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.
21407 Commits
32 Branches
283MB
Tree: b711aaa2d8
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from 'b711aaa2d8'
${ noResults }
FFmpeg/libswscale/swscale.c

3612 lines
126KB
Raw Blame History 

  1. /*

  2.  * Copyright (C) 2001-2003 Michael Niedermayer <michaelni@gmx.at>

  3.  *

  4.  * This file is part of FFmpeg.

  5.  *

  6.  * FFmpeg is free software; you can redistribute it and/or modify

  7.  * it under the terms of the GNU General Public License as published by

  8.  * the Free Software Foundation; either version 2 of the License, or

  9.  * (at your option) any later version.

  10.  *

  11.  * FFmpeg 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

  14.  * GNU General Public License for more details.

  15.  *

  16.  * You should have received a copy of the GNU General Public License

  17.  * along with FFmpeg; if not, write to the Free Software

  18.  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA

  19.  *

  20.  * the C code (not assembly, mmx, ...) of this file can be used

  21.  * under the LGPL license too

  22.  */

  23. 

  24. /*

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

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

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

  28. 

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

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

  31.   x -> x

  32.   YUV9 -> YV12

  33.   YUV9/YV12 -> Y800

  34.   Y800 -> YUV9/YV12

  35.   BGR24 -> BGR32 & RGB24 -> RGB32

  36.   BGR32 -> BGR24 & RGB32 -> RGB24

  37.   BGR15 -> BGR16

  38. */

  39. 

  40. /*

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

  42.  YV12 -> BGR16

  43.  YV12 -> YV12

  44.  BGR15 -> BGR16

  45.  BGR16 -> BGR16

  46.  YVU9 -> YV12

  47. 

  48. untested special converters

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

  50.   YV12/I420 -> YV12/I420

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

  52.   BGR24 -> BGR32 & RGB24 -> RGB32

  53.   BGR32 -> BGR24 & RGB32 -> RGB24

  54.   BGR24 -> YV12

  55. */

  56. 

  57. #define _SVID_SOURCE //needed for MAP_ANONYMOUS

  58. #include <inttypes.h>

  59. #include <string.h>

  60. #include <math.h>

  61. #include <stdio.h>

  62. #include "config.h"

  63. #include <assert.h>

  64. #if HAVE_SYS_MMAN_H

  65. #include <sys/mman.h>

  66. #if defined(MAP_ANON) && !defined(MAP_ANONYMOUS)

  67. #define MAP_ANONYMOUS MAP_ANON

  68. #endif

  69. #endif

  70. #if HAVE_VIRTUALALLOC

  71. #define WIN32_LEAN_AND_MEAN

  72. #include <windows.h>

  73. #endif

  74. #include "swscale.h"

  75. #include "swscale_internal.h"

  76. #include "rgb2rgb.h"

  77. #include "libavutil/intreadwrite.h"

  78. #include "libavutil/x86_cpu.h"

  79. #include "libavutil/avutil.h"

  80. #include "libavutil/bswap.h"

  81. 

  82. unsigned swscale_version(void)

  83. {

  84.     return LIBSWSCALE_VERSION_INT;

  85. }

  86. 

  87. const char * swscale_configuration(void)

  88. {

  89.     return FFMPEG_CONFIGURATION;

  90. }

  91. 

  92. const char * swscale_license(void)

  93. {

  94. #define LICENSE_PREFIX "libswscale license: "

  95.     return LICENSE_PREFIX FFMPEG_LICENSE + sizeof(LICENSE_PREFIX) - 1;

  96. }

  97. 

  98. #undef MOVNTQ

  99. #undef PAVGB

  100. 

  101. //#undef HAVE_MMX2

  102. //#define HAVE_AMD3DNOW

  103. //#undef HAVE_MMX

  104. //#undef ARCH_X86

  105. #define DITHER1XBPP

  106. 

  107. #define FAST_BGR2YV12 // use 7 bit coefficients instead of 15 bit

  108. 

  109. #define RET 0xC3 //near return opcode for x86

  110. 

  111. #ifdef M_PI

  112. #define PI M_PI

  113. #else

  114. #define PI 3.14159265358979323846

  115. #endif

  116. 

  117. #define isSupportedIn(x)    (       \

  118.            (x)==PIX_FMT_YUV420P     \

  119.         || (x)==PIX_FMT_YUVA420P    \

  120.         || (x)==PIX_FMT_YUYV422     \

  121.         || (x)==PIX_FMT_UYVY422     \

  122.         || (x)==PIX_FMT_RGB48BE     \

  123.         || (x)==PIX_FMT_RGB48LE     \

  124.         || (x)==PIX_FMT_RGB32       \

  125.         || (x)==PIX_FMT_RGB32_1     \

  126.         || (x)==PIX_FMT_BGR24       \

  127.         || (x)==PIX_FMT_BGR565      \

  128.         || (x)==PIX_FMT_BGR555      \

  129.         || (x)==PIX_FMT_BGR32       \

  130.         || (x)==PIX_FMT_BGR32_1     \

  131.         || (x)==PIX_FMT_RGB24       \

  132.         || (x)==PIX_FMT_RGB565      \

  133.         || (x)==PIX_FMT_RGB555      \

  134.         || (x)==PIX_FMT_GRAY8       \

  135.         || (x)==PIX_FMT_YUV410P     \

  136.         || (x)==PIX_FMT_YUV440P     \

  137.         || (x)==PIX_FMT_GRAY16BE    \

  138.         || (x)==PIX_FMT_GRAY16LE    \

  139.         || (x)==PIX_FMT_YUV444P     \

  140.         || (x)==PIX_FMT_YUV422P     \

  141.         || (x)==PIX_FMT_YUV411P     \

  142.         || (x)==PIX_FMT_PAL8        \

  143.         || (x)==PIX_FMT_BGR8        \

  144.         || (x)==PIX_FMT_RGB8        \

  145.         || (x)==PIX_FMT_BGR4_BYTE   \

  146.         || (x)==PIX_FMT_RGB4_BYTE   \

  147.         || (x)==PIX_FMT_YUV440P     \

  148.         || (x)==PIX_FMT_MONOWHITE   \

  149.         || (x)==PIX_FMT_MONOBLACK   \

  150.         || (x)==PIX_FMT_YUV420P16LE   \

  151.         || (x)==PIX_FMT_YUV422P16LE   \

  152.         || (x)==PIX_FMT_YUV444P16LE   \

  153.         || (x)==PIX_FMT_YUV420P16BE   \

  154.         || (x)==PIX_FMT_YUV422P16BE   \

  155.         || (x)==PIX_FMT_YUV444P16BE   \

  156.     )

  157. #define isSupportedOut(x)   (       \

  158.            (x)==PIX_FMT_YUV420P     \

  159.         || (x)==PIX_FMT_YUVA420P    \

  160.         || (x)==PIX_FMT_YUYV422     \

  161.         || (x)==PIX_FMT_UYVY422     \

  162.         || (x)==PIX_FMT_YUV444P     \

  163.         || (x)==PIX_FMT_YUV422P     \

  164.         || (x)==PIX_FMT_YUV411P     \

  165.         || isRGB(x)                 \

  166.         || isBGR(x)                 \

  167.         || (x)==PIX_FMT_NV12        \

  168.         || (x)==PIX_FMT_NV21        \

  169.         || (x)==PIX_FMT_GRAY16BE    \

  170.         || (x)==PIX_FMT_GRAY16LE    \

  171.         || (x)==PIX_FMT_GRAY8       \

  172.         || (x)==PIX_FMT_YUV410P     \

  173.         || (x)==PIX_FMT_YUV440P     \

  174.         || (x)==PIX_FMT_YUV420P16LE   \

  175.         || (x)==PIX_FMT_YUV422P16LE   \

  176.         || (x)==PIX_FMT_YUV444P16LE   \

  177.         || (x)==PIX_FMT_YUV420P16BE   \

  178.         || (x)==PIX_FMT_YUV422P16BE   \

  179.         || (x)==PIX_FMT_YUV444P16BE   \

  180.     )

  181. #define isPacked(x)         (       \

  182.            (x)==PIX_FMT_PAL8        \

  183.         || (x)==PIX_FMT_YUYV422     \

  184.         || (x)==PIX_FMT_UYVY422     \

  185.         || isRGB(x)                 \

  186.         || isBGR(x)                 \

  187.     )

  188. #define usePal(x)           (       \

  189.            (x)==PIX_FMT_PAL8        \

  190.         || (x)==PIX_FMT_BGR4_BYTE   \

  191.         || (x)==PIX_FMT_RGB4_BYTE   \

  192.         || (x)==PIX_FMT_BGR8        \

  193.         || (x)==PIX_FMT_RGB8        \

  194.     )

  195. 

  196. #define RGB2YUV_SHIFT 15

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

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

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

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

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

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

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

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

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

  206. 

  207. extern const int32_t ff_yuv2rgb_coeffs[8][4];

  208. 

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

  210.     {0.7152, 0.0722, 0.2126, -0.386, 0.5, -0.115, -0.454, -0.046, 0.5},

  211.     {0.7152, 0.0722, 0.2126, -0.386, 0.5, -0.115, -0.454, -0.046, 0.5},

  212.     {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5},

  213.     {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5},

  214.     {0.59  , 0.11  , 0.30  , -0.331, 0.5, -0.169, -0.421, -0.079, 0.5}, //FCC

  215.     {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5},

  216.     {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5}, //SMPTE 170M

  217.     {0.701 , 0.087 , 0.212 , -0.384, 0.5  -0.116, -0.445, -0.055, 0.5}, //SMPTE 240M

  218. };

  219. 

  220. /*

  221. NOTES

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

  223. 

  224. TODO

  225. more intelligent misalignment avoidance for the horizontal scaler

  226. write special vertical cubic upscale version

  227. optimize C code (YV12 / minmax)

  228. add support for packed pixel YUV input & output

  229. add support for Y8 output

  230. optimize BGR24 & BGR32

  231. add BGR4 output support

  232. write special BGR->BGR scaler

  233. */

  234. 

  235. #if ARCH_X86 && CONFIG_GPL

  236. DECLARE_ASM_CONST(8, uint64_t, bF8)=       0xF8F8F8F8F8F8F8F8LL;

  237. DECLARE_ASM_CONST(8, uint64_t, bFC)=       0xFCFCFCFCFCFCFCFCLL;

  238. DECLARE_ASM_CONST(8, uint64_t, w10)=       0x0010001000100010LL;

  239. DECLARE_ASM_CONST(8, uint64_t, w02)=       0x0002000200020002LL;

  240. DECLARE_ASM_CONST(8, uint64_t, bm00001111)=0x00000000FFFFFFFFLL;

  241. DECLARE_ASM_CONST(8, uint64_t, bm00000111)=0x0000000000FFFFFFLL;

  242. DECLARE_ASM_CONST(8, uint64_t, bm11111000)=0xFFFFFFFFFF000000LL;

  243. DECLARE_ASM_CONST(8, uint64_t, bm01010101)=0x00FF00FF00FF00FFLL;

  244. 

  245. const DECLARE_ALIGNED(8, uint64_t, ff_dither4[2]) = {

  246.         0x0103010301030103LL,

  247.         0x0200020002000200LL,};

  248. 

  249. const DECLARE_ALIGNED(8, uint64_t, ff_dither8[2]) = {

  250.         0x0602060206020602LL,

  251.         0x0004000400040004LL,};

  252. 

  253. DECLARE_ASM_CONST(8, uint64_t, b16Mask)=   0x001F001F001F001FLL;

  254. DECLARE_ASM_CONST(8, uint64_t, g16Mask)=   0x07E007E007E007E0LL;

  255. DECLARE_ASM_CONST(8, uint64_t, r16Mask)=   0xF800F800F800F800LL;

  256. DECLARE_ASM_CONST(8, uint64_t, b15Mask)=   0x001F001F001F001FLL;

  257. DECLARE_ASM_CONST(8, uint64_t, g15Mask)=   0x03E003E003E003E0LL;

  258. DECLARE_ASM_CONST(8, uint64_t, r15Mask)=   0x7C007C007C007C00LL;

  259. 

  260. DECLARE_ALIGNED(8, const uint64_t, ff_M24A)         = 0x00FF0000FF0000FFLL;

  261. DECLARE_ALIGNED(8, const uint64_t, ff_M24B)         = 0xFF0000FF0000FF00LL;

  262. DECLARE_ALIGNED(8, const uint64_t, ff_M24C)         = 0x0000FF0000FF0000LL;

  263. 

  264. #ifdef FAST_BGR2YV12

  265. DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YCoeff)   = 0x000000210041000DULL;

  266. DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UCoeff)   = 0x0000FFEEFFDC0038ULL;

  267. DECLARE_ALIGNED(8, const uint64_t, ff_bgr2VCoeff)   = 0x00000038FFD2FFF8ULL;

  268. #else

  269. DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YCoeff)   = 0x000020E540830C8BULL;

  270. DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UCoeff)   = 0x0000ED0FDAC23831ULL;

  271. DECLARE_ALIGNED(8, const uint64_t, ff_bgr2VCoeff)   = 0x00003831D0E6F6EAULL;

  272. #endif /* FAST_BGR2YV12 */

  273. DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YOffset)  = 0x1010101010101010ULL;

  274. DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UVOffset) = 0x8080808080808080ULL;

  275. DECLARE_ALIGNED(8, const uint64_t, ff_w1111)        = 0x0001000100010001ULL;

  276. 

  277. DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toY1Coeff) = 0x0C88000040870C88ULL;

  278. DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toY2Coeff) = 0x20DE4087000020DEULL;

  279. DECLARE_ASM_CONST(8, uint64_t, ff_rgb24toY1Coeff) = 0x20DE0000408720DEULL;

  280. DECLARE_ASM_CONST(8, uint64_t, ff_rgb24toY2Coeff) = 0x0C88408700000C88ULL;

  281. DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toYOffset) = 0x0008400000084000ULL;

  282. 

  283. DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toUV[2][4]) = {

  284.     {0x38380000DAC83838ULL, 0xECFFDAC80000ECFFULL, 0xF6E40000D0E3F6E4ULL, 0x3838D0E300003838ULL},

  285.     {0xECFF0000DAC8ECFFULL, 0x3838DAC800003838ULL, 0x38380000D0E33838ULL, 0xF6E4D0E30000F6E4ULL},

  286. };

  287. 

  288. DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toUVOffset)= 0x0040400000404000ULL;

  289. 

  290. #endif /* ARCH_X86 && CONFIG_GPL */

  291. 

  292. // clipping helper table for C implementations:

  293. static unsigned char clip_table[768];

  294. 

  295. static SwsVector *sws_getConvVec(SwsVector *a, SwsVector *b);

  296. 

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

  298. {  1,   3,   1,   3,   1,   3,   1,   3, },

  299. {  2,   0,   2,   0,   2,   0,   2,   0, },

  300. };

  301. 

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

  303. {  6,   2,   6,   2,   6,   2,   6,   2, },

  304. {  0,   4,   0,   4,   0,   4,   0,   4, },

  305. };

  306. 

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

  308. { 17,   9,  23,  15,  16,   8,  22,  14, },

  309. {  5,  29,   3,  27,   4,  28,   2,  26, },

  310. { 21,  13,  19,  11,  20,  12,  18,  10, },

  311. {  0,  24,   6,  30,   1,  25,   7,  31, },

  312. { 16,   8,  22,  14,  17,   9,  23,  15, },

  313. {  4,  28,   2,  26,   5,  29,   3,  27, },

  314. { 20,  12,  18,  10,  21,  13,  19,  11, },

  315. {  1,  25,   7,  31,   0,  24,   6,  30, },

  316. };

  317. 

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

  319. {  0,  55,  14,  68,   3,  58,  17,  72, },

  320. { 37,  18,  50,  32,  40,  22,  54,  35, },

  321. {  9,  64,   5,  59,  13,  67,   8,  63, },

  322. { 46,  27,  41,  23,  49,  31,  44,  26, },

  323. {  2,  57,  16,  71,   1,  56,  15,  70, },

  324. { 39,  21,  52,  34,  38,  19,  51,  33, },

  325. { 11,  66,   7,  62,  10,  65,   6,  60, },

  326. { 48,  30,  43,  25,  47,  29,  42,  24, },

  327. };

  328. 

  329. #if 1

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

  331. {117,  62, 158, 103, 113,  58, 155, 100, },

  332. { 34, 199,  21, 186,  31, 196,  17, 182, },

  333. {144,  89, 131,  76, 141,  86, 127,  72, },

  334. {  0, 165,  41, 206,  10, 175,  52, 217, },

  335. {110,  55, 151,  96, 120,  65, 162, 107, },

  336. { 28, 193,  14, 179,  38, 203,  24, 189, },

  337. {138,  83, 124,  69, 148,  93, 134,  79, },

  338. {  7, 172,  48, 213,   3, 168,  45, 210, },

  339. };

  340. #elif 1

  341. // tries to correct a gamma of 1.5

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

  343. {  0, 143,  18, 200,   2, 156,  25, 215, },

  344. { 78,  28, 125,  64,  89,  36, 138,  74, },

  345. { 10, 180,   3, 161,  16, 195,   8, 175, },

  346. {109,  51,  93,  38, 121,  60, 105,  47, },

  347. {  1, 152,  23, 210,   0, 147,  20, 205, },

  348. { 85,  33, 134,  71,  81,  30, 130,  67, },

  349. { 14, 190,   6, 171,  12, 185,   5, 166, },

  350. {117,  57, 101,  44, 113,  54,  97,  41, },

  351. };

  352. #elif 1

  353. // tries to correct a gamma of 2.0

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

  355. {  0, 124,   8, 193,   0, 140,  12, 213, },

  356. { 55,  14, 104,  42,  66,  19, 119,  52, },

  357. {  3, 168,   1, 145,   6, 187,   3, 162, },

  358. { 86,  31,  70,  21,  99,  39,  82,  28, },

  359. {  0, 134,  11, 206,   0, 129,   9, 200, },

  360. { 62,  17, 114,  48,  58,  16, 109,  45, },

  361. {  5, 181,   2, 157,   4, 175,   1, 151, },

  362. { 95,  36,  78,  26,  90,  34,  74,  24, },

  363. };

  364. #else

  365. // tries to correct a gamma of 2.5

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

  367. {  0, 107,   3, 187,   0, 125,   6, 212, },

  368. { 39,   7,  86,  28,  49,  11, 102,  36, },

  369. {  1, 158,   0, 131,   3, 180,   1, 151, },

  370. { 68,  19,  52,  12,  81,  25,  64,  17, },

  371. {  0, 119,   5, 203,   0, 113,   4, 195, },

  372. { 45,   9,  96,  33,  42,   8,  91,  30, },

  373. {  2, 172,   1, 144,   2, 165,   0, 137, },

  374. { 77,  23,  60,  15,  72,  21,  56,  14, },

  375. };

  376. #endif

  377. 

  378. const char *sws_format_name(enum PixelFormat format)

  379. {

  380.     switch (format) {

  381.     case PIX_FMT_YUV420P:

  382.         return "yuv420p";

  383.     case PIX_FMT_YUVA420P:

  384.         return "yuva420p";

  385.     case PIX_FMT_YUYV422:

  386.         return "yuyv422";

  387.     case PIX_FMT_RGB24:

  388.         return "rgb24";

  389.     case PIX_FMT_BGR24:

  390.         return "bgr24";

  391.     case PIX_FMT_YUV422P:

  392.         return "yuv422p";

  393.     case PIX_FMT_YUV444P:

  394.         return "yuv444p";

  395.     case PIX_FMT_RGB32:

  396.         return "rgb32";

  397.     case PIX_FMT_YUV410P:

  398.         return "yuv410p";

  399.     case PIX_FMT_YUV411P:

  400.         return "yuv411p";

  401.     case PIX_FMT_RGB565:

  402.         return "rgb565";

  403.     case PIX_FMT_RGB555:

  404.         return "rgb555";

  405.     case PIX_FMT_GRAY16BE:

  406.         return "gray16be";

  407.     case PIX_FMT_GRAY16LE:

  408.         return "gray16le";

  409.     case PIX_FMT_GRAY8:

  410.         return "gray8";

  411.     case PIX_FMT_MONOWHITE:

  412.         return "mono white";

  413.     case PIX_FMT_MONOBLACK:

  414.         return "mono black";

  415.     case PIX_FMT_PAL8:

  416.         return "Palette";

  417.     case PIX_FMT_YUVJ420P:

  418.         return "yuvj420p";

  419.     case PIX_FMT_YUVJ422P:

  420.         return "yuvj422p";

  421.     case PIX_FMT_YUVJ444P:

  422.         return "yuvj444p";

  423.     case PIX_FMT_XVMC_MPEG2_MC:

  424.         return "xvmc_mpeg2_mc";

  425.     case PIX_FMT_XVMC_MPEG2_IDCT:

  426.         return "xvmc_mpeg2_idct";

  427.     case PIX_FMT_UYVY422:

  428.         return "uyvy422";

  429.     case PIX_FMT_UYYVYY411:

  430.         return "uyyvyy411";

  431.     case PIX_FMT_RGB32_1:

  432.         return "rgb32x";

  433.     case PIX_FMT_BGR32_1:

  434.         return "bgr32x";

  435.     case PIX_FMT_BGR32:

  436.         return "bgr32";

  437.     case PIX_FMT_BGR565:

  438.         return "bgr565";

  439.     case PIX_FMT_BGR555:

  440.         return "bgr555";

  441.     case PIX_FMT_BGR8:

  442.         return "bgr8";

  443.     case PIX_FMT_BGR4:

  444.         return "bgr4";

  445.     case PIX_FMT_BGR4_BYTE:

  446.         return "bgr4 byte";

  447.     case PIX_FMT_RGB8:

  448.         return "rgb8";

  449.     case PIX_FMT_RGB4:

  450.         return "rgb4";

  451.     case PIX_FMT_RGB4_BYTE:

  452.         return "rgb4 byte";

  453.     case PIX_FMT_RGB48BE:

  454.         return "rgb48be";

  455.     case PIX_FMT_RGB48LE:

  456.         return "rgb48le";

  457.     case PIX_FMT_NV12:

  458.         return "nv12";

  459.     case PIX_FMT_NV21:

  460.         return "nv21";

  461.     case PIX_FMT_YUV440P:

  462.         return "yuv440p";

  463.     case PIX_FMT_VDPAU_H264:

  464.         return "vdpau_h264";

  465.     case PIX_FMT_VDPAU_MPEG1:

  466.         return "vdpau_mpeg1";

  467.     case PIX_FMT_VDPAU_MPEG2:

  468.         return "vdpau_mpeg2";

  469.     case PIX_FMT_VDPAU_WMV3:

  470.         return "vdpau_wmv3";

  471.     case PIX_FMT_VDPAU_VC1:

  472.         return "vdpau_vc1";

  473.     case PIX_FMT_VDPAU_MPEG4:

  474.         return "vdpau_mpeg4";

  475.     case PIX_FMT_YUV420P16LE:

  476.         return "yuv420p16le";

  477.     case PIX_FMT_YUV422P16LE:

  478.         return "yuv422p16le";

  479.     case PIX_FMT_YUV444P16LE:

  480.         return "yuv444p16le";

  481.     case PIX_FMT_YUV420P16BE:

  482.         return "yuv420p16be";

  483.     case PIX_FMT_YUV422P16BE:

  484.         return "yuv422p16be";

  485.     case PIX_FMT_YUV444P16BE:

  486.         return "yuv444p16be";

  487.     default:

  488.         return "Unknown format";

  489.     }

  490. }

  491. 

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

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

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

  495.                                                     int dstW, int chrDstW, int big_endian)

  496. {

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

  498.     int i;

  499. 

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

  501.         int val = 1 << 10;

  502.         int j;

  503. 

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

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

  506. 

  507.         if (big_endian) {

  508.             AV_WB16(&dest[i], av_clip_uint16(val >> 11));

  509.         } else {

  510.             AV_WL16(&dest[i], av_clip_uint16(val >> 11));

  511.         }

  512.     }

  513. 

  514.     if (uDest) {

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

  516.             int u = 1 << 10;

  517.             int v = 1 << 10;

  518.             int j;

  519. 

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

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

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

  523.             }

  524. 

  525.             if (big_endian) {

  526.                 AV_WB16(&uDest[i], av_clip_uint16(u >> 11));

  527.                 AV_WB16(&vDest[i], av_clip_uint16(v >> 11));

  528.             } else {

  529.                 AV_WL16(&uDest[i], av_clip_uint16(u >> 11));

  530.                 AV_WL16(&vDest[i], av_clip_uint16(v >> 11));

  531.             }

  532.         }

  533.     }

  534. 

  535.     if (CONFIG_SWSCALE_ALPHA && aDest) {

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

  537.             int val = 1 << 10;

  538.             int j;

  539. 

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

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

  542. 

  543.             if (big_endian) {

  544.                 AV_WB16(&aDest[i], av_clip_uint16(val >> 11));

  545.             } else {

  546.                 AV_WL16(&aDest[i], av_clip_uint16(val >> 11));

  547.             }

  548.         }

  549.     }

  550. }

  551. 

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

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

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

  555.                                  enum PixelFormat dstFormat)

  556. {

  557.     if (isBE(dstFormat)) {

  558.         yuv2yuvX16inC_template(lumFilter, lumSrc, lumFilterSize,

  559.                                chrFilter, chrSrc, chrFilterSize,

  560.                                alpSrc,

  561.                                dest, uDest, vDest, aDest,

  562.                                dstW, chrDstW, 1);

  563.     } else {

  564.         yuv2yuvX16inC_template(lumFilter, lumSrc, lumFilterSize,

  565.                                chrFilter, chrSrc, chrFilterSize,

  566.                                alpSrc,

  567.                                dest, uDest, vDest, aDest,

  568.                                dstW, chrDstW, 0);

  569.     }

  570. }

  571. 

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

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

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

  575. {

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

  577.     int i;

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

  579.         int val=1<<18;

  580.         int j;

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

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

  583. 

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

  585.     }

  586. 

  587.     if (uDest)

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

  589.             int u=1<<18;

  590.             int v=1<<18;

  591.             int j;

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

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

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

  595.             }

  596. 

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

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

  599.         }

  600. 

  601.     if (CONFIG_SWSCALE_ALPHA && aDest)

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

  603.             int val=1<<18;

  604.             int j;

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

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

  607. 

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

  609.         }

  610. 

  611. }

  612. 

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

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

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

  616. {

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

  618.     int i;

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

  620.         int val=1<<18;

  621.         int j;

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

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

  624. 

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

  626.     }

  627. 

  628.     if (!uDest)

  629.         return;

  630. 

  631.     if (dstFormat == PIX_FMT_NV12)

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

  633.             int u=1<<18;

  634.             int v=1<<18;

  635.             int j;

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

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

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

  639.             }

  640. 

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

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

  643.         }

  644.     else

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

  646.             int u=1<<18;

  647.             int v=1<<18;

  648.             int j;

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

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

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

  652.             }

  653. 

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

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

  656.         }

  657. }

  658. 

  659. #define YSCALE_YUV_2_PACKEDX_NOCLIP_C(type,alpha) \

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

  661.         int j;\

  662.         int Y1 = 1<<18;\

  663.         int Y2 = 1<<18;\

  664.         int U  = 1<<18;\

  665.         int V  = 1<<18;\

  666.         int av_unused A1, A2;\

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

  668.         const int i2= 2*i;\

  669.         \

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

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

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

  673.         }\

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

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

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

  677.         }\

  678.         Y1>>=19;\

  679.         Y2>>=19;\

  680.         U >>=19;\

  681.         V >>=19;\

  682.         if (alpha) {\

  683.             A1 = 1<<18;\

  684.             A2 = 1<<18;\

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

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

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

  688.             }\

  689.             A1>>=19;\

  690.             A2>>=19;\

  691.         }\

  692. 

  693. #define YSCALE_YUV_2_PACKEDX_C(type,alpha) \

  694.         YSCALE_YUV_2_PACKEDX_NOCLIP_C(type,alpha)\

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

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

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

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

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

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

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

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

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

  704.         }\

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

  706.             A1=av_clip_uint8(A1);\

  707.             A2=av_clip_uint8(A2);\

  708.         }

  709. 

  710. #define YSCALE_YUV_2_PACKEDX_FULL_C(rnd,alpha) \

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

  712.         int j;\

  713.         int Y = 0;\

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

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

  716.         int av_unused A;\

  717.         int R,G,B;\

  718.         \

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

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

  721.         }\

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

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

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

  725.         }\

  726.         Y >>=10;\

  727.         U >>=10;\

  728.         V >>=10;\

  729.         if (alpha) {\

  730.             A = rnd;\

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

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

  733.             A >>=19;\

  734.             if (A&256)\

  735.                 A = av_clip_uint8(A);\

  736.         }\

  737. 

  738. #define YSCALE_YUV_2_RGBX_FULL_C(rnd,alpha) \

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

  740.         Y-= c->yuv2rgb_y_offset;\

  741.         Y*= c->yuv2rgb_y_coeff;\

  742.         Y+= rnd;\

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

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

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

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

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

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

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

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

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

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

  753.         }\

  754. 

  755. 

  756. #define YSCALE_YUV_2_GRAY16_C \

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

  758.         int j;\

  759.         int Y1 = 1<<18;\

  760.         int Y2 = 1<<18;\

  761.         int U  = 1<<18;\

  762.         int V  = 1<<18;\

  763.         \

  764.         const int i2= 2*i;\

  765.         \

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

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

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

  769.         }\

  770.         Y1>>=11;\

  771.         Y2>>=11;\

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

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

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

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

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

  777.         }

  778. 

  779. #define YSCALE_YUV_2_RGBX_C(type,alpha) \

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

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

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

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

  784. 

  785. #define YSCALE_YUV_2_PACKED2_C(type,alpha)   \

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

  787.         const int i2= 2*i;       \

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

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

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

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

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

  793.         int av_unused A1, A2;                                         \

  794.         if (alpha) {\

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

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

  797.         }\

  798. 

  799. #define YSCALE_YUV_2_GRAY16_2_C   \

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

  801.         const int i2= 2*i;       \

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

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

  804. 

  805. #define YSCALE_YUV_2_RGB2_C(type,alpha) \

  806.     YSCALE_YUV_2_PACKED2_C(type,alpha)\

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

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

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

  810. 

  811. #define YSCALE_YUV_2_PACKED1_C(type,alpha) \

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

  813.         const int i2= 2*i;\

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

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

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

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

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

  819.         int av_unused A1, A2;\

  820.         if (alpha) {\

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

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

  823.         }\

  824. 

  825. #define YSCALE_YUV_2_GRAY16_1_C \

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

  827.         const int i2= 2*i;\

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

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

  830. 

  831. #define YSCALE_YUV_2_RGB1_C(type,alpha) \

  832.     YSCALE_YUV_2_PACKED1_C(type,alpha)\

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

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

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

  836. 

  837. #define YSCALE_YUV_2_PACKED1B_C(type,alpha) \

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

  839.         const int i2= 2*i;\

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

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

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

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

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

  845.         int av_unused A1, A2;\

  846.         if (alpha) {\

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

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

  849.         }\

  850. 

  851. #define YSCALE_YUV_2_RGB1B_C(type,alpha) \

  852.     YSCALE_YUV_2_PACKED1B_C(type,alpha)\

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

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

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

  856. 

  857. #define YSCALE_YUV_2_MONO2_C \

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

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

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

  861.         int acc;\

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

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

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

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

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

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

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

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

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

  871.         dest++;\

  872.     }\

  873. 

  874. 

  875. #define YSCALE_YUV_2_MONOX_C \

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

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

  878.     int acc=0;\

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

  880.         int j;\

  881.         int Y1=1<<18;\

  882.         int Y2=1<<18;\

  883. \

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

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

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

  887.         }\

  888.         Y1>>=19;\

  889.         Y2>>=19;\

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

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

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

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

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

  895.         }\

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

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

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

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

  900.             dest++;\

  901.         }\

  902.     }

  903. 

  904. 

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

  906.     switch(c->dstFormat) {\

  907.     case PIX_FMT_RGB48BE:\

  908.     case PIX_FMT_RGB48LE:\

  909.         func(uint8_t,0)\

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

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

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

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

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

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

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

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

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

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

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

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

  922.             dest+=12;\

  923.         }\

  924.         break;\

  925.     case PIX_FMT_RGBA:\

  926.     case PIX_FMT_BGRA:\

  927.         if (CONFIG_SMALL) {\

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

  929.             func(uint32_t,needAlpha)\

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

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

  932.             }\

  933.         } else {\

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

  935.                 func(uint32_t,1)\

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

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

  938.                 }\

  939.             } else {\

  940.                 func(uint32_t,0)\

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

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

  943.                 }\

  944.             }\

  945.         }\

  946.         break;\

  947.     case PIX_FMT_ARGB:\

  948.     case PIX_FMT_ABGR:\

  949.         if (CONFIG_SMALL) {\

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

  951.             func(uint32_t,needAlpha)\

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

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

  954.             }\

  955.         } else {\

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

  957.                 func(uint32_t,1)\

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

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

  960.                 }\

  961.             } else {\

  962.                 func(uint32_t,0)\

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

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

  965.                 }\

  966.             }\

  967.         }                \

  968.         break;\

  969.     case PIX_FMT_RGB24:\

  970.         func(uint8_t,0)\

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

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

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

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

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

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

  977.             dest+=6;\

  978.         }\

  979.         break;\

  980.     case PIX_FMT_BGR24:\

  981.         func(uint8_t,0)\

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

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

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

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

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

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

  988.             dest+=6;\

  989.         }\

  990.         break;\

  991.     case PIX_FMT_RGB565:\

  992.     case PIX_FMT_BGR565:\

  993.         {\

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

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

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

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

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

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

  1000.             func(uint16_t,0)\

  1001.                 ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\

  1002.                 ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\

  1003.             }\

  1004.         }\

  1005.         break;\

  1006.     case PIX_FMT_RGB555:\

  1007.     case PIX_FMT_BGR555:\

  1008.         {\

  1009.             const int dr1= dither_2x2_8[y&1    ][0];\

  1010.             const int dg1= dither_2x2_8[y&1    ][1];\

  1011.             const int db1= dither_2x2_8[(y&1)^1][0];\

  1012.             const int dr2= dither_2x2_8[y&1    ][1];\

  1013.             const int dg2= dither_2x2_8[y&1    ][0];\

  1014.             const int db2= dither_2x2_8[(y&1)^1][1];\

  1015.             func(uint16_t,0)\

  1016.                 ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\

  1017.                 ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\

  1018.             }\

  1019.         }\

  1020.         break;\

  1021.     case PIX_FMT_RGB8:\

  1022.     case PIX_FMT_BGR8:\

  1023.         {\

  1024.             const uint8_t * const d64= dither_8x8_73[y&7];\

  1025.             const uint8_t * const d32= dither_8x8_32[y&7];\

  1026.             func(uint8_t,0)\

  1027.                 ((uint8_t*)dest)[i2+0]= r[Y1+d32[(i2+0)&7]] + g[Y1+d32[(i2+0)&7]] + b[Y1+d64[(i2+0)&7]];\

  1028.                 ((uint8_t*)dest)[i2+1]= r[Y2+d32[(i2+1)&7]] + g[Y2+d32[(i2+1)&7]] + b[Y2+d64[(i2+1)&7]];\

  1029.             }\

  1030.         }\

  1031.         break;\

  1032.     case PIX_FMT_RGB4:\

  1033.     case PIX_FMT_BGR4:\

  1034.         {\

  1035.             const uint8_t * const d64= dither_8x8_73 [y&7];\

  1036.             const uint8_t * const d128=dither_8x8_220[y&7];\

  1037.             func(uint8_t,0)\

  1038.                 ((uint8_t*)dest)[i]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]]\

  1039.                                  + ((r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]])<<4);\

  1040.             }\

  1041.         }\

  1042.         break;\

  1043.     case PIX_FMT_RGB4_BYTE:\

  1044.     case PIX_FMT_BGR4_BYTE:\

  1045.         {\

  1046.             const uint8_t * const d64= dither_8x8_73 [y&7];\

  1047.             const uint8_t * const d128=dither_8x8_220[y&7];\

  1048.             func(uint8_t,0)\

  1049.                 ((uint8_t*)dest)[i2+0]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]];\

  1050.                 ((uint8_t*)dest)[i2+1]= r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]];\

  1051.             }\

  1052.         }\

  1053.         break;\

  1054.     case PIX_FMT_MONOBLACK:\

  1055.     case PIX_FMT_MONOWHITE:\

  1056.         {\

  1057.             func_monoblack\

  1058.         }\

  1059.         break;\

  1060.     case PIX_FMT_YUYV422:\

  1061.         func2\

  1062.             ((uint8_t*)dest)[2*i2+0]= Y1;\

  1063.             ((uint8_t*)dest)[2*i2+1]= U;\

  1064.             ((uint8_t*)dest)[2*i2+2]= Y2;\

  1065.             ((uint8_t*)dest)[2*i2+3]= V;\

  1066.         }                \

  1067.         break;\

  1068.     case PIX_FMT_UYVY422:\

  1069.         func2\

  1070.             ((uint8_t*)dest)[2*i2+0]= U;\

  1071.             ((uint8_t*)dest)[2*i2+1]= Y1;\

  1072.             ((uint8_t*)dest)[2*i2+2]= V;\

  1073.             ((uint8_t*)dest)[2*i2+3]= Y2;\

  1074.         }                \

  1075.         break;\

  1076.     case PIX_FMT_GRAY16BE:\

  1077.         func_g16\

  1078.             ((uint8_t*)dest)[2*i2+0]= Y1>>8;\

  1079.             ((uint8_t*)dest)[2*i2+1]= Y1;\

  1080.             ((uint8_t*)dest)[2*i2+2]= Y2>>8;\

  1081.             ((uint8_t*)dest)[2*i2+3]= Y2;\

  1082.         }                \

  1083.         break;\

  1084.     case PIX_FMT_GRAY16LE:\

  1085.         func_g16\

  1086.             ((uint8_t*)dest)[2*i2+0]= Y1;\

  1087.             ((uint8_t*)dest)[2*i2+1]= Y1>>8;\

  1088.             ((uint8_t*)dest)[2*i2+2]= Y2;\

  1089.             ((uint8_t*)dest)[2*i2+3]= Y2>>8;\

  1090.         }                \

  1091.         break;\

  1092.     }\

  1093. 

  1094. 

  1095. static inline void yuv2packedXinC(SwsContext *c, const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,

  1096.                                   const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,

  1097.                                   const int16_t **alpSrc, uint8_t *dest, int dstW, int y)

  1098. {

  1099.     int i;

  1100.     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)

  1101. }

  1102. 

  1103. static inline void yuv2rgbXinC_full(SwsContext *c, const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,

  1104.                                     const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,

  1105.                                     const int16_t **alpSrc, uint8_t *dest, int dstW, int y)

  1106. {

  1107.     int i;

  1108.     int step= fmt_depth(c->dstFormat)/8;

  1109.     int aidx= 3;

  1110. 

  1111.     switch(c->dstFormat) {

  1112.     case PIX_FMT_ARGB:

  1113.         dest++;

  1114.         aidx= 0;

  1115.     case PIX_FMT_RGB24:

  1116.         aidx--;

  1117.     case PIX_FMT_RGBA:

  1118.         if (CONFIG_SMALL) {

  1119.             int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;

  1120.             YSCALE_YUV_2_RGBX_FULL_C(1<<21, needAlpha)

  1121.                 dest[aidx]= needAlpha ? A : 255;

  1122.                 dest[0]= R>>22;

  1123.                 dest[1]= G>>22;

  1124.                 dest[2]= B>>22;

  1125.                 dest+= step;

  1126.             }

  1127.         } else {

  1128.             if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {

  1129.                 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 1)

  1130.                     dest[aidx]= A;

  1131.                     dest[0]= R>>22;

  1132.                     dest[1]= G>>22;

  1133.                     dest[2]= B>>22;

  1134.                     dest+= step;

  1135.                 }

  1136.             } else {

  1137.                 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 0)

  1138.                     dest[aidx]= 255;

  1139.                     dest[0]= R>>22;

  1140.                     dest[1]= G>>22;

  1141.                     dest[2]= B>>22;

  1142.                     dest+= step;

  1143.                 }

  1144.             }

  1145.         }

  1146.         break;

  1147.     case PIX_FMT_ABGR:

  1148.         dest++;

  1149.         aidx= 0;

  1150.     case PIX_FMT_BGR24:

  1151.         aidx--;

  1152.     case PIX_FMT_BGRA:

  1153.         if (CONFIG_SMALL) {

  1154.             int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;

  1155.             YSCALE_YUV_2_RGBX_FULL_C(1<<21, needAlpha)

  1156.                 dest[aidx]= needAlpha ? A : 255;

  1157.                 dest[0]= B>>22;

  1158.                 dest[1]= G>>22;

  1159.                 dest[2]= R>>22;

  1160.                 dest+= step;

  1161.             }

  1162.         } else {

  1163.             if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {

  1164.                 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 1)

  1165.                     dest[aidx]= A;

  1166.                     dest[0]= B>>22;

  1167.                     dest[1]= G>>22;

  1168.                     dest[2]= R>>22;

  1169.                     dest+= step;

  1170.                 }

  1171.             } else {

  1172.                 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 0)

  1173.                     dest[aidx]= 255;

  1174.                     dest[0]= B>>22;

  1175.                     dest[1]= G>>22;

  1176.                     dest[2]= R>>22;

  1177.                     dest+= step;

  1178.                 }

  1179.             }

  1180.         }

  1181.         break;

  1182.     default:

  1183.         assert(0);

  1184.     }

  1185. }

  1186. 

  1187. static void fillPlane(uint8_t* plane, int stride, int width, int height, int y, uint8_t val)

  1188. {

  1189.     int i;

  1190.     uint8_t *ptr = plane + stride*y;

  1191.     for (i=0; i<height; i++) {

  1192.         memset(ptr, val, width);

  1193.         ptr += stride;

  1194.     }

  1195. }

  1196. 

  1197. static inline void rgb48ToY(uint8_t *dst, const uint8_t *src, int width)

  1198. {

  1199.     int i;

  1200.     for (i = 0; i < width; i++) {

  1201.         int r = src[i*6+0];

  1202.         int g = src[i*6+2];

  1203.         int b = src[i*6+4];

  1204. 

  1205.         dst[i] = (RY*r + GY*g + BY*b + (33<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;

  1206.     }

  1207. }

  1208. 

  1209. static inline void rgb48ToUV(uint8_t *dstU, uint8_t *dstV,

  1210.                              uint8_t *src1, uint8_t *src2, int width)

  1211. {

  1212.     int i;

  1213.     assert(src1==src2);

  1214.     for (i = 0; i < width; i++) {

  1215.         int r = src1[6*i + 0];

  1216.         int g = src1[6*i + 2];

  1217.         int b = src1[6*i + 4];

  1218. 

  1219.         dstU[i] = (RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;

  1220.         dstV[i] = (RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;

  1221.     }

  1222. }

  1223. 

  1224. static inline void rgb48ToUV_half(uint8_t *dstU, uint8_t *dstV,

  1225.                                   uint8_t *src1, uint8_t *src2, int width)

  1226. {

  1227.     int i;

  1228.     assert(src1==src2);

  1229.     for (i = 0; i < width; i++) {

  1230.         int r= src1[12*i + 0] + src1[12*i + 6];

  1231.         int g= src1[12*i + 2] + src1[12*i + 8];

  1232.         int b= src1[12*i + 4] + src1[12*i + 10];

  1233. 

  1234.         dstU[i]= (RU*r + GU*g + BU*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);

  1235.         dstV[i]= (RV*r + GV*g + BV*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);

  1236.     }

  1237. }

  1238. 

  1239. #define BGR2Y(type, name, shr, shg, shb, maskr, maskg, maskb, RY, GY, BY, S)\

  1240. static inline void name(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)\

  1241. {\

  1242.     int i;\

  1243.     for (i=0; i<width; i++) {\

  1244.         int b= (((const type*)src)[i]>>shb)&maskb;\

  1245.         int g= (((const type*)src)[i]>>shg)&maskg;\

  1246.         int r= (((const type*)src)[i]>>shr)&maskr;\

  1247. \

  1248.         dst[i]= (((RY)*r + (GY)*g + (BY)*b + (33<<((S)-1)))>>(S));\

  1249.     }\

  1250. }

  1251. 

  1252. BGR2Y(uint32_t, bgr32ToY,16, 0, 0, 0x00FF, 0xFF00, 0x00FF, RY<< 8, GY   , BY<< 8, RGB2YUV_SHIFT+8)

  1253. BGR2Y(uint32_t, rgb32ToY, 0, 0,16, 0x00FF, 0xFF00, 0x00FF, RY<< 8, GY   , BY<< 8, RGB2YUV_SHIFT+8)

  1254. BGR2Y(uint16_t, bgr16ToY, 0, 0, 0, 0x001F, 0x07E0, 0xF800, RY<<11, GY<<5, BY    , RGB2YUV_SHIFT+8)

  1255. BGR2Y(uint16_t, bgr15ToY, 0, 0, 0, 0x001F, 0x03E0, 0x7C00, RY<<10, GY<<5, BY    , RGB2YUV_SHIFT+7)

  1256. BGR2Y(uint16_t, rgb16ToY, 0, 0, 0, 0xF800, 0x07E0, 0x001F, RY    , GY<<5, BY<<11, RGB2YUV_SHIFT+8)

  1257. BGR2Y(uint16_t, rgb15ToY, 0, 0, 0, 0x7C00, 0x03E0, 0x001F, RY    , GY<<5, BY<<10, RGB2YUV_SHIFT+7)

  1258. 

  1259. static inline void abgrToA(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)

  1260. {

  1261.     int i;

  1262.     for (i=0; i<width; i++) {

  1263.         dst[i]= src[4*i];

  1264.     }

  1265. }

  1266. 

  1267. #define BGR2UV(type, name, shr, shg, shb, maska, maskr, maskg, maskb, RU, GU, BU, RV, GV, BV, S)\

  1268. static inline void name(uint8_t *dstU, uint8_t *dstV, const uint8_t *src, const uint8_t *dummy, long width, uint32_t *unused)\

  1269. {\

  1270.     int i;\

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

  1272.         int b= (((const type*)src)[i]&maskb)>>shb;\

  1273.         int g= (((const type*)src)[i]&maskg)>>shg;\

  1274.         int r= (((const type*)src)[i]&maskr)>>shr;\

  1275. \

  1276.         dstU[i]= ((RU)*r + (GU)*g + (BU)*b + (257<<((S)-1)))>>(S);\

  1277.         dstV[i]= ((RV)*r + (GV)*g + (BV)*b + (257<<((S)-1)))>>(S);\

  1278.     }\

  1279. }\

  1280. static inline void name ## _half(uint8_t *dstU, uint8_t *dstV, const uint8_t *src, const uint8_t *dummy, long width, uint32_t *unused)\

  1281. {\

  1282.     int i;\

  1283.     for (i=0; i<width; i++) {\

  1284.         int pix0= ((const type*)src)[2*i+0];\

  1285.         int pix1= ((const type*)src)[2*i+1];\

  1286.         int g= (pix0&~(maskr|maskb))+(pix1&~(maskr|maskb));\

  1287.         int b= ((pix0+pix1-g)&(maskb|(2*maskb)))>>shb;\

  1288.         int r= ((pix0+pix1-g)&(maskr|(2*maskr)))>>shr;\

  1289.         g&= maskg|(2*maskg);\

  1290. \

  1291.         g>>=shg;\

  1292. \

  1293.         dstU[i]= ((RU)*r + (GU)*g + (BU)*b + (257<<(S)))>>((S)+1);\

  1294.         dstV[i]= ((RV)*r + (GV)*g + (BV)*b + (257<<(S)))>>((S)+1);\

  1295.     }\

  1296. }

  1297. 

  1298. BGR2UV(uint32_t, bgr32ToUV,16, 0, 0, 0xFF000000, 0xFF0000, 0xFF00,   0x00FF, RU<< 8, GU   , BU<< 8, RV<< 8, GV   , BV<< 8, RGB2YUV_SHIFT+8)

  1299. BGR2UV(uint32_t, rgb32ToUV, 0, 0,16, 0xFF000000,   0x00FF, 0xFF00, 0xFF0000, RU<< 8, GU   , BU<< 8, RV<< 8, GV   , BV<< 8, RGB2YUV_SHIFT+8)

  1300. BGR2UV(uint16_t, bgr16ToUV, 0, 0, 0,          0,   0x001F, 0x07E0,   0xF800, RU<<11, GU<<5, BU    , RV<<11, GV<<5, BV    , RGB2YUV_SHIFT+8)

  1301. BGR2UV(uint16_t, bgr15ToUV, 0, 0, 0,          0,   0x001F, 0x03E0,   0x7C00, RU<<10, GU<<5, BU    , RV<<10, GV<<5, BV    , RGB2YUV_SHIFT+7)

  1302. BGR2UV(uint16_t, rgb16ToUV, 0, 0, 0,          0,   0xF800, 0x07E0,   0x001F, RU    , GU<<5, BU<<11, RV    , GV<<5, BV<<11, RGB2YUV_SHIFT+8)

  1303. BGR2UV(uint16_t, rgb15ToUV, 0, 0, 0,          0,   0x7C00, 0x03E0,   0x001F, RU    , GU<<5, BU<<10, RV    , GV<<5, BV<<10, RGB2YUV_SHIFT+7)

  1304. 

  1305. static inline void palToY(uint8_t *dst, const uint8_t *src, long width, uint32_t *pal)

  1306. {

  1307.     int i;

  1308.     for (i=0; i<width; i++) {

  1309.         int d= src[i];

  1310. 

  1311.         dst[i]= pal[d] & 0xFF;

  1312.     }

  1313. }

  1314. 

  1315. static inline void palToUV(uint8_t *dstU, uint8_t *dstV,

  1316.                            const uint8_t *src1, const uint8_t *src2,

  1317.                            long width, uint32_t *pal)

  1318. {

  1319.     int i;

  1320.     assert(src1 == src2);

  1321.     for (i=0; i<width; i++) {

  1322.         int p= pal[src1[i]];

  1323. 

  1324.         dstU[i]= p>>8;

  1325.         dstV[i]= p>>16;

  1326.     }

  1327. }

  1328. 

  1329. static inline void monowhite2Y(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)

  1330. {

  1331.     int i, j;

  1332.     for (i=0; i<width/8; i++) {

  1333.         int d= ~src[i];

  1334.         for(j=0; j<8; j++)

  1335.             dst[8*i+j]= ((d>>(7-j))&1)*255;

  1336.     }

  1337. }

  1338. 

  1339. static inline void monoblack2Y(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)

  1340. {

  1341.     int i, j;

  1342.     for (i=0; i<width/8; i++) {

  1343.         int d= src[i];

  1344.         for(j=0; j<8; j++)

  1345.             dst[8*i+j]= ((d>>(7-j))&1)*255;

  1346.     }

  1347. }

  1348. 

  1349. 

  1350. //Note: we have C, MMX, MMX2, 3DNOW versions, there is no 3DNOW+MMX2 one

  1351. //Plain C versions

  1352. #if ((!HAVE_MMX || !CONFIG_GPL) && !HAVE_ALTIVEC) || CONFIG_RUNTIME_CPUDETECT

  1353. #define COMPILE_C

  1354. #endif

  1355. 

  1356. #if ARCH_PPC

  1357. #if HAVE_ALTIVEC || CONFIG_RUNTIME_CPUDETECT

  1358. #define COMPILE_ALTIVEC

  1359. #endif

  1360. #endif //ARCH_PPC

  1361. 

  1362. #if ARCH_X86

  1363. 

  1364. #if ((HAVE_MMX && !HAVE_AMD3DNOW && !HAVE_MMX2) || CONFIG_RUNTIME_CPUDETECT) && CONFIG_GPL

  1365. #define COMPILE_MMX

  1366. #endif

  1367. 

  1368. #if (HAVE_MMX2 || CONFIG_RUNTIME_CPUDETECT) && CONFIG_GPL

  1369. #define COMPILE_MMX2

  1370. #endif

  1371. 

  1372. #if ((HAVE_AMD3DNOW && !HAVE_MMX2) || CONFIG_RUNTIME_CPUDETECT) && CONFIG_GPL

  1373. #define COMPILE_3DNOW

  1374. #endif

  1375. #endif //ARCH_X86

  1376. 

  1377. #define COMPILE_TEMPLATE_MMX 0

  1378. #define COMPILE_TEMPLATE_MMX2 0

  1379. #define COMPILE_TEMPLATE_AMD3DNOW 0

  1380. #define COMPILE_TEMPLATE_ALTIVEC 0

  1381. 

  1382. #ifdef COMPILE_C

  1383. #define RENAME(a) a ## _C

  1384. #include "swscale_template.c"

  1385. #endif

  1386. 

  1387. #ifdef COMPILE_ALTIVEC

  1388. #undef RENAME

  1389. #undef COMPILE_TEMPLATE_ALTIVEC

  1390. #define COMPILE_TEMPLATE_ALTIVEC 1

  1391. #define RENAME(a) a ## _altivec

  1392. #include "swscale_template.c"

  1393. #endif

  1394. 

  1395. #if ARCH_X86

  1396. 

  1397. //MMX versions

  1398. #ifdef COMPILE_MMX

  1399. #undef RENAME

  1400. #undef COMPILE_TEMPLATE_MMX

  1401. #undef COMPILE_TEMPLATE_MMX2

  1402. #undef COMPILE_TEMPLATE_AMD3DNOW

  1403. #define COMPILE_TEMPLATE_MMX 1

  1404. #define COMPILE_TEMPLATE_MMX2 0

  1405. #define COMPILE_TEMPLATE_AMD3DNOW 0

  1406. #define RENAME(a) a ## _MMX

  1407. #include "swscale_template.c"

  1408. #endif

  1409. 

  1410. //MMX2 versions

  1411. #ifdef COMPILE_MMX2

  1412. #undef RENAME

  1413. #undef COMPILE_TEMPLATE_MMX

  1414. #undef COMPILE_TEMPLATE_MMX2

  1415. #undef COMPILE_TEMPLATE_AMD3DNOW

  1416. #define COMPILE_TEMPLATE_MMX 1

  1417. #define COMPILE_TEMPLATE_MMX2 1

  1418. #define COMPILE_TEMPLATE_AMD3DNOW 0

  1419. #define RENAME(a) a ## _MMX2

  1420. #include "swscale_template.c"

  1421. #endif

  1422. 

  1423. //3DNOW versions

  1424. #ifdef COMPILE_3DNOW

  1425. #undef RENAME

  1426. #undef COMPILE_TEMPLATE_MMX

  1427. #undef COMPILE_TEMPLATE_MMX2

  1428. #undef COMPILE_TEMPLATE_AMD3DNOW

  1429. #define COMPILE_TEMPLATE_MMX 1

  1430. #define COMPILE_TEMPLATE_MMX2 0

  1431. #define COMPILE_TEMPLATE_AMD3DNOW 1

  1432. #define RENAME(a) a ## _3DNow

  1433. #include "swscale_template.c"

  1434. #endif

  1435. 

  1436. #endif //ARCH_X86

  1437. 

  1438. static double getSplineCoeff(double a, double b, double c, double d, double dist)

  1439. {

  1440. //    printf("%f %f %f %f %f\n", a,b,c,d,dist);

  1441.     if (dist<=1.0) return ((d*dist + c)*dist + b)*dist +a;

  1442.     else           return getSplineCoeff(        0.0,

  1443.                                           b+ 2.0*c + 3.0*d,

  1444.                                                  c + 3.0*d,

  1445.                                          -b- 3.0*c - 6.0*d,

  1446.                                          dist-1.0);

  1447. }

  1448. 

  1449. static inline int initFilter(int16_t **outFilter, int16_t **filterPos, int *outFilterSize, int xInc,

  1450.                              int srcW, int dstW, int filterAlign, int one, int flags,

  1451.                              SwsVector *srcFilter, SwsVector *dstFilter, double param[2])

  1452. {

  1453.     int i;

  1454.     int filterSize;

  1455.     int filter2Size;

  1456.     int minFilterSize;

  1457.     int64_t *filter=NULL;

  1458.     int64_t *filter2=NULL;

  1459.     const int64_t fone= 1LL<<54;

  1460.     int ret= -1;

  1461. #if ARCH_X86

  1462.     if (flags & SWS_CPU_CAPS_MMX)

  1463.         __asm__ volatile("emms\n\t"::: "memory"); //FIXME this should not be required but it IS (even for non-MMX versions)

  1464. #endif

  1465. 

  1466.     // NOTE: the +1 is for the MMX scaler which reads over the end

  1467.     FF_ALLOC_OR_GOTO(NULL, *filterPos, (dstW+1)*sizeof(int16_t), fail);

  1468. 

  1469.     if (FFABS(xInc - 0x10000) <10) { // unscaled

  1470.         int i;

  1471.         filterSize= 1;

  1472.         FF_ALLOCZ_OR_GOTO(NULL, filter, dstW*sizeof(*filter)*filterSize, fail);

  1473. 

  1474.         for (i=0; i<dstW; i++) {

  1475.             filter[i*filterSize]= fone;

  1476.             (*filterPos)[i]=i;

  1477.         }

  1478. 

  1479.     } else if (flags&SWS_POINT) { // lame looking point sampling mode

  1480.         int i;

  1481.         int xDstInSrc;

  1482.         filterSize= 1;

  1483.         FF_ALLOC_OR_GOTO(NULL, filter, dstW*sizeof(*filter)*filterSize, fail);

  1484. 

  1485.         xDstInSrc= xInc/2 - 0x8000;

  1486.         for (i=0; i<dstW; i++) {

  1487.             int xx= (xDstInSrc - ((filterSize-1)<<15) + (1<<15))>>16;

  1488. 

  1489.             (*filterPos)[i]= xx;

  1490.             filter[i]= fone;

  1491.             xDstInSrc+= xInc;

  1492.         }

  1493.     } else if ((xInc <= (1<<16) && (flags&SWS_AREA)) || (flags&SWS_FAST_BILINEAR)) { // bilinear upscale

  1494.         int i;

  1495.         int xDstInSrc;

  1496.         filterSize= 2;

  1497.         FF_ALLOC_OR_GOTO(NULL, filter, dstW*sizeof(*filter)*filterSize, fail);

  1498. 

  1499.         xDstInSrc= xInc/2 - 0x8000;

  1500.         for (i=0; i<dstW; i++) {

  1501.             int xx= (xDstInSrc - ((filterSize-1)<<15) + (1<<15))>>16;

  1502.             int j;

  1503. 

  1504.             (*filterPos)[i]= xx;

  1505.             //bilinear upscale / linear interpolate / area averaging

  1506.             for (j=0; j<filterSize; j++) {

  1507.                 int64_t coeff= fone - FFABS((xx<<16) - xDstInSrc)*(fone>>16);

  1508.                 if (coeff<0) coeff=0;

  1509.                 filter[i*filterSize + j]= coeff;

  1510.                 xx++;

  1511.             }

  1512.             xDstInSrc+= xInc;

  1513.         }

  1514.     } else {

  1515.         int xDstInSrc;

  1516.         int sizeFactor;

  1517. 

  1518.         if      (flags&SWS_BICUBIC)      sizeFactor=  4;

  1519.         else if (flags&SWS_X)            sizeFactor=  8;

  1520.         else if (flags&SWS_AREA)         sizeFactor=  1; //downscale only, for upscale it is bilinear

  1521.         else if (flags&SWS_GAUSS)        sizeFactor=  8;   // infinite ;)

  1522.         else if (flags&SWS_LANCZOS)      sizeFactor= param[0] != SWS_PARAM_DEFAULT ? ceil(2*param[0]) : 6;

  1523.         else if (flags&SWS_SINC)         sizeFactor= 20; // infinite ;)

  1524.         else if (flags&SWS_SPLINE)       sizeFactor= 20;  // infinite ;)

  1525.         else if (flags&SWS_BILINEAR)     sizeFactor=  2;

  1526.         else {

  1527.             sizeFactor= 0; //GCC warning killer

  1528.             assert(0);

  1529.         }

  1530. 

  1531.         if (xInc <= 1<<16)      filterSize= 1 + sizeFactor; // upscale

  1532.         else                    filterSize= 1 + (sizeFactor*srcW + dstW - 1)/ dstW;

  1533. 

  1534.         if (filterSize > srcW-2) filterSize=srcW-2;

  1535. 

  1536.         FF_ALLOC_OR_GOTO(NULL, filter, dstW*sizeof(*filter)*filterSize, fail);

  1537. 

  1538.         xDstInSrc= xInc - 0x10000;

  1539.         for (i=0; i<dstW; i++) {

  1540.             int xx= (xDstInSrc - ((filterSize-2)<<16)) / (1<<17);

  1541.             int j;

  1542.             (*filterPos)[i]= xx;

  1543.             for (j=0; j<filterSize; j++) {

  1544.                 int64_t d= ((int64_t)FFABS((xx<<17) - xDstInSrc))<<13;

  1545.                 double floatd;

  1546.                 int64_t coeff;

  1547. 

  1548.                 if (xInc > 1<<16)

  1549.                     d= d*dstW/srcW;

  1550.                 floatd= d * (1.0/(1<<30));

  1551. 

  1552.                 if (flags & SWS_BICUBIC) {

  1553.                     int64_t B= (param[0] != SWS_PARAM_DEFAULT ? param[0] :   0) * (1<<24);

  1554.                     int64_t C= (param[1] != SWS_PARAM_DEFAULT ? param[1] : 0.6) * (1<<24);

  1555.                     int64_t dd = ( d*d)>>30;

  1556.                     int64_t ddd= (dd*d)>>30;

  1557. 

  1558.                     if      (d < 1LL<<30)

  1559.                         coeff = (12*(1<<24)-9*B-6*C)*ddd + (-18*(1<<24)+12*B+6*C)*dd + (6*(1<<24)-2*B)*(1<<30);

  1560.                     else if (d < 1LL<<31)

  1561.                         coeff = (-B-6*C)*ddd + (6*B+30*C)*dd + (-12*B-48*C)*d + (8*B+24*C)*(1<<30);

  1562.                     else

  1563.                         coeff=0.0;

  1564.                     coeff *= fone>>(30+24);

  1565.                 }

  1566. /*                else if (flags & SWS_X) {

  1567.                     double p= param ? param*0.01 : 0.3;

  1568.                     coeff = d ? sin(d*PI)/(d*PI) : 1.0;

  1569.                     coeff*= pow(2.0, - p*d*d);

  1570.                 }*/

  1571.                 else if (flags & SWS_X) {

  1572.                     double A= param[0] != SWS_PARAM_DEFAULT ? param[0] : 1.0;

  1573.                     double c;

  1574. 

  1575.                     if (floatd<1.0)

  1576.                         c = cos(floatd*PI);

  1577.                     else

  1578.                         c=-1.0;

  1579.                     if (c<0.0)      c= -pow(-c, A);

  1580.                     else            c=  pow( c, A);

  1581.                     coeff= (c*0.5 + 0.5)*fone;

  1582.                 } else if (flags & SWS_AREA) {

  1583.                     int64_t d2= d - (1<<29);

  1584.                     if      (d2*xInc < -(1LL<<(29+16))) coeff= 1.0 * (1LL<<(30+16));

  1585.                     else if (d2*xInc <  (1LL<<(29+16))) coeff= -d2*xInc + (1LL<<(29+16));

  1586.                     else coeff=0.0;

  1587.                     coeff *= fone>>(30+16);

  1588.                 } else if (flags & SWS_GAUSS) {

  1589.                     double p= param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;

  1590.                     coeff = (pow(2.0, - p*floatd*floatd))*fone;

  1591.                 } else if (flags & SWS_SINC) {

  1592.                     coeff = (d ? sin(floatd*PI)/(floatd*PI) : 1.0)*fone;

  1593.                 } else if (flags & SWS_LANCZOS) {

  1594.                     double p= param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;

  1595.                     coeff = (d ? sin(floatd*PI)*sin(floatd*PI/p)/(floatd*floatd*PI*PI/p) : 1.0)*fone;

  1596.                     if (floatd>p) coeff=0;

  1597.                 } else if (flags & SWS_BILINEAR) {

  1598.                     coeff= (1<<30) - d;

  1599.                     if (coeff<0) coeff=0;

  1600.                     coeff *= fone >> 30;

  1601.                 } else if (flags & SWS_SPLINE) {

  1602.                     double p=-2.196152422706632;

  1603.                     coeff = getSplineCoeff(1.0, 0.0, p, -p-1.0, floatd) * fone;

  1604.                 } else {

  1605.                     coeff= 0.0; //GCC warning killer

  1606.                     assert(0);

  1607.                 }

  1608. 

  1609.                 filter[i*filterSize + j]= coeff;

  1610.                 xx++;

  1611.             }

  1612.             xDstInSrc+= 2*xInc;

  1613.         }

  1614.     }

  1615. 

  1616.     /* apply src & dst Filter to filter -> filter2

  1617.        av_free(filter);

  1618.     */

  1619.     assert(filterSize>0);

  1620.     filter2Size= filterSize;

  1621.     if (srcFilter) filter2Size+= srcFilter->length - 1;

  1622.     if (dstFilter) filter2Size+= dstFilter->length - 1;

  1623.     assert(filter2Size>0);

  1624.     FF_ALLOCZ_OR_GOTO(NULL, filter2, filter2Size*dstW*sizeof(*filter2), fail);

  1625. 

  1626.     for (i=0; i<dstW; i++) {

  1627.         int j, k;

  1628. 

  1629.         if(srcFilter) {

  1630.             for (k=0; k<srcFilter->length; k++) {

  1631.                 for (j=0; j<filterSize; j++)

  1632.                     filter2[i*filter2Size + k + j] += srcFilter->coeff[k]*filter[i*filterSize + j];

  1633.             }

  1634.         } else {

  1635.             for (j=0; j<filterSize; j++)

  1636.                 filter2[i*filter2Size + j]= filter[i*filterSize + j];

  1637.         }

  1638.         //FIXME dstFilter

  1639. 

  1640.         (*filterPos)[i]+= (filterSize-1)/2 - (filter2Size-1)/2;

  1641.     }

  1642.     av_freep(&filter);

  1643. 

  1644.     /* try to reduce the filter-size (step1 find size and shift left) */

  1645.     // Assume it is near normalized (*0.5 or *2.0 is OK but * 0.001 is not).

  1646.     minFilterSize= 0;

  1647.     for (i=dstW-1; i>=0; i--) {

  1648.         int min= filter2Size;

  1649.         int j;

  1650.         int64_t cutOff=0.0;

  1651. 

  1652.         /* get rid off near zero elements on the left by shifting left */

  1653.         for (j=0; j<filter2Size; j++) {

  1654.             int k;

  1655.             cutOff += FFABS(filter2[i*filter2Size]);

  1656. 

  1657.             if (cutOff > SWS_MAX_REDUCE_CUTOFF*fone) break;

  1658. 

  1659.             /* preserve monotonicity because the core can't handle the filter otherwise */

  1660.             if (i<dstW-1 && (*filterPos)[i] >= (*filterPos)[i+1]) break;

  1661. 

  1662.             // move filter coefficients left

  1663.             for (k=1; k<filter2Size; k++)

  1664.                 filter2[i*filter2Size + k - 1]= filter2[i*filter2Size + k];

  1665.             filter2[i*filter2Size + k - 1]= 0;

  1666.             (*filterPos)[i]++;

  1667.         }

  1668. 

  1669.         cutOff=0;

  1670.         /* count near zeros on the right */

  1671.         for (j=filter2Size-1; j>0; j--) {

  1672.             cutOff += FFABS(filter2[i*filter2Size + j]);

  1673. 

  1674.             if (cutOff > SWS_MAX_REDUCE_CUTOFF*fone) break;

  1675.             min--;

  1676.         }

  1677. 

  1678.         if (min>minFilterSize) minFilterSize= min;

  1679.     }

  1680. 

  1681.     if (flags & SWS_CPU_CAPS_ALTIVEC) {

  1682.         // we can handle the special case 4,

  1683.         // so we don't want to go to the full 8

  1684.         if (minFilterSize < 5)

  1685.             filterAlign = 4;

  1686. 

  1687.         // We really don't want to waste our time

  1688.         // doing useless computation, so fall back on

  1689.         // the scalar C code for very small filters.

  1690.         // Vectorizing is worth it only if you have a

  1691.         // decent-sized vector.

  1692.         if (minFilterSize < 3)

  1693.             filterAlign = 1;

  1694.     }

  1695. 

  1696.     if (flags & SWS_CPU_CAPS_MMX) {

  1697.         // special case for unscaled vertical filtering

  1698.         if (minFilterSize == 1 && filterAlign == 2)

  1699.             filterAlign= 1;

  1700.     }

  1701. 

  1702.     assert(minFilterSize > 0);

  1703.     filterSize= (minFilterSize +(filterAlign-1)) & (~(filterAlign-1));

  1704.     assert(filterSize > 0);

  1705.     filter= av_malloc(filterSize*dstW*sizeof(*filter));

  1706.     if (filterSize >= MAX_FILTER_SIZE*16/((flags&SWS_ACCURATE_RND) ? APCK_SIZE : 16) || !filter)

  1707.         goto fail;

  1708.     *outFilterSize= filterSize;

  1709. 

  1710.     if (flags&SWS_PRINT_INFO)

  1711.         av_log(NULL, AV_LOG_VERBOSE, "SwScaler: reducing / aligning filtersize %d -> %d\n", filter2Size, filterSize);

  1712.     /* try to reduce the filter-size (step2 reduce it) */

  1713.     for (i=0; i<dstW; i++) {

  1714.         int j;

  1715. 

  1716.         for (j=0; j<filterSize; j++) {

  1717.             if (j>=filter2Size) filter[i*filterSize + j]= 0;

  1718.             else               filter[i*filterSize + j]= filter2[i*filter2Size + j];

  1719.             if((flags & SWS_BITEXACT) && j>=minFilterSize)

  1720.                 filter[i*filterSize + j]= 0;

  1721.         }

  1722.     }

  1723. 

  1724. 

  1725.     //FIXME try to align filterPos if possible

  1726. 

  1727.     //fix borders

  1728.     for (i=0; i<dstW; i++) {

  1729.         int j;

  1730.         if ((*filterPos)[i] < 0) {

  1731.             // move filter coefficients left to compensate for filterPos

  1732.             for (j=1; j<filterSize; j++) {

  1733.                 int left= FFMAX(j + (*filterPos)[i], 0);

  1734.                 filter[i*filterSize + left] += filter[i*filterSize + j];

  1735.                 filter[i*filterSize + j]=0;

  1736.             }

  1737.             (*filterPos)[i]= 0;

  1738.         }

  1739. 

  1740.         if ((*filterPos)[i] + filterSize > srcW) {

  1741.             int shift= (*filterPos)[i] + filterSize - srcW;

  1742.             // move filter coefficients right to compensate for filterPos

  1743.             for (j=filterSize-2; j>=0; j--) {

  1744.                 int right= FFMIN(j + shift, filterSize-1);

  1745.                 filter[i*filterSize +right] += filter[i*filterSize +j];

  1746.                 filter[i*filterSize +j]=0;

  1747.             }

  1748.             (*filterPos)[i]= srcW - filterSize;

  1749.         }

  1750.     }

  1751. 

  1752.     // Note the +1 is for the MMX scaler which reads over the end

  1753.     /* align at 16 for AltiVec (needed by hScale_altivec_real) */

  1754.     FF_ALLOCZ_OR_GOTO(NULL, *outFilter, *outFilterSize*(dstW+1)*sizeof(int16_t), fail);

  1755. 

  1756.     /* normalize & store in outFilter */

  1757.     for (i=0; i<dstW; i++) {

  1758.         int j;

  1759.         int64_t error=0;

  1760.         int64_t sum=0;

  1761. 

  1762.         for (j=0; j<filterSize; j++) {

  1763.             sum+= filter[i*filterSize + j];

  1764.         }

  1765.         sum= (sum + one/2)/ one;

  1766.         for (j=0; j<*outFilterSize; j++) {

  1767.             int64_t v= filter[i*filterSize + j] + error;

  1768.             int intV= ROUNDED_DIV(v, sum);

  1769.             (*outFilter)[i*(*outFilterSize) + j]= intV;

  1770.             error= v - intV*sum;

  1771.         }

  1772.     }

  1773. 

  1774.     (*filterPos)[dstW]= (*filterPos)[dstW-1]; // the MMX scaler will read over the end

  1775.     for (i=0; i<*outFilterSize; i++) {

  1776.         int j= dstW*(*outFilterSize);

  1777.         (*outFilter)[j + i]= (*outFilter)[j + i - (*outFilterSize)];

  1778.     }

  1779. 

  1780.     ret=0;

  1781. fail:

  1782.     av_free(filter);

  1783.     av_free(filter2);

  1784.     return ret;

  1785. }

  1786. 

  1787. #ifdef COMPILE_MMX2

  1788. static int initMMX2HScaler(int dstW, int xInc, uint8_t *filterCode, int16_t *filter, int32_t *filterPos, int numSplits)

  1789. {

  1790.     uint8_t *fragmentA;

  1791.     x86_reg imm8OfPShufW1A;

  1792.     x86_reg imm8OfPShufW2A;

  1793.     x86_reg fragmentLengthA;

  1794.     uint8_t *fragmentB;

  1795.     x86_reg imm8OfPShufW1B;

  1796.     x86_reg imm8OfPShufW2B;

  1797.     x86_reg fragmentLengthB;

  1798.     int fragmentPos;

  1799. 

  1800.     int xpos, i;

  1801. 

  1802.     // create an optimized horizontal scaling routine

  1803. 

  1804.     //code fragment

  1805. 

  1806.     __asm__ volatile(

  1807.         "jmp                         9f                 \n\t"

  1808.     // Begin

  1809.         "0:                                             \n\t"

  1810.         "movq    (%%"REG_d", %%"REG_a"), %%mm3          \n\t"

  1811.         "movd    (%%"REG_c", %%"REG_S"), %%mm0          \n\t"

  1812.         "movd   1(%%"REG_c", %%"REG_S"), %%mm1          \n\t"

  1813.         "punpcklbw                %%mm7, %%mm1          \n\t"

  1814.         "punpcklbw                %%mm7, %%mm0          \n\t"

  1815.         "pshufw                   $0xFF, %%mm1, %%mm1   \n\t"

  1816.         "1:                                             \n\t"

  1817.         "pshufw                   $0xFF, %%mm0, %%mm0   \n\t"

  1818.         "2:                                             \n\t"

  1819.         "psubw                    %%mm1, %%mm0          \n\t"

  1820.         "movl   8(%%"REG_b", %%"REG_a"), %%esi          \n\t"

  1821.         "pmullw                   %%mm3, %%mm0          \n\t"

  1822.         "psllw                       $7, %%mm1          \n\t"

  1823.         "paddw                    %%mm1, %%mm0          \n\t"

  1824. 

  1825.         "movq                     %%mm0, (%%"REG_D", %%"REG_a") \n\t"

  1826. 

  1827.         "add                         $8, %%"REG_a"      \n\t"

  1828.     // End

  1829.         "9:                                             \n\t"

  1830. //        "int $3                                         \n\t"

  1831.         "lea                 " LOCAL_MANGLE(0b) ", %0   \n\t"

  1832.         "lea                 " LOCAL_MANGLE(1b) ", %1   \n\t"

  1833.         "lea                 " LOCAL_MANGLE(2b) ", %2   \n\t"

  1834.         "dec                         %1                 \n\t"

  1835.         "dec                         %2                 \n\t"

  1836.         "sub                         %0, %1             \n\t"

  1837.         "sub                         %0, %2             \n\t"

  1838.         "lea                 " LOCAL_MANGLE(9b) ", %3   \n\t"

  1839.         "sub                         %0, %3             \n\t"

  1840. 

  1841. 

  1842.         :"=r" (fragmentA), "=r" (imm8OfPShufW1A), "=r" (imm8OfPShufW2A),

  1843.         "=r" (fragmentLengthA)

  1844.     );

  1845. 

  1846.     __asm__ volatile(

  1847.         "jmp                         9f                 \n\t"

  1848.     // Begin

  1849.         "0:                                             \n\t"

  1850.         "movq    (%%"REG_d", %%"REG_a"), %%mm3          \n\t"

  1851.         "movd    (%%"REG_c", %%"REG_S"), %%mm0          \n\t"

  1852.         "punpcklbw                %%mm7, %%mm0          \n\t"

  1853.         "pshufw                   $0xFF, %%mm0, %%mm1   \n\t"

  1854.         "1:                                             \n\t"

  1855.         "pshufw                   $0xFF, %%mm0, %%mm0   \n\t"

  1856.         "2:                                             \n\t"

  1857.         "psubw                    %%mm1, %%mm0          \n\t"

  1858.         "movl   8(%%"REG_b", %%"REG_a"), %%esi          \n\t"

  1859.         "pmullw                   %%mm3, %%mm0          \n\t"

  1860.         "psllw                       $7, %%mm1          \n\t"

  1861.         "paddw                    %%mm1, %%mm0          \n\t"

  1862. 

  1863.         "movq                     %%mm0, (%%"REG_D", %%"REG_a") \n\t"

  1864. 

  1865.         "add                         $8, %%"REG_a"      \n\t"

  1866.     // End

  1867.         "9:                                             \n\t"

  1868. //        "int                       $3                   \n\t"

  1869.         "lea                 " LOCAL_MANGLE(0b) ", %0   \n\t"

  1870.         "lea                 " LOCAL_MANGLE(1b) ", %1   \n\t"

  1871.         "lea                 " LOCAL_MANGLE(2b) ", %2   \n\t"

  1872.         "dec                         %1                 \n\t"

  1873.         "dec                         %2                 \n\t"

  1874.         "sub                         %0, %1             \n\t"

  1875.         "sub                         %0, %2             \n\t"

  1876.         "lea                 " LOCAL_MANGLE(9b) ", %3   \n\t"

  1877.         "sub                         %0, %3             \n\t"

  1878. 

  1879. 

  1880.         :"=r" (fragmentB), "=r" (imm8OfPShufW1B), "=r" (imm8OfPShufW2B),

  1881.         "=r" (fragmentLengthB)

  1882.     );

  1883. 

  1884.     xpos= 0; //lumXInc/2 - 0x8000; // difference between pixel centers

  1885.     fragmentPos=0;

  1886. 

  1887.     for (i=0; i<dstW/numSplits; i++) {

  1888.         int xx=xpos>>16;

  1889. 

  1890.         if ((i&3) == 0) {

  1891.             int a=0;

  1892.             int b=((xpos+xInc)>>16) - xx;

  1893.             int c=((xpos+xInc*2)>>16) - xx;

  1894.             int d=((xpos+xInc*3)>>16) - xx;

  1895.             int inc                = (d+1<4);

  1896.             uint8_t *fragment      = (d+1<4) ? fragmentB       : fragmentA;

  1897.             x86_reg imm8OfPShufW1  = (d+1<4) ? imm8OfPShufW1B  : imm8OfPShufW1A;

  1898.             x86_reg imm8OfPShufW2  = (d+1<4) ? imm8OfPShufW2B  : imm8OfPShufW2A;

  1899.             x86_reg fragmentLength = (d+1<4) ? fragmentLengthB : fragmentLengthA;

  1900.             int maxShift= 3-(d+inc);

  1901.             int shift=0;

  1902. 

  1903.             if (filterCode) {

  1904.                 filter[i  ] = (( xpos         & 0xFFFF) ^ 0xFFFF)>>9;

  1905.                 filter[i+1] = (((xpos+xInc  ) & 0xFFFF) ^ 0xFFFF)>>9;

  1906.                 filter[i+2] = (((xpos+xInc*2) & 0xFFFF) ^ 0xFFFF)>>9;

  1907.                 filter[i+3] = (((xpos+xInc*3) & 0xFFFF) ^ 0xFFFF)>>9;

  1908.                 filterPos[i/2]= xx;

  1909. 

  1910.                 memcpy(filterCode + fragmentPos, fragment, fragmentLength);

  1911. 

  1912.                 filterCode[fragmentPos + imm8OfPShufW1]=

  1913.                     (a+inc) | ((b+inc)<<2) | ((c+inc)<<4) | ((d+inc)<<6);

  1914.                 filterCode[fragmentPos + imm8OfPShufW2]=

  1915.                     a | (b<<2) | (c<<4) | (d<<6);

  1916. 

  1917.                 if (i+4-inc>=dstW) shift=maxShift; //avoid overread

  1918.                 else if ((filterPos[i/2]&3) <= maxShift) shift=filterPos[i/2]&3; //Align

  1919. 

  1920.                 if (shift && i>=shift) {

  1921.                     filterCode[fragmentPos + imm8OfPShufW1]+= 0x55*shift;

  1922.                     filterCode[fragmentPos + imm8OfPShufW2]+= 0x55*shift;

  1923.                     filterPos[i/2]-=shift;

  1924.                 }

  1925.             }

  1926. 

  1927.             fragmentPos+= fragmentLength;

  1928. 

  1929.             if (filterCode)

  1930.                 filterCode[fragmentPos]= RET;

  1931.         }

  1932.         xpos+=xInc;

  1933.     }

  1934.     if (filterCode)

  1935.         filterPos[((i/2)+1)&(~1)]= xpos>>16; // needed to jump to the next part

  1936. 

  1937.     return fragmentPos + 1;

  1938. }

  1939. #endif /* COMPILE_MMX2 */

  1940. 

  1941. static void globalInit(void)

  1942. {

  1943.     // generating tables:

  1944.     int i;

  1945.     for (i=0; i<768; i++) {

  1946.         int c= av_clip_uint8(i-256);

  1947.         clip_table[i]=c;

  1948.     }

  1949. }

  1950. 

  1951. static SwsFunc getSwsFunc(SwsContext *c)

  1952. {

  1953. #if CONFIG_RUNTIME_CPUDETECT

  1954.     int flags = c->flags;

  1955. 

  1956. #if ARCH_X86 && CONFIG_GPL

  1957.     // ordered per speed fastest first

  1958.     if (flags & SWS_CPU_CAPS_MMX2) {

  1959.         sws_init_swScale_MMX2(c);

  1960.         return swScale_MMX2;

  1961.     } else if (flags & SWS_CPU_CAPS_3DNOW) {

  1962.         sws_init_swScale_3DNow(c);

  1963.         return swScale_3DNow;

  1964.     } else if (flags & SWS_CPU_CAPS_MMX) {

  1965.         sws_init_swScale_MMX(c);

  1966.         return swScale_MMX;

  1967.     } else {

  1968.         sws_init_swScale_C(c);

  1969.         return swScale_C;

  1970.     }

  1971. 

  1972. #else

  1973. #if ARCH_PPC

  1974.     if (flags & SWS_CPU_CAPS_ALTIVEC) {

  1975.         sws_init_swScale_altivec(c);

  1976.         return swScale_altivec;

  1977.     } else {

  1978.         sws_init_swScale_C(c);

  1979.         return swScale_C;

  1980.     }

  1981. #endif

  1982.     sws_init_swScale_C(c);

  1983.     return swScale_C;

  1984. #endif /* ARCH_X86 && CONFIG_GPL */

  1985. #else //CONFIG_RUNTIME_CPUDETECT

  1986. #if   COMPILE_TEMPLATE_MMX2

  1987.     sws_init_swScale_MMX2(c);

  1988.     return swScale_MMX2;

  1989. #elif COMPILE_TEMPLATE_AMD3DNOW

  1990.     sws_init_swScale_3DNow(c);

  1991.     return swScale_3DNow;

  1992. #elif COMPILE_TEMPLATE_MMX

  1993.     sws_init_swScale_MMX(c);

  1994.     return swScale_MMX;

  1995. #elif COMPILE_TEMPLATE_ALTIVEC

  1996.     sws_init_swScale_altivec(c);

  1997.     return swScale_altivec;

  1998. #else

  1999.     sws_init_swScale_C(c);

  2000.     return swScale_C;

  2001. #endif

  2002. #endif //!CONFIG_RUNTIME_CPUDETECT

  2003. }

  2004. 

  2005. static int PlanarToNV12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  2006.                                int srcSliceH, uint8_t* dstParam[], int dstStride[])

  2007. {

  2008.     uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;

  2009.     /* Copy Y plane */

  2010.     if (dstStride[0]==srcStride[0] && srcStride[0] > 0)

  2011.         memcpy(dst, src[0], srcSliceH*dstStride[0]);

  2012.     else {

  2013.         int i;

  2014.         const uint8_t *srcPtr= src[0];

  2015.         uint8_t *dstPtr= dst;

  2016.         for (i=0; i<srcSliceH; i++) {

  2017.             memcpy(dstPtr, srcPtr, c->srcW);

  2018.             srcPtr+= srcStride[0];

  2019.             dstPtr+= dstStride[0];

  2020.         }

  2021.     }

  2022.     dst = dstParam[1] + dstStride[1]*srcSliceY/2;

  2023.     if (c->dstFormat == PIX_FMT_NV12)

  2024.         interleaveBytes(src[1], src[2], dst, c->srcW/2, srcSliceH/2, srcStride[1], srcStride[2], dstStride[0]);

  2025.     else

  2026.         interleaveBytes(src[2], src[1], dst, c->srcW/2, srcSliceH/2, srcStride[2], srcStride[1], dstStride[0]);

  2027. 

  2028.     return srcSliceH;

  2029. }

  2030. 

  2031. static int PlanarToYuy2Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  2032.                                int srcSliceH, uint8_t* dstParam[], int dstStride[])

  2033. {

  2034.     uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;

  2035. 

  2036.     yv12toyuy2(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);

  2037. 

  2038.     return srcSliceH;

  2039. }

  2040. 

  2041. static int PlanarToUyvyWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  2042.                                int srcSliceH, uint8_t* dstParam[], int dstStride[])

  2043. {

  2044.     uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;

  2045. 

  2046.     yv12touyvy(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);

  2047. 

  2048.     return srcSliceH;

  2049. }

  2050. 

  2051. static int YUV422PToYuy2Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  2052.                                 int srcSliceH, uint8_t* dstParam[], int dstStride[])

  2053. {

  2054.     uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;

  2055. 

  2056.     yuv422ptoyuy2(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);

  2057. 

  2058.     return srcSliceH;

  2059. }

  2060. 

  2061. static int YUV422PToUyvyWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  2062.                                 int srcSliceH, uint8_t* dstParam[], int dstStride[])

  2063. {

  2064.     uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;

  2065. 

  2066.     yuv422ptouyvy(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);

  2067. 

  2068.     return srcSliceH;

  2069. }

  2070. 

  2071. static int YUYV2YUV420Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  2072.                                int srcSliceH, uint8_t* dstParam[], int dstStride[])

  2073. {

  2074.     uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;

  2075.     uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY/2;

  2076.     uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY/2;

  2077. 

  2078.     yuyvtoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);

  2079. 

  2080.     if (dstParam[3])

  2081.         fillPlane(dstParam[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);

  2082. 

  2083.     return srcSliceH;

  2084. }

  2085. 

  2086. static int YUYV2YUV422Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  2087.                                int srcSliceH, uint8_t* dstParam[], int dstStride[])

  2088. {

  2089.     uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;

  2090.     uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY;

  2091.     uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY;

  2092. 

  2093.     yuyvtoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);

  2094. 

  2095.     return srcSliceH;

  2096. }

  2097. 

  2098. static int UYVY2YUV420Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  2099.                                int srcSliceH, uint8_t* dstParam[], int dstStride[])

  2100. {

  2101.     uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;

  2102.     uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY/2;

  2103.     uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY/2;

  2104. 

  2105.     uyvytoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);

  2106. 

  2107.     if (dstParam[3])

  2108.         fillPlane(dstParam[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);

  2109. 

  2110.     return srcSliceH;

  2111. }

  2112. 

  2113. static int UYVY2YUV422Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  2114.                                int srcSliceH, uint8_t* dstParam[], int dstStride[])

  2115. {

  2116.     uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;

  2117.     uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY;

  2118.     uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY;

  2119. 

  2120.     uyvytoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);

  2121. 

  2122.     return srcSliceH;

  2123. }

  2124. 

  2125. static int pal2rgbWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  2126.                           int srcSliceH, uint8_t* dst[], int dstStride[])

  2127. {

  2128.     const enum PixelFormat srcFormat= c->srcFormat;

  2129.     const enum PixelFormat dstFormat= c->dstFormat;

  2130.     void (*conv)(const uint8_t *src, uint8_t *dst, long num_pixels,

  2131.                  const uint8_t *palette)=NULL;

  2132.     int i;

  2133.     uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;

  2134.     uint8_t *srcPtr= src[0];

  2135. 

  2136.     if (!usePal(srcFormat))

  2137.         av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",

  2138.                sws_format_name(srcFormat), sws_format_name(dstFormat));

  2139. 

  2140.     switch(dstFormat) {

  2141.     case PIX_FMT_RGB32  : conv = palette8topacked32; break;

  2142.     case PIX_FMT_BGR32  : conv = palette8topacked32; break;

  2143.     case PIX_FMT_BGR32_1: conv = palette8topacked32; break;

  2144.     case PIX_FMT_RGB32_1: conv = palette8topacked32; break;

  2145.     case PIX_FMT_RGB24  : conv = palette8topacked24; break;

  2146.     case PIX_FMT_BGR24  : conv = palette8topacked24; break;

  2147.     default: av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",

  2148.                     sws_format_name(srcFormat), sws_format_name(dstFormat)); break;

  2149.     }

  2150. 

  2151. 

  2152.     for (i=0; i<srcSliceH; i++) {

  2153.         conv(srcPtr, dstPtr, c->srcW, (uint8_t *) c->pal_rgb);

  2154.         srcPtr+= srcStride[0];

  2155.         dstPtr+= dstStride[0];

  2156.     }

  2157. 

  2158.     return srcSliceH;

  2159. }

  2160. 

  2161. /* {RGB,BGR}{15,16,24,32,32_1} -> {RGB,BGR}{15,16,24,32} */

  2162. static int rgb2rgbWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  2163.                           int srcSliceH, uint8_t* dst[], int dstStride[])

  2164. {

  2165.     const enum PixelFormat srcFormat= c->srcFormat;

  2166.     const enum PixelFormat dstFormat= c->dstFormat;

  2167.     const int srcBpp= (fmt_depth(srcFormat) + 7) >> 3;

  2168.     const int dstBpp= (fmt_depth(dstFormat) + 7) >> 3;

  2169.     const int srcId= fmt_depth(srcFormat) >> 2; /* 1:0, 4:1, 8:2, 15:3, 16:4, 24:6, 32:8 */

  2170.     const int dstId= fmt_depth(dstFormat) >> 2;

  2171.     void (*conv)(const uint8_t *src, uint8_t *dst, long src_size)=NULL;

  2172. 

  2173.     /* BGR -> BGR */

  2174.     if (  (isBGR(srcFormat) && isBGR(dstFormat))

  2175.        || (isRGB(srcFormat) && isRGB(dstFormat))) {

  2176.         switch(srcId | (dstId<<4)) {

  2177.         case 0x34: conv= rgb16to15; break;

  2178.         case 0x36: conv= rgb24to15; break;

  2179.         case 0x38: conv= rgb32to15; break;

  2180.         case 0x43: conv= rgb15to16; break;

  2181.         case 0x46: conv= rgb24to16; break;

  2182.         case 0x48: conv= rgb32to16; break;

  2183.         case 0x63: conv= rgb15to24; break;

  2184.         case 0x64: conv= rgb16to24; break;

  2185.         case 0x68: conv= rgb32to24; break;

  2186.         case 0x83: conv= rgb15to32; break;

  2187.         case 0x84: conv= rgb16to32; break;

  2188.         case 0x86: conv= rgb24to32; break;

  2189.         default: av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",

  2190.                         sws_format_name(srcFormat), sws_format_name(dstFormat)); break;

  2191.         }

  2192.     } else if (  (isBGR(srcFormat) && isRGB(dstFormat))

  2193.              || (isRGB(srcFormat) && isBGR(dstFormat))) {

  2194.         switch(srcId | (dstId<<4)) {

  2195.         case 0x33: conv= rgb15tobgr15; break;

  2196.         case 0x34: conv= rgb16tobgr15; break;

  2197.         case 0x36: conv= rgb24tobgr15; break;

  2198.         case 0x38: conv= rgb32tobgr15; break;

  2199.         case 0x43: conv= rgb15tobgr16; break;

  2200.         case 0x44: conv= rgb16tobgr16; break;

  2201.         case 0x46: conv= rgb24tobgr16; break;

  2202.         case 0x48: conv= rgb32tobgr16; break;

  2203.         case 0x63: conv= rgb15tobgr24; break;

  2204.         case 0x64: conv= rgb16tobgr24; break;

  2205.         case 0x66: conv= rgb24tobgr24; break;

  2206.         case 0x68: conv= rgb32tobgr24; break;

  2207.         case 0x83: conv= rgb15tobgr32; break;

  2208.         case 0x84: conv= rgb16tobgr32; break;

  2209.         case 0x86: conv= rgb24tobgr32; break;

  2210.         case 0x88: conv= rgb32tobgr32; break;

  2211.         default: av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",

  2212.                         sws_format_name(srcFormat), sws_format_name(dstFormat)); break;

  2213.         }

  2214.     } else {

  2215.         av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",

  2216.                sws_format_name(srcFormat), sws_format_name(dstFormat));

  2217.     }

  2218. 

  2219.     if(conv) {

  2220.         uint8_t *srcPtr= src[0];

  2221.         if(srcFormat == PIX_FMT_RGB32_1 || srcFormat == PIX_FMT_BGR32_1)

  2222.             srcPtr += ALT32_CORR;

  2223. 

  2224.         if (dstStride[0]*srcBpp == srcStride[0]*dstBpp && srcStride[0] > 0)

  2225.             conv(srcPtr, dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]);

  2226.         else {

  2227.             int i;

  2228.             uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;

  2229. 

  2230.             for (i=0; i<srcSliceH; i++) {

  2231.                 conv(srcPtr, dstPtr, c->srcW*srcBpp);

  2232.                 srcPtr+= srcStride[0];

  2233.                 dstPtr+= dstStride[0];

  2234.             }

  2235.         }

  2236.     }

  2237.     return srcSliceH;

  2238. }

  2239. 

  2240. static int bgr24toyv12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  2241.                               int srcSliceH, uint8_t* dst[], int dstStride[])

  2242. {

  2243. 

  2244.     rgb24toyv12(

  2245.         src[0],

  2246.         dst[0]+ srcSliceY    *dstStride[0],

  2247.         dst[1]+(srcSliceY>>1)*dstStride[1],

  2248.         dst[2]+(srcSliceY>>1)*dstStride[2],

  2249.         c->srcW, srcSliceH,

  2250.         dstStride[0], dstStride[1], srcStride[0]);

  2251.     if (dst[3])

  2252.         fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);

  2253.     return srcSliceH;

  2254. }

  2255. 

  2256. static int yvu9toyv12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  2257.                              int srcSliceH, uint8_t* dst[], int dstStride[])

  2258. {

  2259.     int i;

  2260. 

  2261.     /* copy Y */

  2262.     if (srcStride[0]==dstStride[0] && srcStride[0] > 0)

  2263.         memcpy(dst[0]+ srcSliceY*dstStride[0], src[0], srcStride[0]*srcSliceH);

  2264.     else {

  2265.         uint8_t *srcPtr= src[0];

  2266.         uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;

  2267. 

  2268.         for (i=0; i<srcSliceH; i++) {

  2269.             memcpy(dstPtr, srcPtr, c->srcW);

  2270.             srcPtr+= srcStride[0];

  2271.             dstPtr+= dstStride[0];

  2272.         }

  2273.     }

  2274. 

  2275.     if (c->dstFormat==PIX_FMT_YUV420P || c->dstFormat==PIX_FMT_YUVA420P) {

  2276.         planar2x(src[1], dst[1] + dstStride[1]*(srcSliceY >> 1), c->chrSrcW,

  2277.                  srcSliceH >> 2, srcStride[1], dstStride[1]);

  2278.         planar2x(src[2], dst[2] + dstStride[2]*(srcSliceY >> 1), c->chrSrcW,

  2279.                  srcSliceH >> 2, srcStride[2], dstStride[2]);

  2280.     } else {

  2281.         planar2x(src[1], dst[2] + dstStride[2]*(srcSliceY >> 1), c->chrSrcW,

  2282.                  srcSliceH >> 2, srcStride[1], dstStride[2]);

  2283.         planar2x(src[2], dst[1] + dstStride[1]*(srcSliceY >> 1), c->chrSrcW,

  2284.                  srcSliceH >> 2, srcStride[2], dstStride[1]);

  2285.     }

  2286.     if (dst[3])

  2287.         fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);

  2288.     return srcSliceH;

  2289. }

  2290. 

  2291. /* unscaled copy like stuff (assumes nearly identical formats) */

  2292. static int packedCopy(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  2293.                       int srcSliceH, uint8_t* dst[], int dstStride[])

  2294. {

  2295.     if (dstStride[0]==srcStride[0] && srcStride[0] > 0)

  2296.         memcpy(dst[0] + dstStride[0]*srcSliceY, src[0], srcSliceH*dstStride[0]);

  2297.     else {

  2298.         int i;

  2299.         uint8_t *srcPtr= src[0];

  2300.         uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;

  2301.         int length=0;

  2302. 

  2303.         /* universal length finder */

  2304.         while(length+c->srcW <= FFABS(dstStride[0])

  2305.            && length+c->srcW <= FFABS(srcStride[0])) length+= c->srcW;

  2306.         assert(length!=0);

  2307. 

  2308.         for (i=0; i<srcSliceH; i++) {

  2309.             memcpy(dstPtr, srcPtr, length);

  2310.             srcPtr+= srcStride[0];

  2311.             dstPtr+= dstStride[0];

  2312.         }

  2313.     }

  2314.     return srcSliceH;

  2315. }

  2316. 

  2317. static int planarCopy(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  2318.                       int srcSliceH, uint8_t* dst[], int dstStride[])

  2319. {

  2320.     int plane, i, j;

  2321.     for (plane=0; plane<4; plane++) {

  2322.         int length= (plane==0 || plane==3) ? c->srcW  : -((-c->srcW  )>>c->chrDstHSubSample);

  2323.         int y=      (plane==0 || plane==3) ? srcSliceY: -((-srcSliceY)>>c->chrDstVSubSample);

  2324.         int height= (plane==0 || plane==3) ? srcSliceH: -((-srcSliceH)>>c->chrDstVSubSample);

  2325.         uint8_t *srcPtr= src[plane];

  2326.         uint8_t *dstPtr= dst[plane] + dstStride[plane]*y;

  2327. 

  2328.         if (!dst[plane]) continue;

  2329.         // ignore palette for GRAY8

  2330.         if (plane == 1 && !dst[2]) continue;

  2331.         if (!src[plane] || (plane == 1 && !src[2])) {

  2332.             if(is16BPS(c->dstFormat))

  2333.                 length*=2;

  2334.             fillPlane(dst[plane], dstStride[plane], length, height, y, (plane==3) ? 255 : 128);

  2335.         } else {

  2336.             if(is16BPS(c->srcFormat) && !is16BPS(c->dstFormat)) {

  2337.                 if (!isBE(c->srcFormat)) srcPtr++;

  2338.                 for (i=0; i<height; i++) {

  2339.                     for (j=0; j<length; j++) dstPtr[j] = srcPtr[j<<1];

  2340.                     srcPtr+= srcStride[plane];

  2341.                     dstPtr+= dstStride[plane];

  2342.                 }

  2343.             } else if(!is16BPS(c->srcFormat) && is16BPS(c->dstFormat)) {

  2344.                 for (i=0; i<height; i++) {

  2345.                     for (j=0; j<length; j++) {

  2346.                         dstPtr[ j<<1   ] = srcPtr[j];

  2347.                         dstPtr[(j<<1)+1] = srcPtr[j];

  2348.                     }

  2349.                     srcPtr+= srcStride[plane];

  2350.                     dstPtr+= dstStride[plane];

  2351.                 }

  2352.             } else if(is16BPS(c->srcFormat) && is16BPS(c->dstFormat)

  2353.                   && isBE(c->srcFormat) != isBE(c->dstFormat)) {

  2354. 

  2355.                 for (i=0; i<height; i++) {

  2356.                     for (j=0; j<length; j++)

  2357.                         ((uint16_t*)dstPtr)[j] = bswap_16(((uint16_t*)srcPtr)[j]);

  2358.                     srcPtr+= srcStride[plane];

  2359.                     dstPtr+= dstStride[plane];

  2360.                 }

  2361.             } else if (dstStride[plane]==srcStride[plane] && srcStride[plane] > 0)

  2362.                 memcpy(dst[plane] + dstStride[plane]*y, src[plane], height*dstStride[plane]);

  2363.             else {

  2364.                 if(is16BPS(c->srcFormat) && is16BPS(c->dstFormat))

  2365.                     length*=2;

  2366.                 for (i=0; i<height; i++) {

  2367.                     memcpy(dstPtr, srcPtr, length);

  2368.                     srcPtr+= srcStride[plane];

  2369.                     dstPtr+= dstStride[plane];

  2370.                 }

  2371.             }

  2372.         }

  2373.     }

  2374.     return srcSliceH;

  2375. }

  2376. 

  2377. 

  2378. static void getSubSampleFactors(int *h, int *v, int format)

  2379. {

  2380.     switch(format) {

  2381.     case PIX_FMT_UYVY422:

  2382.     case PIX_FMT_YUYV422:

  2383.         *h=1;

  2384.         *v=0;

  2385.         break;

  2386.     case PIX_FMT_YUV420P:

  2387.     case PIX_FMT_YUV420P16LE:

  2388.     case PIX_FMT_YUV420P16BE:

  2389.     case PIX_FMT_YUVA420P:

  2390.     case PIX_FMT_GRAY16BE:

  2391.     case PIX_FMT_GRAY16LE:

  2392.     case PIX_FMT_GRAY8: //FIXME remove after different subsamplings are fully implemented

  2393.     case PIX_FMT_NV12:

  2394.     case PIX_FMT_NV21:

  2395.         *h=1;

  2396.         *v=1;

  2397.         break;

  2398.     case PIX_FMT_YUV440P:

  2399.         *h=0;

  2400.         *v=1;

  2401.         break;

  2402.     case PIX_FMT_YUV410P:

  2403.         *h=2;

  2404.         *v=2;

  2405.         break;

  2406.     case PIX_FMT_YUV444P:

  2407.     case PIX_FMT_YUV444P16LE:

  2408.     case PIX_FMT_YUV444P16BE:

  2409.         *h=0;

  2410.         *v=0;

  2411.         break;

  2412.     case PIX_FMT_YUV422P:

  2413.     case PIX_FMT_YUV422P16LE:

  2414.     case PIX_FMT_YUV422P16BE:

  2415.         *h=1;

  2416.         *v=0;

  2417.         break;

  2418.     case PIX_FMT_YUV411P:

  2419.         *h=2;

  2420.         *v=0;

  2421.         break;

  2422.     default:

  2423.         *h=0;

  2424.         *v=0;

  2425.         break;

  2426.     }

  2427. }

  2428. 

  2429. static uint16_t roundToInt16(int64_t f)

  2430. {

  2431.     int r= (f + (1<<15))>>16;

  2432.          if (r<-0x7FFF) return 0x8000;

  2433.     else if (r> 0x7FFF) return 0x7FFF;

  2434.     else                return r;

  2435. }

  2436. 

  2437. int sws_setColorspaceDetails(SwsContext *c, const int inv_table[4], int srcRange, const int table[4], int dstRange, int brightness, int contrast, int saturation)

  2438. {

  2439.     int64_t crv =  inv_table[0];

  2440.     int64_t cbu =  inv_table[1];

  2441.     int64_t cgu = -inv_table[2];

  2442.     int64_t cgv = -inv_table[3];

  2443.     int64_t cy  = 1<<16;

  2444.     int64_t oy  = 0;

  2445. 

  2446.     memcpy(c->srcColorspaceTable, inv_table, sizeof(int)*4);

  2447.     memcpy(c->dstColorspaceTable,     table, sizeof(int)*4);

  2448. 

  2449.     c->brightness= brightness;

  2450.     c->contrast  = contrast;

  2451.     c->saturation= saturation;

  2452.     c->srcRange  = srcRange;

  2453.     c->dstRange  = dstRange;

  2454.     if (isYUV(c->dstFormat) || isGray(c->dstFormat)) return -1;

  2455. 

  2456.     c->uOffset=   0x0400040004000400LL;

  2457.     c->vOffset=   0x0400040004000400LL;

  2458. 

  2459.     if (!srcRange) {

  2460.         cy= (cy*255) / 219;

  2461.         oy= 16<<16;

  2462.     } else {

  2463.         crv= (crv*224) / 255;

  2464.         cbu= (cbu*224) / 255;

  2465.         cgu= (cgu*224) / 255;

  2466.         cgv= (cgv*224) / 255;

  2467.     }

  2468. 

  2469.     cy = (cy *contrast             )>>16;

  2470.     crv= (crv*contrast * saturation)>>32;

  2471.     cbu= (cbu*contrast * saturation)>>32;

  2472.     cgu= (cgu*contrast * saturation)>>32;

  2473.     cgv= (cgv*contrast * saturation)>>32;

  2474. 

  2475.     oy -= 256*brightness;

  2476. 

  2477.     c->yCoeff=    roundToInt16(cy *8192) * 0x0001000100010001ULL;

  2478.     c->vrCoeff=   roundToInt16(crv*8192) * 0x0001000100010001ULL;

  2479.     c->ubCoeff=   roundToInt16(cbu*8192) * 0x0001000100010001ULL;

  2480.     c->vgCoeff=   roundToInt16(cgv*8192) * 0x0001000100010001ULL;

  2481.     c->ugCoeff=   roundToInt16(cgu*8192) * 0x0001000100010001ULL;

  2482.     c->yOffset=   roundToInt16(oy *   8) * 0x0001000100010001ULL;

  2483. 

  2484.     c->yuv2rgb_y_coeff  = (int16_t)roundToInt16(cy <<13);

  2485.     c->yuv2rgb_y_offset = (int16_t)roundToInt16(oy << 9);

  2486.     c->yuv2rgb_v2r_coeff= (int16_t)roundToInt16(crv<<13);

  2487.     c->yuv2rgb_v2g_coeff= (int16_t)roundToInt16(cgv<<13);

  2488.     c->yuv2rgb_u2g_coeff= (int16_t)roundToInt16(cgu<<13);

  2489.     c->yuv2rgb_u2b_coeff= (int16_t)roundToInt16(cbu<<13);

  2490. 

  2491.     ff_yuv2rgb_c_init_tables(c, inv_table, srcRange, brightness, contrast, saturation);

  2492.     //FIXME factorize

  2493. 

  2494. #ifdef COMPILE_ALTIVEC

  2495.     if (c->flags & SWS_CPU_CAPS_ALTIVEC)

  2496.         ff_yuv2rgb_init_tables_altivec(c, inv_table, brightness, contrast, saturation);

  2497. #endif

  2498.     return 0;

  2499. }

  2500. 

  2501. int sws_getColorspaceDetails(SwsContext *c, int **inv_table, int *srcRange, int **table, int *dstRange, int *brightness, int *contrast, int *saturation)

  2502. {

  2503.     if (isYUV(c->dstFormat) || isGray(c->dstFormat)) return -1;

  2504. 

  2505.     *inv_table = c->srcColorspaceTable;

  2506.     *table     = c->dstColorspaceTable;

  2507.     *srcRange  = c->srcRange;

  2508.     *dstRange  = c->dstRange;

  2509.     *brightness= c->brightness;

  2510.     *contrast  = c->contrast;

  2511.     *saturation= c->saturation;

  2512. 

  2513.     return 0;

  2514. }

  2515. 

  2516. static int handle_jpeg(enum PixelFormat *format)

  2517. {

  2518.     switch (*format) {

  2519.     case PIX_FMT_YUVJ420P:

  2520.         *format = PIX_FMT_YUV420P;

  2521.         return 1;

  2522.     case PIX_FMT_YUVJ422P:

  2523.         *format = PIX_FMT_YUV422P;

  2524.         return 1;

  2525.     case PIX_FMT_YUVJ444P:

  2526.         *format = PIX_FMT_YUV444P;

  2527.         return 1;

  2528.     case PIX_FMT_YUVJ440P:

  2529.         *format = PIX_FMT_YUV440P;

  2530.         return 1;

  2531.     default:

  2532.         return 0;

  2533.     }

  2534. }

  2535. 

  2536. SwsContext *sws_getContext(int srcW, int srcH, enum PixelFormat srcFormat, int dstW, int dstH, enum PixelFormat dstFormat, int flags,

  2537.                            SwsFilter *srcFilter, SwsFilter *dstFilter, const double *param)

  2538. {

  2539. 

  2540.     SwsContext *c;

  2541.     int i;

  2542.     int usesVFilter, usesHFilter;

  2543.     int unscaled, needsDither;

  2544.     int srcRange, dstRange;

  2545.     SwsFilter dummyFilter= {NULL, NULL, NULL, NULL};

  2546. #if ARCH_X86

  2547.     if (flags & SWS_CPU_CAPS_MMX)

  2548.         __asm__ volatile("emms\n\t"::: "memory");

  2549. #endif

  2550. 

  2551. #if !CONFIG_RUNTIME_CPUDETECT //ensure that the flags match the compiled variant if cpudetect is off

  2552.     flags &= ~(SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_MMX2|SWS_CPU_CAPS_3DNOW|SWS_CPU_CAPS_ALTIVEC|SWS_CPU_CAPS_BFIN);

  2553. #if   COMPILE_TEMPLATE_MMX2

  2554.     flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_MMX2;

  2555. #elif COMPILE_TEMPLATE_AMD3DNOW

  2556.     flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_3DNOW;

  2557. #elif COMPILE_TEMPLATE_MMX

  2558.     flags |= SWS_CPU_CAPS_MMX;

  2559. #elif COMPILE_TEMPLATE_ALTIVEC

  2560.     flags |= SWS_CPU_CAPS_ALTIVEC;

  2561. #elif ARCH_BFIN

  2562.     flags |= SWS_CPU_CAPS_BFIN;

  2563. #endif

  2564. #endif /* CONFIG_RUNTIME_CPUDETECT */

  2565.     if (clip_table[512] != 255) globalInit();

  2566.     if (!rgb15to16) sws_rgb2rgb_init(flags);

  2567. 

  2568.     unscaled = (srcW == dstW && srcH == dstH);

  2569.     needsDither= (isBGR(dstFormat) || isRGB(dstFormat))

  2570.         && (fmt_depth(dstFormat))<24

  2571.         && ((fmt_depth(dstFormat))<(fmt_depth(srcFormat)) || (!(isRGB(srcFormat) || isBGR(srcFormat))));

  2572. 

  2573.     srcRange = handle_jpeg(&srcFormat);

  2574.     dstRange = handle_jpeg(&dstFormat);

  2575. 

  2576.     if (!isSupportedIn(srcFormat)) {

  2577.         av_log(NULL, AV_LOG_ERROR, "swScaler: %s is not supported as input pixel format\n", sws_format_name(srcFormat));

  2578.         return NULL;

  2579.     }

  2580.     if (!isSupportedOut(dstFormat)) {

  2581.         av_log(NULL, AV_LOG_ERROR, "swScaler: %s is not supported as output pixel format\n", sws_format_name(dstFormat));

  2582.         return NULL;

  2583.     }

  2584. 

  2585.     i= flags & ( SWS_POINT

  2586.                 |SWS_AREA

  2587.                 |SWS_BILINEAR

  2588.                 |SWS_FAST_BILINEAR

  2589.                 |SWS_BICUBIC

  2590.                 |SWS_X

  2591.                 |SWS_GAUSS

  2592.                 |SWS_LANCZOS

  2593.                 |SWS_SINC

  2594.                 |SWS_SPLINE

  2595.                 |SWS_BICUBLIN);

  2596.     if(!i || (i & (i-1))) {

  2597.         av_log(NULL, AV_LOG_ERROR, "swScaler: Exactly one scaler algorithm must be chosen\n");

  2598.         return NULL;

  2599.     }

  2600. 

  2601.     /* sanity check */

  2602.     if (srcW<4 || srcH<1 || dstW<8 || dstH<1) { //FIXME check if these are enough and try to lowwer them after fixing the relevant parts of the code

  2603.         av_log(NULL, AV_LOG_ERROR, "swScaler: %dx%d -> %dx%d is invalid scaling dimension\n",

  2604.                srcW, srcH, dstW, dstH);

  2605.         return NULL;

  2606.     }

  2607.     if(srcW > VOFW || dstW > VOFW) {

  2608.         av_log(NULL, AV_LOG_ERROR, "swScaler: Compile-time maximum width is "AV_STRINGIFY(VOFW)" change VOF/VOFW and recompile\n");

  2609.         return NULL;

  2610.     }

  2611. 

  2612.     if (!dstFilter) dstFilter= &dummyFilter;

  2613.     if (!srcFilter) srcFilter= &dummyFilter;

  2614. 

  2615.     FF_ALLOCZ_OR_GOTO(NULL, c, sizeof(SwsContext), fail);

  2616. 

  2617.     c->av_class = &sws_context_class;

  2618.     c->srcW= srcW;

  2619.     c->srcH= srcH;

  2620.     c->dstW= dstW;

  2621.     c->dstH= dstH;

  2622.     c->lumXInc= ((srcW<<16) + (dstW>>1))/dstW;

  2623.     c->lumYInc= ((srcH<<16) + (dstH>>1))/dstH;

  2624.     c->flags= flags;

  2625.     c->dstFormat= dstFormat;

  2626.     c->srcFormat= srcFormat;

  2627.     c->vRounder= 4* 0x0001000100010001ULL;

  2628. 

  2629.     usesHFilter= usesVFilter= 0;

  2630.     if (dstFilter->lumV && dstFilter->lumV->length>1) usesVFilter=1;

  2631.     if (dstFilter->lumH && dstFilter->lumH->length>1) usesHFilter=1;

  2632.     if (dstFilter->chrV && dstFilter->chrV->length>1) usesVFilter=1;

  2633.     if (dstFilter->chrH && dstFilter->chrH->length>1) usesHFilter=1;

  2634.     if (srcFilter->lumV && srcFilter->lumV->length>1) usesVFilter=1;

  2635.     if (srcFilter->lumH && srcFilter->lumH->length>1) usesHFilter=1;

  2636.     if (srcFilter->chrV && srcFilter->chrV->length>1) usesVFilter=1;

  2637.     if (srcFilter->chrH && srcFilter->chrH->length>1) usesHFilter=1;

  2638. 

  2639.     getSubSampleFactors(&c->chrSrcHSubSample, &c->chrSrcVSubSample, srcFormat);

  2640.     getSubSampleFactors(&c->chrDstHSubSample, &c->chrDstVSubSample, dstFormat);

  2641. 

  2642.     // reuse chroma for 2 pixels RGB/BGR unless user wants full chroma interpolation

  2643.     if ((isBGR(dstFormat) || isRGB(dstFormat)) && !(flags&SWS_FULL_CHR_H_INT)) c->chrDstHSubSample=1;

  2644. 

  2645.     // drop some chroma lines if the user wants it

  2646.     c->vChrDrop= (flags&SWS_SRC_V_CHR_DROP_MASK)>>SWS_SRC_V_CHR_DROP_SHIFT;

  2647.     c->chrSrcVSubSample+= c->vChrDrop;

  2648. 

  2649.     // drop every other pixel for chroma calculation unless user wants full chroma

  2650.     if ((isBGR(srcFormat) || isRGB(srcFormat)) && !(flags&SWS_FULL_CHR_H_INP)

  2651.       && srcFormat!=PIX_FMT_RGB8      && srcFormat!=PIX_FMT_BGR8

  2652.       && srcFormat!=PIX_FMT_RGB4      && srcFormat!=PIX_FMT_BGR4

  2653.       && srcFormat!=PIX_FMT_RGB4_BYTE && srcFormat!=PIX_FMT_BGR4_BYTE

  2654.       && ((dstW>>c->chrDstHSubSample) <= (srcW>>1) || (flags&(SWS_FAST_BILINEAR|SWS_POINT))))

  2655.         c->chrSrcHSubSample=1;

  2656. 

  2657.     if (param) {

  2658.         c->param[0] = param[0];

  2659.         c->param[1] = param[1];

  2660.     } else {

  2661.         c->param[0] =

  2662.         c->param[1] = SWS_PARAM_DEFAULT;

  2663.     }

  2664. 

  2665.     // Note the -((-x)>>y) is so that we always round toward +inf.

  2666.     c->chrSrcW= -((-srcW) >> c->chrSrcHSubSample);

  2667.     c->chrSrcH= -((-srcH) >> c->chrSrcVSubSample);

  2668.     c->chrDstW= -((-dstW) >> c->chrDstHSubSample);

  2669.     c->chrDstH= -((-dstH) >> c->chrDstVSubSample);

  2670. 

  2671.     sws_setColorspaceDetails(c, ff_yuv2rgb_coeffs[SWS_CS_DEFAULT], srcRange, ff_yuv2rgb_coeffs[SWS_CS_DEFAULT] /* FIXME*/, dstRange, 0, 1<<16, 1<<16);

  2672. 

  2673.     /* unscaled special cases */

  2674.     if (unscaled && !usesHFilter && !usesVFilter && (srcRange == dstRange || isBGR(dstFormat) || isRGB(dstFormat))) {

  2675.         /* yv12_to_nv12 */

  2676.         if ((srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUVA420P) && (dstFormat == PIX_FMT_NV12 || dstFormat == PIX_FMT_NV21)) {

  2677.             c->swScale= PlanarToNV12Wrapper;

  2678.         }

  2679.         /* yuv2bgr */

  2680.         if ((srcFormat==PIX_FMT_YUV420P || srcFormat==PIX_FMT_YUV422P || srcFormat==PIX_FMT_YUVA420P) && (isBGR(dstFormat) || isRGB(dstFormat))

  2681.             && !(flags & SWS_ACCURATE_RND) && !(dstH&1)) {

  2682.             c->swScale= ff_yuv2rgb_get_func_ptr(c);

  2683.         }

  2684. 

  2685.         if (srcFormat==PIX_FMT_YUV410P && (dstFormat==PIX_FMT_YUV420P || dstFormat==PIX_FMT_YUVA420P) && !(flags & SWS_BITEXACT)) {

  2686.             c->swScale= yvu9toyv12Wrapper;

  2687.         }

  2688. 

  2689.         /* bgr24toYV12 */

  2690.         if (srcFormat==PIX_FMT_BGR24 && (dstFormat==PIX_FMT_YUV420P || dstFormat==PIX_FMT_YUVA420P) && !(flags & SWS_ACCURATE_RND))

  2691.             c->swScale= bgr24toyv12Wrapper;

  2692. 

  2693.         /* RGB/BGR -> RGB/BGR (no dither needed forms) */

  2694.         if (  (isBGR(srcFormat) || isRGB(srcFormat))

  2695.            && (isBGR(dstFormat) || isRGB(dstFormat))

  2696.            && srcFormat != PIX_FMT_BGR8      && dstFormat != PIX_FMT_BGR8

  2697.            && srcFormat != PIX_FMT_RGB8      && dstFormat != PIX_FMT_RGB8

  2698.            && srcFormat != PIX_FMT_BGR4      && dstFormat != PIX_FMT_BGR4

  2699.            && srcFormat != PIX_FMT_RGB4      && dstFormat != PIX_FMT_RGB4

  2700.            && srcFormat != PIX_FMT_BGR4_BYTE && dstFormat != PIX_FMT_BGR4_BYTE

  2701.            && srcFormat != PIX_FMT_RGB4_BYTE && dstFormat != PIX_FMT_RGB4_BYTE

  2702.            && srcFormat != PIX_FMT_MONOBLACK && dstFormat != PIX_FMT_MONOBLACK

  2703.            && srcFormat != PIX_FMT_MONOWHITE && dstFormat != PIX_FMT_MONOWHITE

  2704.                                              && dstFormat != PIX_FMT_RGB32_1

  2705.                                              && dstFormat != PIX_FMT_BGR32_1

  2706.            && srcFormat != PIX_FMT_RGB48LE   && dstFormat != PIX_FMT_RGB48LE

  2707.            && srcFormat != PIX_FMT_RGB48BE   && dstFormat != PIX_FMT_RGB48BE

  2708.            && (!needsDither || (c->flags&(SWS_FAST_BILINEAR|SWS_POINT))))

  2709.              c->swScale= rgb2rgbWrapper;

  2710. 

  2711.         if ((usePal(srcFormat) && (

  2712.                  dstFormat == PIX_FMT_RGB32   ||

  2713.                  dstFormat == PIX_FMT_RGB32_1 ||

  2714.                  dstFormat == PIX_FMT_RGB24   ||

  2715.                  dstFormat == PIX_FMT_BGR32   ||

  2716.                  dstFormat == PIX_FMT_BGR32_1 ||

  2717.                  dstFormat == PIX_FMT_BGR24)))

  2718.              c->swScale= pal2rgbWrapper;

  2719. 

  2720.         if (srcFormat == PIX_FMT_YUV422P) {

  2721.             if (dstFormat == PIX_FMT_YUYV422)

  2722.                 c->swScale= YUV422PToYuy2Wrapper;

  2723.             else if (dstFormat == PIX_FMT_UYVY422)

  2724.                 c->swScale= YUV422PToUyvyWrapper;

  2725.         }

  2726. 

  2727.         /* LQ converters if -sws 0 or -sws 4*/

  2728.         if (c->flags&(SWS_FAST_BILINEAR|SWS_POINT)) {

  2729.             /* yv12_to_yuy2 */

  2730.             if (srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUVA420P) {

  2731.                 if (dstFormat == PIX_FMT_YUYV422)

  2732.                     c->swScale= PlanarToYuy2Wrapper;

  2733.                 else if (dstFormat == PIX_FMT_UYVY422)

  2734.                     c->swScale= PlanarToUyvyWrapper;

  2735.             }

  2736.         }

  2737.         if(srcFormat == PIX_FMT_YUYV422 && (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P))

  2738.             c->swScale= YUYV2YUV420Wrapper;

  2739.         if(srcFormat == PIX_FMT_UYVY422 && (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P))

  2740.             c->swScale= UYVY2YUV420Wrapper;

  2741.         if(srcFormat == PIX_FMT_YUYV422 && dstFormat == PIX_FMT_YUV422P)

  2742.             c->swScale= YUYV2YUV422Wrapper;

  2743.         if(srcFormat == PIX_FMT_UYVY422 && dstFormat == PIX_FMT_YUV422P)

  2744.             c->swScale= UYVY2YUV422Wrapper;

  2745. 

  2746. #ifdef COMPILE_ALTIVEC

  2747.         if ((c->flags & SWS_CPU_CAPS_ALTIVEC) &&

  2748.             !(c->flags & SWS_BITEXACT) &&

  2749.             srcFormat == PIX_FMT_YUV420P) {

  2750.           // unscaled YV12 -> packed YUV, we want speed

  2751.           if (dstFormat == PIX_FMT_YUYV422)

  2752.               c->swScale= yv12toyuy2_unscaled_altivec;

  2753.           else if (dstFormat == PIX_FMT_UYVY422)

  2754.               c->swScale= yv12touyvy_unscaled_altivec;

  2755.         }

  2756. #endif

  2757. 

  2758.         /* simple copy */

  2759.         if (  srcFormat == dstFormat

  2760.             || (srcFormat == PIX_FMT_YUVA420P && dstFormat == PIX_FMT_YUV420P)

  2761.             || (srcFormat == PIX_FMT_YUV420P && dstFormat == PIX_FMT_YUVA420P)

  2762.             || (isPlanarYUV(srcFormat) && isGray(dstFormat))

  2763.             || (isPlanarYUV(dstFormat) && isGray(srcFormat))

  2764.             || (isGray(dstFormat) && isGray(srcFormat))

  2765.             || (isPlanarYUV(srcFormat) && isPlanarYUV(dstFormat)

  2766.                 && c->chrDstHSubSample == c->chrSrcHSubSample

  2767.                 && c->chrDstVSubSample == c->chrSrcVSubSample

  2768.                 && dstFormat != PIX_FMT_NV12 && dstFormat != PIX_FMT_NV21

  2769.                 && srcFormat != PIX_FMT_NV12 && srcFormat != PIX_FMT_NV21))

  2770.         {

  2771.             if (isPacked(c->srcFormat))

  2772.                 c->swScale= packedCopy;

  2773.             else /* Planar YUV or gray */

  2774.                 c->swScale= planarCopy;

  2775.         }

  2776. #if ARCH_BFIN

  2777.         if (flags & SWS_CPU_CAPS_BFIN)

  2778.             ff_bfin_get_unscaled_swscale (c);

  2779. #endif

  2780. 

  2781.         if (c->swScale) {

  2782.             if (flags&SWS_PRINT_INFO)

  2783.                 av_log(c, AV_LOG_INFO, "using unscaled %s -> %s special converter\n",

  2784.                        sws_format_name(srcFormat), sws_format_name(dstFormat));

  2785.             return c;

  2786.         }

  2787.     }

  2788. 

  2789.     if (flags & SWS_CPU_CAPS_MMX2) {

  2790.         c->canMMX2BeUsed= (dstW >=srcW && (dstW&31)==0 && (srcW&15)==0) ? 1 : 0;

  2791.         if (!c->canMMX2BeUsed && dstW >=srcW && (srcW&15)==0 && (flags&SWS_FAST_BILINEAR)) {

  2792.             if (flags&SWS_PRINT_INFO)

  2793.                 av_log(c, AV_LOG_INFO, "output width is not a multiple of 32 -> no MMX2 scaler\n");

  2794.         }

  2795.         if (usesHFilter) c->canMMX2BeUsed=0;

  2796.     }

  2797.     else

  2798.         c->canMMX2BeUsed=0;

  2799. 

  2800.     c->chrXInc= ((c->chrSrcW<<16) + (c->chrDstW>>1))/c->chrDstW;

  2801.     c->chrYInc= ((c->chrSrcH<<16) + (c->chrDstH>>1))/c->chrDstH;

  2802. 

  2803.     // match pixel 0 of the src to pixel 0 of dst and match pixel n-2 of src to pixel n-2 of dst

  2804.     // but only for the FAST_BILINEAR mode otherwise do correct scaling

  2805.     // n-2 is the last chrominance sample available

  2806.     // this is not perfect, but no one should notice the difference, the more correct variant

  2807.     // would be like the vertical one, but that would require some special code for the

  2808.     // first and last pixel

  2809.     if (flags&SWS_FAST_BILINEAR) {

  2810.         if (c->canMMX2BeUsed) {

  2811.             c->lumXInc+= 20;

  2812.             c->chrXInc+= 20;

  2813.         }

  2814.         //we don't use the x86 asm scaler if MMX is available

  2815.         else if (flags & SWS_CPU_CAPS_MMX) {

  2816.             c->lumXInc = ((srcW-2)<<16)/(dstW-2) - 20;

  2817.             c->chrXInc = ((c->chrSrcW-2)<<16)/(c->chrDstW-2) - 20;

  2818.         }

  2819.     }

  2820. 

  2821.     /* precalculate horizontal scaler filter coefficients */

  2822.     {

  2823.         const int filterAlign=

  2824.             (flags & SWS_CPU_CAPS_MMX) ? 4 :

  2825.             (flags & SWS_CPU_CAPS_ALTIVEC) ? 8 :

  2826.             1;

  2827. 

  2828.         if (initFilter(&c->hLumFilter, &c->hLumFilterPos, &c->hLumFilterSize, c->lumXInc,

  2829.                        srcW      ,       dstW, filterAlign, 1<<14,

  2830.                        (flags&SWS_BICUBLIN) ? (flags|SWS_BICUBIC)  : flags,

  2831.                        srcFilter->lumH, dstFilter->lumH, c->param) < 0)

  2832.             goto fail;

  2833.         if (initFilter(&c->hChrFilter, &c->hChrFilterPos, &c->hChrFilterSize, c->chrXInc,

  2834.                        c->chrSrcW, c->chrDstW, filterAlign, 1<<14,

  2835.                        (flags&SWS_BICUBLIN) ? (flags|SWS_BILINEAR) : flags,

  2836.                        srcFilter->chrH, dstFilter->chrH, c->param) < 0)

  2837.             goto fail;

  2838. 

  2839. #if defined(COMPILE_MMX2)

  2840. // can't downscale !!!

  2841.         if (c->canMMX2BeUsed && (flags & SWS_FAST_BILINEAR)) {

  2842.             c->lumMmx2FilterCodeSize = initMMX2HScaler(      dstW, c->lumXInc, NULL, NULL, NULL, 8);

  2843.             c->chrMmx2FilterCodeSize = initMMX2HScaler(c->chrDstW, c->chrXInc, NULL, NULL, NULL, 4);

  2844. 

  2845. #ifdef MAP_ANONYMOUS

  2846.             c->lumMmx2FilterCode = mmap(NULL, c->lumMmx2FilterCodeSize, PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, 0, 0);

  2847.             c->chrMmx2FilterCode = mmap(NULL, c->chrMmx2FilterCodeSize, PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, 0, 0);

  2848. #elif HAVE_VIRTUALALLOC

  2849.             c->lumMmx2FilterCode = VirtualAlloc(NULL, c->lumMmx2FilterCodeSize, MEM_COMMIT, PAGE_EXECUTE_READWRITE);

  2850.             c->chrMmx2FilterCode = VirtualAlloc(NULL, c->chrMmx2FilterCodeSize, MEM_COMMIT, PAGE_EXECUTE_READWRITE);

  2851. #else

  2852.             c->lumMmx2FilterCode = av_malloc(c->lumMmx2FilterCodeSize);

  2853.             c->chrMmx2FilterCode = av_malloc(c->chrMmx2FilterCodeSize);

  2854. #endif

  2855. 

  2856.             FF_ALLOCZ_OR_GOTO(c, c->lumMmx2Filter   , (dstW        /8+8)*sizeof(int16_t), fail);

  2857.             FF_ALLOCZ_OR_GOTO(c, c->chrMmx2Filter   , (c->chrDstW  /4+8)*sizeof(int16_t), fail);

  2858.             FF_ALLOCZ_OR_GOTO(c, c->lumMmx2FilterPos, (dstW      /2/8+8)*sizeof(int32_t), fail);

  2859.             FF_ALLOCZ_OR_GOTO(c, c->chrMmx2FilterPos, (c->chrDstW/2/4+8)*sizeof(int32_t), fail);

  2860. 

  2861.             initMMX2HScaler(      dstW, c->lumXInc, c->lumMmx2FilterCode, c->lumMmx2Filter, c->lumMmx2FilterPos, 8);

  2862.             initMMX2HScaler(c->chrDstW, c->chrXInc, c->chrMmx2FilterCode, c->chrMmx2Filter, c->chrMmx2FilterPos, 4);

  2863. 

  2864. #ifdef MAP_ANONYMOUS

  2865.             mprotect(c->lumMmx2FilterCode, c->lumMmx2FilterCodeSize, PROT_EXEC | PROT_READ);

  2866.             mprotect(c->chrMmx2FilterCode, c->chrMmx2FilterCodeSize, PROT_EXEC | PROT_READ);

  2867. #endif

  2868.         }

  2869. #endif /* defined(COMPILE_MMX2) */

  2870.     } // initialize horizontal stuff

  2871. 

  2872. 

  2873. 

  2874.     /* precalculate vertical scaler filter coefficients */

  2875.     {

  2876.         const int filterAlign=

  2877.             (flags & SWS_CPU_CAPS_MMX) && (flags & SWS_ACCURATE_RND) ? 2 :

  2878.             (flags & SWS_CPU_CAPS_ALTIVEC) ? 8 :

  2879.             1;

  2880. 

  2881.         if (initFilter(&c->vLumFilter, &c->vLumFilterPos, &c->vLumFilterSize, c->lumYInc,

  2882.                    srcH      ,        dstH, filterAlign, (1<<12),

  2883.                    (flags&SWS_BICUBLIN) ? (flags|SWS_BICUBIC)  : flags,

  2884.                    srcFilter->lumV, dstFilter->lumV, c->param) < 0)

  2885.             goto fail;

  2886.         if (initFilter(&c->vChrFilter, &c->vChrFilterPos, &c->vChrFilterSize, c->chrYInc,

  2887.                    c->chrSrcH, c->chrDstH, filterAlign, (1<<12),

  2888.                    (flags&SWS_BICUBLIN) ? (flags|SWS_BILINEAR) : flags,

  2889.                    srcFilter->chrV, dstFilter->chrV, c->param) < 0)

  2890.             goto fail;

  2891. 

  2892. #ifdef COMPILE_ALTIVEC

  2893.         FF_ALLOC_OR_GOTO(c, c->vYCoeffsBank, sizeof (vector signed short)*c->vLumFilterSize*c->dstH, fail);

  2894.         FF_ALLOC_OR_GOTO(c, c->vCCoeffsBank, sizeof (vector signed short)*c->vChrFilterSize*c->chrDstH, fail);

  2895. 

  2896.         for (i=0;i<c->vLumFilterSize*c->dstH;i++) {

  2897.             int j;

  2898.             short *p = (short *)&c->vYCoeffsBank[i];

  2899.             for (j=0;j<8;j++)

  2900.                 p[j] = c->vLumFilter[i];

  2901.         }

  2902. 

  2903.         for (i=0;i<c->vChrFilterSize*c->chrDstH;i++) {

  2904.             int j;

  2905.             short *p = (short *)&c->vCCoeffsBank[i];

  2906.             for (j=0;j<8;j++)

  2907.                 p[j] = c->vChrFilter[i];

  2908.         }

  2909. #endif

  2910.     }

  2911. 

  2912.     // calculate buffer sizes so that they won't run out while handling these damn slices

  2913.     c->vLumBufSize= c->vLumFilterSize;

  2914.     c->vChrBufSize= c->vChrFilterSize;

  2915.     for (i=0; i<dstH; i++) {

  2916.         int chrI= i*c->chrDstH / dstH;

  2917.         int nextSlice= FFMAX(c->vLumFilterPos[i   ] + c->vLumFilterSize - 1,

  2918.                            ((c->vChrFilterPos[chrI] + c->vChrFilterSize - 1)<<c->chrSrcVSubSample));

  2919. 

  2920.         nextSlice>>= c->chrSrcVSubSample;

  2921.         nextSlice<<= c->chrSrcVSubSample;

  2922.         if (c->vLumFilterPos[i   ] + c->vLumBufSize < nextSlice)

  2923.             c->vLumBufSize= nextSlice - c->vLumFilterPos[i];

  2924.         if (c->vChrFilterPos[chrI] + c->vChrBufSize < (nextSlice>>c->chrSrcVSubSample))

  2925.             c->vChrBufSize= (nextSlice>>c->chrSrcVSubSample) - c->vChrFilterPos[chrI];

  2926.     }

  2927. 

  2928.     // allocate pixbufs (we use dynamic allocation because otherwise we would need to

  2929.     // allocate several megabytes to handle all possible cases)

  2930.     FF_ALLOC_OR_GOTO(c, c->lumPixBuf, c->vLumBufSize*2*sizeof(int16_t*), fail);

  2931.     FF_ALLOC_OR_GOTO(c, c->chrPixBuf, c->vChrBufSize*2*sizeof(int16_t*), fail);

  2932.     if (CONFIG_SWSCALE_ALPHA && isALPHA(c->srcFormat) && isALPHA(c->dstFormat))

  2933.         FF_ALLOCZ_OR_GOTO(c, c->alpPixBuf, c->vLumBufSize*2*sizeof(int16_t*), fail);

  2934.     //Note we need at least one pixel more at the end because of the MMX code (just in case someone wanna replace the 4000/8000)

  2935.     /* align at 16 bytes for AltiVec */

  2936.     for (i=0; i<c->vLumBufSize; i++) {

  2937.         FF_ALLOCZ_OR_GOTO(c, c->lumPixBuf[i+c->vLumBufSize], VOF+1, fail);

  2938.         c->lumPixBuf[i] = c->lumPixBuf[i+c->vLumBufSize];

  2939.     }

  2940.     for (i=0; i<c->vChrBufSize; i++) {

  2941.         FF_ALLOC_OR_GOTO(c, c->chrPixBuf[i+c->vChrBufSize], (VOF+1)*2, fail);

  2942.         c->chrPixBuf[i] = c->chrPixBuf[i+c->vChrBufSize];

  2943.     }

  2944.     if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf)

  2945.         for (i=0; i<c->vLumBufSize; i++) {

  2946.             FF_ALLOCZ_OR_GOTO(c, c->alpPixBuf[i+c->vLumBufSize], VOF+1, fail);

  2947.             c->alpPixBuf[i] = c->alpPixBuf[i+c->vLumBufSize];

  2948.         }

  2949. 

  2950.     //try to avoid drawing green stuff between the right end and the stride end

  2951.     for (i=0; i<c->vChrBufSize; i++) memset(c->chrPixBuf[i], 64, (VOF+1)*2);

  2952. 

  2953.     assert(2*VOFW == VOF);

  2954. 

  2955.     assert(c->chrDstH <= dstH);

  2956. 

  2957.     if (flags&SWS_PRINT_INFO) {

  2958. #ifdef DITHER1XBPP

  2959.         const char *dither= " dithered";

  2960. #else

  2961.         const char *dither= "";

  2962. #endif

  2963.         if (flags&SWS_FAST_BILINEAR)

  2964.             av_log(c, AV_LOG_INFO, "FAST_BILINEAR scaler, ");

  2965.         else if (flags&SWS_BILINEAR)

  2966.             av_log(c, AV_LOG_INFO, "BILINEAR scaler, ");

  2967.         else if (flags&SWS_BICUBIC)

  2968.             av_log(c, AV_LOG_INFO, "BICUBIC scaler, ");

  2969.         else if (flags&SWS_X)

  2970.             av_log(c, AV_LOG_INFO, "Experimental scaler, ");

  2971.         else if (flags&SWS_POINT)

  2972.             av_log(c, AV_LOG_INFO, "Nearest Neighbor / POINT scaler, ");

  2973.         else if (flags&SWS_AREA)

  2974.             av_log(c, AV_LOG_INFO, "Area Averageing scaler, ");

  2975.         else if (flags&SWS_BICUBLIN)

  2976.             av_log(c, AV_LOG_INFO, "luma BICUBIC / chroma BILINEAR scaler, ");

  2977.         else if (flags&SWS_GAUSS)

  2978.             av_log(c, AV_LOG_INFO, "Gaussian scaler, ");

  2979.         else if (flags&SWS_SINC)

  2980.             av_log(c, AV_LOG_INFO, "Sinc scaler, ");

  2981.         else if (flags&SWS_LANCZOS)

  2982.             av_log(c, AV_LOG_INFO, "Lanczos scaler, ");

  2983.         else if (flags&SWS_SPLINE)

  2984.             av_log(c, AV_LOG_INFO, "Bicubic spline scaler, ");

  2985.         else

  2986.             av_log(c, AV_LOG_INFO, "ehh flags invalid?! ");

  2987. 

  2988.         if (dstFormat==PIX_FMT_BGR555 || dstFormat==PIX_FMT_BGR565)

  2989.             av_log(c, AV_LOG_INFO, "from %s to%s %s ",

  2990.                    sws_format_name(srcFormat), dither, sws_format_name(dstFormat));

  2991.         else

  2992.             av_log(c, AV_LOG_INFO, "from %s to %s ",

  2993.                    sws_format_name(srcFormat), sws_format_name(dstFormat));

  2994. 

  2995.         if (flags & SWS_CPU_CAPS_MMX2)

  2996.             av_log(c, AV_LOG_INFO, "using MMX2\n");

  2997.         else if (flags & SWS_CPU_CAPS_3DNOW)

  2998.             av_log(c, AV_LOG_INFO, "using 3DNOW\n");

  2999.         else if (flags & SWS_CPU_CAPS_MMX)

  3000.             av_log(c, AV_LOG_INFO, "using MMX\n");

  3001.         else if (flags & SWS_CPU_CAPS_ALTIVEC)

  3002.             av_log(c, AV_LOG_INFO, "using AltiVec\n");

  3003.         else

  3004.             av_log(c, AV_LOG_INFO, "using C\n");

  3005.     }

  3006. 

  3007.     if (flags & SWS_PRINT_INFO) {

  3008.         if (flags & SWS_CPU_CAPS_MMX) {

  3009.             if (c->canMMX2BeUsed && (flags&SWS_FAST_BILINEAR))

  3010.                 av_log(c, AV_LOG_VERBOSE, "using FAST_BILINEAR MMX2 scaler for horizontal scaling\n");

  3011.             else {

  3012.                 if (c->hLumFilterSize==4)

  3013.                     av_log(c, AV_LOG_VERBOSE, "using 4-tap MMX scaler for horizontal luminance scaling\n");

  3014.                 else if (c->hLumFilterSize==8)

  3015.                     av_log(c, AV_LOG_VERBOSE, "using 8-tap MMX scaler for horizontal luminance scaling\n");

  3016.                 else

  3017.                     av_log(c, AV_LOG_VERBOSE, "using n-tap MMX scaler for horizontal luminance scaling\n");

  3018. 

  3019.                 if (c->hChrFilterSize==4)

  3020.                     av_log(c, AV_LOG_VERBOSE, "using 4-tap MMX scaler for horizontal chrominance scaling\n");

  3021.                 else if (c->hChrFilterSize==8)

  3022.                     av_log(c, AV_LOG_VERBOSE, "using 8-tap MMX scaler for horizontal chrominance scaling\n");

  3023.                 else

  3024.                     av_log(c, AV_LOG_VERBOSE, "using n-tap MMX scaler for horizontal chrominance scaling\n");

  3025.             }

  3026.         } else {

  3027. #if ARCH_X86

  3028.             av_log(c, AV_LOG_VERBOSE, "using x86 asm scaler for horizontal scaling\n");

  3029. #else

  3030.             if (flags & SWS_FAST_BILINEAR)

  3031.                 av_log(c, AV_LOG_VERBOSE, "using FAST_BILINEAR C scaler for horizontal scaling\n");

  3032.             else

  3033.                 av_log(c, AV_LOG_VERBOSE, "using C scaler for horizontal scaling\n");

  3034. #endif

  3035.         }

  3036.         if (isPlanarYUV(dstFormat)) {

  3037.             if (c->vLumFilterSize==1)

  3038.                 av_log(c, AV_LOG_VERBOSE, "using 1-tap %s \"scaler\" for vertical scaling (YV12 like)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  3039.             else

  3040.                 av_log(c, AV_LOG_VERBOSE, "using n-tap %s scaler for vertical scaling (YV12 like)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  3041.         } else {

  3042.             if (c->vLumFilterSize==1 && c->vChrFilterSize==2)

  3043.                 av_log(c, AV_LOG_VERBOSE, "using 1-tap %s \"scaler\" for vertical luminance scaling (BGR)\n"

  3044.                        "      2-tap scaler for vertical chrominance scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  3045.             else if (c->vLumFilterSize==2 && c->vChrFilterSize==2)

  3046.                 av_log(c, AV_LOG_VERBOSE, "using 2-tap linear %s scaler for vertical scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  3047.             else

  3048.                 av_log(c, AV_LOG_VERBOSE, "using n-tap %s scaler for vertical scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  3049.         }

  3050. 

  3051.         if (dstFormat==PIX_FMT_BGR24)

  3052.             av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR24 converter\n",

  3053.                    (flags & SWS_CPU_CAPS_MMX2) ? "MMX2" : ((flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C"));

  3054.         else if (dstFormat==PIX_FMT_RGB32)

  3055.             av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR32 converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  3056.         else if (dstFormat==PIX_FMT_BGR565)

  3057.             av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR16 converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  3058.         else if (dstFormat==PIX_FMT_BGR555)

  3059.             av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR15 converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  3060. 

  3061.         av_log(c, AV_LOG_VERBOSE, "%dx%d -> %dx%d\n", srcW, srcH, dstW, dstH);

  3062.     }

  3063.     if (flags & SWS_PRINT_INFO) {

  3064.         av_log(c, AV_LOG_DEBUG, "lum srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",

  3065.                c->srcW, c->srcH, c->dstW, c->dstH, c->lumXInc, c->lumYInc);

  3066.         av_log(c, AV_LOG_DEBUG, "chr srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",

  3067.                c->chrSrcW, c->chrSrcH, c->chrDstW, c->chrDstH, c->chrXInc, c->chrYInc);

  3068.     }

  3069. 

  3070.     c->swScale= getSwsFunc(c);

  3071.     return c;

  3072. 

  3073. fail:

  3074.     sws_freeContext(c);

  3075.     return NULL;

  3076. }

  3077. 

  3078. static void reset_ptr(uint8_t* src[], int format)

  3079. {

  3080.     if(!isALPHA(format))

  3081.         src[3]=NULL;

  3082.     if(!isPlanarYUV(format)) {

  3083.         src[3]=src[2]=NULL;

  3084.         if(   format != PIX_FMT_PAL8

  3085.            && format != PIX_FMT_RGB8

  3086.            && format != PIX_FMT_BGR8

  3087.            && format != PIX_FMT_RGB4_BYTE

  3088.            && format != PIX_FMT_BGR4_BYTE

  3089.           )

  3090.             src[1]= NULL;

  3091.     }

  3092. }

  3093. 

  3094. /**

  3095.  * swscale wrapper, so we don't need to export the SwsContext.

  3096.  * Assumes planar YUV to be in YUV order instead of YVU.

  3097.  */

  3098. int sws_scale(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  3099.               int srcSliceH, uint8_t* dst[], int dstStride[])

  3100. {

  3101.     int i;

  3102.     uint8_t* src2[4]= {src[0], src[1], src[2], src[3]};

  3103.     uint8_t* dst2[4]= {dst[0], dst[1], dst[2], dst[3]};

  3104. 

  3105.     if (c->sliceDir == 0 && srcSliceY != 0 && srcSliceY + srcSliceH != c->srcH) {

  3106.         av_log(c, AV_LOG_ERROR, "Slices start in the middle!\n");

  3107.         return 0;

  3108.     }

  3109.     if (c->sliceDir == 0) {

  3110.         if (srcSliceY == 0) c->sliceDir = 1; else c->sliceDir = -1;

  3111.     }

  3112. 

  3113.     if (usePal(c->srcFormat)) {

  3114.         for (i=0; i<256; i++) {

  3115.             int p, r, g, b,y,u,v;

  3116.             if(c->srcFormat == PIX_FMT_PAL8) {

  3117.                 p=((uint32_t*)(src[1]))[i];

  3118.                 r= (p>>16)&0xFF;

  3119.                 g= (p>> 8)&0xFF;

  3120.                 b=  p     &0xFF;

  3121.             } else if(c->srcFormat == PIX_FMT_RGB8) {

  3122.                 r= (i>>5    )*36;

  3123.                 g= ((i>>2)&7)*36;

  3124.                 b= (i&3     )*85;

  3125.             } else if(c->srcFormat == PIX_FMT_BGR8) {

  3126.                 b= (i>>6    )*85;

  3127.                 g= ((i>>3)&7)*36;

  3128.                 r= (i&7     )*36;

  3129.             } else if(c->srcFormat == PIX_FMT_RGB4_BYTE) {

  3130.                 r= (i>>3    )*255;

  3131.                 g= ((i>>1)&3)*85;

  3132.                 b= (i&1     )*255;

  3133.             } else {

  3134.                 assert(c->srcFormat == PIX_FMT_BGR4_BYTE);

  3135.                 b= (i>>3    )*255;

  3136.                 g= ((i>>1)&3)*85;

  3137.                 r= (i&1     )*255;

  3138.             }

  3139.             y= av_clip_uint8((RY*r + GY*g + BY*b + ( 33<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);

  3140.             u= av_clip_uint8((RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);

  3141.             v= av_clip_uint8((RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);

  3142.             c->pal_yuv[i]= y + (u<<8) + (v<<16);

  3143. 

  3144. 

  3145.             switch(c->dstFormat) {

  3146.             case PIX_FMT_BGR32:

  3147. #if !HAVE_BIGENDIAN

  3148.             case PIX_FMT_RGB24:

  3149. #endif

  3150.                 c->pal_rgb[i]=  r + (g<<8) + (b<<16);

  3151.                 break;

  3152.             case PIX_FMT_BGR32_1:

  3153. #if HAVE_BIGENDIAN

  3154.             case PIX_FMT_BGR24:

  3155. #endif

  3156.                 c->pal_rgb[i]= (r + (g<<8) + (b<<16)) << 8;

  3157.                 break;

  3158.             case PIX_FMT_RGB32_1:

  3159. #if HAVE_BIGENDIAN

  3160.             case PIX_FMT_RGB24:

  3161. #endif

  3162.                 c->pal_rgb[i]= (b + (g<<8) + (r<<16)) << 8;

  3163.                 break;

  3164.             case PIX_FMT_RGB32:

  3165. #if !HAVE_BIGENDIAN

  3166.             case PIX_FMT_BGR24:

  3167. #endif

  3168.             default:

  3169.                 c->pal_rgb[i]=  b + (g<<8) + (r<<16);

  3170.             }

  3171.         }

  3172.     }

  3173. 

  3174.     // copy strides, so they can safely be modified

  3175.     if (c->sliceDir == 1) {

  3176.         // slices go from top to bottom

  3177.         int srcStride2[4]= {srcStride[0], srcStride[1], srcStride[2], srcStride[3]};

  3178.         int dstStride2[4]= {dstStride[0], dstStride[1], dstStride[2], dstStride[3]};

  3179. 

  3180.         reset_ptr(src2, c->srcFormat);

  3181.         reset_ptr(dst2, c->dstFormat);

  3182. 

  3183.         /* reset slice direction at end of frame */

  3184.         if (srcSliceY + srcSliceH == c->srcH)

  3185.             c->sliceDir = 0;

  3186. 

  3187.         return c->swScale(c, src2, srcStride2, srcSliceY, srcSliceH, dst2, dstStride2);

  3188.     } else {

  3189.         // slices go from bottom to top => we flip the image internally

  3190.         int srcStride2[4]= {-srcStride[0], -srcStride[1], -srcStride[2], -srcStride[3]};

  3191.         int dstStride2[4]= {-dstStride[0], -dstStride[1], -dstStride[2], -dstStride[3]};

  3192. 

  3193.         src2[0] += (srcSliceH-1)*srcStride[0];

  3194.         if (!usePal(c->srcFormat))

  3195.             src2[1] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[1];

  3196.         src2[2] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[2];

  3197.         src2[3] += (srcSliceH-1)*srcStride[3];

  3198.         dst2[0] += ( c->dstH                      -1)*dstStride[0];

  3199.         dst2[1] += ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[1];

  3200.         dst2[2] += ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[2];

  3201.         dst2[3] += ( c->dstH                      -1)*dstStride[3];

  3202. 

  3203.         reset_ptr(src2, c->srcFormat);

  3204.         reset_ptr(dst2, c->dstFormat);

  3205. 

  3206.         /* reset slice direction at end of frame */

  3207.         if (!srcSliceY)

  3208.             c->sliceDir = 0;

  3209. 

  3210.         return c->swScale(c, src2, srcStride2, c->srcH-srcSliceY-srcSliceH, srcSliceH, dst2, dstStride2);

  3211.     }

  3212. }

  3213. 

  3214. #if LIBSWSCALE_VERSION_MAJOR < 1

  3215. int sws_scale_ordered(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  3216.                       int srcSliceH, uint8_t* dst[], int dstStride[])

  3217. {

  3218.     return sws_scale(c, src, srcStride, srcSliceY, srcSliceH, dst, dstStride);

  3219. }

  3220. #endif

  3221. 

  3222. SwsFilter *sws_getDefaultFilter(float lumaGBlur, float chromaGBlur,

  3223.                                 float lumaSharpen, float chromaSharpen,

  3224.                                 float chromaHShift, float chromaVShift,

  3225.                                 int verbose)

  3226. {

  3227.     SwsFilter *filter= av_malloc(sizeof(SwsFilter));

  3228.     if (!filter)

  3229.         return NULL;

  3230. 

  3231.     if (lumaGBlur!=0.0) {

  3232.         filter->lumH= sws_getGaussianVec(lumaGBlur, 3.0);

  3233.         filter->lumV= sws_getGaussianVec(lumaGBlur, 3.0);

  3234.     } else {

  3235.         filter->lumH= sws_getIdentityVec();

  3236.         filter->lumV= sws_getIdentityVec();

  3237.     }

  3238. 

  3239.     if (chromaGBlur!=0.0) {

  3240.         filter->chrH= sws_getGaussianVec(chromaGBlur, 3.0);

  3241.         filter->chrV= sws_getGaussianVec(chromaGBlur, 3.0);

  3242.     } else {

  3243.         filter->chrH= sws_getIdentityVec();

  3244.         filter->chrV= sws_getIdentityVec();

  3245.     }

  3246. 

  3247.     if (chromaSharpen!=0.0) {

  3248.         SwsVector *id= sws_getIdentityVec();

  3249.         sws_scaleVec(filter->chrH, -chromaSharpen);

  3250.         sws_scaleVec(filter->chrV, -chromaSharpen);

  3251.         sws_addVec(filter->chrH, id);

  3252.         sws_addVec(filter->chrV, id);

  3253.         sws_freeVec(id);

  3254.     }

  3255. 

  3256.     if (lumaSharpen!=0.0) {

  3257.         SwsVector *id= sws_getIdentityVec();

  3258.         sws_scaleVec(filter->lumH, -lumaSharpen);

  3259.         sws_scaleVec(filter->lumV, -lumaSharpen);

  3260.         sws_addVec(filter->lumH, id);

  3261.         sws_addVec(filter->lumV, id);

  3262.         sws_freeVec(id);

  3263.     }

  3264. 

  3265.     if (chromaHShift != 0.0)

  3266.         sws_shiftVec(filter->chrH, (int)(chromaHShift+0.5));

  3267. 

  3268.     if (chromaVShift != 0.0)

  3269.         sws_shiftVec(filter->chrV, (int)(chromaVShift+0.5));

  3270. 

  3271.     sws_normalizeVec(filter->chrH, 1.0);

  3272.     sws_normalizeVec(filter->chrV, 1.0);

  3273.     sws_normalizeVec(filter->lumH, 1.0);

  3274.     sws_normalizeVec(filter->lumV, 1.0);

  3275. 

  3276.     if (verbose) sws_printVec2(filter->chrH, NULL, AV_LOG_DEBUG);

  3277.     if (verbose) sws_printVec2(filter->lumH, NULL, AV_LOG_DEBUG);

  3278. 

  3279.     return filter;

  3280. }

  3281. 

  3282. SwsVector *sws_allocVec(int length)

  3283. {

  3284.     SwsVector *vec = av_malloc(sizeof(SwsVector));

  3285.     if (!vec)

  3286.         return NULL;

  3287.     vec->length = length;

  3288.     vec->coeff  = av_malloc(sizeof(double) * length);

  3289.     if (!vec->coeff)

  3290.         av_freep(&vec);

  3291.     return vec;

  3292. }

  3293. 

  3294. SwsVector *sws_getGaussianVec(double variance, double quality)

  3295. {

  3296.     const int length= (int)(variance*quality + 0.5) | 1;

  3297.     int i;

  3298.     double middle= (length-1)*0.5;

  3299.     SwsVector *vec= sws_allocVec(length);

  3300. 

  3301.     if (!vec)

  3302.         return NULL;

  3303. 

  3304.     for (i=0; i<length; i++) {

  3305.         double dist= i-middle;

  3306.         vec->coeff[i]= exp(-dist*dist/(2*variance*variance)) / sqrt(2*variance*PI);

  3307.     }

  3308. 

  3309.     sws_normalizeVec(vec, 1.0);

  3310. 

  3311.     return vec;

  3312. }

  3313. 

  3314. SwsVector *sws_getConstVec(double c, int length)

  3315. {

  3316.     int i;

  3317.     SwsVector *vec= sws_allocVec(length);

  3318. 

  3319.     if (!vec)

  3320.         return NULL;

  3321. 

  3322.     for (i=0; i<length; i++)

  3323.         vec->coeff[i]= c;

  3324. 

  3325.     return vec;

  3326. }

  3327. 

  3328. 

  3329. SwsVector *sws_getIdentityVec(void)

  3330. {

  3331.     return sws_getConstVec(1.0, 1);

  3332. }

  3333. 

  3334. double sws_dcVec(SwsVector *a)

  3335. {

  3336.     int i;

  3337.     double sum=0;

  3338. 

  3339.     for (i=0; i<a->length; i++)

  3340.         sum+= a->coeff[i];

  3341. 

  3342.     return sum;

  3343. }

  3344. 

  3345. void sws_scaleVec(SwsVector *a, double scalar)

  3346. {

  3347.     int i;

  3348. 

  3349.     for (i=0; i<a->length; i++)

  3350.         a->coeff[i]*= scalar;

  3351. }

  3352. 

  3353. void sws_normalizeVec(SwsVector *a, double height)

  3354. {

  3355.     sws_scaleVec(a, height/sws_dcVec(a));

  3356. }

  3357. 

  3358. static SwsVector *sws_getConvVec(SwsVector *a, SwsVector *b)

  3359. {

  3360.     int length= a->length + b->length - 1;

  3361.     int i, j;

  3362.     SwsVector *vec= sws_getConstVec(0.0, length);

  3363. 

  3364.     if (!vec)

  3365.         return NULL;

  3366. 

  3367.     for (i=0; i<a->length; i++) {

  3368.         for (j=0; j<b->length; j++) {

  3369.             vec->coeff[i+j]+= a->coeff[i]*b->coeff[j];

  3370.         }

  3371.     }

  3372. 

  3373.     return vec;

  3374. }

  3375. 

  3376. static SwsVector *sws_sumVec(SwsVector *a, SwsVector *b)

  3377. {

  3378.     int length= FFMAX(a->length, b->length);

  3379.     int i;

  3380.     SwsVector *vec= sws_getConstVec(0.0, length);

  3381. 

  3382.     if (!vec)

  3383.         return NULL;

  3384. 

  3385.     for (i=0; i<a->length; i++) vec->coeff[i + (length-1)/2 - (a->length-1)/2]+= a->coeff[i];

  3386.     for (i=0; i<b->length; i++) vec->coeff[i + (length-1)/2 - (b->length-1)/2]+= b->coeff[i];

  3387. 

  3388.     return vec;

  3389. }

  3390. 

  3391. static SwsVector *sws_diffVec(SwsVector *a, SwsVector *b)

  3392. {

  3393.     int length= FFMAX(a->length, b->length);

  3394.     int i;

  3395.     SwsVector *vec= sws_getConstVec(0.0, length);

  3396. 

  3397.     if (!vec)

  3398.         return NULL;

  3399. 

  3400.     for (i=0; i<a->length; i++) vec->coeff[i + (length-1)/2 - (a->length-1)/2]+= a->coeff[i];

  3401.     for (i=0; i<b->length; i++) vec->coeff[i + (length-1)/2 - (b->length-1)/2]-= b->coeff[i];

  3402. 

  3403.     return vec;

  3404. }

  3405. 

  3406. /* shift left / or right if "shift" is negative */

  3407. static SwsVector *sws_getShiftedVec(SwsVector *a, int shift)

  3408. {

  3409.     int length= a->length + FFABS(shift)*2;

  3410.     int i;

  3411.     SwsVector *vec= sws_getConstVec(0.0, length);

  3412. 

  3413.     if (!vec)

  3414.         return NULL;

  3415. 

  3416.     for (i=0; i<a->length; i++) {

  3417.         vec->coeff[i + (length-1)/2 - (a->length-1)/2 - shift]= a->coeff[i];

  3418.     }

  3419. 

  3420.     return vec;

  3421. }

  3422. 

  3423. void sws_shiftVec(SwsVector *a, int shift)

  3424. {

  3425.     SwsVector *shifted= sws_getShiftedVec(a, shift);

  3426.     av_free(a->coeff);

  3427.     a->coeff= shifted->coeff;

  3428.     a->length= shifted->length;

  3429.     av_free(shifted);

  3430. }

  3431. 

  3432. void sws_addVec(SwsVector *a, SwsVector *b)

  3433. {

  3434.     SwsVector *sum= sws_sumVec(a, b);

  3435.     av_free(a->coeff);

  3436.     a->coeff= sum->coeff;

  3437.     a->length= sum->length;

  3438.     av_free(sum);

  3439. }

  3440. 

  3441. void sws_subVec(SwsVector *a, SwsVector *b)

  3442. {

  3443.     SwsVector *diff= sws_diffVec(a, b);

  3444.     av_free(a->coeff);

  3445.     a->coeff= diff->coeff;

  3446.     a->length= diff->length;

  3447.     av_free(diff);

  3448. }

  3449. 

  3450. void sws_convVec(SwsVector *a, SwsVector *b)

  3451. {

  3452.     SwsVector *conv= sws_getConvVec(a, b);

  3453.     av_free(a->coeff);

  3454.     a->coeff= conv->coeff;

  3455.     a->length= conv->length;

  3456.     av_free(conv);

  3457. }

  3458. 

  3459. SwsVector *sws_cloneVec(SwsVector *a)

  3460. {

  3461.     int i;

  3462.     SwsVector *vec= sws_allocVec(a->length);

  3463. 

  3464.     if (!vec)

  3465.         return NULL;

  3466. 

  3467.     for (i=0; i<a->length; i++) vec->coeff[i]= a->coeff[i];

  3468. 

  3469.     return vec;

  3470. }

  3471. 

  3472. void sws_printVec2(SwsVector *a, AVClass *log_ctx, int log_level)

  3473. {

  3474.     int i;

  3475.     double max=0;

  3476.     double min=0;

  3477.     double range;

  3478. 

  3479.     for (i=0; i<a->length; i++)

  3480.         if (a->coeff[i]>max) max= a->coeff[i];

  3481. 

  3482.     for (i=0; i<a->length; i++)

  3483.         if (a->coeff[i]<min) min= a->coeff[i];

  3484. 

  3485.     range= max - min;

  3486. 

  3487.     for (i=0; i<a->length; i++) {

  3488.         int x= (int)((a->coeff[i]-min)*60.0/range +0.5);

  3489.         av_log(log_ctx, log_level, "%1.3f ", a->coeff[i]);

  3490.         for (;x>0; x--) av_log(log_ctx, log_level, " ");

  3491.         av_log(log_ctx, log_level, "|\n");

  3492.     }

  3493. }

  3494. 

  3495. #if LIBSWSCALE_VERSION_MAJOR < 1

  3496. void sws_printVec(SwsVector *a)

  3497. {

  3498.     sws_printVec2(a, NULL, AV_LOG_DEBUG);

  3499. }

  3500. #endif

  3501. 

  3502. void sws_freeVec(SwsVector *a)

  3503. {

  3504.     if (!a) return;

  3505.     av_freep(&a->coeff);

  3506.     a->length=0;

  3507.     av_free(a);

  3508. }

  3509. 

  3510. void sws_freeFilter(SwsFilter *filter)

  3511. {

  3512.     if (!filter) return;

  3513. 

  3514.     if (filter->lumH) sws_freeVec(filter->lumH);

  3515.     if (filter->lumV) sws_freeVec(filter->lumV);

  3516.     if (filter->chrH) sws_freeVec(filter->chrH);

  3517.     if (filter->chrV) sws_freeVec(filter->chrV);

  3518.     av_free(filter);

  3519. }

  3520. 

  3521. 

  3522. void sws_freeContext(SwsContext *c)

  3523. {

  3524.     int i;

  3525.     if (!c) return;

  3526. 

  3527.     if (c->lumPixBuf) {

  3528.         for (i=0; i<c->vLumBufSize; i++)

  3529.             av_freep(&c->lumPixBuf[i]);

  3530.         av_freep(&c->lumPixBuf);

  3531.     }

  3532. 

  3533.     if (c->chrPixBuf) {

  3534.         for (i=0; i<c->vChrBufSize; i++)

  3535.             av_freep(&c->chrPixBuf[i]);

  3536.         av_freep(&c->chrPixBuf);

  3537.     }

  3538. 

  3539.     if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {

  3540.         for (i=0; i<c->vLumBufSize; i++)

  3541.             av_freep(&c->alpPixBuf[i]);

  3542.         av_freep(&c->alpPixBuf);

  3543.     }

  3544. 

  3545.     av_freep(&c->vLumFilter);

  3546.     av_freep(&c->vChrFilter);

  3547.     av_freep(&c->hLumFilter);

  3548.     av_freep(&c->hChrFilter);

  3549. #ifdef COMPILE_ALTIVEC

  3550.     av_freep(&c->vYCoeffsBank);

  3551.     av_freep(&c->vCCoeffsBank);

  3552. #endif

  3553. 

  3554.     av_freep(&c->vLumFilterPos);

  3555.     av_freep(&c->vChrFilterPos);

  3556.     av_freep(&c->hLumFilterPos);

  3557.     av_freep(&c->hChrFilterPos);

  3558. 

  3559. #if ARCH_X86 && CONFIG_GPL

  3560. #ifdef MAP_ANONYMOUS

  3561.     if (c->lumMmx2FilterCode) munmap(c->lumMmx2FilterCode, c->lumMmx2FilterCodeSize);

  3562.     if (c->chrMmx2FilterCode) munmap(c->chrMmx2FilterCode, c->chrMmx2FilterCodeSize);

  3563. #elif HAVE_VIRTUALALLOC

  3564.     if (c->lumMmx2FilterCode) VirtualFree(c->lumMmx2FilterCode, c->lumMmx2FilterCodeSize, MEM_RELEASE);

  3565.     if (c->chrMmx2FilterCode) VirtualFree(c->chrMmx2FilterCode, c->chrMmx2FilterCodeSize, MEM_RELEASE);

  3566. #else

  3567.     av_free(c->lumMmx2FilterCode);

  3568.     av_free(c->chrMmx2FilterCode);

  3569. #endif

  3570.     c->lumMmx2FilterCode=NULL;

  3571.     c->chrMmx2FilterCode=NULL;

  3572. #endif /* ARCH_X86 && CONFIG_GPL */

  3573. 

  3574.     av_freep(&c->lumMmx2Filter);

  3575.     av_freep(&c->chrMmx2Filter);

  3576.     av_freep(&c->lumMmx2FilterPos);

  3577.     av_freep(&c->chrMmx2FilterPos);

  3578.     av_freep(&c->yuvTable);

  3579. 

  3580.     av_free(c);

  3581. }

  3582. 

  3583. struct SwsContext *sws_getCachedContext(struct SwsContext *context,

  3584.                                         int srcW, int srcH, enum PixelFormat srcFormat,

  3585.                                         int dstW, int dstH, enum PixelFormat dstFormat, int flags,

  3586.                                         SwsFilter *srcFilter, SwsFilter *dstFilter, const double *param)

  3587. {

  3588.     static const double default_param[2] = {SWS_PARAM_DEFAULT, SWS_PARAM_DEFAULT};

  3589. 

  3590.     if (!param)

  3591.         param = default_param;

  3592. 

  3593.     if (context) {

  3594.         if (context->srcW != srcW || context->srcH != srcH ||

  3595.             context->srcFormat != srcFormat ||

  3596.             context->dstW != dstW || context->dstH != dstH ||

  3597.             context->dstFormat != dstFormat || context->flags != flags ||

  3598.             context->param[0] != param[0] || context->param[1] != param[1])

  3599.         {

  3600.             sws_freeContext(context);

  3601.             context = NULL;

  3602.         }

  3603.     }

  3604.     if (!context) {

  3605.         return sws_getContext(srcW, srcH, srcFormat,

  3606.                               dstW, dstH, dstFormat, flags,

  3607.                               srcFilter, dstFilter, param);

  3608.     }

  3609.     return context;

  3610. }