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.
28535 Commits
32 Branches
944MB
Tree: 7e19a6e868
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from '7e19a6e868'
${ noResults }
FFmpeg/libswscale/swscale.c

2164 lines
77KB
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/avutil.h"

  66. #include "libavutil/mathematics.h"

  67. #include "libavutil/bswap.h"

  68. #include "libavutil/pixdesc.h"

  69. 

  70. #undef MOVNTQ

  71. #undef PAVGB

  72. 

  73. //#undef HAVE_MMX2

  74. //#define HAVE_AMD3DNOW

  75. //#undef HAVE_MMX

  76. //#undef ARCH_X86

  77. #define DITHER1XBPP

  78. 

  79. #define isPacked(x)         (       \

  80.            (x)==PIX_FMT_PAL8        \

  81.         || (x)==PIX_FMT_YUYV422     \

  82.         || (x)==PIX_FMT_UYVY422     \

  83.         || (x)==PIX_FMT_Y400A       \

  84.         || isAnyRGB(x)              \

  85.     )

  86. 

  87. #define RGB2YUV_SHIFT 15

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

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

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

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

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

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

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

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

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

  97. 

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

  99.     {0.7152, 0.0722, 0.2126, -0.386, 0.5, -0.115, -0.454, -0.046, 0.5}, //ITU709

  100.     {0.7152, 0.0722, 0.2126, -0.386, 0.5, -0.115, -0.454, -0.046, 0.5}, //ITU709

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

  104.     {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5}, //DEFAULT / ITU601 / ITU624 / SMPTE 170M

  105.     {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5}, //DEFAULT / ITU601 / ITU624 / SMPTE 170M

  106.     {0.701 , 0.087 , 0.212 , -0.384, 0.5, -0.116, -0.445, -0.055, 0.5}, //SMPTE 240M

  107. };

  108. 

  109. /*

  110. NOTES

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

  112. 

  113. TODO

  114. more intelligent misalignment avoidance for the horizontal scaler

  115. write special vertical cubic upscale version

  116. optimize C code (YV12 / minmax)

  117. add support for packed pixel YUV input & output

  118. add support for Y8 output

  119. optimize BGR24 & BGR32

  120. add BGR4 output support

  121. write special BGR->BGR scaler

  122. */

  123. 

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

  125. {  1,   3,   1,   3,   1,   3,   1,   3, },

  126. {  2,   0,   2,   0,   2,   0,   2,   0, },

  127. };

  128. 

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

  130. {  6,   2,   6,   2,   6,   2,   6,   2, },

  131. {  0,   4,   0,   4,   0,   4,   0,   4, },

  132. };

  133. 

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

  135. {  8,   4,  11,   7,   8,   4,  11,   7, },

  136. {  2,  14,   1,  13,   2,  14,   1,  13, },

  137. { 10,   6,   9,   5,  10,   6,   9,   5, },

  138. {  0,  12,   3,  15,   0,  12,   3,  15, },

  139. };

  140. 

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

  142. { 17,   9,  23,  15,  16,   8,  22,  14, },

  143. {  5,  29,   3,  27,   4,  28,   2,  26, },

  144. { 21,  13,  19,  11,  20,  12,  18,  10, },

  145. {  0,  24,   6,  30,   1,  25,   7,  31, },

  146. { 16,   8,  22,  14,  17,   9,  23,  15, },

  147. {  4,  28,   2,  26,   5,  29,   3,  27, },

  148. { 20,  12,  18,  10,  21,  13,  19,  11, },

  149. {  1,  25,   7,  31,   0,  24,   6,  30, },

  150. };

  151. 

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

  153. {  0,  55,  14,  68,   3,  58,  17,  72, },

  154. { 37,  18,  50,  32,  40,  22,  54,  35, },

  155. {  9,  64,   5,  59,  13,  67,   8,  63, },

  156. { 46,  27,  41,  23,  49,  31,  44,  26, },

  157. {  2,  57,  16,  71,   1,  56,  15,  70, },

  158. { 39,  21,  52,  34,  38,  19,  51,  33, },

  159. { 11,  66,   7,  62,  10,  65,   6,  60, },

  160. { 48,  30,  43,  25,  47,  29,  42,  24, },

  161. };

  162. 

  163. #if 1

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

  165. {117,  62, 158, 103, 113,  58, 155, 100, },

  166. { 34, 199,  21, 186,  31, 196,  17, 182, },

  167. {144,  89, 131,  76, 141,  86, 127,  72, },

  168. {  0, 165,  41, 206,  10, 175,  52, 217, },

  169. {110,  55, 151,  96, 120,  65, 162, 107, },

  170. { 28, 193,  14, 179,  38, 203,  24, 189, },

  171. {138,  83, 124,  69, 148,  93, 134,  79, },

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

  173. };

  174. #elif 1

  175. // tries to correct a gamma of 1.5

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

  177. {  0, 143,  18, 200,   2, 156,  25, 215, },

  178. { 78,  28, 125,  64,  89,  36, 138,  74, },

  179. { 10, 180,   3, 161,  16, 195,   8, 175, },

  180. {109,  51,  93,  38, 121,  60, 105,  47, },

  181. {  1, 152,  23, 210,   0, 147,  20, 205, },

  182. { 85,  33, 134,  71,  81,  30, 130,  67, },

  183. { 14, 190,   6, 171,  12, 185,   5, 166, },

  184. {117,  57, 101,  44, 113,  54,  97,  41, },

  185. };

  186. #elif 1

  187. // tries to correct a gamma of 2.0

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

  189. {  0, 124,   8, 193,   0, 140,  12, 213, },

  190. { 55,  14, 104,  42,  66,  19, 119,  52, },

  191. {  3, 168,   1, 145,   6, 187,   3, 162, },

  192. { 86,  31,  70,  21,  99,  39,  82,  28, },

  193. {  0, 134,  11, 206,   0, 129,   9, 200, },

  194. { 62,  17, 114,  48,  58,  16, 109,  45, },

  195. {  5, 181,   2, 157,   4, 175,   1, 151, },

  196. { 95,  36,  78,  26,  90,  34,  74,  24, },

  197. };

  198. #else

  199. // tries to correct a gamma of 2.5

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

  201. {  0, 107,   3, 187,   0, 125,   6, 212, },

  202. { 39,   7,  86,  28,  49,  11, 102,  36, },

  203. {  1, 158,   0, 131,   3, 180,   1, 151, },

  204. { 68,  19,  52,  12,  81,  25,  64,  17, },

  205. {  0, 119,   5, 203,   0, 113,   4, 195, },

  206. { 45,   9,  96,  33,  42,   8,  91,  30, },

  207. {  2, 172,   1, 144,   2, 165,   0, 137, },

  208. { 77,  23,  60,  15,  72,  21,  56,  14, },

  209. };

  210. #endif

  211. 

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

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

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

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

  216. {

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

  218.     int i;

  219.     int shift = 11 + 16 - output_bits;

  220. 

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

  222.         int val = 1 << 10;

  223.         int j;

  224. 

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

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

  227. 

  228.         if (big_endian) {

  229.             AV_WB16(&dest[i], av_clip_uint16(val >> shift));

  230.         } else {

  231.             AV_WL16(&dest[i], av_clip_uint16(val >> shift));

  232.         }

  233.     }

  234. 

  235.     if (uDest) {

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

  237.             int u = 1 << 10;

  238.             int v = 1 << 10;

  239.             int j;

  240. 

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

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

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

  244.             }

  245. 

  246.             if (big_endian) {

  247.                 AV_WB16(&uDest[i], av_clip_uint16(u >> shift));

  248.                 AV_WB16(&vDest[i], av_clip_uint16(v >> shift));

  249.             } else {

  250.                 AV_WL16(&uDest[i], av_clip_uint16(u >> shift));

  251.                 AV_WL16(&vDest[i], av_clip_uint16(v >> shift));

  252.             }

  253.         }

  254.     }

  255. 

  256.     if (CONFIG_SWSCALE_ALPHA && aDest) {

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

  258.             int val = 1 << 10;

  259.             int j;

  260. 

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

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

  263. 

  264.             if (big_endian) {

  265.                 AV_WB16(&aDest[i], av_clip_uint16(val >> shift));

  266.             } else {

  267.                 AV_WL16(&aDest[i], av_clip_uint16(val >> shift));

  268.             }

  269.         }

  270.     }

  271. }

  272. 

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

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

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

  276.                                  enum PixelFormat dstFormat)

  277. {

  278. #define conv16(bits) \

  279.     if (isBE(dstFormat)) { \

  280.         yuv2yuvX16inC_template(lumFilter, lumSrc, lumFilterSize, \

  281.                                chrFilter, chrSrc, chrFilterSize, \

  282.                                alpSrc, \

  283.                                dest, uDest, vDest, aDest, \

  284.                                dstW, chrDstW, 1, bits); \

  285.     } else { \

  286.         yuv2yuvX16inC_template(lumFilter, lumSrc, lumFilterSize, \

  287.                                chrFilter, chrSrc, chrFilterSize, \

  288.                                alpSrc, \

  289.                                dest, uDest, vDest, aDest, \

  290.                                dstW, chrDstW, 0, bits); \

  291.     }

  292.     if (is16BPS(dstFormat)) {

  293.         conv16(16);

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

  295.         conv16(9);

  296.     } else {

  297.         conv16(10);

  298.     }

  299. #undef conv16

  300. }

  301. 

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

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

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

  305. {

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

  307.     int i;

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

  309.         int val=1<<18;

  310.         int j;

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

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

  313. 

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

  315.     }

  316. 

  317.     if (uDest)

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

  319.             int u=1<<18;

  320.             int v=1<<18;

  321.             int j;

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

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

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

  325.             }

  326. 

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

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

  329.         }

  330. 

  331.     if (CONFIG_SWSCALE_ALPHA && aDest)

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

  333.             int val=1<<18;

  334.             int j;

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

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

  337. 

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

  339.         }

  340. 

  341. }

  342. 

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

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

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

  346. {

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

  348.     int i;

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

  350.         int val=1<<18;

  351.         int j;

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

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

  354. 

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

  356.     }

  357. 

  358.     if (!uDest)

  359.         return;

  360. 

  361.     if (dstFormat == PIX_FMT_NV12)

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

  363.             int u=1<<18;

  364.             int v=1<<18;

  365.             int j;

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

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

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

  369.             }

  370. 

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

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

  373.         }

  374.     else

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

  376.             int u=1<<18;

  377.             int v=1<<18;

  378.             int j;

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

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

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

  382.             }

  383. 

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

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

  386.         }

  387. }

  388. 

  389. #define YSCALE_YUV_2_PACKEDX_NOCLIP_C(type,alpha) \

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

  391.         int j;\

  392.         int Y1 = 1<<18;\

  393.         int Y2 = 1<<18;\

  394.         int U  = 1<<18;\

  395.         int V  = 1<<18;\

  396.         int av_unused A1, A2;\

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

  398.         const int i2= 2*i;\

  399.         \

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

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

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

  403.         }\

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

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

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

  407.         }\

  408.         Y1>>=19;\

  409.         Y2>>=19;\

  410.         U >>=19;\

  411.         V >>=19;\

  412.         if (alpha) {\

  413.             A1 = 1<<18;\

  414.             A2 = 1<<18;\

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

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

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

  418.             }\

  419.             A1>>=19;\

  420.             A2>>=19;\

  421.         }

  422. 

  423. #define YSCALE_YUV_2_PACKEDX_C(type,alpha) \

  424.         YSCALE_YUV_2_PACKEDX_NOCLIP_C(type,alpha)\

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

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

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

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

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

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

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

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

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

  434.         }\

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

  436.             A1=av_clip_uint8(A1);\

  437.             A2=av_clip_uint8(A2);\

  438.         }

  439. 

  440. #define YSCALE_YUV_2_PACKEDX_FULL_C(rnd,alpha) \

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

  442.         int j;\

  443.         int Y = 0;\

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

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

  446.         int av_unused A;\

  447.         int R,G,B;\

  448.         \

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

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

  451.         }\

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

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

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

  455.         }\

  456.         Y >>=10;\

  457.         U >>=10;\

  458.         V >>=10;\

  459.         if (alpha) {\

  460.             A = rnd;\

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

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

  463.             A >>=19;\

  464.             if (A&256)\

  465.                 A = av_clip_uint8(A);\

  466.         }

  467. 

  468. #define YSCALE_YUV_2_RGBX_FULL_C(rnd,alpha) \

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

  470.         Y-= c->yuv2rgb_y_offset;\

  471.         Y*= c->yuv2rgb_y_coeff;\

  472.         Y+= rnd;\

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

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

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

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

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

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

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

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

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

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

  483.         }

  484. 

  485. #define YSCALE_YUV_2_GRAY16_C \

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

  487.         int j;\

  488.         int Y1 = 1<<18;\

  489.         int Y2 = 1<<18;\

  490.         int U  = 1<<18;\

  491.         int V  = 1<<18;\

  492.         \

  493.         const int i2= 2*i;\

  494.         \

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

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

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

  498.         }\

  499.         Y1>>=11;\

  500.         Y2>>=11;\

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

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

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

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

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

  506.         }

  507. 

  508. #define YSCALE_YUV_2_RGBX_C(type,alpha) \

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

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

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

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

  513. 

  514. #define YSCALE_YUV_2_PACKED2_C(type,alpha)   \

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

  516.         const int i2= 2*i;       \

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

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

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

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

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

  522.         int av_unused A1, A2;                                         \

  523.         if (alpha) {\

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

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

  526.         }

  527. 

  528. #define YSCALE_YUV_2_GRAY16_2_C   \

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

  530.         const int i2= 2*i;       \

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

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

  533. 

  534. #define YSCALE_YUV_2_RGB2_C(type,alpha) \

  535.     YSCALE_YUV_2_PACKED2_C(type,alpha)\

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

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

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

  539. 

  540. #define YSCALE_YUV_2_PACKED1_C(type,alpha) \

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

  542.         const int i2= 2*i;\

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

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

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

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

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

  548.         int av_unused A1, A2;\

  549.         if (alpha) {\

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

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

  552.         }

  553. 

  554. #define YSCALE_YUV_2_GRAY16_1_C \

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

  556.         const int i2= 2*i;\

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

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

  559. 

  560. #define YSCALE_YUV_2_RGB1_C(type,alpha) \

  561.     YSCALE_YUV_2_PACKED1_C(type,alpha)\

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

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

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

  565. 

  566. #define YSCALE_YUV_2_PACKED1B_C(type,alpha) \

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

  568.         const int i2= 2*i;\

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

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

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

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

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

  574.         int av_unused A1, A2;\

  575.         if (alpha) {\

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

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

  578.         }

  579. 

  580. #define YSCALE_YUV_2_RGB1B_C(type,alpha) \

  581.     YSCALE_YUV_2_PACKED1B_C(type,alpha)\

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

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

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

  585. 

  586. #define YSCALE_YUV_2_MONO2_C \

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

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

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

  590.         int acc;\

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

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

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

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

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

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

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

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

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

  600.         dest++;\

  601.     }

  602. 

  603. #define YSCALE_YUV_2_MONOX_C \

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

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

  606.     int acc=0;\

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

  608.         int j;\

  609.         int Y1=1<<18;\

  610.         int Y2=1<<18;\

  611. \

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

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

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

  615.         }\

  616.         Y1>>=19;\

  617.         Y2>>=19;\

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

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

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

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

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

  623.         }\

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

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

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

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

  628.             dest++;\

  629.         }\

  630.     }

  631. 

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

  633.     switch(c->dstFormat) {\

  634.     case PIX_FMT_RGB48BE:\

  635.     case PIX_FMT_RGB48LE:\

  636.         func(uint8_t,0)\

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

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

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

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

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

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

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

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

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

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

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

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

  649.             dest+=12;\

  650.         }\

  651.         break;\

  652.     case PIX_FMT_BGR48BE:\

  653.     case PIX_FMT_BGR48LE:\

  654.         func(uint8_t,0)\

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

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

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

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

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

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

  661.             dest+=12;\

  662.         }\

  663.         break;\

  664.     case PIX_FMT_RGBA:\

  665.     case PIX_FMT_BGRA:\

  666.         if (CONFIG_SMALL) {\

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

  668.             func(uint32_t,needAlpha)\

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

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

  671.             }\

  672.         } else {\

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

  674.                 func(uint32_t,1)\

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

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

  677.                 }\

  678.             } else {\

  679.                 func(uint32_t,0)\

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

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

  682.                 }\

  683.             }\

  684.         }\

  685.         break;\

  686.     case PIX_FMT_ARGB:\

  687.     case PIX_FMT_ABGR:\

  688.         if (CONFIG_SMALL) {\

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

  690.             func(uint32_t,needAlpha)\

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

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

  693.             }\

  694.         } else {\

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

  696.                 func(uint32_t,1)\

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

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

  699.                 }\

  700.             } else {\

  701.                 func(uint32_t,0)\

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

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

  704.                 }\

  705.             }\

  706.         }                \

  707.         break;\

  708.     case PIX_FMT_RGB24:\

  709.         func(uint8_t,0)\

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

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

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

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

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

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

  716.             dest+=6;\

  717.         }\

  718.         break;\

  719.     case PIX_FMT_BGR24:\

  720.         func(uint8_t,0)\

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

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

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

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

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

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

  727.             dest+=6;\

  728.         }\

  729.         break;\

  730.     case PIX_FMT_RGB565BE:\

  731.     case PIX_FMT_RGB565LE:\

  732.     case PIX_FMT_BGR565BE:\

  733.     case PIX_FMT_BGR565LE:\

  734.         {\

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

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

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

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

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

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

  741.             func(uint16_t,0)\

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

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

  744.             }\

  745.         }\

  746.         break;\

  747.     case PIX_FMT_RGB555BE:\

  748.     case PIX_FMT_RGB555LE:\

  749.     case PIX_FMT_BGR555BE:\

  750.     case PIX_FMT_BGR555LE:\

  751.         {\

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

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

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

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

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

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

  758.             func(uint16_t,0)\

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

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

  761.             }\

  762.         }\

  763.         break;\

  764.     case PIX_FMT_RGB444BE:\

  765.     case PIX_FMT_RGB444LE:\

  766.     case PIX_FMT_BGR444BE:\

  767.     case PIX_FMT_BGR444LE:\

  768.         {\

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

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

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

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

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

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

  775.             func(uint16_t,0)\

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

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

  778.             }\

  779.         }\

  780.         break;\

  781.     case PIX_FMT_RGB8:\

  782.     case PIX_FMT_BGR8:\

  783.         {\

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

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

  786.             func(uint8_t,0)\

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

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

  789.             }\

  790.         }\

  791.         break;\

  792.     case PIX_FMT_RGB4:\

  793.     case PIX_FMT_BGR4:\

  794.         {\

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

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

  797.             func(uint8_t,0)\

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

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

  800.             }\

  801.         }\

  802.         break;\

  803.     case PIX_FMT_RGB4_BYTE:\

  804.     case PIX_FMT_BGR4_BYTE:\

  805.         {\

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

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

  808.             func(uint8_t,0)\

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

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

  811.             }\

  812.         }\

  813.         break;\

  814.     case PIX_FMT_MONOBLACK:\

  815.     case PIX_FMT_MONOWHITE:\

  816.         {\

  817.             func_monoblack\

  818.         }\

  819.         break;\

  820.     case PIX_FMT_YUYV422:\

  821.         func2\

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

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

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

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

  826.         }                \

  827.         break;\

  828.     case PIX_FMT_UYVY422:\

  829.         func2\

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

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

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

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

  834.         }                \

  835.         break;\

  836.     case PIX_FMT_GRAY16BE:\

  837.         func_g16\

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

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

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

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

  842.         }                \

  843.         break;\

  844.     case PIX_FMT_GRAY16LE:\

  845.         func_g16\

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

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

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

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

  850.         }                \

  851.         break;\

  852.     }

  853. 

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

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

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

  857. {

  858.     int i;

  859.     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)

  860. }

  861. 

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

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

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

  865. {

  866.     int i;

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

  868.     int aidx= 3;

  869. 

  870.     switch(c->dstFormat) {

  871.     case PIX_FMT_ARGB:

  872.         dest++;

  873.         aidx= 0;

  874.     case PIX_FMT_RGB24:

  875.         aidx--;

  876.     case PIX_FMT_RGBA:

  877.         if (CONFIG_SMALL) {

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

  879.             YSCALE_YUV_2_RGBX_FULL_C(1<<21, needAlpha)

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

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

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

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

  884.                 dest+= step;

  885.             }

  886.         } else {

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

  888.                 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 1)

  889.                     dest[aidx]= A;

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

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

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

  893.                     dest+= step;

  894.                 }

  895.             } else {

  896.                 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 0)

  897.                     dest[aidx]= 255;

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

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

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

  901.                     dest+= step;

  902.                 }

  903.             }

  904.         }

  905.         break;

  906.     case PIX_FMT_ABGR:

  907.         dest++;

  908.         aidx= 0;

  909.     case PIX_FMT_BGR24:

  910.         aidx--;

  911.     case PIX_FMT_BGRA:

  912.         if (CONFIG_SMALL) {

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

  914.             YSCALE_YUV_2_RGBX_FULL_C(1<<21, needAlpha)

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

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

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

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

  919.                 dest+= step;

  920.             }

  921.         } else {

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

  923.                 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 1)

  924.                     dest[aidx]= A;

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

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

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

  928.                     dest+= step;

  929.                 }

  930.             } else {

  931.                 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 0)

  932.                     dest[aidx]= 255;

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

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

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

  936.                     dest+= step;

  937.                 }

  938.             }

  939.         }

  940.         break;

  941.     default:

  942.         assert(0);

  943.     }

  944. }

  945. 

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

  947. {

  948.     int i;

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

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

  951.         memset(ptr, val, width);

  952.         ptr += stride;

  953.     }

  954. }

  955. 

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

  957.                             uint32_t *unused)

  958. {

  959.     int i;

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

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

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

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

  964. 

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

  966.     }

  967. }

  968. 

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

  970.                              const uint8_t *src1, const uint8_t *src2,

  971.                              long width, uint32_t *unused)

  972. {

  973.     int i;

  974.     assert(src1==src2);

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

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

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

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

  979. 

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

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

  982.     }

  983. }

  984. 

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

  986.                                   const uint8_t *src1, const uint8_t *src2,

  987.                                   long width, uint32_t *unused)

  988. {

  989.     int i;

  990.     assert(src1==src2);

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

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

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

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

  995. 

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

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

  998.     }

  999. }

  1000. 

  1001. static inline void bgr48ToY(uint8_t *dst, const uint8_t *src, long width,

  1002.                             uint32_t *unused)

  1003. {

  1004.     int i;

  1005.     for (i = 0; i < width; i++) {

  1006.         int b = src[i*6+0];

  1007.         int g = src[i*6+2];

  1008.         int r = src[i*6+4];

  1009. 

  1010.         dst[i] = (RY*r + GY*g + BY*b + (33<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;

  1011.     }

  1012. }

  1013. 

  1014. static inline void bgr48ToUV(uint8_t *dstU, uint8_t *dstV,

  1015.                              const uint8_t *src1, const uint8_t *src2,

  1016.                              long width, uint32_t *unused)

  1017. {

  1018.     int i;

  1019.     for (i = 0; i < width; i++) {

  1020.         int b = src1[6*i + 0];

  1021.         int g = src1[6*i + 2];

  1022.         int r = src1[6*i + 4];

  1023. 

  1024.         dstU[i] = (RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;

  1025.         dstV[i] = (RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;

  1026.     }

  1027. }

  1028. 

  1029. static inline void bgr48ToUV_half(uint8_t *dstU, uint8_t *dstV,

  1030.                                   const uint8_t *src1, const uint8_t *src2,

  1031.                                   long width, uint32_t *unused)

  1032. {

  1033.     int i;

  1034.     for (i = 0; i < width; i++) {

  1035.         int b= src1[12*i + 0] + src1[12*i + 6];

  1036.         int g= src1[12*i + 2] + src1[12*i + 8];

  1037.         int r= src1[12*i + 4] + src1[12*i + 10];

  1038. 

  1039.         dstU[i]= (RU*r + GU*g + BU*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);

  1040.         dstV[i]= (RV*r + GV*g + BV*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);

  1041.     }

  1042. }

  1043. 

  1044. #define BGR2Y(type, name, shr, shg, shb, maskr, maskg, maskb, RY, GY, BY, S)\

  1045. static inline void name(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)\

  1046. {\

  1047.     int i;\

  1048.     for (i=0; i<width; i++) {\

  1049.         int b= (((const type*)src)[i]>>shb)&maskb;\

  1050.         int g= (((const type*)src)[i]>>shg)&maskg;\

  1051.         int r= (((const type*)src)[i]>>shr)&maskr;\

  1052. \

  1053.         dst[i]= (((RY)*r + (GY)*g + (BY)*b + (33<<((S)-1)))>>(S));\

  1054.     }\

  1055. }

  1056. 

  1057. BGR2Y(uint32_t, bgr32ToY,16, 0, 0, 0x00FF, 0xFF00, 0x00FF, RY<< 8, GY   , BY<< 8, RGB2YUV_SHIFT+8)

  1058. BGR2Y(uint32_t,bgr321ToY,16,16, 0, 0xFF00, 0x00FF, 0xFF00, RY    , GY<<8, BY    , RGB2YUV_SHIFT+8)

  1059. BGR2Y(uint32_t, rgb32ToY, 0, 0,16, 0x00FF, 0xFF00, 0x00FF, RY<< 8, GY   , BY<< 8, RGB2YUV_SHIFT+8)

  1060. BGR2Y(uint32_t,rgb321ToY, 0,16,16, 0xFF00, 0x00FF, 0xFF00, RY    , GY<<8, BY    , RGB2YUV_SHIFT+8)

  1061. BGR2Y(uint16_t, bgr16ToY, 0, 0, 0, 0x001F, 0x07E0, 0xF800, RY<<11, GY<<5, BY    , RGB2YUV_SHIFT+8)

  1062. BGR2Y(uint16_t, bgr15ToY, 0, 0, 0, 0x001F, 0x03E0, 0x7C00, RY<<10, GY<<5, BY    , RGB2YUV_SHIFT+7)

  1063. BGR2Y(uint16_t, rgb16ToY, 0, 0, 0, 0xF800, 0x07E0, 0x001F, RY    , GY<<5, BY<<11, RGB2YUV_SHIFT+8)

  1064. BGR2Y(uint16_t, rgb15ToY, 0, 0, 0, 0x7C00, 0x03E0, 0x001F, RY    , GY<<5, BY<<10, RGB2YUV_SHIFT+7)

  1065. 

  1066. static inline void abgrToA(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)

  1067. {

  1068.     int i;

  1069.     for (i=0; i<width; i++) {

  1070.         dst[i]= src[4*i];

  1071.     }

  1072. }

  1073. 

  1074. #define BGR2UV(type, name, shr, shg, shb, shp, maskr, maskg, maskb, RU, GU, BU, RV, GV, BV, S) \

  1075. static inline void name(uint8_t *dstU, uint8_t *dstV, const uint8_t *src, const uint8_t *dummy, long width, uint32_t *unused)\

  1076. {\

  1077.     int i;\

  1078.     for (i=0; i<width; i++) {\

  1079.         int b= ((((const type*)src)[i]>>shp)&maskb)>>shb;\

  1080.         int g= ((((const type*)src)[i]>>shp)&maskg)>>shg;\

  1081.         int r= ((((const type*)src)[i]>>shp)&maskr)>>shr;\

  1082. \

  1083.         dstU[i]= ((RU)*r + (GU)*g + (BU)*b + (257<<((S)-1)))>>(S);\

  1084.         dstV[i]= ((RV)*r + (GV)*g + (BV)*b + (257<<((S)-1)))>>(S);\

  1085.     }\

  1086. }\

  1087. static inline void name ## _half(uint8_t *dstU, uint8_t *dstV, const uint8_t *src, const uint8_t *dummy, long width, uint32_t *unused)\

  1088. {\

  1089.     int i;\

  1090.     for (i=0; i<width; i++) {\

  1091.         int pix0= ((const type*)src)[2*i+0]>>shp;\

  1092.         int pix1= ((const type*)src)[2*i+1]>>shp;\

  1093.         int g= (pix0&~(maskr|maskb))+(pix1&~(maskr|maskb));\

  1094.         int b= ((pix0+pix1-g)&(maskb|(2*maskb)))>>shb;\

  1095.         int r= ((pix0+pix1-g)&(maskr|(2*maskr)))>>shr;\

  1096.         g&= maskg|(2*maskg);\

  1097. \

  1098.         g>>=shg;\

  1099. \

  1100.         dstU[i]= ((RU)*r + (GU)*g + (BU)*b + (257<<(S)))>>((S)+1);\

  1101.         dstV[i]= ((RV)*r + (GV)*g + (BV)*b + (257<<(S)))>>((S)+1);\

  1102.     }\

  1103. }

  1104. 

  1105. BGR2UV(uint32_t, bgr32ToUV,16, 0, 0, 0, 0xFF0000, 0xFF00,   0x00FF, RU<< 8, GU   , BU<< 8, RV<< 8, GV   , BV<< 8, RGB2YUV_SHIFT+8)

  1106. BGR2UV(uint32_t,bgr321ToUV,16, 0, 0, 8, 0xFF0000, 0xFF00,   0x00FF, RU<< 8, GU   , BU<< 8, RV<< 8, GV   , BV<< 8, RGB2YUV_SHIFT+8)

  1107. BGR2UV(uint32_t, rgb32ToUV, 0, 0,16, 0,   0x00FF, 0xFF00, 0xFF0000, RU<< 8, GU   , BU<< 8, RV<< 8, GV   , BV<< 8, RGB2YUV_SHIFT+8)

  1108. BGR2UV(uint32_t,rgb321ToUV, 0, 0,16, 8,   0x00FF, 0xFF00, 0xFF0000, RU<< 8, GU   , BU<< 8, RV<< 8, GV   , BV<< 8, RGB2YUV_SHIFT+8)

  1109. BGR2UV(uint16_t, bgr16ToUV, 0, 0, 0, 0,   0x001F, 0x07E0,   0xF800, RU<<11, GU<<5, BU    , RV<<11, GV<<5, BV    , RGB2YUV_SHIFT+8)

  1110. BGR2UV(uint16_t, bgr15ToUV, 0, 0, 0, 0,   0x001F, 0x03E0,   0x7C00, RU<<10, GU<<5, BU    , RV<<10, GV<<5, BV    , RGB2YUV_SHIFT+7)

  1111. BGR2UV(uint16_t, rgb16ToUV, 0, 0, 0, 0,   0xF800, 0x07E0,   0x001F, RU    , GU<<5, BU<<11, RV    , GV<<5, BV<<11, RGB2YUV_SHIFT+8)

  1112. BGR2UV(uint16_t, rgb15ToUV, 0, 0, 0, 0,   0x7C00, 0x03E0,   0x001F, RU    , GU<<5, BU<<10, RV    , GV<<5, BV<<10, RGB2YUV_SHIFT+7)

  1113. 

  1114. static inline void palToY(uint8_t *dst, const uint8_t *src, long width, uint32_t *pal)

  1115. {

  1116.     int i;

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

  1118.         int d= src[i];

  1119. 

  1120.         dst[i]= pal[d] & 0xFF;

  1121.     }

  1122. }

  1123. 

  1124. static inline void palToUV(uint8_t *dstU, uint8_t *dstV,

  1125.                            const uint8_t *src1, const uint8_t *src2,

  1126.                            long width, uint32_t *pal)

  1127. {

  1128.     int i;

  1129.     assert(src1 == src2);

  1130.     for (i=0; i<width; i++) {

  1131.         int p= pal[src1[i]];

  1132. 

  1133.         dstU[i]= p>>8;

  1134.         dstV[i]= p>>16;

  1135.     }

  1136. }

  1137. 

  1138. static inline void monowhite2Y(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)

  1139. {

  1140.     int i, j;

  1141.     for (i=0; i<width/8; i++) {

  1142.         int d= ~src[i];

  1143.         for(j=0; j<8; j++)

  1144.             dst[8*i+j]= ((d>>(7-j))&1)*255;

  1145.     }

  1146. }

  1147. 

  1148. static inline void monoblack2Y(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)

  1149. {

  1150.     int i, j;

  1151.     for (i=0; i<width/8; i++) {

  1152.         int d= src[i];

  1153.         for(j=0; j<8; j++)

  1154.             dst[8*i+j]= ((d>>(7-j))&1)*255;

  1155.     }

  1156. }

  1157. 

  1158. //Note: we have C, MMX, MMX2, 3DNOW versions, there is no 3DNOW+MMX2 one

  1159. //Plain C versions

  1160. #if CONFIG_RUNTIME_CPUDETECT

  1161. #  define COMPILE_C 1

  1162. #  if   ARCH_X86

  1163. #    define COMPILE_MMX     HAVE_MMX

  1164. #    define COMPILE_MMX2    HAVE_MMX2

  1165. #    define COMPILE_3DNOW   HAVE_AMD3DNOW

  1166. #  elif ARCH_PPC

  1167. #    define COMPILE_ALTIVEC HAVE_ALTIVEC

  1168. #  endif

  1169. #else /* CONFIG_RUNTIME_CPUDETECT */

  1170. #  if   ARCH_X86

  1171. #    if   HAVE_MMX2

  1172. #      define COMPILE_MMX2  1

  1173. #    elif HAVE_AMD3DNOW

  1174. #      define COMPILE_3DNOW 1

  1175. #    elif HAVE_MMX

  1176. #      define COMPILE_MMX   1

  1177. #    else

  1178. #      define COMPILE_C     1

  1179. #    endif

  1180. #  elif ARCH_PPC && HAVE_ALTIVEC

  1181. #    define COMPILE_ALTIVEC 1

  1182. #  else

  1183. #    define COMPILE_C       1

  1184. #  endif

  1185. #endif

  1186. 

  1187. #ifndef COMPILE_C

  1188. #  define COMPILE_C 0

  1189. #endif

  1190. #ifndef COMPILE_MMX

  1191. #  define COMPILE_MMX 0

  1192. #endif

  1193. #ifndef COMPILE_MMX2

  1194. #  define COMPILE_MMX2 0

  1195. #endif

  1196. #ifndef COMPILE_3DNOW

  1197. #  define COMPILE_3DNOW 0

  1198. #endif

  1199. #ifndef COMPILE_ALTIVEC

  1200. #  define COMPILE_ALTIVEC 0

  1201. #endif

  1202. 

  1203. #define COMPILE_TEMPLATE_MMX 0

  1204. #define COMPILE_TEMPLATE_MMX2 0

  1205. #define COMPILE_TEMPLATE_AMD3DNOW 0

  1206. #define COMPILE_TEMPLATE_ALTIVEC 0

  1207. 

  1208. #include "swscale_template.c"

  1209. 

  1210. #if COMPILE_ALTIVEC

  1211. #undef RENAME

  1212. #undef COMPILE_TEMPLATE_ALTIVEC

  1213. #define COMPILE_TEMPLATE_ALTIVEC 1

  1214. #define RENAME(a) a ## _altivec

  1215. #include "ppc/swscale_template.c"

  1216. #endif

  1217. 

  1218. #if ARCH_X86

  1219. 

  1220. //MMX versions

  1221. #if COMPILE_MMX

  1222. #undef RENAME

  1223. #undef COMPILE_TEMPLATE_MMX

  1224. #undef COMPILE_TEMPLATE_MMX2

  1225. #undef COMPILE_TEMPLATE_AMD3DNOW

  1226. #define COMPILE_TEMPLATE_MMX 1

  1227. #define COMPILE_TEMPLATE_MMX2 0

  1228. #define COMPILE_TEMPLATE_AMD3DNOW 0

  1229. #define RENAME(a) a ## _MMX

  1230. #include "x86/swscale_template.c"

  1231. #endif

  1232. 

  1233. //MMX2 versions

  1234. #if COMPILE_MMX2

  1235. #undef RENAME

  1236. #undef COMPILE_TEMPLATE_MMX

  1237. #undef COMPILE_TEMPLATE_MMX2

  1238. #undef COMPILE_TEMPLATE_AMD3DNOW

  1239. #define COMPILE_TEMPLATE_MMX 1

  1240. #define COMPILE_TEMPLATE_MMX2 1

  1241. #define COMPILE_TEMPLATE_AMD3DNOW 0

  1242. #define RENAME(a) a ## _MMX2

  1243. #include "x86/swscale_template.c"

  1244. #endif

  1245. 

  1246. //3DNOW versions

  1247. #if COMPILE_3DNOW

  1248. #undef RENAME

  1249. #undef COMPILE_TEMPLATE_MMX

  1250. #undef COMPILE_TEMPLATE_MMX2

  1251. #undef COMPILE_TEMPLATE_AMD3DNOW

  1252. #define COMPILE_TEMPLATE_MMX 1

  1253. #define COMPILE_TEMPLATE_MMX2 0

  1254. #define COMPILE_TEMPLATE_AMD3DNOW 1

  1255. #define RENAME(a) a ## _3DNow

  1256. #include "x86/swscale_template.c"

  1257. #endif

  1258. 

  1259. #endif //ARCH_X86

  1260. 

  1261. SwsFunc ff_getSwsFunc(SwsContext *c)

  1262. {

  1263.     sws_init_swScale_c(c);

  1264. 

  1265. #if CONFIG_RUNTIME_CPUDETECT

  1266. #if ARCH_X86

  1267.     // ordered per speed fastest first

  1268.     if (c->flags & SWS_CPU_CAPS_MMX2) {

  1269.         sws_init_swScale_MMX2(c);

  1270.         return swScale_MMX2;

  1271.     } else if (c->flags & SWS_CPU_CAPS_3DNOW) {

  1272.         sws_init_swScale_3DNow(c);

  1273.         return swScale_3DNow;

  1274.     } else if (c->flags & SWS_CPU_CAPS_MMX) {

  1275.         sws_init_swScale_MMX(c);

  1276.         return swScale_MMX;

  1277.     }

  1278. 

  1279. #else

  1280. #if COMPILE_ALTIVEC

  1281.     if (c->flags & SWS_CPU_CAPS_ALTIVEC) {

  1282.         sws_init_swScale_altivec(c);

  1283.         return swScale_altivec;

  1284.     }

  1285. #endif

  1286. #endif /* ARCH_X86 */

  1287. #else //CONFIG_RUNTIME_CPUDETECT

  1288. #if   COMPILE_TEMPLATE_MMX2

  1289.     sws_init_swScale_MMX2(c);

  1290.     return swScale_MMX2;

  1291. #elif COMPILE_TEMPLATE_AMD3DNOW

  1292.     sws_init_swScale_3DNow(c);

  1293.     return swScale_3DNow;

  1294. #elif COMPILE_TEMPLATE_MMX

  1295.     sws_init_swScale_MMX(c);

  1296.     return swScale_MMX;

  1297. #elif COMPILE_TEMPLATE_ALTIVEC

  1298.     sws_init_swScale_altivec(c);

  1299.     return swScale_altivec;

  1300. #endif

  1301. #endif //!CONFIG_RUNTIME_CPUDETECT

  1302. 

  1303.     return swScale_c;

  1304. }

  1305. 

  1306. static void copyPlane(const uint8_t *src, int srcStride,

  1307.                       int srcSliceY, int srcSliceH, int width,

  1308.                       uint8_t *dst, int dstStride)

  1309. {

  1310.     dst += dstStride * srcSliceY;

  1311.     if (dstStride == srcStride && srcStride > 0) {

  1312.         memcpy(dst, src, srcSliceH * dstStride);

  1313.     } else {

  1314.         int i;

  1315.         for (i=0; i<srcSliceH; i++) {

  1316.             memcpy(dst, src, width);

  1317.             src += srcStride;

  1318.             dst += dstStride;

  1319.         }

  1320.     }

  1321. }

  1322. 

  1323. static int planarToNv12Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,

  1324.                                int srcSliceH, uint8_t* dstParam[], int dstStride[])

  1325. {

  1326.     uint8_t *dst = dstParam[1] + dstStride[1]*srcSliceY/2;

  1327. 

  1328.     copyPlane(src[0], srcStride[0], srcSliceY, srcSliceH, c->srcW,

  1329.               dstParam[0], dstStride[0]);

  1330. 

  1331.     if (c->dstFormat == PIX_FMT_NV12)

  1332.         interleaveBytes(src[1], src[2], dst, c->srcW/2, srcSliceH/2, srcStride[1], srcStride[2], dstStride[0]);

  1333.     else

  1334.         interleaveBytes(src[2], src[1], dst, c->srcW/2, srcSliceH/2, srcStride[2], srcStride[1], dstStride[0]);

  1335. 

  1336.     return srcSliceH;

  1337. }

  1338. 

  1339. static int planarToYuy2Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,

  1340.                                int srcSliceH, uint8_t* dstParam[], int dstStride[])

  1341. {

  1342.     uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;

  1343. 

  1344.     yv12toyuy2(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);

  1345. 

  1346.     return srcSliceH;

  1347. }

  1348. 

  1349. static int planarToUyvyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,

  1350.                                int srcSliceH, uint8_t* dstParam[], int dstStride[])

  1351. {

  1352.     uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;

  1353. 

  1354.     yv12touyvy(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);

  1355. 

  1356.     return srcSliceH;

  1357. }

  1358. 

  1359. static int yuv422pToYuy2Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,

  1360.                                 int srcSliceH, uint8_t* dstParam[], int dstStride[])

  1361. {

  1362.     uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;

  1363. 

  1364.     yuv422ptoyuy2(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);

  1365. 

  1366.     return srcSliceH;

  1367. }

  1368. 

  1369. static int yuv422pToUyvyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,

  1370.                                 int srcSliceH, uint8_t* dstParam[], int dstStride[])

  1371. {

  1372.     uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;

  1373. 

  1374.     yuv422ptouyvy(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);

  1375. 

  1376.     return srcSliceH;

  1377. }

  1378. 

  1379. static int yuyvToYuv420Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,

  1380.                                int srcSliceH, uint8_t* dstParam[], int dstStride[])

  1381. {

  1382.     uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;

  1383.     uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY/2;

  1384.     uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY/2;

  1385. 

  1386.     yuyvtoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);

  1387. 

  1388.     if (dstParam[3])

  1389.         fillPlane(dstParam[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);

  1390. 

  1391.     return srcSliceH;

  1392. }

  1393. 

  1394. static int yuyvToYuv422Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,

  1395.                                int srcSliceH, uint8_t* dstParam[], int dstStride[])

  1396. {

  1397.     uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;

  1398.     uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY;

  1399.     uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY;

  1400. 

  1401.     yuyvtoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);

  1402. 

  1403.     return srcSliceH;

  1404. }

  1405. 

  1406. static int uyvyToYuv420Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,

  1407.                                int srcSliceH, uint8_t* dstParam[], int dstStride[])

  1408. {

  1409.     uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;

  1410.     uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY/2;

  1411.     uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY/2;

  1412. 

  1413.     uyvytoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);

  1414. 

  1415.     if (dstParam[3])

  1416.         fillPlane(dstParam[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);

  1417. 

  1418.     return srcSliceH;

  1419. }

  1420. 

  1421. static int uyvyToYuv422Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,

  1422.                                int srcSliceH, uint8_t* dstParam[], int dstStride[])

  1423. {

  1424.     uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;

  1425.     uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY;

  1426.     uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY;

  1427. 

  1428.     uyvytoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);

  1429. 

  1430.     return srcSliceH;

  1431. }

  1432. 

  1433. static void gray8aToPacked32(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)

  1434. {

  1435.     long i;

  1436.     for (i=0; i<num_pixels; i++)

  1437.         ((uint32_t *) dst)[i] = ((const uint32_t *)palette)[src[i<<1]] | (src[(i<<1)+1] << 24);

  1438. }

  1439. 

  1440. static void gray8aToPacked32_1(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)

  1441. {

  1442.     long i;

  1443. 

  1444.     for (i=0; i<num_pixels; i++)

  1445.         ((uint32_t *) dst)[i] = ((const uint32_t *)palette)[src[i<<1]] | src[(i<<1)+1];

  1446. }

  1447. 

  1448. static void gray8aToPacked24(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)

  1449. {

  1450.     long i;

  1451. 

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

  1453.         //FIXME slow?

  1454.         dst[0]= palette[src[i<<1]*4+0];

  1455.         dst[1]= palette[src[i<<1]*4+1];

  1456.         dst[2]= palette[src[i<<1]*4+2];

  1457.         dst+= 3;

  1458.     }

  1459. }

  1460. 

  1461. static int palToRgbWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,

  1462.                            int srcSliceH, uint8_t* dst[], int dstStride[])

  1463. {

  1464.     const enum PixelFormat srcFormat= c->srcFormat;

  1465.     const enum PixelFormat dstFormat= c->dstFormat;

  1466.     void (*conv)(const uint8_t *src, uint8_t *dst, long num_pixels,

  1467.                  const uint8_t *palette)=NULL;

  1468.     int i;

  1469.     uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;

  1470.     const uint8_t *srcPtr= src[0];

  1471. 

  1472.     if (srcFormat == PIX_FMT_Y400A) {

  1473.         switch (dstFormat) {

  1474.         case PIX_FMT_RGB32  : conv = gray8aToPacked32; break;

  1475.         case PIX_FMT_BGR32  : conv = gray8aToPacked32; break;

  1476.         case PIX_FMT_BGR32_1: conv = gray8aToPacked32_1; break;

  1477.         case PIX_FMT_RGB32_1: conv = gray8aToPacked32_1; break;

  1478.         case PIX_FMT_RGB24  : conv = gray8aToPacked24; break;

  1479.         case PIX_FMT_BGR24  : conv = gray8aToPacked24; break;

  1480.         }

  1481.     } else if (usePal(srcFormat)) {

  1482.         switch (dstFormat) {

  1483.         case PIX_FMT_RGB32  : conv = sws_convertPalette8ToPacked32; break;

  1484.         case PIX_FMT_BGR32  : conv = sws_convertPalette8ToPacked32; break;

  1485.         case PIX_FMT_BGR32_1: conv = sws_convertPalette8ToPacked32; break;

  1486.         case PIX_FMT_RGB32_1: conv = sws_convertPalette8ToPacked32; break;

  1487.         case PIX_FMT_RGB24  : conv = sws_convertPalette8ToPacked24; break;

  1488.         case PIX_FMT_BGR24  : conv = sws_convertPalette8ToPacked24; break;

  1489.         }

  1490.     }

  1491. 

  1492.     if (!conv)

  1493.         av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",

  1494.                sws_format_name(srcFormat), sws_format_name(dstFormat));

  1495.     else {

  1496.         for (i=0; i<srcSliceH; i++) {

  1497.             conv(srcPtr, dstPtr, c->srcW, (uint8_t *) c->pal_rgb);

  1498.             srcPtr+= srcStride[0];

  1499.             dstPtr+= dstStride[0];

  1500.         }

  1501.     }

  1502. 

  1503.     return srcSliceH;

  1504. }

  1505. 

  1506. #define isRGBA32(x) (            \

  1507.            (x) == PIX_FMT_ARGB   \

  1508.         || (x) == PIX_FMT_RGBA   \

  1509.         || (x) == PIX_FMT_BGRA   \

  1510.         || (x) == PIX_FMT_ABGR   \

  1511.         )

  1512. 

  1513. /* {RGB,BGR}{15,16,24,32,32_1} -> {RGB,BGR}{15,16,24,32} */

  1514. static int rgbToRgbWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,

  1515.                            int srcSliceH, uint8_t* dst[], int dstStride[])

  1516. {

  1517.     const enum PixelFormat srcFormat= c->srcFormat;

  1518.     const enum PixelFormat dstFormat= c->dstFormat;

  1519.     const int srcBpp= (c->srcFormatBpp + 7) >> 3;

  1520.     const int dstBpp= (c->dstFormatBpp + 7) >> 3;

  1521.     const int srcId= c->srcFormatBpp >> 2; /* 1:0, 4:1, 8:2, 15:3, 16:4, 24:6, 32:8 */

  1522.     const int dstId= c->dstFormatBpp >> 2;

  1523.     void (*conv)(const uint8_t *src, uint8_t *dst, long src_size)=NULL;

  1524. 

  1525. #define CONV_IS(src, dst) (srcFormat == PIX_FMT_##src && dstFormat == PIX_FMT_##dst)

  1526. 

  1527.     if (isRGBA32(srcFormat) && isRGBA32(dstFormat)) {

  1528.         if (     CONV_IS(ABGR, RGBA)

  1529.               || CONV_IS(ARGB, BGRA)

  1530.               || CONV_IS(BGRA, ARGB)

  1531.               || CONV_IS(RGBA, ABGR)) conv = shuffle_bytes_3210;

  1532.         else if (CONV_IS(ABGR, ARGB)

  1533.               || CONV_IS(ARGB, ABGR)) conv = shuffle_bytes_0321;

  1534.         else if (CONV_IS(ABGR, BGRA)

  1535.               || CONV_IS(ARGB, RGBA)) conv = shuffle_bytes_1230;

  1536.         else if (CONV_IS(BGRA, RGBA)

  1537.               || CONV_IS(RGBA, BGRA)) conv = shuffle_bytes_2103;

  1538.         else if (CONV_IS(BGRA, ABGR)

  1539.               || CONV_IS(RGBA, ARGB)) conv = shuffle_bytes_3012;

  1540.     } else

  1541.     /* BGR -> BGR */

  1542.     if (  (isBGRinInt(srcFormat) && isBGRinInt(dstFormat))

  1543.        || (isRGBinInt(srcFormat) && isRGBinInt(dstFormat))) {

  1544.         switch(srcId | (dstId<<4)) {

  1545.         case 0x34: conv= rgb16to15; break;

  1546.         case 0x36: conv= rgb24to15; break;

  1547.         case 0x38: conv= rgb32to15; break;

  1548.         case 0x43: conv= rgb15to16; break;

  1549.         case 0x46: conv= rgb24to16; break;

  1550.         case 0x48: conv= rgb32to16; break;

  1551.         case 0x63: conv= rgb15to24; break;

  1552.         case 0x64: conv= rgb16to24; break;

  1553.         case 0x68: conv= rgb32to24; break;

  1554.         case 0x83: conv= rgb15to32; break;

  1555.         case 0x84: conv= rgb16to32; break;

  1556.         case 0x86: conv= rgb24to32; break;

  1557.         }

  1558.     } else if (  (isBGRinInt(srcFormat) && isRGBinInt(dstFormat))

  1559.              || (isRGBinInt(srcFormat) && isBGRinInt(dstFormat))) {

  1560.         switch(srcId | (dstId<<4)) {

  1561.         case 0x33: conv= rgb15tobgr15; break;

  1562.         case 0x34: conv= rgb16tobgr15; break;

  1563.         case 0x36: conv= rgb24tobgr15; break;

  1564.         case 0x38: conv= rgb32tobgr15; break;

  1565.         case 0x43: conv= rgb15tobgr16; break;

  1566.         case 0x44: conv= rgb16tobgr16; break;

  1567.         case 0x46: conv= rgb24tobgr16; break;

  1568.         case 0x48: conv= rgb32tobgr16; break;

  1569.         case 0x63: conv= rgb15tobgr24; break;

  1570.         case 0x64: conv= rgb16tobgr24; break;

  1571.         case 0x66: conv= rgb24tobgr24; break;

  1572.         case 0x68: conv= rgb32tobgr24; break;

  1573.         case 0x83: conv= rgb15tobgr32; break;

  1574.         case 0x84: conv= rgb16tobgr32; break;

  1575.         case 0x86: conv= rgb24tobgr32; break;

  1576.         }

  1577.     }

  1578. 

  1579.     if (!conv) {

  1580.         av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",

  1581.                sws_format_name(srcFormat), sws_format_name(dstFormat));

  1582.     } else {

  1583.         const uint8_t *srcPtr= src[0];

  1584.               uint8_t *dstPtr= dst[0];

  1585.         if ((srcFormat == PIX_FMT_RGB32_1 || srcFormat == PIX_FMT_BGR32_1) && !isRGBA32(dstFormat))

  1586.             srcPtr += ALT32_CORR;

  1587. 

  1588.         if ((dstFormat == PIX_FMT_RGB32_1 || dstFormat == PIX_FMT_BGR32_1) && !isRGBA32(srcFormat))

  1589.             dstPtr += ALT32_CORR;

  1590. 

  1591.         if (dstStride[0]*srcBpp == srcStride[0]*dstBpp && srcStride[0] > 0)

  1592.             conv(srcPtr, dstPtr + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]);

  1593.         else {

  1594.             int i;

  1595.             dstPtr += dstStride[0]*srcSliceY;

  1596. 

  1597.             for (i=0; i<srcSliceH; i++) {

  1598.                 conv(srcPtr, dstPtr, c->srcW*srcBpp);

  1599.                 srcPtr+= srcStride[0];

  1600.                 dstPtr+= dstStride[0];

  1601.             }

  1602.         }

  1603.     }

  1604.     return srcSliceH;

  1605. }

  1606. 

  1607. static int bgr24ToYv12Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,

  1608.                               int srcSliceH, uint8_t* dst[], int dstStride[])

  1609. {

  1610.     rgb24toyv12(

  1611.         src[0],

  1612.         dst[0]+ srcSliceY    *dstStride[0],

  1613.         dst[1]+(srcSliceY>>1)*dstStride[1],

  1614.         dst[2]+(srcSliceY>>1)*dstStride[2],

  1615.         c->srcW, srcSliceH,

  1616.         dstStride[0], dstStride[1], srcStride[0]);

  1617.     if (dst[3])

  1618.         fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);

  1619.     return srcSliceH;

  1620. }

  1621. 

  1622. static int yvu9ToYv12Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,

  1623.                              int srcSliceH, uint8_t* dst[], int dstStride[])

  1624. {

  1625.     copyPlane(src[0], srcStride[0], srcSliceY, srcSliceH, c->srcW,

  1626.               dst[0], dstStride[0]);

  1627. 

  1628.     planar2x(src[1], dst[1] + dstStride[1]*(srcSliceY >> 1), c->chrSrcW,

  1629.              srcSliceH >> 2, srcStride[1], dstStride[1]);

  1630.     planar2x(src[2], dst[2] + dstStride[2]*(srcSliceY >> 1), c->chrSrcW,

  1631.              srcSliceH >> 2, srcStride[2], dstStride[2]);

  1632.     if (dst[3])

  1633.         fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);

  1634.     return srcSliceH;

  1635. }

  1636. 

  1637. /* unscaled copy like stuff (assumes nearly identical formats) */

  1638. static int packedCopyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,

  1639.                              int srcSliceH, uint8_t* dst[], int dstStride[])

  1640. {

  1641.     if (dstStride[0]==srcStride[0] && srcStride[0] > 0)

  1642.         memcpy(dst[0] + dstStride[0]*srcSliceY, src[0], srcSliceH*dstStride[0]);

  1643.     else {

  1644.         int i;

  1645.         const uint8_t *srcPtr= src[0];

  1646.         uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;

  1647.         int length=0;

  1648. 

  1649.         /* universal length finder */

  1650.         while(length+c->srcW <= FFABS(dstStride[0])

  1651.            && length+c->srcW <= FFABS(srcStride[0])) length+= c->srcW;

  1652.         assert(length!=0);

  1653. 

  1654.         for (i=0; i<srcSliceH; i++) {

  1655.             memcpy(dstPtr, srcPtr, length);

  1656.             srcPtr+= srcStride[0];

  1657.             dstPtr+= dstStride[0];

  1658.         }

  1659.     }

  1660.     return srcSliceH;

  1661. }

  1662. 

  1663. static int planarCopyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,

  1664.                              int srcSliceH, uint8_t* dst[], int dstStride[])

  1665. {

  1666.     int plane, i, j;

  1667.     for (plane=0; plane<4; plane++) {

  1668.         int length= (plane==0 || plane==3) ? c->srcW  : -((-c->srcW  )>>c->chrDstHSubSample);

  1669.         int y=      (plane==0 || plane==3) ? srcSliceY: -((-srcSliceY)>>c->chrDstVSubSample);

  1670.         int height= (plane==0 || plane==3) ? srcSliceH: -((-srcSliceH)>>c->chrDstVSubSample);

  1671.         const uint8_t *srcPtr= src[plane];

  1672.         uint8_t *dstPtr= dst[plane] + dstStride[plane]*y;

  1673. 

  1674.         if (!dst[plane]) continue;

  1675.         // ignore palette for GRAY8

  1676.         if (plane == 1 && !dst[2]) continue;

  1677.         if (!src[plane] || (plane == 1 && !src[2])) {

  1678.             if(is16BPS(c->dstFormat))

  1679.                 length*=2;

  1680.             fillPlane(dst[plane], dstStride[plane], length, height, y, (plane==3) ? 255 : 128);

  1681.         } else {

  1682.             if(is9_OR_10BPS(c->srcFormat)) {

  1683.                 const int src_depth = av_pix_fmt_descriptors[c->srcFormat].comp[plane].depth_minus1+1;

  1684.                 const int dst_depth = av_pix_fmt_descriptors[c->dstFormat].comp[plane].depth_minus1+1;

  1685.                 const uint16_t *srcPtr2 = (const uint16_t*)srcPtr;

  1686. 

  1687.                 if (is16BPS(c->dstFormat)) {

  1688.                     uint16_t *dstPtr2 = (uint16_t*)dstPtr;

  1689. #define COPY9_OR_10TO16(rfunc, wfunc) \

  1690.                     for (i = 0; i < height; i++) { \

  1691.                         for (j = 0; j < length; j++) { \

  1692.                             int srcpx = rfunc(&srcPtr2[j]); \

  1693.                             wfunc(&dstPtr2[j], (srcpx<<(16-src_depth)) | (srcpx>>(2*src_depth-16))); \

  1694.                         } \

  1695.                         dstPtr2 += dstStride[plane]/2; \

  1696.                         srcPtr2 += srcStride[plane]/2; \

  1697.                     }

  1698.                     if (isBE(c->dstFormat)) {

  1699.                         if (isBE(c->srcFormat)) {

  1700.                             COPY9_OR_10TO16(AV_RB16, AV_WB16);

  1701.                         } else {

  1702.                             COPY9_OR_10TO16(AV_RL16, AV_WB16);

  1703.                         }

  1704.                     } else {

  1705.                         if (isBE(c->srcFormat)) {

  1706.                             COPY9_OR_10TO16(AV_RB16, AV_WL16);

  1707.                         } else {

  1708.                             COPY9_OR_10TO16(AV_RL16, AV_WL16);

  1709.                         }

  1710.                     }

  1711.                 } else if (is9_OR_10BPS(c->dstFormat)) {

  1712.                     uint16_t *dstPtr2 = (uint16_t*)dstPtr;

  1713. #define COPY9_OR_10TO9_OR_10(loop) \

  1714.                     for (i = 0; i < height; i++) { \

  1715.                         for (j = 0; j < length; j++) { \

  1716.                             loop; \

  1717.                         } \

  1718.                         dstPtr2 += dstStride[plane]/2; \

  1719.                         srcPtr2 += srcStride[plane]/2; \

  1720.                     }

  1721. #define COPY9_OR_10TO9_OR_10_2(rfunc, wfunc) \

  1722.                     if (dst_depth > src_depth) { \

  1723.                         COPY9_OR_10TO9_OR_10(int srcpx = rfunc(&srcPtr2[j]); \

  1724.                             wfunc(&dstPtr2[j], (srcpx << 1) | (srcpx >> 9))); \

  1725.                     } else if (dst_depth < src_depth) { \

  1726.                         COPY9_OR_10TO9_OR_10(wfunc(&dstPtr2[j], rfunc(&srcPtr2[j]) >> 1)); \

  1727.                     } else { \

  1728.                         COPY9_OR_10TO9_OR_10(wfunc(&dstPtr2[j], rfunc(&srcPtr2[j]))); \

  1729.                     }

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

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

  1732.                             COPY9_OR_10TO9_OR_10_2(AV_RB16, AV_WB16);

  1733.                         } else {

  1734.                             COPY9_OR_10TO9_OR_10_2(AV_RL16, AV_WB16);

  1735.                         }

  1736.                     } else {

  1737.                         if (isBE(c->srcFormat)) {

  1738.                             COPY9_OR_10TO9_OR_10_2(AV_RB16, AV_WL16);

  1739.                         } else {

  1740.                             COPY9_OR_10TO9_OR_10_2(AV_RL16, AV_WL16);

  1741.                         }

  1742.                     }

  1743.                 } else {

  1744.                     // FIXME Maybe dither instead.

  1745. #define COPY9_OR_10TO8(rfunc) \

  1746.                     for (i = 0; i < height; i++) { \

  1747.                         for (j = 0; j < length; j++) { \

  1748.                             dstPtr[j] = rfunc(&srcPtr2[j])>>(src_depth-8); \

  1749.                         } \

  1750.                         dstPtr  += dstStride[plane]; \

  1751.                         srcPtr2 += srcStride[plane]/2; \

  1752.                     }

  1753.                     if (isBE(c->srcFormat)) {

  1754.                         COPY9_OR_10TO8(AV_RB16);

  1755.                     } else {

  1756.                         COPY9_OR_10TO8(AV_RL16);

  1757.                     }

  1758.                 }

  1759.             } else if(is9_OR_10BPS(c->dstFormat)) {

  1760.                 const int dst_depth = av_pix_fmt_descriptors[c->dstFormat].comp[plane].depth_minus1+1;

  1761.                 uint16_t *dstPtr2 = (uint16_t*)dstPtr;

  1762. 

  1763.                 if (is16BPS(c->srcFormat)) {

  1764.                     const uint16_t *srcPtr2 = (const uint16_t*)srcPtr;

  1765. #define COPY16TO9_OR_10(rfunc, wfunc) \

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

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

  1768.                             wfunc(&dstPtr2[j], rfunc(&srcPtr2[j])>>(16-dst_depth)); \

  1769.                         } \

  1770.                         dstPtr2 += dstStride[plane]/2; \

  1771.                         srcPtr2 += srcStride[plane]/2; \

  1772.                     }

  1773.                     if (isBE(c->dstFormat)) {

  1774.                         if (isBE(c->srcFormat)) {

  1775.                             COPY16TO9_OR_10(AV_RB16, AV_WB16);

  1776.                         } else {

  1777.                             COPY16TO9_OR_10(AV_RL16, AV_WB16);

  1778.                         }

  1779.                     } else {

  1780.                         if (isBE(c->srcFormat)) {

  1781.                             COPY16TO9_OR_10(AV_RB16, AV_WL16);

  1782.                         } else {

  1783.                             COPY16TO9_OR_10(AV_RL16, AV_WL16);

  1784.                         }

  1785.                     }

  1786.                 } else /* 8bit */ {

  1787. #define COPY8TO9_OR_10(wfunc) \

  1788.                     for (i = 0; i < height; i++) { \

  1789.                         for (j = 0; j < length; j++) { \

  1790.                             const int srcpx = srcPtr[j]; \

  1791.                             wfunc(&dstPtr2[j], (srcpx<<(dst_depth-8)) | (srcpx >> (16-dst_depth))); \

  1792.                         } \

  1793.                         dstPtr2 += dstStride[plane]/2; \

  1794.                         srcPtr  += srcStride[plane]; \

  1795.                     }

  1796.                     if (isBE(c->dstFormat)) {

  1797.                         COPY8TO9_OR_10(AV_WB16);

  1798.                     } else {

  1799.                         COPY8TO9_OR_10(AV_WL16);

  1800.                     }

  1801.                 }

  1802.             } else if(is16BPS(c->srcFormat) && !is16BPS(c->dstFormat)) {

  1803.                 if (!isBE(c->srcFormat)) srcPtr++;

  1804.                 for (i=0; i<height; i++) {

  1805.                     for (j=0; j<length; j++) dstPtr[j] = srcPtr[j<<1];

  1806.                     srcPtr+= srcStride[plane];

  1807.                     dstPtr+= dstStride[plane];

  1808.                 }

  1809.             } else if(!is16BPS(c->srcFormat) && is16BPS(c->dstFormat)) {

  1810.                 for (i=0; i<height; i++) {

  1811.                     for (j=0; j<length; j++) {

  1812.                         dstPtr[ j<<1   ] = srcPtr[j];

  1813.                         dstPtr[(j<<1)+1] = srcPtr[j];

  1814.                     }

  1815.                     srcPtr+= srcStride[plane];

  1816.                     dstPtr+= dstStride[plane];

  1817.                 }

  1818.             } else if(is16BPS(c->srcFormat) && is16BPS(c->dstFormat)

  1819.                   && isBE(c->srcFormat) != isBE(c->dstFormat)) {

  1820. 

  1821.                 for (i=0; i<height; i++) {

  1822.                     for (j=0; j<length; j++)

  1823.                         ((uint16_t*)dstPtr)[j] = av_bswap16(((const uint16_t*)srcPtr)[j]);

  1824.                     srcPtr+= srcStride[plane];

  1825.                     dstPtr+= dstStride[plane];

  1826.                 }

  1827.             } else if (dstStride[plane] == srcStride[plane] &&

  1828.                        srcStride[plane] > 0 && srcStride[plane] == length) {

  1829.                 memcpy(dst[plane] + dstStride[plane]*y, src[plane],

  1830.                        height*dstStride[plane]);

  1831.             } else {

  1832.                 if(is16BPS(c->srcFormat) && is16BPS(c->dstFormat))

  1833.                     length*=2;

  1834.                 for (i=0; i<height; i++) {

  1835.                     memcpy(dstPtr, srcPtr, length);

  1836.                     srcPtr+= srcStride[plane];

  1837.                     dstPtr+= dstStride[plane];

  1838.                 }

  1839.             }

  1840.         }

  1841.     }

  1842.     return srcSliceH;

  1843. }

  1844. 

  1845. int ff_hardcodedcpuflags(void)

  1846. {

  1847.     int flags = 0;

  1848. #if   COMPILE_TEMPLATE_MMX2

  1849.     flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_MMX2;

  1850. #elif COMPILE_TEMPLATE_AMD3DNOW

  1851.     flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_3DNOW;

  1852. #elif COMPILE_TEMPLATE_MMX

  1853.     flags |= SWS_CPU_CAPS_MMX;

  1854. #elif COMPILE_TEMPLATE_ALTIVEC

  1855.     flags |= SWS_CPU_CAPS_ALTIVEC;

  1856. #elif ARCH_BFIN

  1857.     flags |= SWS_CPU_CAPS_BFIN;

  1858. #endif

  1859.     return flags;

  1860. }

  1861. 

  1862. void ff_get_unscaled_swscale(SwsContext *c)

  1863. {

  1864.     const enum PixelFormat srcFormat = c->srcFormat;

  1865.     const enum PixelFormat dstFormat = c->dstFormat;

  1866.     const int flags = c->flags;

  1867.     const int dstH = c->dstH;

  1868.     int needsDither;

  1869. 

  1870.     needsDither= isAnyRGB(dstFormat)

  1871.         &&  c->dstFormatBpp < 24

  1872.         && (c->dstFormatBpp < c->srcFormatBpp || (!isAnyRGB(srcFormat)));

  1873. 

  1874.     /* yv12_to_nv12 */

  1875.     if ((srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUVA420P) && (dstFormat == PIX_FMT_NV12 || dstFormat == PIX_FMT_NV21)) {

  1876.         c->swScale= planarToNv12Wrapper;

  1877.     }

  1878.     /* yuv2bgr */

  1879.     if ((srcFormat==PIX_FMT_YUV420P || srcFormat==PIX_FMT_YUV422P || srcFormat==PIX_FMT_YUVA420P) && isAnyRGB(dstFormat)

  1880.         && !(flags & SWS_ACCURATE_RND) && !(dstH&1)) {

  1881.         c->swScale= ff_yuv2rgb_get_func_ptr(c);

  1882.     }

  1883. 

  1884.     if (srcFormat==PIX_FMT_YUV410P && (dstFormat==PIX_FMT_YUV420P || dstFormat==PIX_FMT_YUVA420P) && !(flags & SWS_BITEXACT)) {

  1885.         c->swScale= yvu9ToYv12Wrapper;

  1886.     }

  1887. 

  1888.     /* bgr24toYV12 */

  1889.     if (srcFormat==PIX_FMT_BGR24 && (dstFormat==PIX_FMT_YUV420P || dstFormat==PIX_FMT_YUVA420P) && !(flags & SWS_ACCURATE_RND))

  1890.         c->swScale= bgr24ToYv12Wrapper;

  1891. 

  1892.     /* RGB/BGR -> RGB/BGR (no dither needed forms) */

  1893.     if (   isAnyRGB(srcFormat)

  1894.         && isAnyRGB(dstFormat)

  1895.         && srcFormat != PIX_FMT_BGR8      && dstFormat != PIX_FMT_BGR8

  1896.         && srcFormat != PIX_FMT_RGB8      && dstFormat != PIX_FMT_RGB8

  1897.         && srcFormat != PIX_FMT_BGR4      && dstFormat != PIX_FMT_BGR4

  1898.         && srcFormat != PIX_FMT_RGB4      && dstFormat != PIX_FMT_RGB4

  1899.         && srcFormat != PIX_FMT_BGR4_BYTE && dstFormat != PIX_FMT_BGR4_BYTE

  1900.         && srcFormat != PIX_FMT_RGB4_BYTE && dstFormat != PIX_FMT_RGB4_BYTE

  1901.         && srcFormat != PIX_FMT_MONOBLACK && dstFormat != PIX_FMT_MONOBLACK

  1902.         && srcFormat != PIX_FMT_MONOWHITE && dstFormat != PIX_FMT_MONOWHITE

  1903.         && srcFormat != PIX_FMT_RGB48LE   && dstFormat != PIX_FMT_RGB48LE

  1904.         && srcFormat != PIX_FMT_RGB48BE   && dstFormat != PIX_FMT_RGB48BE

  1905.         && srcFormat != PIX_FMT_BGR48LE   && dstFormat != PIX_FMT_BGR48LE

  1906.         && srcFormat != PIX_FMT_BGR48BE   && dstFormat != PIX_FMT_BGR48BE

  1907.         && (!needsDither || (c->flags&(SWS_FAST_BILINEAR|SWS_POINT))))

  1908.         c->swScale= rgbToRgbWrapper;

  1909. 

  1910.     if ((usePal(srcFormat) && (

  1911.         dstFormat == PIX_FMT_RGB32   ||

  1912.         dstFormat == PIX_FMT_RGB32_1 ||

  1913.         dstFormat == PIX_FMT_RGB24   ||

  1914.         dstFormat == PIX_FMT_BGR32   ||

  1915.         dstFormat == PIX_FMT_BGR32_1 ||

  1916.         dstFormat == PIX_FMT_BGR24)))

  1917.         c->swScale= palToRgbWrapper;

  1918. 

  1919.     if (srcFormat == PIX_FMT_YUV422P) {

  1920.         if (dstFormat == PIX_FMT_YUYV422)

  1921.             c->swScale= yuv422pToYuy2Wrapper;

  1922.         else if (dstFormat == PIX_FMT_UYVY422)

  1923.             c->swScale= yuv422pToUyvyWrapper;

  1924.     }

  1925. 

  1926.     /* LQ converters if -sws 0 or -sws 4*/

  1927.     if (c->flags&(SWS_FAST_BILINEAR|SWS_POINT)) {

  1928.         /* yv12_to_yuy2 */

  1929.         if (srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUVA420P) {

  1930.             if (dstFormat == PIX_FMT_YUYV422)

  1931.                 c->swScale= planarToYuy2Wrapper;

  1932.             else if (dstFormat == PIX_FMT_UYVY422)

  1933.                 c->swScale= planarToUyvyWrapper;

  1934.         }

  1935.     }

  1936.     if(srcFormat == PIX_FMT_YUYV422 && (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P))

  1937.         c->swScale= yuyvToYuv420Wrapper;

  1938.     if(srcFormat == PIX_FMT_UYVY422 && (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P))

  1939.         c->swScale= uyvyToYuv420Wrapper;

  1940.     if(srcFormat == PIX_FMT_YUYV422 && dstFormat == PIX_FMT_YUV422P)

  1941.         c->swScale= yuyvToYuv422Wrapper;

  1942.     if(srcFormat == PIX_FMT_UYVY422 && dstFormat == PIX_FMT_YUV422P)

  1943.         c->swScale= uyvyToYuv422Wrapper;

  1944. 

  1945. #if COMPILE_ALTIVEC

  1946.     if ((c->flags & SWS_CPU_CAPS_ALTIVEC) &&

  1947.         !(c->flags & SWS_BITEXACT) &&

  1948.         srcFormat == PIX_FMT_YUV420P) {

  1949.         // unscaled YV12 -> packed YUV, we want speed

  1950.         if (dstFormat == PIX_FMT_YUYV422)

  1951.             c->swScale= yv12toyuy2_unscaled_altivec;

  1952.         else if (dstFormat == PIX_FMT_UYVY422)

  1953.             c->swScale= yv12touyvy_unscaled_altivec;

  1954.     }

  1955. #endif

  1956. 

  1957.     /* simple copy */

  1958.     if (  srcFormat == dstFormat

  1959.         || (srcFormat == PIX_FMT_YUVA420P && dstFormat == PIX_FMT_YUV420P)

  1960.         || (srcFormat == PIX_FMT_YUV420P && dstFormat == PIX_FMT_YUVA420P)

  1961.         || (isPlanarYUV(srcFormat) && isGray(dstFormat))

  1962.         || (isPlanarYUV(dstFormat) && isGray(srcFormat))

  1963.         || (isGray(dstFormat) && isGray(srcFormat))

  1964.         || (isPlanarYUV(srcFormat) && isPlanarYUV(dstFormat)

  1965.             && c->chrDstHSubSample == c->chrSrcHSubSample

  1966.             && c->chrDstVSubSample == c->chrSrcVSubSample

  1967.             && dstFormat != PIX_FMT_NV12 && dstFormat != PIX_FMT_NV21

  1968.             && srcFormat != PIX_FMT_NV12 && srcFormat != PIX_FMT_NV21))

  1969.     {

  1970.         if (isPacked(c->srcFormat))

  1971.             c->swScale= packedCopyWrapper;

  1972.         else /* Planar YUV or gray */

  1973.             c->swScale= planarCopyWrapper;

  1974.     }

  1975. #if ARCH_BFIN

  1976.     if (flags & SWS_CPU_CAPS_BFIN)

  1977.         ff_bfin_get_unscaled_swscale (c);

  1978. #endif

  1979. }

  1980. 

  1981. static void reset_ptr(const uint8_t* src[], int format)

  1982. {

  1983.     if(!isALPHA(format))

  1984.         src[3]=NULL;

  1985.     if(!isPlanarYUV(format)) {

  1986.         src[3]=src[2]=NULL;

  1987. 

  1988.         if (!usePal(format))

  1989.             src[1]= NULL;

  1990.     }

  1991. }

  1992. 

  1993. static int check_image_pointers(uint8_t *data[4], enum PixelFormat pix_fmt,

  1994.                                 const int linesizes[4])

  1995. {

  1996.     const AVPixFmtDescriptor *desc = &av_pix_fmt_descriptors[pix_fmt];

  1997.     int i;

  1998. 

  1999.     for (i = 0; i < 4; i++) {

  2000.         int plane = desc->comp[i].plane;

  2001.         if (!data[plane] || !linesizes[plane])

  2002.             return 0;

  2003.     }

  2004. 

  2005.     return 1;

  2006. }

  2007. 

  2008. /**

  2009.  * swscale wrapper, so we don't need to export the SwsContext.

  2010.  * Assumes planar YUV to be in YUV order instead of YVU.

  2011.  */

  2012. int sws_scale(SwsContext *c, const uint8_t* const src[], const int srcStride[], int srcSliceY,

  2013.               int srcSliceH, uint8_t* const dst[], const int dstStride[])

  2014. {

  2015.     int i;

  2016.     const uint8_t* src2[4]= {src[0], src[1], src[2], src[3]};

  2017.     uint8_t* dst2[4]= {dst[0], dst[1], dst[2], dst[3]};

  2018. 

  2019.     // do not mess up sliceDir if we have a "trailing" 0-size slice

  2020.     if (srcSliceH == 0)

  2021.         return 0;

  2022. 

  2023.     if (!check_image_pointers(src, c->srcFormat, srcStride)) {

  2024.         av_log(c, AV_LOG_ERROR, "bad src image pointers\n");

  2025.         return 0;

  2026.     }

  2027.     if (!check_image_pointers(dst, c->dstFormat, dstStride)) {

  2028.         av_log(c, AV_LOG_ERROR, "bad dst image pointers\n");

  2029.         return 0;

  2030.     }

  2031. 

  2032.     if (c->sliceDir == 0 && srcSliceY != 0 && srcSliceY + srcSliceH != c->srcH) {

  2033.         av_log(c, AV_LOG_ERROR, "Slices start in the middle!\n");

  2034.         return 0;

  2035.     }

  2036.     if (c->sliceDir == 0) {

  2037.         if (srcSliceY == 0) c->sliceDir = 1; else c->sliceDir = -1;

  2038.     }

  2039. 

  2040.     if (usePal(c->srcFormat)) {

  2041.         for (i=0; i<256; i++) {

  2042.             int p, r, g, b,y,u,v;

  2043.             if(c->srcFormat == PIX_FMT_PAL8) {

  2044.                 p=((const uint32_t*)(src[1]))[i];

  2045.                 r= (p>>16)&0xFF;

  2046.                 g= (p>> 8)&0xFF;

  2047.                 b=  p     &0xFF;

  2048.             } else if(c->srcFormat == PIX_FMT_RGB8) {

  2049.                 r= (i>>5    )*36;

  2050.                 g= ((i>>2)&7)*36;

  2051.                 b= (i&3     )*85;

  2052.             } else if(c->srcFormat == PIX_FMT_BGR8) {

  2053.                 b= (i>>6    )*85;

  2054.                 g= ((i>>3)&7)*36;

  2055.                 r= (i&7     )*36;

  2056.             } else if(c->srcFormat == PIX_FMT_RGB4_BYTE) {

  2057.                 r= (i>>3    )*255;

  2058.                 g= ((i>>1)&3)*85;

  2059.                 b= (i&1     )*255;

  2060.             } else if(c->srcFormat == PIX_FMT_GRAY8 || c->srcFormat == PIX_FMT_Y400A) {

  2061.                 r = g = b = i;

  2062.             } else {

  2063.                 assert(c->srcFormat == PIX_FMT_BGR4_BYTE);

  2064.                 b= (i>>3    )*255;

  2065.                 g= ((i>>1)&3)*85;

  2066.                 r= (i&1     )*255;

  2067.             }

  2068.             y= av_clip_uint8((RY*r + GY*g + BY*b + ( 33<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);

  2069.             u= av_clip_uint8((RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);

  2070.             v= av_clip_uint8((RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);

  2071.             c->pal_yuv[i]= y + (u<<8) + (v<<16);

  2072. 

  2073.             switch(c->dstFormat) {

  2074.             case PIX_FMT_BGR32:

  2075. #if !HAVE_BIGENDIAN

  2076.             case PIX_FMT_RGB24:

  2077. #endif

  2078.                 c->pal_rgb[i]=  r + (g<<8) + (b<<16);

  2079.                 break;

  2080.             case PIX_FMT_BGR32_1:

  2081. #if HAVE_BIGENDIAN

  2082.             case PIX_FMT_BGR24:

  2083. #endif

  2084.                 c->pal_rgb[i]= (r + (g<<8) + (b<<16)) << 8;

  2085.                 break;

  2086.             case PIX_FMT_RGB32_1:

  2087. #if HAVE_BIGENDIAN

  2088.             case PIX_FMT_RGB24:

  2089. #endif

  2090.                 c->pal_rgb[i]= (b + (g<<8) + (r<<16)) << 8;

  2091.                 break;

  2092.             case PIX_FMT_RGB32:

  2093. #if !HAVE_BIGENDIAN

  2094.             case PIX_FMT_BGR24:

  2095. #endif

  2096.             default:

  2097.                 c->pal_rgb[i]=  b + (g<<8) + (r<<16);

  2098.             }

  2099.         }

  2100.     }

  2101. 

  2102.     // copy strides, so they can safely be modified

  2103.     if (c->sliceDir == 1) {

  2104.         // slices go from top to bottom

  2105.         int srcStride2[4]= {srcStride[0], srcStride[1], srcStride[2], srcStride[3]};

  2106.         int dstStride2[4]= {dstStride[0], dstStride[1], dstStride[2], dstStride[3]};

  2107. 

  2108.         reset_ptr(src2, c->srcFormat);

  2109.         reset_ptr((const uint8_t**)dst2, c->dstFormat);

  2110. 

  2111.         /* reset slice direction at end of frame */

  2112.         if (srcSliceY + srcSliceH == c->srcH)

  2113.             c->sliceDir = 0;

  2114. 

  2115.         return c->swScale(c, src2, srcStride2, srcSliceY, srcSliceH, dst2, dstStride2);

  2116.     } else {

  2117.         // slices go from bottom to top => we flip the image internally

  2118.         int srcStride2[4]= {-srcStride[0], -srcStride[1], -srcStride[2], -srcStride[3]};

  2119.         int dstStride2[4]= {-dstStride[0], -dstStride[1], -dstStride[2], -dstStride[3]};

  2120. 

  2121.         src2[0] += (srcSliceH-1)*srcStride[0];

  2122.         if (!usePal(c->srcFormat))

  2123.             src2[1] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[1];

  2124.         src2[2] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[2];

  2125.         src2[3] += (srcSliceH-1)*srcStride[3];

  2126.         dst2[0] += ( c->dstH                      -1)*dstStride[0];

  2127.         dst2[1] += ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[1];

  2128.         dst2[2] += ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[2];

  2129.         dst2[3] += ( c->dstH                      -1)*dstStride[3];

  2130. 

  2131.         reset_ptr(src2, c->srcFormat);

  2132.         reset_ptr((const uint8_t**)dst2, c->dstFormat);

  2133. 

  2134.         /* reset slice direction at end of frame */

  2135.         if (!srcSliceY)

  2136.             c->sliceDir = 0;

  2137. 

  2138.         return c->swScale(c, src2, srcStride2, c->srcH-srcSliceY-srcSliceH, srcSliceH, dst2, dstStride2);

  2139.     }

  2140. }

  2141. 

  2142. /* Convert the palette to the same packed 32-bit format as the palette */

  2143. void sws_convertPalette8ToPacked32(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)

  2144. {

  2145.     long i;

  2146. 

  2147.     for (i=0; i<num_pixels; i++)

  2148.         ((uint32_t *) dst)[i] = ((const uint32_t *) palette)[src[i]];

  2149. }

  2150. 

  2151. /* Palette format: ABCD -> dst format: ABC */

  2152. void sws_convertPalette8ToPacked24(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)

  2153. {

  2154.     long i;

  2155. 

  2156.     for (i=0; i<num_pixels; i++) {

  2157.         //FIXME slow?

  2158.         dst[0]= palette[src[i]*4+0];

  2159.         dst[1]= palette[src[i]*4+1];

  2160.         dst[2]= palette[src[i]*4+2];

  2161.         dst+= 3;

  2162.     }

  2163. }