF, RY<< 8, GY   , BY<< 8, RGB2YUV_SHIFT+8)

  1085. BGR2Y(uint32_t,bgr321ToY,16,16, 0, 0xFF00, 0x00FF, 0xFF00, RY    , GY<<8, BY    , RGB2YUV_SHIFT+8)

  1086. BGR2Y(uint32_t, rgb32ToY, 0, 0,16, 0x00FF, 0xFF00, 0x00FF, RY<< 8, GY   , BY<< 8, RGB2YUV_SHIFT+8)

  1087. BGR2Y(uint32_t,rgb321ToY, 0,16,16, 0xFF00, 0x00FF, 0xFF00, RY    , GY<<8, BY    , RGB2YUV_SHIFT+8)

  1088. BGR2Y(uint16_t, bgr16ToY, 0, 0, 0, 0x001F, 0x07E0, 0xF800, RY<<11, GY<<5, BY    , RGB2YUV_SHIFT+8)

  1089. BGR2Y(uint16_t, bgr15ToY, 0, 0, 0, 0x001F, 0x03E0, 0x7C00, RY<<10, GY<<5, BY    , RGB2YUV_SHIFT+7)

  1090. BGR2Y(uint16_t, rgb16ToY, 0, 0, 0, 0xF800, 0x07E0, 0x001F, RY    , GY<<5, BY<<11, RGB2YUV_SHIFT+8)

  1091. BGR2Y(uint16_t, rgb15ToY, 0, 0, 0, 0x7C00, 0x03E0, 0x001F, RY    , GY<<5, BY<<10, RGB2YUV_SHIFT+7)

  1092. 

  1093. static inline void abgrToA(uint8_t *dst, const uint8_t *src, int width, uint32_t *unused)

  1094. {

  1095.     int i;

  1096.     for (i=0; i<width; i++) {

  1097.         dst[i]= src[4*i];

  1098.     }

  1099. }

  1100. 

  1101. #define BGR2UV(type, name, shr, shg, shb, shp, maskr, maskg, maskb, RU, GU, BU, RV, GV, BV, S) \

  1102. static inline void name(uint8_t *dstU, uint8_t *dstV, const uint8_t *src, const uint8_t *dummy, int width, uint32_t *unused)\

  1103. {\

  1104.     int i;\

  1105.     for (i=0; i<width; i++) {\

  1106.         int b= ((((const type*)src)[i]>>shp)&maskb)>>shb;\

  1107.         int g= ((((const type*)src)[i]>>shp)&maskg)>>shg;\

  1108.         int r= ((((const type*)src)[i]>>shp)&maskr)>>shr;\

  1109. \

  1110.         dstU[i]= ((RU)*r + (GU)*g + (BU)*b + (257<<((S)-1)))>>(S);\

  1111.         dstV[i]= ((RV)*r + (GV)*g + (BV)*b + (257<<((S)-1)))>>(S);\

  1112.     }\

  1113. }\

  1114. static inline void name ## _half(uint8_t *dstU, uint8_t *dstV, const uint8_t *src, const uint8_t *dummy, int width, uint32_t *unused)\

  1115. {\

  1116.     int i;\

  1117.     for (i=0; i<width; i++) {\

  1118.         int pix0= ((const type*)src)[2*i+0]>>shp;\

  1119.         int pix1= ((const type*)src)[2*i+1]>>shp;\

  1120.         int g= (pix0&~(maskr|maskb))+(pix1&~(maskr|maskb));\

  1121.         int b= ((pix0+pix1-g)&(maskb|(2*maskb)))>>shb;\

  1122.         int r= ((pix0+pix1-g)&(maskr|(2*maskr)))>>shr;\

  1123.         g&= maskg|(2*maskg);\

  1124. \

  1125.         g>>=shg;\

  1126. \

  1127.         dstU[i]= ((RU)*r + (GU)*g + (BU)*b + (257<<(S)))>>((S)+1);\

  1128.         dstV[i]= ((RV)*r + (GV)*g + (BV)*b + (257<<(S)))>>((S)+1);\

  1129.     }\

  1130. }

  1131. 

  1132. BGR2UV(uint32_t, bgr32ToUV,16, 0, 0, 0, 0xFF0000, 0xFF00,   0x00FF, RU<< 8, GU   , BU<< 8, RV<< 8, GV   , BV<< 8, RGB2YUV_SHIFT+8)

  1133. BGR2UV(uint32_t,bgr321ToUV,16, 0, 0, 8, 0xFF0000, 0xFF00,   0x00FF, RU<< 8, GU   , BU<< 8, RV<< 8, GV   , BV<< 8, RGB2YUV_SHIFT+8)

  1134. BGR2UV(uint32_t, rgb32ToUV, 0, 0,16, 0,   0x00FF, 0xFF00, 0xFF0000, RU<< 8, GU   , BU<< 8, RV<< 8, GV   , BV<< 8, RGB2YUV_SHIFT+8)

  1135. BGR2UV(uint32_t,rgb321ToUV, 0, 0,16, 8,   0x00FF, 0xFF00, 0xFF0000, RU<< 8, GU   , BU<< 8, RV<< 8, GV   , BV<< 8, RGB2YUV_SHIFT+8)

  1136. BGR2UV(uint16_t, bgr16ToUV, 0, 0, 0, 0,   0x001F, 0x07E0,   0xF800, RU<<11, GU<<5, BU    , RV<<11, GV<<5, BV    , RGB2YUV_SHIFT+8)

  1137. BGR2UV(uint16_t, bgr15ToUV, 0, 0, 0, 0,   0x001F, 0x03E0,   0x7C00, RU<<10, GU<<5, BU    , RV<<10, GV<<5, BV    , RGB2YUV_SHIFT+7)

  1138. BGR2UV(uint16_t, rgb16ToUV, 0, 0, 0, 0,   0xF800, 0x07E0,   0x001F, RU    , GU<<5, BU<<11, RV    , GV<<5, BV<<11, RGB2YUV_SHIFT+8)

  1139. BGR2UV(uint16_t, rgb15ToUV, 0, 0, 0, 0,   0x7C00, 0x03E0,   0x001F, RU    , GU<<5, BU<<10, RV    , GV<<5, BV<<10, RGB2YUV_SHIFT+7)

  1140. 

  1141. static inline void palToY(uint8_t *dst, const uint8_t *src, int width, uint32_t *pal)

  1142. {

  1143.     int i;

  1144.     for (i=0; i<width; i++) {

  1145.         int d= src[i];

  1146. 

  1147.         dst[i]= pal[d] & 0xFF;

  1148.     }

  1149. }

  1150. 

  1151. static inline void palToUV(uint8_t *dstU, uint8_t *dstV,

  1152.                            const uint8_t *src1, const uint8_t *src2,

  1153.                            int width, uint32_t *pal)

  1154. {

  1155.     int i;

  1156.     assert(src1 == src2);

  1157.     for (i=0; i<width; i++) {

  1158.         int p= pal[src1[i]];

  1159. 

  1160.         dstU[i]= p>>8;

  1161.         dstV[i]= p>>16;

  1162.     }

  1163. }

  1164. 

  1165. static inline void monowhite2Y(uint8_t *dst, const uint8_t *src, int width, uint32_t *unused)

  1166. {

  1167.     int i, j;

  1168.     for (i=0; i<width/8; i++) {

  1169.         int d= ~src[i];

  1170.         for(j=0; j<8; j++)

  1171.             dst[8*i+j]= ((d>>(7-j))&1)*255;

  1172.     }

  1173. }

  1174. 

  1175. static inline void monoblack2Y(uint8_t *dst, const uint8_t *src, int width, uint32_t *unused)

  1176. {

  1177.     int i, j;

  1178.     for (i=0; i<width/8; i++) {

  1179.         int d= src[i];

  1180.         for(j=0; j<8; j++)

  1181.             dst[8*i+j]= ((d>>(7-j))&1)*255;

  1182.     }

  1183. }

  1184. 

  1185. //Note: we have C, MMX, MMX2, 3DNOW versions, there is no 3DNOW+MMX2 one

  1186. //Plain C versions

  1187. 

  1188. #define COMPILE_TEMPLATE_MMX2 0

  1189. #define COMPILE_TEMPLATE_ALTIVEC 0

  1190. 

  1191. #include "swscale_template.c"

  1192. 

  1193. #if HAVE_ALTIVEC

  1194. #undef RENAME

  1195. #undef COMPILE_TEMPLATE_ALTIVEC

  1196. #define COMPILE_TEMPLATE_ALTIVEC 1

  1197. #define RENAME(a) a ## _altivec

  1198. #include "ppc/swscale_template.c"

  1199. #endif

  1200. 

  1201. //MMX versions

  1202. #if HAVE_MMX

  1203. #undef RENAME

  1204. #undef COMPILE_TEMPLATE_MMX2

  1205. #define COMPILE_TEMPLATE_MMX2 0

  1206. #define RENAME(a) a ## _MMX

  1207. #include "x86/swscale_template.c"

  1208. #endif

  1209. 

  1210. //MMX2 versions

  1211. #if HAVE_MMX2

  1212. #undef RENAME

  1213. #undef COMPILE_TEMPLATE_MMX2

  1214. #define COMPILE_TEMPLATE_MMX2 1

  1215. #define RENAME(a) a ## _MMX2

  1216. #include "x86/swscale_template.c"

  1217. #endif

  1218. 

  1219. SwsFunc ff_getSwsFunc(SwsContext *c)

  1220. {

  1221.     int cpu_flags = av_get_cpu_flags();

  1222. 

  1223.     sws_init_swScale_c(c);

  1224. 

  1225. #if HAVE_MMX

  1226.     if (cpu_flags & AV_CPU_FLAG_MMX)

  1227.         sws_init_swScale_MMX(c);

  1228. #endif

  1229. #if HAVE_MMX2

  1230.     if (cpu_flags & AV_CPU_FLAG_MMX2)

  1231.         sws_init_swScale_MMX2(c);

  1232. #endif

  1233. #if HAVE_ALTIVEC

  1234.     if (cpu_flags & AV_CPU_FLAG_ALTIVEC)

  1235.         sws_init_swScale_altivec(c);

  1236. #endif

  1237. 

  1238.     return swScale_c;

  1239. }

  1240. 

  1241. static void copyPlane(const uint8_t *src, int srcStride,

  1242.                       int srcSliceY, int srcSliceH, int width,

  1243.                       uint8_t *dst, int dstStride)

  1244. {

  1245.     dst += dstStride * srcSliceY;

  1246.     if (dstStride == srcStride && srcStride > 0) {

  1247.         memcpy(dst, src, srcSliceH * dstStride);

  1248.     } else {

  1249.         int i;

  1250.         for (i=0; i<srcSliceH; i++) {

  1251.             memcpy(dst, src, width);

  1252.             src += srcStride;

  1253.             dst += dstStride;

  1254.         }

  1255.     }

  1256. }

  1257. 

  1258. static int planarToNv12Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,

  1259.                                int srcSliceH, uint8_t* dstParam[], int dstStride[])

  1260. {

  1261.     uint8_t *dst = dstParam[1] + dstStride[1]*srcSliceY/2;

  1262. 

  1263.     copyPlane(src[0], srcStride[0], srcSliceY, srcSliceH, c->srcW,

  1264.               dstParam[0], dstStride[0]);

  1265. 

  1266.     if (c->dstFormat == PIX_FMT_NV12)

  1267.         interleaveBytes(src[1], src[2], dst, c->srcW/2, srcSliceH/2, srcStride[1], srcStride[2], dstStride[0]);

  1268.     else

  1269.         interleaveBytes(src[2], src[1], dst, c->srcW/2, srcSliceH/2, srcStride[2], srcStride[1], dstStride[0]);

  1270. 

  1271.     return srcSliceH;

  1272. }

  1273. 

  1274. static int planarToYuy2Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,

  1275.                                int srcSliceH, uint8_t* dstParam[], int dstStride[])

  1276. {

  1277.     uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;

  1278. 

  1279.     yv12toyuy2(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);

  1280. 

  1281.     return srcSliceH;

  1282. }

  1283. 

  1284. static int planarToUyvyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,

  1285.                                int srcSliceH, uint8_t* dstParam[], int dstStride[])

  1286. {

  1287.     uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;

  1288. 

  1289.     yv12touyvy(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);

  1290. 

  1291.     return srcSliceH;

  1292. }

  1293. 

  1294. static int yuv422pToYuy2Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,

  1295.                                 int srcSliceH, uint8_t* dstParam[], int dstStride[])

  1296. {

  1297.     uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;

  1298. 

  1299.     yuv422ptoyuy2(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);

  1300. 

  1301.     return srcSliceH;

  1302. }

  1303. 

  1304. static int yuv422pToUyvyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,

  1305.                                 int srcSliceH, uint8_t* dstParam[], int dstStride[])

  1306. {

  1307.     uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;

  1308. 

  1309.     yuv422ptouyvy(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);

  1310. 

  1311.     return srcSliceH;

  1312. }

  1313. 

  1314. static int yuyvToYuv420Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,

  1315.                                int srcSliceH, uint8_t* dstParam[], int dstStride[])

  1316. {

  1317.     uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;

  1318.     uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY/2;

  1319.     uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY/2;

  1320. 

  1321.     yuyvtoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);

  1322. 

  1323.     if (dstParam[3])

  1324.         fillPlane(dstParam[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);

  1325. 

  1326.     return srcSliceH;

  1327. }

  1328. 

  1329. static int yuyvToYuv422Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,

  1330.                                int srcSliceH, uint8_t* dstParam[], int dstStride[])

  1331. {

  1332.     uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;

  1333.     uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY;

  1334.     uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY;

  1335. 

  1336.     yuyvtoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);

  1337. 

  1338.     return srcSliceH;

  1339. }

  1340. 

  1341. static int uyvyToYuv420Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,

  1342.                                int srcSliceH, uint8_t* dstParam[], int dstStride[])

  1343. {

  1344.     uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;

  1345.     uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY/2;

  1346.     uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY/2;

  1347. 

  1348.     uyvytoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);

  1349. 

  1350.     if (dstParam[3])

  1351.         fillPlane(dstParam[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);

  1352. 

  1353.     return srcSliceH;

  1354. }

  1355. 

  1356. static int uyvyToYuv422Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,

  1357.                                int srcSliceH, uint8_t* dstParam[], int dstStride[])

  1358. {

  1359.     uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;

  1360.     uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY;

  1361.     uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY;

  1362. 

  1363.     uyvytoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);

  1364. 

  1365.     return srcSliceH;

  1366. }

  1367. 

  1368. static void gray8aToPacked32(const uint8_t *src, uint8_t *dst, int num_pixels, const uint8_t *palette)

  1369. {

  1370.     int i;

  1371.     for (i=0; i<num_pixels; i++)

  1372.         ((uint32_t *) dst)[i] = ((const uint32_t *)palette)[src[i<<1]] | (src[(i<<1)+1] << 24);

  1373. }

  1374. 

  1375. static void gray8aToPacked32_1(const uint8_t *src, uint8_t *dst, int num_pixels, const uint8_t *palette)

  1376. {

  1377.     int i;

  1378. 

  1379.     for (i=0; i<num_pixels; i++)

  1380.         ((uint32_t *) dst)[i] = ((const uint32_t *)palette)[src[i<<1]] | src[(i<<1)+1];

  1381. }

  1382. 

  1383. static void gray8aToPacked24(const uint8_t *src, uint8_t *dst, int num_pixels, const uint8_t *palette)

  1384. {

  1385.     int i;

  1386. 

  1387.     for (i=0; i<num_pixels; i++) {

  1388.         //FIXME slow?

  1389.         dst[0]= palette[src[i<<1]*4+0];

  1390.         dst[1]= palette[src[i<<1]*4+1];

  1391.         dst[2]= palette[src[i<<1]*4+2];

  1392.         dst+= 3;

  1393.     }

  1394. }

  1395. 

  1396. static int palToRgbWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,

  1397.                            int srcSliceH, uint8_t* dst[], int dstStride[])

  1398. {

  1399.     const enum PixelFormat srcFormat= c->srcFormat;

  1400.     const enum PixelFormat dstFormat= c->dstFormat;

  1401.     void (*conv)(const uint8_t *src, uint8_t *dst, int num_pixels,

  1402.                  const uint8_t *palette)=NULL;

  1403.     int i;

  1404.     uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;

  1405.     const uint8_t *srcPtr= src[0];

  1406. 

  1407.     if (srcFormat == PIX_FMT_Y400A) {

  1408.         switch (dstFormat) {

  1409.         case PIX_FMT_RGB32  : conv = gray8aToPacked32; break;

  1410.         case PIX_FMT_BGR32  : conv = gray8aToPacked32; break;

  1411.         case PIX_FMT_BGR32_1: conv = gray8aToPacked32_1; break;

  1412.         case PIX_FMT_RGB32_1: conv = gray8aToPacked32_1; break;

  1413.         case PIX_FMT_RGB24  : conv = gray8aToPacked24; break;

  1414.         case PIX_FMT_BGR24  : conv = gray8aToPacked24; break;

  1415.         }

  1416.     } else if (usePal(srcFormat)) {

  1417.         switch (dstFormat) {

  1418.         case PIX_FMT_RGB32  : conv = sws_convertPalette8ToPacked32; break;

  1419.         case PIX_FMT_BGR32  : conv = sws_convertPalette8ToPacked32; break;

  1420.         case PIX_FMT_BGR32_1: conv = sws_convertPalette8ToPacked32; break;

  1421.         case PIX_FMT_RGB32_1: conv = sws_convertPalette8ToPacked32; break;

  1422.         case PIX_FMT_RGB24  : conv = sws_convertPalette8ToPacked24; break;

  1423.         case PIX_FMT_BGR24  : conv = sws_convertPalette8ToPacked24; break;

  1424.         }

  1425.     }

  1426. 

  1427.     if (!conv)

  1428.         av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",

  1429.                sws_format_name(srcFormat), sws_format_name(dstFormat));

  1430.     else {

  1431.         for (i=0; i<srcSliceH; i++) {

  1432.             conv(srcPtr, dstPtr, c->srcW, (uint8_t *) c->pal_rgb);

  1433.             srcPtr+= srcStride[0];

  1434.             dstPtr+= dstStride[0];

  1435.         }

  1436.     }

  1437. 

  1438.     return srcSliceH;

  1439. }

  1440. 

  1441. #define isRGBA32(x) (            \

  1442.            (x) == PIX_FMT_ARGB   \

  1443.         || (x) == PIX_FMT_RGBA   \

  1444.         || (x) == PIX_FMT_BGRA   \

  1445.         || (x) == PIX_FMT_ABGR   \

  1446.         )

  1447. 

  1448. /* {RGB,BGR}{15,16,24,32,32_1} -> {RGB,BGR}{15,16,24,32} */

  1449. static int rgbToRgbWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,

  1450.                            int srcSliceH, uint8_t* dst[], int dstStride[])

  1451. {

  1452.     const enum PixelFormat srcFormat= c->srcFormat;

  1453.     const enum PixelFormat dstFormat= c->dstFormat;

  1454.     const int srcBpp= (c->srcFormatBpp + 7) >> 3;

  1455.     const int dstBpp= (c->dstFormatBpp + 7) >> 3;

  1456.     const int srcId= c->srcFormatBpp >> 2; /* 1:0, 4:1, 8:2, 15:3, 16:4, 24:6, 32:8 */

  1457.     const int dstId= c->dstFormatBpp >> 2;

  1458.     void (*conv)(const uint8_t *src, uint8_t *dst, int src_size)=NULL;

  1459. 

  1460. #define CONV_IS(src, dst) (srcFormat == PIX_FMT_##src && dstFormat == PIX_FMT_##dst)

  1461. 

  1462.     if (isRGBA32(srcFormat) && isRGBA32(dstFormat)) {

  1463.         if (     CONV_IS(ABGR, RGBA)

  1464.               || CONV_IS(ARGB, BGRA)

  1465.               || CONV_IS(BGRA, ARGB)

  1466.               || CONV_IS(RGBA, ABGR)) conv = shuffle_bytes_3210;

  1467.         else if (CONV_IS(ABGR, ARGB)

  1468.               || CONV_IS(ARGB, ABGR)) conv = shuffle_bytes_0321;

  1469.         else if (CONV_IS(ABGR, BGRA)

  1470.               || CONV_IS(ARGB, RGBA)) conv = shuffle_bytes_1230;

  1471.         else if (CONV_IS(BGRA, RGBA)

  1472.               || CONV_IS(RGBA, BGRA)) conv = shuffle_bytes_2103;

  1473.         else if (CONV_IS(BGRA, ABGR)

  1474.               || CONV_IS(RGBA, ARGB)) conv = shuffle_bytes_3012;

  1475.     } else

  1476.     /* BGR -> BGR */

  1477.     if (  (isBGRinInt(srcFormat) && isBGRinInt(dstFormat))

  1478.        || (isRGBinInt(srcFormat) && isRGBinInt(dstFormat))) {

  1479.         switch(srcId | (dstId<<4)) {

  1480.         case 0x34: conv= rgb16to15; break;

  1481.         case 0x36: conv= rgb24to15; break;

  1482.         case 0x38: conv= rgb32to15; break;

  1483.         case 0x43: conv= rgb15to16; break;

  1484.         case 0x46: conv= rgb24to16; break;

  1485.         case 0x48: conv= rgb32to16; break;

  1486.         case 0x63: conv= rgb15to24; break;

  1487.         case 0x64: conv= rgb16to24; break;

  1488.         case 0x68: conv= rgb32to24; break;

  1489.         case 0x83: conv= rgb15to32; break;

  1490.         case 0x84: conv= rgb16to32; break;

  1491.         case 0x86: conv= rgb24to32; break;

  1492.         }

  1493.     } else if (  (isBGRinInt(srcFormat) && isRGBinInt(dstFormat))

  1494.              || (isRGBinInt(srcFormat) && isBGRinInt(dstFormat))) {

  1495.         switch(srcId | (dstId<<4)) {

  1496.         case 0x33: conv= rgb15tobgr15; break;

  1497.         case 0x34: conv= rgb16tobgr15; break;

  1498.         case 0x36: conv= rgb24tobgr15; break;

  1499.         case 0x38: conv= rgb32tobgr15; break;

  1500.         case 0x43: conv= rgb15tobgr16; break;

  1501.         case 0x44: conv= rgb16tobgr16; break;

  1502.         case 0x46: conv= rgb24tobgr16; break;

  1503.         case 0x48: conv= rgb32tobgr16; break;

  1504.         case 0x63: conv= rgb15tobgr24; break;

  1505.         case 0x64: conv= rgb16tobgr24; break;

  1506.         case 0x66: conv= rgb24tobgr24; break;

  1507.         case 0x68: conv= rgb32tobgr24; break;

  1508.         case 0x83: conv= rgb15tobgr32; break;

  1509.         case 0x84: conv= rgb16tobgr32; break;

  1510.         case 0x86: conv= rgb24tobgr32; break;

  1511.         }

  1512.     }

  1513. 

  1514.     if (!conv) {

  1515.         av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",

  1516.                sws_format_name(srcFormat), sws_format_name(dstFormat));

  1517.     } else {

  1518.         const uint8_t *srcPtr= src[0];

  1519.               uint8_t *dstPtr= dst[0];

  1520.         if ((srcFormat == PIX_FMT_RGB32_1 || srcFormat == PIX_FMT_BGR32_1) && !isRGBA32(dstFormat))

  1521.             srcPtr += ALT32_CORR;

  1522. 

  1523.         if ((dstFormat == PIX_FMT_RGB32_1 || dstFormat == PIX_FMT_BGR32_1) && !isRGBA32(srcFormat))

  1524.             dstPtr += ALT32_CORR;

  1525. 

  1526.         if (dstStride[0]*srcBpp == srcStride[0]*dstBpp && srcStride[0] > 0)

  1527.             conv(srcPtr, dstPtr + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]);

  1528.         else {

  1529.             int i;

  1530.             dstPtr += dstStride[0]*srcSliceY;

  1531. 

  1532.             for (i=0; i<srcSliceH; i++) {

  1533.                 conv(srcPtr, dstPtr, c->srcW*srcBpp);

  1534.                 srcPtr+= srcStride[0];

  1535.                 dstPtr+= dstStride[0];

  1536.             }

  1537.         }

  1538.     }

  1539.     return srcSliceH;

  1540. }

  1541. 

  1542. static int bgr24ToYv12Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,

  1543.                               int srcSliceH, uint8_t* dst[], int dstStride[])

  1544. {

  1545.     rgb24toyv12(

  1546.         src[0],

  1547.         dst[0]+ srcSliceY    *dstStride[0],

  1548.         dst[1]+(srcSliceY>>1)*dstStride[1],

  1549.         dst[2]+(srcSliceY>>1)*dstStride[2],

  1550.         c->srcW, srcSliceH,

  1551.         dstStride[0], dstStride[1], srcStride[0]);

  1552.     if (dst[3])

  1553.         fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);

  1554.     return srcSliceH;

  1555. }

  1556. 

  1557. static int yvu9ToYv12Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,

  1558.                              int srcSliceH, uint8_t* dst[], int dstStride[])

  1559. {

  1560.     copyPlane(src[0], srcStride[0], srcSliceY, srcSliceH, c->srcW,

  1561.               dst[0], dstStride[0]);

  1562. 

  1563.     planar2x(src[1], dst[1] + dstStride[1]*(srcSliceY >> 1), c->chrSrcW,

  1564.              srcSliceH >> 2, srcStride[1], dstStride[1]);

  1565.     planar2x(src[2], dst[2] + dstStride[2]*(srcSliceY >> 1), c->chrSrcW,

  1566.              srcSliceH >> 2, srcStride[2], dstStride[2]);

  1567.     if (dst[3])

  1568.         fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);

  1569.     return srcSliceH;

  1570. }

  1571. 

  1572. /* unscaled copy like stuff (assumes nearly identical formats) */

  1573. static int packedCopyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,

  1574.                              int srcSliceH, uint8_t* dst[], int dstStride[])

  1575. {

  1576.     if (dstStride[0]==srcStride[0] && srcStride[0] > 0)

  1577.         memcpy(dst[0] + dstStride[0]*srcSliceY, src[0], srcSliceH*dstStride[0]);

  1578.     else {

  1579.         int i;

  1580.         const uint8_t *srcPtr= src[0];

  1581.         uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;

  1582.         int length=0;

  1583. 

  1584.         /* universal length finder */

  1585.         while(length+c->srcW <= FFABS(dstStride[0])

  1586.            && length+c->srcW <= FFABS(srcStride[0])) length+= c->srcW;

  1587.         assert(length!=0);

  1588. 

  1589.         for (i=0; i<srcSliceH; i++) {

  1590.             memcpy(dstPtr, srcPtr, length);

  1591.             srcPtr+= srcStride[0];

  1592.             dstPtr+= dstStride[0];

  1593.         }

  1594.     }

  1595.     return srcSliceH;

  1596. }

  1597. 

  1598. static int planarCopyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,

  1599.                              int srcSliceH, uint8_t* dst[], int dstStride[])

  1600. {

  1601.     int plane, i, j;

  1602.     for (plane=0; plane<4; plane++) {

  1603.         int length= (plane==0 || plane==3) ? c->srcW  : -((-c->srcW  )>>c->chrDstHSubSample);

  1604.         int y=      (plane==0 || plane==3) ? srcSliceY: -((-srcSliceY)>>c->chrDstVSubSample);

  1605.         int height= (plane==0 || plane==3) ? srcSliceH: -((-srcSliceH)>>c->chrDstVSubSample);

  1606.         const uint8_t *srcPtr= src[plane];

  1607.         uint8_t *dstPtr= dst[plane] + dstStride[plane]*y;

  1608. 

  1609.         if (!dst[plane]) continue;

  1610.         // ignore palette for GRAY8

  1611.         if (plane == 1 && !dst[2]) continue;

  1612.         if (!src[plane] || (plane == 1 && !src[2])) {

  1613.             if(is16BPS(c->dstFormat))

  1614.                 length*=2;

  1615.             fillPlane(dst[plane], dstStride[plane], length, height, y, (plane==3) ? 255 : 128);

  1616.         } else {

  1617.             if(is9_OR_10BPS(c->srcFormat)) {

  1618.                 const int src_depth = av_pix_fmt_descriptors[c->srcFormat].comp[plane].depth_minus1+1;

  1619.                 const int dst_depth = av_pix_fmt_descriptors[c->dstFormat].comp[plane].depth_minus1+1;

  1620.                 const uint16_t *srcPtr2 = (const uint16_t*)srcPtr;

  1621. 

  1622.                 if (is16BPS(c->dstFormat)) {

  1623.                     uint16_t *dstPtr2 = (uint16_t*)dstPtr;

  1624. #define COPY9_OR_10TO16(rfunc, wfunc) \

  1625.                     for (i = 0; i < height; i++) { \

  1626.                         for (j = 0; j < length; j++) { \

  1627.                             int srcpx = rfunc(&srcPtr2[j]); \

  1628.                             wfunc(&dstPtr2[j], (srcpx<<(16-src_depth)) | (srcpx>>(2*src_depth-16))); \

  1629.                         } \

  1630.                         dstPtr2 += dstStride[plane]/2; \

  1631.                         srcPtr2 += srcStride[plane]/2; \

  1632.                     }

  1633.                     if (isBE(c->dstFormat)) {

  1634.                         if (isBE(c->srcFormat)) {

  1635.                             COPY9_OR_10TO16(AV_RB16, AV_WB16);

  1636.                         } else {

  1637.                             COPY9_OR_10TO16(AV_RL16, AV_WB16);

  1638.                         }

  1639.                     } else {

  1640.                         if (isBE(c->srcFormat)) {

  1641.                             COPY9_OR_10TO16(AV_RB16, AV_WL16);

  1642.                         } else {

  1643.                             COPY9_OR_10TO16(AV_RL16, AV_WL16);

  1644.                         }

  1645.                     }

  1646.                 } else if (is9_OR_10BPS(c->dstFormat)) {

  1647.                     uint16_t *dstPtr2 = (uint16_t*)dstPtr;

  1648. #define COPY9_OR_10TO9_OR_10(loop) \

  1649.                     for (i = 0; i < height; i++) { \

  1650.                         for (j = 0; j < length; j++) { \

  1651.                             loop; \

  1652.                         } \

  1653.                         dstPtr2 += dstStride[plane]/2; \

  1654.                         srcPtr2 += srcStride[plane]/2; \

  1655.                     }

  1656. #define COPY9_OR_10TO9_OR_10_2(rfunc, wfunc) \

  1657.                     if (dst_depth > src_depth) { \

  1658.                         COPY9_OR_10TO9_OR_10(int srcpx = rfunc(&srcPtr2[j]); \

  1659.                             wfunc(&dstPtr2[j], (srcpx << 1) | (srcpx >> 9))); \

  1660.                     } else if (dst_depth < src_depth) { \

  1661.                         COPY9_OR_10TO9_OR_10(wfunc(&dstPtr2[j], rfunc(&srcPtr2[j]) >> 1)); \

  1662.                     } else { \

  1663.                         COPY9_OR_10TO9_OR_10(wfunc(&dstPtr2[j], rfunc(&srcPtr2[j]))); \

  1664.                     }

  1665.                     if (isBE(c->dstFormat)) {

  1666.                         if (isBE(c->srcFormat)) {

  1667.                             COPY9_OR_10TO9_OR_10_2(AV_RB16, AV_WB16);

  1668.                         } else {

  1669.                             COPY9_OR_10TO9_OR_10_2(AV_RL16, AV_WB16);

  1670.                         }

  1671.                     } else {

  1672.                         if (isBE(c->srcFormat)) {

  1673.                             COPY9_OR_10TO9_OR_10_2(AV_RB16, AV_WL16);

  1674.                         } else {

  1675.                             COPY9_OR_10TO9_OR_10_2(AV_RL16, AV_WL16);

  1676.                         }

  1677.                     }

  1678.                 } else {

  1679.                     // FIXME Maybe dither instead.

  1680. #define COPY9_OR_10TO8(rfunc) \

  1681.                     for (i = 0; i < height; i++) { \

  1682.                         for (j = 0; j < length; j++) { \

  1683.                             dstPtr[j] = rfunc(&srcPtr2[j])>>(src_depth-8); \

  1684.                         } \

  1685.                         dstPtr  += dstStride[plane]; \

  1686.                         srcPtr2 += srcStride[plane]/2; \

  1687.                     }

  1688.                     if (isBE(c->srcFormat)) {

  1689.                         COPY9_OR_10TO8(AV_RB16);

  1690.                     } else {

  1691.                         COPY9_OR_10TO8(AV_RL16);

  1692.                     }

  1693.                 }

  1694.             } else if(is9_OR_10BPS(c->dstFormat)) {

  1695.                 const int dst_depth = av_pix_fmt_descriptors[c->dstFormat].comp[plane].depth_minus1+1;

  1696.                 uint16_t *dstPtr2 = (uint16_t*)dstPtr;

  1697. 

  1698.                 if (is16BPS(c->srcFormat)) {

  1699.                     const uint16_t *srcPtr2 = (const uint16_t*)srcPtr;

  1700. #define COPY16TO9_OR_10(rfunc, wfunc) \

  1701.                     for (i = 0; i < height; i++) { \

  1702.                         for (j = 0; j < length; j++) { \

  1703.                             wfunc(&dstPtr2[j], rfunc(&srcPtr2[j])>>(16-dst_depth)); \

  1704.                         } \

  1705.                         dstPtr2 += dstStride[plane]/2; \

  1706.                         srcPtr2 += srcStride[plane]/2; \

  1707.                     }

  1708.                     if (isBE(c->dstFormat)) {

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

  1710.                             COPY16TO9_OR_10(AV_RB16, AV_WB16);

  1711.                         } else {

  1712.                             COPY16TO9_OR_10(AV_RL16, AV_WB16);

  1713.                         }

  1714.                     } else {

  1715.                         if (isBE(c->srcFormat)) {

  1716.                             COPY16TO9_OR_10(AV_RB16, AV_WL16);

  1717.                         } else {

  1718.                             COPY16TO9_OR_10(AV_RL16, AV_WL16);

  1719.                         }

  1720.                     }

  1721.                 } else /* 8bit */ {

  1722. #define COPY8TO9_OR_10(wfunc) \

  1723.                     for (i = 0; i < height; i++) { \

  1724.                         for (j = 0; j < length; j++) { \

  1725.                             const int srcpx = srcPtr[j]; \

  1726.                             wfunc(&dstPtr2[j], (srcpx<<(dst_depth-8)) | (srcpx >> (16-dst_depth))); \

  1727.                         } \

  1728.                         dstPtr2 += dstStride[plane]/2; \

  1729.                         srcPtr  += srcStride[plane]; \

  1730.                     }

  1731.                     if (isBE(c->dstFormat)) {

  1732.                         COPY8TO9_OR_10(AV_WB16);

  1733.                     } else {

  1734.                         COPY8TO9_OR_10(AV_WL16);

  1735.                     }

  1736.                 }

  1737.             } else if(is16BPS(c->srcFormat) && !is16BPS(c->dstFormat)) {

  1738.                 if (!isBE(c->srcFormat)) srcPtr++;

  1739.                 for (i=0; i<height; i++) {

  1740.                     for (j=0; j<length; j++) dstPtr[j] = srcPtr[j<<1];

  1741.                     srcPtr+= srcStride[plane];

  1742.                     dstPtr+= dstStride[plane];

  1743.                 }

  1744.             } else if(!is16BPS(c->srcFormat) && is16BPS(c->dstFormat)) {

  1745.                 for (i=0; i<height; i++) {

  1746.                     for (j=0; j<length; j++) {

  1747.                         dstPtr[ j<<1   ] = srcPtr[j];

  1748.                         dstPtr[(j<<1)+1] = srcPtr[j];

  1749.                     }

  1750.                     srcPtr+= srcStride[plane];

  1751.                     dstPtr+= dstStride[plane];

  1752.                 }

  1753.             } else if(is16BPS(c->srcFormat) && is16BPS(c->dstFormat)

  1754.                   && isBE(c->srcFormat) != isBE(c->dstFormat)) {

  1755. 

  1756.                 for (i=0; i<height; i++) {

  1757.                     for (j=0; j<length; j++)

  1758.                         ((uint16_t*)dstPtr)[j] = av_bswap16(((const uint16_t*)srcPtr)[j]);

  1759.                     srcPtr+= srcStride[plane];

  1760.                     dstPtr+= dstStride[plane];

  1761.                 }

  1762.             } else if (dstStride[plane] == srcStride[plane] &&

  1763.                        srcStride[plane] > 0 && srcStride[plane] == length) {

  1764.                 memcpy(dst[plane] + dstStride[plane]*y, src[plane],

  1765.                        height*dstStride[plane]);

  1766.             } else {

  1767.                 if(is16BPS(c->srcFormat) && is16BPS(c->dstFormat))

  1768.                     length*=2;

  1769.                 for (i=0; i<height; i++) {

  1770.                     memcpy(dstPtr, srcPtr, length);

  1771.                     srcPtr+= srcStride[plane];

  1772.                     dstPtr+= dstStride[plane];

  1773.                 }

  1774.             }

  1775.         }

  1776.     }

  1777.     return srcSliceH;

  1778. }

  1779. 

  1780. void ff_get_unscaled_swscale(SwsContext *c)

  1781. {

  1782.     const enum PixelFormat srcFormat = c->srcFormat;

  1783.     const enum PixelFormat dstFormat = c->dstFormat;

  1784.     const int flags = c->flags;

  1785.     const int dstH = c->dstH;

  1786.     int needsDither;

  1787. 

  1788.     needsDither= isAnyRGB(dstFormat)

  1789.         &&  c->dstFormatBpp < 24

  1790.         && (c->dstFormatBpp < c->srcFormatBpp || (!isAnyRGB(srcFormat)));

  1791. 

  1792.     /* yv12_to_nv12 */

  1793.     if ((srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUVA420P) && (dstFormat == PIX_FMT_NV12 || dstFormat == PIX_FMT_NV21)) {

  1794.         c->swScale= planarToNv12Wrapper;

  1795.     }

  1796.     /* yuv2bgr */

  1797.     if ((srcFormat==PIX_FMT_YUV420P || srcFormat==PIX_FMT_YUV422P || srcFormat==PIX_FMT_YUVA420P) && isAnyRGB(dstFormat)

  1798.         && !(flags & SWS_ACCURATE_RND) && !(dstH&1)) {

  1799.         c->swScale= ff_yuv2rgb_get_func_ptr(c);

  1800.     }

  1801. 

  1802.     if (srcFormat==PIX_FMT_YUV410P && (dstFormat==PIX_FMT_YUV420P || dstFormat==PIX_FMT_YUVA420P) && !(flags & SWS_BITEXACT)) {

  1803.         c->swScale= yvu9ToYv12Wrapper;

  1804.     }

  1805. 

  1806.     /* bgr24toYV12 */

  1807.     if (srcFormat==PIX_FMT_BGR24 && (dstFormat==PIX_FMT_YUV420P || dstFormat==PIX_FMT_YUVA420P) && !(flags & SWS_ACCURATE_RND))

  1808.         c->swScale= bgr24ToYv12Wrapper;

  1809. 

  1810.     /* RGB/BGR -> RGB/BGR (no dither needed forms) */

  1811.     if (   isAnyRGB(srcFormat)

  1812.         && isAnyRGB(dstFormat)

  1813.         && srcFormat != PIX_FMT_BGR8      && dstFormat != PIX_FMT_BGR8

  1814.         && srcFormat != PIX_FMT_RGB8      && dstFormat != PIX_FMT_RGB8

  1815.         && srcFormat != PIX_FMT_BGR4      && dstFormat != PIX_FMT_BGR4

  1816.         && srcFormat != PIX_FMT_RGB4      && dstFormat != PIX_FMT_RGB4

  1817.         && srcFormat != PIX_FMT_BGR4_BYTE && dstFormat != PIX_FMT_BGR4_BYTE

  1818.         && srcFormat != PIX_FMT_RGB4_BYTE && dstFormat != PIX_FMT_RGB4_BYTE

  1819.         && srcFormat != PIX_FMT_MONOBLACK && dstFormat != PIX_FMT_MONOBLACK

  1820.         && srcFormat != PIX_FMT_MONOWHITE && dstFormat != PIX_FMT_MONOWHITE

  1821.         && srcFormat != PIX_FMT_RGB48LE   && dstFormat != PIX_FMT_RGB48LE

  1822.         && srcFormat != PIX_FMT_RGB48BE   && dstFormat != PIX_FMT_RGB48BE

  1823.         && srcFormat != PIX_FMT_BGR48LE   && dstFormat != PIX_FMT_BGR48LE

  1824.         && srcFormat != PIX_FMT_BGR48BE   && dstFormat != PIX_FMT_BGR48BE

  1825.         && (!needsDither || (c->flags&(SWS_FAST_BILINEAR|SWS_POINT))))

  1826.         c->swScale= rgbToRgbWrapper;

  1827. 

  1828.     if ((usePal(srcFormat) && (

  1829.         dstFormat == PIX_FMT_RGB32   ||

  1830.         dstFormat == PIX_FMT_RGB32_1 ||

  1831.         dstFormat == PIX_FMT_RGB24   ||

  1832.         dstFormat == PIX_FMT_BGR32   ||

  1833.         dstFormat == PIX_FMT_BGR32_1 ||

  1834.         dstFormat == PIX_FMT_BGR24)))

  1835.         c->swScale= palToRgbWrapper;

  1836. 

  1837.     if (srcFormat == PIX_FMT_YUV422P) {

  1838.         if (dstFormat == PIX_FMT_YUYV422)

  1839.             c->swScale= yuv422pToYuy2Wrapper;

  1840.         else if (dstFormat == PIX_FMT_UYVY422)

  1841.             c->swScale= yuv422pToUyvyWrapper;

  1842.     }

  1843. 

  1844.     /* LQ converters if -sws 0 or -sws 4*/

  1845.     if (c->flags&(SWS_FAST_BILINEAR|SWS_POINT)) {

  1846.         /* yv12_to_yuy2 */

  1847.         if (srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUVA420P) {

  1848.             if (dstFormat == PIX_FMT_YUYV422)

  1849.                 c->swScale= planarToYuy2Wrapper;

  1850.             else if (dstFormat == PIX_FMT_UYVY422)

  1851.                 c->swScale= planarToUyvyWrapper;

  1852.         }

  1853.     }

  1854.     if(srcFormat == PIX_FMT_YUYV422 && (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P))

  1855.         c->swScale= yuyvToYuv420Wrapper;

  1856.     if(srcFormat == PIX_FMT_UYVY422 && (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P))

  1857.         c->swScale= uyvyToYuv420Wrapper;

  1858.     if(srcFormat == PIX_FMT_YUYV422 && dstFormat == PIX_FMT_YUV422P)

  1859.         c->swScale= yuyvToYuv422Wrapper;

  1860.     if(srcFormat == PIX_FMT_UYVY422 && dstFormat == PIX_FMT_YUV422P)

  1861.         c->swScale= uyvyToYuv422Wrapper;

  1862. 

  1863. #if HAVE_ALTIVEC

  1864.     if ((av_get_cpu_flags() & AV_CPU_FLAG_ALTIVEC) &&

  1865.         !(c->flags & SWS_BITEXACT) &&

  1866.         srcFormat == PIX_FMT_YUV420P) {

  1867.         // unscaled YV12 -> packed YUV, we want speed

  1868.         if (dstFormat == PIX_FMT_YUYV422)

  1869.             c->swScale= yv12toyuy2_unscaled_altivec;

  1870.         else if (dstFormat == PIX_FMT_UYVY422)

  1871.             c->swScale= yv12touyvy_unscaled_altivec;

  1872.     }

  1873. #endif

  1874. 

  1875.     /* simple copy */

  1876.     if (  srcFormat == dstFormat

  1877.         || (srcFormat == PIX_FMT_YUVA420P && dstFormat == PIX_FMT_YUV420P)

  1878.         || (srcFormat == PIX_FMT_YUV420P && dstFormat == PIX_FMT_YUVA420P)

  1879.         || (isPlanarYUV(srcFormat) && isGray(dstFormat))

  1880.         || (isPlanarYUV(dstFormat) && isGray(srcFormat))

  1881.         || (isGray(dstFormat) && isGray(srcFormat))

  1882.         || (isPlanarYUV(srcFormat) && isPlanarYUV(dstFormat)

  1883.             && c->chrDstHSubSample == c->chrSrcHSubSample

  1884.             && c->chrDstVSubSample == c->chrSrcVSubSample

  1885.             && dstFormat != PIX_FMT_NV12 && dstFormat != PIX_FMT_NV21

  1886.             && srcFormat != PIX_FMT_NV12 && srcFormat != PIX_FMT_NV21))

  1887.     {

  1888.         if (isPacked(c->srcFormat))

  1889.             c->swScale= packedCopyWrapper;

  1890.         else /* Planar YUV or gray */

  1891.             c->swScale= planarCopyWrapper;

  1892.     }

  1893. #if ARCH_BFIN

  1894.     ff_bfin_get_unscaled_swscale (c);

  1895. #endif

  1896. }

  1897. 

  1898. static void reset_ptr(const uint8_t* src[], int format)

  1899. {

  1900.     if(!isALPHA(format))

  1901.         src[3]=NULL;

  1902.     if(!isPlanarYUV(format)) {

  1903.         src[3]=src[2]=NULL;

  1904. 

  1905.         if (!usePal(format))

  1906.             src[1]= NULL;

  1907.     }

  1908. }

  1909. 

  1910. static int check_image_pointers(uint8_t *data[4], enum PixelFormat pix_fmt,

  1911.                                 const int linesizes[4])

  1912. {

  1913.     const AVPixFmtDescriptor *desc = &av_pix_fmt_descriptors[pix_fmt];

  1914.     int i;

  1915. 

  1916.     for (i = 0; i < 4; i++) {

  1917.         int plane = desc->comp[i].plane;

  1918.         if (!data[plane] || !linesizes[plane])

  1919.             return 0;

  1920.     }

  1921. 

  1922.     return 1;

  1923. }

  1924. 

  1925. /**

  1926.  * swscale wrapper, so we don't need to export the SwsContext.

  1927.  * Assumes planar YUV to be in YUV order instead of YVU.

  1928.  */

  1929. int sws_scale(SwsContext *c, const uint8_t* const src[], const int srcStride[], int srcSliceY,

  1930.               int srcSliceH, uint8_t* const dst[], const int dstStride[])

  1931. {

  1932.     int i;

  1933.     const uint8_t* src2[4]= {src[0], src[1], src[2], src[3]};

  1934.     uint8_t* dst2[4]= {dst[0], dst[1], dst[2], dst[3]};

  1935. 

  1936.     // do not mess up sliceDir if we have a "trailing" 0-size slice

  1937.     if (srcSliceH == 0)

  1938.         return 0;

  1939. 

  1940.     if (!check_image_pointers(src, c->srcFormat, srcStride)) {

  1941.         av_log(c, AV_LOG_ERROR, "bad src image pointers\n");

  1942.         return 0;

  1943.     }

  1944.     if (!check_image_pointers(dst, c->dstFormat, dstStride)) {

  1945.         av_log(c, AV_LOG_ERROR, "bad dst image pointers\n");

  1946.         return 0;

  1947.     }

  1948. 

  1949.     if (c->sliceDir == 0 && srcSliceY != 0 && srcSliceY + srcSliceH != c->srcH) {

  1950.         av_log(c, AV_LOG_ERROR, "Slices start in the middle!\n");

  1951.         return 0;

  1952.     }

  1953.     if (c->sliceDir == 0) {

  1954.         if (srcSliceY == 0) c->sliceDir = 1; else c->sliceDir = -1;

  1955.     }

  1956. 

  1957.     if (usePal(c->srcFormat)) {

  1958.         for (i=0; i<256; i++) {

  1959.             int p, r, g, b,y,u,v;

  1960.             if(c->srcFormat == PIX_FMT_PAL8) {

  1961.                 p=((const uint32_t*)(src[1]))[i];

  1962.                 r= (p>>16)&0xFF;

  1963.                 g= (p>> 8)&0xFF;

  1964.                 b=  p     &0xFF;

  1965.             } else if(c->srcFormat == PIX_FMT_RGB8) {

  1966.                 r= (i>>5    )*36;

  1967.                 g= ((i>>2)&7)*36;

  1968.                 b= (i&3     )*85;

  1969.             } else if(c->srcFormat == PIX_FMT_BGR8) {

  1970.                 b= (i>>6    )*85;

  1971.                 g= ((i>>3)&7)*36;

  1972.                 r= (i&7     )*36;

  1973.             } else if(c->srcFormat == PIX_FMT_RGB4_BYTE) {

  1974.                 r= (i>>3    )*255;

  1975.                 g= ((i>>1)&3)*85;

  1976.                 b= (i&1     )*255;

  1977.             } else if(c->srcFormat == PIX_FMT_GRAY8 || c->srcFormat == PIX_FMT_Y400A) {

  1978.                 r = g = b = i;

  1979.             } else {

  1980.                 assert(c->srcFormat == PIX_FMT_BGR4_BYTE);

  1981.                 b= (i>>3    )*255;

  1982.                 g= ((i>>1)&3)*85;

  1983.                 r= (i&1     )*255;

  1984.             }

  1985.             y= av_clip_uint8((RY*r + GY*g + BY*b + ( 33<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);

  1986.             u= av_clip_uint8((RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);

  1987.             v= av_clip_uint8((RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);

  1988.             c->pal_yuv[i]= y + (u<<8) + (v<<16);

  1989. 

  1990.             switch(c->dstFormat) {

  1991.             case PIX_FMT_BGR32:

  1992. #if !HAVE_BIGENDIAN

  1993.             case PIX_FMT_RGB24:

  1994. #endif

  1995.                 c->pal_rgb[i]=  r + (g<<8) + (b<<16);

  1996.                 break;

  1997.             case PIX_FMT_BGR32_1:

  1998. #if HAVE_BIGENDIAN

  1999.             case PIX_FMT_BGR24:

  2000. #endif

  2001.                 c->pal_rgb[i]= (r + (g<<8) + (b<<16)) << 8;

  2002.                 break;

  2003.             case PIX_FMT_RGB32_1:

  2004. #if HAVE_BIGENDIAN

  2005.             case PIX_FMT_RGB24:

  2006. #endif

  2007.                 c->pal_rgb[i]= (b + (g<<8) + (r<<16)) << 8;

  2008.                 break;

  2009.             case PIX_FMT_RGB32:

  2010. #if !HAVE_BIGENDIAN

  2011.             case PIX_FMT_BGR24:

  2012. #endif

  2013.             default:

  2014.                 c->pal_rgb[i]=  b + (g<<8) + (r<<16);

  2015.             }

  2016.         }

  2017.     }

  2018. 

  2019.     // copy strides, so they can safely be modified

  2020.     if (c->sliceDir == 1) {

  2021.         // slices go from top to bottom

  2022.         int srcStride2[4]= {srcStride[0], srcStride[1], srcStride[2], srcStride[3]};

  2023.         int dstStride2[4]= {dstStride[0], dstStride[1], dstStride[2], dstStride[3]};

  2024. 

  2025.         reset_ptr(src2, c->srcFormat);

  2026.         reset_ptr((const uint8_t**)dst2, c->dstFormat);

  2027. 

  2028.         /* reset slice direction at end of frame */

  2029.         if (srcSliceY + srcSliceH == c->srcH)

  2030.             c->sliceDir = 0;

  2031. 

  2032.         return c->swScale(c, src2, srcStride2, srcSliceY, srcSliceH, dst2, dstStride2);

  2033.     } else {

  2034.         // slices go from bottom to top => we flip the image internally

  2035.         int srcStride2[4]= {-srcStride[0], -srcStride[1], -srcStride[2], -srcStride[3]};

  2036.         int dstStride2[4]= {-dstStride[0], -dstStride[1], -dstStride[2], -dstStride[3]};

  2037. 

  2038.         src2[0] += (srcSliceH-1)*srcStride[0];

  2039.         if (!usePal(c->srcFormat))

  2040.             src2[1] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[1];

  2041.         src2[2] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[2];

  2042.         src2[3] += (srcSliceH-1)*srcStride[3];

  2043.         dst2[0] += ( c->dstH                      -1)*dstStride[0];

  2044.         dst2[1] += ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[1];

  2045.         dst2[2] += ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[2];

  2046.         dst2[3] += ( c->dstH                      -1)*dstStride[3];

  2047. 

  2048.         reset_ptr(src2, c->srcFormat);

  2049.         reset_ptr((const uint8_t**)dst2, c->dstFormat);

  2050. 

  2051.         /* reset slice direction at end of frame */

  2052.         if (!srcSliceY)

  2053.             c->sliceDir = 0;

  2054. 

  2055.         return c->swScale(c, src2, srcStride2, c->srcH-srcSliceY-srcSliceH, srcSliceH, dst2, dstStride2);

  2056.     }

  2057. }

  2058. 

  2059. /* Convert the palette to the same packed 32-bit format as the palette */

  2060. void sws_convertPalette8ToPacked32(const uint8_t *src, uint8_t *dst, int num_pixels, const uint8_t *palette)

  2061. {

  2062.     int i;

  2063. 

  2064.     for (i=0; i<num_pixels; i++)

  2065.         ((uint32_t *) dst)[i] = ((const uint32_t *) palette)[src[i]];

  2066. }

  2067. 

  2068. /* Palette format: ABCD -> dst format: ABC */

  2069. void sws_convertPalette8ToPacked24(const uint8_t *src, uint8_t *dst, int num_pixels, const uint8_t *palette)

  2070. {

  2071.     int i;

  2072. 

  2073.     for (i=0; i<num_pixels; i++) {

  2074.         //FIXME slow?

  2075.         dst[0]= palette[src[i]*4+0];

  2076.         dst[1]= palette[src[i]*4+1];

  2077.         dst[2]= palette[src[i]*4+2];

  2078.         dst+= 3;

  2079.     }

  2080. }