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.
19159 Commits
32 Branches
802MB
Tree: cdf38a17fb
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from 'cdf38a17fb'
${ noResults }
FFmpeg/libswscale/swscale.c

3430 lines
116KB
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 <unistd.h>

  63. #include "config.h"

  64. #include <assert.h>

  65. #if HAVE_SYS_MMAN_H

  66. #include <sys/mman.h>

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

  68. #define MAP_ANONYMOUS MAP_ANON

  69. #endif

  70. #endif

  71. #if HAVE_VIRTUALALLOC

  72. #define WIN32_LEAN_AND_MEAN

  73. #include <windows.h>

  74. #endif

  75. #include "swscale.h"

  76. #include "swscale_internal.h"

  77. #include "rgb2rgb.h"

  78. #include "libavutil/x86_cpu.h"

  79. #include "libavutil/bswap.h"

  80. 

  81. unsigned swscale_version(void)

  82. {

  83.     return LIBSWSCALE_VERSION_INT;

  84. }

  85. 

  86. #undef MOVNTQ

  87. #undef PAVGB

  88. 

  89. //#undef HAVE_MMX2

  90. //#define HAVE_AMD3DNOW

  91. //#undef HAVE_MMX

  92. //#undef ARCH_X86

  93. //#define WORDS_BIGENDIAN

  94. #define DITHER1XBPP

  95. 

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

  97. 

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

  99. 

  100. #ifdef M_PI

  101. #define PI M_PI

  102. #else

  103. #define PI 3.14159265358979323846

  104. #endif

  105. 

  106. #define isSupportedIn(x)    (       \

  107.            (x)==PIX_FMT_YUV420P     \

  108.         || (x)==PIX_FMT_YUVA420P    \

  109.         || (x)==PIX_FMT_YUYV422     \

  110.         || (x)==PIX_FMT_UYVY422     \

  111.         || (x)==PIX_FMT_RGB32       \

  112.         || (x)==PIX_FMT_RGB32_1     \

  113.         || (x)==PIX_FMT_BGR24       \

  114.         || (x)==PIX_FMT_BGR565      \

  115.         || (x)==PIX_FMT_BGR555      \

  116.         || (x)==PIX_FMT_BGR32       \

  117.         || (x)==PIX_FMT_BGR32_1     \

  118.         || (x)==PIX_FMT_RGB24       \

  119.         || (x)==PIX_FMT_RGB565      \

  120.         || (x)==PIX_FMT_RGB555      \

  121.         || (x)==PIX_FMT_GRAY8       \

  122.         || (x)==PIX_FMT_YUV410P     \

  123.         || (x)==PIX_FMT_YUV440P     \

  124.         || (x)==PIX_FMT_GRAY16BE    \

  125.         || (x)==PIX_FMT_GRAY16LE    \

  126.         || (x)==PIX_FMT_YUV444P     \

  127.         || (x)==PIX_FMT_YUV422P     \

  128.         || (x)==PIX_FMT_YUV411P     \

  129.         || (x)==PIX_FMT_PAL8        \

  130.         || (x)==PIX_FMT_BGR8        \

  131.         || (x)==PIX_FMT_RGB8        \

  132.         || (x)==PIX_FMT_BGR4_BYTE   \

  133.         || (x)==PIX_FMT_RGB4_BYTE   \

  134.         || (x)==PIX_FMT_YUV440P     \

  135.         || (x)==PIX_FMT_MONOWHITE   \

  136.         || (x)==PIX_FMT_MONOBLACK   \

  137.     )

  138. #define isSupportedOut(x)   (       \

  139.            (x)==PIX_FMT_YUV420P     \

  140.         || (x)==PIX_FMT_YUVA420P    \

  141.         || (x)==PIX_FMT_YUYV422     \

  142.         || (x)==PIX_FMT_UYVY422     \

  143.         || (x)==PIX_FMT_YUV444P     \

  144.         || (x)==PIX_FMT_YUV422P     \

  145.         || (x)==PIX_FMT_YUV411P     \

  146.         || isRGB(x)                 \

  147.         || isBGR(x)                 \

  148.         || (x)==PIX_FMT_NV12        \

  149.         || (x)==PIX_FMT_NV21        \

  150.         || (x)==PIX_FMT_GRAY16BE    \

  151.         || (x)==PIX_FMT_GRAY16LE    \

  152.         || (x)==PIX_FMT_GRAY8       \

  153.         || (x)==PIX_FMT_YUV410P     \

  154.         || (x)==PIX_FMT_YUV440P     \

  155.     )

  156. #define isPacked(x)         (       \

  157.            (x)==PIX_FMT_PAL8        \

  158.         || (x)==PIX_FMT_YUYV422     \

  159.         || (x)==PIX_FMT_UYVY422     \

  160.         || isRGB(x)                 \

  161.         || isBGR(x)                 \

  162.     )

  163. #define usePal(x)           (       \

  164.            (x)==PIX_FMT_PAL8        \

  165.         || (x)==PIX_FMT_BGR4_BYTE   \

  166.         || (x)==PIX_FMT_RGB4_BYTE   \

  167.         || (x)==PIX_FMT_BGR8        \

  168.         || (x)==PIX_FMT_RGB8        \

  169.     )

  170. 

  171. #define RGB2YUV_SHIFT 15

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

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

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

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

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

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

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

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

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

  181. 

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

  183. 

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

  185.     {0.7152, 0.0722, 0.2126, -0.386, 0.5, -0.115, -0.454, -0.046, 0.5},

  186.     {0.7152, 0.0722, 0.2126, -0.386, 0.5, -0.115, -0.454, -0.046, 0.5},

  187.     {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5},

  188.     {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5},

  189.     {0.59  , 0.11  , 0.30  , -0.331, 0.5, -0.169, -0.421, -0.079, 0.5}, //FCC

  190.     {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5},

  191.     {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5}, //SMPTE 170M

  192.     {0.701 , 0.087 , 0.212 , -0.384, 0.5  -0.116, -0.445, -0.055, 0.5}, //SMPTE 240M

  193. };

  194. 

  195. /*

  196. NOTES

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

  198. 

  199. TODO

  200. more intelligent misalignment avoidance for the horizontal scaler

  201. write special vertical cubic upscale version

  202. optimize C code (YV12 / minmax)

  203. add support for packed pixel YUV input & output

  204. add support for Y8 output

  205. optimize BGR24 & BGR32

  206. add BGR4 output support

  207. write special BGR->BGR scaler

  208. */

  209. 

  210. #if ARCH_X86 && CONFIG_GPL

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

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

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

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

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

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

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

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

  219. 

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

  221.         0x0103010301030103LL,

  222.         0x0200020002000200LL,};

  223. 

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

  225.         0x0602060206020602LL,

  226.         0x0004000400040004LL,};

  227. 

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

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

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

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

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

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

  234. 

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

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

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

  238. 

  239. #ifdef FAST_BGR2YV12

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

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

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

  243. #else

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

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

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

  247. #endif /* FAST_BGR2YV12 */

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

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

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

  251. 

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

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

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

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

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

  257. 

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

  259.     {0x38380000DAC83838ULL, 0xECFFDAC80000ECFFULL, 0xF6E40000D0E3F6E4ULL, 0x3838D0E300003838ULL},

  260.     {0xECFF0000DAC8ECFFULL, 0x3838DAC800003838ULL, 0x38380000D0E33838ULL, 0xF6E4D0E30000F6E4ULL},

  261. };

  262. 

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

  264. 

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

  266. 

  267. // clipping helper table for C implementations:

  268. static unsigned char clip_table[768];

  269. 

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

  271. 

  272. static const uint8_t  __attribute__((aligned(8))) dither_2x2_4[2][8]={

  273. {  1,   3,   1,   3,   1,   3,   1,   3, },

  274. {  2,   0,   2,   0,   2,   0,   2,   0, },

  275. };

  276. 

  277. static const uint8_t  __attribute__((aligned(8))) dither_2x2_8[2][8]={

  278. {  6,   2,   6,   2,   6,   2,   6,   2, },

  279. {  0,   4,   0,   4,   0,   4,   0,   4, },

  280. };

  281. 

  282. const uint8_t  __attribute__((aligned(8))) dither_8x8_32[8][8]={

  283. { 17,   9,  23,  15,  16,   8,  22,  14, },

  284. {  5,  29,   3,  27,   4,  28,   2,  26, },

  285. { 21,  13,  19,  11,  20,  12,  18,  10, },

  286. {  0,  24,   6,  30,   1,  25,   7,  31, },

  287. { 16,   8,  22,  14,  17,   9,  23,  15, },

  288. {  4,  28,   2,  26,   5,  29,   3,  27, },

  289. { 20,  12,  18,  10,  21,  13,  19,  11, },

  290. {  1,  25,   7,  31,   0,  24,   6,  30, },

  291. };

  292. 

  293. #if 0

  294. const uint8_t  __attribute__((aligned(8))) dither_8x8_64[8][8]={

  295. {  0,  48,  12,  60,   3,  51,  15,  63, },

  296. { 32,  16,  44,  28,  35,  19,  47,  31, },

  297. {  8,  56,   4,  52,  11,  59,   7,  55, },

  298. { 40,  24,  36,  20,  43,  27,  39,  23, },

  299. {  2,  50,  14,  62,   1,  49,  13,  61, },

  300. { 34,  18,  46,  30,  33,  17,  45,  29, },

  301. { 10,  58,   6,  54,   9,  57,   5,  53, },

  302. { 42,  26,  38,  22,  41,  25,  37,  21, },

  303. };

  304. #endif

  305. 

  306. const uint8_t  __attribute__((aligned(8))) dither_8x8_73[8][8]={

  307. {  0,  55,  14,  68,   3,  58,  17,  72, },

  308. { 37,  18,  50,  32,  40,  22,  54,  35, },

  309. {  9,  64,   5,  59,  13,  67,   8,  63, },

  310. { 46,  27,  41,  23,  49,  31,  44,  26, },

  311. {  2,  57,  16,  71,   1,  56,  15,  70, },

  312. { 39,  21,  52,  34,  38,  19,  51,  33, },

  313. { 11,  66,   7,  62,  10,  65,   6,  60, },

  314. { 48,  30,  43,  25,  47,  29,  42,  24, },

  315. };

  316. 

  317. #if 0

  318. const uint8_t  __attribute__((aligned(8))) dither_8x8_128[8][8]={

  319. { 68,  36,  92,  60,  66,  34,  90,  58, },

  320. { 20, 116,  12, 108,  18, 114,  10, 106, },

  321. { 84,  52,  76,  44,  82,  50,  74,  42, },

  322. {  0,  96,  24, 120,   6, 102,  30, 126, },

  323. { 64,  32,  88,  56,  70,  38,  94,  62, },

  324. { 16, 112,   8, 104,  22, 118,  14, 110, },

  325. { 80,  48,  72,  40,  86,  54,  78,  46, },

  326. {  4, 100,  28, 124,   2,  98,  26, 122, },

  327. };

  328. #endif

  329. 

  330. #if 1

  331. const uint8_t  __attribute__((aligned(8))) dither_8x8_220[8][8]={

  332. {117,  62, 158, 103, 113,  58, 155, 100, },

  333. { 34, 199,  21, 186,  31, 196,  17, 182, },

  334. {144,  89, 131,  76, 141,  86, 127,  72, },

  335. {  0, 165,  41, 206,  10, 175,  52, 217, },

  336. {110,  55, 151,  96, 120,  65, 162, 107, },

  337. { 28, 193,  14, 179,  38, 203,  24, 189, },

  338. {138,  83, 124,  69, 148,  93, 134,  79, },

  339. {  7, 172,  48, 213,   3, 168,  45, 210, },

  340. };

  341. #elif 1

  342. // tries to correct a gamma of 1.5

  343. const uint8_t  __attribute__((aligned(8))) dither_8x8_220[8][8]={

  344. {  0, 143,  18, 200,   2, 156,  25, 215, },

  345. { 78,  28, 125,  64,  89,  36, 138,  74, },

  346. { 10, 180,   3, 161,  16, 195,   8, 175, },

  347. {109,  51,  93,  38, 121,  60, 105,  47, },

  348. {  1, 152,  23, 210,   0, 147,  20, 205, },

  349. { 85,  33, 134,  71,  81,  30, 130,  67, },

  350. { 14, 190,   6, 171,  12, 185,   5, 166, },

  351. {117,  57, 101,  44, 113,  54,  97,  41, },

  352. };

  353. #elif 1

  354. // tries to correct a gamma of 2.0

  355. const uint8_t  __attribute__((aligned(8))) dither_8x8_220[8][8]={

  356. {  0, 124,   8, 193,   0, 140,  12, 213, },

  357. { 55,  14, 104,  42,  66,  19, 119,  52, },

  358. {  3, 168,   1, 145,   6, 187,   3, 162, },

  359. { 86,  31,  70,  21,  99,  39,  82,  28, },

  360. {  0, 134,  11, 206,   0, 129,   9, 200, },

  361. { 62,  17, 114,  48,  58,  16, 109,  45, },

  362. {  5, 181,   2, 157,   4, 175,   1, 151, },

  363. { 95,  36,  78,  26,  90,  34,  74,  24, },

  364. };

  365. #else

  366. // tries to correct a gamma of 2.5

  367. const uint8_t  __attribute__((aligned(8))) dither_8x8_220[8][8]={

  368. {  0, 107,   3, 187,   0, 125,   6, 212, },

  369. { 39,   7,  86,  28,  49,  11, 102,  36, },

  370. {  1, 158,   0, 131,   3, 180,   1, 151, },

  371. { 68,  19,  52,  12,  81,  25,  64,  17, },

  372. {  0, 119,   5, 203,   0, 113,   4, 195, },

  373. { 45,   9,  96,  33,  42,   8,  91,  30, },

  374. {  2, 172,   1, 144,   2, 165,   0, 137, },

  375. { 77,  23,  60,  15,  72,  21,  56,  14, },

  376. };

  377. #endif

  378. 

  379. const char *sws_format_name(enum PixelFormat format)

  380. {

  381.     switch (format) {

  382.         case PIX_FMT_YUV420P:

  383.             return "yuv420p";

  384.         case PIX_FMT_YUVA420P:

  385.             return "yuva420p";

  386.         case PIX_FMT_YUYV422:

  387.             return "yuyv422";

  388.         case PIX_FMT_RGB24:

  389.             return "rgb24";

  390.         case PIX_FMT_BGR24:

  391.             return "bgr24";

  392.         case PIX_FMT_YUV422P:

  393.             return "yuv422p";

  394.         case PIX_FMT_YUV444P:

  395.             return "yuv444p";

  396.         case PIX_FMT_RGB32:

  397.             return "rgb32";

  398.         case PIX_FMT_YUV410P:

  399.             return "yuv410p";

  400.         case PIX_FMT_YUV411P:

  401.             return "yuv411p";

  402.         case PIX_FMT_RGB565:

  403.             return "rgb565";

  404.         case PIX_FMT_RGB555:

  405.             return "rgb555";

  406.         case PIX_FMT_GRAY16BE:

  407.             return "gray16be";

  408.         case PIX_FMT_GRAY16LE:

  409.             return "gray16le";

  410.         case PIX_FMT_GRAY8:

  411.             return "gray8";

  412.         case PIX_FMT_MONOWHITE:

  413.             return "mono white";

  414.         case PIX_FMT_MONOBLACK:

  415.             return "mono black";

  416.         case PIX_FMT_PAL8:

  417.             return "Palette";

  418.         case PIX_FMT_YUVJ420P:

  419.             return "yuvj420p";

  420.         case PIX_FMT_YUVJ422P:

  421.             return "yuvj422p";

  422.         case PIX_FMT_YUVJ444P:

  423.             return "yuvj444p";

  424.         case PIX_FMT_XVMC_MPEG2_MC:

  425.             return "xvmc_mpeg2_mc";

  426.         case PIX_FMT_XVMC_MPEG2_IDCT:

  427.             return "xvmc_mpeg2_idct";

  428.         case PIX_FMT_UYVY422:

  429.             return "uyvy422";

  430.         case PIX_FMT_UYYVYY411:

  431.             return "uyyvyy411";

  432.         case PIX_FMT_RGB32_1:

  433.             return "rgb32x";

  434.         case PIX_FMT_BGR32_1:

  435.             return "bgr32x";

  436.         case PIX_FMT_BGR32:

  437.             return "bgr32";

  438.         case PIX_FMT_BGR565:

  439.             return "bgr565";

  440.         case PIX_FMT_BGR555:

  441.             return "bgr555";

  442.         case PIX_FMT_BGR8:

  443.             return "bgr8";

  444.         case PIX_FMT_BGR4:

  445.             return "bgr4";

  446.         case PIX_FMT_BGR4_BYTE:

  447.             return "bgr4 byte";

  448.         case PIX_FMT_RGB8:

  449.             return "rgb8";

  450.         case PIX_FMT_RGB4:

  451.             return "rgb4";

  452.         case PIX_FMT_RGB4_BYTE:

  453.             return "rgb4 byte";

  454.         case PIX_FMT_NV12:

  455.             return "nv12";

  456.         case PIX_FMT_NV21:

  457.             return "nv21";

  458.         case PIX_FMT_YUV440P:

  459.             return "yuv440p";

  460.         case PIX_FMT_VDPAU_H264:

  461.             return "vdpau_h264";

  462.         case PIX_FMT_VDPAU_MPEG1:

  463.             return "vdpau_mpeg1";

  464.         case PIX_FMT_VDPAU_MPEG2:

  465.             return "vdpau_mpeg2";

  466.         case PIX_FMT_VDPAU_WMV3:

  467.             return "vdpau_wmv3";

  468.         case PIX_FMT_VDPAU_VC1:

  469.             return "vdpau_vc1";

  470.         default:

  471.             return "Unknown format";

  472.     }

  473. }

  474. 

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

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

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

  478. {

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

  480.     int i;

  481.     for (i=0; i<dstW; i++)

  482.     {

  483.         int val=1<<18;

  484.         int j;

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

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

  487. 

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

  489.     }

  490. 

  491.     if (uDest)

  492.         for (i=0; i<chrDstW; i++)

  493.         {

  494.             int u=1<<18;

  495.             int v=1<<18;

  496.             int j;

  497.             for (j=0; j<chrFilterSize; j++)

  498.             {

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

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

  501.             }

  502. 

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

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

  505.         }

  506. 

  507.     if (CONFIG_SWSCALE_ALPHA && aDest)

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

  509.             int val=1<<18;

  510.             int j;

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

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

  513. 

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

  515.         }

  516. 

  517. }

  518. 

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

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

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

  522. {

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

  524.     int i;

  525.     for (i=0; i<dstW; i++)

  526.     {

  527.         int val=1<<18;

  528.         int j;

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

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

  531. 

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

  533.     }

  534. 

  535.     if (!uDest)

  536.         return;

  537. 

  538.     if (dstFormat == PIX_FMT_NV12)

  539.         for (i=0; i<chrDstW; i++)

  540.         {

  541.             int u=1<<18;

  542.             int v=1<<18;

  543.             int j;

  544.             for (j=0; j<chrFilterSize; j++)

  545.             {

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

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

  548.             }

  549. 

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

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

  552.         }

  553.     else

  554.         for (i=0; i<chrDstW; i++)

  555.         {

  556.             int u=1<<18;

  557.             int v=1<<18;

  558.             int j;

  559.             for (j=0; j<chrFilterSize; j++)

  560.             {

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

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

  563.             }

  564. 

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

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

  567.         }

  568. }

  569. 

  570. #define YSCALE_YUV_2_PACKEDX_NOCLIP_C(type,alpha) \

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

  572.         int j;\

  573.         int Y1 = 1<<18;\

  574.         int Y2 = 1<<18;\

  575.         int U  = 1<<18;\

  576.         int V  = 1<<18;\

  577.         int av_unused A1, A2;\

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

  579.         const int i2= 2*i;\

  580.         \

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

  582.         {\

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

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

  585.         }\

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

  587.         {\

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

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

  590.         }\

  591.         Y1>>=19;\

  592.         Y2>>=19;\

  593.         U >>=19;\

  594.         V >>=19;\

  595.         if (alpha){\

  596.             A1 = 1<<18;\

  597.             A2 = 1<<18;\

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

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

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

  601.             }\

  602.             A1>>=19;\

  603.             A2>>=19;\

  604.         }\

  605. 

  606. #define YSCALE_YUV_2_PACKEDX_C(type,alpha) \

  607.         YSCALE_YUV_2_PACKEDX_NOCLIP_C(type,alpha)\

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

  609.         {\

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

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

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

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

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

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

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

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

  618.         }\

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

  620.             A1=av_clip_uint8(A1);\

  621.             A2=av_clip_uint8(A2);\

  622.         }

  623. 

  624. #define YSCALE_YUV_2_PACKEDX_FULL_C(rnd,alpha) \

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

  626.         int j;\

  627.         int Y = 0;\

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

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

  630.         int av_unused A;\

  631.         int R,G,B;\

  632.         \

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

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

  635.         }\

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

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

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

  639.         }\

  640.         Y >>=10;\

  641.         U >>=10;\

  642.         V >>=10;\

  643.         if (alpha){\

  644.             A = rnd;\

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

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

  647.             A >>=19;\

  648.             if (A&256)\

  649.                 A = av_clip_uint8(A);\

  650.         }\

  651. 

  652. #define YSCALE_YUV_2_RGBX_FULL_C(rnd,alpha) \

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

  654.         Y-= c->yuv2rgb_y_offset;\

  655.         Y*= c->yuv2rgb_y_coeff;\

  656.         Y+= rnd;\

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

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

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

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

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

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

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

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

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

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

  667.         }\

  668. 

  669. 

  670. #define YSCALE_YUV_2_GRAY16_C \

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

  672.         int j;\

  673.         int Y1 = 1<<18;\

  674.         int Y2 = 1<<18;\

  675.         int U  = 1<<18;\

  676.         int V  = 1<<18;\

  677.         \

  678.         const int i2= 2*i;\

  679.         \

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

  681.         {\

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

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

  684.         }\

  685.         Y1>>=11;\

  686.         Y2>>=11;\

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

  688.         {\

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

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

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

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

  693.         }

  694. 

  695. #define YSCALE_YUV_2_RGBX_C(type,alpha) \

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

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

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

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

  700. 

  701. #define YSCALE_YUV_2_PACKED2_C(type,alpha)   \

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

  703.         const int i2= 2*i;       \

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

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

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

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

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

  709.         int av_unused A1, A2;                                         \

  710.         if (alpha){\

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

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

  713.         }\

  714. 

  715. #define YSCALE_YUV_2_GRAY16_2_C   \

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

  717.         const int i2= 2*i;       \

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

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

  720. 

  721. #define YSCALE_YUV_2_RGB2_C(type,alpha) \

  722.     YSCALE_YUV_2_PACKED2_C(type,alpha)\

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

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

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

  726. 

  727. #define YSCALE_YUV_2_PACKED1_C(type,alpha) \

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

  729.         const int i2= 2*i;\

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

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

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

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

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

  735.         int av_unused A1, A2;\

  736.         if (alpha){\

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

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

  739.         }\

  740. 

  741. #define YSCALE_YUV_2_GRAY16_1_C \

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

  743.         const int i2= 2*i;\

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

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

  746. 

  747. #define YSCALE_YUV_2_RGB1_C(type,alpha) \

  748.     YSCALE_YUV_2_PACKED1_C(type,alpha)\

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

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

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

  752. 

  753. #define YSCALE_YUV_2_PACKED1B_C(type,alpha) \

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

  755.         const int i2= 2*i;\

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

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

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

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

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

  761.         int av_unused A1, A2;\

  762.         if (alpha){\

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

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

  765.         }\

  766. 

  767. #define YSCALE_YUV_2_RGB1B_C(type,alpha) \

  768.     YSCALE_YUV_2_PACKED1B_C(type,alpha)\

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

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

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

  772. 

  773. #define YSCALE_YUV_2_MONO2_C \

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

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

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

  777.         int acc;\

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

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

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

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

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

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

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

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

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

  787.         dest++;\

  788.     }\

  789. 

  790. 

  791. #define YSCALE_YUV_2_MONOX_C \

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

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

  794.     int acc=0;\

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

  796.         int j;\

  797.         int Y1=1<<18;\

  798.         int Y2=1<<18;\

  799. \

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

  801.         {\

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

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

  804.         }\

  805.         Y1>>=19;\

  806.         Y2>>=19;\

  807.         if ((Y1|Y2)&256)\

  808.         {\

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

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

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

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

  813.         }\

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

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

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

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

  818.             dest++;\

  819.         }\

  820.     }

  821. 

  822. 

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

  824.     switch(c->dstFormat)\

  825.     {\

  826.     case PIX_FMT_RGBA:\

  827.     case PIX_FMT_BGRA:\

  828.         if (CONFIG_SMALL){\

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

  830.             func(uint32_t,needAlpha)\

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

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

  833.             }\

  834.         }else{\

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

  836.                 func(uint32_t,1)\

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

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

  839.                 }\

  840.             }else{\

  841.                 func(uint32_t,0)\

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

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

  844.                 }\

  845.             }\

  846.         }\

  847.         break;\

  848.     case PIX_FMT_ARGB:\

  849.     case PIX_FMT_ABGR:\

  850.         if (CONFIG_SMALL){\

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

  852.             func(uint32_t,needAlpha)\

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

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

  855.             }\

  856.         }else{\

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

  858.                 func(uint32_t,1)\

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

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

  861.                 }\

  862.             }else{\

  863.                 func(uint32_t,0)\

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

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

  866.                 }\

  867.             }\

  868.         }                \

  869.         break;\

  870.     case PIX_FMT_RGB24:\

  871.         func(uint8_t,0)\

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

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

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

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

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

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

  878.             dest+=6;\

  879.         }\

  880.         break;\

  881.     case PIX_FMT_BGR24:\

  882.         func(uint8_t,0)\

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

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

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

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

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

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

  889.             dest+=6;\

  890.         }\

  891.         break;\

  892.     case PIX_FMT_RGB565:\

  893.     case PIX_FMT_BGR565:\

  894.         {\

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

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

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

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

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

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

  901.             func(uint16_t,0)\

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

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

  904.             }\

  905.         }\

  906.         break;\

  907.     case PIX_FMT_RGB555:\

  908.     case PIX_FMT_BGR555:\

  909.         {\

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

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

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

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

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

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

  916.             func(uint16_t,0)\

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

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

  919.             }\

  920.         }\

  921.         break;\

  922.     case PIX_FMT_RGB8:\

  923.     case PIX_FMT_BGR8:\

  924.         {\

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

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

  927.             func(uint8_t,0)\

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

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

  930.             }\

  931.         }\

  932.         break;\

  933.     case PIX_FMT_RGB4:\

  934.     case PIX_FMT_BGR4:\

  935.         {\

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

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

  938.             func(uint8_t,0)\

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

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

  941.             }\

  942.         }\

  943.         break;\

  944.     case PIX_FMT_RGB4_BYTE:\

  945.     case PIX_FMT_BGR4_BYTE:\

  946.         {\

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

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

  949.             func(uint8_t,0)\

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

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

  952.             }\

  953.         }\

  954.         break;\

  955.     case PIX_FMT_MONOBLACK:\

  956.     case PIX_FMT_MONOWHITE:\

  957.         {\

  958.             func_monoblack\

  959.         }\

  960.         break;\

  961.     case PIX_FMT_YUYV422:\

  962.         func2\

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

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

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

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

  967.         }                \

  968.         break;\

  969.     case PIX_FMT_UYVY422:\

  970.         func2\

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

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

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

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

  975.         }                \

  976.         break;\

  977.     case PIX_FMT_GRAY16BE:\

  978.         func_g16\

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

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

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

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

  983.         }                \

  984.         break;\

  985.     case PIX_FMT_GRAY16LE:\

  986.         func_g16\

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

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

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

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

  991.         }                \

  992.         break;\

  993.     }\

  994. 

  995. 

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

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

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

  999. {

  1000.     int i;

  1001.     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)

  1002. }

  1003. 

  1004. static inline void yuv2rgbXinC_full(SwsContext *c, const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,

  1005.                                     const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,

  1006.                                     const int16_t **alpSrc, uint8_t *dest, int dstW, int y)

  1007. {

  1008.     int i;

  1009.     int step= fmt_depth(c->dstFormat)/8;

  1010.     int aidx= 3;

  1011. 

  1012.     switch(c->dstFormat){

  1013.     case PIX_FMT_ARGB:

  1014.         dest++;

  1015.         aidx= 0;

  1016.     case PIX_FMT_RGB24:

  1017.         aidx--;

  1018.     case PIX_FMT_RGBA:

  1019.         if (CONFIG_SMALL){

  1020.             int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;

  1021.             YSCALE_YUV_2_RGBX_FULL_C(1<<21, needAlpha)

  1022.                 dest[aidx]= needAlpha ? A : 255;

  1023.                 dest[0]= R>>22;

  1024.                 dest[1]= G>>22;

  1025.                 dest[2]= B>>22;

  1026.                 dest+= step;

  1027.             }

  1028.         }else{

  1029.             if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf){

  1030.                 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 1)

  1031.                     dest[aidx]= A;

  1032.                     dest[0]= R>>22;

  1033.                     dest[1]= G>>22;

  1034.                     dest[2]= B>>22;

  1035.                     dest+= step;

  1036.                 }

  1037.             }else{

  1038.                 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 0)

  1039.                     dest[aidx]= 255;

  1040.                     dest[0]= R>>22;

  1041.                     dest[1]= G>>22;

  1042.                     dest[2]= B>>22;

  1043.                     dest+= step;

  1044.                 }

  1045.             }

  1046.         }

  1047.         break;

  1048.     case PIX_FMT_ABGR:

  1049.         dest++;

  1050.         aidx= 0;

  1051.     case PIX_FMT_BGR24:

  1052.         aidx--;

  1053.     case PIX_FMT_BGRA:

  1054.         if (CONFIG_SMALL){

  1055.             int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;

  1056.             YSCALE_YUV_2_RGBX_FULL_C(1<<21, needAlpha)

  1057.                 dest[aidx]= needAlpha ? A : 255;

  1058.                 dest[0]= B>>22;

  1059.                 dest[1]= G>>22;

  1060.                 dest[2]= R>>22;

  1061.                 dest+= step;

  1062.             }

  1063.         }else{

  1064.             if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf){

  1065.                 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 1)

  1066.                     dest[aidx]= A;

  1067.                     dest[0]= B>>22;

  1068.                     dest[1]= G>>22;

  1069.                     dest[2]= R>>22;

  1070.                     dest+= step;

  1071.                 }

  1072.             }else{

  1073.                 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 0)

  1074.                     dest[aidx]= 255;

  1075.                     dest[0]= B>>22;

  1076.                     dest[1]= G>>22;

  1077.                     dest[2]= R>>22;

  1078.                     dest+= step;

  1079.                 }

  1080.             }

  1081.         }

  1082.         break;

  1083.     default:

  1084.         assert(0);

  1085.     }

  1086. }

  1087. 

  1088. static void fillPlane(uint8_t* plane, int stride, int width, int height, int y, uint8_t val){

  1089.     int i;

  1090.     uint8_t *ptr = plane + stride*y;

  1091.     for (i=0; i<height; i++){

  1092.         memset(ptr, val, width);

  1093.         ptr += stride;

  1094.     }

  1095. }

  1096. 

  1097. //Note: we have C, MMX, MMX2, 3DNOW versions, there is no 3DNOW+MMX2 one

  1098. //Plain C versions

  1099. #if !HAVE_MMX || CONFIG_RUNTIME_CPUDETECT || !CONFIG_GPL

  1100. #define COMPILE_C

  1101. #endif

  1102. 

  1103. #if ARCH_PPC

  1104. #if (HAVE_ALTIVEC || CONFIG_RUNTIME_CPUDETECT) && CONFIG_GPL

  1105. #undef COMPILE_C

  1106. #define COMPILE_ALTIVEC

  1107. #endif

  1108. #endif //ARCH_PPC

  1109. 

  1110. #if ARCH_X86

  1111. 

  1112. #if ((HAVE_MMX && !HAVE_AMD3DNOW && !HAVE_MMX2) || CONFIG_RUNTIME_CPUDETECT) && CONFIG_GPL

  1113. #define COMPILE_MMX

  1114. #endif

  1115. 

  1116. #if (HAVE_MMX2 || CONFIG_RUNTIME_CPUDETECT) && CONFIG_GPL

  1117. #define COMPILE_MMX2

  1118. #endif

  1119. 

  1120. #if ((HAVE_AMD3DNOW && !HAVE_MMX2) || CONFIG_RUNTIME_CPUDETECT) && CONFIG_GPL

  1121. #define COMPILE_3DNOW

  1122. #endif

  1123. #endif //ARCH_X86

  1124. 

  1125. #undef HAVE_MMX

  1126. #undef HAVE_MMX2

  1127. #undef HAVE_AMD3DNOW

  1128. #undef HAVE_ALTIVEC

  1129. #define HAVE_MMX 0

  1130. #define HAVE_MMX2 0

  1131. #define HAVE_AMD3DNOW 0

  1132. #define HAVE_ALTIVEC 0

  1133. 

  1134. #ifdef COMPILE_C

  1135. #define RENAME(a) a ## _C

  1136. #include "swscale_template.c"

  1137. #endif

  1138. 

  1139. #ifdef COMPILE_ALTIVEC

  1140. #undef RENAME

  1141. #undef HAVE_ALTIVEC

  1142. #define HAVE_ALTIVEC 1

  1143. #define RENAME(a) a ## _altivec

  1144. #include "swscale_template.c"

  1145. #endif

  1146. 

  1147. #if ARCH_X86

  1148. 

  1149. //MMX versions

  1150. #ifdef COMPILE_MMX

  1151. #undef RENAME

  1152. #undef HAVE_MMX

  1153. #undef HAVE_MMX2

  1154. #undef HAVE_AMD3DNOW

  1155. #define HAVE_MMX 1

  1156. #define HAVE_MMX2 0

  1157. #define HAVE_AMD3DNOW 0

  1158. #define RENAME(a) a ## _MMX

  1159. #include "swscale_template.c"

  1160. #endif

  1161. 

  1162. //MMX2 versions

  1163. #ifdef COMPILE_MMX2

  1164. #undef RENAME

  1165. #undef HAVE_MMX

  1166. #undef HAVE_MMX2

  1167. #undef HAVE_AMD3DNOW

  1168. #define HAVE_MMX 1

  1169. #define HAVE_MMX2 1

  1170. #define HAVE_AMD3DNOW 0

  1171. #define RENAME(a) a ## _MMX2

  1172. #include "swscale_template.c"

  1173. #endif

  1174. 

  1175. //3DNOW versions

  1176. #ifdef COMPILE_3DNOW

  1177. #undef RENAME

  1178. #undef HAVE_MMX

  1179. #undef HAVE_MMX2

  1180. #undef HAVE_AMD3DNOW

  1181. #define HAVE_MMX 1

  1182. #define HAVE_MMX2 0

  1183. #define HAVE_AMD3DNOW 1

  1184. #define RENAME(a) a ## _3DNow

  1185. #include "swscale_template.c"

  1186. #endif

  1187. 

  1188. #endif //ARCH_X86

  1189. 

  1190. // minor note: the HAVE_xyz are messed up after this line so don't use them

  1191. 

  1192. static double getSplineCoeff(double a, double b, double c, double d, double dist)

  1193. {

  1194. //    printf("%f %f %f %f %f\n", a,b,c,d,dist);

  1195.     if (dist<=1.0)      return ((d*dist + c)*dist + b)*dist +a;

  1196.     else                return getSplineCoeff(        0.0,

  1197.                                              b+ 2.0*c + 3.0*d,

  1198.                                                     c + 3.0*d,

  1199.                                             -b- 3.0*c - 6.0*d,

  1200.                                             dist-1.0);

  1201. }

  1202. 

  1203. static inline int initFilter(int16_t **outFilter, int16_t **filterPos, int *outFilterSize, int xInc,

  1204.                              int srcW, int dstW, int filterAlign, int one, int flags,

  1205.                              SwsVector *srcFilter, SwsVector *dstFilter, double param[2])

  1206. {

  1207.     int i;

  1208.     int filterSize;

  1209.     int filter2Size;

  1210.     int minFilterSize;

  1211.     int64_t *filter=NULL;

  1212.     int64_t *filter2=NULL;

  1213.     const int64_t fone= 1LL<<54;

  1214.     int ret= -1;

  1215. #if ARCH_X86

  1216.     if (flags & SWS_CPU_CAPS_MMX)

  1217.         __asm__ volatile("emms\n\t"::: "memory"); //FIXME this should not be required but it IS (even for non-MMX versions)

  1218. #endif

  1219. 

  1220.     // NOTE: the +1 is for the MMX scaler which reads over the end

  1221.     *filterPos = av_malloc((dstW+1)*sizeof(int16_t));

  1222. 

  1223.     if (FFABS(xInc - 0x10000) <10) // unscaled

  1224.     {

  1225.         int i;

  1226.         filterSize= 1;

  1227.         filter= av_mallocz(dstW*sizeof(*filter)*filterSize);

  1228. 

  1229.         for (i=0; i<dstW; i++)

  1230.         {

  1231.             filter[i*filterSize]= fone;

  1232.             (*filterPos)[i]=i;

  1233.         }

  1234. 

  1235.     }

  1236.     else if (flags&SWS_POINT) // lame looking point sampling mode

  1237.     {

  1238.         int i;

  1239.         int xDstInSrc;

  1240.         filterSize= 1;

  1241.         filter= av_malloc(dstW*sizeof(*filter)*filterSize);

  1242. 

  1243.         xDstInSrc= xInc/2 - 0x8000;

  1244.         for (i=0; i<dstW; i++)

  1245.         {

  1246.             int xx= (xDstInSrc - ((filterSize-1)<<15) + (1<<15))>>16;

  1247. 

  1248.             (*filterPos)[i]= xx;

  1249.             filter[i]= fone;

  1250.             xDstInSrc+= xInc;

  1251.         }

  1252.     }

  1253.     else if ((xInc <= (1<<16) && (flags&SWS_AREA)) || (flags&SWS_FAST_BILINEAR)) // bilinear upscale

  1254.     {

  1255.         int i;

  1256.         int xDstInSrc;

  1257.         if      (flags&SWS_BICUBIC) filterSize= 4;

  1258.         else if (flags&SWS_X      ) filterSize= 4;

  1259.         else                        filterSize= 2; // SWS_BILINEAR / SWS_AREA

  1260.         filter= av_malloc(dstW*sizeof(*filter)*filterSize);

  1261. 

  1262.         xDstInSrc= xInc/2 - 0x8000;

  1263.         for (i=0; i<dstW; i++)

  1264.         {

  1265.             int xx= (xDstInSrc - ((filterSize-1)<<15) + (1<<15))>>16;

  1266.             int j;

  1267. 

  1268.             (*filterPos)[i]= xx;

  1269.                 //bilinear upscale / linear interpolate / area averaging

  1270.                 for (j=0; j<filterSize; j++)

  1271.                 {

  1272.                     int64_t coeff= fone - FFABS((xx<<16) - xDstInSrc)*(fone>>16);

  1273.                     if (coeff<0) coeff=0;

  1274.                     filter[i*filterSize + j]= coeff;

  1275.                     xx++;

  1276.                 }

  1277.             xDstInSrc+= xInc;

  1278.         }

  1279.     }

  1280.     else

  1281.     {

  1282.         int xDstInSrc;

  1283.         int sizeFactor;

  1284. 

  1285.         if      (flags&SWS_BICUBIC)      sizeFactor=  4;

  1286.         else if (flags&SWS_X)            sizeFactor=  8;

  1287.         else if (flags&SWS_AREA)         sizeFactor=  1; //downscale only, for upscale it is bilinear

  1288.         else if (flags&SWS_GAUSS)        sizeFactor=  8;   // infinite ;)

  1289.         else if (flags&SWS_LANCZOS)      sizeFactor= param[0] != SWS_PARAM_DEFAULT ? ceil(2*param[0]) : 6;

  1290.         else if (flags&SWS_SINC)         sizeFactor= 20; // infinite ;)

  1291.         else if (flags&SWS_SPLINE)       sizeFactor= 20;  // infinite ;)

  1292.         else if (flags&SWS_BILINEAR)     sizeFactor=  2;

  1293.         else {

  1294.             sizeFactor= 0; //GCC warning killer

  1295.             assert(0);

  1296.         }

  1297. 

  1298.         if (xInc <= 1<<16)      filterSize= 1 + sizeFactor; // upscale

  1299.         else                    filterSize= 1 + (sizeFactor*srcW + dstW - 1)/ dstW;

  1300. 

  1301.         if (filterSize > srcW-2) filterSize=srcW-2;

  1302. 

  1303.         filter= av_malloc(dstW*sizeof(*filter)*filterSize);

  1304. 

  1305.         xDstInSrc= xInc - 0x10000;

  1306.         for (i=0; i<dstW; i++)

  1307.         {

  1308.             int xx= (xDstInSrc - ((filterSize-2)<<16)) / (1<<17);

  1309.             int j;

  1310.             (*filterPos)[i]= xx;

  1311.             for (j=0; j<filterSize; j++)

  1312.             {

  1313.                 int64_t d= ((int64_t)FFABS((xx<<17) - xDstInSrc))<<13;

  1314.                 double floatd;

  1315.                 int64_t coeff;

  1316. 

  1317.                 if (xInc > 1<<16)

  1318.                     d= d*dstW/srcW;

  1319.                 floatd= d * (1.0/(1<<30));

  1320. 

  1321.                 if (flags & SWS_BICUBIC)

  1322.                 {

  1323.                     int64_t B= (param[0] != SWS_PARAM_DEFAULT ? param[0] :   0) * (1<<24);

  1324.                     int64_t C= (param[1] != SWS_PARAM_DEFAULT ? param[1] : 0.6) * (1<<24);

  1325.                     int64_t dd = ( d*d)>>30;

  1326.                     int64_t ddd= (dd*d)>>30;

  1327. 

  1328.                     if      (d < 1LL<<30)

  1329.                         coeff = (12*(1<<24)-9*B-6*C)*ddd + (-18*(1<<24)+12*B+6*C)*dd + (6*(1<<24)-2*B)*(1<<30);

  1330.                     else if (d < 1LL<<31)

  1331.                         coeff = (-B-6*C)*ddd + (6*B+30*C)*dd + (-12*B-48*C)*d + (8*B+24*C)*(1<<30);

  1332.                     else

  1333.                         coeff=0.0;

  1334.                     coeff *= fone>>(30+24);

  1335.                 }

  1336. /*                else if (flags & SWS_X)

  1337.                 {

  1338.                     double p= param ? param*0.01 : 0.3;

  1339.                     coeff = d ? sin(d*PI)/(d*PI) : 1.0;

  1340.                     coeff*= pow(2.0, - p*d*d);

  1341.                 }*/

  1342.                 else if (flags & SWS_X)

  1343.                 {

  1344.                     double A= param[0] != SWS_PARAM_DEFAULT ? param[0] : 1.0;

  1345.                     double c;

  1346. 

  1347.                     if (floatd<1.0)

  1348.                         c = cos(floatd*PI);

  1349.                     else

  1350.                         c=-1.0;

  1351.                     if (c<0.0)      c= -pow(-c, A);

  1352.                     else            c=  pow( c, A);

  1353.                     coeff= (c*0.5 + 0.5)*fone;

  1354.                 }

  1355.                 else if (flags & SWS_AREA)

  1356.                 {

  1357.                     int64_t d2= d - (1<<29);

  1358.                     if      (d2*xInc < -(1LL<<(29+16))) coeff= 1.0 * (1LL<<(30+16));

  1359.                     else if (d2*xInc <  (1LL<<(29+16))) coeff= -d2*xInc + (1LL<<(29+16));

  1360.                     else coeff=0.0;

  1361.                     coeff *= fone>>(30+16);

  1362.                 }

  1363.                 else if (flags & SWS_GAUSS)

  1364.                 {

  1365.                     double p= param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;

  1366.                     coeff = (pow(2.0, - p*floatd*floatd))*fone;

  1367.                 }

  1368.                 else if (flags & SWS_SINC)

  1369.                 {

  1370.                     coeff = (d ? sin(floatd*PI)/(floatd*PI) : 1.0)*fone;

  1371.                 }

  1372.                 else if (flags & SWS_LANCZOS)

  1373.                 {

  1374.                     double p= param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;

  1375.                     coeff = (d ? sin(floatd*PI)*sin(floatd*PI/p)/(floatd*floatd*PI*PI/p) : 1.0)*fone;

  1376.                     if (floatd>p) coeff=0;

  1377.                 }

  1378.                 else if (flags & SWS_BILINEAR)

  1379.                 {

  1380.                     coeff= (1<<30) - d;

  1381.                     if (coeff<0) coeff=0;

  1382.                     coeff *= fone >> 30;

  1383.                 }

  1384.                 else if (flags & SWS_SPLINE)

  1385.                 {

  1386.                     double p=-2.196152422706632;

  1387.                     coeff = getSplineCoeff(1.0, 0.0, p, -p-1.0, floatd) * fone;

  1388.                 }

  1389.                 else {

  1390.                     coeff= 0.0; //GCC warning killer

  1391.                     assert(0);

  1392.                 }

  1393. 

  1394.                 filter[i*filterSize + j]= coeff;

  1395.                 xx++;

  1396.             }

  1397.             xDstInSrc+= 2*xInc;

  1398.         }

  1399.     }

  1400. 

  1401.     /* apply src & dst Filter to filter -> filter2

  1402.        av_free(filter);

  1403.     */

  1404.     assert(filterSize>0);

  1405.     filter2Size= filterSize;

  1406.     if (srcFilter) filter2Size+= srcFilter->length - 1;

  1407.     if (dstFilter) filter2Size+= dstFilter->length - 1;

  1408.     assert(filter2Size>0);

  1409.     filter2= av_mallocz(filter2Size*dstW*sizeof(*filter2));

  1410. 

  1411.     for (i=0; i<dstW; i++)

  1412.     {

  1413.         int j, k;

  1414. 

  1415.         if(srcFilter){

  1416.             for (k=0; k<srcFilter->length; k++){

  1417.                 for (j=0; j<filterSize; j++)

  1418.                     filter2[i*filter2Size + k + j] += srcFilter->coeff[k]*filter[i*filterSize + j];

  1419.             }

  1420.         }else{

  1421.             for (j=0; j<filterSize; j++)

  1422.                 filter2[i*filter2Size + j]= filter[i*filterSize + j];

  1423.         }

  1424.         //FIXME dstFilter

  1425. 

  1426.         (*filterPos)[i]+= (filterSize-1)/2 - (filter2Size-1)/2;

  1427.     }

  1428.     av_freep(&filter);

  1429. 

  1430.     /* try to reduce the filter-size (step1 find size and shift left) */

  1431.     // Assume it is near normalized (*0.5 or *2.0 is OK but * 0.001 is not).

  1432.     minFilterSize= 0;

  1433.     for (i=dstW-1; i>=0; i--)

  1434.     {

  1435.         int min= filter2Size;

  1436.         int j;

  1437.         int64_t cutOff=0.0;

  1438. 

  1439.         /* get rid off near zero elements on the left by shifting left */

  1440.         for (j=0; j<filter2Size; j++)

  1441.         {

  1442.             int k;

  1443.             cutOff += FFABS(filter2[i*filter2Size]);

  1444. 

  1445.             if (cutOff > SWS_MAX_REDUCE_CUTOFF*fone) break;

  1446. 

  1447.             /* preserve monotonicity because the core can't handle the filter otherwise */

  1448.             if (i<dstW-1 && (*filterPos)[i] >= (*filterPos)[i+1]) break;

  1449. 

  1450.             // move filter coefficients left

  1451.             for (k=1; k<filter2Size; k++)

  1452.                 filter2[i*filter2Size + k - 1]= filter2[i*filter2Size + k];

  1453.             filter2[i*filter2Size + k - 1]= 0;

  1454.             (*filterPos)[i]++;

  1455.         }

  1456. 

  1457.         cutOff=0;

  1458.         /* count near zeros on the right */

  1459.         for (j=filter2Size-1; j>0; j--)

  1460.         {

  1461.             cutOff += FFABS(filter2[i*filter2Size + j]);

  1462. 

  1463.             if (cutOff > SWS_MAX_REDUCE_CUTOFF*fone) break;

  1464.             min--;

  1465.         }

  1466. 

  1467.         if (min>minFilterSize) minFilterSize= min;

  1468.     }

  1469. 

  1470.     if (flags & SWS_CPU_CAPS_ALTIVEC) {

  1471.         // we can handle the special case 4,

  1472.         // so we don't want to go to the full 8

  1473.         if (minFilterSize < 5)

  1474.             filterAlign = 4;

  1475. 

  1476.         // We really don't want to waste our time

  1477.         // doing useless computation, so fall back on

  1478.         // the scalar C code for very small filters.

  1479.         // Vectorizing is worth it only if you have a

  1480.         // decent-sized vector.

  1481.         if (minFilterSize < 3)

  1482.             filterAlign = 1;

  1483.     }

  1484. 

  1485.     if (flags & SWS_CPU_CAPS_MMX) {

  1486.         // special case for unscaled vertical filtering

  1487.         if (minFilterSize == 1 && filterAlign == 2)

  1488.             filterAlign= 1;

  1489.     }

  1490. 

  1491.     assert(minFilterSize > 0);

  1492.     filterSize= (minFilterSize +(filterAlign-1)) & (~(filterAlign-1));

  1493.     assert(filterSize > 0);

  1494.     filter= av_malloc(filterSize*dstW*sizeof(*filter));

  1495.     if (filterSize >= MAX_FILTER_SIZE*16/((flags&SWS_ACCURATE_RND) ? APCK_SIZE : 16) || !filter)

  1496.         goto error;

  1497.     *outFilterSize= filterSize;

  1498. 

  1499.     if (flags&SWS_PRINT_INFO)

  1500.         av_log(NULL, AV_LOG_VERBOSE, "SwScaler: reducing / aligning filtersize %d -> %d\n", filter2Size, filterSize);

  1501.     /* try to reduce the filter-size (step2 reduce it) */

  1502.     for (i=0; i<dstW; i++)

  1503.     {

  1504.         int j;

  1505. 

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

  1507.         {

  1508.             if (j>=filter2Size) filter[i*filterSize + j]= 0;

  1509.             else               filter[i*filterSize + j]= filter2[i*filter2Size + j];

  1510.             if((flags & SWS_BITEXACT) && j>=minFilterSize)

  1511.                 filter[i*filterSize + j]= 0;

  1512.         }

  1513.     }

  1514. 

  1515. 

  1516.     //FIXME try to align filterPos if possible

  1517. 

  1518.     //fix borders

  1519.     for (i=0; i<dstW; i++)

  1520.     {

  1521.         int j;

  1522.         if ((*filterPos)[i] < 0)

  1523.         {

  1524.             // move filter coefficients left to compensate for filterPos

  1525.             for (j=1; j<filterSize; j++)

  1526.             {

  1527.                 int left= FFMAX(j + (*filterPos)[i], 0);

  1528.                 filter[i*filterSize + left] += filter[i*filterSize + j];

  1529.                 filter[i*filterSize + j]=0;

  1530.             }

  1531.             (*filterPos)[i]= 0;

  1532.         }

  1533. 

  1534.         if ((*filterPos)[i] + filterSize > srcW)

  1535.         {

  1536.             int shift= (*filterPos)[i] + filterSize - srcW;

  1537.             // move filter coefficients right to compensate for filterPos

  1538.             for (j=filterSize-2; j>=0; j--)

  1539.             {

  1540.                 int right= FFMIN(j + shift, filterSize-1);

  1541.                 filter[i*filterSize +right] += filter[i*filterSize +j];

  1542.                 filter[i*filterSize +j]=0;

  1543.             }

  1544.             (*filterPos)[i]= srcW - filterSize;

  1545.         }

  1546.     }

  1547. 

  1548.     // Note the +1 is for the MMX scaler which reads over the end

  1549.     /* align at 16 for AltiVec (needed by hScale_altivec_real) */

  1550.     *outFilter= av_mallocz(*outFilterSize*(dstW+1)*sizeof(int16_t));

  1551. 

  1552.     /* normalize & store in outFilter */

  1553.     for (i=0; i<dstW; i++)

  1554.     {

  1555.         int j;

  1556.         int64_t error=0;

  1557.         int64_t sum=0;

  1558. 

  1559.         for (j=0; j<filterSize; j++)

  1560.         {

  1561.             sum+= filter[i*filterSize + j];

  1562.         }

  1563.         sum= (sum + one/2)/ one;

  1564.         for (j=0; j<*outFilterSize; j++)

  1565.         {

  1566.             int64_t v= filter[i*filterSize + j] + error;

  1567.             int intV= ROUNDED_DIV(v, sum);

  1568.             (*outFilter)[i*(*outFilterSize) + j]= intV;

  1569.             error= v - intV*sum;

  1570.         }

  1571.     }

  1572. 

  1573.     (*filterPos)[dstW]= (*filterPos)[dstW-1]; // the MMX scaler will read over the end

  1574.     for (i=0; i<*outFilterSize; i++)

  1575.     {

  1576.         int j= dstW*(*outFilterSize);

  1577.         (*outFilter)[j + i]= (*outFilter)[j + i - (*outFilterSize)];

  1578.     }

  1579. 

  1580.     ret=0;

  1581. error:

  1582.     av_free(filter);

  1583.     av_free(filter2);

  1584.     return ret;

  1585. }

  1586. 

  1587. #ifdef COMPILE_MMX2

  1588. static void initMMX2HScaler(int dstW, int xInc, uint8_t *funnyCode, int16_t *filter, int32_t *filterPos, int numSplits)

  1589. {

  1590.     uint8_t *fragmentA;

  1591.     x86_reg imm8OfPShufW1A;

  1592.     x86_reg imm8OfPShufW2A;

  1593.     x86_reg fragmentLengthA;

  1594.     uint8_t *fragmentB;

  1595.     x86_reg imm8OfPShufW1B;

  1596.     x86_reg imm8OfPShufW2B;

  1597.     x86_reg fragmentLengthB;

  1598.     int fragmentPos;

  1599. 

  1600.     int xpos, i;

  1601. 

  1602.     // create an optimized horizontal scaling routine

  1603. 

  1604.     //code fragment

  1605. 

  1606.     __asm__ volatile(

  1607.         "jmp                         9f                 \n\t"

  1608.     // Begin

  1609.         "0:                                             \n\t"

  1610.         "movq    (%%"REG_d", %%"REG_a"), %%mm3          \n\t"

  1611.         "movd    (%%"REG_c", %%"REG_S"), %%mm0          \n\t"

  1612.         "movd   1(%%"REG_c", %%"REG_S"), %%mm1          \n\t"

  1613.         "punpcklbw                %%mm7, %%mm1          \n\t"

  1614.         "punpcklbw                %%mm7, %%mm0          \n\t"

  1615.         "pshufw                   $0xFF, %%mm1, %%mm1   \n\t"

  1616.         "1:                                             \n\t"

  1617.         "pshufw                   $0xFF, %%mm0, %%mm0   \n\t"

  1618.         "2:                                             \n\t"

  1619.         "psubw                    %%mm1, %%mm0          \n\t"

  1620.         "movl   8(%%"REG_b", %%"REG_a"), %%esi          \n\t"

  1621.         "pmullw                   %%mm3, %%mm0          \n\t"

  1622.         "psllw                       $7, %%mm1          \n\t"

  1623.         "paddw                    %%mm1, %%mm0          \n\t"

  1624. 

  1625.         "movq                     %%mm0, (%%"REG_D", %%"REG_a") \n\t"

  1626. 

  1627.         "add                         $8, %%"REG_a"      \n\t"

  1628.     // End

  1629.         "9:                                             \n\t"

  1630. //        "int $3                                         \n\t"

  1631.         "lea                 " LOCAL_MANGLE(0b) ", %0   \n\t"

  1632.         "lea                 " LOCAL_MANGLE(1b) ", %1   \n\t"

  1633.         "lea                 " LOCAL_MANGLE(2b) ", %2   \n\t"

  1634.         "dec                         %1                 \n\t"

  1635.         "dec                         %2                 \n\t"

  1636.         "sub                         %0, %1             \n\t"

  1637.         "sub                         %0, %2             \n\t"

  1638.         "lea                 " LOCAL_MANGLE(9b) ", %3   \n\t"

  1639.         "sub                         %0, %3             \n\t"

  1640. 

  1641. 

  1642.         :"=r" (fragmentA), "=r" (imm8OfPShufW1A), "=r" (imm8OfPShufW2A),

  1643.         "=r" (fragmentLengthA)

  1644.     );

  1645. 

  1646.     __asm__ volatile(

  1647.         "jmp                         9f                 \n\t"

  1648.     // Begin

  1649.         "0:                                             \n\t"

  1650.         "movq    (%%"REG_d", %%"REG_a"), %%mm3          \n\t"

  1651.         "movd    (%%"REG_c", %%"REG_S"), %%mm0          \n\t"

  1652.         "punpcklbw                %%mm7, %%mm0          \n\t"

  1653.         "pshufw                   $0xFF, %%mm0, %%mm1   \n\t"

  1654.         "1:                                             \n\t"

  1655.         "pshufw                   $0xFF, %%mm0, %%mm0   \n\t"

  1656.         "2:                                             \n\t"

  1657.         "psubw                    %%mm1, %%mm0          \n\t"

  1658.         "movl   8(%%"REG_b", %%"REG_a"), %%esi          \n\t"

  1659.         "pmullw                   %%mm3, %%mm0          \n\t"

  1660.         "psllw                       $7, %%mm1          \n\t"

  1661.         "paddw                    %%mm1, %%mm0          \n\t"

  1662. 

  1663.         "movq                     %%mm0, (%%"REG_D", %%"REG_a") \n\t"

  1664. 

  1665.         "add                         $8, %%"REG_a"      \n\t"

  1666.     // End

  1667.         "9:                                             \n\t"

  1668. //        "int                       $3                   \n\t"

  1669.         "lea                 " LOCAL_MANGLE(0b) ", %0   \n\t"

  1670.         "lea                 " LOCAL_MANGLE(1b) ", %1   \n\t"

  1671.         "lea                 " LOCAL_MANGLE(2b) ", %2   \n\t"

  1672.         "dec                         %1                 \n\t"

  1673.         "dec                         %2                 \n\t"

  1674.         "sub                         %0, %1             \n\t"

  1675.         "sub                         %0, %2             \n\t"

  1676.         "lea                 " LOCAL_MANGLE(9b) ", %3   \n\t"

  1677.         "sub                         %0, %3             \n\t"

  1678. 

  1679. 

  1680.         :"=r" (fragmentB), "=r" (imm8OfPShufW1B), "=r" (imm8OfPShufW2B),

  1681.         "=r" (fragmentLengthB)

  1682.     );

  1683. 

  1684.     xpos= 0; //lumXInc/2 - 0x8000; // difference between pixel centers

  1685.     fragmentPos=0;

  1686. 

  1687.     for (i=0; i<dstW/numSplits; i++)

  1688.     {

  1689.         int xx=xpos>>16;

  1690. 

  1691.         if ((i&3) == 0)

  1692.         {

  1693.             int a=0;

  1694.             int b=((xpos+xInc)>>16) - xx;

  1695.             int c=((xpos+xInc*2)>>16) - xx;

  1696.             int d=((xpos+xInc*3)>>16) - xx;

  1697. 

  1698.             filter[i  ] = (( xpos         & 0xFFFF) ^ 0xFFFF)>>9;

  1699.             filter[i+1] = (((xpos+xInc  ) & 0xFFFF) ^ 0xFFFF)>>9;

  1700.             filter[i+2] = (((xpos+xInc*2) & 0xFFFF) ^ 0xFFFF)>>9;

  1701.             filter[i+3] = (((xpos+xInc*3) & 0xFFFF) ^ 0xFFFF)>>9;

  1702.             filterPos[i/2]= xx;

  1703. 

  1704.             if (d+1<4)

  1705.             {

  1706.                 int maxShift= 3-(d+1);

  1707.                 int shift=0;

  1708. 

  1709.                 memcpy(funnyCode + fragmentPos, fragmentB, fragmentLengthB);

  1710. 

  1711.                 funnyCode[fragmentPos + imm8OfPShufW1B]=

  1712.                     (a+1) | ((b+1)<<2) | ((c+1)<<4) | ((d+1)<<6);

  1713.                 funnyCode[fragmentPos + imm8OfPShufW2B]=

  1714.                     a | (b<<2) | (c<<4) | (d<<6);

  1715. 

  1716.                 if (i+3>=dstW) shift=maxShift; //avoid overread

  1717.                 else if ((filterPos[i/2]&3) <= maxShift) shift=filterPos[i/2]&3; //Align

  1718. 

  1719.                 if (shift && i>=shift)

  1720.                 {

  1721.                     funnyCode[fragmentPos + imm8OfPShufW1B]+= 0x55*shift;

  1722.                     funnyCode[fragmentPos + imm8OfPShufW2B]+= 0x55*shift;

  1723.                     filterPos[i/2]-=shift;

  1724.                 }

  1725. 

  1726.                 fragmentPos+= fragmentLengthB;

  1727.             }

  1728.             else

  1729.             {

  1730.                 int maxShift= 3-d;

  1731.                 int shift=0;

  1732. 

  1733.                 memcpy(funnyCode + fragmentPos, fragmentA, fragmentLengthA);

  1734. 

  1735.                 funnyCode[fragmentPos + imm8OfPShufW1A]=

  1736.                 funnyCode[fragmentPos + imm8OfPShufW2A]=

  1737.                     a | (b<<2) | (c<<4) | (d<<6);

  1738. 

  1739.                 if (i+4>=dstW) shift=maxShift; //avoid overread

  1740.                 else if ((filterPos[i/2]&3) <= maxShift) shift=filterPos[i/2]&3; //partial align

  1741. 

  1742.                 if (shift && i>=shift)

  1743.                 {

  1744.                     funnyCode[fragmentPos + imm8OfPShufW1A]+= 0x55*shift;

  1745.                     funnyCode[fragmentPos + imm8OfPShufW2A]+= 0x55*shift;

  1746.                     filterPos[i/2]-=shift;

  1747.                 }

  1748. 

  1749.                 fragmentPos+= fragmentLengthA;

  1750.             }

  1751. 

  1752.             funnyCode[fragmentPos]= RET;

  1753.         }

  1754.         xpos+=xInc;

  1755.     }

  1756.     filterPos[((i/2)+1)&(~1)]= xpos>>16; // needed to jump to the next part

  1757. }

  1758. #endif /* COMPILE_MMX2 */

  1759. 

  1760. static void globalInit(void){

  1761.     // generating tables:

  1762.     int i;

  1763.     for (i=0; i<768; i++){

  1764.         int c= av_clip_uint8(i-256);

  1765.         clip_table[i]=c;

  1766.     }

  1767. }

  1768. 

  1769. static SwsFunc getSwsFunc(SwsContext *c)

  1770. {

  1771.     int flags = c->flags;

  1772. 

  1773. #if CONFIG_RUNTIME_CPUDETECT && CONFIG_GPL

  1774. #if ARCH_X86

  1775.     // ordered per speed fastest first

  1776.     if (flags & SWS_CPU_CAPS_MMX2) {

  1777.         sws_init_swScale_MMX2(c);

  1778.         return swScale_MMX2;

  1779.     } else if (flags & SWS_CPU_CAPS_3DNOW) {

  1780.         sws_init_swScale_3DNow(c);

  1781.         return swScale_3DNow;

  1782.     } else if (flags & SWS_CPU_CAPS_MMX) {

  1783.         sws_init_swScale_MMX(c);

  1784.         return swScale_MMX;

  1785.     } else {

  1786.         sws_init_swScale_C(c);

  1787.         return swScale_C;

  1788.     }

  1789. 

  1790. #else

  1791. #if ARCH_PPC

  1792.     if (flags & SWS_CPU_CAPS_ALTIVEC) {

  1793.         sws_init_swScale_altivec(c);

  1794.         return swScale_altivec;

  1795.     } else {

  1796.         sws_init_swScale_C(c);

  1797.         return swScale_C;

  1798.     }

  1799. #endif

  1800.     sws_init_swScale_C(c);

  1801.     return swScale_C;

  1802. #endif /* ARCH_X86 */

  1803. #else //CONFIG_RUNTIME_CPUDETECT

  1804. #if   HAVE_MMX2

  1805.     sws_init_swScale_MMX2(c);

  1806.     return swScale_MMX2;

  1807. #elif HAVE_AMD3DNOW

  1808.     sws_init_swScale_3DNow(c);

  1809.     return swScale_3DNow;

  1810. #elif HAVE_MMX

  1811.     sws_init_swScale_MMX(c);

  1812.     return swScale_MMX;

  1813. #elif HAVE_ALTIVEC

  1814.     sws_init_swScale_altivec(c);

  1815.     return swScale_altivec;

  1816. #else

  1817.     sws_init_swScale_C(c);

  1818.     return swScale_C;

  1819. #endif

  1820. #endif //!CONFIG_RUNTIME_CPUDETECT

  1821. }

  1822. 

  1823. static int PlanarToNV12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1824.                                int srcSliceH, uint8_t* dstParam[], int dstStride[]){

  1825.     uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;

  1826.     /* Copy Y plane */

  1827.     if (dstStride[0]==srcStride[0] && srcStride[0] > 0)

  1828.         memcpy(dst, src[0], srcSliceH*dstStride[0]);

  1829.     else

  1830.     {

  1831.         int i;

  1832.         const uint8_t *srcPtr= src[0];

  1833.         uint8_t *dstPtr= dst;

  1834.         for (i=0; i<srcSliceH; i++)

  1835.         {

  1836.             memcpy(dstPtr, srcPtr, c->srcW);

  1837.             srcPtr+= srcStride[0];

  1838.             dstPtr+= dstStride[0];

  1839.         }

  1840.     }

  1841.     dst = dstParam[1] + dstStride[1]*srcSliceY/2;

  1842.     if (c->dstFormat == PIX_FMT_NV12)

  1843.         interleaveBytes(src[1], src[2], dst, c->srcW/2, srcSliceH/2, srcStride[1], srcStride[2], dstStride[0]);

  1844.     else

  1845.         interleaveBytes(src[2], src[1], dst, c->srcW/2, srcSliceH/2, srcStride[2], srcStride[1], dstStride[0]);

  1846. 

  1847.     return srcSliceH;

  1848. }

  1849. 

  1850. static int PlanarToYuy2Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1851.                                int srcSliceH, uint8_t* dstParam[], int dstStride[]){

  1852.     uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;

  1853. 

  1854.     yv12toyuy2(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);

  1855. 

  1856.     return srcSliceH;

  1857. }

  1858. 

  1859. static int PlanarToUyvyWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1860.                                int srcSliceH, uint8_t* dstParam[], int dstStride[]){

  1861.     uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;

  1862. 

  1863.     yv12touyvy(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);

  1864. 

  1865.     return srcSliceH;

  1866. }

  1867. 

  1868. static int YUV422PToYuy2Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1869.                                 int srcSliceH, uint8_t* dstParam[], int dstStride[]){

  1870.     uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;

  1871. 

  1872.     yuv422ptoyuy2(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);

  1873. 

  1874.     return srcSliceH;

  1875. }

  1876. 

  1877. static int YUV422PToUyvyWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1878.                                 int srcSliceH, uint8_t* dstParam[], int dstStride[]){

  1879.     uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;

  1880. 

  1881.     yuv422ptouyvy(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);

  1882. 

  1883.     return srcSliceH;

  1884. }

  1885. 

  1886. static int YUYV2YUV420Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1887.                                int srcSliceH, uint8_t* dstParam[], int dstStride[]){

  1888.     uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;

  1889.     uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY/2;

  1890.     uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY/2;

  1891. 

  1892.     yuyvtoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);

  1893. 

  1894.     if (dstParam[3])

  1895.         fillPlane(dstParam[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);

  1896. 

  1897.     return srcSliceH;

  1898. }

  1899. 

  1900. static int YUYV2YUV422Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1901.                                int srcSliceH, uint8_t* dstParam[], int dstStride[]){

  1902.     uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;

  1903.     uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY;

  1904.     uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY;

  1905. 

  1906.     yuyvtoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);

  1907. 

  1908.     return srcSliceH;

  1909. }

  1910. 

  1911. static int UYVY2YUV420Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1912.                                int srcSliceH, uint8_t* dstParam[], int dstStride[]){

  1913.     uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;

  1914.     uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY/2;

  1915.     uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY/2;

  1916. 

  1917.     uyvytoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);

  1918. 

  1919.     if (dstParam[3])

  1920.         fillPlane(dstParam[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);

  1921. 

  1922.     return srcSliceH;

  1923. }

  1924. 

  1925. static int UYVY2YUV422Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1926.                                int srcSliceH, uint8_t* dstParam[], int dstStride[]){

  1927.     uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;

  1928.     uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY;

  1929.     uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY;

  1930. 

  1931.     uyvytoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);

  1932. 

  1933.     return srcSliceH;

  1934. }

  1935. 

  1936. static int pal2rgbWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1937.                           int srcSliceH, uint8_t* dst[], int dstStride[]){

  1938.     const enum PixelFormat srcFormat= c->srcFormat;

  1939.     const enum PixelFormat dstFormat= c->dstFormat;

  1940.     void (*conv)(const uint8_t *src, uint8_t *dst, long num_pixels,

  1941.                  const uint8_t *palette)=NULL;

  1942.     int i;

  1943.     uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;

  1944.     uint8_t *srcPtr= src[0];

  1945. 

  1946.     if (!usePal(srcFormat))

  1947.         av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",

  1948.                sws_format_name(srcFormat), sws_format_name(dstFormat));

  1949. 

  1950.     switch(dstFormat){

  1951.     case PIX_FMT_RGB32  : conv = palette8topacked32; break;

  1952.     case PIX_FMT_BGR32  : conv = palette8topacked32; break;

  1953.     case PIX_FMT_BGR32_1: conv = palette8topacked32; break;

  1954.     case PIX_FMT_RGB32_1: conv = palette8topacked32; break;

  1955.     case PIX_FMT_RGB24  : conv = palette8topacked24; break;

  1956.     case PIX_FMT_BGR24  : conv = palette8topacked24; break;

  1957.     default: av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",

  1958.                     sws_format_name(srcFormat), sws_format_name(dstFormat)); break;

  1959.     }

  1960. 

  1961. 

  1962.     for (i=0; i<srcSliceH; i++) {

  1963.         conv(srcPtr, dstPtr, c->srcW, (uint8_t *) c->pal_rgb);

  1964.         srcPtr+= srcStride[0];

  1965.         dstPtr+= dstStride[0];

  1966.     }

  1967. 

  1968.     return srcSliceH;

  1969. }

  1970. 

  1971. /* {RGB,BGR}{15,16,24,32,32_1} -> {RGB,BGR}{15,16,24,32} */

  1972. static int rgb2rgbWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  1973.                           int srcSliceH, uint8_t* dst[], int dstStride[]){

  1974.     const enum PixelFormat srcFormat= c->srcFormat;

  1975.     const enum PixelFormat dstFormat= c->dstFormat;

  1976.     const int srcBpp= (fmt_depth(srcFormat) + 7) >> 3;

  1977.     const int dstBpp= (fmt_depth(dstFormat) + 7) >> 3;

  1978.     const int srcId= fmt_depth(srcFormat) >> 2; /* 1:0, 4:1, 8:2, 15:3, 16:4, 24:6, 32:8 */

  1979.     const int dstId= fmt_depth(dstFormat) >> 2;

  1980.     void (*conv)(const uint8_t *src, uint8_t *dst, long src_size)=NULL;

  1981. 

  1982.     /* BGR -> BGR */

  1983.     if (  (isBGR(srcFormat) && isBGR(dstFormat))

  1984.        || (isRGB(srcFormat) && isRGB(dstFormat))){

  1985.         switch(srcId | (dstId<<4)){

  1986.         case 0x34: conv= rgb16to15; break;

  1987.         case 0x36: conv= rgb24to15; break;

  1988.         case 0x38: conv= rgb32to15; break;

  1989.         case 0x43: conv= rgb15to16; break;

  1990.         case 0x46: conv= rgb24to16; break;

  1991.         case 0x48: conv= rgb32to16; break;

  1992.         case 0x63: conv= rgb15to24; break;

  1993.         case 0x64: conv= rgb16to24; break;

  1994.         case 0x68: conv= rgb32to24; break;

  1995.         case 0x83: conv= rgb15to32; break;

  1996.         case 0x84: conv= rgb16to32; break;

  1997.         case 0x86: conv= rgb24to32; break;

  1998.         default: av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",

  1999.                         sws_format_name(srcFormat), sws_format_name(dstFormat)); break;

  2000.         }

  2001.     }else if (  (isBGR(srcFormat) && isRGB(dstFormat))

  2002.              || (isRGB(srcFormat) && isBGR(dstFormat))){

  2003.         switch(srcId | (dstId<<4)){

  2004.         case 0x33: conv= rgb15tobgr15; break;

  2005.         case 0x34: conv= rgb16tobgr15; break;

  2006.         case 0x36: conv= rgb24tobgr15; break;

  2007.         case 0x38: conv= rgb32tobgr15; break;

  2008.         case 0x43: conv= rgb15tobgr16; break;

  2009.         case 0x44: conv= rgb16tobgr16; break;

  2010.         case 0x46: conv= rgb24tobgr16; break;

  2011.         case 0x48: conv= rgb32tobgr16; break;

  2012.         case 0x63: conv= rgb15tobgr24; break;

  2013.         case 0x64: conv= rgb16tobgr24; break;

  2014.         case 0x66: conv= rgb24tobgr24; break;

  2015.         case 0x68: conv= rgb32tobgr24; break;

  2016.         case 0x83: conv= rgb15tobgr32; break;

  2017.         case 0x84: conv= rgb16tobgr32; break;

  2018.         case 0x86: conv= rgb24tobgr32; break;

  2019.         case 0x88: conv= rgb32tobgr32; break;

  2020.         default: av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",

  2021.                         sws_format_name(srcFormat), sws_format_name(dstFormat)); break;

  2022.         }

  2023.     }else{

  2024.         av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",

  2025.                sws_format_name(srcFormat), sws_format_name(dstFormat));

  2026.     }

  2027. 

  2028.     if(conv)

  2029.     {

  2030.         uint8_t *srcPtr= src[0];

  2031.         if(srcFormat == PIX_FMT_RGB32_1 || srcFormat == PIX_FMT_BGR32_1)

  2032.             srcPtr += ALT32_CORR;

  2033. 

  2034.         if (dstStride[0]*srcBpp == srcStride[0]*dstBpp && srcStride[0] > 0)

  2035.             conv(srcPtr, dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]);

  2036.         else

  2037.         {

  2038.             int i;

  2039.             uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;

  2040. 

  2041.             for (i=0; i<srcSliceH; i++)

  2042.             {

  2043.                 conv(srcPtr, dstPtr, c->srcW*srcBpp);

  2044.                 srcPtr+= srcStride[0];

  2045.                 dstPtr+= dstStride[0];

  2046.             }

  2047.         }

  2048.     }

  2049.     return srcSliceH;

  2050. }

  2051. 

  2052. static int bgr24toyv12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  2053.                               int srcSliceH, uint8_t* dst[], int dstStride[]){

  2054. 

  2055.     rgb24toyv12(

  2056.         src[0],

  2057.         dst[0]+ srcSliceY    *dstStride[0],

  2058.         dst[1]+(srcSliceY>>1)*dstStride[1],

  2059.         dst[2]+(srcSliceY>>1)*dstStride[2],

  2060.         c->srcW, srcSliceH,

  2061.         dstStride[0], dstStride[1], srcStride[0]);

  2062.     if (dst[3])

  2063.         fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);

  2064.     return srcSliceH;

  2065. }

  2066. 

  2067. static int yvu9toyv12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  2068.                              int srcSliceH, uint8_t* dst[], int dstStride[]){

  2069.     int i;

  2070. 

  2071.     /* copy Y */

  2072.     if (srcStride[0]==dstStride[0] && srcStride[0] > 0)

  2073.         memcpy(dst[0]+ srcSliceY*dstStride[0], src[0], srcStride[0]*srcSliceH);

  2074.     else{

  2075.         uint8_t *srcPtr= src[0];

  2076.         uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;

  2077. 

  2078.         for (i=0; i<srcSliceH; i++)

  2079.         {

  2080.             memcpy(dstPtr, srcPtr, c->srcW);

  2081.             srcPtr+= srcStride[0];

  2082.             dstPtr+= dstStride[0];

  2083.         }

  2084.     }

  2085. 

  2086.     if (c->dstFormat==PIX_FMT_YUV420P || c->dstFormat==PIX_FMT_YUVA420P){

  2087.         planar2x(src[1], dst[1], c->chrSrcW, c->chrSrcH, srcStride[1], dstStride[1]);

  2088.         planar2x(src[2], dst[2], c->chrSrcW, c->chrSrcH, srcStride[2], dstStride[2]);

  2089.     }else{

  2090.         planar2x(src[1], dst[2], c->chrSrcW, c->chrSrcH, srcStride[1], dstStride[2]);

  2091.         planar2x(src[2], dst[1], c->chrSrcW, c->chrSrcH, srcStride[2], dstStride[1]);

  2092.     }

  2093.     if (dst[3])

  2094.         fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);

  2095.     return srcSliceH;

  2096. }

  2097. 

  2098. /* unscaled copy like stuff (assumes nearly identical formats) */

  2099. static int packedCopy(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  2100.                       int srcSliceH, uint8_t* dst[], int dstStride[])

  2101. {

  2102.     if (dstStride[0]==srcStride[0] && srcStride[0] > 0)

  2103.         memcpy(dst[0] + dstStride[0]*srcSliceY, src[0], srcSliceH*dstStride[0]);

  2104.     else

  2105.     {

  2106.         int i;

  2107.         uint8_t *srcPtr= src[0];

  2108.         uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;

  2109.         int length=0;

  2110. 

  2111.         /* universal length finder */

  2112.         while(length+c->srcW <= FFABS(dstStride[0])

  2113.            && length+c->srcW <= FFABS(srcStride[0])) length+= c->srcW;

  2114.         assert(length!=0);

  2115. 

  2116.         for (i=0; i<srcSliceH; i++)

  2117.         {

  2118.             memcpy(dstPtr, srcPtr, length);

  2119.             srcPtr+= srcStride[0];

  2120.             dstPtr+= dstStride[0];

  2121.         }

  2122.     }

  2123.     return srcSliceH;

  2124. }

  2125. 

  2126. static int planarCopy(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  2127.                       int srcSliceH, uint8_t* dst[], int dstStride[])

  2128. {

  2129.     int plane;

  2130.     for (plane=0; plane<4; plane++)

  2131.     {

  2132.         int length= (plane==0 || plane==3) ? c->srcW  : -((-c->srcW  )>>c->chrDstHSubSample);

  2133.         int y=      (plane==0 || plane==3) ? srcSliceY: -((-srcSliceY)>>c->chrDstVSubSample);

  2134.         int height= (plane==0 || plane==3) ? srcSliceH: -((-srcSliceH)>>c->chrDstVSubSample);

  2135. 

  2136.         if (!dst[plane]) continue;

  2137.         // ignore palette for GRAY8

  2138.         if (plane == 1 && !dst[2]) continue;

  2139.         if (!src[plane] || (plane == 1 && !src[2]))

  2140.             fillPlane(dst[plane], dstStride[plane], length, height, y, (plane==3) ? 255 : 128);

  2141.         else

  2142.         {

  2143.             if (dstStride[plane]==srcStride[plane] && srcStride[plane] > 0)

  2144.                 memcpy(dst[plane] + dstStride[plane]*y, src[plane], height*dstStride[plane]);

  2145.             else

  2146.             {

  2147.                 int i;

  2148.                 uint8_t *srcPtr= src[plane];

  2149.                 uint8_t *dstPtr= dst[plane] + dstStride[plane]*y;

  2150.                 for (i=0; i<height; i++)

  2151.                 {

  2152.                     memcpy(dstPtr, srcPtr, length);

  2153.                     srcPtr+= srcStride[plane];

  2154.                     dstPtr+= dstStride[plane];

  2155.                 }

  2156.             }

  2157.         }

  2158.     }

  2159.     return srcSliceH;

  2160. }

  2161. 

  2162. static int gray16togray(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  2163.                         int srcSliceH, uint8_t* dst[], int dstStride[]){

  2164. 

  2165.     int length= c->srcW;

  2166.     int y=      srcSliceY;

  2167.     int height= srcSliceH;

  2168.     int i, j;

  2169.     uint8_t *srcPtr= src[0];

  2170.     uint8_t *dstPtr= dst[0] + dstStride[0]*y;

  2171. 

  2172.     if (!isGray(c->dstFormat)){

  2173.         int height= -((-srcSliceH)>>c->chrDstVSubSample);

  2174.         memset(dst[1], 128, dstStride[1]*height);

  2175.         memset(dst[2], 128, dstStride[2]*height);

  2176.     }

  2177.     if (c->srcFormat == PIX_FMT_GRAY16LE) srcPtr++;

  2178.     for (i=0; i<height; i++)

  2179.     {

  2180.         for (j=0; j<length; j++) dstPtr[j] = srcPtr[j<<1];

  2181.         srcPtr+= srcStride[0];

  2182.         dstPtr+= dstStride[0];

  2183.     }

  2184.     if (dst[3])

  2185.         fillPlane(dst[3], dstStride[3], length, height, y, 255);

  2186.     return srcSliceH;

  2187. }

  2188. 

  2189. static int graytogray16(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  2190.                         int srcSliceH, uint8_t* dst[], int dstStride[]){

  2191. 

  2192.     int length= c->srcW;

  2193.     int y=      srcSliceY;

  2194.     int height= srcSliceH;

  2195.     int i, j;

  2196.     uint8_t *srcPtr= src[0];

  2197.     uint8_t *dstPtr= dst[0] + dstStride[0]*y;

  2198.     for (i=0; i<height; i++)

  2199.     {

  2200.         for (j=0; j<length; j++)

  2201.         {

  2202.             dstPtr[j<<1] = srcPtr[j];

  2203.             dstPtr[(j<<1)+1] = srcPtr[j];

  2204.         }

  2205.         srcPtr+= srcStride[0];

  2206.         dstPtr+= dstStride[0];

  2207.     }

  2208.     return srcSliceH;

  2209. }

  2210. 

  2211. static int gray16swap(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  2212.                       int srcSliceH, uint8_t* dst[], int dstStride[]){

  2213. 

  2214.     int length= c->srcW;

  2215.     int y=      srcSliceY;

  2216.     int height= srcSliceH;

  2217.     int i, j;

  2218.     uint16_t *srcPtr= (uint16_t*)src[0];

  2219.     uint16_t *dstPtr= (uint16_t*)(dst[0] + dstStride[0]*y/2);

  2220.     for (i=0; i<height; i++)

  2221.     {

  2222.         for (j=0; j<length; j++) dstPtr[j] = bswap_16(srcPtr[j]);

  2223.         srcPtr+= srcStride[0]/2;

  2224.         dstPtr+= dstStride[0]/2;

  2225.     }

  2226.     return srcSliceH;

  2227. }

  2228. 

  2229. 

  2230. static void getSubSampleFactors(int *h, int *v, int format){

  2231.     switch(format){

  2232.     case PIX_FMT_UYVY422:

  2233.     case PIX_FMT_YUYV422:

  2234.         *h=1;

  2235.         *v=0;

  2236.         break;

  2237.     case PIX_FMT_YUV420P:

  2238.     case PIX_FMT_YUVA420P:

  2239.     case PIX_FMT_GRAY16BE:

  2240.     case PIX_FMT_GRAY16LE:

  2241.     case PIX_FMT_GRAY8: //FIXME remove after different subsamplings are fully implemented

  2242.     case PIX_FMT_NV12:

  2243.     case PIX_FMT_NV21:

  2244.         *h=1;

  2245.         *v=1;

  2246.         break;

  2247.     case PIX_FMT_YUV440P:

  2248.         *h=0;

  2249.         *v=1;

  2250.         break;

  2251.     case PIX_FMT_YUV410P:

  2252.         *h=2;

  2253.         *v=2;

  2254.         break;

  2255.     case PIX_FMT_YUV444P:

  2256.         *h=0;

  2257.         *v=0;

  2258.         break;

  2259.     case PIX_FMT_YUV422P:

  2260.         *h=1;

  2261.         *v=0;

  2262.         break;

  2263.     case PIX_FMT_YUV411P:

  2264.         *h=2;

  2265.         *v=0;

  2266.         break;

  2267.     default:

  2268.         *h=0;

  2269.         *v=0;

  2270.         break;

  2271.     }

  2272. }

  2273. 

  2274. static uint16_t roundToInt16(int64_t f){

  2275.     int r= (f + (1<<15))>>16;

  2276.          if (r<-0x7FFF) return 0x8000;

  2277.     else if (r> 0x7FFF) return 0x7FFF;

  2278.     else                return r;

  2279. }

  2280. 

  2281. /**

  2282.  * @param inv_table the yuv2rgb coefficients, normally ff_yuv2rgb_coeffs[x]

  2283.  * @param fullRange if 1 then the luma range is 0..255 if 0 it is 16..235

  2284.  * @return -1 if not supported

  2285.  */

  2286. int sws_setColorspaceDetails(SwsContext *c, const int inv_table[4], int srcRange, const int table[4], int dstRange, int brightness, int contrast, int saturation){

  2287.     int64_t crv =  inv_table[0];

  2288.     int64_t cbu =  inv_table[1];

  2289.     int64_t cgu = -inv_table[2];

  2290.     int64_t cgv = -inv_table[3];

  2291.     int64_t cy  = 1<<16;

  2292.     int64_t oy  = 0;

  2293. 

  2294.     memcpy(c->srcColorspaceTable, inv_table, sizeof(int)*4);

  2295.     memcpy(c->dstColorspaceTable,     table, sizeof(int)*4);

  2296. 

  2297.     c->brightness= brightness;

  2298.     c->contrast  = contrast;

  2299.     c->saturation= saturation;

  2300.     c->srcRange  = srcRange;

  2301.     c->dstRange  = dstRange;

  2302.     if (isYUV(c->dstFormat) || isGray(c->dstFormat)) return 0;

  2303. 

  2304.     c->uOffset=   0x0400040004000400LL;

  2305.     c->vOffset=   0x0400040004000400LL;

  2306. 

  2307.     if (!srcRange){

  2308.         cy= (cy*255) / 219;

  2309.         oy= 16<<16;

  2310.     }else{

  2311.         crv= (crv*224) / 255;

  2312.         cbu= (cbu*224) / 255;

  2313.         cgu= (cgu*224) / 255;

  2314.         cgv= (cgv*224) / 255;

  2315.     }

  2316. 

  2317.     cy = (cy *contrast             )>>16;

  2318.     crv= (crv*contrast * saturation)>>32;

  2319.     cbu= (cbu*contrast * saturation)>>32;

  2320.     cgu= (cgu*contrast * saturation)>>32;

  2321.     cgv= (cgv*contrast * saturation)>>32;

  2322. 

  2323.     oy -= 256*brightness;

  2324. 

  2325.     c->yCoeff=    roundToInt16(cy *8192) * 0x0001000100010001ULL;

  2326.     c->vrCoeff=   roundToInt16(crv*8192) * 0x0001000100010001ULL;

  2327.     c->ubCoeff=   roundToInt16(cbu*8192) * 0x0001000100010001ULL;

  2328.     c->vgCoeff=   roundToInt16(cgv*8192) * 0x0001000100010001ULL;

  2329.     c->ugCoeff=   roundToInt16(cgu*8192) * 0x0001000100010001ULL;

  2330.     c->yOffset=   roundToInt16(oy *   8) * 0x0001000100010001ULL;

  2331. 

  2332.     c->yuv2rgb_y_coeff  = (int16_t)roundToInt16(cy <<13);

  2333.     c->yuv2rgb_y_offset = (int16_t)roundToInt16(oy << 9);

  2334.     c->yuv2rgb_v2r_coeff= (int16_t)roundToInt16(crv<<13);

  2335.     c->yuv2rgb_v2g_coeff= (int16_t)roundToInt16(cgv<<13);

  2336.     c->yuv2rgb_u2g_coeff= (int16_t)roundToInt16(cgu<<13);

  2337.     c->yuv2rgb_u2b_coeff= (int16_t)roundToInt16(cbu<<13);

  2338. 

  2339.     ff_yuv2rgb_c_init_tables(c, inv_table, srcRange, brightness, contrast, saturation);

  2340.     //FIXME factorize

  2341. 

  2342. #ifdef COMPILE_ALTIVEC

  2343.     if (c->flags & SWS_CPU_CAPS_ALTIVEC)

  2344.         ff_yuv2rgb_init_tables_altivec(c, inv_table, brightness, contrast, saturation);

  2345. #endif

  2346.     return 0;

  2347. }

  2348. 

  2349. /**

  2350.  * @return -1 if not supported

  2351.  */

  2352. int sws_getColorspaceDetails(SwsContext *c, int **inv_table, int *srcRange, int **table, int *dstRange, int *brightness, int *contrast, int *saturation){

  2353.     if (isYUV(c->dstFormat) || isGray(c->dstFormat)) return -1;

  2354. 

  2355.     *inv_table = c->srcColorspaceTable;

  2356.     *table     = c->dstColorspaceTable;

  2357.     *srcRange  = c->srcRange;

  2358.     *dstRange  = c->dstRange;

  2359.     *brightness= c->brightness;

  2360.     *contrast  = c->contrast;

  2361.     *saturation= c->saturation;

  2362. 

  2363.     return 0;

  2364. }

  2365. 

  2366. static int handle_jpeg(enum PixelFormat *format)

  2367. {

  2368.     switch (*format) {

  2369.         case PIX_FMT_YUVJ420P:

  2370.             *format = PIX_FMT_YUV420P;

  2371.             return 1;

  2372.         case PIX_FMT_YUVJ422P:

  2373.             *format = PIX_FMT_YUV422P;

  2374.             return 1;

  2375.         case PIX_FMT_YUVJ444P:

  2376.             *format = PIX_FMT_YUV444P;

  2377.             return 1;

  2378.         case PIX_FMT_YUVJ440P:

  2379.             *format = PIX_FMT_YUV440P;

  2380.             return 1;

  2381.         default:

  2382.             return 0;

  2383.     }

  2384. }

  2385. 

  2386. SwsContext *sws_getContext(int srcW, int srcH, enum PixelFormat srcFormat, int dstW, int dstH, enum PixelFormat dstFormat, int flags,

  2387.                            SwsFilter *srcFilter, SwsFilter *dstFilter, double *param){

  2388. 

  2389.     SwsContext *c;

  2390.     int i;

  2391.     int usesVFilter, usesHFilter;

  2392.     int unscaled, needsDither;

  2393.     int srcRange, dstRange;

  2394.     SwsFilter dummyFilter= {NULL, NULL, NULL, NULL};

  2395. #if ARCH_X86

  2396.     if (flags & SWS_CPU_CAPS_MMX)

  2397.         __asm__ volatile("emms\n\t"::: "memory");

  2398. #endif

  2399. 

  2400. #if !CONFIG_RUNTIME_CPUDETECT || !CONFIG_GPL //ensure that the flags match the compiled variant if cpudetect is off

  2401.     flags &= ~(SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_MMX2|SWS_CPU_CAPS_3DNOW|SWS_CPU_CAPS_ALTIVEC|SWS_CPU_CAPS_BFIN);

  2402. #if   HAVE_MMX2

  2403.     flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_MMX2;

  2404. #elif HAVE_AMD3DNOW

  2405.     flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_3DNOW;

  2406. #elif HAVE_MMX

  2407.     flags |= SWS_CPU_CAPS_MMX;

  2408. #elif HAVE_ALTIVEC

  2409.     flags |= SWS_CPU_CAPS_ALTIVEC;

  2410. #elif ARCH_BFIN

  2411.     flags |= SWS_CPU_CAPS_BFIN;

  2412. #endif

  2413. #endif /* CONFIG_RUNTIME_CPUDETECT */

  2414.     if (clip_table[512] != 255) globalInit();

  2415.     if (!rgb15to16) sws_rgb2rgb_init(flags);

  2416. 

  2417.     unscaled = (srcW == dstW && srcH == dstH);

  2418.     needsDither= (isBGR(dstFormat) || isRGB(dstFormat))

  2419.         && (fmt_depth(dstFormat))<24

  2420.         && ((fmt_depth(dstFormat))<(fmt_depth(srcFormat)) || (!(isRGB(srcFormat) || isBGR(srcFormat))));

  2421. 

  2422.     srcRange = handle_jpeg(&srcFormat);

  2423.     dstRange = handle_jpeg(&dstFormat);

  2424. 

  2425.     if (!isSupportedIn(srcFormat))

  2426.     {

  2427.         av_log(NULL, AV_LOG_ERROR, "swScaler: %s is not supported as input pixel format\n", sws_format_name(srcFormat));

  2428.         return NULL;

  2429.     }

  2430.     if (!isSupportedOut(dstFormat))

  2431.     {

  2432.         av_log(NULL, AV_LOG_ERROR, "swScaler: %s is not supported as output pixel format\n", sws_format_name(dstFormat));

  2433.         return NULL;

  2434.     }

  2435. 

  2436.     i= flags & ( SWS_POINT

  2437.                 |SWS_AREA

  2438.                 |SWS_BILINEAR

  2439.                 |SWS_FAST_BILINEAR

  2440.                 |SWS_BICUBIC

  2441.                 |SWS_X

  2442.                 |SWS_GAUSS

  2443.                 |SWS_LANCZOS

  2444.                 |SWS_SINC

  2445.                 |SWS_SPLINE

  2446.                 |SWS_BICUBLIN);

  2447.     if(!i || (i & (i-1)))

  2448.     {

  2449.         av_log(NULL, AV_LOG_ERROR, "swScaler: Exactly one scaler algorithm must be chosen\n");

  2450.         return NULL;

  2451.     }

  2452. 

  2453.     /* sanity check */

  2454.     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

  2455.     {

  2456.         av_log(NULL, AV_LOG_ERROR, "swScaler: %dx%d -> %dx%d is invalid scaling dimension\n",

  2457.                srcW, srcH, dstW, dstH);

  2458.         return NULL;

  2459.     }

  2460.     if(srcW > VOFW || dstW > VOFW){

  2461.         av_log(NULL, AV_LOG_ERROR, "swScaler: Compile-time maximum width is "AV_STRINGIFY(VOFW)" change VOF/VOFW and recompile\n");

  2462.         return NULL;

  2463.     }

  2464. 

  2465.     if (!dstFilter) dstFilter= &dummyFilter;

  2466.     if (!srcFilter) srcFilter= &dummyFilter;

  2467. 

  2468.     c= av_mallocz(sizeof(SwsContext));

  2469. 

  2470.     c->av_class = &sws_context_class;

  2471.     c->srcW= srcW;

  2472.     c->srcH= srcH;

  2473.     c->dstW= dstW;

  2474.     c->dstH= dstH;

  2475.     c->lumXInc= ((srcW<<16) + (dstW>>1))/dstW;

  2476.     c->lumYInc= ((srcH<<16) + (dstH>>1))/dstH;

  2477.     c->flags= flags;

  2478.     c->dstFormat= dstFormat;

  2479.     c->srcFormat= srcFormat;

  2480.     c->vRounder= 4* 0x0001000100010001ULL;

  2481. 

  2482.     usesHFilter= usesVFilter= 0;

  2483.     if (dstFilter->lumV && dstFilter->lumV->length>1) usesVFilter=1;

  2484.     if (dstFilter->lumH && dstFilter->lumH->length>1) usesHFilter=1;

  2485.     if (dstFilter->chrV && dstFilter->chrV->length>1) usesVFilter=1;

  2486.     if (dstFilter->chrH && dstFilter->chrH->length>1) usesHFilter=1;

  2487.     if (srcFilter->lumV && srcFilter->lumV->length>1) usesVFilter=1;

  2488.     if (srcFilter->lumH && srcFilter->lumH->length>1) usesHFilter=1;

  2489.     if (srcFilter->chrV && srcFilter->chrV->length>1) usesVFilter=1;

  2490.     if (srcFilter->chrH && srcFilter->chrH->length>1) usesHFilter=1;

  2491. 

  2492.     getSubSampleFactors(&c->chrSrcHSubSample, &c->chrSrcVSubSample, srcFormat);

  2493.     getSubSampleFactors(&c->chrDstHSubSample, &c->chrDstVSubSample, dstFormat);

  2494. 

  2495.     // reuse chroma for 2 pixels RGB/BGR unless user wants full chroma interpolation

  2496.     if ((isBGR(dstFormat) || isRGB(dstFormat)) && !(flags&SWS_FULL_CHR_H_INT)) c->chrDstHSubSample=1;

  2497. 

  2498.     // drop some chroma lines if the user wants it

  2499.     c->vChrDrop= (flags&SWS_SRC_V_CHR_DROP_MASK)>>SWS_SRC_V_CHR_DROP_SHIFT;

  2500.     c->chrSrcVSubSample+= c->vChrDrop;

  2501. 

  2502.     // drop every other pixel for chroma calculation unless user wants full chroma

  2503.     if ((isBGR(srcFormat) || isRGB(srcFormat)) && !(flags&SWS_FULL_CHR_H_INP)

  2504.       && srcFormat!=PIX_FMT_RGB8      && srcFormat!=PIX_FMT_BGR8

  2505.       && srcFormat!=PIX_FMT_RGB4      && srcFormat!=PIX_FMT_BGR4

  2506.       && srcFormat!=PIX_FMT_RGB4_BYTE && srcFormat!=PIX_FMT_BGR4_BYTE

  2507.       && ((dstW>>c->chrDstHSubSample) <= (srcW>>1) || (flags&(SWS_FAST_BILINEAR|SWS_POINT))))

  2508.         c->chrSrcHSubSample=1;

  2509. 

  2510.     if (param){

  2511.         c->param[0] = param[0];

  2512.         c->param[1] = param[1];

  2513.     }else{

  2514.         c->param[0] =

  2515.         c->param[1] = SWS_PARAM_DEFAULT;

  2516.     }

  2517. 

  2518.     c->chrIntHSubSample= c->chrDstHSubSample;

  2519.     c->chrIntVSubSample= c->chrSrcVSubSample;

  2520. 

  2521.     // Note the -((-x)>>y) is so that we always round toward +inf.

  2522.     c->chrSrcW= -((-srcW) >> c->chrSrcHSubSample);

  2523.     c->chrSrcH= -((-srcH) >> c->chrSrcVSubSample);

  2524.     c->chrDstW= -((-dstW) >> c->chrDstHSubSample);

  2525.     c->chrDstH= -((-dstH) >> c->chrDstVSubSample);

  2526. 

  2527.     sws_setColorspaceDetails(c, ff_yuv2rgb_coeffs[SWS_CS_DEFAULT], srcRange, ff_yuv2rgb_coeffs[SWS_CS_DEFAULT] /* FIXME*/, dstRange, 0, 1<<16, 1<<16);

  2528. 

  2529.     /* unscaled special cases */

  2530.     if (unscaled && !usesHFilter && !usesVFilter && (srcRange == dstRange || isBGR(dstFormat) || isRGB(dstFormat)))

  2531.     {

  2532.         /* yv12_to_nv12 */

  2533.         if ((srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUVA420P) && (dstFormat == PIX_FMT_NV12 || dstFormat == PIX_FMT_NV21))

  2534.         {

  2535.             c->swScale= PlanarToNV12Wrapper;

  2536.         }

  2537.         /* yuv2bgr */

  2538.         if ((srcFormat==PIX_FMT_YUV420P || srcFormat==PIX_FMT_YUV422P || srcFormat==PIX_FMT_YUVA420P) && (isBGR(dstFormat) || isRGB(dstFormat))

  2539.             && !(flags & SWS_ACCURATE_RND) && !(dstH&1))

  2540.         {

  2541.             c->swScale= ff_yuv2rgb_get_func_ptr(c);

  2542.         }

  2543. 

  2544.         if (srcFormat==PIX_FMT_YUV410P && (dstFormat==PIX_FMT_YUV420P || dstFormat==PIX_FMT_YUVA420P) && !(flags & SWS_BITEXACT))

  2545.         {

  2546.             c->swScale= yvu9toyv12Wrapper;

  2547.         }

  2548. 

  2549.         /* bgr24toYV12 */

  2550.         if (srcFormat==PIX_FMT_BGR24 && (dstFormat==PIX_FMT_YUV420P || dstFormat==PIX_FMT_YUVA420P) && !(flags & SWS_ACCURATE_RND))

  2551.             c->swScale= bgr24toyv12Wrapper;

  2552. 

  2553.         /* RGB/BGR -> RGB/BGR (no dither needed forms) */

  2554.         if (  (isBGR(srcFormat) || isRGB(srcFormat))

  2555.            && (isBGR(dstFormat) || isRGB(dstFormat))

  2556.            && srcFormat != PIX_FMT_BGR8      && dstFormat != PIX_FMT_BGR8

  2557.            && srcFormat != PIX_FMT_RGB8      && dstFormat != PIX_FMT_RGB8

  2558.            && srcFormat != PIX_FMT_BGR4      && dstFormat != PIX_FMT_BGR4

  2559.            && srcFormat != PIX_FMT_RGB4      && dstFormat != PIX_FMT_RGB4

  2560.            && srcFormat != PIX_FMT_BGR4_BYTE && dstFormat != PIX_FMT_BGR4_BYTE

  2561.            && srcFormat != PIX_FMT_RGB4_BYTE && dstFormat != PIX_FMT_RGB4_BYTE

  2562.            && srcFormat != PIX_FMT_MONOBLACK && dstFormat != PIX_FMT_MONOBLACK

  2563.            && srcFormat != PIX_FMT_MONOWHITE && dstFormat != PIX_FMT_MONOWHITE

  2564.                                              && dstFormat != PIX_FMT_RGB32_1

  2565.                                              && dstFormat != PIX_FMT_BGR32_1

  2566.            && (!needsDither || (c->flags&(SWS_FAST_BILINEAR|SWS_POINT))))

  2567.              c->swScale= rgb2rgbWrapper;

  2568. 

  2569.         if ((usePal(srcFormat) && (

  2570.                  dstFormat == PIX_FMT_RGB32   ||

  2571.                  dstFormat == PIX_FMT_RGB32_1 ||

  2572.                  dstFormat == PIX_FMT_RGB24   ||

  2573.                  dstFormat == PIX_FMT_BGR32   ||

  2574.                  dstFormat == PIX_FMT_BGR32_1 ||

  2575.                  dstFormat == PIX_FMT_BGR24)))

  2576.              c->swScale= pal2rgbWrapper;

  2577. 

  2578.         if (srcFormat == PIX_FMT_YUV422P)

  2579.         {

  2580.             if (dstFormat == PIX_FMT_YUYV422)

  2581.                 c->swScale= YUV422PToYuy2Wrapper;

  2582.             else if (dstFormat == PIX_FMT_UYVY422)

  2583.                 c->swScale= YUV422PToUyvyWrapper;

  2584.         }

  2585. 

  2586.         /* LQ converters if -sws 0 or -sws 4*/

  2587.         if (c->flags&(SWS_FAST_BILINEAR|SWS_POINT)){

  2588.             /* yv12_to_yuy2 */

  2589.             if (srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUVA420P)

  2590.             {

  2591.                 if (dstFormat == PIX_FMT_YUYV422)

  2592.                     c->swScale= PlanarToYuy2Wrapper;

  2593.                 else if (dstFormat == PIX_FMT_UYVY422)

  2594.                     c->swScale= PlanarToUyvyWrapper;

  2595.             }

  2596.         }

  2597.         if(srcFormat == PIX_FMT_YUYV422 && (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P))

  2598.             c->swScale= YUYV2YUV420Wrapper;

  2599.         if(srcFormat == PIX_FMT_UYVY422 && (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P))

  2600.             c->swScale= UYVY2YUV420Wrapper;

  2601.         if(srcFormat == PIX_FMT_YUYV422 && dstFormat == PIX_FMT_YUV422P)

  2602.             c->swScale= YUYV2YUV422Wrapper;

  2603.         if(srcFormat == PIX_FMT_UYVY422 && dstFormat == PIX_FMT_YUV422P)

  2604.             c->swScale= UYVY2YUV422Wrapper;

  2605. 

  2606. #ifdef COMPILE_ALTIVEC

  2607.         if ((c->flags & SWS_CPU_CAPS_ALTIVEC) &&

  2608.             !(c->flags & SWS_BITEXACT) &&

  2609.             srcFormat == PIX_FMT_YUV420P) {

  2610.           // unscaled YV12 -> packed YUV, we want speed

  2611.           if (dstFormat == PIX_FMT_YUYV422)

  2612.               c->swScale= yv12toyuy2_unscaled_altivec;

  2613.           else if (dstFormat == PIX_FMT_UYVY422)

  2614.               c->swScale= yv12touyvy_unscaled_altivec;

  2615.         }

  2616. #endif

  2617. 

  2618.         /* simple copy */

  2619.         if (  srcFormat == dstFormat

  2620.             || (srcFormat == PIX_FMT_YUVA420P && dstFormat == PIX_FMT_YUV420P)

  2621.             || (srcFormat == PIX_FMT_YUV420P && dstFormat == PIX_FMT_YUVA420P)

  2622.             || (isPlanarYUV(srcFormat) && isGray(dstFormat))

  2623.             || (isPlanarYUV(dstFormat) && isGray(srcFormat)))

  2624.         {

  2625.             if (isPacked(c->srcFormat))

  2626.                 c->swScale= packedCopy;

  2627.             else /* Planar YUV or gray */

  2628.                 c->swScale= planarCopy;

  2629.         }

  2630. 

  2631.         /* gray16{le,be} conversions */

  2632.         if (isGray16(srcFormat) && (isPlanarYUV(dstFormat) || (dstFormat == PIX_FMT_GRAY8)))

  2633.         {

  2634.             c->swScale= gray16togray;

  2635.         }

  2636.         if ((isPlanarYUV(srcFormat) || (srcFormat == PIX_FMT_GRAY8)) && isGray16(dstFormat))

  2637.         {

  2638.             c->swScale= graytogray16;

  2639.         }

  2640.         if (srcFormat != dstFormat && isGray16(srcFormat) && isGray16(dstFormat))

  2641.         {

  2642.             c->swScale= gray16swap;

  2643.         }

  2644. 

  2645. #if ARCH_BFIN

  2646.         if (flags & SWS_CPU_CAPS_BFIN)

  2647.             ff_bfin_get_unscaled_swscale (c);

  2648. #endif

  2649. 

  2650.         if (c->swScale){

  2651.             if (flags&SWS_PRINT_INFO)

  2652.                 av_log(c, AV_LOG_INFO, "using unscaled %s -> %s special converter\n",

  2653.                                 sws_format_name(srcFormat), sws_format_name(dstFormat));

  2654.             return c;

  2655.         }

  2656.     }

  2657. 

  2658.     if (flags & SWS_CPU_CAPS_MMX2)

  2659.     {

  2660.         c->canMMX2BeUsed= (dstW >=srcW && (dstW&31)==0 && (srcW&15)==0) ? 1 : 0;

  2661.         if (!c->canMMX2BeUsed && dstW >=srcW && (srcW&15)==0 && (flags&SWS_FAST_BILINEAR))

  2662.         {

  2663.             if (flags&SWS_PRINT_INFO)

  2664.                 av_log(c, AV_LOG_INFO, "output width is not a multiple of 32 -> no MMX2 scaler\n");

  2665.         }

  2666.         if (usesHFilter) c->canMMX2BeUsed=0;

  2667.     }

  2668.     else

  2669.         c->canMMX2BeUsed=0;

  2670. 

  2671.     c->chrXInc= ((c->chrSrcW<<16) + (c->chrDstW>>1))/c->chrDstW;

  2672.     c->chrYInc= ((c->chrSrcH<<16) + (c->chrDstH>>1))/c->chrDstH;

  2673. 

  2674.     // match pixel 0 of the src to pixel 0 of dst and match pixel n-2 of src to pixel n-2 of dst

  2675.     // but only for the FAST_BILINEAR mode otherwise do correct scaling

  2676.     // n-2 is the last chrominance sample available

  2677.     // this is not perfect, but no one should notice the difference, the more correct variant

  2678.     // would be like the vertical one, but that would require some special code for the

  2679.     // first and last pixel

  2680.     if (flags&SWS_FAST_BILINEAR)

  2681.     {

  2682.         if (c->canMMX2BeUsed)

  2683.         {

  2684.             c->lumXInc+= 20;

  2685.             c->chrXInc+= 20;

  2686.         }

  2687.         //we don't use the x86 asm scaler if MMX is available

  2688.         else if (flags & SWS_CPU_CAPS_MMX)

  2689.         {

  2690.             c->lumXInc = ((srcW-2)<<16)/(dstW-2) - 20;

  2691.             c->chrXInc = ((c->chrSrcW-2)<<16)/(c->chrDstW-2) - 20;

  2692.         }

  2693.     }

  2694. 

  2695.     /* precalculate horizontal scaler filter coefficients */

  2696.     {

  2697.         const int filterAlign=

  2698.             (flags & SWS_CPU_CAPS_MMX) ? 4 :

  2699.             (flags & SWS_CPU_CAPS_ALTIVEC) ? 8 :

  2700.             1;

  2701. 

  2702.         initFilter(&c->hLumFilter, &c->hLumFilterPos, &c->hLumFilterSize, c->lumXInc,

  2703.                    srcW      ,       dstW, filterAlign, 1<<14,

  2704.                    (flags&SWS_BICUBLIN) ? (flags|SWS_BICUBIC)  : flags,

  2705.                    srcFilter->lumH, dstFilter->lumH, c->param);

  2706.         initFilter(&c->hChrFilter, &c->hChrFilterPos, &c->hChrFilterSize, c->chrXInc,

  2707.                    c->chrSrcW, c->chrDstW, filterAlign, 1<<14,

  2708.                    (flags&SWS_BICUBLIN) ? (flags|SWS_BILINEAR) : flags,

  2709.                    srcFilter->chrH, dstFilter->chrH, c->param);

  2710. 

  2711. #define MAX_FUNNY_CODE_SIZE 10000

  2712. #if defined(COMPILE_MMX2)

  2713. // can't downscale !!!

  2714.         if (c->canMMX2BeUsed && (flags & SWS_FAST_BILINEAR))

  2715.         {

  2716. #ifdef MAP_ANONYMOUS

  2717.             c->funnyYCode  = mmap(NULL, MAX_FUNNY_CODE_SIZE, PROT_EXEC | PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, 0, 0);

  2718.             c->funnyUVCode = mmap(NULL, MAX_FUNNY_CODE_SIZE, PROT_EXEC | PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, 0, 0);

  2719. #elif HAVE_VIRTUALALLOC

  2720.             c->funnyYCode  = VirtualAlloc(NULL, MAX_FUNNY_CODE_SIZE, MEM_COMMIT, PAGE_EXECUTE_READWRITE);

  2721.             c->funnyUVCode = VirtualAlloc(NULL, MAX_FUNNY_CODE_SIZE, MEM_COMMIT, PAGE_EXECUTE_READWRITE);

  2722. #else

  2723.             c->funnyYCode  = av_malloc(MAX_FUNNY_CODE_SIZE);

  2724.             c->funnyUVCode = av_malloc(MAX_FUNNY_CODE_SIZE);

  2725. #endif

  2726. 

  2727.             c->lumMmx2Filter   = av_malloc((dstW        /8+8)*sizeof(int16_t));

  2728.             c->chrMmx2Filter   = av_malloc((c->chrDstW  /4+8)*sizeof(int16_t));

  2729.             c->lumMmx2FilterPos= av_malloc((dstW      /2/8+8)*sizeof(int32_t));

  2730.             c->chrMmx2FilterPos= av_malloc((c->chrDstW/2/4+8)*sizeof(int32_t));

  2731. 

  2732.             initMMX2HScaler(      dstW, c->lumXInc, c->funnyYCode , c->lumMmx2Filter, c->lumMmx2FilterPos, 8);

  2733.             initMMX2HScaler(c->chrDstW, c->chrXInc, c->funnyUVCode, c->chrMmx2Filter, c->chrMmx2FilterPos, 4);

  2734.         }

  2735. #endif /* defined(COMPILE_MMX2) */

  2736.     } // initialize horizontal stuff

  2737. 

  2738. 

  2739. 

  2740.     /* precalculate vertical scaler filter coefficients */

  2741.     {

  2742.         const int filterAlign=

  2743.             (flags & SWS_CPU_CAPS_MMX) && (flags & SWS_ACCURATE_RND) ? 2 :

  2744.             (flags & SWS_CPU_CAPS_ALTIVEC) ? 8 :

  2745.             1;

  2746. 

  2747.         initFilter(&c->vLumFilter, &c->vLumFilterPos, &c->vLumFilterSize, c->lumYInc,

  2748.                    srcH      ,        dstH, filterAlign, (1<<12),

  2749.                    (flags&SWS_BICUBLIN) ? (flags|SWS_BICUBIC)  : flags,

  2750.                    srcFilter->lumV, dstFilter->lumV, c->param);

  2751.         initFilter(&c->vChrFilter, &c->vChrFilterPos, &c->vChrFilterSize, c->chrYInc,

  2752.                    c->chrSrcH, c->chrDstH, filterAlign, (1<<12),

  2753.                    (flags&SWS_BICUBLIN) ? (flags|SWS_BILINEAR) : flags,

  2754.                    srcFilter->chrV, dstFilter->chrV, c->param);

  2755. 

  2756. #if HAVE_ALTIVEC

  2757.         c->vYCoeffsBank = av_malloc(sizeof (vector signed short)*c->vLumFilterSize*c->dstH);

  2758.         c->vCCoeffsBank = av_malloc(sizeof (vector signed short)*c->vChrFilterSize*c->chrDstH);

  2759. 

  2760.         for (i=0;i<c->vLumFilterSize*c->dstH;i++) {

  2761.             int j;

  2762.             short *p = (short *)&c->vYCoeffsBank[i];

  2763.             for (j=0;j<8;j++)

  2764.                 p[j] = c->vLumFilter[i];

  2765.         }

  2766. 

  2767.         for (i=0;i<c->vChrFilterSize*c->chrDstH;i++) {

  2768.             int j;

  2769.             short *p = (short *)&c->vCCoeffsBank[i];

  2770.             for (j=0;j<8;j++)

  2771.                 p[j] = c->vChrFilter[i];

  2772.         }

  2773. #endif

  2774.     }

  2775. 

  2776.     // calculate buffer sizes so that they won't run out while handling these damn slices

  2777.     c->vLumBufSize= c->vLumFilterSize;

  2778.     c->vChrBufSize= c->vChrFilterSize;

  2779.     for (i=0; i<dstH; i++)

  2780.     {

  2781.         int chrI= i*c->chrDstH / dstH;

  2782.         int nextSlice= FFMAX(c->vLumFilterPos[i   ] + c->vLumFilterSize - 1,

  2783.                            ((c->vChrFilterPos[chrI] + c->vChrFilterSize - 1)<<c->chrSrcVSubSample));

  2784. 

  2785.         nextSlice>>= c->chrSrcVSubSample;

  2786.         nextSlice<<= c->chrSrcVSubSample;

  2787.         if (c->vLumFilterPos[i   ] + c->vLumBufSize < nextSlice)

  2788.             c->vLumBufSize= nextSlice - c->vLumFilterPos[i];

  2789.         if (c->vChrFilterPos[chrI] + c->vChrBufSize < (nextSlice>>c->chrSrcVSubSample))

  2790.             c->vChrBufSize= (nextSlice>>c->chrSrcVSubSample) - c->vChrFilterPos[chrI];

  2791.     }

  2792. 

  2793.     // allocate pixbufs (we use dynamic allocation because otherwise we would need to

  2794.     c->lumPixBuf= av_malloc(c->vLumBufSize*2*sizeof(int16_t*));

  2795.     c->chrPixBuf= av_malloc(c->vChrBufSize*2*sizeof(int16_t*));

  2796.     if (CONFIG_SWSCALE_ALPHA && isALPHA(c->srcFormat) && isALPHA(c->dstFormat))

  2797.         c->alpPixBuf= av_malloc(c->vLumBufSize*2*sizeof(int16_t*));

  2798.     //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)

  2799.     /* align at 16 bytes for AltiVec */

  2800.     for (i=0; i<c->vLumBufSize; i++)

  2801.         c->lumPixBuf[i]= c->lumPixBuf[i+c->vLumBufSize]= av_mallocz(VOF+1);

  2802.     for (i=0; i<c->vChrBufSize; i++)

  2803.         c->chrPixBuf[i]= c->chrPixBuf[i+c->vChrBufSize]= av_malloc((VOF+1)*2);

  2804.     if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf)

  2805.         for (i=0; i<c->vLumBufSize; i++)

  2806.             c->alpPixBuf[i]= c->alpPixBuf[i+c->vLumBufSize]= av_mallocz(VOF+1);

  2807. 

  2808.     //try to avoid drawing green stuff between the right end and the stride end

  2809.     for (i=0; i<c->vChrBufSize; i++) memset(c->chrPixBuf[i], 64, (VOF+1)*2);

  2810. 

  2811.     assert(2*VOFW == VOF);

  2812. 

  2813.     assert(c->chrDstH <= dstH);

  2814. 

  2815.     if (flags&SWS_PRINT_INFO)

  2816.     {

  2817. #ifdef DITHER1XBPP

  2818.         const char *dither= " dithered";

  2819. #else

  2820.         const char *dither= "";

  2821. #endif

  2822.         if (flags&SWS_FAST_BILINEAR)

  2823.             av_log(c, AV_LOG_INFO, "FAST_BILINEAR scaler, ");

  2824.         else if (flags&SWS_BILINEAR)

  2825.             av_log(c, AV_LOG_INFO, "BILINEAR scaler, ");

  2826.         else if (flags&SWS_BICUBIC)

  2827.             av_log(c, AV_LOG_INFO, "BICUBIC scaler, ");

  2828.         else if (flags&SWS_X)

  2829.             av_log(c, AV_LOG_INFO, "Experimental scaler, ");

  2830.         else if (flags&SWS_POINT)

  2831.             av_log(c, AV_LOG_INFO, "Nearest Neighbor / POINT scaler, ");

  2832.         else if (flags&SWS_AREA)

  2833.             av_log(c, AV_LOG_INFO, "Area Averageing scaler, ");

  2834.         else if (flags&SWS_BICUBLIN)

  2835.             av_log(c, AV_LOG_INFO, "luma BICUBIC / chroma BILINEAR scaler, ");

  2836.         else if (flags&SWS_GAUSS)

  2837.             av_log(c, AV_LOG_INFO, "Gaussian scaler, ");

  2838.         else if (flags&SWS_SINC)

  2839.             av_log(c, AV_LOG_INFO, "Sinc scaler, ");

  2840.         else if (flags&SWS_LANCZOS)

  2841.             av_log(c, AV_LOG_INFO, "Lanczos scaler, ");

  2842.         else if (flags&SWS_SPLINE)

  2843.             av_log(c, AV_LOG_INFO, "Bicubic spline scaler, ");

  2844.         else

  2845.             av_log(c, AV_LOG_INFO, "ehh flags invalid?! ");

  2846. 

  2847.         if (dstFormat==PIX_FMT_BGR555 || dstFormat==PIX_FMT_BGR565)

  2848.             av_log(c, AV_LOG_INFO, "from %s to%s %s ",

  2849.                    sws_format_name(srcFormat), dither, sws_format_name(dstFormat));

  2850.         else

  2851.             av_log(c, AV_LOG_INFO, "from %s to %s ",

  2852.                    sws_format_name(srcFormat), sws_format_name(dstFormat));

  2853. 

  2854.         if (flags & SWS_CPU_CAPS_MMX2)

  2855.             av_log(c, AV_LOG_INFO, "using MMX2\n");

  2856.         else if (flags & SWS_CPU_CAPS_3DNOW)

  2857.             av_log(c, AV_LOG_INFO, "using 3DNOW\n");

  2858.         else if (flags & SWS_CPU_CAPS_MMX)

  2859.             av_log(c, AV_LOG_INFO, "using MMX\n");

  2860.         else if (flags & SWS_CPU_CAPS_ALTIVEC)

  2861.             av_log(c, AV_LOG_INFO, "using AltiVec\n");

  2862.         else

  2863.             av_log(c, AV_LOG_INFO, "using C\n");

  2864.     }

  2865. 

  2866.     if (flags & SWS_PRINT_INFO)

  2867.     {

  2868.         if (flags & SWS_CPU_CAPS_MMX)

  2869.         {

  2870.             if (c->canMMX2BeUsed && (flags&SWS_FAST_BILINEAR))

  2871.                 av_log(c, AV_LOG_VERBOSE, "using FAST_BILINEAR MMX2 scaler for horizontal scaling\n");

  2872.             else

  2873.             {

  2874.                 if (c->hLumFilterSize==4)

  2875.                     av_log(c, AV_LOG_VERBOSE, "using 4-tap MMX scaler for horizontal luminance scaling\n");

  2876.                 else if (c->hLumFilterSize==8)

  2877.                     av_log(c, AV_LOG_VERBOSE, "using 8-tap MMX scaler for horizontal luminance scaling\n");

  2878.                 else

  2879.                     av_log(c, AV_LOG_VERBOSE, "using n-tap MMX scaler for horizontal luminance scaling\n");

  2880. 

  2881.                 if (c->hChrFilterSize==4)

  2882.                     av_log(c, AV_LOG_VERBOSE, "using 4-tap MMX scaler for horizontal chrominance scaling\n");

  2883.                 else if (c->hChrFilterSize==8)

  2884.                     av_log(c, AV_LOG_VERBOSE, "using 8-tap MMX scaler for horizontal chrominance scaling\n");

  2885.                 else

  2886.                     av_log(c, AV_LOG_VERBOSE, "using n-tap MMX scaler for horizontal chrominance scaling\n");

  2887.             }

  2888.         }

  2889.         else

  2890.         {

  2891. #if ARCH_X86

  2892.             av_log(c, AV_LOG_VERBOSE, "using x86 asm scaler for horizontal scaling\n");

  2893. #else

  2894.             if (flags & SWS_FAST_BILINEAR)

  2895.                 av_log(c, AV_LOG_VERBOSE, "using FAST_BILINEAR C scaler for horizontal scaling\n");

  2896.             else

  2897.                 av_log(c, AV_LOG_VERBOSE, "using C scaler for horizontal scaling\n");

  2898. #endif

  2899.         }

  2900.         if (isPlanarYUV(dstFormat))

  2901.         {

  2902.             if (c->vLumFilterSize==1)

  2903.                 av_log(c, AV_LOG_VERBOSE, "using 1-tap %s \"scaler\" for vertical scaling (YV12 like)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  2904.             else

  2905.                 av_log(c, AV_LOG_VERBOSE, "using n-tap %s scaler for vertical scaling (YV12 like)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  2906.         }

  2907.         else

  2908.         {

  2909.             if (c->vLumFilterSize==1 && c->vChrFilterSize==2)

  2910.                 av_log(c, AV_LOG_VERBOSE, "using 1-tap %s \"scaler\" for vertical luminance scaling (BGR)\n"

  2911.                        "      2-tap scaler for vertical chrominance scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  2912.             else if (c->vLumFilterSize==2 && c->vChrFilterSize==2)

  2913.                 av_log(c, AV_LOG_VERBOSE, "using 2-tap linear %s scaler for vertical scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  2914.             else

  2915.                 av_log(c, AV_LOG_VERBOSE, "using n-tap %s scaler for vertical scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  2916.         }

  2917. 

  2918.         if (dstFormat==PIX_FMT_BGR24)

  2919.             av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR24 converter\n",

  2920.                    (flags & SWS_CPU_CAPS_MMX2) ? "MMX2" : ((flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C"));

  2921.         else if (dstFormat==PIX_FMT_RGB32)

  2922.             av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR32 converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  2923.         else if (dstFormat==PIX_FMT_BGR565)

  2924.             av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR16 converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  2925.         else if (dstFormat==PIX_FMT_BGR555)

  2926.             av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR15 converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

  2927. 

  2928.         av_log(c, AV_LOG_VERBOSE, "%dx%d -> %dx%d\n", srcW, srcH, dstW, dstH);

  2929.     }

  2930.     if (flags & SWS_PRINT_INFO)

  2931.     {

  2932.         av_log(c, AV_LOG_DEBUG, "lum srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",

  2933.                c->srcW, c->srcH, c->dstW, c->dstH, c->lumXInc, c->lumYInc);

  2934.         av_log(c, AV_LOG_DEBUG, "chr srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",

  2935.                c->chrSrcW, c->chrSrcH, c->chrDstW, c->chrDstH, c->chrXInc, c->chrYInc);

  2936.     }

  2937. 

  2938.     c->swScale= getSwsFunc(c);

  2939.     return c;

  2940. }

  2941. 

  2942. /**

  2943.  * swscale wrapper, so we don't need to export the SwsContext.

  2944.  * Assumes planar YUV to be in YUV order instead of YVU.

  2945.  */

  2946. int sws_scale(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  2947.               int srcSliceH, uint8_t* dst[], int dstStride[]){

  2948.     int i;

  2949.     uint8_t* src2[4]= {src[0], src[1], src[2], src[3]};

  2950. 

  2951.     if (c->sliceDir == 0 && srcSliceY != 0 && srcSliceY + srcSliceH != c->srcH) {

  2952.         av_log(c, AV_LOG_ERROR, "Slices start in the middle!\n");

  2953.         return 0;

  2954.     }

  2955.     if (c->sliceDir == 0) {

  2956.         if (srcSliceY == 0) c->sliceDir = 1; else c->sliceDir = -1;

  2957.     }

  2958. 

  2959.     if (usePal(c->srcFormat)){

  2960.         for (i=0; i<256; i++){

  2961.             int p, r, g, b,y,u,v;

  2962.             if(c->srcFormat == PIX_FMT_PAL8){

  2963.                 p=((uint32_t*)(src[1]))[i];

  2964.                 r= (p>>16)&0xFF;

  2965.                 g= (p>> 8)&0xFF;

  2966.                 b=  p     &0xFF;

  2967.             }else if(c->srcFormat == PIX_FMT_RGB8){

  2968.                 r= (i>>5    )*36;

  2969.                 g= ((i>>2)&7)*36;

  2970.                 b= (i&3     )*85;

  2971.             }else if(c->srcFormat == PIX_FMT_BGR8){

  2972.                 b= (i>>6    )*85;

  2973.                 g= ((i>>3)&7)*36;

  2974.                 r= (i&7     )*36;

  2975.             }else if(c->srcFormat == PIX_FMT_RGB4_BYTE){

  2976.                 r= (i>>3    )*255;

  2977.                 g= ((i>>1)&3)*85;

  2978.                 b= (i&1     )*255;

  2979.             }else {

  2980.                 assert(c->srcFormat == PIX_FMT_BGR4_BYTE);

  2981.                 b= (i>>3    )*255;

  2982.                 g= ((i>>1)&3)*85;

  2983.                 r= (i&1     )*255;

  2984.             }

  2985.             y= av_clip_uint8((RY*r + GY*g + BY*b + ( 33<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);

  2986.             u= av_clip_uint8((RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);

  2987.             v= av_clip_uint8((RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);

  2988.             c->pal_yuv[i]= y + (u<<8) + (v<<16);

  2989. 

  2990. 

  2991.             switch(c->dstFormat) {

  2992.             case PIX_FMT_BGR32:

  2993. #ifndef WORDS_BIGENDIAN

  2994.             case PIX_FMT_RGB24:

  2995. #endif

  2996.                 c->pal_rgb[i]=  r + (g<<8) + (b<<16);

  2997.                 break;

  2998.             case PIX_FMT_BGR32_1:

  2999. #ifdef  WORDS_BIGENDIAN

  3000.             case PIX_FMT_BGR24:

  3001. #endif

  3002.                 c->pal_rgb[i]= (r + (g<<8) + (b<<16)) << 8;

  3003.                 break;

  3004.             case PIX_FMT_RGB32_1:

  3005. #ifdef  WORDS_BIGENDIAN

  3006.             case PIX_FMT_RGB24:

  3007. #endif

  3008.                 c->pal_rgb[i]= (b + (g<<8) + (r<<16)) << 8;

  3009.                 break;

  3010.             case PIX_FMT_RGB32:

  3011. #ifndef WORDS_BIGENDIAN

  3012.             case PIX_FMT_BGR24:

  3013. #endif

  3014.             default:

  3015.                 c->pal_rgb[i]=  b + (g<<8) + (r<<16);

  3016.             }

  3017.         }

  3018.     }

  3019. 

  3020.     // copy strides, so they can safely be modified

  3021.     if (c->sliceDir == 1) {

  3022.         // slices go from top to bottom

  3023.         int srcStride2[4]= {srcStride[0], srcStride[1], srcStride[2], srcStride[3]};

  3024.         int dstStride2[4]= {dstStride[0], dstStride[1], dstStride[2], dstStride[3]};

  3025.         return c->swScale(c, src2, srcStride2, srcSliceY, srcSliceH, dst, dstStride2);

  3026.     } else {

  3027.         // slices go from bottom to top => we flip the image internally

  3028.         uint8_t* dst2[4]= {dst[0] + (c->dstH-1)*dstStride[0],

  3029.                            dst[1] + ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[1],

  3030.                            dst[2] + ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[2],

  3031.                            dst[3] + (c->dstH-1)*dstStride[3]};

  3032.         int srcStride2[4]= {-srcStride[0], -srcStride[1], -srcStride[2], -srcStride[3]};

  3033.         int dstStride2[4]= {-dstStride[0], -dstStride[1], -dstStride[2], -dstStride[3]};

  3034. 

  3035.         src2[0] += (srcSliceH-1)*srcStride[0];

  3036.         if (!usePal(c->srcFormat))

  3037.             src2[1] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[1];

  3038.         src2[2] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[2];

  3039.         src2[3] += (srcSliceH-1)*srcStride[3];

  3040. 

  3041.         return c->swScale(c, src2, srcStride2, c->srcH-srcSliceY-srcSliceH, srcSliceH, dst2, dstStride2);

  3042.     }

  3043. }

  3044. 

  3045. #if LIBSWSCALE_VERSION_MAJOR < 1

  3046. int sws_scale_ordered(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,

  3047.                       int srcSliceH, uint8_t* dst[], int dstStride[]){

  3048.     return sws_scale(c, src, srcStride, srcSliceY, srcSliceH, dst, dstStride);

  3049. }

  3050. #endif

  3051. 

  3052. SwsFilter *sws_getDefaultFilter(float lumaGBlur, float chromaGBlur,

  3053.                                 float lumaSharpen, float chromaSharpen,

  3054.                                 float chromaHShift, float chromaVShift,

  3055.                                 int verbose)

  3056. {

  3057.     SwsFilter *filter= av_malloc(sizeof(SwsFilter));

  3058. 

  3059.     if (lumaGBlur!=0.0){

  3060.         filter->lumH= sws_getGaussianVec(lumaGBlur, 3.0);

  3061.         filter->lumV= sws_getGaussianVec(lumaGBlur, 3.0);

  3062.     }else{

  3063.         filter->lumH= sws_getIdentityVec();

  3064.         filter->lumV= sws_getIdentityVec();

  3065.     }

  3066. 

  3067.     if (chromaGBlur!=0.0){

  3068.         filter->chrH= sws_getGaussianVec(chromaGBlur, 3.0);

  3069.         filter->chrV= sws_getGaussianVec(chromaGBlur, 3.0);

  3070.     }else{

  3071.         filter->chrH= sws_getIdentityVec();

  3072.         filter->chrV= sws_getIdentityVec();

  3073.     }

  3074. 

  3075.     if (chromaSharpen!=0.0){

  3076.         SwsVector *id= sws_getIdentityVec();

  3077.         sws_scaleVec(filter->chrH, -chromaSharpen);

  3078.         sws_scaleVec(filter->chrV, -chromaSharpen);

  3079.         sws_addVec(filter->chrH, id);

  3080.         sws_addVec(filter->chrV, id);

  3081.         sws_freeVec(id);

  3082.     }

  3083. 

  3084.     if (lumaSharpen!=0.0){

  3085.         SwsVector *id= sws_getIdentityVec();

  3086.         sws_scaleVec(filter->lumH, -lumaSharpen);

  3087.         sws_scaleVec(filter->lumV, -lumaSharpen);

  3088.         sws_addVec(filter->lumH, id);

  3089.         sws_addVec(filter->lumV, id);

  3090.         sws_freeVec(id);

  3091.     }

  3092. 

  3093.     if (chromaHShift != 0.0)

  3094.         sws_shiftVec(filter->chrH, (int)(chromaHShift+0.5));

  3095. 

  3096.     if (chromaVShift != 0.0)

  3097.         sws_shiftVec(filter->chrV, (int)(chromaVShift+0.5));

  3098. 

  3099.     sws_normalizeVec(filter->chrH, 1.0);

  3100.     sws_normalizeVec(filter->chrV, 1.0);

  3101.     sws_normalizeVec(filter->lumH, 1.0);

  3102.     sws_normalizeVec(filter->lumV, 1.0);

  3103. 

  3104.     if (verbose) sws_printVec2(filter->chrH, NULL, AV_LOG_DEBUG);

  3105.     if (verbose) sws_printVec2(filter->lumH, NULL, AV_LOG_DEBUG);

  3106. 

  3107.     return filter;

  3108. }

  3109. 

  3110. SwsVector *sws_getGaussianVec(double variance, double quality){

  3111.     const int length= (int)(variance*quality + 0.5) | 1;

  3112.     int i;

  3113.     double *coeff= av_malloc(length*sizeof(double));

  3114.     double middle= (length-1)*0.5;

  3115.     SwsVector *vec= av_malloc(sizeof(SwsVector));

  3116. 

  3117.     vec->coeff= coeff;

  3118.     vec->length= length;

  3119. 

  3120.     for (i=0; i<length; i++)

  3121.     {

  3122.         double dist= i-middle;

  3123.         coeff[i]= exp(-dist*dist/(2*variance*variance)) / sqrt(2*variance*PI);

  3124.     }

  3125. 

  3126.     sws_normalizeVec(vec, 1.0);

  3127. 

  3128.     return vec;

  3129. }

  3130. 

  3131. SwsVector *sws_getConstVec(double c, int length){

  3132.     int i;

  3133.     double *coeff= av_malloc(length*sizeof(double));

  3134.     SwsVector *vec= av_malloc(sizeof(SwsVector));

  3135. 

  3136.     vec->coeff= coeff;

  3137.     vec->length= length;

  3138. 

  3139.     for (i=0; i<length; i++)

  3140.         coeff[i]= c;

  3141. 

  3142.     return vec;

  3143. }

  3144. 

  3145. 

  3146. SwsVector *sws_getIdentityVec(void){

  3147.     return sws_getConstVec(1.0, 1);

  3148. }

  3149. 

  3150. double sws_dcVec(SwsVector *a){

  3151.     int i;

  3152.     double sum=0;

  3153. 

  3154.     for (i=0; i<a->length; i++)

  3155.         sum+= a->coeff[i];

  3156. 

  3157.     return sum;

  3158. }

  3159. 

  3160. void sws_scaleVec(SwsVector *a, double scalar){

  3161.     int i;

  3162. 

  3163.     for (i=0; i<a->length; i++)

  3164.         a->coeff[i]*= scalar;

  3165. }

  3166. 

  3167. void sws_normalizeVec(SwsVector *a, double height){

  3168.     sws_scaleVec(a, height/sws_dcVec(a));

  3169. }

  3170. 

  3171. static SwsVector *sws_getConvVec(SwsVector *a, SwsVector *b){

  3172.     int length= a->length + b->length - 1;

  3173.     double *coeff= av_malloc(length*sizeof(double));

  3174.     int i, j;

  3175.     SwsVector *vec= av_malloc(sizeof(SwsVector));

  3176. 

  3177.     vec->coeff= coeff;

  3178.     vec->length= length;

  3179. 

  3180.     for (i=0; i<length; i++) coeff[i]= 0.0;

  3181. 

  3182.     for (i=0; i<a->length; i++)

  3183.     {

  3184.         for (j=0; j<b->length; j++)

  3185.         {

  3186.             coeff[i+j]+= a->coeff[i]*b->coeff[j];

  3187.         }

  3188.     }

  3189. 

  3190.     return vec;

  3191. }

  3192. 

  3193. static SwsVector *sws_sumVec(SwsVector *a, SwsVector *b){

  3194.     int length= FFMAX(a->length, b->length);

  3195.     double *coeff= av_malloc(length*sizeof(double));

  3196.     int i;

  3197.     SwsVector *vec= av_malloc(sizeof(SwsVector));

  3198. 

  3199.     vec->coeff= coeff;

  3200.     vec->length= length;

  3201. 

  3202.     for (i=0; i<length; i++) coeff[i]= 0.0;

  3203. 

  3204.     for (i=0; i<a->length; i++) coeff[i + (length-1)/2 - (a->length-1)/2]+= a->coeff[i];

  3205.     for (i=0; i<b->length; i++) coeff[i + (length-1)/2 - (b->length-1)/2]+= b->coeff[i];

  3206. 

  3207.     return vec;

  3208. }

  3209. 

  3210. static SwsVector *sws_diffVec(SwsVector *a, SwsVector *b){

  3211.     int length= FFMAX(a->length, b->length);

  3212.     double *coeff= av_malloc(length*sizeof(double));

  3213.     int i;

  3214.     SwsVector *vec= av_malloc(sizeof(SwsVector));

  3215. 

  3216.     vec->coeff= coeff;

  3217.     vec->length= length;

  3218. 

  3219.     for (i=0; i<length; i++) coeff[i]= 0.0;

  3220. 

  3221.     for (i=0; i<a->length; i++) coeff[i + (length-1)/2 - (a->length-1)/2]+= a->coeff[i];

  3222.     for (i=0; i<b->length; i++) coeff[i + (length-1)/2 - (b->length-1)/2]-= b->coeff[i];

  3223. 

  3224.     return vec;

  3225. }

  3226. 

  3227. /* shift left / or right if "shift" is negative */

  3228. static SwsVector *sws_getShiftedVec(SwsVector *a, int shift){

  3229.     int length= a->length + FFABS(shift)*2;

  3230.     double *coeff= av_malloc(length*sizeof(double));

  3231.     int i;

  3232.     SwsVector *vec= av_malloc(sizeof(SwsVector));

  3233. 

  3234.     vec->coeff= coeff;

  3235.     vec->length= length;

  3236. 

  3237.     for (i=0; i<length; i++) coeff[i]= 0.0;

  3238. 

  3239.     for (i=0; i<a->length; i++)

  3240.     {

  3241.         coeff[i + (length-1)/2 - (a->length-1)/2 - shift]= a->coeff[i];

  3242.     }

  3243. 

  3244.     return vec;

  3245. }

  3246. 

  3247. void sws_shiftVec(SwsVector *a, int shift){

  3248.     SwsVector *shifted= sws_getShiftedVec(a, shift);

  3249.     av_free(a->coeff);

  3250.     a->coeff= shifted->coeff;

  3251.     a->length= shifted->length;

  3252.     av_free(shifted);

  3253. }

  3254. 

  3255. void sws_addVec(SwsVector *a, SwsVector *b){

  3256.     SwsVector *sum= sws_sumVec(a, b);

  3257.     av_free(a->coeff);

  3258.     a->coeff= sum->coeff;

  3259.     a->length= sum->length;

  3260.     av_free(sum);

  3261. }

  3262. 

  3263. void sws_subVec(SwsVector *a, SwsVector *b){

  3264.     SwsVector *diff= sws_diffVec(a, b);

  3265.     av_free(a->coeff);

  3266.     a->coeff= diff->coeff;

  3267.     a->length= diff->length;

  3268.     av_free(diff);

  3269. }

  3270. 

  3271. void sws_convVec(SwsVector *a, SwsVector *b){

  3272.     SwsVector *conv= sws_getConvVec(a, b);

  3273.     av_free(a->coeff);

  3274.     a->coeff= conv->coeff;

  3275.     a->length= conv->length;

  3276.     av_free(conv);

  3277. }

  3278. 

  3279. SwsVector *sws_cloneVec(SwsVector *a){

  3280.     double *coeff= av_malloc(a->length*sizeof(double));

  3281.     int i;

  3282.     SwsVector *vec= av_malloc(sizeof(SwsVector));

  3283. 

  3284.     vec->coeff= coeff;

  3285.     vec->length= a->length;

  3286. 

  3287.     for (i=0; i<a->length; i++) coeff[i]= a->coeff[i];

  3288. 

  3289.     return vec;

  3290. }

  3291. 

  3292. void sws_printVec2(SwsVector *a, AVClass *log_ctx, int log_level){

  3293.     int i;

  3294.     double max=0;

  3295.     double min=0;

  3296.     double range;

  3297. 

  3298.     for (i=0; i<a->length; i++)

  3299.         if (a->coeff[i]>max) max= a->coeff[i];

  3300. 

  3301.     for (i=0; i<a->length; i++)

  3302.         if (a->coeff[i]<min) min= a->coeff[i];

  3303. 

  3304.     range= max - min;

  3305. 

  3306.     for (i=0; i<a->length; i++)

  3307.     {

  3308.         int x= (int)((a->coeff[i]-min)*60.0/range +0.5);

  3309.         av_log(log_ctx, log_level, "%1.3f ", a->coeff[i]);

  3310.         for (;x>0; x--) av_log(log_ctx, log_level, " ");

  3311.         av_log(log_ctx, log_level, "|\n");

  3312.     }

  3313. }

  3314. 

  3315. #if LIBSWSCALE_VERSION_MAJOR < 1

  3316. void sws_printVec(SwsVector *a){

  3317.     sws_printVec2(a, NULL, AV_LOG_DEBUG);

  3318. }

  3319. #endif

  3320. 

  3321. void sws_freeVec(SwsVector *a){

  3322.     if (!a) return;

  3323.     av_freep(&a->coeff);

  3324.     a->length=0;

  3325.     av_free(a);

  3326. }

  3327. 

  3328. void sws_freeFilter(SwsFilter *filter){

  3329.     if (!filter) return;

  3330. 

  3331.     if (filter->lumH) sws_freeVec(filter->lumH);

  3332.     if (filter->lumV) sws_freeVec(filter->lumV);

  3333.     if (filter->chrH) sws_freeVec(filter->chrH);

  3334.     if (filter->chrV) sws_freeVec(filter->chrV);

  3335.     av_free(filter);

  3336. }

  3337. 

  3338. 

  3339. void sws_freeContext(SwsContext *c){

  3340.     int i;

  3341.     if (!c) return;

  3342. 

  3343.     if (c->lumPixBuf)

  3344.     {

  3345.         for (i=0; i<c->vLumBufSize; i++)

  3346.             av_freep(&c->lumPixBuf[i]);

  3347.         av_freep(&c->lumPixBuf);

  3348.     }

  3349. 

  3350.     if (c->chrPixBuf)

  3351.     {

  3352.         for (i=0; i<c->vChrBufSize; i++)

  3353.             av_freep(&c->chrPixBuf[i]);

  3354.         av_freep(&c->chrPixBuf);

  3355.     }

  3356. 

  3357.     if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf){

  3358.         for (i=0; i<c->vLumBufSize; i++)

  3359.             av_freep(&c->alpPixBuf[i]);

  3360.         av_freep(&c->alpPixBuf);

  3361.     }

  3362. 

  3363.     av_freep(&c->vLumFilter);

  3364.     av_freep(&c->vChrFilter);

  3365.     av_freep(&c->hLumFilter);

  3366.     av_freep(&c->hChrFilter);

  3367. #if HAVE_ALTIVEC

  3368.     av_freep(&c->vYCoeffsBank);

  3369.     av_freep(&c->vCCoeffsBank);

  3370. #endif

  3371. 

  3372.     av_freep(&c->vLumFilterPos);

  3373.     av_freep(&c->vChrFilterPos);

  3374.     av_freep(&c->hLumFilterPos);

  3375.     av_freep(&c->hChrFilterPos);

  3376. 

  3377. #if ARCH_X86 && CONFIG_GPL

  3378. #ifdef MAP_ANONYMOUS

  3379.     if (c->funnyYCode ) munmap(c->funnyYCode , MAX_FUNNY_CODE_SIZE);

  3380.     if (c->funnyUVCode) munmap(c->funnyUVCode, MAX_FUNNY_CODE_SIZE);

  3381. #elif HAVE_VIRTUALALLOC

  3382.     if (c->funnyYCode ) VirtualFree(c->funnyYCode , MAX_FUNNY_CODE_SIZE, MEM_RELEASE);

  3383.     if (c->funnyUVCode) VirtualFree(c->funnyUVCode, MAX_FUNNY_CODE_SIZE, MEM_RELEASE);

  3384. #else

  3385.     av_free(c->funnyYCode );

  3386.     av_free(c->funnyUVCode);

  3387. #endif

  3388.     c->funnyYCode=NULL;

  3389.     c->funnyUVCode=NULL;

  3390. #endif /* ARCH_X86 && CONFIG_GPL */

  3391. 

  3392.     av_freep(&c->lumMmx2Filter);

  3393.     av_freep(&c->chrMmx2Filter);

  3394.     av_freep(&c->lumMmx2FilterPos);

  3395.     av_freep(&c->chrMmx2FilterPos);

  3396.     av_freep(&c->yuvTable);

  3397. 

  3398.     av_free(c);

  3399. }

  3400. 

  3401. struct SwsContext *sws_getCachedContext(struct SwsContext *context,

  3402.                                         int srcW, int srcH, enum PixelFormat srcFormat,

  3403.                                         int dstW, int dstH, enum PixelFormat dstFormat, int flags,

  3404.                                         SwsFilter *srcFilter, SwsFilter *dstFilter, double *param)

  3405. {

  3406.     static const double default_param[2] = {SWS_PARAM_DEFAULT, SWS_PARAM_DEFAULT};

  3407. 

  3408.     if (!param)

  3409.         param = default_param;

  3410. 

  3411.     if (context) {

  3412.         if (context->srcW != srcW || context->srcH != srcH ||

  3413.             context->srcFormat != srcFormat ||

  3414.             context->dstW != dstW || context->dstH != dstH ||

  3415.             context->dstFormat != dstFormat || context->flags != flags ||

  3416.             context->param[0] != param[0] || context->param[1] != param[1])

  3417.         {

  3418.             sws_freeContext(context);

  3419.             context = NULL;

  3420.         }

  3421.     }

  3422.     if (!context) {

  3423.         return sws_getContext(srcW, srcH, srcFormat,

  3424.                               dstW, dstH, dstFormat, flags,

  3425.                               srcFilter, dstFilter, param);

  3426.     }

  3427.     return context;

  3428. }